CzcR-CzcS, a Two-component System Involved in Heavy Metal and Carbapenem Resistance in Pseudomonas aeruginosa*

Pseudomonas aeruginosa is an environmental bacterium involved in mineralization of organic matter. It is also an opportunistic pathogen able to cause serious infections in immunocompromised hosts. As such, it is exposed to xenobiotics including solvents, heavy metals, and antimicrobials. We studied the response of P. aeruginosa upon exposure to heavy metals or antibiotics to investigate whether common regulatory mechanisms govern resistance to both types of compounds. We showed that sublethal zinc concentrations induced resistance to zinc, cadmium, and cobalt, while lethal zinc concentrations selected mutants constitutively resistant to these heavy metals. Both zinc-induced and stable zinc-resistant strains were also resistant to the carbapenem antibiotic imipenem. On the other hand, only 20% of clones selected on imipenem were also resistant to zinc. Heavy metal resistance in the mutants could be correlated by quantitative real time PCR with increased expression of the heavy metal efflux pump CzcCBA and its cognate two-component regulator genes czcR-czcS. Western blot analysis revealed reduced expression of the basic amino acid and carbapenem-specific OprD porin in all imipenem-resistant mutants. Sequencing of the czcR-czcS DNA region in eight independent zinc- and imipenem-resistant mutants revealed the presence of the same V194L mutation in the CzcS sensor protein. Overexpression in a susceptible wild type strain of the mutated CzsS protein, but not of the wild type form, resulted in decreased oprD and increased czcC expression. We further show that zinc is released from latex urinary catheters into urine in amounts sufficient to induce carbapenem resistance in P. aeruginosa, possibly compromising treatment of urinary tract infections by this class of antibiotics.

Pseudomonas aeruginosa is a Gram-negative bacterium thriving in environments polluted with organic matter. It is also an opportunistic pathogen frequently encountered in the hospital, causing morbidity and mortality in immunocompromised and cystic fibrosis patients (1). P. aeruginosa is characterized by an intrinsically high level of resistance to xenobiotics including antimicrobial agents, solvents, and heavy metals (2), which can be accounted for by a combination of its low outer membrane permeability and the presence of multiple efflux pumps (3). These pumps belong to the resistance, nodulation, cell division (RND) 1 transporter family, present in many Gramnegative bacteria (4). To extrude substrates from the cytoplasm across the two membranes, these systems are composed of a proton antiporter located in the cytoplasmic membrane, a membrane fusion protein spanning the periplasmic space, and an outer membrane protein (5). Members of the RND family, namely the Mex pumps, have recently gained increasing interest. In particular, the constitutively expressed MexAB-OprM (6,7) and the inducible MexXY (8) efflux pumps endow the PAO1 reference strain and other clinical isolates (9) with a natural resistance to a wide range of antimicrobial agents. Proton-driven RND type efflux pumps conferring heavy metal resistance have been described in Ralstonia metallidurans (for a recent review, see Ref. 10) and include the Cnr system (nickel/cobalt) (11), the Ncc system (nickel/cobalt/cadmium) (12), and the Czc system (cobalt/zinc/cadmium) (13,14). In P. aeruginosa, an RND type efflux pump called CzrCBA was recently described in an environmental isolate where it contributes to the intrinsic resistance to zinc and cadmium (15). Cross-resistance between heavy metal and antibiotic pumps has not been reported so far. In the few cases where associations have been observed they were either plasmid-mediated (16) or resulted from uncharacterized multiple resistance mechanisms (17,18). In P. aeruginosa, the question of heavy metal and antibiotic resistance is of particular concern since this organism is a possible candidate for bioremediation processes where selection of antibiotic resistance upon heavy metal exposure is undesirable. On the other hand, zinc was found to be released from urinary catheters resulting in antibiotic resistance (19,20). Therefore the possibility of cross-resistance selection by either heavy metals or antibiotics is of concern for both environmental and clinical issues. In the present study, we addressed this question by exposing the P. aeruginosa reference strain PAO1 to either zinc or to the antibiotic imipenem. Surprisingly exposure to zinc selected strains that were resistant to both heavy metals (zinc, cadmium, and cobalt) and to the carbapenem antibiotic imipenem. Analysis of the underlying mechanism revealed a co-regulation between carbapenem influx and heavy metal efflux. A single amino acid change located in the two-component sensor protein CzcS, regulating heavy metal efflux pump expression, was found to be responsible for the observed cross-resistance in P. aeruginosa.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Bacterial strains used in this study are listed in Table I. Strain BdB4 was isolated from soil of a heavy metal-contaminated site (zinc and lead) near Geneva (Bois de Bay) by successive subculturing in tryptone-yeast extract-glucose (TYG) medium (21) provided with increasing ZnCl 2 concentrations (10 -35 mM; dose increase, 5 mM). Strain BdB4 was confirmed as P. aeruginosa using an API20NE gallery (Biomérieux, Marcy l'Etoile, France). Luria-Bertani (LB) (22) and TYG medium were used as rich media. Some experiments were performed in liquid mineral medium (13) supplemented with 0.4% glucose and 50 mM sodium Hepes, pH 7.0, instead of Tris buffer. The phosphate content (0.64 mM) of this Hepesbuffered minimal medium minimized the interference of phosphate with heavy metals. Cultures were grown at 37°C on a rotary shaker in 200-ml Erlenmeyer flasks containing 25 ml of minimal medium.
Isolation of Spontaneous Mutants Resistant to Heavy Metals or Imipenem-For selection of zinc-resistant mutants, PT5 was inoculated at a final concentration of about 2 ϫ 10 7 cells/ml in 5 ml of liquid TYG medium in test tubes that were incubated for 5 days at 30°C. The medium was supplemented with 15 (maximal tolerable concentration (MTC)), 20, 25, or 30 mM ZnCl 2 . From each condition, clones were isolated on TYG medium by serial dilutions. 50 clones were tested for zinc resistance after growth for 48 h at 30°C on TYG plates containing 20 mM ZnCl 2 . 16 independent clones were then tested for antibiotic resistance and tolerance to other heavy metals.
PT5 was also inoculated on Mueller-Hinton agar plates and exposed to imipenem-impregnated disks (10 g, Biomérieux). After overnight incubation at 37°C, 42 imipenem-resistant clones appearing inside the inhibition zone were picked, streaked on TYG medium, and tested for zinc resistance. Resistant mutants were then tested for tolerance to other metals (cadmium, cobalt, copper, and nickel) and antibiotics (imipenem, ticarcillin, carbenicillin, nalidixic acid, ciprofloxacin, norfloxacin, tetracycline, amikacin, chloramphenicol, and polymyxin B).
Determination of Heavy Metal and Antibiotic Resistance-The MTCs of heavy metals were determined on solidified TYG medium containing different concentrations of heavy metal salts. The MTCs were scored in 10 ϫ 10 ϫ 2-cm Sterilin plates (Bibby Sterilin Ltd., Stone, Staffs, UK) provided with 25 compartments. 2-ml of medium were poured into each compartment, plates were air-dried, and subdivisions were inoculated with 10 l of an overnight culture. The MTC is defined as the highest metal concentration at which growth was still observed after 48 h of incubation at 30°C.
Resistance to antibiotics was determined by the Kirby-Bauer method on Mueller-Hinton agar plates using antibiotic-impregnated disks (Biomérieux). An overnight culture grown at 37°C in TYG medium was diluted 1:500 in 0.9% NaCl, and 2 ml of the suspension were spread on Mueller-Hinton agar plates. The plate was air-dried, and antibiotic disks were applied. After 1 h of diffusion at room temperature, plates were incubated overnight at 37°C. Inhibition zones were measured and compared with those of the reference strain PT5.
Killing Curves in Presence of Zinc-PT5 was grown for 48 h at 37°C in phosphate-deficient (0.1 mM) minimal medium to increase bioavailability of heavy metals. Cultures were performed in the absence (uninduced control cells) or in the presence of 1 mM ZnCl 2 (metal-induced cells). Treated cultures were found to exhibit a lag phase of about 24 h followed by a period of active growth. Both cultures were then diluted 1:200 in fresh medium containing either no zinc or a lethal zinc concentration of 5 mM. Cultures were incubated at 37°C with shaking (150 rpm) for 5 h. Kinetics of killing of uninduced versus metal-induced cells was then followed. At defined intervals, samples were removed, diluted in TYG medium, and plated on the same medium for viable counts.
DNA Manipulations-Standard techniques were used for DNA manipulation (22). The gene encoding the CzcR protein was amplified by PCR from PT5 genomic DNA (Pfu DNA polymerase, Promega) with primers czcR-F and czcR-R (see below). The 900-bp product was cloned into the SmaI site of vector pMMB66EH (23) under the tac promoter, yielding plasmid pRWT. The czcS gene from PT5 and PT1105 was amplified by PCR using primers S58 (5Ј-cggaattcgcggcgtcggctacgtcc) and S59 (5Ј-cgggatcctgcggcgagtaccggctgtggc) containing an EcoRI and a HindIII site, respectively. The 1,600-bp product was EcoRI/HindIIIdigested and cloned into the EcoRI/HindIII sites of pMMB66EH under the tac promoter, yielding plasmids pSWT and pSV194L, respectively. Correct orientation of the genes was verified by sequencing. Plasmids were electroporated into P. aeruginosa strain PT5 (24). To delete the czcA gene from strain PT5, a 600-bp PCR fragment corresponding to the 5Ј-end of czcA was generated with primers 48 (5Ј-cccaagcttcgaacgcat-catccaattcg) and 49 (5Ј-gcttcttcggatccggggcg) and ligated to an 800-bp PCR fragment containing the 3Ј-end of czcA generated with primers 50 (5Ј-cgggatcctgttcgagggcgaccgcc) and 51 (5Ј-gctctagatccagcgatagagcaccggc). The gentamicin resistance cassette from plasmid pPS858 (25) was inserted between the two parts in the BamHI site. This construct was cloned as a PCR fragment into the HindIII-cleaved plasmid pEX18Ap (25). After transfer and homologous recombination into strain PT5, excision of the gentamicin cassette was performed as described previously (25). The resulting strain PT1173 carries a 1,700-bp deletion inside the czcA gene as verified by PCR.
Western Blot Analysis-An overnight preculture of P. aeruginosa grown in minimal medium was diluted 100-fold in the same medium and grown at 37°C to an A 600 of 1. Total protein was solubilized by resuspending a bacterial pellet in 1ϫ SDS gel loading buffer. Samples were boiled 5 min and centrifuged 10 min in a microcentrifuge to remove bacterial debris. A duplicate sample was used for protein quantification with the Lowry method (27) on NaOH-solubilized extracts (28). 10 g of total protein were separated on a 12% SDS-polyacrylamide gel and transferred to nitrocellulose membranes. Blots were incubated with anti-OprD, anti-OprF, or anti-Hsp70 antibodies and revealed by chemiluminescence. All antibody incubations and washes were performed in TBS-T (20 mM Tris, 137 mM NaCl, 0.1% Tween 20, pH 7.6) supplemented with 5% powdered milk.
Real Time PCR Analysis-For RNA isolation, strains were cultured in 5 ml of minimal glucose medium described above or in LB medium and grown at 37°C to midexponential growth phase. 0.25 ml of this culture, corresponding to 5 ϫ 10 8 cells, was added to 0.5 ml of RNeasy Protect bacteria solution (Qiagen, Hildesheim, Germany), and total RNA was isolated with RNeasy columns according to the instructions of the supplier. Residual DNA was eliminated by DNase treatment using 20 units of RQ1 RNase-free DNase (Promega). After removal of DNase by phenol/chloroform extraction, RNA was precipitated, and the pellet resuspended in 30 l of RNase-free H 2 O. For cDNA synthesis, 1 g of RNA was reverse-transcribed using random hexamer primers (Promega) and Improm-II reverse transcriptase (Promega) according to the supplier's instructions. Reverse transcriptase was inactivated by incubation at 70°C for 15 min, and the obtained cDNAs were stored at Ϫ20°C until use.
The following primer sequences for the PCR amplification of cDNA were designed using the Primer3 program 2 : czcR-1 (5Ј-gtcatcacccggacgcagatcat) and czcR-2 (5Ј-gtagccgacgccgcgaatggtat), czcS-1 (5Ј-tacgcagctctcgcagttctcc) and czcS-2 (5Ј-tgtccacctgcaccaggaacagc), czcC-1 (5Јggtcagcatcggcagcaagtacg) and czcC-2 (5Ј-ggtcgtaggcctgtaccgcttcg), and rpsL-3 (5Ј-gcaactatcaaccagctggtg) and rpsL-5 (5Ј-gctgtgctcttgcaggttgtg). A RotorGene real time PCR machine (model RG3000, software version 4.6.67) was used for the quantification of cDNA. PCRs were performed using a Sybr Green Quantitect kit (Qiagen, Hilden, Germany) according to the specifications of the supplier. To check for residual contaminating genomic DNA, control reactions without reverse transcriptase were analyzed in the real time PCR apparatus using the rpsL primer set. No amplification signal above the non-template control was detected indicating that the RNA samples were free of contaminating DNA. The cDNA samples were diluted 10-fold, and 3 l of this dilution served as the template in the PCRs that were performed in duplicate for each gene and sample. To correct for differences in the amount of starting material, the ribosomal rpsL gene was chosen as a reference gene. Results are presented as ratios of gene expression between the target gene (target) and the reference gene (rpsL), which were obtained according to the following equation: ratio ϭ (E target gene ) ⌬Ct target(PT5 Ϫ test strain) / (E rpsL ) ⌬Ct rpsL (PT5 Ϫ test strain) (29) where E is the real time PCR efficiency for a given gene and C t the crossing point of the amplification curve with the threshold. An effect on gene transcription was considered significant when the corresponding ratios were Ն2.0 or Յ0.5.

Growth in Latex Urinary
Catheters-Filter-sterilized urine (0.45-m Nalgene 150-ml filters) obtained from healthy volunteers was inoculated with PT5 at a concentration of about 5 ϫ 10 6 bacteria/ml and introduced into latex urinary catheters (Silkolatex Rü sch Gold, Rü sch, Kamunting, Malaysia). Catheters were incubated for 48 h at 37°C in 13-cm Petri dishes. Cell densities reached 1.5 ϫ 10 9 bacteria/ml. Control PT5 bacteria unexposed to latex catheters were grown in urine in test tubes at 37°C. To assay for the MTC of zinc, PT5 cells grown in urine were centrifuged and resuspended in the same volume of TYG medium before inoculation. Zinc concentrations in urine incubated in latex urinary catheters were measured by atomic absorption spectroscopy (Philips-Pye-Unicam SP9). Urine was diluted 1:400 in Milli-Q water before measurements.

Characterization of the CzcCBA Heavy Metal Efflux
Pump from PAO1-An efflux pump of the RND transporter family, called CzrCBA, was recently shown to produce tolerance to cadmium and zinc in the P. aeruginosa strain CMG103 isolated from a metal-polluted river in Pakistan (15). A highly similar efflux pump, annotated as CzcCBA, 3 is present on the completely sequenced genome of the P. aeruginosa reference strain PAO1 (26). Sequence alignment between the czrCBA genes from strain CMG103 (15) and the czcCBA genes in the PAO1 genome (26) revealed a Ͼ99% amino acid identity between the CzrA and CzcA efflux pump proteins. However, sequence variations were present in the N termini of the outer membrane proteins CzrC (CMG103) and CzcC (PAO1). Furthermore the membrane fusion protein CzrB from strain CMG103 lacks a stretch of 31 amino acids present in CzcB from PAO1. This 31-amino acid stretch is present in the CzcB protein from R. metallidurans and is 70% identical to that of PAO1 (data not shown). To analyze the contribution of the CzcCBA system to heavy metal resistance, we inactivated by homologous recombination the efflux pump gene, annotated as czcA (PA2520) in our PAO1 strain called PT5 (Table I). The resulting mutant, termed PT1173, was indeed more susceptible than its parent to zinc and cadmium as well as to cobalt (Table II). A similar increase in susceptibility to these three heavy metals was observed in strain CMG103-13 containing a transposon insertion in the czrCBA region of strain CMG103 (15). With respect to the hypersusceptibility of the czcA deletion mutant PT1173 to zinc, cadmium, and cobalt, we decided to comply with the current annotation of the PAO1 genome 3 and adopted the designation CzcCBA for this heavy metal efflux pump.
To further characterize the CzcCBA pump, we tested its inducibility by heavy metals. Without induction, the PT5 wild type strain reached a MTC of 15, 6, and 2-3 mM for zinc, cadmium, and cobalt, respectively (Table II). After preculturing PT5 in the presence of zinc (1 mM) or cadmium (0.4 mM) the MTCs increased to 30, 8, and 3 mM for zinc, cadmium, and cobalt, respectively (Table II). MTCs of copper and nickel were not affected. A killing rate experiment showed that uninduced PT5 cells were killed after 3 h in the presence of 5 mM ZnCl 2 , while about 50% of cells, preincubated in the presence of 1 mM ZnCl 2 , stayed viable after 5 h of incubation (Fig. 1). Therefore, the induced cells are strongly protected by the CzcCBA efflux pump against a challenge with lethal heavy metal concentrations.
Selection of Mutants Resistant to Either Zinc or Imipenem-We attempted to select stable, heavy metal-resistant mutants from our wild type strain. We therefore exposed PT5 to elevated zinc concentrations (20 -25 mM). Sixteen independent clones displayed resistance to zinc, cadmium, and cobalt with MTCs comparable to those of the induced PT5 strain (data not shown). This resistance phenotype was stable even after several passages in the absence of zinc. Since the CzcCBA pump belongs to the RND transporter family that also includes all the Mex multidrug efflux pumps, we tested the susceptibilities of the 16 zinc-selected mutants to various antibiotics (see "Experimental Procedures"). Surprisingly all the tested mutants displayed increased imipenem resistance, while susceptibility to the other antibiotics was unaffected (data not shown). One zinc-and imipenem-resistant mutant, called PT1108, was selected for further experiments (Table II).
To investigate a possible link between heavy metal and carbapenem resistance, the reciprocal experiment was performed by exposing strain PT5 to imipenem. Spontaneous mutants appearing after 24 h in the inhibition zone around the imipenem disk were picked, and their resistance profiles were analyzed. Two groups of mutants were identified. Members of the first group were resistant only to imipenem (24 of 30 mutants), while those of the second group were resistant to both heavy metals (zinc, cadmium, and cobalt) and imipenem (6 of 30 mutants). One mutant from the first group (PT1102) and one from the second (PT1105) were selected for subsequent experiments. All imipenem-resistant mutants showed resistance levels between those of the wild type and the oprD knockout strain PT364 (Table II). Zinc-and imipenem-resistant mutants, like PT1105 and PT1108, were termed zir. The fact that mutants selected independently on either zinc or imipenem may display the same resistance profile suggested the existence of a common regulatory mechanism in P. aeruginosa.
Expression of OprD in Zinc-induced Strains and in Heavy Metal-resistant Mutants-In P. aeruginosa carbapenem resistance is known to result mainly from mutations affecting expression of the OprD porin (30) involved in the facilitated diffusion of basic amino acids (31) and carbapenem antibiotics (32). We therefore investigated OprD expression in the wild type strain PT5 in the presence of zinc and subsequently in the mutants obtained under zinc or imipenem selection. Western blot analysis revealed that in the wild type strain OprD expression strongly decreased at 3 M ZnCl 2 and was undetectable at 3 See www.pseudomonas.com. 10 M ZnCl 2 ( Fig. 2A). Expression of the major outer membrane protein OprF was not affected. The amount of oprD mRNA was determined by quantitative real time PCR (qRT-PCR). In the presence of 1 M ZnCl 2 , oprD mRNA levels dropped to 60% and decreased to 20% when 10 M ZnCl 2 was added. These results demonstrate that ZnCl 2 has a negative effect on oprD transcription in a dose-dependent manner, resulting in carbapenem resistance in P. aeruginosa. However, heavy metal resistance is not caused by OprD down-regulation per se since PT364, an oprD knock-out mutant, displayed the same heavy metal susceptibility as its parent strain PT5 (Table II). We further investigated by Western blot analysis the amount of OprD in the three strains obtained by either zinc or imipenem selection. A strongly reduced but still detectable amount of OprD was observed in all three strains (Fig. 2B). These results were confirmed by oprD mRNA transcript analysis using qRT-PCR. The level of oprD mRNA in strains PT1102, PT1105, and PT1108 dropped, respectively, to 30, 2, and 24% of those of the wild type. From these data, we conclude that imipenem resistance in the three mutant strains can be directly attributed to decreased oprD expression.

Involvement of CzcCBA Efflux Pump in Heavy Metal
Resistance-We investigated the possible involvement of the CzcCBA efflux system in the heavy metal resistance by qRT-PCR. When wild type strain PT5 was grown in the presence of 5 mM zinc, the amount of czcC mRNA was increased 215-fold (Table III). Thus, zinc is able to strongly induce the czcCBA efflux system. Cadmium was less effective than zinc as an inducer, while, in contrast to R. metallidurans (33), cobalt and nickel had no measurable effect on transcription of czcC (data not shown).
Interestingly zinc not only induced transcription of the czcCBA structural genes but also induced transcription of the twocomponent regulator genes czcR (20-fold) and czcS (4.5-fold), suggesting an autoamplification mechanism for the pump gene expression. As expected, czcC, czcR, and czcS expressions in mutant PT1102, which is resistant to imipenem only, were comparable to those in the wild type (Table III). In contrast, the zir mutants PT1105 and PT1108 showed strongly increased amounts of czcC, czcR, and czcS mRNAs in the absence of zinc. The expression levels for czcC in the mutants were comparable to those resulting from zinc induction in the wild type (Table  III). Hence overexpression of czcCBA upon zinc induction in the wild type and in the zir mutants seems to result in both cases from increased expression of the two-component czcR-czcS regulator genes.
CzcS Protein Is Altered in zir Mutants-The similar phenotype of mutants obtained from either zinc or imipenem selection prompted us to examine the regulatory region of the czc-CBA operon. We thus sequenced, in the two independently obtained zir mutants PT1105 and PT1108, the czcR and czcS genes as well as the intergenic region between czcR and czcC, which should encode the promoter of these two genes. A unique GTG to TTG transversion was detected in the czcS gene from both strains, resulting in a V194L substitution in the sensor protein. This mutation was also found in six other independently obtained zir mutants. Furthermore sequencing of the czcS gene in the environmental derivative BdB4-3, obtained upon zinc exposure of BdB4 (Table II), revealed a GGC to GAC transition, resulting in a G197D substitution in the CzcS kinase protein. Analysis of the CzcS protein with the transmem-  brane prediction program TMHMM-2.0 4 suggested the presence of two membrane-spanning segments (Fig. 3). According to this program, residues Val-194 and Gly-197 are located just outside the second transmembrane segment on the cytosolic side.
To verify that the V194L mutation in the CzcS sensor was responsible for the observed zir phenotype, we cloned the mutated czcS gene as well as the wild type czcS and czcR genes under the control of the inducible tac promoter in plasmid pMMB66EH. When the mutated CzcS protein was overexpressed in the wild type strain PT5, the level of OprD decreased significantly, while overexpression of the wild type CzcS protein had no effect (Fig. 4). In parallel, an increase in heavy metal resistance occurred upon expression of the CzcS protein carrying the V194L mutation, while expression of the normal CzcS protein had no influence (Table IV). MTCs of nickel were not affected. In agreement with decreased OprD expression, resistance to imipenem increased when CzcS(V194L) was expressed in the wild type strain, leading to a decrease of the inhibition zone from 26 to 19 mm (Table IV). Heavy metal and imipenem resistance levels obtained upon plasmid-directed expression of the mutated CzcS protein were lower than those in the mutant strains (PT1105 and PT1108), a fact that can be explained by competition with the wild type CzcS protein encoded on the chromosome. These results clearly demonstrate that the resistance to both heavy metals and imipenem in the selected mutants is due to the V194L mutation in CzcS (zirS mutants).
Effect of CzcR Overexpression-Since czcR transcription was strongly increased in the zirS mutants and upon zinc induction in the wild type strain, we also cloned the wild type czcR gene in vector pMMB66EH. When overexpressed in strain PT5, CzcR reduced OprD expression (Fig. 4) and rendered cells resistant to imipenem (Table IV). Surprisingly CzcR overexpression did not result in increased resistance to zinc, cadmium, or cobalt but instead rendered bacteria more susceptible to these heavy metals (Table III). Thus, overexpression of wild type CzcR per se allows repression of OprD but is not sufficient to cause increased expression of the CzcCBA efflux pump.
Imipenem Resistance Induced by Zinc Release from Urinary Catheters-Recently Conejo et al. (20) showed that zinc eluted from siliconized latex urinary catheters decreases OprD expression in P. aeruginosa causing carbapenem resistance. The authors carried out their experiments using segments of siliconized latex urinary catheters incubated in Mueller-Hinton broth (34). To determine whether zinc was also released from latex urinary catheters (LUCs) into urine, we grew strain PT5 for 48 h at 37°C in LUCs containing filter-sterilized urine. As determined by atomic absorption spectroscopy, zinc reached a concentration of about 1 mM in urine (Table V). This concentration was sufficient to induce wild type cells to resist a subsequent challenge by 20 mM zinc (Table V). This tolerance was accompanied by a 33-fold increased expression of the Czc-CBA efflux system as determined by qRT-PCR (Table V). As expected, zinc released from LUCs decreased OprD expression (Table V) causing carbapenem resistance, suggesting that zincinduced imipenem resistance might persist as long as zinc is released from the LUCs. Using these growth conditions for 6 days at 37°C, we did not detect appearance of any stable mutants resistant to both imipenem and zinc (98 clones tested, data not shown). DISCUSSION In the present study, we report on the characterization of the CzcCBA efflux pump from the P. aeruginosa reference strain PT5. We demonstrate the role played by the CzcS-CzcR twocomponent system in the control of czcCBA expression as well as its surprising involvement in the expression of the OprD porin leading to carbapenem resistance. To our knowledge this is the first report on a co-regulation mechanism between heavy metal and antibiotic resistance.
Our results clearly show that the CzcCBA efflux system in PT5 confers resistance to zinc, cadmium, and cobalt. The reductions in MTCs of zinc, cadmium, and cobalt in the czcA deletion mutant were rather modest compared with the more than 10-fold differences in MTCs conferred by the CzcCBA system in R. metallidurans (35). It cannot be excluded that other transporters present in PT5 may contribute to the intrinsic resistance of this organism to heavy metals. Overexpression  of the heavy metal pumps could be obtained upon exposure to zinc and cadmium (36) or by constitutive expression resulting, as we show here, from mutations in the two-component sensor CzcS. Zinc in the growth medium not only induced the czcCBA efflux pump operon but at the same time transcription of the two-component regulator genes czcR and czcS. This positive autoregulation loop could allow the cells to respond rapidly to the presence of small amounts of heavy metals in the environment.
Analysis of mutants constitutively expressing the CzcCBA pump surprisingly revealed that all of them were also resistant to carbapenem antibiotics. This resulted from decreased expression of the porin OprD mediating facilitated diffusion of this class of antibiotics. However, OprD is not directly involved in heavy metal resistance since an oprD knock-out mutant was as susceptible as the parent to heavy metals. Hence OprD decrease is a secondary effect linked to CzcCBA overexpression. While only 20% of the analyzed strains selected on imipenem were co-resistant to heavy metals, it nevertheless strongly suggests the existence of a common regulatory mechanism that connects czcCBA to oprD expression. This hypothesis is supported by the identification of two mutational events found in the same region of the CzcS sensor kinase in zincselected derivatives from PT5 and from an environmental isolate. Moreover the V194L mutation was selected in PT5 either upon zinc or imipenem exposure (zirS mutants). Therefore, the sensor kinase CzcS via its cognate regulator CzcR is involved in the regulation of both heavy metal and antibiotic resistance. How could the mutations in CzcS lead to constitutive expression of the CzcCBA efflux pump and reduced expression of OprD? According to sequence analysis, CzcS, like EnvZ in Escherichia coli, belongs to class I histidine kinases. In this class the catalytic and ATP-binding domain is separated in two regions (37). The A domain is involved in the dimerization, phosphotransfer, and phosphatase activity, while the B domain binds ATP (38,39). These histidine kinases transfer a phosphoryl group on an aspartate residue in the cognate regulator protein (OmpR in the case of EnvZ). The activated regulator protein then triggers expression of specific target genes. Many histidine kinases also possess phosphatase activities that allow dephosphorylation and hence inactivation of the regulator protein (37). In the case of EnvZ/OmpR, the level of phosphorylation of OmpR is important for the tight control of osmoregulation (40). Mutations in EnvZ that specifically affect either kinase or phosphatase activities were localized in separate structural domains (41). This bifunctional enzymatic activity of EnvZ is further regulated by a linker region containing two helices (42). Upon signal input, the interaction between these two helices is modified, thereby altering the relative position of domains A and B and hence the ratio of kinase/phosphatase FIG. 5. Model representing the similarities in co-regulation between MexEF-OprN and CzcCBA efflux systems and the OprD porin. Regulators MexT and CzcR could repress OprD expression either directly or via a yet unknown mechanism, leading to imipenem resistance. MexT activates the transcription of mexEF-oprN, conferring resistance to quinolones, chloramphenicol, and trimethoprim, while CzcR activates transcription of czcCBA conferring resistance to zinc, cadmium, and cobalt (see text). activity of the catalytic and ATP-binding domain (42). The mutated Val-194 and Gly-197 residues in CzcS are located in this linker region that connects the putative catalytic and ATP-binding domain to the second transmembrane segment (Fig. 4). The V194L and the G197D mutations could therefore cause a conformational change in CzcS that is normally induced by the signal (heavy metals), causing CzcS to be permanently autophosphorylated. Alternatively the change in conformation could decrease the phosphatase activity of CzcS, thereby preventing or reducing dephosphorylation of CzcR.
Since overexpression of wild type CzcR only reduced OprD expression without increasing heavy metal resistance, it can be concluded that at least the unphosphorylated form of CzcR negatively regulates OprD, while the amount of active phosphorylated CzcR is critical for triggering czcCBA transcription. Simultaneous overexpression of an efflux pump and downregulation of a porin pathway seems to be an efficient mechanism to prevent intracellular accumulation of toxic molecules. This type of co-regulation has been described in nfxC type mutants of P. aeruginosa (43). These mutants overexpress the multidrug efflux pump MexEF-OprN (44), conferring resistance to quinolones, chloramphenicol, and trimethoprim, and are resistant to carbapenems due to decreased OprD expression (44,45). Interestingly this resistance phenotype can be obtained in the susceptible wild type PT5 upon plasmid-mediated overexpression of MexT, the transcriptional activator of the MexEF-OprN efflux pump (46,47). MexT overexpression was shown to decrease both transcriptional (46,47) and posttranscriptional expression of oprD (46). Hence both CzcR and MexT are able to repress OprD expression, which could occur either directly or via another regulator (Fig. 5). Since in PT5 as well as in other PAO1 strains the mexT gene is inactivated by insertion of a stretch of 8 bp, resulting in a frameshift and premature stop of translation (48), one can exclude CzcR-mediated repression of OprD via MexT.
Another example of this type of co-regulation has been demonstrated in E. coli where the micF regulatory RNA (49) simultaneously decreases expression of the porin OmpF and increases expression of the multidrug efflux pump AcrAB (50). Expression of micF in E. coli is controlled by environmental and internal stress signals (51). The presence of a micF homologue in P. aeruginosa was already suggested by Southern hybridization with the E. coli micF gene (52). Exposure to heavy metals or antibiotics could therefore be considered as external stress signals and trigger a micF-like response in P. aeruginosa.
Co-regulation between heavy metal and antibiotic resistance might have important diagnostic and clinical implications. It was known for a long time that MICs of carbapenems for P. aeruginosa are increased in the presence of trace amounts of zinc in the medium (53,54). While direct inactivation of imipenem by zinc compounds was hypothesized (55), it seems clear from our results and those of Conejo et al. (20) that zincmediated repression of OprD is mainly responsible for imipenem resistance. More importantly, zinc, which is used as a catalyst during production of LUCs (56), is released into growth media (19), and as we show here also into urine. The amount of zinc released from LUCs into urine (1.1 mM) is higher than the zinc concentration released into Mueller-Hinton medium (0.56 mM) (20). These concentrations are sufficient to induce imipenem resistance in a clinical isolate (34) and in the wild type strain PT5. Whether this transient increase in carbapenem resistance might compromise treatment of urinary tract infections by P. aeruginosa remains to be determined. While no stable, constitutively resistant mutants to zinc and imipenem could be obtained after 6 days of growth in LUCs, one cannot exclude that such mutants might appear at low frequency in vivo. Interestingly polyvinylchloride urinary catheters, which were used for comparison, released undetectable amounts of zinc (data not shown). Therefore, the choice of catheter used should take into account an eventual colonization by P. aeruginosa and subsequent infection.