INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Multidrug efflux transporters are membrane proteins found in both prokaryotes and eukaryotes that confer resistance to a wide range of structurally unrelated cytotoxic compounds, typically hydrophobic cations. The mechanisms used by these transporters to bind and export such a broad range of substrates remain unknown, largely due to the difficulties posed by the structural analysis of integral membrane proteins. In the case of bacteria, a fruitful alternative approach has been the study of cytosolic multidrug-binding proteins that regulate the expression of specific multidrug transporters at the local level. Examples of both activators (1, 2) and repressors (7, 16, 17) of transcription have been described and can be typically found encoded adjacent to the gene encoding the membrane pump. Structural analysis of BmrR, a dimeric Bacillus subtilis transcriptional activator that binds a range of ligands similar to its cognate transporter, Bmr (18, 19, 35), revealed a crucial negatively charged residue buried at the base of an internal drug-binding pocket lined with hydrophobic residues (38, 39).

Although such an arrangement represents an ideal solution to the problem of binding structurally diverse, hydrophobic, cationic ligands, it remains to be seen if different regulatory proteins, and the efflux pumps themselves, employ a similar, functionally analogous mode of multidrug binding. However, it is somewhat surprising that the multidrug-binding domains of these bacterial regulatory proteins show no apparent homology even though many of them bind a number of common ligands. Thus, further analysis of distinct multidrug-binding proteins is required in order to dissect whether they employ any common themes in their interactions with ligands or instead possess discrete substrate-binding mechanisms.

For the important human pathogen Staphylococcus aureus, a number of plasmid-encoded multidrug resistance transporters have been described, including the closely related major facilitator superfamily members QacA and QacB and the small multidrug resistance protein Smr (15, 22). Expression of both qacA and qacB is regulated by a divergently encoded transcriptional repressor, QacR, a member of the TetR family of repressors (25). IR1, a large inverted repeat located immediately adjacent to and downstream from the qacA and qacB promoters, has been shown to be the site of QacR binding (7). IR1 is unusually large for an operator sequence bound by a TetR family regulator, comprising 15-bp half-sites separated by a 6-bp spacer region (7). In contrast, the DNA-bound structure of TetR indicated that a dimer of TetR, or a similar protein such as QacR, would be unlikely to span a 6-bp spacer (20). Furthermore, TetR binding is prevented by a 1-bp increase or decrease in the single-base-pair spacing between the two tet operator half-sites (37).

Addition of structurally diverse QacA substrates from a wide range of chemical classes, including the compounds benzalkonium, dequalinium, ethidium, proflavine, and rhodamine 6G, has been demonstrated to result in both derepression of the qacA promoter in vivo and dissociation of QacR from qacA promoter-operator DNA in vitro (7). In order to accommodate this range of ligands, it has been proposed that a QacR binding pocket would have to differ substantially from that of the BmrR multidrug-binding protein (13). However, further biochemical and structural investigations of the DNA- and ligand-binding properties of QacR have been seriously hampered by rapid postpurification formation of nonnative disulfide-bonded monomers and oligomeric aggregates (7).

In this paper, the significance of the two cysteine residues within QacR, C72 and C141, to DNA and ligand binding is examined, with a view to generating a fully functional cysteineless QacR derivative for further in vitro studies. Analysis of a cysteineless QacR derivative resulted in the demonstration of an intriguing oligomerization state for DNA-bound QacR, which was found to be dependent on correctly spaced operator half-sites.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains and plasmids. Escherichia coli strain DH5 (26), used as the host for all the procedures described throughout this work, was cultured at 37°C in Luria-Bertani (LB) medium containing ampicillin (100 µg ml¹) to select for plasmids. The plasmid pSK5203, containing the qacA promoter (P_qacA) fused to a chloramphenicol acetyltransferase (cat) reporter gene, and the related plasmid pSK5212, in which qacR is present in cis to the P_qacA-cat fusion, have been described previously (7), as has pSK5210, a clone of wild-type qacR in the expression vector pTTQ18 (32). Site-directed mutagenesis reactions were done with either single-stranded template DNA and a single primer (10) or the QuickChange kit (Stratagene), employing a pair of complementary oligonucleotides and a double-stranded DNA template.

View larger version (29K): [in this window] [in a new window]   FIG. 1.   The cysteineless QacR derivative encoded by pSK5637 and relevant details of the cloned HindIII fragment in this plasmid. The QacR polypeptide encoded by pSK5637 is represented as a rectangular box showing the positions of the amino-terminal (N) HTH DNA-binding motif, the carboxy-terminal (C) hexahistidine tag (Histag), and the two mutations at cysteine residues 72 and 141 that have been introduced into this QacR derivative. Below this is a linear map of the cloned qacR-encoding fragment in plasmid pSK5637, depicting the locations of the qacR and qacA ribosome-binding sites (RBS), the qacR promoter, PqacR, and the divergent qacA promoter, PqacA, which in this construct is fused to a cat reporter gene. Also shown are the AatII, MluI, BsrGI, and BglII restriction endonuclease sites introduced during the course of this work and the position of the QacR binding site, IR1. The sequence of the PqacA 10 hexamer (overlined) is given, downstream from which the highlighted IR1 bases represent nucleotides protected in DNase I footprinting experiments (7). The sequence of the 28-bp IR1 annealed oligonucleotide duplex used throughout this work corresponds to positions 14 to +14 of the IR1 sequence. The nucleotide numbering is the same as for the originally published qacA-qacR sequence (25).

CAT assays. Chloramphenicol acetyltransferase (CAT) assays to ascertain the level of transcription from P_qacA were performed as described previously (7), except that the level of CAT activity was adjusted for the protein concentration in the cell lysates, determined using the Coomassie protein assay reagent kit (Pierce).

Protein purification and stability. Overexpression and purification of wild-type (encoded by pSK5210) and the C72A/C141S cysteineless QacR derivative (encoded by pSK5676) were carried out with a C-terminal His tag as described previously (7), with the following minor modifications. All the buffers were pH 7.5 and contained 20 mM 2-mercaptoethanol and 20 mM Tris-HCl or, in the case of the sonication buffer, 40 mM Tris-HCl. Exchange of proteins into alternative buffers was achieved by passage through Sephadex G50 (Amersham Pharmacia Biotech) columns previously equilibrated with the new buffer.

Gel filtration. A fast protein liquid chromatography Superose 12 HR 10/30 column (Amersham Pharmacia Biotech) with a mobile phase of 300 mM NaCl, 5% (vol/vol) glycerol, and 20 mM Tris-HCl, pH 7.5, was used in all gel filtration experiments. Blue dextran (Sigma) was used to determine the column void volume, and proteins for use as gel filtration molecular mass standards were purchased from Sigma (carbonic anhydrase and -amylase) and Bio-Rad (myoglobin, ovalbumin, and gamma globulin). The molecular weights of the experimental samples were determined following the protocols supplied by the manufacturers. Gel filtration was also used to separate QacR from high-molecular-mass nonspecific protein aggregates before the resultant fractions were stored in dilute single-use aliquots at 70°C at a concentration of approximately 10 µg ml¹ for use in cross-linking reactions. Subsequent chromatography failed to detect the formation of any further aggregates following storage of these low-concentration samples.

Protein cross-linking. Cross-linking reactions were performed in 1× cross-linking buffer (100 mM KCl, 15 mM Tris-HCl, pH 7.5) in a total volume of 40 µl, containing the specified amounts of C72A/C141S QacR and, where indicated, annealed 28-bp gel-purified IR1 oligonucleotide duplexes (Fig. 1). Reactions were incubated at 22°C for 15 min before the addition of gluteraldehyde to a final concentration of 0.01%, followed by a 1.5-min incubation to cross-link any QacR complexes. Alternatively, formaldehyde was used as the cross-linking agent at a final concentration of 1%, with a 10-min incubation at 22°C. Cross-linking reactions were stopped by the addition of 20 µl of 2× SDS gel loading buffer, and the samples were heated at 95°C for 3 min before separation by SDS-12.5% PAGE and transfer to a Hybond-C membrane. QacR was detected immunologically using a 1:1,500 dilution of a rabbit polyclonal antibody raised against the purified C72A/C141S QacR derivative at the Institute of Medical and Veterinary Science, South Australia. The resultant serum was used in standard Western blotting procedures with Blotto as the blocking agent (26).

DLS. The concentration of apo-QacR used in dynamic light-scattering (DLS) experiments was 10 µM (monomer concentration) in a solution of 300 mM NaCl, 1 M imidazole, 5% glycerol, and 50 mM Tris, pH 7.5. When either the 28-bp IR1 or a 33-bp noncognate site was included, the QacR and DNA concentrations were both 40 µM (monomer concentration), and the measurements were conducted in solutions of 50 mM NaCl and 20 mM Tris, pH 7.5. Each experiment used 100-µl samples, which were microfiltered using a 0.02-µm-diameter filter. DLS was measured using a DynaPro 801 dynamic light scattering instrument (Protein Solutions). Data were adjusted for glycerol content and analyzed using the instrument control software package Dynamics 4.0 provided by the manufacturer. Up to 40 readings were recorded during each DLS experiment. Readings in which the baseline error of the data was less than 1.005 and the sum of squares error was below 5.000 (indicative of a monodisperse solution) were fit with a monomodal analysis; all other data were fit using a bimodal analysis.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Site-directed mutagenesis of QacR cysteine residues. The formation of disulfide bonds during the purification and storage of proteins results from the oxidation of free thiol groups, a process catalyzed by divalent metal cations (36), as has been demonstrated for QacR (7). Relatively high levels of reducing agents and the addition of EDTA only served to partially slow the formation of the disulfide-bonded forms of QacR (7). Therefore, site-directed mutagenesis was used to alter the QacR cysteine codons, producing eight QacR single-amino-acid substitution mutants, with C72 and C141 changed separately to alanine (A), serine (S), threonine (T), or proline (P).

View larger version (44K): [in this window] [in a new window]   FIG. 2.   In vivo stability and DNA-binding abilities of QacR variants. The cysteine residues (C141 and C72) in wild-type (WT) QacR were individually replaced with proline (P), serine (S), alanine (A), or threonine (T), in addition to the production of two cysteineless QacR derivatives, C72A/C141S and C72A/C141A. Equal amounts of whole-cell lysates, as judged by Coomassie staining, were separated by SDS-15% PAGE before being transferred to nitrocellulose, and a Western blot was performed with a C-terminal His tag-specific primary antibody to determine the in vivo stability of the QacR derivatives. The position of migration of QacR (23 kDa) relative to molecular size standards is indicated on the left-hand side. The ability of the QacR mutants to continue to bind IR1 in vivo and hence repress transcription from PqacA was assessed by CAT assays with the divergent fusion of PqacA to a cat reporter gene that is available in the plasmids encoding these proteins (Fig. 1). CAT activity is reported relative to the level of expression from PqacA in the presence of wild-type QacR, which was set at 1.

Cysteineless QacR derivative is fully functional. CAT assays were used to determine the in vivo inducibility of the most promising cysteine substitutions in response to the presence of a monovalent or bivalent inducing compound, ethidium or dequalinium, respectively. Following the addition of these compounds, wild-type levels of induction from P_qacA were observed in the presence of the divergently encoded C72A, C141A, or C141S mutant QacR proteins (data not shown). These results indicated that the cysteine residues in QacR could be individually replaced by these alternative amino acids without adversely affecting the ability of the protein to bind structurally diverse ligands. Therefore, two cysteineless QacR mutants, with C72 changed to alanine and C141 changed to alanine (pSK5638) or serine (pSK5637), were produced. For the C72A/C141S double mutant, both the in vivo protein stability (Fig. 2) and the amount of transcription it permitted from P_qacA, as determined by CAT assays (0.93-fold the activity found for wild-type QacR; Fig. 2), were essentially indistinguishable from that of QacR containing two cysteine residues. However, even though the QacR C72A/C141A double mutant was slightly less stable in vivo, it exhibited enhanced repression of P_qacA (0.85-fold) (Fig. 2).

Purified cysteineless QacR has greatly enhanced in vitro stability. Examination by SDS-PAGE of purified C72A/C141S and wild-type QacR proteins after a 4-h incubation at room temperature indicated that the cysteineless derivative exhibited greatly enhanced in vitro stability compared to the wild-type protein (Fig. 3). The C72A/C141S QacR protein lacked the many oligomeric forms that were observed for wild-type QacR in the absence of reducing agents (Fig. 3).

View larger version (77K): [in this window] [in a new window]   FIG. 3.   In vitro stability of purified wild-type and cysteineless QacR. Following the purification of wild-type (WT) QacR or the cysteineless (CL) C72A/C141S derivative, the indicated amounts of purified protein were incubated for 4 h at 22°C before the samples were heated without () or with (+) 33 mM DTT. After separation by SDS-15% PAGE, the proteins were visualized by silver staining. Samples labeled A were in elution buffer containing 20 mM 2-mercaptoethanol, whereas samples labeled B had first been exchanged into a buffer devoid of reducing agents. Arrows indicate the migration positions of the two monomeric forms of QacR that were detected and also the position of the small amounts of a QacR species equivalent in size to a dimer that was detected in the first two lanes for the cysteineless derivative. The positions of migration of size standards (in kilodaltons) are shown on the right-hand side.

Detection of QacR dimers in solution. Passage of the C72A/C141S QacR protein through a gel filtration column produced two peaks, one consistent with a monomeric form of QacR, while the second, smaller peak eluted close to the void volume of the column, indicating highly aggregated protein. Attempts to concentrate QacR beyond 5 mg ml¹, even in the presence of 300 mM NaCl, resulted in the protein precipitating out of solution. However, by addition of IR1 DNA, the solubility of purified cysteineless QacR could be markedly increased, similar to the prevention of aggregate formation that has been observed for the BmrR multidrug-binding regulatory protein (18).

View larger version (81K): [in this window] [in a new window]   FIG. 4.   Detection of cross-linked QacR dimers. Purified C72A/C141S QacR protein, 80 ng (lanes 1 to 4) or 160 ng (lanes 5 and 6), was preincubated for 15 min with no () DNA (lanes 1, 3, and 5) or with (+) 15 ng of 28-bp IR1 DNA (lanes 2 and 4) or 30 ng of 28-bp IR1 DNA (lane 6). Samples in lanes 3 and 4 were cross-linked with 0.01% gluteraldehyde for 1.5 min and those in lanes 5 and 6 were cross-linked with 1% formaldehyde for 10 min before the reactions were stopped by the addition of one-half volume of 2× SDS-PAGE loading buffer, heated at 95°C for 3 min, and separated by SDS-12.5% PAGE. After transfer to nitrocellulose, proteins were detected with an anti-QacR polyclonal antibody. On the left-hand side, the positions of migration of monomeric and dimeric forms of QacR and molecular size markers (in kilodaltons) are indicated.

Two QacR dimers bind per IR1 operator site. In contrast to the cross-linking results presented above, gel filtration indicated that four QacR molecules bound each IR1 DNA site. In three independent gel filtration experiments, using the C72A/C141S QacR derivative preincubated with molar ratios of between 0.25 and 0.5 of the purified, complementary, annealed 28-mer IR1 oligonucleotides (Fig. 1), an average molecular mass of 105.7 ± 3.1 kDa for a QacR-DNA complex was obtained. This value is in good agreement with the theoretical mass of 109.3 kDa for four QacR molecules bound to the 28-bp IR1 DNA fragment. The results of a representative gel filtration experiment are depicted in Fig. 5.

View larger version (20K): [in this window] [in a new window]   FIG. 5.   Gel filtration experiment demonstrating that four QacR molecules are bound per operator site. The x axis values (Ve/V0) were calculated by dividing the elution volume (Ve) of each standard or experimental sample by the void volume of the column (V0). Standards used were: A, -amylase (Mr 200,000); B, gamma globulin (Mr 158,000); C, ovalbumin (Mr 44,000); D, carbonic anhydrase (Mr 29,000); and E, myoglobin (Mr 17,000). The experimentally determined values for purified C72A/C141S QacR and the complex it forms with the 28-bp IR1 DNA fragment are indicated.

Binding of QacR to DNA requires correctly spaced IR1 half-sites. A mutagenic approach was used to investigate the significance of the IR1 spacer region to the ability of QacR dimers to bind operator DNA. Due to the availability of the previously constructed plasmid pSK5688, in which P_qacA had been repositioned so that the 6 bp normally separating the two IR1 half-sites had been replaced with the 10 hexamer of this promoter (Fig. 6), analysis of this alternative P_qacA/IR1 arrangement provided an initial indication of the importance of the 6-bp spacer.

View larger version (37K): [in this window] [in a new window]   FIG. 6.   Relevant portion of the wild-type PqacA/IR1 sequence carried by pSK5249 and pSK5203 (A) and the alternative PqacA/IR1 arrangement in pSK5688, in which the 10 and 35 hexamers (underlined) of the qacA promoter have been repositioned so that the 10 region is located between the two half-sites of IR1 (B). (C) The sequence of the central region separating the two IR1 half-sites is given for plasmids that contain either the wild-type 6 bp (pSK5203 and pSK5249), substitutions to all 6 bp (pSK5857), a 2-bp increase (pSK5856) or decrease (pSK5834) in the spacing between the two half-sites, transversions in the 1 and +1 positions (pSK5858), or transversions in the 3 and 2 positions (pSK5859). A line delineates the center of the twofold symmetry between the IR1 half-sites. The bases are numbered as for Fig. 1, with non-wild-type IR1 nucleotides indicated in bold. The absolute CAT activities for fusions of these various PqacA/IR1 sequences to a cat reporter gene (expressed as micromoles of chloramphenicol acetylated per minute per milligram of protein) were determined for plasmids with qacR in cis, with the exception of pSK5203, which lacked the qacR gene.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

A number of proteins, including human fibroblast growth factor 1 (5), T4 lysozyme (23), and Bacillus -amylases (33), have, like QacR, been observed to form disulfide-bonded multimeric degradation products following purification. Of particular concern, in the case of QacR, was the large amount of the monomeric species containing an intramolecular disulfide bond (Fig. 3), which appeared to form either during or immediately upon purification. Circumvention of this problem required the removal of the two cysteine residues, which resulted in the production of an engineered, fully functional, cysteineless QacR (C72A/C141S) derivative with greatly enhanced in vitro stability.

The cysteine residues of some regulatory proteins have been found to be essential, e.g., forming part of the DNA-binding domain in zinc finger proteins (21) or playing a fundamental role in the binding of cationic metal ligands by other regulatory proteins, such as MerR (34) and ArsR (28). Retention of full DNA-binding capabilities by the cysteineless derivative was validated both in vivo (Fig. 2) and in vitro, demonstrating that the C72 and C141 residues were not required for the binding of QacR to IR1 DNA. Likewise, the ability of cysteineless QacR to continue to mediate wild-type levels of induction from P_qacA in vivo (Table 1) indicated that the two cysteine residues do not play an important role in drug binding or the communication of the ligand-binding state to the DNA-reading head.

The gel filtration data (Fig. 5), taken together with the cross-linking (Fig. 4) and DLS results, show that two separate QacR dimers bind to IR1 DNA either independently or, alternatively, with only limited dimer-dimer interactions. A revised model for the regulation of qacA expression by QacR is therefore proposed, in which induction results in the dissociation of the two dimers bound to an IR1 site, to form two separate ligand-bound dimers. DLS experiments provided proof of this event, demonstrating that the four molecules of QacR bound to operator DNA again assumed dimeric forms in the presence of inducing compounds. This result also strongly supported the postulate that QacR does not self-assemble into a tetrameric form, as did the failure of the cross-linking experiments to produce any covalently linked complexes larger in size than a dimeric form of QacR. As both formaldehyde and gluteraldehyde are generally considered to be zero-length cross-linking agents, the two QacR dimers bound to each IR1 site are therefore likely to have no or only very limited direct dimer-dimer interactions, such that no side-chains are available for cross-linking by these reagents. Formaldehyde and gluteraldehyde also did not produce any covalently linked QacR-DNA complexes that contained four protein molecules, which is consistent with the failure to obtain cross-linked protein-operator complexes for other sequence-specific DNA-binding regulatory proteins (29). The apparent anomaly of apo-QacR eluting from the Superose gel filtration column as a monomer is likely to reflect the ability of this matrix on rare occasions to temporarily monomerize native dimers, a situation previously observed for the Toxoplasma gondii uracil phosphoribosyltransferase (M. A. Schumacher and R. G. Brennan, unpublished data).

Despite all the evidence obtained in this study indicating that QacR does not self-assemble into a tetrameric form prior to DNA binding, no intermediate forms equivalent to one QacR dimer bound per IR1 site were detected in gel mobility shift assays, even at protein concentrations that did not bind all the available operator DNA (7). This result is not surprising, considering that both in vivo (7) and in vitro QacR failed to bind to DNA that contained only a single IR1 half-site. Some form of distance-dependent cooperativity in the binding of a pair of dimers to IR1 was clearly supported by the almost complete inability of QacR to bind operator sequences in which the spacing of the IR1 half-sites had been increased or decreased by 2 bp (Fig. 6C), particularly in light of the continued ability of QacR to bind operator sequences which contained substitutions to all of the central 6 bp.

A mechanism for cooperative binding of QacR dimers to IR1 that involved indirect interactions through the DNA would be consistent with all the results presented in this work. Such a scenario is not unwarranted, as monomers of the human regulatory protein RFX1 have been shown to bind DNA cooperatively, forming dimers that have no direct protein-protein interactions; instead, the cooperativity appears to occur via protein-induced deformation of the binding site (4, 6). Thus, the actual sequence of the central 6 bp of IR1 may make a contribution to the cooperative binding of QacR dimers via influencing the local structure of the operator DNA, in addition to their primary role of correctly spacing the half-sites. Alternatively, data from the DNase I footprint (Fig. 1), together with that from the mutations in the central 6 bp (Fig. 6C), could be taken to indicate that QacR makes some direct DNA contacts to the central spacer that are for the most part sequence nonspecific.

Of interest was the finding that a simple 2-bp mutation introduced at positions 3 and 2 of IR1 produced an operator that exhibited significantly improved repression (pSK5859; Fig. 6C). This observation provides some corroboration for an earlier proposal that the natural QacR/IR1 system is designed to provide a significant basal level of qacA expression in order to protect the cell against compounds that are substrates of the QacA multidrug pump but not ligands of the QacR transcriptional repressor (7).

The demonstration that two QacR dimers bind to IR1 was unexpected, considering that the other TetR family members for which the DNA-binding stoichiometry is known, bind as one dimer per operator, e.g., TetR (9) and CamR (3). However, in contrast to the above proteins, QacR binds to an exceptionally large, 36-bp inverted repeat (Fig. 1). The DNase I footprint (7), in combination with the spacer region mutations (Fig. 6), suggests that QacR interacts directly with a region of DNA consisting of 20 to 28 bp, or at least 10 bp per half-site, which is in stark contrast to the 6-bp DNA-binding capacity of a typical helix-turn-helix (HTH) motif (8, 31). Therefore, in order to bind a single IR1 half-site, the DNA-reading heads of both molecules in one QacR dimer appear to be required, which would account for the observed QacR DNA-binding stoichiometry (Fig. 5). Although the DNA "recognition" helix of TetR compensates for its abnormally short length by making an exceptionally large number of contacts with tet operator DNA (20), QacR may have acquired an alternative solution to ensure that an adequate number of DNA contacts are made by employing a dimer to bind to each operator half-site.

Although MetJ family regulatory proteins also employ one dimer to bind each of their operator half-sites, unlike QacR, these proteins have been shown to form tetramers that involve substantial dimer-dimer interactions, in addition to binding DNA with an antiparallel -sheet motif and not an HTH (30). The lack of any apparent internal symmetry within an IR1 half-site (Fig. 1) suggests that the contacts a QacR dimer makes to DNA may be largely nonsymmetrical, as has been found for a MetJ family member, the bacteriophage P22 repressor Arc (24). Further prokaryotic regulatory proteins that bind DNA as higher-order oligomers have been shown to self-assemble from dimers into tetramers, as in the case of LacR (14), and from dimers into octamers, in the case of the lambda repressor (27). In both of the above examples, the dimeric units making up the higher-order oligomers are used to bind distinct operators.

Analysis of the region of DNA encoding the qacA and qacR genes failed to detect any additional sequences exhibiting similarity to IR1, and it has been established that QacR does not interact with the dissimilar IR2 inverted repeat located in the vicinity of the qacR promoter (7). This further supports a model in which separate QacR dimers are used to bind the two IR1 operator half-sites by a cooperative process that does not require the prior self-assembly of QacR into tetramers.