The CTXϕ Repressor RstR Binds DNA Cooperatively to Form Tetrameric Repressor-Operator Complexes*

CTXϕ is a filamentous bacteriophage that encodes cholera toxin and integrates into the Vibrio cholerae genome to form stable lysogens. In CTXϕ lysogens, gene expression originating from the rstA phage promoter is repressed by the phage-encoded repressor RstR. The N-terminal region of RstR contains a helix-turn-helix DNA-binding element similar to the helix-turn-helix of the cI/Cro family of phage repressors, whereas the short C-terminal region is unrelated to the oligomerization domain of cI repressor. Purified His-tagged RstR bound to three extended 50-bp operator sites in the rstA promoter region. Each of the RstR footprints exhibited a characteristic staggered pattern of DNase I-accessible regions that suggested RstR binds DNA as a dimer-of-dimers. In gel permeation chromatography and cross-linking experiments, RstR oligomerized to form dimers and tetramers. RstR was shown to be tetrameric when bound to operator DNA by performing mobility shift experiments with mixtures of RstR and a lengthened active variant of RstR. Binding of RstR to the high affinity O1 site could be fit to a cooperative model of operator binding in which two RstR dimers associate to form tetrameric RstR-operator complexes. The binding of RstR dimers to the left or right halves of O1 operator DNA was not observed in mobility shift assays. These observations support a model in which protein-protein contacts between neighboring RstR dimers contribute to strong operator binding.

The molecular mechanisms that regulate bacteriophage lysogeny have been most extensively studied in phage lambda and its close relatives that infect Escherichia coli and Salmonella enterica serovar Typhimurium (1,2). For these bacteriophages, the bistable switch from lysogenic to lytic development involves the interplay of two antagonistic transcriptional repressors, CI and Cro in the case of , that bind to the same set of regulatory sites in the bacteriophage control regions. A similar regulatory network controls lysogeny in Ø80 (3), HK022 (4), and the unrelated bacteriophage P2 (5). However, the regulation of lysogeny in a wide variety of other bacteriophages has not been investigated.
CTX encodes the genes for cholera toxin, the virulence factor primarily responsible for the watery diarrhea characteristic of the disease cholera. Nontoxigenic strains of Vibrio cholerae can be readily transduced to toxin-producing strains by lysogenization with CTX, a process known as lysogenic con-version (6). Like other filamentous bacteriophage, CTX does not have a truly lytic growth phase. However, CTX lysogens exhibit several characteristics found in other temperate bacteriophages: 1) CTX lysogens contain the CTX prophage stably integrated into the V. cholerae genome and 2) the CTX prophage expresses a transcriptional repressor, RstR, that represses the expression of CTX replication genes and provides immunity to secondary infection by CTX (7,8).
CTX shares genetic and morphological similarities with the E. coli filamentous bacteriophage fd. Similarities in gene sequence and gene order indicate that many elements of the pathway for phage assembly and secretion are conserved (6,9). A novel region of the CTX genome known as the "RS region" encodes rstA and rstB, two genes whose products are required for the replication and integration of the CTX chromosome, respectively (7). The RS region also encodes the transcriptional repressor RstR, which is divergently transcribed from rstA. rstA is expressed from a strong promoter in intergenic region 2 (ig-2), the 138-bp region separating rstA from rstR (8) (see Fig.  1A). Expression of rstA is strongly repressed in CTX lysogens, and RstR is the only V. cholerae factor required for rstA repression in E. coli (7). RstR encodes a 13-kDa protein with an N-terminal helix-turn-helix (HTH) 1 element similar to the HTH present in the Cro/cI superfamily of repressors (7). This sequence similarity extends to regions surrounding the bi-helical HTH motif, suggesting that the entire N-terminal region of RstR folds into an ␣-helical domain similar to that of CI repressor (10). The short C-terminal region of RstR is unrelated to the oligomerization domain of CI repressor. Three distinct CTX variants have been described. CTX ET (derived from O1 El Tor V. cholerae), CTX CL (derived from classical V. cholerae), and CTX Calc (derived from O139 Calcutta V. cholerae) are largely identical, but each encodes a unique RstR repressor and adjacent ig-2 region (8,11). Each RstR allele specifically represses expression of its neighboring rstA gene but is unable to repress rstA expression from a heterologous CTX. These observations led to the proposal that each RstR repressor binds exclusively and specifically to cognate operators in its neighboring ig-2 sequences (8).
Here we show that purified RstR ET binds to three operator sites in the rstA promoter region. Unlike many bacterial HTH repressors, RstR binds to extended 50-bp operators, forming tetrameric repressor-operator complexes. The pathway for formation of these complexes appears to involve the cooperative interaction of two RstR dimers bound to neighboring sites in operator DNA. Our model for RstR binding bears similarities, as well as interesting differences, to the "pairwise" cooperative binding of lambda CI dimers to adjacent operator sites in the lambda control region.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-E. coli XL1-Blue (recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lacZ⌬M15/FЈ::Tn10 lacI Q ) (Stratagene, La Jolla CA) was used for routine cloning and for the overexpression of RstR6H. The ig-2 fragments 312, 309, and 301 were generated by the polymerase chain reaction with the appropriate primers and pCTX-Kn as template DNA. The resulting PCR products were cloned into plasmid PCRII (Invitrogen), and the DNA sequence of the resulting plasmids was determined.
RstR6H Expression Vector-The rstR coding sequence was amplified by PCR from pCTX-Kn using the primers RSTR1 (5Ј-CCCCATGGCG-AAGATAAAAGAA) and RSTR4 (5Ј-GCGGATCCAGCACCATGATTT) and ligated to PCR2.1 (Invitrogen). The resulting plasmid, pHK297, was digested with NcoI and BamHI, and the rstR-containing fragment was isolated and ligated to the expression vector pQE60 (Qiagen) previously digested with NcoI and BamHI. The resulting plasmid, pH-K300, encodes RstR6H with an additional alanine residue after the initiator methionine (because of the introduction of the NcoI restriction site) and the sequence GSRSHHHHHH-COOH after the terminal alanine residue of RstR. RstR-CBD-6H was constructed by cloning a PCR product encoding the 52-amino acid chitin-binding domain of Bacillus circulans from plasmid pTXB1 (New England Biolabs, Beverly, MA) into the unique BamHI site in pHK300.
RstR6H Purification-A 10-liter fermentor culture of E. coli strain XL1-blue (Stratagene) containing pHK300 was grown with aeration at 37°C in Luria broth supplemented with 50 g/ml carbenicillin. RstR6H expression was induced at an A 600 of 0.8 by the addition of 1 mM isopropyl-␤-D-thiogalactopyranoside. After 16 h, the cells were collected by centrifugation, resuspended in 100 ml of chilled buffer A (0.05 M NaH 2 PO 4 , 0.5 M NaCl, 0.05% Tween 20, pH 7.0), and lysed by two passes in a French pressure cell at 24,000 psi. Cell debris was sedimented at 100,000 ϫ g for 1 h (S100 fraction), and RstR6H was purified by Ni 2ϩ chelation affinity chromatography as follows. Two ml of a 50:50 slurry of Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) in buffer A was added to 35 ml of S100 extract, and the mixture was gently mixed on ice for 1 h. The slurry was transferred to a small chromatography column and washed with ϳ150 ml of buffer A, followed by 35 ml of buffer A containing 0.1 M imidazole. RstR6H was eluted in 5 ml of buffer A containing 0.3 M imidazole (pH unadjusted). The RstR6H-containing fraction (2.5 ml) was dialyzed overnight at 4°C versus 200 ml of buffer A. RstR6H was further purified by a second round of affinity chromatography using the above procedure. The final eluate (1 ml) was dialyzed overnight versus buffer B (20 mM Tris-HCl, 0.25 M NaCl, 0.05% Tween 20, pH 8.0), and the aliquots were stored frozen at Ϫ70°C. The lengthened RstR-CBD-6H protein was purified similarly. RstR6H concentrations were determined by UV absorption at 280 nm, using a molar extinction coefficient of 8,370 M Ϫ1 cm Ϫ1 (12). After SDS-PAGE, the proteins bands were transferred to a polyvinylidene difluoride membrane and subjected to N-terminal sequencing. CD spectra were obtained at 25°C on a Jasco 810 spectropolarimeter, using a 0.1-cm quartz sample cell. For CD analysis, RstR6H was first dialyzed against 10 mM sodium phosphate buffer, pH 7.5, 0.25 M NaF.
Gel Mobility Shift Assay-The ig-2 DNA fragments were liberated from the PCRII cloning vector by EcoRI digestion, fractionated by agarose gel electrophoresis, and purified using a Qiaquick gel extraction kit (Qiagen). DNA was radiolabeled using T4 DNA polymerase and [␣-32 P]dATP. Synthetic O1 DNA probes were made by radiolabeling one oligonucleotide with T4 polynucleotide kinase and [␥-32 P]ATP and annealing to a complementary unlabeled oligonucleotide. The resulting DNA fragments were purified on nondenaturing polyacrylamide gels and extracted using the "soak and crush" method (13). DNA binding reactions were as follows: 20-l binding reactions contained 20 mM Tris-HCl, pH 7.5, 20 mM KCl, 1 mM MgCl 2 , 5 mM dithiothreitol, 5% glycerol, 50 g/ml bovine serum albumin, 50 g/ml sonicated salmon sperm DNA, and 5,000 cpm probe DNA (1-2 fmol). RstR6H dilutions were prepared in cold reaction buffer immediately prior to use. The binding reactions were incubated at room temperature for 30 min. In experiments with 309 probe, an incubation on ice for 20 min prior to gel electrophoresis was found to significantly improve the stability of RstR-DNA complexes. Where probes were digested with AluI, the digested DNA was added directly to binding reactions without further purification. For mixing experiments with RstR6H and RstR-CBD-6H, the protein mixtures were prepared in binding buffer plus 50 mM dithiothreitol to prevent covalent dimer formation through the single cysteine residue present in RstR6H and RstR-CBD-6H. The RstR mixtures were incubated at 30°C for 30 min prior to the addition to DNA binding reactions, as described above. One microliter of a buffered dye solution (0.1% bromphenol blue) was added to each sample, and the aliquots were loaded onto 6% acrylamide DNA retardation gels (0.5ϫ TBE) and electrophoresed at 10V/cm. The gels were dried directly to blotting paper and exposed to autoradiographic film. For quantitation of bound and free DNA, the dried gels were scanned with a Molecular Dynamics STORM PhosphorImager and quantitated using ImageQuant software. For DNA binding experiments, the fraction of total DNA bound was calculated as counts bound/(counts bound ϩ counts free) without any corrections. Phosphorimaging data from a preliminary DNA binding experiment were collected after 16 and 62 h of exposure. These data yielded essentially identical binding curves, indicating that the response of the phosphorimaging device was linear over a wide range of exposure levels.
DNase I Footprint Assay-DNase I footprint assays were performed as previously described (14). End-labeled DNA probes were generated by PCR with one 5Ј-radiolabeled primer and a second nonradiolabeled primer. DNA probes were purified by nondenaturing PAGE and eluted using the crush and soak method. The reactions were carried out in 30 l containing 50,000 cpm of labeled DNA fragment and purified RstR6H, as described above for gel shift experiments. Binding proceeded for 30 min at room temperature. 0.2 unit of DNase I (Ambion, Austin, TX) was added, and the incubation was continued for 1 min. The GϩA sequencing ladders were generated as previously described (15). Dried DNA pellets were resuspended in loading buffer (95% formamide, 18 mM Na2EDTA, 0.025% SDS, 0.01% xylene cyanol, and bromphenol blue) and heated to 90°C for 2 min prior to loading onto pre-run 8% sequencing gels. The gels were dried to blotting paper and exposed to autoradiographic film.
Gel Permeation Chromatography-Gel permeation was performed on a Superdex 75 HR 10/30 column (total volume, 24 ml), using an Amersham Biosciences fast protein liquid chromatography system. The column was equilibrated in buffer B plus 5 mM dithiothreitol and calibrated with a set of molecular weight markers (Fluka) as shown in Fig.  4. RstR6H samples (0.2 ml) were preincubated at 30°C for 30 min and chromatographed at a flow rate of 0.5 ml/min. 0.5-ml fractions were collected, and RstR6H was detected in an immobilized immunoassay using mouse anti-RstR antisera. To confirm the presence of RstR, the column fractions were analyzed by SDS-PAGE followed by immunoblotting with anti-RstR antisera.
Formaldehyde Cross-linking-Formaldehyde was added to RstR6H (0.1-0.3 mg/ml) in buffer B to a final concentration of 1% and incubated for 10 min at 22°C. Cross-linking was stopped by the addition of an equal volume of 2ϫ SDS sample buffer and heating to 95°C for 5 min. The samples were fractionated by SDS-PAGE under reducing conditions (0.1 M dithiothreitol) on 10% polyacrylamide gels. The gels were stained with a colloidal Coomassie G-250 staining kit (Invitrogen).
Calculations-A reaction describing the equilibrium binding of two RstR dimers to DNA to form tetrameric complexes with no intermediates is as follows, where R 2 is a free RstR dimer, O is free operator DNA, and R 4 O is the bound complex. K A , the apparent equilibrium association constant, is as follows.
For a highly cooperative binding reaction, it is useful to define the following equation, where ⌰ is the fraction of total binding sites occupied in DNA binding experiments (16). Preferring to describe binding in terms of the apparent dissociation constant, K D , we determined the following equation.
Rearranging Equation 4 gives We approximated [R 2 ] Ϸ [R Total ]/2, because the dimer:tetramer equilibrium constant is in the micromolar range, whereas DNA binding occurs at nanomolar RstR concentrations. A monomer-dimer equilibrium was not considered here but could play a role in operator binding. Trial-anderror fitting of theoretical binding curves and linear regression analyses of the Hill plot was carried out using GraphPad Prism 4.0 (GraphPad Software, San Diego CA).

RESULTS
RstR Binds to Multiple Sites in the rstA Promoter Region-RstR containing a C-terminal polyhistidinyl tag (RstR6H) was overexpressed and purified from E. coli by two rounds of Ni 2ϩnitrilotriacetic acid affinity chromatography (see "Experimental Procedures"). This material was about 95% pure, as estimated from SDS-PAGE and Coomassie Blue staining (see Fig.  4B). As predicted from the gene sequence, the final material yielded a single band of ϳ14 kDa after SDS-PAGE. The Nterminal sequence of the 14-kDa band was found to be AKIKER. Residues 2-6 were identical to positions 2-6 of the predicted RstR sequence, whereas the alanine at position 1 is the result of cloning the rstR gene into the expression plasmid vector. The apparent lack of an N-terminal methionine residue indicates that the initiator formyl-methionine is proteolytically removed in E. coli. Far UV circular dichroism spectra of RstR6H showed clear minima at 208 and 222 nM, indicating that RstR contains ␣-helical secondary structure.
By analyzing ␤-galactosidase expression from an rstA::lacZ reporter plasmid, we previously showed that a 290-bp DNA fragment containing the ig-2 region from CTX ET contained both the P rstA promoter and sequences sufficient for rstR-mediated transcription repression of P rstA (8). A similar DNA fragment, extending from Ϫ168 to ϩ88 relative to the rstA transcription start site, was used to investigate DNA binding by purified RstR6H in mobility shift assays. As shown in Fig. 1, RstR6H retarded the migration of the 256-bp 312 probe, giving rise to three more slowly migrating bands. The more quickly migrating shifted band probably represents repressor binding to a high affinity binding site, whereas the second and third shifted bands indicate successive binding of RstR6H to two additional sites. In further mobility shift experiments, RstR6H did not bind to ig-2 DNA probes derived from CTX Cl or CTX Calc (data not shown). The rstR-ig-2 regions of all three CTX genomes are widely divergent, and individual RstR re-pressors exclusively repress expression of their neighboring rstA promoter (8,17). We conclude from these observations that RstR6H binds specifically to regulatory sequences within the ig-2 region of CTX ET .
To map RstR6H binding sites in ig-2, mobility shift assays were performed with two smaller, overlapping probes. RstR6H retarded the mobility of labeled 301 DNA (Ϫ77 to ϩ43) to a single more slowly migrating species (Fig. 1B). Binding occurred over a relatively small range of repressor concentrations (Ͻ10-fold), with half-maximal binding occurring at ϳ3.5 nM RstR6H. RstR6H shifted the 309 probe (Ϫ168 to Ϫ7) to two more slowly migrating bands. Upon titration of RstR6H into the 309 binding reactions, the more quickly migrating shifted form (band 1) diminished in intensity and the more slowly migrating species (band 2) accumulated, suggesting that RstR6H binds sequentially to two sites in 309. Half-maximal binding to 309 occurred at ϳ20 nM RstR6H. To refine the mapping of RstR binding sites, end-labeled 301 and 309 probes were digested with the restriction endonuclease AluI and the resulting subfragments were used in mobility shift assays. AluI cuts 301 once at position Ϫ4, resulting in two labeled subfragments. Neither 301 subfragment was shifted by RstR6H, indicating that the RstR binding site in 301 overlaps or is very close to this AluI site (data not shown). Similarly, AluI digestion of 309 at position Ϫ80 resulted in two subfragments, neither of which was shifted by RstR6H (data not shown). One interpretation of this experiment is that the two potential RstR binding sites in 309 both overlap the AluI site at Ϫ80. Alternatively, only one binding site overlaps the (Ϫ80)AluI site, but RstR binding to a neighboring site requires that RstR first bind to the AluI-containing site. These experiments physically map three RstR binding sites in ig-2; one high affinity binding site overlaps the AluI restriction site at position Ϫ4, whereas two lower affinity binding sites are located near the AluI site at position Ϫ80. The high affinity binding site in 301 was named O1, and the two promoter-distal binding sites were named O2 and O3.
DNase I Protection Analysis of RstR6H Binding-The precise positions of RstR binding sites in ig-2 was determined by DNase I protection assay (18). Because full-length ig-2 DNA was too large for high resolution footprint analysis, the abovementioned 301 and 309 DNA fragments were radiolabeled and used as probes in these experiments. Fig. 2A shows DNase I protection data for each strand of the 301 probe. Fig. 3 summarizes all of the DNase I protection experiments. RstR6H protected an unusually large 50-bp region (ϩ17 to Ϫ32) in 301 corresponding to the high affinity O1 binding site. Complete occupancy of O1 occurred abruptly with increasing RstR6H concentrations, similar to the binding pattern observed in mobility shift assays (Fig. 1). Two footprinted regions, each ϳ50 bp in length, were identified in 309 as corresponding to the promoter-distal sites O2 and O3. These footprints extended from positions Ϫ40 to Ϫ150. The region between O2 and O3 contained an unusual number of DNase I hypersensitive sites on both strands, suggesting that this region of the DNA is bent or distorted upon RstR binding. Interestingly, the footprints show that only the O2 site spans the single AluI restriction site in 309 (Fig. 3). Therefore, our previous mobility shift results indicate that RstR cannot stably bind to the isolated O3 site present on one of the AluI subfragments. Taken together, these data suggest that RstR can only bind to O3 if RstR is first bound to the neighboring O2 site.
The O1, O2, and O3 footprints included sites that remained DNase I-accessible and positions that became hypersensitive to DNase I attack after RstR6H binding. These DNase I-exposed sites clustered in short 2-3-bp regions and were interspersed with strongly protected regions 6 -11 bp in length (Fig. 3). A similar pattern was observed on the opposite DNA strand, although shifted 3 bp in the 3Ј direction. This offset pattern is expected if the minor groove of relatively short regions remain exposed after RstR binding, because DNase I cuts in the minor groove and the nearest backbone ribose-phosphate groups are about 3 bp apart in the 3Ј direction (19). This protection pattern is seen more clearly when the footprint data are mapped onto a planar projection of DNA, as shown in Fig. 2B. Displaying the data in this way revealed another feature of the RstR binding sites: DNase I-exposed regions occurred on one face of the DNA helix in one half of the footprint, as shown by the exposed patches forming a line parallel to the primary DNA helical axis. This pattern shifted ϳ3 bp, representing a 100°turn about the DNA helical axis, to a different face of the DNA helix in the remaining half of the footprint. A similar DNase I protection pattern was observed in the O2 and O3 footprints (data not shown). This pattern is most easily explained if one RstR moiety, probably a dimer, bound to the face of the DNA helix opposite each set of exposed sites. Each half-region is sufficiently large (20 -25 bp) to present two consecutive major groove elements required for binding the two HTH elements of an RstR dimer.
RstR Is Tetrameric When Bound to Operator DNA-After affinity purification, the oligomeric state of RstR6H was probed by gel permeation chromatography and formaldehyde crosslinking. As shown in Fig. 4A, RstR6H eluted as two peaks with M r values of ϳ24,000 and ϳ46,000. We suggest that these peaks represent a mixture of dimers and tetramers (predicted M r values of ϳ28,000 and 56,000 for dimers and tetramers, respectively). The apparent molecular weights deviated from the predicted values for dimers and tetramers, a discrepancy that may be due to RstR6H folding into a more compact shape than a typical globular protein. When increasingly dilute RstR6H samples were chromatographed, the peak corresponding to dimers increased in size relative to the tetramer peak, indicating that tetramers dissociate into dimers. We estimated the apparent dimer/tetramer dissociation constant, K tet , to be ϳ0.5 M monomer RstR6H. Therefore, RstR6H would be primarily dimeric, or possibly monomeric, at the nanomolar concentrations where DNA binding is observed. As shown in Fig.  4B, RstR6H could be cross-linked to dimers using formaldehyde (apparent M r ϭ ϳ28 kDa).
The size and pattern of RstR footprints suggested to us that RstR binds operator DNA as a dimer-of-dimers. To determine precisely the oligomeric state of RstR when bound to operator DNA, we employed the mixed oligomer technique devised by Hope and Struhl (20). A lengthened variant of RstR repressor, RstR-CBD-6H, was mixed with RstR6H and used in mobility shift assays of O1 operator binding. RstR-CBD-6H contains 52 additional C-terminal residues and retards O1 DNA to a greater extent than RstR6H in mobility shift assays (Fig. 5). RstR activity was unaffected by the C-terminal addition in RstR-CBD-6H, because the purified fusion protein bound O1 DNA with an affinity comparable with that of RstR6H.
As shown in Fig. 5, a total of five shifted bands were observed when mixtures of RstR6H(short) and RstR-CBD-6H(long) were incubated with labeled O1 DNA. The slowest and fastest migrating complexes co-migrated with the RstR-CBD-6H and RstR6H complexes, respectively. The bands of intermediate mobility represent complexes that contain distinct mixtures of "short" and "long" forms of RstR. If RstR bound to its operator as a tetramer, five complexes corresponding to 4:0, 3:1, 2:2, 1:3, and 0:4 mixtures of short and long RstR would be expected to form. Assuming random assortment of repressor subunits, an equimolar mixture of short and long RstR would be predicted to give rise to these five complexes with relative molar ratios of 1:3:6:3:1. The band intensities that resulted from mixing ap- proximately equimolar amounts of short and long RstR generally fit the predicted pattern of band intensities for tetrameric binding (Fig. 5). The simplest conclusion from these data is that RstR is tetrameric when bound to operator DNA. A tetrameric binding model was also supported by direct measurements of RstR-DNA stoichiometry. In mobility shift experiments with O1 DNA present at high concentrations (one-tenth micromolar) and using our most active RstR6H preparations, we observed that ϳ21 pmol of RstR6H monomer was required to completely shift 5 pmol of O1 DNA (data not shown).
Cooperative Binding of RstR to Operator DNA-Although RstR is dimeric at concentrations where operator binding is observed, RstR is tetrameric when bound to operator DNA. Operator binding could occur by the sequential, independent binding of RstR dimers or by the cooperative binding of two RstR dimers. We favor a cooperative binding model, because mobility shift assays and DNase I footprinting studies show that RstR binds operator DNA over a narrow range (Ͻ10-fold) of repressor concentrations (Figs. 1 and 2). Also, we have never observed an intermediate shifted band in mobility shift experiments that would indicate dimer-bound complexes with O1 DNA. To examine the possibility of cooperative binding further, detailed mobility shift analyses were carried out, and the resulting data were fit to quantitative models of repressor binding. The results of these experiments are shown in Fig. 6. The fraction of O1 DNA bound by repressor increased from 0.1 to near unity over an ϳ10-fold range of RstR6H concentrations, exhibiting half-maximal binding at ϳ3.9 nM RstR6H. These RstR binding data were successfully modeled by a theoretical binding curve that describes two free repressor dimers binding DNA cooperatively to form tetrameric repressor-operator complexes, with a dissociation constant of 1.5 ϫ 10 Ϫ17 M 2 (see "Calculations" under "Experimental Procedures"). A Hill plot of the data in Fig. 6 yielded a best fit line with a slope of 2.0 Ϯ 0.1, indicating that operator binding occurs cooperatively and that a significant fraction of RstR is dimeric at concentrations where DNA binding occurs.
Finally, we investigated whether RstR6H dimers alone could form stable complexes with left half or right half fragments of O1. Measuring the equilibrium binding constant for dimer binding would allow us to determine the coupling factor or the energetic contribution made by dimer-dimer protein interactions to the overall DNA binding reaction (21) (assuming that dimer binding affinity for the left and right halves of O1 are identical). However, binding of RstR6H to a 33-bp left half or a 28-bp right half operator fragment was not detected in mobility shift experiments, even at micromolar RstR concentrations (data not shown). The apparent low affinity of RstR dimers for left and right halves of O1 DNA suggests that protein-protein interactions between DNA-bound dimers make a significant energetic contribution to operator binding. DISCUSSION We have found that the RstR repressor of CTX ET binds specifically to three DNA sites surrounding the rstA promoter. The high affinity O1 site encompasses almost the entire promoter region (Fig. 3), whereas the two lower affinity sites, O2 and O3, are located Ϫ40 to Ϫ150 upstream of the rstA promoter. Preliminary in vivo studies with rstA::lacZ fusions indicate that O1 and O2-O3 function independently to repress expression of P rstA . A second function for O2-O3 is suggested by the observation that O2 overlaps the divergent rstR promoter, indicating that O2-O3 may also play a role in the autoregulation of rstR expression.
In mobility shift assays, RstR binding to O3 exhibited an unusual dependence upon a neighboring intact O2 site, indicating that RstR is recruited to bind O3 by RstR bound to O2. It will be interesting to investigate whether this recruitment is mediated by direct protein-protein contacts or through an altered DNA conformation resulting from RstR binding to O2. It is unlikely that the pattern of RstR binding to O2-O3 is due to nonspecific repressor binding to neighboring DNA, a process  Fig. 1. A, representative data obtained from one mobility shift experiment. The faint band migrating just above the free probe is a minor contaminant in the probe preparation. B, the fraction of DNA bound, , was measured at numerous repressor concentrations in two independent experiments. The solid line is a theoretical binding curve describing the dissociation of tetrameric RstR-DNA complexes to two free dimers and free operator DNA with a dissociation constant of 1.5 ϫ 10 Ϫ17 M 2 (see Equation 5). The inset depicts a Hill plot of RstR6H binding to O1 DNA. Data from duplicate experiments were averaged and plotted as a single point. The line is a linear regression best fit to the data. also known as "phasing" (22). The DNase I footprint at O3 was similar to the footprint at the high affinity site O1, indicating that O3 binding involves similar protein-DNA contacts. Also, phased binding of RstR was not observed adjacent to O1, even at high repressor concentrations.
The primary conclusion of this work is that RstR tetramerizes on operator DNA, using protein-protein interactions between neighboring dimers to stabilize repressor-operator complexes. The ϳ50 bp size of each RstR footprint, together with the staggered pattern of DNase I-accessible regions within each footprint (Fig. 4), suggested that RstR binds DNA as a dimer-of-dimers, with the binding face of one dimer rotated ϳ100°about the DNA helical axis from the neighboring dimer. RstR was indeed tetrameric when bound to operator DNA, as shown by the ability of the short and long forms of RstR6H to form five distinct complexes with O1 DNA in mobility shift experiments (Fig. 4). This method was previously used to show that Arc repressor of bacteriophage P22 and the mammalian transcription factor LSF also bind DNA as tetramers (23,24). Better studied HTH repressors, such as LacI, GalR, or CI and Cro of phage , bind to their operator sites as homodimers and footprint 20 -25-bp regions or about two full turns of the DNA helix (18,25,26). The two HTH elements (one/repressor monomer) of a CI dimer, for example, are spaced such that each recognition helix can contact adjacent major groove elements on one face of the DNA helix (27). Our model for the staggered arrangement of RstR dimers on operator DNA is strikingly similar to the arrangement of two CI dimers bound to O R 1 and O R 2 of bacteriophage (1,28). Despite this similarity, the protein-protein contacts that mediate the cooperative binding of RstR dimers are probably different from the contacts that mediate the pairwise cooperative binding of CI dimers to O R 1 and O R 2 (29). The C-terminal region of RstR is only 52 residues long and does not share sequence similarity to the C-terminal domain of CI repressor, which mediates CI oligomerization. Also, secondary structure predictions indicate that the RstR C-terminal region is largely ␣-helical, whereas the C-terminal domain of CI consists largely of ␤-sheet elements (29).
Several lines of evidence suggest that it is unlikely that each RstR operator is merely two classically sized operators situated close together. First, RstR6H bound strongly to a synthetic 50-bp DNA corresponding to the footprinted region at O1 but failed to bind to any subfragments of O1 in mobility shift assays, including 33-and 28-bp DNAs corresponding to the left and right halves of O1, respectively. Also, RstR failed to stably bind to any of the AluI subfragments of 301 and 309 DNA that contain large portions of O1, O2, or parts of both sites. Finally, ␤-galactosidase expression from an rstA::lacZ fusion that contains the leftmost 26 bp of O1 and the rightmost 39 bp of O2 was not repressed by RstR in vivo (data not shown). The entire 50-bp footprinted region is required for strong repressor binding.
To identify potentially important RstR-DNA contacts, we searched the O1, O2, and O3 sequences for an RstR consensus binding site. At O1, two sets of inverted repeats were identified that contained the potential "half-site" sequence CTNN(A/ C)AAG (Fig. 3). However, this half-site sequence was not readily identified at the expected positions in O2 or O3. A survey of other HTH transcription regulators that have extended binding sites, such as OxyR, OccR, and Rns, indicates that consensus binding sites for these regulatory factors were also difficult to identify, in some cases requiring the identification of numerous natural binding sites or the in vitro generation of many synthetic binding sites (30 -32). As others have pointed out, such extended binding sites could be biologically advantageous, allowing for sites with large variations in transcription factor binding affinities (30).
The steep nature of the DNA binding curve (Fig. 6), plus the absence of any detectable dimer-bound complexes in mobility shift experiments, indicate that RstR binds operator DNA cooperatively. These observations account for the overall reaction but ignore the possible role of dimeric RstR-operator complexes as intermediates in the binding reaction. We imagine two possible pathways for the assembly of tetrameric complexes. In one, dimers first combine to form free tetramers, which then bind to operator DNA in a coupled reaction.
Although our gel filtration experiments indicate that tetramers only form at high repressor concentrations, the small amount of tetramers formed at low repressor concentrations could be rapidly trapped by operator binding. Alternatively, two RstR dimers could bind sequentially to operator DNA to form the final tetrameric complex.
Our mobility shift experiments did not show evidence of dimer complexes. However, a dimer complex might be a very shortlived intermediate or sufficiently unstable to be seen in our mobility shift assays. Kinetic binding experiments carried out under conditions where RstR is primarily dimeric or tetrameric would aid in determining the active form of RstR.
Gene transcription is frequently regulated by factors that bind cooperatively to DNA. The energetic contribution of cooperativity is often, although not always, mediated by proteinprotein interactions. These interactions fall into two broad classes. In one class are factors that utilize protein-protein interactions to loop DNA between distant sites, as in the case of the LacI and GalR repressors (33,34) and the AraC regulator (35). In these systems the spacing of distant operator sites can often be altered without disrupting normal regulation. The second class consists of regulatory proteins that bind and oligomerize to closely spaced sites on the DNA. Examples include the Arc and Mnt repressors of phage P22 (21,23,36), LexA (37), and the MCM1/␣2 complex of yeast (38). RstR may belong to this later class of regulators. On-going studies of mutant binding sites indicate that as little as a 1-bp deletion near the center of O1 drastically reduces RstR binding, suggesting that RstR dimers must bind to correctly spaced sites for dimer-dimer interactions to stabilize the tetrameric complexes.