QacR Is a Repressor Protein That Regulates Expression of theStaphylococcus aureus Multidrug Efflux Pump QacA*

The Staphylococcus aureus QacA protein is a multidrug transporter that confers resistance to a broad range of antimicrobial agents via proton motive force-dependent efflux of the compounds. Primer extension analysis was performed to map the transcription start points of theqacA and divergently transcribed qacRmRNAs. Each gene utilized a single promoter element, the locations of which were confirmed by site-directed mutagenesis. Fusions of theqacA and qacR promoters to a chloramphenicol acetyl transferase reporter gene were used to demonstrate that QacR is a trans-acting repressor of qacA transcription that does not autoregulate its own expression. An inverted repeat overlapping the qacA transcription start site was shown to be the operator sequence for control of qacA gene expression. Removal of one half of the operator prevented QacR-mediated repression of the qacA promoter. Purified QacR protein bound specifically to this operator sequence in DNase I-footprinting experiments. Importantly, addition of diverse QacA substrates was shown to induce qacA expression in vivo, as well as inhibit binding of QacR to operator DNA in vitro, by using gel-mobility shift assays. QacR therefore appears to interact directly with structurally dissimilar inducing compounds that are substrates of the QacA multidrug efflux pump.

Closely following the discovery of mammalian P-glycoprotein (1,2), the phenomenon of multidrug resistance was also described for a bacterial system, the Staphylococcus aureus QacA pump (3), and has since been found to be widespread among both Gram-negative and Gram-positive bacteria (4 -6). Resistance involves the active transport of a structurally diverse range of toxic compounds, typically hydrophobic cations, from the cell by a single efflux system. In the case of P-glycoprotein, the ability to export many anticancer agents (7) has prompted investigations into its mode of action. Biochemical studies and the generation of mutants with altered drug binding capabilities have suggested P-glycoprotein interacts directly with various substrates (7,8). Additionally, recent progress has been made toward determining the structure of P-glycoprotein (9). Despite these advances, the basis of multidrug recognition by P-glycoprotein is still not understood. The functional similarities of bacterial multidrug efflux systems with P-glycoprotein, together with their presence in significant human pathogens, such as S. aureus (3), Pseudomonas aeruginosa (10), Neisseria gonorrhoeae (11), and Mycobacterium tuberculosis (12), makes elucidation of their molecular mechanisms an important research goal. Some progress has been made toward delineating the significance that various motifs and individual amino acids hold for determining the specificity of transport and overall mechanism of action for the QacA (13) and Bacillus subtilis Bmr (14) multidrug transporter proteins (reviewed in Ref. 6), but the strong association of efflux pumps with the membrane makes isolation and in depth analysis of these proteins difficult.
Considerable effort has also been directed toward identifying the factors involved in regulation of multidrug transporter gene expression. For mammalian P-glycoprotein (15,16), as well as the Bmr (17), and Escherichia coli EmrB (18) bacterial multidrug efflux systems, increased gene expression followed the addition of some of the structurally diverse compounds exported by these pumps. Indeed, a certain degree of regulatory control over the genes for membrane transport proteins is to be expected, given their toxic nature toward the cell if overexpressed, as has been observed for the Gram-negative bacterial tetracycline resistance gene, tetA (19). Control of tetA expression by the specific repressor protein, TetR, is the best understood example of the regulation of a gene encoding a drug transporter (19). Induction of tetA expression occurs when TetR binds a tetracycline/divalent metal cation complex, inducing a conformational change in the protein such that TetR no longer binds the tetA operator, thereby liberating the promoter site (20). A similar style of regulation may also be responsible for the induction of expression observed for some bacterial multidrug efflux genes. However, for this to be effective, both the transporter and the regulatory protein would need to recognize the structurally diverse compounds that these systems efflux from the cell. Despite this, regulation of some bacterial multidrug efflux genes involves specific trans-acting regulatory proteins, such as the B. subtilis BmrR (17), E. coli EmrR (18), and N. gonorrhoeae MtrR (21) proteins. Importantly, the B. subtilis BmrR transcriptional activator protein binds directly to at least some of the compounds that both induce bmr expression and are also substrates for the Bmr transporter (17,22,23).
Analysis of the region immediately upstream from the S. aureus qacA gene revealed a putative regulatory element, qacR, previously called orf188 (24). Based on homology comparisons, QacR, together with TetR, belong to a family of regulatory proteins which all share common features associated with a multi-helical DNA-binding domain at their N-terminal ends and have highly divergent C termini postulated to be involved in the binding of inducing compounds (24,25). In this paper, we make the first steps toward understanding how QacR regulates the expression of qacA, by demonstrating it is a negative regulator that both binds to an operator site adjacent to the qacA promoter and also appears to interact directly with a diverse range of inducing compounds.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-The bacterial strains, plasmids, and primers used in this study are described in Table I. For all experiments performed in S. aureus, strain SK982 was the host; E. coli strain DH5␣ was used for CAT 1 assays, QacR overexpression, and in all cloning manipulations. All strains were cultured at 37°C in LB media containing, where appropriate, 100 g of ampicillin, 50 g of kanamycin, 30 g of chloramphenicol, or 20 g of gentamicin per ml. The plasmid pSK616, containing qacA-qacR DNA originally derived from pSK1 (24), was the starting material for all manipulations, unless otherwise specified. The vector pSK5201, used to generate promoter fusions to the chloramphenicol acetyltransferase cat reporter gene, was constructed by replacing the 1.4-kilobase pair PvuI-NdeI fragment in pKK232-8 (26) with the 1.9-kilobase pair PvuI-NdeI M13 origin-containing fragment from pMAL-p2 (New England Biolabs), to allow rescue of single-stranded DNA from subsequent constructs.
PCR utilizing the primers 671HindIII and 865BamHI (Table I), which contained additional sequences at their 5Ј ends encoding HindIII or BamHI sites, respectively, were performed to generate a fragment encoding the qacR-qacA intergenic region (see Fig. 1). This was then cloned into pSK5201, producing pSK5202, with the predicted promoter for qacR, P qacR , placed upstream of the cat gene ribosome binding site. pSK5203 was constructed with cat gene expression under the control of the predicted promoter for qacA, P qacA . Mutations were created in the putative "Ϫ10" regions of the qacR (pSK5202) and qacA (pSK5203) promoters by site-directed mutagenesis producing pSK5211 and pSK5206, respectively. The primer used to mutate P qacR was 5Ј-GAG-TATTAAGTACTATCAAGATCTGCACAAAAAATTCAAAGAAT-3Ј (corresponding to qacA sense nucleotides, mutations underlined), and that for P qacA was 5Ј-GCGATCGGTCTATAAGGGTACCAATCTATTTT-TTTACATATTACAAC-3Ј (corresponding to qacA antisense nucleotides, mutations underlined). An additional plasmid, pSK5212, with qacR in cis to P qacA controlling cat, was constructed by cloning the product of a PCR reaction with the primers 108HindIIIH6 and 867HindIII into pSK5201. To delete one half of the putative qacA operator sequence (nucleotides 818 -835; Fig. 1), a 750-bp fragment from pSK5212 was obtained using the Sau3AI site separating the two halves of IR1 and another Sau3AI site in the polylinker adjacent to the 3Ј end of qacR in pSK5212. This fragment was ligated into the BamHI site of pSK5201 and a resultant clone, with cat transcription under the control of P qacA , designated pSK5215.
The plasmid pSK4238, which was constructed to enable the expression of qacR in trans, contained a PCR-generated qacR fragment preceded by a strong ribosome binding site inserted behind the lacZ promoter of pK184 (27), a vector compatible with the pSK5201-based constructs. To facilitate the overexpression and purification of QacR, the primers 678BamHI and 108HindIIIH6 were used to obtain a PCR product that consisted of the qacR gene preceded by a strong ribosome binding site and also containing a hexahistidine-encoding sequence (His tag) at the 3Ј end. This fragment was ligated into the E. coli 1 The abbreviations used are: CAT, chloramphenicol acetyltransferase; bp, base pair(s); DTT, dithiothreitol; Eb, ethidium bromide; IPTG, isopropyl-1-thio-␤-D-galactopyranoside; IR, inverted repeat; MIC, minimum inhibitory concentration; NTA, nitrilotriacetic acid; PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction; QAC, quaternary ammonium compound; tsp, transcription start point.
a The number of each primer refers to the 5Ј most base in each instance that corresponds to a nucleotide in the previously published qacR-qacA sequence (24), which is the same numbering as used in Fig. 1. Some primers contain additional nucleotides (underlined) at their 5Ј ends for the incorporation of restriction endonuclease recognition sites, an improved ribosome binding site, or a hexahistidine encoding tag. expression vector pTTQ18 (28) to produce pSK5210, with qacR expression placed under the control of the strong IPTG-inducible tac promoter. All clones produced from the products of PCR reactions were sequenced to confirm that no errors had been incorporated during amplification.
RNA and DNA Isolation and Manipulations-Total cellular RNA was purified from S. aureus strains SK982 and SK982 containing the qacA ϩ , qacR ϩ plasmid pSK1 by the hot phenol method as described by Miller (29), modified for use with S. aureus by an initial incubation on ice in protoplast lysis buffer (15 mM Tris-HCl, 6 mM EDTA, 450 mM sucrose, pH 8.0) containing 0.25 mg/ml lysostaphin (Sigma) for 45 min to generate protoplasts. Double-and single-stranded plasmid DNA preparations, DNA manipulations, transformations, and site-directed mutagenesis were by standard procedures (30). For dideoxy sequencing of double-stranded DNA. the Sequitherm kit (Epicentre Technologies) was utilized. PCR was performed using Pfu enzyme (Stratagene), according to the manufacturer's instructions.
Identification of Promoters: Primer Extension-Primer extension analysis was performed essentially as described by Ausubel et al. (31), utilizing the primers 899EcoRI for qacA and 557 and 672BamHI for qacR ( Fig. 1). Primers were end-labeled with [␥-32 P]ATP (Bresatec), mixed with 50 g of total RNA, denatured by heating at 80°C for 3 min, and then hybridized at 42°C for 90 min before being extended by the addition of dNTPs and Moloney murine leukemia virus reverse transcriptase (Promega). The primer extension products were loaded on a 6% polyacrylamide gel and electrophoresed alongside dideoxy sequencing ladders generated with the same primers.
Determination of Promoter Activity: Chloramphenicol Acetyltransferase Assay-Overnight cultures of E. coli DH5␣ cells harboring a promoter fusion construct alone, or together with pSK4238 providing qacR in trans, were diluted 1:40 in 200 ml of LB media and grown overnight with the appropriate antibiotics and in some cases the addition of a sub-MIC of a potential qacA inducing compound. Cells were collected by centrifugation, resuspended in 10 ml of 1 M Tris-HCl (pH 8.0) and lysed by sonicating (Branson Sonifier B-12) twice for 30 s at 75 watts. The lysate was cleared by centrifugation before the level of CAT activity was determined using acetyl-CoA (Sigma) according to the method of Shaw (32). Triplicate experiments were performed on 3 separate days.
Expression and Purification of QacR from E. coli-An overnight culture of E. coli DH5␣ cells freshly transformed with pSK5210 was diluted 1:50 in 1 liter of prewarmed LB media and grown to an OD 600 of 0.5, at which stage overexpression of QacR was induced by the addition of 0.5 mM IPTG. After another 2-2.5 h, the cells were harvested by centrifugation and resuspended in 20 ml of cold sonication buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 5 mM 2-mercaptoethanol, 20% v/v glycerol, pH 7.8). The cells were frozen overnight at Ϫ70°C and then thawed on ice before addition of lysozyme (Sigma) to 1 mg/ml and incubated on ice for 30 min. Cells were lysed by sonication (75 watts, 30 s, 1 min of cooling, repeated eight times), followed by the addition of RNase A to 10 g/ml and DNase I to 5 g/ml and incubated on ice for 30 min. The lysate was cleared by centrifugation at 15,000 rpm for 20 min before being mixed with 10 ml of 50% Ni 2ϩ -NTA metal chelate affinity resin (Qiagen) for 1.5 h at 4°C. The resin was then packed into a column at 4°C and washed at a flow rate of 15 ml/h with 40 ml of sonication buffer and then overnight with 200 ml of wash buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 5 mM 2-mercaptoethanol, 20% v/v glycerol, 50 mM imidazole, pH 7.3). Purified QacR was eluted in wash buffer containing 300 mM imidazole (pH 8.0). The fractions were analyzed by SDS-PAGE (12.5% polyacrylamide) using the buffer system of Laemmli (33) and proteins visualized by staining with Coomassie Brilliant Blue R. Molecular size standards were purchased from Bio-Rad, and protein concentration was estimated by the method of Bradford (34), using bovine serum albumin (New England Biolabs) as a standard. Gel-filtration chromatography was performed on a FPLC Superose 12 HR 10/30 column (Amersham Pharmacia Biotech) with 100 mM KCl, 20 mM Tris-HCl, 20 mM 2-mercaptoethanol, pH 7.5, as the buffer. Gel-filtration molecular weight markers MW-GF-200 were purchased from Sigma. Low pH native PAGE was performed as described in Ref. 35. Western analysis using a 6ϫHis monoclonal antibody to detect hexahistidine-tagged proteins was performed according to the manufacturer's (CLONTECH) instructions.
Treatment of QacR with Divalent Metal Cations-Purified QacR (6 g) in a final volume of 40 l was incubated in 20 mM Tris-HCl (pH 7.5) containing the indicated amounts of divalent metal cations for 3 h at 22°C. The reaction was stopped by the addition of 20 l of non-reducing SDS-PAGE sample buffer containing 30 mM EDTA. Samples were heated at 100°C for 4 min without reducing agent unless otherwise stated, cooled on ice for 2 min, and 5-l aliquots analyzed by SDS-PAGE followed by silver staining to visualize the proteins.
Gel-mobility Shift Analysis-Labeled probes obtained by PCR amplification using primers end-labeled with [␥-32 P]ATP were purified with the Wizard DNA purification kit (Promega). Initial experiments utilized a 212-bp PCR product corresponding to the whole qacR-qacA intergenic region (bp 672-867; Fig. 1) generated with the primers 672BamHI and 867HindIII. Subsequent experiments in which substrates of QacA were added to some of the binding mixtures used a smaller 137-bp PCR product (bp 738 -867; Fig. 1), which included the qacA promoter and IR1 portions of the intergenic region, produced with the primers 738 and 867HindIII. A 186-bp PCR fragment from the 3Ј end of the qacR gene that acted as a specificity control was generated using the primers 108HindIIIH6 and 262. The QacR fractions obtained from the affinity column were immediately exchanged into GMS buffer (15 mM Tris-HCl, 1 mM EDTA, 100 mM KCl, 7.5% v/v glycerol, pH 7.5) containing 10 mM 2-mercaptoethanol by passage through a Sephadex G-50 column previously equilibrated with this buffer and stored in single-use aliquots at Ϫ70°C until required. The 32 P-labeled DNA fragments, approximately 2000 cpm/reaction, were incubated with the indicated amounts of QacR in GMS buffer containing 0.3 mM DTT, 75 g of bovine serum albumin per ml, and 75 g of poly(dI-dC) (Sigma) per ml in a total volume of 20 l. Some reactions also had substrates of QacA or divalent metal cations added at the indicated concentrations. After 15 min of incubation at 22°C, the reaction mixtures were analyzed by high ionic strength PAGE and autoradiographed as described by Ausubel et al. (31).
DNase I Footprinting-Footprinting of the qacA coding strand utilized the 137-bp DNA fragment used in gel-mobility shift experiments, generated using primers 738 (end-labeled with [␥-32 P]ATP) and 867HindIII. For footprinting of the qacA non-coding strand, a 154-bp fragment (bp 738 -899; Fig. 1) was prepared using the primers 899EcoRI (end-labeled with [␥-32 P]ATP) and 738. Approximately 40,000 cpm/reaction of labeled DNA was incubated in GMS buffer containing 0.3 mM DTT, 75 g/ml bovine serum albumin, and 60 g/ml poly(dI-dC) for 15 min at 22°C in a total volume of 20 l. Selected reactions also contained 600 ng of purified QacR. One unit of DNase I (Promega) in 20 l of DNase I buffer (10 mM Tris-HCl, 10 mM MgCl 2 , 2 mM CaCl 2 , 1 mM DTT, 15 mM NaCl, pH 7.5) was then added, mixed, and the digestion allowed to proceed for exactly 1 min at 22°C before being terminated with the addition of 40 l of stop solution (0.6 M sodium acetate, 25 mM EDTA, pH 4.8). The reactions were extracted once with phenol-chloroform (3:1 v/v) and the DNA ethanol precipitated overnight at Ϫ20°C. The resulting pellet was washed twice with 70% ethanol and resuspended in 6.5 l of loading dye (49% formamide, 0.0125% xylene cyanol, 0.0125% bromphenol blue). After heating at 95°C for 3 min, 2-l aliquots of the DNA were analyzed on a 6% polyacrylamide gel. Sequencing ladders were generated in each case with the same primer as was end-labeled for the PCR reaction.

Localization of the Promoter Elements and Transcription
Start Points for qacA and qacR-Previous inspection of the sequence upstream of qacA had revealed a divergently transcribed open reading frame, qacR, postulated to regulate qacA transcription by binding to a region of dyad symmetry (IR1; Fig. 1) that partially overlapped the most likely sequence for the qacA promoter (P qacA ; Fig. 1) (24). Additionally, a potential promoter for qacR (bp 794 -827 in Fig. 1) was identified, which would overlap IR1 and possibly subject qacR expression to auto-regulation (24). However, on further analysis of the qacR-qacA intergenic sequence, we identified another potential promoter-like element for qacR (P qacR ; bp 702-729 in Fig. 1) that lies closer to the start of the gene and has its own distinct region of dyad symmetry (IR2; Fig. 1), suggesting that expression of qacR could be initiated from multiple promoters.
To ascertain the actual promoter(s) used by each gene, the tsp for both qacA and qacR were determined in the S. aureus strain SK982 (36) harboring the qacA ϩ , qacR ϩ multiresistance plasmid pSK1 (24) using primer extension analysis. For each reaction, only one extension product was observed. The tsp for qacA in S. aureus corresponded to a G at position 810 and that for qacR to be a C at position 693 (Fig. 2). Therefore, for each gene, transcription is initiated from a single promoter element, each partially overlapped by potential operator sites for DNAbinding proteins (IR1 and IR2; Fig. 1). To confirm the same tsp was used when the cloned qac region was present in E. coli, primer extension analyses, using the same primers as described in Fig. 2, were performed on RNA isolated from E. coli strain DH5␣ harboring pSK616. The results obtained indicated that the same promoters were utilized in both species, although for both genes the tsp in E. coli was found to be an A, one nucleotide upstream from the tsp determined for S. aureus (data not shown). This finding may reflect a real difference, since A is the preferred start nucleotide for the initiation of transcription by E. coli 70 RNA polymerase (37). Alternatively, the observed change may merely be an artifact of preparing RNA from quite distinct species.
The expression of some regulatory genes, like tetR (19), from multiple, autoregulated promoters, suggested the originally proposed qacR promoter region (24) might also be functional as an inducible promoter. To test this, primer extension analysis using primer 557 ( Fig. 1) was also performed on RNA isolated from cultures of S. aureus containing pSK1 that had been treated with Eb, a compound that acts as an inducer of QacRmediated qacA gene expression (see below). A further tsp was not observed for qacR, supporting the finding that the qacR gene is transcribed from a single promoter. This was confirmed by failure to detect any primer extension product (data not shown) for another primer, 672BamHI (Fig. 1), which was at a closer, more ideal distance to the area at which transcription would be initiated for the originally proposed qacR promoter.
Various constructs in the promoter cloning and singlestranded DNA rescue vector pSK5201 were utilized to substantiate the location of the promoter elements. The plasmids pSK5202 and pSK5203 were constructed such that P qacR and P qacA , respectively, would be expected to control transcription of the cat reporter gene. Site-directed mutagenesis employing the primers described under "Experimental Procedures" was used to change 4 and 5 nucleotides of the putative P qacA and P qacR "Ϫ10" regions, from TATAAT and TATCAT to TGGTAC (pSK5206) and AGATCT (pSK5211), respectively, producing mutant sequences that were predicted to have severely reduced abilities to function as efficient promoters. Analysis of the above constructs by CAT assays demonstrated that the qacR promoter element was greater than 4-fold stronger than P qacA (Table II). The mutations generated in the nominated Ϫ10 regions reduced expression of the cat reporter gene 15-and 50-fold for the qacR and qacA promoter mutants, respectively, when compared with the wild-type promoter clones (Table II), thereby confirming the location of the promoters.
The qacR Gene Encodes a Repressor of qacA Expression That Acts via IR1-To analyze the potential regulation of qacA by the qacR gene product, CAT assays were carried out on E. coli DH5␣ cells harboring pSK5203 or pSK5212. It was clearly demonstrated that in the case of pSK5212, the presence of the qacR gene in cis to qacA resulted in a significant reduction in transcription levels from P qacA when compared with pSK5203 (6.41 versus 1.37 units; Table II). When pSK5203 and the compatible plasmid pSK4238, which provided qacR in trans, were present in the same cell, QacR completely repressed expression from the qacA promoter, resulting in no detectable cat activity (Table II). As expected, the presence of the pSK4238   1. Regulatory region of the qacA and qacR genes. The respective promoters, P qacA and P qacR , their putative "Ϫ10" and "Ϫ35" regions, and the tsp for each gene are shown. Bold type delineates the most likely ribosome binding sites. Inverted repeats, IR1 and IR2, that partially overlap each promoter are indicated by bold arrows. Thin arrowed lines denote selected oligonucleotide primers (Table I)  vector, pK184, lacking qacR, had no effect on transcription from P qacA (data not shown). Taken together, these data indicate that the observed high level of repression afforded by pSK4238 was a result of qacR being placed under the control of the strong lac promoter, leading to abnormally high levels of QacR. The presence of pSK4238 was found to have no significant effect on the level of transcription from the qacR promoter in pSK5202 (Table II), indicating that the expression of qacR is not autoregulatory.
To investigate the potential of the region of dyad symmetry (IR1) overlapping the qacA promoter to act as an operator site, CAT assays were performed with cells carrying pSK5215, which possessed only one half of the dyad symmetry (bp 800 -817, Fig. 1), the other being deleted as described under "Experimental Procedures." Expression from the qacA promoter in pSK5215 was, relative to pSK5212 (which possessed the full IR1), restored to the level observed for pSK5203, which lacked qacR (Table II). Further experiments defining the role of IR1 as a potential operator to which QacR binds are detailed below.
Purification of QacR-To obtain large quantities of QacR for in vitro studies, the qacR gene was first cloned downstream from a strong ribosome binding site and the IPTG-inducible tac promoter in the E. coli vector pTTQ18, resulting in inducible overexpression of QacR (Fig. 3A, lane 2). Utilization of a His tag incorporated at the C terminus of QacR allowed purification close to homogeneity by Ni 2ϩ -NTA chromatography (Fig. 3A,  lanes 5 and 6). The purified QacR protein migrated in reducing SDS-PAGE gels with an observed molecular mass of approximately 23 kDa, the size predicted for a QacR monomer (24) with a hexahistidine tag. N-terminal amino acid sequencing of the first 12 residues of the purified protein confirmed its identity. The molecular mass of freshly purified QacR was determined by gel-filtration chromatography to be 22.5 kDa, in close agreement with the predicted molecular mass for a QacR-His tag monomer.
Long term storage of the purified QacR-containing fractions at either 4°C or room temperature resulted in the formation of higher molecular weight aggregates, the predominant band corresponding to the expected size for a QacR dimer (Fig. 3A,  lane 7; QacRII). QacR monomer could be partially regenerated from these higher molecular weight forms by treatment with 7 M guanidine-HCl followed by dialysis, but only with the addition of a reducing agent, 20 mM DTT (data not shown). This suggested that formation of disulfide bonds between the two cysteine residues in QacR (Cys 72 and Cys 141 ) (24) was respon-sible for the oligomerization process. A recognized problem during the purification and storage of proteins is the formation of disulfide bonds resulting from the oxidation of free thiol groups catalyzed by divalent metal cations and dissolved oxygen (38). To minimize this problem, the major QacR-containing fractions were exchanged into GMS buffer containing 10 mM 2-mercaptoethanol immediately after their elution from the Ni 2ϩ -NTA column, followed by freezing in single-use aliquots at Ϫ70°C.
When subjected to electrophoresis under native conditions, freshly purified QacR migrated as a monomer following incubation at 37°C for 30 min in 20 mM 2-mercaptoethanol. However, incubation at room temperature in the absence of 2-mercaptoethanol resulted in the appearance of trace amounts of oligomeric forms. Thus the native PAGE and gel filtration data, taken in conjunction with the SDS-PAGE results presented in Fig. 3, indicate that QacR is purified from the cell as a monomer and the only oligomeric forms observed are the disulfidebonded multimers, which appear following prolonged storage. Western blots using a 6ϫHis monoclonal antibody to detect hexahistidine-tagged proteins were performed on crude cell lysates separated by electrophoresis through SDS-PAGE gels. Only monomeric QacR was detected (data not shown), confirming that the disulfide-bonded oligomers which form during storage of the purified protein (Fig. 3A) are unlikely to be physiologically relevant.
Involvement of Metal Cations in the Formation of Disulfidebonded Oligomers-Addition of EDTA to a final concentration of 5-10 mM immediately after elution of QacR from the Ni 2ϩ -NTA column was able to partially suppress the formation of oligomers during storage (data not shown), supporting the involvement of divalent metal cations in catalyzing disulfidebond formation between purified QacR molecules. To test the role of divalent metal cations, QacR monomer was incubated with various concentrations of NiCl 2 and CuCl 2 , which resulted in an increased rate of oligomer formation, reaching a maximum at 10 mM for CuCl 2 and at 1 mM for NiCl 2 (Fig. 3B). In particular, treatment with low concentrations of Ni 2ϩ resulted in the appearance of a significant dimer band (QacRII; Fig. 3B), corresponding to the predominant oligomeric species observed after long term storage, which suggested that leaching of Ni 2ϩ ions from the affinity column during the purification process could be contributing to the oligomer formation. To reverse the formation of the oligomers, heating in the presence of 100 mM DTT was required, confirming their disulfide-bonded nature. A faster migrating species, presumably representing QacR containing an intramolecular disulfide bond, was also reduced to the non disulfide-bonded monomer (Fig. 3B). Interestingly, addition of CuCl 2 to a concentration of 1 mM or more largely abolished the intramolecular disulfide-bonded species, in favor of oligomeric forms (Fig. 3B). Two further metals tested, ZnSO 4 and MgSO 4 , had no effect on the formation of disulfide bonds (data not shown).
QacR Binds to a DNA Fragment Containing IR1 and the qacA Promoter-Gel-mobility shift assays were used to show QacR bound specifically to a DNA fragment corresponding to the qacA-qacR intergenic region (bp 672-867; Fig. 1), and not to a control fragment from the 3Ј end of the qacR gene (Fig. 4). Only excess, unlabeled, intergenic competitor DNA, and not excess control DNA, was able to partially titrate out QacR from gel-mobility shift binding reactions (Fig. 4), confirming the specificity of QacR for the intergenic DNA fragment. Additionally, cleavage of the intergenic fragment at the DraI site (bp 751; Fig. 1), to yield two singly end-labeled fragments, resulted in only the qacA promoter-containing fragment being retarded (Fig. 4). This demonstrated that QacR is specific for the qacA promoter and did not bind to a DNA fragment containing the qacR promoter and IR2, lending further support to the postulate that expression of qacR is not autoregulated. In addition, only the crude protein extract from IPTG-induced E. coli DH5␣ cells harboring the QacR overexpressing plasmid pSK5210, but not the control plasmid pTTQ18, was able to shift labeled intergenic DNA (data not shown), confirming that QacR alone is responsible for the observed retardation of intergenic DNA. The addition of CuCl 2 or NiCl 2 had no observable effect on gel-mobility shift binding reactions (data not shown), indicat-ing that at concentrations which increased the rate of oligomer formation (Fig. 3B) these divalent metal cations do not appear to have any direct effect on the binding of QacR to IR1.
DNase I-footprinting experiments confirmed that purified QacR bound specifically to nucleotides in the inverted repeat (IR1) overlapping the qacA tsp (Fig. 5). As shown in Fig. 5B, QacR protected from DNase I digestion all the nucleotides in the IR1 region, with the exception of a few bases at each end, on both the qacA sense and antisense strands.
Substrates of QacA Act as Inducers of qacA Expression-The QacA protein confers resistance to a wide range of structurally dissimilar compounds from four distinct classes of chemicals; the dyes, biguanidines, diamidines, and QACs. To investigate if these substrates potentiate the expression of qacA, CAT assays after overnight incubation in sub-MICs of potential inducing compounds were carried out with E. coli DH5␣ cells harboring pSK4150, which contained the qacR gene in cis to the qacA promoter. A significant increase in CAT activity was observed for many of the QacA substrates (Table III). Conversely, except for Eb, which showed a small degree of induction independent of QacR, no increase in CAT activity was observed when the qacA promoter alone was tested against a range of inducing compounds (data not shown). Also tested for their ability to increase qacA gene expression were another diamidine, propamidine isethionate, and another QAC, cetyltrimethylammonium bromide, both of which showed no induction (data not shown). None of the compounds tested were able to overcome the strong repression of the qacA promoter in pSK5203 afforded by qacR being overexpressed in trans from the plasmid pSK4238 (data not shown).
Structurally Diverse Substrates of QacA Inhibit Binding of QacR to a qacA Operator-containing DNA Fragment-As for the in vivo induction of expression, many QacA substrates were also able to show an in vitro effect on binding of QacR to the qacA operator site. Fig. 6 demonstrates a strong dissociation of QacR from the operator-containing DNA fragment when any of the dyes Eb, rhodamine 6G, proflavin, or crystal violet were included in gel-mobility shift experiments. To dissociate QacR from the operator containing DNA, a concentration 50 times Autoradiograph of purified QacR added in the indicated amounts to; the 212-bp labeled DNA fragment corresponding to the entire qacR-qacA intergenic region (a), the 186-bp labeled specificity control DNA fragment from the 3Ј end of the qacR gene (b), or the labeled 212-bp intergenic fragment cleaved with DraI ( Fig. 1) to yield a 121-bp P qacAcontaining fragment (c) and a 91-bp P qacR -containing fragment (d). A 50-fold excess of cold, unlabeled, 212-bp intergenic (ϩI) or 186-bp control (ϩC) competitor DNA was also added to some assays, prior to incubation for 15 min and non-denaturing PAGE as described under "Experimental Procedures." The position of protein-DNA complexes e and f, that resulted from QacR binding to fragments a and c, respectively, are also indicated.
the MIC for S. aureus of the biguanidine compound chlorhexidine digluconate was required, whereas none of the diamidines, pentamidine isethionate (Fig. 6), propamidine isethionate, or diamidinodiphenylamine-hydrochloric acid, as well as a QAC, cetyltrimethylammonium bromide (data not shown) had any effect. The QACs dequalinium chloride and cetylpyridinium chloride both strongly inhibited binding of QacR to the operator, at a sub-MIC, and 10-fold greater than the MIC for S. aureus, respectively (Fig. 6). Benzalkonium chloride was shown to partially inhibit binding of QacR to operator DNA at a concentration that was 4 times the MIC of this compound for S. aureus (Fig. 6). The addition of QacA substrates at the concentrations used in the above experiments to DNA alone did not affect its mobility. DISCUSSION This paper provides the first experimental evidence to demonstrate that the expression of the S. aureus multidrug efflux gene qacA is regulated by a repressor protein, QacR, the product of a divergently transcribed open reading frame. By both in vitro gel-mobility shift (Fig. 4) and DNase I protection experiments (Fig. 5), and in vivo analysis by deletion of one half of the dyad symmetry which resulted in constitutive expression of P qacA -cat fusions (Table II), the operator site for QacR binding was shown to have been correctly identified; located immediately downstream from the qacA promoter (IR1; Fig. 1). By analogy with studies on promoter/operator systems such as that of the E. coli lac operon (39,40), the binding of QacR to its operator may not inhibit the binding of RNA polymerase, but would prevent the transition of the RNA polymerase-promoter complex into a productively transcribing state, insuring adequate repression of qacA transcription.
The inability of QacR to autoregulate expression of its own gene was demonstrated by qacR overexpressed in trans being able to completely abolish transcription from the qacA promoter, yet having no effect on its own promoter fused to a reporter gene (Table II). This finding was confirmed by showing that QacR did not bind to a DNA fragment containing the qacR promoter (Fig. 4). Most other members of the family of regulatory proteins sharing homology with QacR that are divergently transcribed appear to regulate the expression of their own genes, such as actII-ORF1 from a Streptomyces antibiotic exporting complex (41), CamR, a repressor of D-camphor degra- A, end-labeled qacA promoter/operator DNA fragments were incubated without (Ϫ) and with (ϩ) 600 ng of purified QacR before being subjected to DNase I digestion and electrophoresis as described under "Experimental Procedures." Footprinting of the qacA sense strand used a PCR fragment generated with labeled primer 738 and unlabeled primer 867HindIII. Footprinting of the qacA antisense strand utilized a PCR fragment generated with labeled primer 899EcoRI and unlabeled primer 738. The regions that QacR protects from DNase I digestion are bracketed in each case. Sequencing reactions were performed with the same primer as labeled for the relevant PCR reaction. B, sequence of the qacA Ϫ10 region and the downstream inverted repeat, IR1, with the region QacR protects from DNase I digestion highlighted (white), the qacA tsp circled, the qacA Ϫ10 region boxed, and the location of IR1 shown as bold arrows. Nucleotides are numbered as described in Fig. 1.  dation in Pseudomonas putida (42), TcmR, the repressor of the Streptomyces glaucescens tetracenomycin C resistance gene tcmA (43), and TetR (19). QacR appears to be unusual for this family of proteins in that expression of its own gene was not subject to autoregulation. Although we demonstrated IR2 was not bound by QacR, an alternative role for this potential operator sequence in the control of qacR, and hence qacA gene expression, awaits further detailed investigation.
The repression of qacA transcription by QacR was able to be overcome by the addition of a range of structurally dissimilar QacA substrates, resulting in induction of qacA expression (Table  III). Gel-mobility shift assays (Fig. 6) suggested that for many QacA substrates, induction of qacA expression involved QacR interacting directly with the substrates. Direct recognition of structurally dissimilar compounds, rather than the involvement of a secondary messenger signaling cellular damage, has also been shown for BmrR (17,22,23). Some in vivo inducers of qacA expression inhibited in vitro binding of QacR to qacA operator DNA only at concentrations greatly exceeding their MIC for S. aureus. This may reflect the fact that all QacA substrates appear to be hydrophobic cations; such hydrophobicity may result in these compounds having significantly elevated intracellular levels compared with the concentration at which they were added to the surrounding medium (17). Unexpectedly, two QacA substrates which showed no in vivo induction of qacA gene expression, chlorhexidine digluconate and cetylpyridinium chloride, were able to inhibit binding of QacR to operator DNA, but only at concentrations vastly in excess of their MIC for S. aureus (Fig. 6). For diamidinodiphenylamine-hydrochloric acid, pentamidine isethionate, and propamidine isethionate (all diamidines), and cetyltrimethylammonium bromide (a QAC), no notable effects on the expression of qacA, or the binding of QacR to operator DNA were observed. These results suggest that the level of transcription occurring from the qacA promoter when qacR is present in cis (Table II) enables QacA-mediated efflux of substrates such as the diamidines, compounds not recognized by the repressor protein QacR.
On the basis of the gel-filtration and native PAGE results, it would appear that QacR is purified as a monomer. The failure to detect any disulfide-bonded oligomers in Western blots of crude cell lysates is a strong indication that the multimers which form during long term storage of purified QacR do not occur naturally in the cell and are therefore unlikely to have any physiological significance. The ability of Cu 2ϩ and Ni 2ϩ to induce oligomerization (Fig. 3B), together with the requirement for a reducing agent to reverse the process, supports the involvement of divalent metal cations in the formation of intraand intermolecular disulfide bonds subsequent to the isolation of QacR. Formation of disulfide-bonded oligomers during the storage of purified proteins has been shown to be responsible for the inactivation of human fibroblast growth factor-1 (44), T4 lysozyme (45), and Bacillus ␣-amylases (46). The observed heterogeneity in the banding pattern of both QacR monomers and oligomers is most likely the result of different combinations of intermolecular and intramolecular disulfide bonds, which also occurs for human fibroblast growth factor (44). Furthermore, rapid formation of disulfide-bonded dimers following the purification of the bacterial mercury resistance regulator, MerR, has also been a problem (47). Interestingly, for MerR (48) and also ArsR, the repressor of the E. coli arsenical and antimonial resistance ars operon (49), cysteine residues are required for the binding of the metal cations that act as inducers of these systems, resulting in conformational changes in the respective regulatory proteins, thereby inducing expression. The cysteine residues in QacR could likewise play a role in the apparent ability of this protein to bind the divalent and monovalent cationic compounds that act as inducers of qacA gene expression.
The ability of QacR to interact with multiple and chemically diverse compounds makes it an attractive candidate for future studies. Because of their location in the soluble cytoplasmic fraction of the cell, multidrug efflux regulatory proteins are much easier targets than the corresponding membrane bound transporter for initial studies directed at understanding how structurally diverse compounds are recognized by a single protein. Further work involving the demonstration of a direct interaction between QacR and inducing compounds, mutational analysis of individual residues, and refinement of the purification process toward x-ray crystallography studies on the structure of QacR bound to its operator DNA or inducing compounds is in progress.