Characterizing the DNA Contacts and Cooperative Binding of F Plasmid TraM to Its Cognate Sites at oriT *

TraM is a DNA binding protein required for conjugative transfer of the self-transmissible IncF group of plasmids, including F, R1, and R100. F TraM binds to three sites in ForiT: two high affinity binding sites, sbmA andsbmB, which are direct repeats of nearly identical sequence involved in the autoregulation of the traM gene; and a lower affinity site, sbmC, an inverted repeat important for transfer, which is situated nearest to the nic site where transfer originates. TraM bound cooperatively to its binding sites atoriT; the presence of sbmA and sbmBincreased the affinity for sbmC 10-fold. Bending oforiT DNA by TraM was minimal, suggesting that TraM, a tetramer, was able to loop the DNA when bound to sbmA andsbmB simultaneously. Hydroxyl radical footprinting of DNA of sbmA and sbmC revealed that TraM contacted the DNA within a region previously delineated by DNase I footprinting. TraM protected the CT bases within the sequence CTAG, which occurred at 12-base intervals on the top and bottom strand of sbmA, most consistently with other protected bases. The footprint onsbmC revealed that the predicted inverted repeats were protected by TraM with a pattern that began at the center of the repeats and radiated outward at 11–12 base intervals toward the 5′-ends of either strand. At high protein concentrations, this pattern extended beyond the footprint defined by DNase I, suggesting that the DNA was wrapped around the protein forming a nucleosome-like structure, which could aid in preparing the DNA for transfer.

TraM is a DNA binding protein required for conjugative transfer of the self-transmissible IncF group of plasmids, including F, R1, and R100. F TraM binds to three sites in F oriT: two high affinity binding sites, sbmA and sbmB, which are direct repeats of nearly identical sequence involved in the autoregulation of the traM gene; and a lower affinity site, sbmC, an inverted repeat important for transfer, which is situated nearest to the nic site where transfer originates. TraM bound cooperatively to its binding sites at oriT; the presence of sbmA and sbmB increased the affinity for sbmC 10-fold. Bending of oriT DNA by TraM was minimal, suggesting that TraM, a tetramer, was able to loop the DNA when bound to sbmA and sbmB simultaneously. Hydroxyl radical footprinting of DNA of sbmA and sbmC revealed that TraM contacted the DNA within a region previously delineated by DNase I footprinting. TraM protected the CT bases within the sequence CTAG, which occurred at 12-base intervals on the top and bottom strand of sbmA, most consistently with other protected bases. The footprint on sbmC revealed that the predicted inverted repeats were protected by TraM with a pattern that began at the center of the repeats and radiated outward at 11-12 base intervals toward the 5-ends of either strand. At high protein concentrations, this pattern extended beyond the footprint defined by DNase I, suggesting that the DNA was wrapped around the protein forming a nucleosome-like structure, which could aid in preparing the DNA for transfer.
The F plasmid is a 100-kbp circular plasmid 1 found in Escherichia coli and is a paradigm for pilus-mediated conjugation. Transfer proteins required for conjugation have been divided into five categories based on relaxosome formation and DNA transport, pilus assembly, mating pair stabilization, surface exclusion, and regulation (1). The relaxosome is a nucleoprotein complex of TraI, TraY, and TraM bound to the origin of transfer (oriT), which interacts with the DNA transport protein TraD, located in the inner membrane. TraI is both a relaxase, which cleaves at the nic site within oriT in a strand-and sequence-specific manner, and a helicase, which is essential for transfer (2,3). TraY facilitates TraI relaxase activity at nic (4,5), whereas TraM is a DNA binding protein that binds to three sites in the oriT region (sbmA, -B, -C). TraM is not required for cleavage at oriT (5,6), although it is essential for conjugation (7). Because TraM is thought to bind TraD (8), it might function by linking the relaxosome to the transferosome, a multiprotein complex that spans the cell envelope and is responsible for pilus assembly and mating pair formation.
The highest affinity binding sites for F TraM are sbmA and -B, which are involved in the autoregulation of the traM gene (9), whereas the lowest affinity site, sbmC, is located nearest to nic and appears to have an important role in DNA metabolism during transfer. Removal of sbmA and -B from a cloned oriT segment decreases transfer by 100-fold, whereas removal of sbmC further decreases transfer another 100-fold (10). A second function of sbmA and sbmB may be to assist binding of TraM to sbmC, the weakest binding site, which is also the most critical for conjugation. This is supported by a similar mutational analysis of R100 (an F-like plasmid), where mutagenesis of the proximal half of sbmC (closest to nic) decreased mating 120-fold, whereas mutagenesis of any of the other TraM binding sites decreased mating to a maximum of 7-fold (11).
The multiple binding sites may function in a cooperative fashion (12) whereby low concentrations of protein act synergistically to control DNA-related events. Cooperative binding of TraM to its binding sites was determined using electrophoretic mobility shift assays (EMSA) 2 (2). TraM⅐DNA contacts within the sbmA and -C sites were mapped using hydroxyl radical footprinting. The results suggest that TraM contacts slightly different bases than predicted and that the pattern of contact by TraM on the DNA suggests the formation of a nucleosome-like structure, which might aid in readying the DNA for transfer.

EXPERIMENTAL PROCEDURES
Recombinant DNA Techniques-Restriction enzymes, calf intestinal alkaline phosphatase, Klenow fragment, T4 polynucleotide kinase, and T4 DNA ligase were supplied by Roche Diagnostics. All procedures were as described in Ausubel et al. (13) except where noted. Plasmids were transformed using CaCl 2 competent cells (14) or by electroporation using a Bio-Rad Gene Pulser at 2.5 V, 25 microfarads, and 200 ⍀. DNA fragments used to create plasmid constructs were isolated from acrylamide (14) or agarose gels (Qiagen gel extraction kit). All gels were run in 89 mM Tris, 89 mM boric acid, 2 mM EDTA, pH 8.0, buffer unless otherwise indicated. PCR was performed using Vent polymerase (New England BioLabs) with 20 mmol of dNTP (Roche) and ϳ500 pmol of each primer in a 100-l volume for 30 cycles. Fill-in reactions were done using Klenow polymerase and 500 pmol of dNTP in 30 l at 37°C for 30 min. Plasmids were isolated using the method of Birnboim and Doly (15) 1 The nucleotide sequence of the F plasmid is available in the Gen-Bank TM database under accession number NC002483.
nase. The primers were heated to 95°C, cooled to room temperature, digested with BlnI for 30 min, and ethanol-precipitated. This fragment was ligated to 0.3 pmol of pBEND2 (16) digested with XbaI, dephosphorylated with alkaline phosphatase, and purified from a 0.6% agarose gel. pRF918 was constructed by annealing 400 pmol of LFR49 (CTAGAG-CAGCGCCCCTAGCGG) and LFR50 (CTAGCCGCTAGGGGCGCTGCT) and ligating to 0.3 pmol of pBEND2 digested with XhoI. pRF920 was constructed by digesting pNY300 with RsaI and DraI and purifying the 58-bp band from an agarose gel. This was ligated to pBEND2, which had been digested with SalI, dephosphorylated with alkaline phosphatase, and filled in with Klenow polymerase. pRF940 was constructed by digesting pNY300 with DraI and SalI, and purifying the 300-bp fragment from a 5% polyacrylamide gel. This was then digested with BstBI and filled in with Klenow polymerase, and the 190-bp fragment was purified from an 8% acrylamide gel, which was ligated to pBEND2, which had been digested with XbaI, filled in with Klenow polymerase, dephosphorylated with alkaline phosphatase, and isolated from a 0.6% agarose gel. pRF930 was constructed by digesting pRF940 with BamHI, ClaI, and RsaI, filling in with Klenow polymerase, and purifying the 144-bp fragment from an 8% acrylamide gel. This was ligated to pBEND2 digested with XbaI, dephosphorylated with alkaline phosphatase, filled in with Klenow polymerase, and isolated from a 0.6% agarose gel.
Electrophoretic Mobility Shift Assays-PCR was used to amplify sequences from pRF911, pRF912, pRF918, pRF920, pRF930, and pRF940 with primers RFE16 (GGTGCCTGACTGCGTTGCA) and RFE17 (TAGGCGTATCACGAGGCCCT). DNA was radiolabeled by including ϳ50 Ci of [␣ 32 P]ATP in the PCR reactions. Reactions were concentrated in a Savant SpeedVac, and the DNA was isolated from a 1.5% agarose gel. A Beckman LS3801 scintillation counter was used to determine the concentration of labeled DNA using the specific activity of [␣ 32 P]ATP. Approximately 3000 cpm of DNA was used in each reaction with the specified amount of purified TraM. Binding reactions were conducted in binding buffer (final concentration 50 mM Tris, 10% glycerol, 1 mM dithiothreitol, 30 g/ml bovine serum albumin) with 1 g of poly(dI⅐dC) in a final volume of 15 l. After addition of protein, binding reactions were incubated at 37°C for 15 min before loading on a 8% acrylamide gel containing 89 mM Tris, 89 mM boric acid which had been pre-run at 4°C at 30 mA. Gels were run at 4°C at 30 mA until the bromphenol blue marker reached the bottom of the gel. Gels were analyzed using a Molecular Dynamics PhosphorImager 445SI using ImageQuaNT version 4.2a software.
Bending of TraM-bound DNA-PCR-amplified DNA (ϳ50,000 cpm) was digested with a series of restriction enzymes specific for the MCS of pBEND2 for 8 h. Digests were extracted with phenol and chloroform, ethanol-precipitated, and resuspended in 20 l of water. 2 l of the digested DNA was run in each lane of an 8% polyacrylamide gel following the same protocol as for EMSA. The relative mobility of each retarded band was calculated by comparing its mobility to that of the BamHI-digested DNA, which represented the most linear fragment with the highest predicted mobility. The maximal bend angle was then estimated using the formula ␣ ϭ 2[cos Ϫ1 (relative mobility)] (16).
Determination of Cooperativity of Binding-The presence of cooperativity of binding was assessed using four different but related approaches: 1) Hill plots (12) were constructed by graphing the log of the free DNA concentration divided by that of the total complexed DNA versus the log of the protein concentration. The Hill coefficient (n H ), was calculated from the maximum slope of the curve and used to determine cooperative binding (n H greater than 1). 2) The breadth of the binding curve was used to determine affinity of TraM for its binding sites according to the method of Freifelder (12) with cooperativity being defined as less than 1.81 log units between the TraM concentrations that define 10 and 90% bound DNA (17). 3) The cooperativity factor, , was used to determine cooperativity when more than one species was bound during EMSA using the equation, ϭ 4͑unbound species)(fully bound species) (intermediate species) 2 (Eq. 1) Cooperativity was assumed if values of exceeded 1. 4. Cooperativity was also measured by determining the value of (18,19) using the association constants for two individual sites in a given fragment of DNA. For example, if the individual sites are X and Z, If the association constants for each site are identical, then, ϭ ͑1/K a of smaller site) 2 (free protein at 50% binding of larger site) 2 (Eq. 3) Values of greater than 1 indicate cooperativity. Competition Assays to Determine the Specificity of Binding-The binding sites cloned into pRF911 (sbmA) and pRF940 (sbmABC) were amplified by PCR using primers RFE16 and RFE17 and used as competitor DNA in competitive electrophoretic mobility assays. The nonradioactive DNA was purified from a 1.5% agarose gel and quantified by absorbance at 260 nm. 0.5 fmol of radiolabeled template was incubated with various amounts of TraM for 15 min at 37°C. Non-radioactive sbmABC (0.5 amol to 500 fmol) was then added in a 1-l volume and incubated for an additional 15 min at 37°C. Reactions were resolved on a 5% acrylamide gel, and the intensity of bands was determined by phosphorimaging the dried gel.
Hydroxyl Radical Footprinting-Approximately 0.5 g of plasmid DNA (pRF911, pRF920) was digested with PvuII and run on a 1.8% agarose gel. The 156-and 183-bp (respectively) fragments were excised from the gel and purified. Sequencing reactions were performed using RFE9 (GCTGCCCGGGAGGCCTTC) or RFE10 (GCTGGATATCTTTA-AACTCGAG), Sequenase (United States Biochemical), and 33 P-labeled ddNTP (Amersham Biosciences, Inc.). 125 pmol of RFE9 or RFE10 were end-labeled using 5 l (50 Ci) of [␥-32 P]ATP (ICN) and polynucleotide kinase for 1 h. Primers were purified using Quick Spin Oligo columns (Roche Molecular Biochemicals). PCR reactions using the end-labeled primer and 125 pmol of either RFE9 or RFE10 were performed, and the product was run out on a 1.8% agarose gel. The 158-and 185-bp fragments were excised and purified. Radioactivity was quantified using a scintillation counter and ϳ5000 cpm was used in each reaction. Purified TraM was allowed to bind to the DNA in a 15-l volume in 100 mM Tris, 1 mM EDTA, pH 8.0, buffer for 15 min at 37°C. 2 l of each of 10 mM sodium ascorbate, 0.2 mM iron ammonium sulfate, 0.4 mM EDTA, and 0.3% H 2 O 2 (v/v) were mixed and added to the binding reaction. After 2 min 12 l of stop solution (32 l of 0.2 M EDTA, 10 l of 0.1 M thiourea, and 1 l of 0.5 mg/ml tRNA) was added with 3 l of 100% glycerol to stop the cleavage reaction. Reactions were loaded onto a 5% TB polyacrylamide gel and run at 28 mA at 4°C until the bromphenol blue dye reached the bottom of the gel. The gel was exposed to X-OMAT AR film (Kodak) for 4 h. After development of the film, bands, which corresponded to the bound and unbound DNA, were excised and the DNA was eluted overnight in 400 l of 0.5 M ammonium acetate, 1 mM EDTA at 37°C. The samples were clarified by centrifugation at 14,000 rpm in a microcentrifuge. 1 ml of 95% ethanol was added to the aqueous layer to precipitate the DNA. The DNA was dissolved in 5 l of Sequencing Stop Solution (United States Biochemical) and loaded onto an 8% polyacrylamide gel containing 8 M urea. As a marker, 1 l of the 10-l G, A, T, and C sequencing reactions was loaded onto the gel. The gel was run at 40 watts, until the xylene cyanol marker was ϳ5 cm from the bottom, dried, and exposed to a Molecular Dynamics Phosphor Screen for up to 5 days. The intensity of each band was determined by scanning each lane using ImageQuaNT software, and a consensus for the pattern of binding was generated from three clearest gels.

RESULTS
TraM Binds sbmA, -B, -C Cooperatively-EMSA was used to determine the affinity of TraM for fragments of DNA containing sbmA, -B, and/or sbmC derived from the F oriT region. The TraM binding sites were cloned individually and in combinations into pBEND2 (Ref. 16, Fig. 1) to give pRF911 (sbmA), pRF918 (half of the sbmA site), pRF920 (sbmC), pRF930 (sbmAB), and pRF940 (sbmABC). Examples of EMSA for these fragments and a negative control (MCS of pBEND2) are shown in Fig. 2, A-F. Shifts above 1700 nM TraM were considered to be the result of nonspecific binding as demonstrated in the negative control (Fig. 2F). TraM bound to sbmA, sbmAB, and sbmABC in an ordered fashion giving one or two retarded species (complexes I and II) below 1700 nM. A finer gradation in increasing TraM concentrations gave more precisely defined patterns of band retardation for sbmAB and sbmABC, which were used to determine association constants and the presence of cooperativity (Fig. 3, A and B). Only two species were initially seen for sbmABC (Fig. 2), which were resolved into three distinct bands in Fig. 3B (complexes I-III). Increased amounts of TraM gradually decreased the mobilities of sbmC and the half sbmA site, with a faint intermediate band appearing in the latter (Fig. 2, B and C). This was interpreted as a lower affinity for these sites and indicated that either TraM requires both half sites in cis for high affinity binding or the smaller repeats within sbmA are not sufficient for TraM recognition. By calculating the percentage of unbound DNA in each lane, association constants (K a ) were estimated (1/[TraM] at 50% unbound, Table I). TraM bound sbmA and sbmAB with similar affinities (1-2 ϫ 10 8 M Ϫ1 ), sbmABC with the highest affinity (5 ϫ 10 8 M Ϫ1 ), and sbmC and half sbmA with ϳ3to 15-fold lower affinities than sbmA. These results reflected the DNase I protection pattern of the F oriT region, which showed that the order of binding was sbmAϾsbmBϾsbmC (6).
Determining Cooperativity of Binding-TraM could bind sbmA, sbmC, sbmAB, or sbmABC cooperatively because of the presence of two half sites in sbmA and sbmC as well as multiple sites in the more complex fragments. Consequently, a variety of methods were used to assess cooperative binding ("Experimental Procedures"). Using the EMSA data (Figs. 2 and 3), Hill plots were constructed for sbmA, sbmC, sbmAB, and sbmABC and used to determine the Hill coefficient, n H (Table I). If n H is greater than 1 and less than n, the number of binding sites, cooperativity is assumed. The n H value for sbmA was 1.7 indi-  cating cooperative binding to sbmA, which has two TraM binding sites. A Hill coefficient of 0.9 was found for sbmC, which contains also two predicted binding sites, indicating non-cooperative binding. Hill coefficients of 3.25 and 2.9 were found for sbmAB (four sites) and sbmABC (six sites), respectively, indicating cooperative binding of TraM to both these fragments. The breadth of the binding curve, and values (Table I), also confirmed that TraM bound sbmA, sbmAB, and sbmABC cooperatively, with TraM binding to sbmABC with the greatest degree of cooperativity. Calculations for were not done for sbmABC, because the presence of three shifted species on an EMSA complicated the calculations.
Competition Assays to Determine Specificity of Binding-The stabilities of TraM⅐DNA complexes described above were determined by performing EMSA in the presence of increasing amounts of unlabeled sbmA or sbmABC competitor DNA. The amount of TraM used to generate either complex I or II(III) was based on the results presented in Fig. 2. Competitive EMSA using sbmA as a competitor gave similar results to assays using sbmABC and are not shown. TraM bound to sbmA (complex I, Fig. 4A), sbmAB (complex II, data not shown), or sbmABC (complex II(III), Fig. 5) could be competed off at high concentrations of either competitor DNA. In the case of sbmAB (data not shown) and sbmABC (Fig. 5), complex II was shifted to the position of complex I at the highest concentrations of competitor DNA. Complex I of sbmAB or sbmABC was extremely stable and could not be competed off over the range of competitor DNA used during a 15-min incubation. TraM bound to sbmC was completely competed off with 100-fold excess of competitor DNA, supporting the idea that TraM has a lower affinity for linear DNA containing sbmC alone (Fig. 4B).
DNA Bending Assays-The ability of TraM to bend DNA was estimated using circular permutation assays (see "Experimental Procedures"). Using primers that bordered the MCS of pBEND2, DNA fragments from pRF911 (sbmA) and pRF940 (sbmABC) were amplified by PCR and digested with restriction enzymes to generate a nest of fragments of equal length but with the TraM binding site positioned at different distances from the ends of the fragments. The fragments, with or without TraM bound, were subject to EMSA, and the relative mobilities of fragments were calculated (Fig. 6). Using the equation from Kim et al. (16), TraM was estimated to bend sbmA by 50°or less. No detectable bending of sbmABC was observed, suggesting that TraM bound to all three sites to form a compact structure possibly involving bends and loops in the DNA that was insensitive to the position of the large sbmABC binding site within the DNA fragment.
Hydroxyl Radical Footprinting-Previously, DNase I footprinting had delineated sequences (smbA, -B, -C) in the oriT region protected by TraM, which suggested that TraM contacted the upper, retained DNA strand more than the lower, transferred strand. These protected regions were used to pre-dict a consensus sequence, which was thought to be important for TraM recognition of its binding sites (NT(A/G)(G/T)(G/ C)GG(C/T)GC(T/G)GCTA (6)). To determine which nucleotides are contacted by TraM within these binding sites, hydroxyl radical footprinting of the upper and lower strands of DNA containing sbmA and sbmC was performed (Fig. 7, A and B), and the results from three different gels are summarized in Fig. 8. Unlike the DNase I protection assay, protection of bases in the top and bottom strands was approximately equal. There also appeared to be protection of bases outside the sbmC foot-  The hydroxyl radical footprint for sbmA did not align precisely with the predicted consensus sequence but was shifted somewhat upstream (solid arrows compared with dashed arrows in Fig. 8A). Although the GC-rich nature of the TraM binding sites is striking, TraM contacted the CT bases within the CTAG motif, repeated twice, 12 bases apart, on both the upper and lower strands of sbmA, suggesting that sequence recognition was dependent on the non-GC bases within this 4-base palindrome.
The pattern of contact between TraM and bases within sbmC appeared to be unrelated to the footprint for TraM and sbmA (Fig. 8B). The strongest binding was in the center of the inverted repeat with both strands of the DNA being protected. This sequence bears no resemblance to the CTAG motif in sbmA suggesting that sequence recognition and protection did not coincide with each other. Radiating out from this central footprint are protected regions spaced approximately every 11 bases toward the 5Ј-ends of the DNA fragment. The bases protected at these intervals do not in any way resemble the sequences protected in sbmA.

DISCUSSION
An improved purification procedure for TraM has allowed better resolution of shifted bands in EMSA, resulting in the estimation of K a values that reflect the affinity of TraM for its binding sites. Although TraM is predominately a tetramer upon purification (6,20,21), monomer and dimer forms are present as measured by gel exclusion and analytical ultracentrifugation. Resizing purified tetrameric TraM resulted in the reappearance of monomers and dimers, suggesting that the three forms of TraM exist in equilibrium (3).
The cooperative nature of TraM binding, especially to sbmABC, was clearly evident and probably reflects the tetrameric nature of the protein. In binding to sbmAB, an initial complex (I) of TraM bound to sbmA could be converted to a single tetramer of TraM bound to both sbmA and sbmB simultaneously (complex II) upon increasing concentrations of TraM. Thus the small amounts of the monomeric or dimeric forms of TraM might initiate the binding process, followed by tetramerization of TraM during cooperative binding to these sites. Preliminary evidence on estimating the number of TraM molecules bound to sbmAB suggested that complex I contains less than a tetramer of TraM whereas complex II does indeed contain a tetramer. More complete studies are currently under way. The number and arrangement of TraM subunits within complex III of sbmABC is not obvious, because TraM is apparently binding up to six sites involving direct and inverted repeats simultaneously.
Analysis of F TraM, bound to its individual binding sites (sbmA and sbmC), showed little bending of the DNA. This is in accord with results for TraM from the F-like plasmid pED208, which had no detectable effect on DNA bending as determined by electron microscopy (22). Instead, IHF, which has two binding sites in oriT bordering sbmC, has been implicated in construction of the relaxosome in F and pED208 (22,23). The sequence between sbmA and sbmB does not contain an IHF binding site (3), suggesting that TraM could bind sbmA and -B simultaneously, folding the DNA back on itself to form a hairpin. Thus, bending the DNA in the sbmABC region could involve a combination of IHF and looping by TraM.
An interesting phenomenon involved the binding patterns of TraM to sbmA, sbmAB, and sbmABC versus sbmC and a half sbmA site. The binding pattern of the first three gave defined species suggesting the formation of distinct complexes that were composed of specified amounts of protein and DNA. However, binding to the latter two binding sites gave species that gradually increased in size as more TraM was added, a phenomenon that was also seen in all EMSAs at high protein concentrations. This suggests that, as more protein was added, the protein complex bound to the DNA increased proportionately in size, with the added TraM being distributed approximately evenly among all the complexes. The smearing of these species during EMSA also suggested that these complexes were not stable and dissociated during electrophoresis. The striking pattern of protection in the hydroxyl radical footprints occurred at approximately every 11-12 bp. This suggests a model whereby TraM could wrap the DNA around itself to form a large complex with the protein contacting 4 -5 bases approximately every 11 bp. This would resemble the nucleosome-like structure proposed for TraK in the RP4 system (24). TraK bends the DNA and wraps it around a multimeric core of protein in a process that is independent of IHF.
Competition assays demonstrated that complex I and II(III) for sbmA, sbmAB, and sbmABC, were found to be highly stable and resist dissociation in the presence of a 1000-fold excess of specific DNA competitor. At very high concentrations of competitor DNA, small amounts of dissociation were seen for complex II of sbmAB and sbmABC, which resulted in formation of complex I. Complex I resisted dissociation in all cases suggesting a very stable complex was formed. Binding to sbmC was not as strong, and dissociation was seen when competitor DNA reached a 100-fold excess. Clearly, the presence of sbmA and sbmB stabilizes TraM binding to sbmC. These experiments were conducted using linear DNA templates, however, supercoiling is predicted to have an effect on TraM activity and possibly increase the degree of cooperativity even further, an aspect of TraM function that is currently under study.
Hydroxyl radical footprinting was also performed on TraM bound to sbmA and sbmC to define which bases were protected within the previously published DNase I footprint (6). Footprinting of sbmA showed that TraM had as many points of contact on the upper and lower strands in contrast to the results for DNase I footprinting, which indicated more contact with the upper, retained strand. Modeling TraM onto sbmA suggested that a dimer initially recognizes the sequence CTAG and binds to two adjacent major grooves 12 bp apart on the same face of the DNA (Fig. 8). This would agree with the tetrameric nature of TraM where each dimer within a tetramer would interact with two major grooves at two binding sites (sbmA and sbmB, for instance), in a cooperative fashion.
Footprinting of sbmC showed that TraM made contact with FIG. 6. Analysis of sbmA (A) and sbmABC (B) bending by TraM. The fragments were digested with the restriction enzymes indicated and electrophoresed with or without 17 and 6.8 nM TraM. The unbound and bound DNA fragments are indicated by the lower and upper arrows, respectively. The bands below the lower arrow represent small labeled fragments resulting from digestion of the PCR products with different restriction enzymes, which were used as negative controls for TraM binding as well as internal markers for the distance each band had migrated in the gel. the top and bottom strands in a symmetrical pattern, although there appeared to be no sequence conservation between the contacted sites. The footprints on the top and bottom strands were approximately equal in intensity and were centered at the ACAACA sequence in the center of the footprint. Interestingly the footprints extended from the central region toward the 5Ј-ends of the DNA fragment. Two dimers can be modeled onto the sbmC site with each dimer occupying two major grooves on the upper and lower strands, respectively. The dimers would overlap at the center of the inverted repeat giving protection on both strands (Fig. 8B). The spacing of 11-12 bp is suggestive of a phase-dependent process with one face of the DNA repeatedly contacting the protein. A consensus sequence was previously generated from the DNase I footprints by compiling the sequences in all three TraM binding sites (6). However, the hydroxyl radical footprint for sbmA demonstrated that contact with the DNA was shifted slightly upstream toward the palindromic sequence CTAG and that all four CTAG sequences in sbmA were involved (Fig. 8). TraM contacted the DNA every 11-12 bases outside the original DNase I footprint when a high concentration of TraM was used (Fig. 7, A and B). This suggested that TraM aggregates, as seen at high concentrations of protein on EMSA gels, wrapping the DNA around the protein and contacting the same face of the DNA upon each turn.
The importance of multiple binding sites was exemplified by the LacI repressor using a reporter gene containing various combinations of Lac operators (25). These experiments showed that O 1 is required for repression, with repression increasing 38-and 24-fold when O 2 or O 3 , respectively, were present. This effect was attributed to cooperativity of binding by Lac repressor tetramers to the DNA. A similar mechanism can be suggested for TraM at oriT. The weak binding sites, sbmC (10) and sbmA in R100 (corresponding to the proximal half of sbmC in F (11)), have been shown to be of critical importance in maximizing transfer efficiency. The increase in affinity of TraM for DNA fragments containing sbmABC versus sbmC alone (almost 10-fold) suggests that sbmAB may assist with loading TraM  sbmC (B). The thick brackets above the sequences represent the DNase I footprints (6). The solid arrows represent the direct and inverted repeats in Fig. 1 whereas the dashed arrows indicate the repeats based on the CTAG motif. The thin brackets represent the 12-bp spacing in sbmA (A) and the 11-bp spacing in sbmC (B). The boldface letters represent bases derived from the vector. The degree of protection is proportional to the shading of the boxes.
onto sbmC, the weakest binding site, a step that might be critical for TraM function and efficient transfer.
We have demonstrated that the intracellular concentration of TraM is critical for transfer (26). As cells enter exponential growth, the transcription of traM diminishes and the levels of TraM in the cell are undetectable by late exponential phase. If sbmC, the site nearest oriT, and ostensibly the most important for transfer, was bound by TraM in a manner that was sensitive to the intracellular concentrations of the protein, the F conjugation apparatus would have engineered a simple, sensitive mechanism for controlling conjugation ability. In wild type F-like plasmids, conjugation is repressed by the fertility inhibition system FinOP (27). F is naturally derepressed and appears to control its fertility via a complex circuitry centered around control of traM transcription by TraY (9) and other host proteins. F could modulate its readiness for transfer at oriT by controlling the amount of TraM in the cell which, in turn, would affect binding of TraM to sbmC, following binding to smbA and sbmB, in a cooperative manner. This model would predict that binding of TraM to the oriT region, especially sbmC, would affect the local superhelical density of the DNA. This would cause unwinding near oriT, most probably in the AT-rich region upstream of sbmC in sbyA, permitting entry of the TraI helicase and initiating large scale DNA unwinding and transport to the recipient cell. Because TraM has been suggested to bind TraD, the coupling protein (8,28), the ability to form a nucleosome-like structure at high intracellular concentrations plus its ability to move the DNA into position for transport by binding TraD suggest that TraM might indeed be the "signaling" protein proposed by Willetts and Wilkins in 1984 (29).