Characterization of a new efflux pump, MexGHI-OpmD, from Pseudomonas aeruginosa that confers resistance to vanadium
- Laboratory of Microbial Interactions, Department of Immunology, Parasitology and Ultrastructure, Flanders Interuniversity Institute of Biotechnology, Vrije Universiteit Brussel, Paardenstraat 65, B-1640 Sint Genesius Rode, Belgium1
- Author for correspondence: Pierre Cornelis. Tel: +32 2 3590221. Fax: +32 2 3590399. e-mail: pcornel{at}vub.ac.be
- Received 6 December 2001.
- Revised 25 April 2002.
- Accepted 2 May 2002.
Abstract
Vanadium has an antibacterial activity against Pseudomonas aeruginosa, especially under conditions of iron limitation. Some degree of resistance to V is inducible by prior exposure to the metal. One mutant (VS1) with a higher sensitivity to V was obtained by transposon mutagenesis of P. aeruginosa PA 59.20, a clinical isolate. This mutant had an insertion in a non-coding region, upstream of a cluster of four genes. Three of them show similarities to genes corresponding to known P. aeruginosa antibiotic efflux systems, including an efflux protein, a membrane fusion protein and an outer-membrane porin. This cluster was named mexGHI-opmD. By allelic exchange, three mutants, ncr (for non-coding region), mexI and opmD were constructed in P. aeruginosa PAO1. Next to V sensitivity, the ncr, mexI and opmD mutants also showed reduced production of elastase, rhamnolipids, pyocyanine, pyoverdine and had reduced swarming motility, phenotypes that are known to be regulated by quorum sensing. All wild-type phenotypes, including growth in the presence of V, were restored by complementation with the complete cluster. The production of N-acyl-homoserine lactones (AHLs) was detected using the Chromobacter violaceum bioassay. Total extracts from the three mutants failed to induce the production of violacein by C. violaceum, although AHLs were detected by TLC and C. violaceum overlay. Violacein production was restored by complementation with mexGHI-opmD. The opmD mutant grew very slowly in LB or CAA medium, indicating that OpmD has an important physiological function for the cell. In conclusion, it is believed that the MexGHI-OpmD pump is probably involved in AHL homeostasis in P. aeruginosa.
INTRODUCTION
Vanadium exerts a bacteriostatic effect on Pseudomonas aeruginosa, especially when the cells are grown under conditions of iron limitation (Baysse et al., 2000⇓ ). In our previous work, we have shown that V(IV) was complexed by the two P. aeruginosa siderophores pyoverdine (Pvd) and pyochelin (Pch), and V-Pch was toxic for the cells (Baysse et al., 2000⇓ ). Some degree of resistance against V could also be acquired after pre-incubation of the cells with subinhibitory concentrations
of VOSO4 (Baysse et al., 2000⇓ ). At least one factor, the production of Fe-dependent superoxide dismutase (SodB), seems to contribute to the development
of resistance against V in P. aeruginosa (Baysse et al., 2000⇓ ). Indeed, it is known that, at neutral pH, V(IV) can be readily oxidized to V(V), generating the toxic superoxide anion
(
) (Liochev & Fridovich, 1987⇓ ). Up to six mechanisms have been recognized by which bacteria can exert resistance against toxic metals, including metal
exclusion by permeability barrier, intracellular or extracellular sequestration, detoxification and active transport of the
metal out of the cell (reviewed by Nies, 1999⇓ ; Bruins et al., 2000⇓ ). So far, no study has been done on the determinants that confer resistance to V. To our knowledge, only one study describes
the isolation and characterization of natural isolates of Escherichia hermannii and Enterobacter cloacae that accumulate both Ni and V, indicating a possible metal sequestration mechanism (Hernández et al., 1998⇓ ). Resistance to V is probably multi-factorial, and a molecular approach is necessary to identify the determinants governing
its expression. To achieve this, we chose a transposon mutagenesis approach that has been successfully used to detect P. aeruginosa and Pseudomonas fluorescens mutants impaired in their resistance against Zn and Cd (Hassan et al., 1999⇓ ; Rossbach et al., 2000a⇓ , b⇓ ).
In this study, we describe one P. aeruginosa mutant with an increased susceptibility to V, with an insertion upstream from a previously undescribed efflux pump, with similarity to Mex efflux systems from P. aeruginosa known to contribute to antibiotic resistance. We also show that this new MexGHI-OpmD pump plays a role in the quorum sensing network of P. aeruginosa.
METHODS
Strains and transposon mutagenesis.
P. aeruginosa strain PA 59.20, a type III pyoverdine-producing strain (Meyer et al., 1997⇓ ) was selected because of the higher stability of Tn5 insertions in this strain, compared to P. aeruginosa PAO1 (Cornelis et al., 1992⇓ ). For mutagenesis, a mini-TnphoA3 transposon construct was used (Pattery et al., 1999⇓ ). This transposon bears a resistance against gentamicin (Gm) on plasmid pGV4692 (Table 1⇓). The procedures for the conjugational transfer of the transposon and the selection of transposon mutants were described by Cornelis et al. (1992)⇓ . For the selection of mutants, Casamino acid medium (CAA), containing 100 μg Gm ml−1, was used. The mutants were afterwards transferred to CAA plates containing 1·5 mM VOSO4 for the detection of V-sensitive clones.
Table 1. Bacterial strains and plasmids used in this study
Identification of transposon insertions.
Genomic DNA from V-sensitive mutants was isolated using the procedure described by Wilson (1990)⇓ . After restriction with SmaI, the DNA was self-ligated under conditions that favour the formation of circular molecules. The ligated DNA was further used as template for a PCR amplification using primers Gm1 and PhoA5 (Table 2⇓) for 30 cycles and the following parameters: 50 s at 94 °C, 30 s at 55 °C and 4 min at 72 °C using ExTakara polymerase (Biowhittaker).
Table 2. Oligonucleotide primers used in this study
After the first PCR reaction, a nested PCR reaction was performed using primers Gm2 and PhoA4 (Table 2⇑). Fragments were further cloned in the pCR2.1 vector of the TA cloning kit (Invitrogen). E. coli DH5α and E. coli Top10F′ (Invitrogen) (Table 1⇑) were used as hosts for cloning.
DNA sequencing was performed by Eurogentec, using universal primers. Sequences obtained were compared to sequences present in the database of the Pseudomonas genome (http://www.pseudomonas.com) using the blastx algorithm.
Mutagenesis by allelic exchange.
From P. aeruginosa strain PAO1 (Stover et al., 2000⇓ ) a 2·6 kb sequence between genes PA4204 and PA4205 was amplified by PCR using primers 4204 and 4205 (Table 2⇑). This fragment was ligated to pCR2.1 using the TA cloning kit (Invitrogen). The recombinant plasmid was further restricted with SacII and the ends polished by treatment with Klenow polymerase. The Gm cassette from the pBBR-1-MCS vector (Kovach et al., 1994⇓ ) was amplified with Taq polymerase (Gibco-BRL) using the primers gent1 and gent2 (Table 2⇑), and cloned into the pCR2.1 vector, resulting in plasmid pGM. The Gm cassette was removed from plasmid pGM by NotI restriction, the fragment was blunt-ended and then ligated to the SacII-digested plasmid. After transformation, the resulting plasmid containing the Gm-interrupted DNA was isolated from a recombinant cloned and digested by HindIII and BclI to liberate the DNA fragment containing the Gm cassette. This fragment was in turn ligated to HindIII/BamHI-restricted pBR325 DNA (Bolivar, 1978⇓ ) before transformation of E. coli DH5α or S17-1. Transformed E. coli S17-1 was used to mobilize the suicide plasmid to P. aeruginosa PAO1 (Simon et al., 1983⇓ ). Selection for double recombination events was done by plating the recombinants on CAA plates containing Gm (100 μg ml−1), and Gm plus chloramphenicol (Cm) (200 μg ml−1), and selecting the clones that were sensitive to Cm. Confirmation of the double recombination event in these clones was done by PCR amplification using primers 4205upf and 4205upr (Table 2⇑), since double recombination should result in the amplification of a unique band of 1·3 kb.
The P. aeruginosa PAO1 opmD gene (PA4208) was amplified by PCR using primers opmD1 and opmD2 (Table 2⇑). The resulting fragment was ligated into vector pCR2.1 (Invitrogen).
After restriction with EcoN1, the Gm cassette was ligated into the opmD gene. The complete DNA fragment with the inserted Gm cassette was liberated by restriction with EcoRI and ligated to EcoRI-restricted pBR325. Similarly, the mexI gene (PA4207) was amplified using primers mexI1 and mexI2 (Table 2⇑). After StuI restriction, the Gm cassette was ligated into mexI. The complete fragment was liberated from pCR2.1 by restriction with HindIII and BclI, and ligated to HindIII/BamHI-restricted pBR325. After transformation of E. coli GJ23, the plasmids were transferred by conjugation to P. aeruginosa PAO1. Transconjugants showing the loss of Cm resistance (200 μg ml−1) were selected and the double recombination confirmed by PCR using the above described primers.
RT-PCR analysis.
RNA was extracted (High Pure RNA Isolation Kit; Roche Diagnostics) from P. aeruginosa PA 59.20 wild-type, PA 59.20 vs1, PAO1 wild-type and PAO1 ncr mutants grown in LB and harvested in the middle of the exponential phase. cDNA, synthesized using the First-Strand cDNA Synthesis Kit (Amersham Pharmacia) was done using two sets of primers: mexI3 and mexI1, and opmD3 and opmD4 (Table 2⇑). The relative position of the primers is shown in Fig. 1(A)⇓. As a control for RNA contamination by DNA, the PCR reaction was done on the same samples, without first strand cDNA synthesis.
Fig. 1. (A) Organization of the mexGHI-opmD cluster, corresponding to genes PA4205–PA4208 (http://www.pseudomonas.com). The place of the TnphoA3 insertion in strain PA 59.20 is indicated by a black triangle, while the place of the insertion of the Gm cassette in strain PAO1 is indicated by an arrow. The hatched squares represent the lux boxes identified upstream of the cluster and in the intergenic region between phzM and phzA (Mavrodi et al., 2001⇓ ; Pattery et al., 2001⇓ ). The thick black lines below correspond to the inserts of PAO1 genomic DNA in pRG930 described in Table 1⇑. Short arrows represent the primers used in the RT-PCR analysis. (B) Sequence of the non-coding region between gene PA4204 and mexG. The two underlined sequences correspond to putative lux boxes as defined by Whiteley & Greenberg (2001)⇓ . The sequence CCGCGG in bold and italic type corresponds to the SacII restriction site used to insert the Gm cassette, to realize an allelic exchange in strain PAO1. The ATG in bold and italic corresponds to the predicted translation initiation codon of mexG (http://www.pseudomonas.com). The ATG in bold and the sequence GGAGG in bold, upstream of the ATG, correspond to the newly predicted initiation codon of mexG (this work). The place of the TnphoA3 insertion in strain PA 59.20 is indicated by an arrow.
Physiological studies.
Growth parameters were measured using a Bioscreen Apparatus (Life Technologies), using the following parameters: shaking for 10 s every 3 min; reading every 20 min; temperature 37 °C; volume of the culture, 300 μl. As inoculum, an overnight culture of PAO1 in CAA was diluted to achieve a final OD600 of 0·001. Each culture was realized in triplicate and each experiment was repeated three times.
Vanadyl oxydo sulfate (VOSO4.5H2O) was prepared as a 100 mM stock solution and kept at 4 °C. Growth curves were also followed in the presence of increasing concentrations (0·5, 1, 1·5, 2 and 2·5 mM) of different metal salts in CAA medium at 37 °C: CoCl2, ZnSO4, NiCl2, CuSO4, CrCl3 and Bi(NO3)3. CdCl2 was used at 0·2 mM, final concentration.
Antibiotic susceptibility.
Antimicrobial susceptibility tests were performed by the disk diffusion method on Mueller–Hinton agar plates. The antibiotics tested (Sanofi Diagnostic Pasteur) included amikacin (30 μg), ticarcillin-clavulanate (75/10 μg), tobramycin (10 μg), tetracycline (Tc; 30 UI), imipenem (10 μg), netilmicin (5 μg), ciprofloxacin (5 μg) and colistin (10 μg). Zones of inhibition were measured after incubation at 37 °C for 24 h.
Production of elastase.
Overnight LB-grown cultures were centrifuged and the supernatants collected. One-hundred microlitres of culture supernatant were added to glass test tubes containing 10 mg elastin Congo Red ml−1 in 0·9 ml 0·1 M Tris/HCl. After 6 h incubation at 37 °C, the tubes were centrifuged and the OD495 of the supernatants were measured (Rust et al., 1994⇓ ). All measurements were repeated three times.
Production of rhamnolipids.
M9-glutamate minimal medium agar plates containing 0·2 g cetyltrimethylammoniumbromide and 5 mg methylene blue l−1 were inoculated with 10 μl of an overnight LB culture of P. aeruginosa strains. After an overnight incubation at 37 °C, the diameter of the clearing zone around the bacterial spots was measured as evidence of rhamnolipid production (Siegmund & Wagner, 1991⇓ ).
Swarming motility.
Two microlitres from overnight agitated cultures in LB (5 ml) were spotted on LB plates containing 1·3 % agarose. The plates were incubated overnight at 37 °C and photographed. Spreading bacteria indicated swarming from the inoculation spot.
Measurement of siderophore and pyocyanine production.
The pyoverdine content in the supernatant was determined by measuring A400 or by spectrofluorimetry (excitation at 405 nm, emission at 460 nm, using a Shimadzu spectrofluorimeter) and normalized by a biomass unit expressed as OD600 of the culture (Höfte et al., 1993⇓ ). Measurements are means of three independent experiments.
Pyocyanine production was visualized by plating the bacteria on Pseudomonas P agar (Difco), followed by 48 h incubation. Pyocyanine production caused the medium to be coloured deep blue. Pyocyanine was measured in culture supernatants according to Mavrodi et al. (2001)⇓ .
Bio-assay for acyl-homoserine lactone (AHL) production.
As indicator strain, a Chromobacter violaceum CV026 AHL-deficient mutant was used as described by McClean et al. (1997)⇓ . TLC was used to separate the AHLs and overlaid with the CV026 indicator strain as described by McClean et al. (1997)⇓ . Total culture supernatants from overnight-agitated cultures grown at 37 °C in LB were extracted with dichloromethane as described by McClean et al. (1997)⇓ to extract the AHLs.
Complementation in trans with wild-type DNA.
A genomic bank of PAO1 DNA was constructed as described previously (Lim et al., 1997⇓ ) in the wide-host range cosmid pRG930Cm (van den Eede et al., 1992⇓ ), using PstI to partially digest the genomic DNA. The bank was screened by colony blotting using the 2·6 kb PCR fragment amplified using primers 4204 and 4205 as probe. The Roche Diagnostics non-radioactive Dig detection system was used, in combination with the chemiluminescent substrate CPD-star (Roche Diagnostics). Positive clones were confirmed by PCR analysis using the following primers: opmD3 and opmD4, 4207f and 4207r, 4206f and 4206r, and 4205upf and 4205upr (Table 2⇑).
RESULTS
Screening of the transposon mutant bank
After mutagenesis with mini-TnphoA3, 3000 separate clones were screened by replica-plating on CAA medium and on CAA containing 1·5 mM VOSO4, a concentration at which the wild-type PA 59.20 could grow. One mutant, vs1, could not grow in the presence of 1 mM VOSO4. These results were confirmed in liquid medium (results not shown).
Localization of the place of transposon insertion in PA 59.20 vs1
DNA from P. aeruginosa PA 59.20 vs1 was isolated and the sequence of the DNA flanking the transposon insertion was determined as described in Methods. After the nested PCR reaction, a fragment of about 600 bp was obtained, which was cloned in pCR2.1 and sequenced. The sequence analysis revealed 94% identity for a 171 bp sequence with a PAO1 sequence (http://www.pseudomonas.com) found between genes PA4204 and PA4205 (Fig. 1A⇑, B⇑). According to the PAO1 genome annotation, gene PA4205 encodes a hypothetical integral membrane protein with four predicted transmembrane helices. The predicted ATG start codon for this protein is at position 4705955, but is not preceded by a convincing ribosome-binding site (AAGGA, Fig. 1B⇑). An alternative ATG start codon is present, 128 nt downstream, in the same frame, just after a ribosome-binding site closer to the consensus (GGAGG, Fig. 1B⇑). From this new start codon, an ORF extends that encodes a protein of 105 aa containing three transmembrane domains (10–28, 35–55, 64–85) (Zhai & Saier, 2001⇓ ). Gene PA4206 encodes a putative RND efflux protein of 370 residues (Tseng et al., 1999⇓ ), while gene PA4207 encodes a protein with highest similarity to P. aeruginosa MexF, a member of the family of membrane fusion proteins (Zgurskaya & Nikaido, 2000⇓ ). The last protein, encoded by gene PA4208, is most similar to the P. aeruginosa efflux porin OprN (Köhler et al., 1997⇓ ). After this gene cluster, and transcribed in the opposite orientation, is gene PA4209 that has recently been shown to correspond to the phzM gene encoding an O-methyltransferase necessary for the biosynthesis of pyocyanine (Mavrodi et al., 2001⇓ ; Pattery et al., 2001⇓ ). This gene cluster, with the exception of gene PA4205, is very similar to the already described complete tripartite gene clusters that encode the MexABOprM, MexCDOprJ and MexEFOprN systems, which function in drug efflux in P. aeruginosa (Poole, 2001a⇓ , b⇓ ).
Generation of new mutants by allelic exchange
Since the transposon was inserted in a non-coding region where its presence could eventually have positively influenced the transcription of downstream genes, it was decided to create a mutant in P. aeruginosa PAO1 with a Gm cassette insertion in the same locus, but downstream of the original Tn5 mutation. This was done by insertion of a Gm cassette into the unique SacII site (Fig. 1B⇑). The double recombinant PAO1 ncr mutant had exactly the same phenotype as the original vs1 mutant in strain PA 59.20. This mutant could not grow in CAA or in LB medium in the presence of 1·5 mM VOSO4, and showed reduced growth in the presence of 1 mM VOSO4 (Fig. 2⇓).
Fig. 2. Growth, in CAA plus 1 mM VOSO4, of wild-type P. aeruginosa PAO1 (•), mutant PAO1 ncr (○), mutant PAO1 mexI (▵), mutant PAO1 opmD (□) and opmD complemented with mexGHI-opmD in trans (+). Growth was measured in the Bioscreen as described in Methods. Each growth curve is the mean of three separate experiments.
The growth of the mutant in the presence of Co, Cr, Ni, Cd, Cu, Bi or Zn was unaffected (results not shown). Given the high similarity of this operon with similar operons encoding efflux systems, we decided to name the genes, mexG-mexH-mexI-opmD (the name opmD for this gene has already been given to this protein; see http: / / www . cmdr . ubc . ca / bobh / OprMfamily . html ). Similarly, we obtained two other mutants, one with the Gm cassette inserted in the mexI gene, the other one with the insertion in the opmD gene. Both mutants were also sensitive to 1 mM VOSO4 (Fig. 2⇑). Mutant opmD grew very slowly in liquid CAA medium and could not form isolated colonies on LB-agar plates (results not shown).
Complementation of the PAO1 ncr, mexI and opmD mutants
A PAO1 DNA genomic bank was constructed by cloning partially PstI-digested DNA into the wide host-range cosmid vector pRG930Cm (van den Eede et al., 1992⇓ ; S. Matthijs, unpublished results). About 3500 colonies were screened by colony hybridization using as probe the 2·5 kb PCR fragment amplified with primers 4204 and 4205. One positive clone was selected and was confirmed by PCR amplifications with primers designed to amplify the mexG, mexH, mexI and opmD genes. Based on the results of these amplifications, we confirmed that the insert contained the genes PA4200–4209 (Table 1⇑ and Fig. 1A⇑). Subclones were obtained that contained only mexG, or mexG-mexH, or mexG-mexH-mexI (Table 1⇑ and Fig. 1A⇑). These subclones, as well as the original cosmid, were transferred by conjugation to PAO1 ncr, mexI or opmD and the transconjugants were tested for growth in the presence of VOSO4. From the results, it was clear that only recombinants containing mexG-mexH-mexI-opmD could grow in the presence of 1 mM V, indicating that the complete pump is needed to confer resistance to the metal (results shown only for opmD complemented with mexG-mexH-mexI-opmD in Fig. 2⇑).
RT-PCR analysis of the mexGHI-opmD transcript
To confirm the operon organization of the cluster and to evaluate the effect of the two mutations in the intergenic region (the original transposon insertion in strain PA 59.20 and the PAO1 ncr mutation) on transcription efficiency, we did an RT-PCR analysis on mRNAs extracted from wild-type PA 59.20, PA 59.20 vs1, PAO1 and PAO1 ncr. Different primer combinations were used (Fig. 1A⇑). Transcripts were clearly less abundant in the PA 59.20 vs1 transposon mutant and in the PAO1 ncr mutant compared to the wild-type (see Fig. 3a⇓ for mexI primers, Fig. 3b⇑ for opmD primers). When amplifications were performed using the opmD primer combinations, almost no transcript could be amplified from the two mutant mRNA preparations.
Fig. 3. RT-PCR analysis of abundance of transcripts in P. aeruginosa PA 59.20 and PA 59.20 vs1 mutant, and in PAO1 and PAO1 ncr mutant. (a) Amplification with primers mexI1 and mexI3. Lanes: 1, PA 59.20 vs1; 2, PA 59.20 wild-type; 3, PAO1 ncr; 4, PAO1 wild-type; 5, markers. (b) Amplification with primers opmD3 and opmD4. Lanes: 1, markers; 2, PA 59.20 vs1; 3, PA 59.20 wild-type; 4, PAO1 ncr; 5, PAO1 wild-type.
Other phenotypes of the PAO ncr, mexI and opmD mutants
We first observed that the PAO ncr mutant produced less pyocyanine on P agar medium and in liquid medium (50% less compared to the wild-type; results not shown). Pyocyanine production is known to be under the control of the RhlR activator, together with the AHL cell-to-cell signal molecule C4-AHL, the production of which is in turn controlled by the 3-oxo-C12-AHL molecule, together with the activator LasR (Latifi et al., 1996⇓ ; Pesci & Iglewski, 1997⇓ ). We therefore decided to compare the PAO1 wild-type, the ncr, mexI and opmD mutants and the complemented mutants for different quorum-sensing-regulated traits. These include the production of the LasB elastase, controlled by LasR, and RhlR rhamnolipids, controlled by RhlR (reviewed by de Kievit & Iglewski, 2000⇓ ), pyoverdine, controlled by LasR (Stintzi et al., 1998⇓ ), and the capacity to swarm (Köhler et al., 2000⇓ ; Reimmann et al., 2002⇓ ). Table 3⇓ shows the results for the measurement of elastase activity, production of rhamnolipids and the siderophore pyoverdine. In the case of elastase, the ncr, mexI and opmD mutants showed a drastic reduction of activity, while the ncr and opmD mutants complemented by the mexGHI-opmD cluster displayed wild-type elastase activity. The production of rhamnolipids, as judged by the diameter of the clearing zone around the inoculated spot, was also reduced in the three mutants, but less severely in the case of the ncr mutant. Again in this case, complementation by the complete gene cluster restored wild-type levels of rhamnolipid production. Finally, pyoverdine production also decreased to about 50% of the level of the wild-type in the case of the mexI and the opmD mutants, while the two complemented mutants produced as much pyoverdine as the wild-type. Fig. 4⇓ shows that the three mutants have a much reduced swarming capacity, compared to the wild-type. This is particularly true for the mexI mutant, which does not swarm at all. Interestingly, we regularly observed that the opmD mutant produced sectors around the inoculation zone, an indication of the selection of phenotypic revertants. In this case, similar to what was observed for the other traits, the ncr mutant was the least affected. Complementation of the opmD mutant resulted in restoration of normal swarming behaviour.
Table 3. Production of elastase, rhamnolipids and pyoverdine by wild-type PAO1, ncr, mexI, opmD mutants and complemented ncr and opmD mutants
Fig. 4. Swarming motility assay. Clockwise from the top right: PAO1, PAO1 ncr, PAO1 mexI, PAO1 opmD and PAO1 opmD complemented with mexGHI-opmD.
Production of AHLs
Because of the down-regulation of several typical quorum-sensing-regulated traits, we decided to look for the production of signalling AHL molecules by wild-type and mutants without and with complementation. C. violaceum CV026 was used as a biosensor. This strain is unable to produce the violet pigment violacein because of a Tn5 insertion in the cviI gene, which encodes a homoserine lactone synthase (McClean et al., 1997⇓ ). In this strain, violacein production is inducible by AHLs with N-acyl side chains from C4 to C8 in length; on the other hand, the presence of C10–C14 AHLs inhibits the production of violacein when mixed with the stimulatory shorter chain AHLs (McClean et al., 1997⇓ ). As seen in Fig. 5(a)⇓, the production of violacein is stimulated by total AHLs from P. aeruginosa, but not by extracts from mexI or opmD mutants, while a faint pigment production can be visualized around the well containing extracts from the ncr mutant. Extracts from ncr and opmD mutants complemented with mexGHI-opmD were in turn strongly stimulatory. Although little or non-stimulatory, extracts from ncr and opmD contain at least one faint AHL spot as detected by TLC and C. violaceum overlay (Fig. 5b⇑). In the case of the opmD mutant, this AHL spot could not be detected. When the opmD mutant was complemented with the mexGHI-opmD genes, one strong AHL spot could be detected in TLC, corresponding to one of the two spots detected in PAO1 extracts.
Fig. 5. (a) Bioassay of violacein production by C. violaceum CV026 as an indication of the presence of short AHLs. Top: left, total AHLs from P. aeruginosa PAO1; middle, total AHLs from PAO1 ncr; right, total AHLs from PAO1 mexI. Bottom: left, total AHLs from PAO1 opmD; middle, total AHLs from PAO1 ncr complemented with mexGHI-opmD; right, total AHLs from PAO1 opmD complemented with mexGHI-opmD. (b) TLC separation of AHLs from PAO1 (lane 1), PAO1 ncr (lane 2), PAO1 mexI (lane 3), PAO1 opmD complemented with mexGHI-opmD (lane 4) and PAO1 opmD (lane 5). The TLC plates were dried and overlaid with C. violaceum CV026 as described by McClean et al. (1997)⇓ .
Susceptibility of the PAO ncr mutant to antibiotics
Since the mexGHI-opmD gene cluster shows high similarity to described efflux pumps that are known to confer multiple drug resistance to P. aeruginosa, we decided to test the susceptibility of the ncr mutant to different antibiotics. Unexpectedly, we repeatedly found that this mutant is fully resistant to Tc while being less sensitive to netilmicin and ticarcillin plus clavulanic acid (Table 4⇓). Interestingly, complementation with the full cluster restored sensitivity to Tc and complete resistance to ticarcillin plus clavulanic acid.
Table 4. Resistance of PAO1 (WT) and PAO1 ncr to antibiotics
In silico analysis of the non-coding region upstream of the gene cluster
Analysis of the non-coding DNA between PA4204 and PA4205 allowed identification of two sequences with the NNCT-(N12)-AGNN motif (Fig. 1B⇑) that is conserved in so-called lux box elements that are generally found about −40 bp upstream of some P. aeruginosa quorum-sensing-regulated genes (Whiteley & Greenberg, 2001⇓ ). Fig. 6⇓ shows the comparison of the motifs upstream of mexG and other motifs described by Whiteley & Greenberg (2001)⇓ .
Fig. 6. Comparison of the two putative lux boxes found upstream of mexG (mexG1 and mexG2) and other promoter-specific elements found upstream of P. aeruginosa quorum-sensing-controlled genes (after Whiteley & Greenberg, 2001⇓ ).
DISCUSSION
Only two reports, so far, have established that V can be toxic for bacteria, in the case of Streptococcus pneumoniae (Fukuda & Yamase, 1997⇓ ) and, by our group in the case of P. aeruginosa (Baysse et al., 2000⇓ ). As mentioned by Nies (1999)⇓ in his review on microbial heavy-metal resistance, nothing is known about the mechanisms of resistance against vanadyl- or vanadate ions. In their study, Hernández et al. (1998)⇓ described that E. hermannii cells grown in the presence of V accumulate the metal and show the induction of a 45 kDa outer-membrane protein, which could be indicative of the induction of an efflux system porin. This study is a first attempt to elucidate the mechanisms of bacterial resistance against V. Using a transposon mutagenesis strategy, we isolated a mutant that showed an increased susceptibility to V. By choosing a mini-TnphoA translational fusion reporter transposon, we hoped to discover some in-frame periplasmic fusions that would respond to V, an approach that was used successfully to detect lacZ fusions responding to Zn and Cd (Rossbach et al., 2000a⇓ , b⇓ ). However, after screening of more than 3000 independent clones containing a transposon insertion, no clear and specific V-responding fusion was found. Yet, we succeeded in isolating one V-sensitive mutant that could not grow at a concentration of 1·5 mM VOSO4. It was found to have an insertion in a non-coding region upstream of a gene cluster with a likely operon structure and bearing resemblance to known antibiotic efflux systems already described in P. aeruginosa (Poole, 2001b⇓ ). It is known that active efflux plays an important role in the development of resistance against some toxic metals (Rosen, 1996⇓ ; Nies, 1999⇓ ). The best known examples are the plasmid-borne czc, cnr and ncc systems of Ralstonia metallidurans (Hassan et al., 1999⇓ ; Silver & Phung, 1996⇓ ). In P. aeruginosa, similar metal efflux systems have been described for cadmium and zinc, and also identified in P. fluorescens (Hassan et al., 1999⇓ ; Rossbach et al., 2000a⇓ , b⇓ ). It is however the first time, to our knowledge, that an operon with clear similarity to antibiotic efflux determinants has been shown to be involved in resistance against a metal. It is interesting to note that the expression of the pump only contributes to resistance against V, and not to other metals. That the efflux pump itself is needed for V-resistance has been confirmed by our complementation experiments.
One of the first obvious phenotypes of our V-sensitive mutants was the clearly reduced production of the phenazine pigment pyocyanine. One possibility is that pyocyanine excretion is directly mediated by this pump. In this regard, it is worth noting that phzM, a gene recently demonstrated to be necessary for the biosynthesis of pyocyanine, is transcribed just downstream of opmD, in the opposite orientation (Mavrodi et al., 2001⇓ ; Pattery et al., 2001⇓ ). Another more likely explanation for the decreased pyocyanine production is that decreased extracellular release of AHLs by the pump mutants results in decreased transcription of the phz phenazine biosynthesis cluster. Indeed, we observed that different, well known, quorum-sensing-regulated traits were affected in our different ncr, mexI and opmD mutants, while the presence of the pump in trans restored the wild-type phenotype in all instances. P. aeruginosa is known to produce two major AHL signal molecules, one long-chain, N-(3-oxododecanoyl)-l-homoserine lactone (OdDHL), and one short-chain, N-butanoyl-l-homoserine lactone (BHL). The C. violaceum assay is known to detect short-chains AHLs and, conversely, to be inhibited by long-chain homoserine lactones, including OdDHL (McClean et al., 1997⇓ ). The fact that total AHL fractions from the mexI and opmD mutants failed to stimulate the production of violacein by C. violaceum may be due to an imbalance in the BHL/OdDHL ratio. If the proportion of long-chain AHLs is higher in these mutant extracts, an inhibition of the biosensor is likely to occur (McClean et al., 1997⇓ ). Supporting this hypothesis is the fact that one stimulating AHL can be detected in the extracts of these mutants when separated first by TLC. This AHL spot is the same as one that is detected by TLC separation of wild-type PAO1 extracts and is likely to be BHL. However, only a more thorough analysis of AHL production in these mutants by HPLC will definitively answer this question. It is interesting to note that the opmD mutant is also affected in its growth in CAA medium. Its morphology is also different on LB plates since it does not form isolated colonies but grows in patches (results not shown). Also, this mutant aggregates in liquid cultures (results not shown). These observations indicate that OpmD is an important outer-membrane porin for P. aeruginosa and that it cannot be replaced by another member of the OprM porin family (http://www.cmdr.ubc.ca/bobh/OprMfamily.html). Recently, Köhler et al. (2001)⇓ showed that overproduction of the MexEF-OprN efflux pump results in decreased production of virulence factors dependent on RhlI-RhlR, including pyocyanine and rhamnolipids. Two groups, Evans et al. (1998)⇓ and Pearson et al. (1999)⇓ , reached opposite conclusions concerning the influence of the MexAB-OprM pump in regard to quorum sensing. The first group found that strains overexpressing mexAB-oprM excrete less OdDHL while the opposite was observed by Pearson and co-workers. The difference between the two observations could be explained by the fact that Evans and co-workers worked with a strain that is a nalB mutant overexpressing mexAB-oprM, while Pearson and co-workers worked with a PAO1 strain that was not mutated for nalB. Köhler et al. (2001)⇓ propose that another quorum sensing molecule, 2-heptyl-3-hydroxy-4-quinolone, termed PQS (Pesci et al., 1999⇓ ; Holden et al., 2000⇓ ), could be the substrate of MexEF-OprN. PQS is indeed known to positively regulate the rhl system (McKnight et al., 2000⇓ ). Interestingly, Whiteley et al. (1999)⇓ previously identified the mexGHI-opmD gene cluster (called qsc133) as being regulated by quorum sensing, in response to both BHL and OdDHL. With this in mind, we looked for putative lux boxes in the intergenic region between genes PA4204 and PA4205. Interestingly, two sequences that match the consensus sequence NNCT-(N12)-AGNN, as defined by Whiteley & Greenberg (2001)⇓ , were found. Although no conclusions can be taken at this stage concerning their localization compared to the transcription start point, it is tempting to assume that one would be a site for binding of LasR while the other would be a sequence where RhlR would bind. In their analysis, Whiteley et al. (1999)⇓ did not find a putative lux box upstream of qsc133, but they did not, at that time, take the PA4205 gene into account.
Curiously, the V-sensitive mutants did not display a decreased resistance to any of the antibiotics that we tested. Conversely, we observed an increased resistance towards Tc, ticarcillin plus clavulanic acid, and netilmicin. It has been shown that loss of mexAB-oprM expression correlates with an increased expression of mexEF-oprN and mexCD-oprJ (Li et al., 2000⇓ ). It is therefore possible that decreased expression of mexGHI-opmD could be compensated by increased expression of other pumps, such as mexCD-oprJ that is known to confer resistance to Tc, Cm and some cephems (Poole et al., 1996⇓ ). The presence of the full mexGHI-opmD in trans results in increased resistance to ticarcillin and clavulanic acid. One possible explanation for this phenomenon is that clavulanic acid (itself a lactone) is excreted by the pump. This, however, will need experimental confirmation. It will therefore be necessary in the future to study the expression of mexGHI-opmD and its interplay with other efflux systems in P. aeruginosa.
While this paper was being reviewed, an article was published (Diggle et al., 2002⇓ ) where a Tn5 insertion in mexI is described as causing a reduction of a lecA::lacZ fusion expression. The lecA gene encodes a lectin that is regulated positively by the RhlI–RhlR system (Winzer et al., 2000⇓ ).
Acknowledgments
S. Aendekerk is recipient of an IWT doctoral fellowship. This work was supported by a grant of the Fund Jean & Alphonse Forton against cystic fibrosis. We wish to thank Drs Paul Williams and Miguel Cámara from Nottingham University for sending us the CV026 strain and for interesting discussions.
