Puriﬁcation and Characterization of the Pseudomonas aeruginosa NfxB Protein, the Negative Regulator of the nfxB Gene

The protein NfxB, involved in conferring resistance to quinolones in Pseudomonas aeruginosa , has a helix- turn-helix motif which is similar to that of other DNA-binding proteins. It appears to affect the membrane-associated energy-driven efﬂux of some antibiotics (H. Nikaido, Science 264:382–388, 1994). We constructed a plasmid that overproduced NfxB in Escherichia coli and puriﬁed the protein. Two species of NfxB (23 and 21 kDa), which are probably translated from different initiation codons, were isolated. Both proteins are also expressed in vivo in P. aeruginosa , with the 23-kDa NfxB being the major species. NfxB speciﬁcally binds upstream of the nfxB coding region as demonstrated by gel retardation and DNase I footprinting. Expression of the (cid:70) ( nfxB (cid:42) -lacZ (cid:49) )(Hyb) gene was repressed in the presence of the nfxB gene product provided by a second compatible plasmid in E. coli . In the P. aeruginosa wild-type strain (PAO2142), NfxB was undetectable by immunoblotting; however, it was detected in the nfxB missense mutant (PK1013E). These results suggested that NfxB negatively autoregulates the expression of nfxB itself.

The intrinsic resistance of Pseudomonas aeruginosa to a large variety of antimicrobial agents was shown to be due mainly to efflux system effects and partly to the low-permeability outer membrane (10,14).
The P. aeruginosa nfxB mutants, which show a 16-fold increase in resistance to norfloxacin, were isolated spontaneously (7). The mutants also show hypersusceptibility to ␤-lactam and aminoglycoside antibiotics (17). Some nfxB mutants overproduce the 54-kDa outer membrane protein, OprJ (11), little of which is produced in the wild-type strain (7,17). Antibiotic resistance in the nfxB mutants is probably not due to altered outer membrane permeability but to multidrug efflux pump effects (14,18).
The nfxB gene cloned from the wild type as well as a mutant (the nfx13E mutation) has been sequenced (16,17). The amino acid sequence of NfxB revealed that it has a helix-turn-helix motif that might be responsible for its ability to bind in a sequence-specific manner to DNA, and it has no significant hydrophobic membrane-spanning regions (16). nfx13E is a missense mutation which replaces an arginine residue in the putative helix-turn-helix domain with a glycine residue in the mutant (16). The wild-type nfxB gene can restore susceptibility to norfloxacin in P. aeruginosa with the nfx13E mutation. This evidence suggests that NfxB binds to DNA and regulates the expression of genes that encode the outer membrane protein(s). It is possible that Nfx13E mutant protein is unable to bind to DNA and thus loses its regulatory function. In addition, nfxB expression might be autoregulated by NfxB protein itself. Like the LysR-type transcriptional regulators, many proteins that have helix-turn-helix motifs bind their regulatory region and negatively autoregulate their expression (4,20).
To examine the DNA-binding activity and the regulatory function of NfxB, we constructed a plasmid system that overproduced the protein and then purified it. The protein specifically bound to the regulatory region of nfxB itself. We also studied the regulation of expression of nfxB in Escherichia coli and P. aeruginosa and found that the gene is negatively autoregulated by its gene product.
pNF253, which contains the entire region of the nfxB gene, and pTM45 (multicopy vector of P. aeruginosa) have been described by Okazaki and Hirai (16). pNF253 was digested with EcoRI and made blunt ended with S1 nuclease. The 1.1-kb fragment which contains the entire region of the nfxB was isolated by digestion with HindIII. The fragment was ligated to pUC18 that had been digested with HincII and HindIII. The resultant plasmid, pNF3, was digested with HindIII, and the protruding HindIII end was filled in with T4 DNA polymerase. EcoRI (5Ј-CCGGAATTCCGG-3Ј) or BamHI (5Ј-CCCGGATCCGGG-3Ј) linkers were inserted between the polymerized ends to replace the HindIII site of pNF3 with an EcoRI (pNF3E) or BamHI site (pNF3B). The resultant plasmids, pNF3E and pNF3B, contain the 1.1-kbp nfxB fragment between the two EcoRI sites and the EcoRI-BamHI site, respectively. pNF20, which also carries the entire region of the nfxB gene on pSU2719 (3), a derivative of pACYC184, was constructed as follows. A BamHI fragment of pNF3B was ligated into the BamHI site of pSU2719. In the resulting plasmid, pNF20, the reading frame of the nfxB gene was placed in the same orientation as that of the lac gene.
To overproduce NfxB, we constructed pNF7 by inserting the 1.1-kbp EcoRI-BamHI fragment of pNF3E containing the nfxB gene into the EcoRI site of pT7-5 (23). The nfxB gene was inserted immediately downstream of the T7 promoter in the correct orientation.
Two ⌽(nfxBЈ-lacZ ϩ )(Hyb) fusion genes were constructed as follows. Two DNA fragments (334 and 272 bp) containing the promoter region and part of the coding region of nfxB were isolated from pNF3E. The 334-bp DNA fragment was isolated by digestion with EcoRI and HincII, and the 272-bp fragment was isolated by EcoRI digestion followed by AvaII digestion and polymerization with T4 DNA polymerase (see Fig. 7). These two fragments were ligated with pMC1403 (2) which had been digested with EcoRI and SmaI. The hybrid plas-mids, pNF334 and pNF272, harbored the 334-bp fragment and the 272-bp fragment, respectively, in frame.
Luria-Bertani (LB) liquid medium and LB plates (19) were used throughout this study.
Purification of NfxB protein. E. coli BL21(DE3) (22) was the host for the NfxB-overproducing plasmid pNF7. The strain was grown in 3 liters of LB medium containing 100 g of ampicillin per ml to an optical density at 600 nm of 0.5. NfxB overproduction was induced with a final concentration of 0.4 mM isopropyl-␤-D-thiogalactopyranoside (IPTG). After incubation for a further 3 h, the cells were harvested by centrifugation and the cell pellet was suspended in 50 ml of buffer A (20 mM Tris-HCl [pH 7.5], 2 mM ␤-mercaptoethanol, 10% glycerol) containing 50 mM NaCl. After the cells were disrupted by sonication (six 20-s bursts), a clear supernatant was obtained by centrifugation at 32,000 ϫ g for 1 h (fraction I). Polymin-P (pH 8.0) was added to the supernatant to a final concentration of 0.135%. After being stirred for 1 h, the suspension was centrifuged at 25,000 ϫ g for 20 min. The pellet was resuspended in buffer A containing 400 mM NaCl, stirred for 30 min, and centrifuged at 25,000 ϫ g for 20 min. The supernatant (fraction II) was brought to a final concentration of 40% (wt/vol) ammonium sulfate (6). This fraction was stirred for 1 h and centrifuged at 20,000 ϫ g for 20 min. The pellet was resuspended in buffer A containing 20 mM NaCl and dialyzed against buffer A containing 20 mM NaCl (fraction III). Fraction III was applied to a DEAE Bio-Gel A (Bio-Rad) column (bed volume, 50 ml) equilibrated with buffer A containing 20 mM NaCl. NfxB was passed through the column with buffer A containing 20 mM NaCl. The pooled flowthrough fraction (fraction IV) was dialyzed against buffer B (10 mM NaPO 4 buffer [pH 7.0], 5 mM ␤-mercaptoethanol, 10% glycerol) containing 20 mM NaCl. The dialyzed fraction IV was applied to a CM Bio-Gel A (Bio-Rad) column (bed volume, 15 ml) equilibrated with buffer B containing 20 mM NaCl. The column was washed with 3 volumes of buffer B containing 20 mM NaCl, and proteins were eluted with a 150-ml linear gradient of 20 to 300 mM NaCl in buffer B. NfxB eluted at about 200 to 250 mM NaCl. These fractions were pooled (fraction V) and directly applied to a phosphocellulose column (bed volume, 8 ml) which had been equilibrated with buffer B containing 250 mM NaCl. The column was washed with 3 column volumes of buffer B containing 250 mM NaCl, and proteins were eluted with a 100-ml linear gradient of 250 to 1,000 mM NaCl. NfxB was eluted at about 400 to 450 mM NaCl and then was pooled and dialyzed against storage buffer (20 mM Tris-HCl [pH 8.0], 80 mM NaCl, 5 mM ␤-mercaptoethanol, 50% glycerol) (fraction VI). The protein was stored at Ϫ20ЊC.
DNA-binding experiments. The interaction of NfxB with the regulatory regions of nfxB was studied by means of a gel retardation assay (24) and by a DNase I protection experiment (8).
␤-Galactosidase assay. E. coli CSH26 carrying appropriate plasmids was cultured to an optical density at 600 nm of 0.6, and ␤-galactosidase activity in the cells was measured as described by Miller (13).
Separation of the 23-and 21-kDa NfxB proteins, and preparation of anti-NfxB serum. Fraction VI (1.5 mg), which contained the 23-and 21-kDa NfxB proteins, was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and purified with a model 491 Prep Cell (Bio-Rad) as described by the manufacturer. The 21-kDa NfxB was injected into rabbits to raise antiserum against NfxB (21).
Other procedures. All DNA manipulations were done by the methods in reference 19. SDS-PAGE was performed as described by Laemmli (9). Proteins were identified by immunoblotting (21), and primer extension experiments were performed as previously described (19). The amino termini of purified 23-kDa (fraction VIb) and 21-kDa (fraction VIs) proteins were sequenced by the Center for Instrumental Analysis (Hokkaido University) with a peptide sequencer (model ABI477A; Applied Biosystems). The intensity of the bands were quantified with an ImageQuant (Molecular Dynamics). The protein concentrations were determined with bovine serum albumin as the standard, following the Bradford method (1).

RESULTS
Overproduction and purification of NfxB. E. coli BL21 (DE3) carrying pNF7 overproduced a 23-kDa protein that constituted about 3% of the total protein after induction with IPTG (Fig. 1, lane 2). Since this protein was not found in the extract of the control cells that carried pT7-5 (the pNF7 parent vector without the nfxB insert [ Fig. 1, lane 1]) and the size of the protein agreed with that of NfxB deduced from the DNA sequence (2,1135 Da), the overproduced protein was probably the product of the nfxB gene.
The 23-kDa protein was purified from BL21(DE3) cells carrying pNF7 after IPTG induction. The purification steps of the protein were monitored by SDS-PAGE (Fig. 1, lanes 3 to 8).
The final purified fraction (fraction VI) contained two proteins (23 and 21 kDa) (Fig. 1, lane 8). These proteins were not separated by sequential chromatography, because the molec-ular masses are very similar. We separated these proteins with a model 491 Prep-Cell, as described in Materials and Methods, and analyzed their amino acid sequences. The sequence of the 21-kDa protein (fraction VIs) from the amino terminus was TLISHDERLI, which completely agreed with the sequence predicted from the nucleotide sequence of the nfxB gene (16). The amino acid sequence of the 23-kDa protein (fraction VIb) was MRTIRK, which corresponded to the sequence translated from nucleotide (nt) 235 of nfxB (see Fig. 7). These results indicated that NfxB is translated from two initiation codons corresponding to nt 235 and 271. Putative Shine-Dalgarno sequences of both coding regions were identified (nt 216 to 219 for the 23-kDa NfxB, and nt 260 to 263 for the 21-kDa NfxB).
We also examined whether these two proteins were pro- on May 2, 2019 by guest http://jb.asm.org/ duced in P. aeruginosa by using anti-NfxB serum which was obtained from rabbits immunized with purified 21-kDa NfxB (fraction VIs). As shown in Fig. 2, the 23-and 21-kDa NfxB proteins were detected in P. aeruginosa PK1013E (lane 4). These two proteins in P. aeruginosa had exactly the same mobility as the purified NfxB fractions (lanes 1 to 3). The ratio of the 23-and 21-kDa proteins was 3.2:1.0 in P. aeruginosa (Fig. 2, lane 4). This value was obtained by measuring the intensity of the bands in Fig. 2 with an ImageQuant. This ratio was similar to that of pNF7 expression in E. coli (data not shown). These results suggested that there are two NfxB proteins in P. aeruginosa that are produced from two translation start sites. However, we cannot rule out the possibility that the 21-kDa species is a degradation product of the 23-kDa protein.
Since both NfxB proteins were found in P. aeruginosa, we used a purified mixture of these proteins (fraction VI) to characterize their functions.

Identification of the mRNA initiation sites of nfxB.
To identify the promoter region of nfxB, we determined the transcription initiation site of this gene by primer extension. The mRNA was reverse transcribed with an oligonucleotide primer which is complementary to the sequence between nt 308 and 327 of nfxB. In this experiment, the initiation site is unclear; however, the cytosine at nt 140 is the most likely candidate ( Fig. 3; also see Fig. 7). Since the initiation site was not accompanied by the Ϫ10 and Ϫ35 consensus sequences of the E. coli promoter, its structure probably differs from that of the general E. coli promoter. Although there were two minor start sites at nt 162 and 167, the intensities of these bands were eight-and sevenfold lower than that of the major band (nt 140), respectively. The major transcript of nfxB results from initiation at nt 140.
NfxB binds to the regulatory regions of nfxB itself. Since NfxB has an amino acid sequence similar to the DNA-binding domain of many well-characterized helix-turn-helix type DNA-FIG. 3. Reverse transcriptase mapping of the nfxB promoter. RNA was prepared from exponentially growing PK1013E. RNA isolation, hybridization, and primer extension proceeded as described by Sambrook et al. (19). The primer had the sequence 5Ј-GACGATAGCGACTGCCAGCG-3Ј and was labeled with [␥-32 P]ATP by polynucleotide kinase. The primer extension product was analyzed by electrophoresis on a 6% polyacrylamide-urea gel (lane R) along with a dideoxy sequencing ladder obtained with the same labeled primer used for primer extension (indicated as 32 P-primer). Another dideoxy sequencing ladder was obtained with the unlabeled primer and [␣-32 P]dCTP (indicated as 32 P-dCTP). Autoradiography was done with a BAS2000 image analyzing system (FUJIX, Tokyo, Japan). binding proteins (4, 16), we examined whether NfxB binds to the regulatory region of nfxB itself and regulates its own transcription. We used gel retardation to study the binding of purified NfxB to the DNA fragment containing the regulatory region of nfxB. The electrophoretic mobility of the DNA fragment that contained the regulatory region of nfxB was significantly retarded by NfxB (Fig. 4A, lane 2). To examine the sequence specificity of binding, a 5-to 100-fold excess of unlabeled DNA fragment containing the regulatory region was mixed with a [ 32 P]DNA fragment as the competitor. The retardation was remarkably inhibited by the unlabeled DNA fragment, and the inhibition was dependent on the amount of the competitor DNA (Fig. 4A, lanes 3 to 6). On the other hand, there was no competition by excess salmon sperm DNA (lanes 8 to 11).
Binding to the DNA fragment containing the nfxB regulatory region was further analyzed by gel retardation by varying the protein concentration (Fig. 4B). As the concentration of NfxB increased, the retardation of the DNA fragment increased (Fig. 4B, lanes 5 to 8). It also appears that there are multiple bands in Fig. 4A, lane 3. This may reflect that NfxB binds to the DNA fragment at more than one site. Since some discrete bands are also observed in Fig. 4B, lanes 4 and 5, some NfxB may dissociate from DNA during the gel electrophoresis at these NfxB concentrations. These results suggest that NfxB specifically binds to the regulatory region of nfxB, possibly at multiple sites.
The region of NfxB binding was determined more precisely by DNase I protection. The 5Ј-and 3Ј-end-labeled nfxB sequences were partially digested with DNase I in the presence or absence of NfxB, and the digest was analyzed by gel electrophoresis. The protection pattern in Fig. 5 shows where protection and enhancement occurred. For a diagram of this pat-FIG. 5. DNase I protection of the nfxB regulatory region by NfxB. An entire nfxB labeled fragment was generated as follows. A BamHI fragment was isolated from pNF3B. The 5Ј end of the fragment was labeled with [␥-32 P]ATP with polynucleotide kinase, and the 3Ј end of the fragment was labeled with [␣-32 P]dCTP with Klenow polymerase. The labeled fragments were digested by XbaI. The 5Ј-end-labeled BamHI-XbaI fragment was used for the footprinting of the antisense strand (A), and the 3Ј-end-labeled fragment was used for the sense strand (B). These two DNA fragments (20 fmol tern, see Fig. 7. In both strands, NfxB protected nfxB from nt 121 to 223, which corresponds to Ϫ20 to ϩ83 of the upstream region of nfxB relative to nt 140 (the putative transcription start site). The protection and enhancement were more obvious when over 0.65 pmol of NfxB (32 nM) was added in each strand (Fig. 6, lanes 3 to 5). Autoregulation of the nfxB gene. To examine whether nfxB expression is regulated by NfxB in vivo, we used two ⌽(nfxBЈ-lacZ ϩ )(Hyb) fusion constructs, pNF334 and pNF272, that conferred the Lac ϩ phenotype when introduced into a E. coli lac deletion strain. Expression of nfxB was measured in terms of ␤-galactosidase activity with E. coli CSH26 (13) as the host strain in the presence or absence of intact nfxB gene.
When the nfxB gene was supplied by the second compatible plasmid, pNF20, about 10 times less ⌽(nfxBЈ-lacZ ϩ )(Hyb) fusion products were expressed than in the absence of nfxB (Fig.  6A). Since the 23-kDa NfxB was detected by immunoblotting in E. coli CSH26 carrying pNF20 (data not shown), NfxB was expressed under the lac promoter in pNF20. In the presence of NfxB, nfxB expression under the nfxB promoter was repressed in E. coli. This phenomenon supports the notion that nfxB expression is negatively autoregulated.
We also examined the autoregulation of nfxB expression in P. aeruginosa by using anti-NfxB serum (Fig. 6B). Both the 23and 21-kDa NfxB proteins were expressed by the nfxB mutant, PK1013E (Fig. 6B, lane 1), which has a point mutation at nt 394 (cytosine to guanine) in the nfxB coding region. This mutation causes an Arg-to-Gly substitution (16). On the other hand, neither the 23-nor 21-kDa NfxB protein was detected in the wild-type strain, PAO2142 (lane 2). The mutant PK1013E produced mutant NfxB, which should not be able to bind to the nfxB regulatory region, because the point mutation in the helix-turnhelix region of NfxB results in a loss of DNA-binding activity.

DISCUSSION
We purified NfxB from E. coli by using an overproduction system constructed with a T7 expression vector. Two species of FIG. 7. Partial sequence of the nfxB gene. The nucleotide sequence of nfxB and the numbering are adapted from reference 16. The numbering is shown to the left of the sequence. The deduced amino acid sequence of nfxB is indicated. Solid lines indicate protected regions, and ϩ signs indicate enhanced regions determined from DNase I protection experiments (Fig. 5). Repeated arrowheads indicate inverted repeat sequences (IR1 and IR2). The 39-bp homologous regions, units 1 and 2, are indicated. The putative mRNA start point identified by primer extension (Fig. 3) is labeled as ϩ1, and transcription direction is indicated by a big arrow. Numbering above the nucleotide sequence is derived from this ϩ1. SD1 and SD2 indicate the putative ribosome-binding sites. M1 and M2 represent initiation codons for the 23and 21-kDa NfxB products, respectively. The homologous regions of the nfxB and protein F genes are presented, and asterisks indicate identical nucleotides. The numbering of the nucleotide sequence of the protein F gene is deduced from the mRNA start site (5).
NfxB proteins were purified with this system. The molecular masses of the major and minor species are estimated by SDS-PAGE as 23 and 21 kDa, respectively. These two species were also identified in P. aeruginosa PK1013E, and the ratio of 23-to 21-kDa protein was 3.2:1.0. Although neither protein was detected in a wild-type strain, PAO2142, by immunoblotting, a trace amount of NfxB was apparently produced; this was sufficient to negatively regulate the expression of nfxB. These proteins might be expressed from different translation initiation sites, because two possible initiation codons with Shine-Dalgarno sequence exist. Further studies are necessary to determine whether both species have the same repressor functions, because only mixtures of these two proteins were studied. However, NfxB appears to negatively regulate its own expression.
In both strands, the region from Ϫ20 to ϩ83 of nfxB was protected from DNase I digestion by purified NfxB. The protein masked over 100 bp of the nfxB regulatory region. It is not likely that only one NfxB protein masks over 100 bp. Furthermore, the protected regions in both strands contain some positions where the digestion was enhanced and some positions that were not protected from the digestion. These positions are thought to be gaps between two adjacent sites of several NfxBbinding sites. Possible multiple sites are consistent with the results obtained from gel retardation assays. As the concentration of the unlabeled DNA fragment was increased, two or more bands were found (Fig. 4A, lane 3). These bands may reflect the number of NfxB-binding sites of the regulatory region.
As shown in Fig. 7, the NfxB-binding region contains two 39-bp repeats (units 1 and 2) that are 59% homologous. Since each unit was almost completely protected from DNase I digestion, each probably contains NfxB-binding site(s). The homology between these units also supports the notion there are two or more NfxB-binding sites. The second unit (nt 164 to ca. 202) contains two inverted repeats (IR1 and IR2 in Fig. 7). These inverted repeats might play a crucial role in NfxB binding.
Interestingly, computer analyses revealed that the 5Ј-flanking region of the protein F gene in P. aeruginosa (5) also has a region highly homologous with the inverted repeat IR1 (16) (Fig. 7). Since protein F is one of the major outer membrane porin proteins (15), NfxB may bind to the regulatory region of the protein F gene and control its expression also. However, Hirai et al. (7) reported that there were no changes in protein F levels in nfxB mutants, which does not support NfxB regulation of the protein F gene. In any case, in vitro analysis (gel retardation or footprinting) will help to clarify this issue.
The gene encoding the 54-kDa outer membrane protein, OprJ (11), is thought to be an outer membrane channel, which forms an integral component of the drug efflux system (14), since OprJ is overproduced in nfxB mutants. After oprJ is cloned and sequenced, it will be interesting to see if it has any of the 39-bp repeats in its regulatory region. If so, NfxB effects on its expression will also be of interest.
Poole et al. (18) cloned an operon, mexA-mexB-oprK, which confers resistance to a broad range of antimicrobial agents and is believed to function in the export of the siderophore pyoverdine in P. aeruginosa. mexA-mexB-oprK and nfxB mutants are susceptible to similar antimicrobial agents. MexA and MexB exhibit homology to previously described bacterial export proteins located in the cytoplasmic membrane, and OprK is thought to be an outer membrane channel like OprJ (11,14,18). This gene organization suggests that the three proteins form a drug efflux complex. From these results, we suspect that nfxB regulates the production of an efflux complex that includes OprJ.