Amino Acid Substitutions Account for Most MexS Alterations in Clinical nfxC Mutants of Pseudomonas aeruginosa

Abstract

Multidrug-resistant mutants of Pseudomonas aeruginosa that overproduce the active efflux system MexEF-OprN (called nfxC mutants) have rarely been characterized in the hospital setting. Screening of 221 clinical strains exhibiting a reduced susceptibility to ciprofloxacin (a substrate of MexEF-OprN) and imipenem (a substrate of the negatively coregulated porin OprD) led to the identification of 43 (19.5%) nfxC mutants. Subsequent analysis of 22 nonredundant mutants showed that, in contrast to their in vitro-selected counterparts, only 3 of them (13.6%) harbored a disrupted mexS gene, which codes for the oxidoreductase MexS, whose inactivation is known to activate the mexEF-oprN operon through a LysR-type regulator, MexT. Nine (40.9%) of the clinical nfxC mutants contained single amino acid mutations in MexS, and these were associated with moderate effects on resistance and virulence factor production in 8/9 strains. Finally, the remaining 10 (45.5%) nfxC mutants did not display mutations in any of the regulators known to control mexEF-oprN expression (the mexS, mexT, mvaT, and ampR genes), confirming that other loci are responsible for pump upregulation in patients. Collectively, these data demonstrate that nfxC mutants are probably more frequent in the hospital than previously thought and have genetic and phenotypic features somewhat different from those of in vitro-selected mutants.

INTRODUCTION

Pseudomonas aeruginosa is a notorious cause of acute and chronic infections in vulnerable patients. The ability of this environmental Gram-negative bacterium to produce a broad range of virulence factors (1) and to become resistant to multiple antimicrobial agents is considered a key to its success in the hospital setting. When overexpressed upon mutation, several efflux systems belonging to the resistance-nodulation-cell division (RND) family of drug transporters are able to decrease the susceptibility of the pathogen to structurally unrelated antibiotics (2). One of these systems, named MexEF-OprN, is quiescent in wild-type strains grown under standard laboratory conditions. Its contribution to the intrinsic resistance of P. aeruginosa is therefore minimal. In contrast, in so-called nfxC mutants, stable overproduction of the pump results in a significant increase in the MICs (4- to 16-fold) of chloramphenicol, trimethoprim, and fluoroquinolones (3). Compared with the susceptibility of wild-type strains, typical nfxC mutants exhibit a hypersusceptibility to some antipseudomonal β-lactams (penicillins, cephalosporins) and aminoglycosides, a phenotype possibly due to the impaired activity of two other RND pumps, namely, MexAB-OprM and MexXY/OprM (4). Furthermore, this typical NfxC phenotype includes a decreased susceptibility to carbapenems, linked to the downregulation of the oprD gene, which codes for the specific porin OprD, allowing the facilitated diffusion of these antibiotics into the cell (3).

In P. aeruginosa, while most RND pumps have their expression modulated by repressors (5), transcription of the mexEF-oprN operon is controlled by a LysR-type activator, MexT, encoded by an adjacent gene (6). In some drug-susceptible laboratory strains of the PAO1 lineage, mexT is inactivated by an 8-bp insert (7). Spontaneous excision of this intragenic fragment restores the open reading frame of mexT with the concomitant overexpression of mexEF-oprN and the development of the typical NfxC phenotype (6). In other strains, mexEF-oprN transcription is triggered by mutations in another gene, mexS, which is divergently transcribed from mexT and encodes an oxidoreductase (8). In any case, a functional MexT is mandatory for the in vitro selection of MexEF-OprN-overproducing mutants. This regulator has been reported to increase mexS expression (6), even if the consensus nod-box DNA sequence, considered the binding site of MexT, remains to be identified in the promoter region of mexS (9). To explain the MexS/MexT-dependent regulation of mexEF-oprN, it has been postulated that MexS is involved in the detoxification of some endogenously produced MexT-activating molecule(s) (10, 11). In this scenario, if it is not processed by MexS, the toxic metabolite(s) would be exported out of the cell by MexEF-OprN as a rescue mechanism.

In clinical strains, nfxC mutations are difficult to characterize because of polymorphic variations in the MexS and MexT protein sequences (http://pseudomonas.com). Moreover, data suggest that still uncharacterized pathways might influence mexEF-oprN expression (12). Supporting this notion, in vitro mutants with alterations in the mvaT, ampR, or mxtR gene have been reported to overexpress mexEF-oprN and to exhibit a multidrug resistance phenotype (13–15). However, the relevance of such mutations in clinical strains awaits confirmation.

In vitro-selected nfxC mutants were found to be deficient in the production of several quorum-sensing-dependent virulence factors (16) without an apparent loss of fitness (17). The mutants derived from reference strain PAO1 typically produce less pyocyanin, rhamnolipids, and elastase than the wild-type parents (3, 16) and less type III secretion system (T3SS) effector toxin ExoS (18). This phenotype was attributed to (i) reduced intracellular levels of the Pseudomonas quinolone signal (PQS), caused by a shortage of a metabolic precursor (kynurenine or 4-hydroxy-2-heptylquinoline [HHQ]) exported by the pump (17, 19), and (ii) MexT acting as a global regulator and indirectly impairing the T3SS in an MexEF-OprN-independent way (18).

Information about the rates and traits of nfxC mutants in cystic fibrosis (CF) patients (20, 21) and non-CF patients (12, 22–24) remains scarce. As a plausible explanation, the low virulence of these mutants would be detrimental to their survival in the host or in the hospital setting and would account for their infrequent isolation from clinical samples. Alternatively, these mutants would be phenotypically and genetically distinct from their in vitro counterparts (i.e., they would keep some degree of pathogenicity or persistence) and thus would be underrecognized. In this study, we show that most clinical nfxC mutants have mild defects in MexS and are still able to produce substantial amounts of virulence factors.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The reference strains and cloning plasmids used in this study are listed in Table 1. Twenty-two clinical nfxC mutants collected between May 2012 and May 2013 at the University Hospital of Besançon, Besançon, France (see Table S1 in the supplemental material), and 7 drug-susceptible strains of P. aeruginosa collected from surface waters (PE1, PE1346, PE1361, PE1393, PE1423, PE1446, and PE1450) were also investigated. All the bacterial cultures were grown in Mueller-Hinton broth (MHB) with adjusted concentrations of Ca²⁺ (range, 20 to 25 mg liter⁻¹) and Mg²⁺ (range, 10 to 12.5 mg liter⁻¹) (Becton Dickinson and Company, Cockeysville, MD) or on Mueller-Hinton agar (MHA; Bio-Rad, Marnes-la-Coquette, France). Escherichia coli transformants were selected on MHA containing 50 μg ml⁻¹ kanamycin (a marker of the vector pCR-Blunt), 15 μg ml⁻¹ tetracycline (a marker of the vector mini-CTX1), or 50 μg ml⁻¹ streptomycin (a marker of the vector pKNG101). Recombinant plasmids were introduced into P. aeruginosa strains by triparental matings and mobilization with broad-host-range vector pRK2013 in E. coli HB101 as a helper strain (25). Transconjugants were selected on Pseudomonas isolation agar (PIA; Becton, Dickinson and Company) supplemented with 200 μg ml⁻¹ tetracycline or 2,000 μg ml⁻¹ streptomycin, as required. Excision of pKNG101 was obtained by selection on M9 minimal medium (8.54 mM NaCl, 25.18 mM NaH₂PO₄, 18.68 mM NH₄Cl, 22 mM KH₂PO₄, 2 mM MgSO₄, pH 7.4) supplemented with 5% sucrose and 0.8% agar.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant characteristics	Source or reference
Strains
Pseudomonas aeruginosa
PAO1	Wild-type reference strain PAO1-UW (University of Washington)	B. Holloway
PAO7H	nfxC mutant derived from wild-type strain PAO1-UW	3
PA14	Wild-type reference strain PA14	B. Ausubel
PA14ΔS	PA14 with a mexS (nfxC-type) deletion	This study
PA14ΔT	PA14 with a mexT deletion	This study
PA14ΔSPA14	PA14 ΔmexS trans-complemented with mexS from reference strain PA14	This study
PA14ΔSPAO1	PA14 ΔmexS trans-complemented with mexS from reference strain PAO1	This study
PA14ΔS1307	PA14 ΔmexS trans-complemented with mexS from clinical strain 1307	This study
PA14ΔS2310	PA14 ΔmexS trans-complemented with mexS from clinical strain 2310	This study
PA14ΔS2505	PA14 ΔmexS trans-complemented with mexS from clinical strain 2505	This study
PA14ΔS3005	PA14 ΔmexS trans-complemented with mexS from clinical strain 3005	This study
PA14ΔS0911	PA14 ΔmexS trans-complemented with mexS from clinical strain 0911	This study
PA14ΔS1009	PA14 ΔmexS trans-complemented with mexS from clinical strain 1009	This study
PA14ΔS0801	PA14 ΔmexS trans-complemented with mexS from clinical strain 0801	This study
PA14ΔS1409	PA14 ΔmexS trans-complemented with mexS from clinical strain 1409	This study
PA14ΔS2311	PA14 ΔmexS trans-complemented with mexS from clinical strain 2311	This study
PA14ΔS2609	PA14 ΔmexS trans-complemented with mexS from clinical strain 2609	This study
PA14ΔS1709	PA14 ΔmexS trans-complemented with mexS from clinical strain 1709	This study
PA14ΔS1711	PA14 ΔmexS trans-complemented with mexS from clinical strain 1711	This study
PA14ΔS0607	PA14 ΔmexS trans-complemented with mexS from clinical strain 0607	This study
PA14ΔT0810	PA14 ΔmexT trans-complemented with mexT from clinical strain 0810	This study
PA14ΔT1510	PA14 ΔmexT trans-complemented with mexT from clinical strain 1510	This study
Escherichia coli
CC118	Δ(ara-leu) araD ΔlacX74 galE galK phoA20 thi-1 rpsE rpoB argE(Am) recA1	43
CC118λpir	CC118 lysogenized with λpir phage	44
DH5α	F− supE44 endA1 hsdR17(rK− mK−) thi-1 recA1 Δ(argF-lacZYA)U169 ϕ80dlacZΔM15 phoA gyrA96 relA1 deoR λ−	Invitrogen
HB101	supE44 hsd(rB− mB−) recA13 ara-14 proA2 lacY1 galK2 rpsL20 xyl-5 mtl-1 leuB6 thi-1	45
Plasmids
pCR-Blunt	Cloning vector for blunt-end PCR products, lacZΔColE1 f1 ori Apr Kmr	Invitrogen
pRK2013	Helper plasmid, ColE1 ori Tra+ Mob+ Kmr	25
mini-CTX1	Self-proficient integration vector, tet Ω-FRT-attP-MCS ori int oriT Tcra	33
pKNG101	Suicide vector in P. aeruginosa, sacB Smr	32

^aFRT, FLP recombination target; MCS, multiple-cloning site.

Antibiotic susceptibility testing.

The MICs of selected antibiotics were determined by the standard serial 2-fold dilution method in MHA with an inoculum of 10⁴ CFU per spot, according to CLSI recommendations (26). Growth was assessed visually after 18 h of incubation at 37°C.

RT-qPCR experiments.

Specific gene expression levels were measured by real-time quantitative PCR (RT-qPCR) after reverse transcription, as described previously (27). Briefly, 2 μg of total RNA was reverse transcribed with ImProm-II reverse transcriptase as specified by the manufacturer (Promega, Madison, WI). The amounts of specific cDNA were assessed in a Rotor Gene RG6000 instrument (Qiagen, Courtaboeuf, France) by using a QuantiFast SYBR PCR green kit (Qiagen). When primers were not already published, the primers used for amplification were designed from the gene sequences available in the Pseudomonas Genome Database, version 2, by using primer3 software (http://bioinfo.ut.ee/primer3-0.4.0/) (see Table S2 in the supplemental material). For each strain, the mRNA levels of the target genes were normalized to those of the rpsL housekeeping gene and expressed as a ratio to the level in wild-type reference strain PA14. Mean gene expression values were calculated from two independent bacterial cultures, each of which was assayed in duplicate. Strain PA14ΔS was used as a positive control for mexE gene overexpression. As shown in preliminary experiments, all mexE transcript levels ≥20-fold above the mexE transcript level of PA14 were associated with a decreased susceptibility (≥2-fold) of the strains to MexEF-OprN substrate antibiotics and considered significant.

Virulence factor analysis.

Biofilm production was assessed by measuring bacterial adhesion to 96-well polystyrene plates (28). Cultures were incubated in triplicate in MHB medium overnight at 30°C and washed twice with 200 μl of distilled water to eliminate planktonic bacteria. Attached bacteria were colored by 1% (wt/vol) crystal violet and solubilized by 99% (vol/vol) ethanol. Attachment was evaluated at 600 nm and considered negative when the optical density (OD) was <1, as previously reported (19).

Swarming motility was tested on freshly prepared M8 medium (42.2 mM Na₂HPO₄, 22 mM KH₂PO₄, 7.8 mM NaCl, pH 7.4) supplemented with 2 mM MgSO₄, 0.5% casein, 0.5% agar, and 1% glucose (29). After 15 min of incubation of the plates at 37°C, 5 μl of culture (7.5 × 10⁵ CFU) was spotted onto the medium surface, with strain PA14 being used as a positive control. The formation of dendrites after 24 h of culture at 37°C was considered a positive result, while a steady spot was considered negative.

Elastase activity was assessed by using MHA plates supplemented with 4 mg ml⁻¹ elastin-Congo red (Sigma-Aldrich, St. Louis, MO) and inoculated with 5-μl volumes of bacterial suspension (7.5 × 10⁵ CFU). Enzymatic degradation of the substrate formed clear halos around the culture spots after 48 h of incubation at 37°C (30). The absence of a visible halo was considered a negative result.

Rhamnolipid production was appreciated using a hemolysis assay. Briefly, after 18 h of growth at 37°C in agitated MHB (A₆₀₀ = 6.7 ± 0.4), bacterial supernatants containing rhamnolipids were collected and mixed with defibrinated horse blood diluted 1/100 (vol/vol) in phosphate-buffered saline. After 30 min of incubation at room temperature, the mixture was centrifuged for 10 min at 950 × g. The concentration of hemoglobin in the supernatants was determined spectrophotometrically at 405 nm. OD values were expressed as the percent hemolysis relative to the complete hemolysis achieved with Triton X-100 (by definition, 100%). The results presented are mean values from two independent experiments. Hemolytic activity was considered to be significantly reduced when it was less than 50% of that for the control.

Finally, pyocyanin assays were carried out on culture supernatants after 18 h of growth at 37°C in a specific broth [120 mM Tris HCl, pH 7.2, 0.1% tryptone, 20 mM (NH₄)₂SO₄, 1.6 mM CaCl₂, 10 mM KCl, 24 mM sodium citrate, 50 mM glucose] (A₆₀₀ = 1.6 ± 0.2). The pigment was extracted from the cultures with chloroform (1 volume) and mixed with 0.1 M HCl (0.06 volume) before spectrophotometric measurement at 520 nm (31). Results are mean values from two independent experiments. Pyocyanin production was considered to be significantly reduced when it was less than 50% of that of reference strain PA14 grown under the same conditions.

The virulence factor production of individual clinical strains was rated by a global score ranging from 0 to 5, which corresponds to the number of positive or significant results obtained by each of the 5 assays mentioned above, the result of each of which was given a value of 1 if it was positive or significant.

Construction of deletion mutants from strain PA14.

Single mexS and mexT deletion mutants were constructed by using overlapping PCRs and recombination events, as described by Kaniga et al. (32). First, the 5′ and 3′ regions flanking mexS (417 and 433 bp, respectively) and mexT (408 and 453 bp, respectively) were individually amplified by PCR with specific primers (see Table S2 in the supplemental material) under the following conditions: 3 min of denaturation at 98°C followed by 30 cycles of amplification, each of which was composed of 10 s at 98°C, 30 s at 60°C, and 30 s at 72°C, and a final extension step of 7 min at 72°C. The resultant amplicons were used as the templates for overlapping PCRs with external pairs of primers to generate the mutagenic DNA fragments. The reaction mixtures contained 1× iProof HF master mix, 3% dimethyl sulfoxide, and 0.5 μM each primer (Bio-Rad). The amplified products were cloned into plasmid pCR-Blunt according to the manufacturer's instructions (Invitrogen, Carlsbad, CA) and next subcloned as BamHI/ApaI fragments into the suicide vector pKNG101 in E. coli CC118λpir (32). The recombinant plasmids were transferred into P. aeruginosa by conjugation and selected on PIA containing 2,000 μg ml⁻¹ streptomycin. The excision of the undesired pKNG101 sequence was performed by plating transformants on M9 minimal medium plates containing 5% (wt/vol) sucrose and 1% (wt/vol) glucose. Negative selection on streptomycin was carried out to confirm the loss of the plasmid. The allelic exchanges were confirmed by PCR. Nucleotide sequencing experiments confirmed deletion of 826 bp in mexS and 929 bp in mexT, yielding strains PA14ΔS and PA14ΔT, respectively.

Chromosomal complementation with full-length mexS and mexT.

A search for mutations in the mexS and mexT genes, as well as in the mexS-mexT and mexT-mexE intergenic regions, was performed with 43 clinical strains by using the primers listed in Table S2 in the supplemental material. The mutated mexS and mexT genes along with their respective promoter regions were amplified from purified genomic DNA by PCR. The resulting DNA fragments were cloned into plasmid pCR-Blunt and next ligated to BamHI/HindIII-linearized plasmid mini-CTX1 (33). The recombinant plasmids were then transferred from E. coli CC118 to P. aeruginosa strains PA14ΔmexS or PA14ΔmexT by conjugation with subsequent selection on PIA plates containing 200 μg ml⁻¹ tetracycline, to allow their chromosomal insertion into the attB site. Chromosomal integration was confirmed by PCR and sequencing.

RESULTS AND DISCUSSION

Wild-type genes mexS and mexT.

Strains PAO1 (3, 8, 34), PA14 (10), and PAK (35) have alternatively been used as wild-type reference strains in studies on the MexEF-OprN efflux pump. However, the mexS and mexT genes in these strains show a nonsilent sequence polymorphism whose impact on the functionality of the encoded proteins, MexS and MexT, respectively, remained to be clarified. For instance, in addition to carrying an 8-bp intragenic fragment inactivating mexT (7), most of the laboratory strains of the PAO1 lineage differ from PA14 (or PAK) by an aspartic acid residue (D) instead of an asparagine (N) at position 249 (D₂₄₉) in MexS (http://pseudomonas.com).

Because the MexS-D₂₄₉ protein was considered either functional (7) or nonfunctional (35), we deleted mexS in both PA14 and PAO1 and compared the effects of this deletion on mexEF-oprN expression and antibiotic resistance. In PA14, suppression of mexS (strain PA14ΔS) resulted in a strong increase in mexE transcription (427-fold) and in a 16- to 32-fold higher resistance to MexEF-OprN substrates, as in typical nfxC mutants (Table 2). As expected, complementation of PA14ΔS with the PA14 mexS allele (PA14ΔS_PA14) restored the drug-susceptible phenotype. In contrast, the PAO1 mexS allele had virtually no impact on the resistance levels of strain PA14ΔS (strain PA14ΔS_PAO1) and failed to reverse the overexpression of mexE, whose transcripts remained 45-fold more abundant in PA14ΔS in comparison with PA14 mexS allele. Consistent with PAO1 producing inactive MexS-D₂₄₉ and MexT peptides, spontaneous excision of the extra 8-bp sequence inserted in mexT is known to trigger MexEF-OprN production in this strain, with MexT recovering its functionality in a nonfunctional MexS background (36). Confirming that MexS-N₂₄₉ (and not MexS-D₂₄₉) is functional, analysis of 7 drug-susceptible strains of P. aeruginosa collected from surface waters (PE1, PE1346, PE1361, PE1393, PE1423, PE1446, and PE1450) showed that the genomes of all of them encoded MexS-N₂₄₉ together with an active MexT (without any insertion in the mexT gene) (data not shown). Based on these results, we therefore used strain PA14 instead of PAO1 in further experiments to investigate the functionality of MexS and MexT from clinical nfxC mutants.

TABLE 2.

Genotypes and resistance profiles of nfxC mutants

Strain	Sequence		mexE transcript levelc	MIC (μg ml−1)d
	MexS or mexSa	MexT or mexTb		CHL	CIP	IMP	TIC	AMK
Reference strains
PA14	WT	WT	1	64	0.12	1	16	2
PA14ΔS	Δ809 bp (bp 1–809)	WT	427	2,048	2	2	8	0.5
PA14ΔSPA14	WT	WT	1.9	64	0.12	1	16	2
PA14ΔSPAO1	N249D	WT	87	2,048	2	2	8	1
PA14ΔT	WT	Δ883 bp (bp 32–915)	0.4	64	0.12	1	16	2
PA14ΔTPA14	WT	WT	6.2	64	0.12	1	16	2
PAO1	N249D	+8 bp (at bp 118)	0.2	16	0.12	1	16	8
PAO7H	N249D	WT	265	2,048	2	4	8	4
Clinical strains with no mutation in mexS and mexT
2502	WT	WT	35	256	0.25	8	64	2
1206	WT	WT	41	512	0.5	2	64	4
0708	WT	WT	53	256	0.25	1	64	4
0309	WT	WT	28	128	1	1	64	2
2607	WT	WT	39	256	0.5	2	64	4
0712	WT	WT	325	2,048	2	4	16	16
0608	WT	WT	25	256	0.25	2	64	8
Clinical strains with mutations in mexS
1307	V104A	WT	29	256	16	2	64	8
2310	F253L	WT	183	1,024	0.5	4	32	8
2505	D44E	WT	212	512	64	16	32	4
3005	S60F	WT	259	2,048	1	4	8	4
0911	F185L	WT	133	1,024	0.5	4	8	2
1009	V73A + L270Q	WT	312	1,024	32	8	128	8
0801	C245G	WT	81	256	0.5	4	128	64
1409	A166P	WT	179	1,024	2	4	8	2
2311	S60P	WT	455	1,024	1	4	4	2
2609	L263Q	WT	534	2,048	1	4	8	2
1709	Δ8 bp (bp 710–718)	WT	552	2,048	2	4	8	2
1711	ΔC293	WT	825	32	1	2	32	8
0607	Δ30 bp (bp 927–956)	WT	556	512	1	8	4	2
Clinical strains with mutations in mexT
0810	WT	G258D	254	512	8	4	128	2
1510	WT	Y138D + G258D	20	256	0.25	16	128	1

^aMexS (339 aa) of PA14 is functional (N₂₄₉) and is considered the wild type (WT), contrary to PAO1-UW (D₂₄₉) (www.pseudomonas.com). aa, amino acid.

^bMexT (304 aa) of PA14 is functional and is considered the wild type, contrary to PAO1-UW (+8 bp [starting at bp 118]) (www.pseudomonas.com).

^cExpressed as a ratio to that of wild-type reference strain PA14. nfxC mutants (the values for which are in bold) have a transcript level of ≥20.

^dCHL, chloramphenicol; CIP, ciprofloxacin; IMP, imipenem; TIC, ticarcillin; AMK, amikacin.

Selection of clinical nfxC mutants.

We screened a collection of 221 clinical isolates of P. aeruginosa exhibiting a reduced susceptibility to ciprofloxacin, a substrate of MexEF-OprN, and imipenem, a substrate of porin OprD, whose expression is inversely coregulated with that of MexEF-OprN (6). The ciprofloxacin and imipenem concentrations used in the screening were equal to the MIC values for reference strain PA14 (0.12 μg ml⁻¹ and 1 μg ml⁻¹, respectively; Table 2). As resistance to these antibiotics may also be due to other efflux pumps (e.g., MexXY/OprM, MexCD-OprJ, MexAB-OprM) as well as other mechanisms (e.g., fluoroquinolone target alterations, mutational loss of porin OprD), the levels of the mexE transcripts were determined in all the strains by RT-qPCR. Forty-three (19.5%) of the 221 selected isolates were found to significantly overexpress mexE (≥20-fold) compared with the level of mexE expression by PA14 (data not shown). According to available clinical data, these 43 nfxC mutants were involved in the colonization or infection of 17 patients (from 1 to 12 isolates per patient) admitted to various medical and surgical units of University Hospital of Besançon, Besançon, France (see Table S1 in the supplemental material). Most of these patients (12/17) were treated with antibiotics prior to the isolation of the nfxC mutant strains, including 7/12 treated with fluoroquinolones known to easily select nfxC mutants (22, 37). Sequencing of the mexS and mexT genes (data not shown) allowed us to identify the redundant mutants in individual patients and to eventually retain 22 strains (1, 2, or 3 different strains per patient) for further investigations (see Table S1 in the supplemental material).

Drug susceptibility of clinical nfxC mutant isolates.

The level of overexpression of the mexE gene was found to vary greatly among the 22 clinical mutants (from 20- to 825-fold the level of expression by PA14; Table 2). These elevated values were associated with an increased resistance of the strains (except strain 1711) to the MexEF-OprN substrates chloramphenicol (MIC range, 2- to 32-fold the MIC for PA14) and ciprofloxacin (MIC range, 2- to 512-fold the MIC for PA14), though the possibility that additional mechanisms may have influenced the drug MICs cannot be excluded. For unclear reasons, one strain, 1711, turned out to be more susceptible (2-fold) to chloramphenicol than PA14, despite the strong upregulation of its mexE gene (825-fold). As indicated in Table 2, 20/22 strains exhibited a 2- to 16-fold decrease in susceptibility to imipenem compared with that of PA14, possibly due to the MexT-dependent downregulation of specific porin OprD (6), a mutational loss of OprD, and/or carbapenemase production (38). Finally, the reported hypersusceptibility of typical in vitro nfxC mutants to the MexAB-OprM substrate ticarcillin and to the MexXY(OprM) substrate amikacin (4) was observed in only 7 strains and 1 strain, respectively, suggesting that this hypersusceptible phenotype either arises in specific genetic backgrounds, such as the PAO1 and PA14 backgrounds, or is masked in most clinical nfxC strains by additional mechanisms. It should be noted that because of this phenotypic variability, MexEF-OprN-overproducing mutants may be difficult to recognize in the medical laboratory unless molecular biology techniques are used.

Amino acid variations in the MexT regulator.

MexT needs to be functional to activate mexEF-oprN operon expression in nfxC mutants (6). Concordant with this, DNA sequencing revealed that 20/22 strains (91%) produced a MexT protein identical to that of PA14 (Table 3). Interestingly, 2/22 strains (9%) harbored mexT genes with point mutations resulting in one (G₂₅₈D; strain 0810) or two (Y₁₃₈D and G₂₅₈D; strain 1510) amino acid substitutions in the effector-binding domain of MexT. In these isolates, the sequence of mexS, as well as the sequences of the mexS-mexT and mexT-mexE intergenic regions, was identical to that of PA14 (Table 2). To investigate the impact of the G₂₅₈D substitution and the Y₁₃₈D plus G₂₅₈D substitutions on MexT activity, we complemented PA14ΔT with the mexT alleles from strains 0810 and 1510. The expression of the mexE gene, the drug resistance, and the virulence factor score of PA14ΔT were unaffected by the complementation (data not shown), indicating that neither Y₁₃₈D nor G₂₅₈D influences MexT activity, as the mutational activation of MexT would have induced mexEF-oprN expression in a functional MexS background. Also, these results indirectly imply that still unknown mutations are involved in the NfxC phenotype of isolates 0810 and 1510.

TABLE 3.

Genotypes and phenotypes of PA14ΔS complemented with mutated mexS alleles from clinical isolates

Strain	MexS sequence (339 aaf)	Transcript levela						MIC (mg ml−1)b						Virulence factor activityc
		mexE	mexS	mexT	oprD	mexB	mexY	CHL	CIP	IMP	MEM	TIC	AMK	Biofilm formation (OD600)	Swarming motility	Elastase production (mm)	Hemolytic activityd (%)	Pyocyanin productiond (%)	Virulence scoree
PA14	WTg	1	1	1	1	1	1	64	0.12	1	0.5	16	2	2.6	+	18	72	100	5
PA14ΔS	Δ809 bp (aa 1–809)	427	NDh	0.9	0.3	0.4	0.4	2,048	4	2	1	8	0.5	0.4	−	12	22	13	1
PA14ΔSPA14	WT	1.9	1.3	1.5	1	1	1	64	0.12	1	0.5	16	2	2.7	+	18	79	92	5
PA14ΔS1307i	V104A	0.8	1.2	1.4	1.8	1.1	1.2	64	0.12	1	0.5	16	2	4.4	+	16	70	114	5
Strains with mild substitutions in MexS
PA14ΔS2310	F253L	21	4	1	0.9	1	0.8	256	0.5	1	0.5	16	1	2.8	+	16	70	84	5
PA14ΔS2505	D44E	35	5.2	1.2	0.4	0.7	0.7	512	1	2	1	16	1	1.7	+	17	70	18	4
PA14ΔS3005	S60F	71	4.8	1.2	0.4	1	0.7	512	1	2	1	16	1	1.7	+	16	68	18	4
PA14ΔS0911	F185L	59	3.7	1.5	0.6	1.4	0.8	512	1	2	1	16	1	2.2	+	17	58	40	4
PA14ΔS1009	V73A + L270Q	43	5.8	1.4	0.5	1.2	0.8	512	1	2	1	16	1	1.8	+	16	53	21	4
PA14ΔS0801	C245G	83	4	1.7	0.5	1.1	0.8	512	1	2	1	16	1	2.3	+	16	71	70	5
PA14ΔS1409	A166P	83	5.1	1.9	0.4	1.2	0.8	1,024	2	2	1	16	1	1.6	+	16	67	22	4
PA14ΔS2311	S60P	230	9.1	1.7	0.4	1	0.8	1,024	2	2	1	16	1	1.2	+	16	59	33	4
Strains in which MexS is inactivated
PA14ΔS2609	L263Q	320	6.9	1.1	0.2	0.8	0.5	2,048	4	2	1	8	0.5	0.2	+/−	16	21	14	1
PA14ΔS1709	Δ8 bp (aa 710–718)	334	5.1	1.7	0.5	0.6	0.4	2,048	4	2	1	8	0.5	0.6	−	13	32	17	1
PA14ΔS1711	ΔC293	404	2.6	1.7	0.7	0.6	0.6	2,048	4	2	1	8	0.5	0.7	−	11	37	13	1
PA14ΔS0607	Δ30 bp (aa 927–956)	492	8	1.9	0.5	0.8	0.6	2,048	4	2	1	8	0.5	0.6	−	13	20	21	1

^aExpressed as a ratio relative to that of wild-type reference strain PA14. nfxC mutants (the values for which are in bold) have a mexE transcript level of ≥20.

^bCHL, chloramphenicol; CIP, ciprofloxacin; IMP, imipenem; MEM, meropenem; TIC, ticarcillin; AMK, amikacin.

^cThe results for virulence factors are in bold when they are positive or considered significant. OD₆₀₀, OD at 600 nm.

^dHemolytic activity and pyocyanin production were measured on stationary-phase cultures after 18 h of growth (see Materials and Methods for details).

^eThe score was determined from the results of each test (biofilm formation, swarming motility, elastase production, hemolytic activity, pyocyanin production), positivity by each of which was given a value of 1. Of note, swarming motility and biofilm formation are correlated with rhamnolipid production.

^faa, amino acid.

^gWT, wild type.

^hND, not detected.

ⁱThe strain tolerated substitutions in MexS.

Impact of alterations in mexS on resistance and virulence.

Experimental results from our university laboratory (unpublished data) and of other research groups (8) indicate that most (77% and 66%, respectively) nfxC mutants selected in vitro on MexEF-OprN substrates, such as ciprofloxacin and chloramphenicol, harbor nucleotide deletions or insertions in the mexS gene that are predicted to result in inactive MexS peptides. In the present study, intriguingly, only 13.6% (n = 3/22) of the strains turned out to carry such indels in mexS, whereas 45.5% (n = 10/22) exhibited point mutations resulting in one (n = 9 strains) or two (n = 1 strain) amino acid substitutions in the MexS oxidoreductase. The remaining 41% (n = 9/22) harbored a PA14-like, wild-type MexS (Table 2). The latter 9 strains were found to produce a MexT identical to that of PA14 (n = 7) or harbor the nonsignificant amino acid variations Y₁₃₈D and G₂₅₈D (n = 2; see above). As the mexS-mexT and mexT-mexE intergenic regions were 100% identical between the 9 strains and PA14, these results unambiguously demonstrate that mutations in still unknown loci (other than mexS and mexT) are able to upregulate mexEF-oprN expression in clinical strains.

Because amino acid substitutions may have less dramatic effects on MexS activity than disruption of the mexS gene, we cloned the 13 mutated mexS alleles in plasmid mini-CTX and complemented mutant PA14ΔS by chromosomal insertion of the cloned genes into the attB site. RT-qPCR experiments showed that all the transconjugants except one (complemented with MexS-V₁₀₄A from strain 1307) significantly overexpressed the efflux operon (Table 3). The mexE mRNA levels were significantly correlated (Spearman's rho = 0.96, P < 0.01) with the ciprofloxacin MICs (Fig. 1A). As expected, complementation with indel-carrying mexS alleles (from strains 1709, 1711, and 0607) failed to decrease the expression levels of mexE or the MICs of chloramphenicol (2,048 μg ml⁻¹) and ciprofloxacin (4 μg ml⁻¹) (compared with those for PA14ΔS; Table 3). Wild-type levels of resistance to imipenem, ticarcillin, and amikacin were also not restored in the null mutant upon complementation. Similar results were obtained with MexS-L₂₆₃Q (from strain 2609), supporting the notion that this mutation is strongly detrimental to MexS activity. As an indication that the remaining mutations (except the well-tolerated variation V₁₀₄A from strain 1307) partially compromise but do not abolish MexS activity, complementation of PA14ΔS with other MexS variants reduced the level of mexE expression from 1.9- to 20.3-fold and its level of resistance to both chloramphenicol and ciprofloxacin from 2- to 8-fold (Table 3). This was accompanied by the restoration of wild-type susceptibility to imipenem, ticarcillin, and amikacin in 1 (strain 2310 with MexS-F₂₅₃L), 8, and 0 complemented mutants, respectively. Upon complementation with MexS-F₂₅₃L, gene oprD expression was increased to wild-type levels (0.9-fold that of PA14), while mexE expression was strongly repressed (20.3-fold that of PA14), providing further evidence that F₂₅₃L only weakly affects MexS activity. The mRNA levels of the oprD gene in PA14ΔS were not impacted or were only marginally impacted by expression of disrupted mexS genes or alleles encoding amino acid variations other than V₁₀₄A and F₂₅₃L, consistent with the unchanged resistance of PA14ΔS to imipenem upon complementation. The hypersusceptibility of in vitro nfxC mutants to some β-lactam antibiotics, such as ticarcillin, has been proposed to result from nfxC-dependent repression of the mexAB-oprM operon (39). As shown in Table 3, inactivation of MexS by indels or L₂₆₃Q failed to restore wild-type ticarcillin susceptibility in transcomplemented strain PA14ΔS (for which the ticarcillin MIC was 2-fold lower than that for PA14), while less severe alterations did. Nevertheless, mexB expression was not significantly different among the transcomplemented mutants, with the level of expression by the mutants ranging from 0.6- to 1.4-fold that by PA14, which suggests the existence of more complex interplays between MexAB-OprM and MexEF-OprN in nfxC mutants, as already evoked (4). None of the mexS alleles except those encoding the V₁₀₄A substitution was able to increase the levels of mexY expression and amikacin MICs up to wild-type levels in PA14ΔS (Table 3). However, the slight effect (a 2-fold increase in the MIC) was visible for strains with alleles with mutations resulting in mild defects but was absent for strains with alleles with mutations resulting in severe defects.

FIG 1.

Correlation between levels of mexE expression, resistance, and virulence factor production in strain PA14ΔS complemented with mexS alleles from 13 clinical isolates. The expression levels of the mexEF-oprN operon, as assessed by RT-qPCR of the mexE gene, are expressed as the ratios to the level of mexEF-oprN expression by wild-type strain PA14 (set at a value of 1, by definition). Ciprofloxacin (Cip) MICs (diamonds) are expressed as ratios relative to the ciprofloxacin MIC for PA14. The virulence factor scores (circles) were determined by the number of positive results by 5 different tests (biofilm formation, swarming motility, elastase production, hemolytic activity, and pyocyanin production), as indicated in Table 3. The negative and positive controls, strains PA14 and PA14ΔS, respectively, are indicated. The relationships between the variables mexE expression and ciprofloxacin MIC (Spearman's rho = 0.96, P < 0.01) and the variables mexE expression and virulence factor scores (Spearman's rho = −0.87, P < 0.01) were found to be significant.

The virulence traits of the transcomplemented PA14ΔS mutants varied greatly according to the different mexS alleles. As for PA14ΔS, mutations leading to the complete inactivation of MexS and strong mexE upregulation (in alleles from strains 2609, 1709, 1711, and 0607) were associated with a low virulence score of 1 (Fig. 1B). Biofilm formation, swarming motility (Fig. 2), hemolytic activity, as well as pyocyanin production remained strongly impaired in the transcomplemented mutants (Table 3). Complementation with the other alleles (from strains 2310, 2505, 3005, 0911, 1009, 0801, 1409, and 2311) partially restored the wild-type virulence traits in PA14ΔS, yielding scores of 4 and 5. However, the level of pyocyanin production remained low in most of these complemented mutants (from 0.2- to 0.8-fold that of PA14ΔS_PA14) and showed no evident correlation with mexE expression levels, as was observed in mutants PA14ΔS₀₈₀₁ and PA14ΔS₁₄₀₉ in Table 3.

FIG 2.

Swarming motility of PA14ΔS complemented with different mutated mexS alleles from clinical isolates. Swarming motility was evaluated as the capacity to give rise to dendrite-like patterns. The patterns for strain PA14 and PA14ΔS_PA14 (positive controls) (A and C, respectively) and PA14ΔS (negative control) (B) are indicated. (F) Inactivation leading to an aberrant MexS protein (in PA14ΔS₁₇₁₁, for example) abolished the swarming. In most cases, substitutions in MexS, for example, V₁₀₄A (D), F₁₈₅L (G), and D₄₄E (H), did not affect the ability of the bacteria to swarm; however, the L₂₆₃Q substitution led to an almost complete loss of motility (E).

Consistent with our previous conclusions, the mexS allele encoding the well-tolerated substitution V₁₀₄A provided PA14ΔS with a wild-type phenotype of resistance and virulence (Fig. 2), indicating that mexEF-oprN overexpression in strain 1307 is due to mexS-independent (and also mexT-independent) genetic events.

Other regulatory genes in nfxC mutants.

As reported above, 10/22 clinical strains (strains 2502, 1206, 0708, 0309, 2607, 0712, 0608, 1307, 0810, and 1510) appeared to produce functional MexS and MexT proteins. Since mutations in genes coding for the global regulators MvaT and AmpR have been reported to activate the mexEF-oprN operon in in vitro-selected nfxC mutants (13, 14), we carried out sequencing experiments, which eventually failed to reveal alterations in these genes. Again, these results clearly indicate that other loci are implicated in pump MexEF-OprN overproduction in the clinical setting.

Conclusion.

This study provides an insight into the genetic events leading to MexEF-OprN overproduction in clinical nfxC isolates. The hypothesis of preferential selection of partially derepressed MexEF-OprN mutants rather than fully derepressed ones in vivo is reinforced by our observation that mexE expression was lower in most clinical nfxC mutants (mean, 205-fold ± 187-fold that of wild-type strain PA14; median, 179-fold that of wild-type strain PA14) than in comparator strain PA14ΔS (427-fold that of wild-type strain PA14) (Table 2). None of the amino acid variations found in the mexT product (2/22 isolates, 9%) proved to be significant, a result consistent with the observation that LysR regulators are rarely constitutively activated by mutations (e.g., BenM in Acinetobacter baylyi and CysB in Salmonella enterica serovar Typhimurium) (40, 41). In contrast, single point mutations in the MexS oxidoreductase (9/22, 40.9%) represent a significant cause of MexEF-OprN upregulation in clinical P. aeruginosa strains. Consistent with these findings, a decrease in ciprofloxacin MICs from 2- to 4-fold was observed in 7 clinical nfxC mutants upon complementation with a plasmid-borne copy of mexS from strain PA14 (see Table S3 in the supplemental material). Of note, another mutation in MexS (A₁₅₅V) leading to multidrug resistance has recently been reported in a clinical isolate (42). Our results demonstrate that most MexEF-OprN-overproducing clinical strains either have a wild-type, PA14-like MexS (10/22, 45.5%) or are only partially deficient in MexS activity (8/22, 36.3%). Mutants harboring these mutations resulting in presumed mild defects display resistance and virulence traits intermediate between those of wild-type strains and strongly defective MexS mutants (4/22, 18.2%), which could account for their emergence in vivo. However, analysis of our clinical strains gave contrasting results (see Table S4 in the supplemental material), reinforcing the idea that the virulence of P. aeruginosa is multifactorial and factors other than those tested in this study may well contribute to the pathogenicity of strongly deficient mexS mutants, some of which were still able to cause infections. The MexEF-OprN overproducers studied here had similar growth rates (data not shown). Finally, this work indirectly demonstrates that still unknown regulators are involved in the activation of mexEF-oprM in 10/22 (45.5%) clinical nfxC mutants. We are currently trying to determine such regulatory pathways.