Interaction of ArmZ with the DNA-Binding Domain of MexZ Induces Expression of mexXY Multidrug Efflux Pump Genes and Antimicrobial Resistance in Pseudomonas aeruginosa

The part of MexZ involved in the interaction with ArmZ was defined by the use of a bacterial two-hybrid system, based on the reconstitution of Bordetella pertussis adenylate cyclase (CyaA) (BACTH) (41). Escherichia coli BTH101 (Δcya) cells were transformed with pairs of complementary BACTH vectors containing cyaT18 and cyaT25 fragments fused with mexZ or armZ. Transformants were selected on MacConkey plates with 1% maltose. Red color of colonies of transformants and high β-galactosidase activity in the corresponding cell extracts, indicating CyaA reconstitution, were observed only for transformants containing plasmids with cyaT18-mexZ plus cyaT25-mexZ, cyaT18-armZ plus cyaT25-mexZ, and cyaT18-mexZ plus cyaT25-armZ but not for remaining combinations, confirming that MexZ can self-interact and that MexZ and ArmZ interact with each other (Fig. 1A). These data indicated that the BACTH system could be used to study molecular details of the MexZ-ArmZ interaction.

To identify the part of MexZ involved in interactions with ArmZ, we designed a series of truncated MexZ variants (Fig. 1B), based on the crystal structure of MexZ (29). mexZ alleles coding for truncated proteins deprived of short unstructured regions or adjacent alpha helices, from either the N terminus or the C terminus, were PCR amplified and cloned into BACTH vectors. Since MexZ-ArmZ interactions were observed only when the Cya fragment was linked to the N terminus of MexZ (data not shown), all shortened mexZ alleles were cloned into pLKB4, facilitating N-terminal translational fusion with CyaT18, and the constructed plasmids were introduced into E. coli BTH101 carrying pKAB25.2 (cyaT25-armZ) or pKAB25.1 (cyaT25-mexZ). Analysis of transformants grown on MacConkey plates and β-galactosidase activity measurements showed that the lack of the N-terminal 8 amino acids (aa) diminished the interaction with ArmZ, whereas the lack of the N-terminal 26 amino acids completely abrogated the MexZ-ArmZ interaction (Fig. 1C and D). Importantly, MexZ ΔN26 showed an interaction with full-length MexZ, proving that the hybrid protein was at least partially functional in the BACTH system (Fig. 1E and F). Concomitantly, the lack of C-terminal alpha helices (ΔC9, ΔC31, and even ΔC61) comprising the MexZ dimerization domain did not block MexZ-ArmZ heterodimerization completely (Fig. 1C and D), while it significantly reduced the self-association properties of MexZ (Fig. 1E and F). All truncated CyaT18-MexZ variants were expressed in the corresponding E. coli BTH101(pKAB25.2) cells, although a decreased level relative to that of full-length CyaT18-MexZ was observed for some of them (Fig. 1G). Remarkably, the reduced protein level not only might be caused by a deleterious effect of the truncation on protein stability but also might be a consequence of the impaired interaction with ArmZ, as we observed that the presence of CyaA reconstitution in E. coli BTH101 triggered by the interaction between MexZ and ArmZ stabilized CyaT18-containing fusion proteins (see Fig. S1 in the supplemental material). To check if the truncations affected the ability of MexZ to repress transcription from the mexXY promoter (mexXYp), we expressed the CyaT18-MexZ proteins in E. coli containing promoter-probe plasmid pKAB10.1 with the mexX-mexZ intergenic region cloned upstream of a promoterless xylE cassette (mexXYp-xylE). Measurement of XylE activity in extracts of cells producing full-length CyaT18-MexZ showed a strong repression of mexXYp-xylE (Fig. 1H). No XylE activity decrease was observed when CyaT18 fusions with truncated MexZ variants were expressed (Fig. 1H), indicating the significance of residues at both the N terminus (DBD) and the C terminus (dimerization domain) for the repressor activity of MexZ.

The truncation analysis pointed out that the N-terminal MexZ region involved in interactions with ArmZ was also crucial for MexZ DNA-binding activity. To search for amino acid substitutions in MexZ that abolish MexZ-ArmZ interactions without interfering with MexZ DNA-binding activity, we developed a BACTH-based screening procedure (Fig. 2A). E. coli BTH101 cells carrying pKAB10.1 (mexXYp-xylE) and pKAB25C.2 (cyaT25-armZ) were transformed with a mexZ mutant library in pKAB18.1 (cyaT18-mexZ*). The library was obtained by in vitro hydroxylamine mutagenesis of the PCR-amplified mexZ gene, followed by high-efficiency cloning of mutated mexZ into the BACTH vector pLKB4. Transformants were plated onto MacConkey medium with 1% maltose and grown for 1 to 2 days, and white colonies (comprising ∼0.4% of all colonies) were restreaked onto MacConkey and L agar plates (Fig. 2A). The white color of streaked bacteria on indicator plates could be caused by an effect of the mexZ mutation on the interaction of the corresponding protein product with ArmZ, but it could also be caused by the lack of a mexZ insert, the presence of a premature stop codon, or a mutation that drastically destabilizes the protein. Hence, the ability of CyaT18-MexZ* to repress mexXYp was tested by spraying the streaks onto L agar plates with a catechol solution to assess XylE activity. The white streaks on both media, MacConkey and L agar plates sprayed with catechol, indicated the production of a MexZ variant that has repressor activity but is defective in the interaction with ArmZ.

Overall, 28 such clones were isolated. Sequencing of mexZ in these plasmids showed nucleotide changes introducing the R3S (2 clones), K6E (2 clones), R13H (2 clones), K53E (8 clones), K53M (4 clones), K53N (8 clones), K53Q (1 clone), and K53T (1 clone) amino acid substitutions into MexZ. The mutations impaired the MexZ interaction with ArmZ; however, only the K53E mutation completely abrogated the interaction (Fig. 2B and C). Concomitantly, all mutants retained the ability to repress mexXYp, as assessed by a plate assay (Fig. 2D) and measurements of XylE activity in extracts from exponentially growing cultures (Fig. 2E). The level of repression varied significantly between MexZ variants despite similar levels of expression of fusion proteins (Fig. 2F). The strongest repression, comparable to that of wild-type (WT) MexZ, was observed for MexZ_K53Q. Altogether, BACTH screening led to the identification of MexZ amino acids important for the interaction with its antirepressor, pointing out the significance of K53. Since only the K53E mutation completely abrogated the MexZ-ArmZ interaction, this variant was chosen for further analysis from the set of substitutions at this position.

To check the impact of R3S, K6E, R13H, and K53E substitutions in MexZ on the interaction with ArmZ in vitro, MexZ-His₆ (WT and mutants) was purified and used as the bait in a pulldown assay. Equal amounts of MexZ-His₆ (WT and mutants) were mixed with extracts from E. coli cells overproducing FLAG-ArmZ and loaded onto Ni-nitrilotriacetic acid (NTA) columns, followed by washing and elution of bound proteins with increasing concentrations of imidazole. Eluates were separated on polyacrylamide gels, which were stained with Coomassie brilliant blue or used for Western blot analysis. MexZ-His₆ eluted at imidazole concentrations of 0.15 M and 0.3 M (Fig. S2A). Concomitantly, these fractions contained large amounts of FLAG-ArmZ in the case of WT MexZ-His₆, observed even after Coomassie staining (Fig. S2A) or in Western blot analysis with anti-FLAG antibodies (Fig. 2G and Fig. S2B), whereas no FLAG-ArmZ was bound to a control column containing the purified His₆ tag (Fig. 2G and Fig. S2). The use of R3S, K6E, and R13H variants of MexZ-His₆ resulted in a significantly reduced amount of FLAG-ArmZ bound, whereas in the case of MexZ_K53E-His₆, FLAG-ArmZ was hardly detectable (Fig. 2G). Overall, these results are consistent with data from BACTH analyses and indicate that R3S, K6E, and R13H substitutions in MexZ disturb the interactions with ArmZ, whereas the K53E amino acid change completely abolishes them.

The N-terminal 8 residues encompassing R3 and K6, important for MexZ-ArmZ interactions, were missing in the solved structure of MexZ (29). Inspection of the structure of the N-terminal part of AcrR from Salmonella enterica serovar Typhimurium (42), the sequence of which is highly similar to that of the N-terminal part of MexZ (Fig. 2H, bottom), showed that the amino acids form two alpha helices with the N terminus bent toward other helices forming the DNA-binding domain (42). To visualize residues on a structural model of MexZ, we used I-TASSER (43) to refine the structure of the N-terminal part of MexZ. Prediction showed that the N terminus of MexZ forms a long alpha helix (Fig. 2H). In this model, R3, K6, R13, and K53 residues were localized in proximity to each other (Fig. 2H). This suggests that the identified amino acids may form an interface for interaction with ArmZ.

The impact of ArmZ on MexZ binding to DNA in vitro was analyzed using an electrophoretic mobility shift assay (EMSA) with purified MexZ-His₆ and the mexXY-mexZ intergenic region (mexXYp) as a probe, with or without His₆-ArmZ. MexZ-His₆ bound to mexXYp DNA but not to a similar-length control fragment (Fig. 3A). mexXYp was not bound by His₆-ArmZ. Strikingly, the addition of His₆-ArmZ to MexZ-His₆ prior to mixing it with mexXYp only partly interfered with MexZ-DNA binding under the conditions tested (Fig. 3A). Concomitantly, to test the impact of ArmZ on MexZ-mediated repression of mexXYp in vivo, we first used the E. coli DH5α(pKAB10.1 [mexXYp-xylE]/pKAB801 [araBADp-mexZ]) strain and determined that induction with 0.005% arabinose resulted in an amount of MexZ that was not capable of full mexXYp repression (Fig. 3B). However, even such moderate MexZ repression of mexXYp was not relieved by the overproduction of ArmZ. Similarly, the overproduction of ArmZ together with PA5470, a putative peptide-releasing factor (36, 37) encoded in the same operon, had no effect on MexZ-mediated repression (Fig. 3C). This strongly suggested that other factors present in P. aeruginosa, but not in the E. coli strain used, might participate in the control of the influence of ArmZ on MexZ binding to DNA.

To analyze the impact of R3S, K6E, R13H, and K53E amino acid substitutions in MexZ on the ArmZ-mediated increase of mexXY expression in P. aeruginosa, we constructed a series of PAO1161 mutants containing the corresponding missense mutations in mexZ as well as control strains with inactivated mexZ (ΔmexZ) or armZ (ΔarmZ) and the ΔmexZ ΔarmZ double-deletion strain. WT and mutant strains were grown in medium with or without a subinhibitory concentration of tetracycline, known to induce mexXY expression (33), and mexY as well as armZ transcript levels were analyzed using reverse transcription-quantitative PCR (RT-qPCR) (Fig. 4A and B). The addition of tetracycline resulted in increased armZ mRNA production in the WT and all mexZ missense mutants, with a slightly higher level of armZ expression observed in strains producing MexZ_K6E and MexZ_K53E than in the WT strain (Fig. 4B). As expected, tetracycline induced mexY expression in WT PAO1161, whereas only a minor increase was observed in the ΔarmZ mutant, suggesting that the bulk of the induction was ArmZ dependent (Fig. 4A). The lack of MexZ (ΔmexZ) or the production of MexZ_R3S led to high mexY expression levels independently of the presence of tetracycline (Fig. 4A). In the absence of an inducer, a significant increase in the mexY transcript level was observed in strains expressing MexZ_K6E and MexZ_R13H but not to the extent observed in ΔmexZ or mexZ_R3S strains, suggesting ongoing partial repression of the mexXY promoter in these two mutants (Fig. 4A). However, these variants were impaired in the ArmZ-dependent response to tetracycline, since the observed increase of the mexY mRNA level was significantly lower than that in the WT strain. The strongest repression of mexXYp in the absence and presence of the inducer was observed in cells expressing MexZ_K53E, confirming a minor effect of this amino acid substitution on the DNA-binding ability of MexZ but a very strong effect on the interaction with ArmZ under native conditions. Altogether, these data indicate that mutations that block the MexZ-ArmZ interaction may hamper both MexZ repression and tetracycline-mediated induction of mexY expression.

To check if excess ArmZ restores the defect in the MexZ response, the empty araC-araBADp expression cassette or the araC-araBADp-armZ cassette, allowing arabinose-induced overexpression of armZ, was introduced into the genomes of PAO1161 and its derivatives carrying mutant alleles of mexZ. Strains were grown in arabinose-containing medium, and RNA was isolated for RT-qPCR analysis (Fig. 4C and D). The extents of armZ overexpression were similar in all tested strains (Fig. 4D). The abundance of ArmZ triggered an increase in the mexY level in cells producing WT MexZ, to the level observed for the control ΔmexZ mutant strain (Fig. 4C). In agreement with the observed partial inhibition of the MexZ-ArmZ interaction by the K6E or R13H substitution in MexZ (Fig. 2), mexY expression was responsive to armZ overexpression in strains producing MexZ_K6E or MexZ_R13H (Fig. 4C). Strikingly, this was not observed in cells expressing MexZ_K53E (Fig. 4C), and akin to the ΔarmZ strain, the mutant producing MexZ_K53E was more susceptible to aminoglycosides and tetracycline than WT strain PAO1161 (Table 1), confirming the critical role of K53 for MexZ interactions with ArmZ. Overall, these data indicate that the mutations that disturb the MexZ-ArmZ interaction negatively affect ArmZ-mediated upregulation of genes encoding the MexXY multidrug efflux system, thus increasing the susceptibility of the cells to antimicrobials extruded by this pump.

To identify other bacterial multidrug efflux systems with expression potentially regulated by large ArmZ-like antirepressors alleviating the downregulation exerted by local TetR family transcription factors, we used MultiGeneBlast (44) to search for clusters of genes encoding proteins with similarity to P. aeruginosa MexX/MexY/MexZ in 1,748 representative and reference bacterial genomes included in the RefSeq database (release 91) (45). Genes encoding all three components in proximity to each other were identified in 141 strains (Data Set S1). Since ArmZ shows sequence similarity to RtcB family ligases, ubiquitous in prokaryotic genomes, the identification of ArmZ orthologs was based on the fact that in P. aeruginosa, armZ is transcriptionally coupled with PA5470 (36), a gene encoding a peptide chain release factor (37). A MultiGeneBlast search using P. aeruginosa ArmZ/PA5470 showed 61 genomes with operons containing putative orthologs (Data Set S2). Concomitantly, genomes of 27 strains contained genes encoding MexX/MexY/MexZ and ArmZ/PA5471 orthologs (Data Set S3). These were classified mostly as Enterobacteriales; however, one other Pseudomonas species (P. protegens CHA0) and seven Betaproteobacteria species were also identified (Fig. 5A and Data Set S3). Among these, Burkholderia ubonensis MSMB22 and Burkholderia cepacia ATCC 25416 contained multiple clusters encoding MexXYZ orthologs (Data Set S3). Phylogenetic analysis of MexZ orthologs showed that P. aeruginosa MexZ grouped with proteins identified in B. cepacia, B. ubonensis, or Achromobacter xylosoxidans (Fig. 5A). The alignment of amino acid sequences of MexZ orthologs showed a very high level of conservation of residues forming the N-terminal DNA-binding domain, including those corresponding to R3, K6, and R13 of P. aeruginosa MexZ, likely forming the MexZ-ArmZ interaction surface (Fig. 5B). Concomitantly, K53, critical for the MexZ-ArmZ interaction in P. aeruginosa, was replaced with serine only in the case of the MexZ ortholog from P. protegens and with alanine in the case of B. cepacia and B. ubonensis in one of the MexZ orthologs encoded by their genomes but not in the others (Fig. 5B). A substitution of arginine, which was unlikely to affect the interaction, was observed in the case of the B. cepacia protein under GenBank accession no. ALK16488.1 (Fig. 5B). A similar comparison of amino acid sequences of MexZ-like regulators from strains lacking ArmZ/PA5470 orthologs also demonstrated a high level of conservation of amino acid residues corresponding to R3, K6, and R13, whereas a lower level of conservation was observed for K53 (Fig. 5B and Fig. S4). These data show that a lysine is preferentially found at the position corresponding to K53 of P. aeruginosa MexZ in MexZ-like regulators in strains whose genomes also encode ArmZ/PA5470 orthologs. Taken together, our results suggest a widespread role of ArmZ-like proteins in the regulation of multidrug efflux pump expression in various bacterial species, including clinically important pathogens.

ArmZ interacts with MexZ to control the expression of genes encoding the MexXY efflux pump, which confers multidrug resistance to P. aeruginosa. The positively charged residues responsible for MexZ-ArmZ interactions are located at the N terminus of MexZ, suggesting an overlap between DNA- and ArmZ-binding domains. Similar cross talk may contribute to the induction of efflux-mediated antimicrobial resistance in other bacterial species. Further structural studies should shed light on the exact molecular mechanism of MexZ inactivation by ArmZ.

Strains used and constructed in this work are listed in Table 2. All P. aeruginosa strains used in this study are derivatives of PAO1161 [leu(r⁻ m⁻)] (restriction-modification defective, leucine-auxotrophic strain; provided by B. M. Holloway, Monash University, Clayton, Victoria, Australia). Escherichia coli strain DH5α was used for plasmid manipulations. Plasmids used and constructed in this study are listed in Tables S1 and S2, respectively, in the supplemental material. Oligonucleotides used in this work are listed in Table S3. All new plasmid constructs were verified by sequencing (Laboratory of DNA Sequencing and Oligonucleotide Synthesis, Institute of Biochemistry and Biophysics, Polish Academy of Sciences [IBB PAS]). Standard DNA manipulations were performed as described previously (61). Gibson assembly reactions were performed by the use of OverLap assembly mix (A&A Biotechnology, Poland).

Mutated alleles to be introduced into the PAO1161 genome by homologous recombination were cloned into suicide vector pAKE600 (62). The E. coli S17-1 strain was transformed with pAKE600 derivatives, and transformants were used as donors for conjugation with PAO1161 Rif^r. The allele exchange method was performed as described previously (63). The presence of the mexZ or armZ deletions as well as insertions in the PA5412-PA5413 intergenic region was verified by PCR, whereas the presence of mexZ missense mutations was verified by restriction analysis of PCR products (Table S2).

Bacteria were grown in L broth or on L agar (L broth with 1.5% [wt/vol] agar) at 37°C or 28°C. MacConkey agar (BD Difco) supplemented with 1% maltose was used in bacterial adenylate cyclase two-hybrid (BACTH) assays. Cation-adjusted Mueller-Hinton II broth (BBL) was used in antimicrobial susceptibility assays. Plasmids in E. coli strains were maintained by the addition of 150 μg ml⁻¹ benzylpenicillin sodium salt in liquid medium (300 μg ml⁻¹ for solid medium), 50 μg ml⁻¹ kanamycin, or 10 μg ml⁻¹ chloramphenicol. P. aeruginosa transconjugants were selected by the addition of 300 μg ml⁻¹ carbenicillin and 300 μg ml⁻¹ rifampin.

Protein-protein interactions were analyzed using the BACTH system (41). CyaT18 and CyaT25 fragments were fused to the N termini of the proteins of interest. Pairs of plasmids encoding the hybrid genes cyaT18-X and cyaT25-Y were transformed into E. coli BTH101 Δcya. The interaction between the proteins leading to the reconstitution of CyaA activity, triggering cAMP synthesis and the transcription of genes involved in the catabolism of carbohydrates (lactose or maltose), was identified as red color of colonies on MacConkey agar and by an increase of the β-galactosidase activity in the cell extracts. Bacteria were plated onto MacConkey medium with 1% maltose supplemented with appropriate antibiotics and grown for 1 day at 28°C. Three random clones for each type of transformant were streaked onto fresh MacConkey plates, incubated for 24 h or 48 h at 28°C, and photographed. The colonies were used to inoculate cultures in liquid L broth (grown for 20 h at 28°C) for β-galactosidase activity measurements as described previously (64).

To obtain the library of mexZ alleles in the pLKB4 vector (ΔcyaT18), the PCR-amplified open reading frame (ORF) was mutagenized in vitro and cloned into pLKB4. This strategy prevented the accumulation of random mutations in the plasmid backbone, including lacp-cyaT18. The PCR product encompassing mexZ, amplified from pKAB18.1 by the use of primers 44 and 45, was gel purified, and 2 μg of DNA in 1 ml of a solution containing 0.1 M potassium phosphate buffer (pH 6.0) and 1 mM EDTA was mixed with 0.8 ml of a solution containing 2 M hydroxylamine and 0.4 M NaOH and incubated for 90 min at 70°C, followed by purification with a Clean-up kit (A&A Biotechnology, Poland). The mutagenized PCR product, digested with EcoRI and SacI, was ligated with pLKB4 digested with same enzymes and treated with FastAP thermosensitive alkaline phosphatase (Thermo). The ligation mixtures were digested with BamHI to remove traces of the empty vector and used for the transformation of electrocompetent E. coli DH5α cells. Approximately 100,000 clones, with more than 99% containing the insert, were pooled, and the plasmid DNA was isolated to form the pKAB18.1 (cyaT18-mexZ*) library.

A search for MexZ variants impaired in interactions with ArmZ but capable of mexXYp repression was performed by transformation of the pKAB18.1 library into E. coli BTH101 Δcya cells carrying pKAB10.1 (mexXYp-xylE) and pKAB25C.2 (cyaT25-armZ) (Fig. 2A). Transformants were selected on MacConkey agar with 1% maltose supplemented with penicillin, kanamycin, and chloramphenicol at 28°C for 24 to 48 h. White colonies were streaked onto a fresh MacConkey agar plate and onto an L agar plate. After overnight growth, the L agar plate was sprayed with 10 mM catechol to assess the ability of the MexZ mutant to repress mexXYp-xylE. Overall, 120,000 transformants were screened, resulting in 28 clones producing proteins with a disturbed interaction with ArmZ but with the ability to repress mexXYp-xylE. Sequencing of mexZ in these clones showed point mutations resulting in 8 single-amino-acid substitutions (plasmids pKAB18.8 to pKAB18.15) (Table S2).

E. coli BL21(DE3) cells with pET28 derivatives encoding MexZ-His₆ (WT and mutants) or His₆-ArmZ were grown to an optical density at 600 nm (OD₆₀₀) of 0.6, induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG), and grown with shaking at 37°C for an additional 3 h. The cells were harvested by centrifugation and sonicated in lysis, elution, wash (LEW) buffer (50 mM NaPi [pH 8.0], 300 mM NaCl). His₆-tagged proteins were captured using affinity chromatography on Macherey-Nagel Protino Ni-TED 1000 columns, followed by washing (10 ml of LEW buffer) and elution using LEW buffer with 300 mM imidazole. The fractions were dialyzed overnight against LEW buffer with 10% glycerol, and proteins were stored at −80°C. Protein yield and purity were estimated by SDS-PAGE using the Pharmacia Phast gel system.

Purified MexZ-His₆ (150 μg) was mixed with 1 mg of proteins from a cell extract obtained by sonication of E. coli BL21(pKAB28.6) cells overproducing FLAG-ArmZ in LEW buffer with 10% glycerol. The mixtures were incubated at 22°C for 10 min and applied to Protino Ni-TED 150 columns (Macherey-Nagel) equilibrated with LEW buffer. The columns were washed with 5 ml of LEW buffer, and protein complexes were eluted by applying LEW buffer with increasing concentrations of imidazole. Samples of E. coli BL21 cells with the empty vector pKAB28.7, treated as described above for the protein purification protocol, were used as a negative control. Eluates were separated on 10% polyacrylamide gels and stained with Coomassie brilliant blue or analyzed by Western blotting by the use of mouse monoclonal anti-FLAG antibodies (catalog no. MA1-91878; Thermo).

Binding of MexZ-His₆ and His₆-ArmZ to DNA was analyzed by an electrophoretic mobility shift assay (EMSA). Probes were prepared by gel purification of PCR products obtained using pKAB10.1 as a template and primers 1 and 2 (mexXYp, 184 bp) or using pUC19 as a template and primers 46 and 7 (control, 204 bp). The recombinant protein(s) at the indicated concentrations was added to 0.1 μM probe in 20 μl of binding buffer (10 mM Tris-HCl [pH 8.5], 10 mM MgCl₂, 100 mM KCl, 0.1 mg ml⁻¹ bovine serum albumin [BSA]) and incubated at 22°C for 15 min, and the mixtures were separated by electrophoresis on a 1.5% agarose gel in 0.5× TBE buffer (45 mM Tris-HCl [pH 8.0], 45 mM boric acid, 1 mM EDTA) at 4°C. Gels were placed in 1 μg ml⁻¹ ethidium bromide in water for 10 min, followed by destaining in water (10 min or as long as required) before visualization of the bands using UV light. In competition experiments, MexZ-His₆ and His₆-ArmZ were mixed and equilibrated for 15 min in binding buffer before the addition of the probe.

RNA isolation was performed with the Total RNA minikit (A&A Biotechnology, Poland) using cells from 2 ml of cultures at an OD₆₀₀ of 0.5. RT-qPCR was performed essentially as described previously (65). Briefly, cDNA was synthesized with a TranScriba cDNA synthesis kit (A&A Biotechnology, Poland) using random-hexamer primers. qPCRs with Hot FIREPol EvaGreen qPCR mix plus (Solis Biodyne) were run in a Roche LightCycler 480 apparatus and analyzed by the relative expression method using Roche LightCycler 480 software (v1.5.1.62). Three technical replicates were used for each gene-primer combination. Oligonucleotides are listed in Table S3. nadB (PA0761) was used as a reference gene.

The MexZ structure, including the N-terminal part missing from the previously solved structure (PDB accession no. 2WUI) (29), was modeled using I-TASSER (43). Swiss PDB Viewer v4.1 (http://www.expasy.org/spdbv/) was used for application of symmetry and visualization of the residues on the model (66). Identification of genes encoding clustered orthologs of MexZ/MexX/MexY or ArmZ/PA5470 was performed using MultiGeneBlast (44). The database used for searches encompassed 1,748 representative and reference bacterial genomes included in RefSeq release 91 (45). The minimal sequence coverage of BLAST hits was set to 25%, and the minimal percent identity was set to 30%. The maximum distance between genes was limited to 20 kb. Protein sequence alignment and phylogenetic tree construction by the use of the neighbor-joining method were conducted using COBALT (https://www.ncbi.nlm.nih.gov/tools/cobalt/) (67). The alignment was visualized using SnapGene viewer (SnapGene software; GSL Biotech).

The activity of catechol 2,3-dioxygenase (XylE) in cell extracts was determined spectrophotometrically as described previously (68) and normalized to the protein content in cell extracts determined by using the Bradford assay (69). One mU is defined as the activity of enzyme resulting in the formation of 1 nmol of 2-hydroxymuconic semialdehyde per min. Western blot analysis of CyaT18-fused proteins was performed by the use of mouse anti-CyaA antibodies (3D1, catalog no. SC-13582; Santa Cruz Biotechnology). For Western blot analysis, cell pellets were resuspended in loading buffer (50 mM Tris-HCl [pH 8.0], 0.1 M dithiothreitol [DTT], 2% SDS, 0.1% bromophenol blue, 10% glycerol) and boiled for 10 min. MICs were evaluated by using a microdilution method in microtiter plates with cation-adjusted Mueller-Hinton broth (70). MICs for the control strain ATCC 27853 were 2 μg ml⁻¹ for amikacin, 1 μg ml⁻¹ for gentamicin, 0.5 μg ml⁻¹ for tobramycin, and 16 μg ml⁻¹ for tetracycline and were in the ranges indicated for this reference strain by the CLSI.