CpxR Activates MexAB-OprM Efflux Pump Expression and Enhances Antibiotic Resistance in Both Laboratory and Clinical nalB-Type Isolates of Pseudomonas aeruginosa

Resistance-Nodulation-Division (RND) efflux pumps are responsible for multidrug resistance in Pseudomonas aeruginosa. In this study, we demonstrate that CpxR, previously identified as a regulator of the cell envelope stress response in Escherichia coli, is directly involved in activation of expression of RND efflux pump MexAB-OprM in P. aeruginosa. A conserved CpxR binding site was identified upstream of the mexA promoter in all genome-sequenced P. aeruginosa strains. CpxR is required to enhance mexAB-oprM expression and drug resistance, in the absence of repressor MexR, in P. aeruginosa strains PA14. As defective mexR is a genetic trait associated with the clinical emergence of nalB-type multidrug resistance in P. aeruginosa during antibiotic treatment, we investigated the involvement of CpxR in regulating multidrug resistance among resistant isolates generated in the laboratory via antibiotic treatment and collected in clinical settings. CpxR is required to activate expression of mexAB-oprM and enhances drug resistance, in the absence or presence of MexR, in ofloxacin-cefsulodin-resistant isolates generated in the laboratory. Furthermore, CpxR was also important in the mexR-defective clinical isolates. The newly identified regulatory linkage between CpxR and the MexAB-OprM efflux pump highlights the presence of a complex regulatory network modulating multidrug resistance in P. aeruginosa.

existence of multiple regulatory components renders the mexAB-oprM operon subject to complex regulation in P. aeruginosa.
As a response regulator, CpxR was first identified as an important regulator for protecting cell envelope and promoting cell survival in Escherichia coli [22][23][24]. Numerous studies have verified the role of CpxR in antibiotic resistance in pathogenic bacteria. In E. coli, overexpression of CpxR confers resistance to β-lactams in a drug-hypersusceptible mutant, in which AcrAB, a major efflux pump, was defective [25]; CpxR is involved in the defence response to aminoglycoside-induced oxidative stress [26,27]; it confers resistance to fosfomycin by directly repressing glpT and uhpT expression in enterohemorrhagic E. coli [28]; Induction of the CpxR pathway directly contributes to tolerance toward certain antimicrobial peptides, including polymyxin B and protamine [29,30]. In Salmonella typhimurium, CpxR also confers resistance to antimicrobial peptides protamine, magainin, and melittin through activation of two Tatdependent peptidoglycan amidases [31]; moreover, it confers strong ceftriaxone resistance by modulating expression of STM1530 and ompD [32]; Laboratory-generated and clinical S. typhimurium strains lacking CpxR show reduced resistance to aminoglycosides and β-lactams [33]. In Klebsiella pneumoniae, CpxR is involved in multidrug resistance through direct promoter binding and activation of ompC and kpnEF [34,35]. In Vibrio cholerae, CpxR can activate expression of RND efflux pumps VexAB and VexGH, which can confer resistance to ampicillin [36]. Erwinia amylovora lacking CpxR show reduced resistance to β-lactams, aminoglycosides, and lincomycin [37]. Although CpxR is widely distributed in the genomes of various gammaproteobacteria, its role in Pseudomonas species remains unknown.
In this study, bioinformatics, biochemical, and genetic analyses identified a regulatory linkage between CpxR and multidrug efflux pump MexAB-OprM in P. aeruginosa. We show that CpxR activates mexAB-oprM expression by directly binding to the distal promoter and is important for multidrug resistance in nalB-type P. aeruginosa isolates under both laboratory and clinical conditions.

Comparative genomic analysis of RND efflux pumps
In order to unravel the regulatory networks responsible for modulating the expression of RND efflux pumps in P. aeruginosa, comparative genomic analysis was carried out to identify novel regulatory sites on the promoters of orthologous RND operons among different Pseudomonas species. In this case, we compared the promoter regions of the orthologous operons of mexAB-oprM, mexEF-oprN, and muxABC-opmB in 15 genome-sequenced Pseudomonas species. The results showed that, apart from the previously identified NalD repressor binding site on the mexA promoter [12] (see S1A Fig), a well conserved DNA motif was identified on the muxA promoter ( Fig 1A). Interestingly, the motif contains a consensus binding site (5 0 -GTAAA-(N) 4-8 -GTAAA-3 0 ) for CpxR, a response regulator of the two-component system in E. coli [38]. The gene locus PA14_22760 has been annotated as cpxR in the genome of P. aeruginosa strain PA14 in the Pseudomonas database [39]; it encodes a protein with the highest BLASTP score (47% identity) with E. coli CpxR among the ORFs of P. aeruginosa strain PA14.
As CpxR is a global regulator of the cell envelope stress response in E. coli [22,42] and might regulate the muxABC-opmB operon (as shown in the inter-species analysis above; Fig  1A), we used its binding site (5 0 -GTAAA-(N) 4-8 -GTAAA-3 0 ) as a probe to perform intra-species analysis of the genome of P. aeruginosa PA14. Because CpxR can exert its activity independent of the orientation of its binding site [38], the existence of potential CpxR binding sites on both strands was assessed. The results showed that a number of genes possess the consensus CpxR binding site on their promoter regions (S1 Table). Among such genes, PA14_22740, which is adjacent to the cpxR locus in the genome of P. aeruginosa strain PA14, encodes a small, putative periplasmic protein with two LTXXQ motifs (S2 Fig), a canonical feature of the protein encoded by cpxP, the cognate target gene of CpxR in E. coli [43,44]. Furthermore, among P. aeruginosa strain PA14 genes, the protein encoded by PA14_22740 showed the highest BLASTP score (25% identity) with E. coli CpxP protein. Thus, we annotated PA14_22740 as a cpxP gene in P. aeruginosa strain PA14. Surprisingly, the promoter of mexAB-oprM in P. aeruginosa PA14 also contains a consensus CpxR binding site (S1 Table and Fig 1B).
The inter-species analysis showed that the conserved CpxR binding site is present on the promoter regions of the identified cpxP orthologues, similar to the case of the muxA orthologues ( Fig  1C, for the details see S1B Fig). In contrast, the presence of the CpxR binding site on the mexA promoter is unique to P. aeruginosa among the 15 Pseudomonas species for which the entire genome has been sequenced. Therefore, for the first time, by using comparative genomic analysis, we have found a potential regulatory linkage between CpxR and mexAB-oprM in P. aeruginosa.
Since the newly identified CpxR binding site is located upstream of the distal promoter of mexA, which is known to be modulated by the MexR repressor, we further investigated the existence of the MexR binding site (5 0 -GTTGA-(N) 5 -TCAAC-3 0 , Fig 1B) [40,41] in the promoter regions of mexA orthologues among Pseudomonas species. The results showed that the presence of the MexR binding site on the mexA promoter is unique to P. aeruginosa (Fig 1C). In contrast, the NalD binding site is well conserved in the promoter region of each mexA orthologue ( Fig 1C and S1A Fig). In fact, the nalD orthologue (ttgR) locus is divergently linked to the mexA orthologue locus in the genomes of other Pseudomonas species, while the mexR locus is divergently linked to the mexA locus, which is completely separated from the nalD locus in the genome of P. aeruginosa ( Fig 1D). These observations indicate a species-specific coupling of multiple regulators (CpxR and MexR) with the MexAB-OprM efflux pump in P. aeruginosa.

Direct activation of mexAB-oprM expression by CpxR
When the mexA, muxA, and cpxP promoters were fused with the lacZ reporter gene and their expression levels were monitored, we found that their activities were under the control of CpxR in P. aeruginosa. In particular, expression of the mexA, muxA, and cpxP promoters was strongly activated by the presence of ectopically expressed CpxR in PA14ΔcpxR (Table 1). In contrast, when the newly identified CpxR binding site on the mexA promoter was altered by site-directed mutagenesis (mexAp M1 , for details see Fig 1B), CpxR-dependent activation was completely abolished. When the conserved phosphorylation site (the 52 nd aspartate residue) of CpxR was mutated to alanine (CpxR D52A ), the ectopically expressed CpxR could not activate expression of target promoters in PA14ΔcpxR (Table 1). As the stability of CpxR D52A is not altered in comparison with that of wild-type CpxR (S3 Fig), these results suggest that phosphorylation of CpxR is important for its activity.
The importance of phosphorylation in CpxR activation is further supported by the fact that the phosphorylated form of CpxR clearly bound to the target promoter region containing the proximal promoters of mexA are underlined [12,40]. The transcriptional start sites of the distal and proximal promoters of mexA are indicated by bent arrows [12,40]. The putative CpxR binding site is shaded, whereas the two MexR binding sites overlapping the -35/-10 region of the distal promoter are indicated in lower-case letters [40,41]. A NalD binding site overlapping the -35/-10 region of the proximal promoter is indicated in italic lower-case letters [12]. The nucleotide substitutions for the mutated mexA promoters (mexAp M1 and mexAp M3 ) are paired with a short vertical line. The downstream deletion boundary of the distal-only mexA promoter (mexAp M2 ) is indicated with a long vertical line. (C) A well-conserved CpxR binding site exists in the promoter region of each cpxP and muxA orthologue, while a well-conserved NalD binding site exists in the promoter region of each mexA orthologue. In contrast, the conserved CpxR and MexR binding site on the mexA promoter is unique to P. aeruginosa. (D) Components that might be involved in regulating mexA orthologue (in blue) expression in Pseudomonas species. The orthologous loci of cpxR and nalD among different Pseudomonas species are marked in green and yellow, respectively. The orthologous nalD locus is separated and replaced by the mexR locus in P. aeruginosa, but divergently linked to the orthologous mexA locus in the genomes of other Pseudomonas species.  (Fig 2A and 2B). Taken together, these results demonstrate that CpxR binds directly to its conserved DNA binding site in a phosphorylation-dependent manner, and such binding is essential for CpxR-dependent activation of the target promoters in P. aeruginosa.
The unique P. aeruginosa-specific regulatory linkage between CpxR and MexAB-OprM was demonstrated by similar experiments in an alternative host, Pseudomonas putida KT2440. In this strain, consensus CpxR binding sites exist on the promoters of PP_4504 and PP_3585, the orthologous genes of cpxP and muxA, respectively, but not on the promoter of ttgA, the orthologous gene of mexA. In P. putida KT2440, ectopically expressed CpxR significantly activates expression of PP_4504 and PP_3585, but does not alter expression of ttgA (Table 1). Therefore, among the Pseudomonas species analysed, CpxR-dependent activation of the promoter of mexAB-oprM is unique in P. aeruginosa.
As CpxR can activate expression of mexAB-oprM and muxABC-opmB, the contributions of these genes to multidrug resistance in P. aeruginosa were investigated. Minimal inhibitory concentrations (MICs) of ciprofloxacin, ofloxacin, ceftazidime, cefsulodin, and aztreonam, but not amikacin, were increased at least 4-fold by ectopically expressed CpxR in PA14 and PA14ΔcpxR strains ( Table 2) in a manner dependent on MexA, but not MuxA. In this case, ectopically expressed CpxR failed to increase the MICs of the tested antibiotics in a mexA null-mutant PA14ΔmexA strain. In contrast, the MIC increases caused by the ectopically expressed CpxR were not altered in a muxA null-mutant PA14ΔmuxA strain (Table 2). These results indicate that CpxR activates expression of mexAB-oprM, which enhances multidrug resistance in P. aeruginosa.
The newly identified CpxR binding site is located upstream of the distal promoter of mexA in P. aeruginosa. To determine which promoter (distal or proximal) is activated by CpxR, two mexA promoter-lacZ reporter systems were constructed. To monitor the expression of the distal promoter, the entire proximal promoter region was excluded in the mexAp M2 ::lacZ construct; to monitor the expression of the proximal promoter, a key nucleotide within the -10 region of  the distal promoter [45] was disrupted in the mexAp M3 ::lacZ construct ( Fig 1B). Ectopically expressed CpxR strongly activated expression of mexAp M2 ::lacZ, but not mexAp M3 ::lacZ, in the PA14ΔcpxR strain (Table 1), indicating that CpxR is involved only in regulation of the distal mexA promoter, which is also directly regulated by the MexR repressor in P. aeruginosa. The possible interplay between CpxR and MexR on expression of the distal mexA promoter prompted us to investigate the involvement of CpxR in the multidrug resistance phenotype of nalB-type P. aeruginosa, which has been associated with defective MexR in laboratory and clinical isolates [13,16,17]. We monitored mexA expression levels in strains of various genetic backgrounds, including mexR null-mutant PA14ΔmexR, cpxR null-mutant PA14ΔcpxR, cpxR/ mexR double-mutant PA14ΔcpxRΔmexR, and wild-type PA14 strains. The mexA expression level of the PA14ΔmexR strain was significantly higher than that of the wild-type PA14 strain (Fig 3), a result similar to that previously reported for nalB-type P. aeruginosa [12]. Moreover, the lack of CpxR in the PA14ΔcpxRΔmexR strain resulted in decreased mexA expression ( Fig  3). To further confirm the regulatory influence of CpxR on expression of MexAB-OprM, the relative transcript level of mexB and protein level of MexA were investigated in the wild-type and mutant strains by quantitative real-time PCR and western blot analysis, respectively. The regulatory patterns of CpxR on the expression of the chromosomal genes were similar to that of the mexAp::lacZ reporter system in the tested strains (Fig 3).
To evaluate the influence of changes in mexA expression on drug resistance, the MICs of antibiotics were determined. In the PA14ΔmexR strain, the MICs of the antibiotics were significantly increased in comparison with those of the parental PA14 strain. Moreover, a lack of cpxR led to decreased MICs in the PA14ΔcpxRΔmexR strain in comparison with those of the PA14ΔmexR strain. These results indicate that the MICs of antibiotics in the tested strains are correlated well with their expression levels of mexA (Table 3). A similar effect was observed when cpxR (PA3204) was deleted in a mexR-deleted mutant of another standard laboratory P. aeruginosa strain, PAO1 (PAO1ΔmexR), indicating that the observed effect was not specific to a particular P. aeruginosa strain (Table 3).
CpxR activates expression of mexAB-oprM and enhances multidrug resistance in nalB-type P. aeruginosa resistant isolates from the laboratory and a clinical setting Defective mexR is a genetic trait associated with the clinical emergence of multidrug resistance in P. aeruginosa during antibiotic treatment [16]. Previously, mexR defective strains were selected in vitro by plating susceptible P. aeruginosa strains on agar medium containing lethal  The expression levels of mexAp::lacZ reporter as measured by β-galactosidase assay in the PA14, PA14ΔcpxR, PA14ΔmexR, and PA14ΔcpxRΔmexR strains. Each bar represents the mean ± SD of levels of a fluorquinolone antibiotic (ofloxacin or ciprofloxacin) and a third-generation cephalosporin antibiotic (cefsulodin or cefoperazone) [10,11,46,47]. In this work, PA14 cells were plated on agar medium containing lethal levels of ofloxacin and cefsulodin antibiotics, after which 40 ofloxacin-cefsulodin resistant (OCR) colonies were randomly collected for further analysis. Among the selected colonies, five isolates exhibited significantly reduced expression levels of mexA, as well as reduced MICs of ciprofloxacin, ceftazidime, and aztreonam, when cpxR was deleted (Table 4). When the mexR sequences of the five isolates were analysed, PA14OCR16, PA14OCR24, PA14OCR28, and PA14OCR32 were found to harbour defective mutations. The plasmid harbouring cpxR, but not cpxR D52A , complemented the phenotype, indicating that CpxR mediated the observed alteration in isolate PA14OCR16 ( Table 4). The pattern of CpxR-dependent activation of mexAB-oprM expression and enhancement of multidrug resistance in the four OCR isolates with defective mutations in mexR was identical to that of the engineered PA14ΔmexR strain ( Table 3), implying that CpxR might perform a common function in mexR-defective nalB-type P. aeruginosa strains. Interestingly, the fifth OCR isolate, PA14OCR36, had an intact mexR gene. In this particular isolate, the expression level of cpxP was drastically increased with respect to those of mexRdefective isolates PA14OCR16, PA14OCR24, PA14OCR28, and PA14OCR32. In parallel, the mexA expression level of isolate PA14OCR36 was comparable to those of the mexR-defective strains ( Table 4). Deletion of cpxR from isolate PA14OCR36 resulted in drastically decreased expression levels of cpxP and mexA. The plasmid harbouring cpxR, but not cpxR D52A , complemented the phenotype, indicating that CpxR is important for the observed alteration in cpxP expression in isolate PA14OCR36 (Table 4). Sequence analysis indicated the cpxR, nalC, and nalD genes of isolate PA14OCR36 were intact, suggesting that this isolate was distinct from constitutively active CpxR mutants or previously known nalC-or nalD-type mutants [12,14]. These results indicate that CpxR could override repression by MexR upon expression of the mexAB-oprM operon in isolate PA14OCR36, a newly identified nalB-phenotype OCR isolate of P. aeruginosa.
two independent experiments (**, p < 0.01). (B) Relative mexB transcript levels determined by quantitative real-time PCR in the PA14, PA14ΔcpxR, PA14ΔmexR, and PA14ΔcpxRΔmexR strains. Data are expressed relative to the quantity of mexB mRNA in the wild-type PA14 strain. Each bar represents the mean ± SD of the relative quantity in three independent experiments (*, p < 0.05). (C) Relative MexA protein levels were determined by western blotting in the PA14, PA14ΔcpxR, PA14ΔmexR, and PA14ΔcpxRΔmexR strains. The intensity of each band was quantified. The results are expressed relative to the quantity of MexA in the wildtype PA14 strain. Each bar represents the mean ± SD of the relative quantity in three independent experiments (*, p < 0.05). OCR isolate PA14OCR33 has a null mutation in the nalD locus and elevated mexA expression; mexA expression in PA14OCR33 is independent of CpxR, because deletion of cpxR did not alter the expression level of mexA in this strain (Table 4). This result confirms that CpxR plays no role in the regulatory influence of the NalD repressor on the expression of mexAB-oprM in P. aeruginosa.
To evaluate the importance of CpxR under clinical conditions, we obtained P. aeruginosa clinical isolates from the Department of Microbiology, Chinese People's Liberation Army General Hospital (Beijing, China). Fifty independent clinical isolates exhibiting ciprofloxacin and ceftazidime resistance were analysed, among which three isolates, LAR005, LAR023, and LAR048, exhibited significantly reduced expression levels of mexA, as well as reduced MICs of ciprofloxacin, ceftazidime and aztreonam, when cpxR was deleted (Table 5). When the mexR sequences of clinical isolates LAR005, LAR023, and LAR048 were analysed, each was found to harbour frameshifted or nonsense mutations at different sites in the mexR coding region (Table 5). Taken together, these results indicate that CpxR plays an important role in modulating multidrug resistance in nalB-type P. aeruginosa isolates generated in the laboratory and collected in the clinic.

Discussion
In this work, we applied comparative genomic analysis to illuminate the regulatory networks responsible for modulating RND efflux pump expression in P. aeruginosa. Similar comparative genomic analysis has been performed to identify novel regulons based on conserved DNA motifs on the promoter regions of potential target genes as binding sites of global regulators [7,48,49]. With the accumulation of whole-genome sequencing and transcriptomic data, comparative genomic analysis has become a powerful approach for identifying common or species-specific genetic regulatory networks among different species. Our work has demonstrated a novel regulatory linkage between CpxR and MexAB-OprM, an important efflux pump in P. aeruginosa. The significance of this regulatory linkage is several-fold: first, the regulatory influence of CpxR on RND efflux pump expression, even for pumps within the same orthologous group, could be very different among bacterial species. The direct regulatory influence of CpxR on expression of the mexAB-oprM orthologous operon is unique in P. aeruginosa (see Fig 1 and Table 1). Furthermore, it has been observed that the mdtABCD operon (encoding a RND efflux pump) possesses CpxR binding sites on its promoter region, while its expression is negatively regulated by CpxR in E. coli under conditions that activate the Cpx system [42]. In contrast, in this work, we demonstrate that the muxABC-opmB operon, an orthologue of the mdtABCD operon from E. coli (see S2 Table), is directly activated by CpxR in P. aeruginosa (see Table 1). Second, given that CpxR is involved in positive regulation of RND efflux pump expression in Vibrio cholerae [36] and P. aeruginosa, bioinformatics studies predict that VexAB/VexGH from V. cholerae [36] and MuxABC/MexAB from P. aeruginosa belong to different orthologous groups (for details, see S2 Table). Taken together, our observations suggest that the involvement of CpxR in regulating RND efflux pump expression may be evolutionarily divergent among bacterial species.
Incorporation of the MexAB-OprM efflux pump into the CpxR regulon reinforces the physiological importance of this efflux pump in P. aeruginosa. Indeed, the MexAB-OprM efflux pump plays profound physiological roles in addition to its role in antibiotic resistance in P. aeruginosa, including quorum sensing signal trafficking [50] and mediating bacterial virulence in hosts [51][52][53]. These functions of MexAB-OprM suggest that the regulatory linkage between CpxR and MexAB-OprM might have other purposes in addition to its role in antibiotic resistance in P. aeruginosa.
In this work, we have demonstrated that CpxR plays an important role in multidrug resistance by directly activating expression of mexAB-oprM in nalB-type P. aeruginosa isolates generated in the laboratory and collected in the clinic. Direct regulation of mexAB-oprM by CpxR suggests the existence of multiple pathways through which the expression level of the MexA-B-OprM efflux pump might be elevated in P. aeruginosa. The Cpx system is involved in the cellular response to misfolded membrane proteins in E. coli [22] and V. cholerae [54]. Recently, several works have demonstrated the existence of a resistome in the genome of P. aeruginosa consisting of a broad array of genes belonging to different functional families, which give rise to decreased susceptibility to antibiotics when they are mutated [55][56][57][58][59]. Interestingly, a number of genes encoding membrane proteins belong to the resistome [58,59]. Future studies should assess whether CpxR is activated in response to misfolded membrane proteins as a means of determining whether it contributes to resistome expression in P. aeruginosa. Unlike the cpxRA operons in E. coli and V. cholerae, the cpxR locus is not directly linked to the sensor kinase gene locus in the genome of P. aeruginosa. Characterization of the signalling mechanism underlying the newly identified regulatory linkage between CpxR and MexAB-OprM, as well as identification of candidate CpxA sensors in P. aeruginosa, is underway. The combined effects of various signals mediated by multiple regulators, including CpxR and MexR, on Mex-AB-OprM expression will be understood in a broader physiological context in the near future.

Comparative genomic analysis
For the determination of putative orthologous proteins, a primary BLASTP search in a given genome was conducted for the gene with the highest similarity. Next, additional searches for conserved functional motifs were conducted based on the literature when appropriate. Sequence retrieval and BLASTP searches related to whole-genome sequenced Pseudomonas species/strains were conducted using the Pseudomonas database (http://www.pseudomonas. com) [39], whereas other species/strains were analysed using the KEGG database (http://www. genome.jp/kegg/). Sequence similarity was determined using the online Pairwise alignment tool (http://www.ebi.ac.uk/Tools/psa/emboss_water/).
For the inter-species analysis, conserved DNA motifs were obtained by alignment of the intergenic regions preceding the orthologous RND efflux pump operons using the MEME suite of online software [60]. For the intra-species analysis for putative CpxR binding sites, an online DNA motif search programme (http://www.pseudomonas.com/replicon/setmotif) was used to scan the entire genome of P. aeruginosa PA14 entering GTAAAN (4,8)GTAAA as the query form.

Deletion of gene loci in P. aeruginosa strains
Generation of gene-locus-deleted P. aeruginosa strains was conducted using a method described previously [61]. For each gene, an upstream region including the start codon (longer than 500 bp) and a downstream region containing the stop codon (longer than 500 bp) were PCR-amplified and linked together. The resulting fragment was cloned into the suicide plasmid pEX18Tc. A fragment containing the FRT gentamicin-resistance (Gm) cassette from plasmid pPS856 was then inserted between flanking regions of the plasmid. The gene locus of each P. aeruginosa strain was then replaced with the plasmid by double-crossover homologous recombination. The Gm-resistance marker in the chromosome was removed by introducing plasmid pFLP2, which carries the Flp recombinase gene. Correct deletion in the constructed mutant was verified by PCR using primers that bound to flanking chromosomal regions of the fragments cloned into pEX18Tc. All DNA primers used in this study are listed in S3 Table. Construction of promoter-lacZ reporter gene fusion products and β-galactosidase assays The promoter region of each gene was PCR-amplified and TA-cloned into the pEASY-T1 vector (TransGen, China). Site-directed mutagenesis was performed using a protocol described previously [62]. Disruption of the CpxR binding site on the mexA promoter (mexAp M1 ) was performed by altering the 5 0 -GTAAACCTAATGTAAA-3 0 sequence to 5 0 -GTAAACCTAA TACAAA-3 0 . Exclusion of the entire proximal promoter of mexA (up to 162 bp from the mexA ATG codon) was performed by PCR-amplifying the distal promoter only (mexAp M2 ). Disruption of the -10 region of the mexA distal promoter (mexAp M3 ) was performed by altering the 5 0 -TATTTT-3 0 sequence to 5 0 -TGTTTT-3 0 . Once confirmed by sequencing, the promoter regions were subcloned into the broad-host, low-copy-number plasmid pMP190 [63]. The resulting plasmids were introduced into Pseudomonas strains by conjugal transfer from E. coli donor strain ST18 [64]. For the β-galactosidase assays, cells were grown overnight in Muller Hinton broth (Oxoid) supplemented with appropriate antibiotics, after which they were diluted 1:50 in 5 mL of fresh medium in 50-mL culture flasks at 37°C (30°C for P. putida KT2440) with mixing at 150 rpm. Cells were recovered during the logarithmic growth phase (OD 600 = 0.5-1.2). β-galactosidase assays were performed as described by Miller [65]. The results are expressed as the mean values from two independent experiments with triplicate samples.

Construction of in trans CpxR expression plasmids and purification of His 6 -CpxR
In order to control in trans CpxR expression, the lacI q -tacP region was PCR-amplified using the pME6032 plasmid [66] as a template and cloned into broad-host plasmid pBBR1MCS5 to replace the original constitutively expressed lac promoter [67]. Next, the CpxR and CpxR D52A in trans expression systems (pCpxR and pCpxR D52A ) were constructed using the altered plasmid. IPTG (0.2 mM) was added to induce CpxR overexpression. It was noted that the basal level of CpxR expression in the absence of IPTG was sufficient to complement the cpxR deletion mutants in P. aeruginosa.
The plasmid used to express the N-terminal His 6 -tagged CpxR proteins was constructed by PCR-amplifying the CpxR coding sequence and cloning it into pET28a (Novagen). The plasmid was transformed into E. coli expression host strain BL21(DE3) and grown to an OD 600 of 0.8 at 37°C with vigorous shaking in 1 L of LB medium containing kanamycin (50 μg/mL). The cells were then induced with 1 mM IPTG and allowed to express overnight at 22°C, after which they were harvested by centrifugation. The resulting pellet was resuspended in 10 mL of precooled buffer A (20 mM Tris-HCl, 200 mM NaCl, 1 mM imidazole, pH 8.0) and centrifuged at 4°C for 10 minutes at 3,500 rpm, after which the pellet was resuspended in 60 mL of pre-cooled buffer A. The cells were disrupted by sonication at 180 W for 8 minutes. The debris and membranes were removed by centrifugation at 4°C for 60 minutes at 15,000 rpm. The soluble fraction was passed through a 0.2-μm filter and loaded onto a 5-mL nickel column which was previously washed with 10 column volumes of ddH 2 O and equilibrated with 10 column volumes of buffer A. CpxR proteins were eluted with a mixture of buffer A and buffer B (20 mM Tris-HCl, 200 mM NaCl, 500 mM imidazole, pH 8.0), in which the proportion of buffer B was gradually increased from 0% to 100%. The flow speed was 1 mL/min during the elution process. The protein was collected when its protein peak appeared. The CpxR protein solution was desalted and concentrated to a final volume of 1.5 mL in buffer C (20 mM Tris-HCl, 200 mM NaCl, pH 8.0). The concentration of CpxR protein was determined by the Bradford method. CpxR protein was stored in buffer C supplemented with 50% glycerol at -80°C.

EMSAs
Purified N-terminal His-tagged CpxR proteins were phosphorylated using acetyl phosphate (AP) as previously described [68]. Briefly, 1.6 μM of purified CpxR was incubated with 50 mM AP in a reaction buffer containing 100 mM Tris-HCl (pH 7.4), 10 mM MgCl 2 , and 125 mM KCl at 30°C for 2 hours. The mobility shift assay was carried out using the 2 nd generation DIG Gel Shift Kit (Roche). Briefly, 165 bp DNA fragments of the promoter region of mexAp and mexAp M1 , 140 bp DNA fragments of the promoter region of cpxPp were PCR-amplified, after which 150 nM of purified PCR product was DIG-labelled according to the manufacturer's instructions. The binding reaction was carried out with different concentrations of phosphorylated CpxR (as great as 160 nM) and 0.2 nM DIG-labelled DNA fragments at 37°C for 30 min. The samples were separated by electrophoresis on 6% native polyacrylamide gels, transferred to Hybond-N blotting membranes (Amersham), and visualized by chemiluminescence.

DNase I footprinting assay
The DNase I footprinting assays were performed as previously reported [69]. Briefly, 400 ng of the probe (final concentration of 50 nM) was incubated with 1.5 μg of phosphorylated CpxR (final concentration of 1.5μM) in a total volume of 40 μL. After incubation for 30 min at 25°C, 10 μL of a solution containing approximately 0.015 units of DNase I (Promega) and 100 nmol of freshly prepared CaCl 2 was added. Following incubation for 1 min at 25°C, the reaction was stopped by adding 140 μL of DNase I stop solution (200 mM unbuffered sodium acetate, 30 mM EDTA, and 0.15% SDS). Samples were extracted with phenol/chloroform and precipitated with ethanol, after which the pellets were dissolved in 30 μL of Milli-Q water. The preparation of the DNA ladder, electrophoresis, and data analysis were performed as described before [69], except that a GeneScan-LIZ500 size standard (Applied Biosystems) was used.

Quantitative real-time PCR assay
An overnight culture (approximately 16 h) was diluted 1:100 in Mueller-Hinton broth and grown to the logarithmic growth phase (OD 600 = 0.4~0.6). Total RNA was extracted from 500 μL of cultured cells using the RNAprep Bacterial Kit (TianGen, China). Residual genomic DNA was digested by RQ1 RNase-Free DNase (Promega, USA). RNA samples were quantified using the Take3 Micro-Volume Plate function of a BioTek Synergy Neo Multi-Mode Reader. cDNA was synthesized from 1 μg of total RNA using TransScript cDNA Synthesis SuperMix (TransGen, China) according to the following procedure: after annealing of the RNA sample and the random hexamer primer for 5 min at 65°C, reverse transcription was carried out for 2 min at 25°C and 55 min at 42°C, followed by reverse transcriptase inactivation for 5 min at 70°C. An Opticon2 Realtime PCR system (Bio-Rad, Hercules, CA, USA) and SuperReal Premix SYBR Green Plus (Tian-Gen, China) were used to perform quantitative PCR on a 1-μL sample of diluted cDNA (1:10) according to the following procedure: one denaturation cycle for 15 min at 94°C and 40 amplification cycles for 10 s at 94°C, annealing for 20 s at 60°C, extension for 20 s at 72°C. Control samples without reverse transcriptase confirmed the absence of contaminating DNA in any of the samples. The housekeeping gene rpsL was used as the internal reference gene. Relative expression of mexB was calculated according to the 2 −ΔΔCT method [70] from three independent experiments. Primers for mexB and rpsL were designed as previously reported [17,71] Western blot assay For the western blot detection of CpxR protein in PA14ΔcpxR containing pCpxR or pCpxR D52A , an overnight culture was diluted 1:100 in Mueller-Hinton broth and grown to the logarithmic growth phase (OD 600 = 0.8-1.2). Total protein was extracted using a Bacterial Protein Extraction Kit (CWBiotech, China). Next, 5 μg of total protein and 10 ng of purified His-tagged CpxR were resolved in 10% SDS-polyacrylamide gels and transferred electrophoretically to PVDF membranes (Millipore, USA). Electrophoretic transfer of proteins was carried out for 1 h at 4°C with 200 mA of constant current. The blotted membranes were subsequently blocked in phosphate-buffered saline containing 0.1% (vol/vol) Tween-20 (PBST) and 5% (wt/vol) skim milk (Difco) for 60 min. The membranes were incubated with primary anti-CpxR rabbit polyclonal antibodies (1:10000) in PBST containing 5% (wt/vol) skim milk at 37°C for 60 min, after which they were washed three times (5 min each) with PBST and three times (5 min each) with PBS. The membranes were incubated with secondary goat anti-rabbit IgG antibodies conjugated to horseradish peroxidase (HRP) in PBST containing 5% (wt/vol) skim milk, after which they were washed three times (5 min each) with PBST and three times (5 min each) with PBS. All washes were carried out at room temperature with agitation. Substrates for HRP were obtained from the Amersham ECL Western Blotting Detection Kit (Amersham) and used according to the manufacturer's instructions. The enzymatic activity of HRP was detected using a 4200 Chemiluminescence Analyzer (Tanon, China).
To allow detection of MexA protein by western blotting, cell membrane proteins were isolated by ultracentrifugation. Briefly, an overnight culture was diluted 1:100 in Mueller-Hinton broth and grown to the logarithmic growth phase (OD 600 = 0.4-0.6). After sonication, total cell membrane protein was extracted from 40 mL of cultured cells using a Bacterial Membrane Protein Isolation Kit (Tiandz, Inc., China) according to the manufacturer's instructions. The ultracentrifuged cell membrane protein pellets were resuspended in 100 μL of H 2 O. The concentrations of the membrane proteins were quantified using the Take3 Micro-Volume Plate function of a BioTek Synergy Neo Multi-Mode Reader. SDS-PAGE and immunoblotting for cell membrane proteins (10 μg of each sample) were performed as described above, except for the following: SDS (0.1% (wt/vol)) was included in the blotting buffer, the transfer was carried out for 16 h at 4°C with 25 mA constant current, and anti-MexB rabbit polyclonal antiserum (1:10000) was used as the primary antibody. Band intensity was quantified in three independent experiments using Image-pro Plus version 6.0.

Antibiotic susceptibility test
The MIC of each antibiotic was determined on Muller Hinton agar by the two-fold dilution method. Mueller-Hinton agar plates containing serial twofold dilutions of each antibiotic (from 0.03125 to 32 μg/mL for ciprofloxacin and ofloxacin, from 0.125 to 128 μg/mL for the other antibiotics) were prepared. Overnight bacterial cultures were diluted 1:100 in fresh Mueller-Hinton broth, grown to the mid-logarithmic phase (OD600 of 0.4-0.6), harvested, and washed in PBS. The Mueller-Hinton agar plates were spotted with 3 μL of the diluted bacterial suspensions (approximately 10 4 cfu). The MIC was defined as the concentration at which bacterial growth was completely inhibited after incubation at 37°C for 24 hours. Ciprofloxacin, ofloxacin, and amikacin were purchased from Bio Basic Inc. Ceftazidime was purchased from Sigma-Aldrich. Cefsulodin was purchased from TOKU-E (Japan). Aztreonam was purchased from Selleck.
Generation of ofloxacin-cefsulodin resistant isolates in the laboratory P. aeruginosa PA14 cells (approximately 4 × 10 9 cells) grown overnight in LB broth medium were plated on LB agar containing 1.2 μg/mL ofloxacin and 2.4 μg/mL cefsulodin. After incubation at 37°C for 72 hours, resistant colonies appeared at a frequency of approximately 10 −7 .  isolates for this study. We are also grateful to Prof. Taji Nakae (Kitasato University, Japan) for providing anti-MexA rabbit polyclonal antiserum. We are also grateful to Prof. Hui Wang (Department of Clinical Laboratory, Peking University People's Hospital, Beijing, China) for helpful discussion.