The Bacillus subtilis Competence Transcription Factor, ComK, Overrides LexA-imposed Transcriptional Inhibition without Physically Displacing LexA*

During the development of competence inBacillus subtilis the recA gene is activated by the competence transcription factor, ComK, which is presumably required to alleviate the transcriptional repression of recA by LexA. To investigate the mechanism by which ComK activatesrecA transcription we examined the binding of ComK and LexA to the recA promoter in vitro. Using hydroxyl radical protection analyses to establish the location of ComK dimer-binding sites within the recA promoter, we identified four AT-boxes in a configuration unique for ComK-regulated promoters. Gel mobility shift experiments showed that all four ComK dimer-binding sites were occupied at ComK concentrations in the physiological range. In addition, occupation of all ComK-binding sites did not prevent LexA from binding to the recA promoter, despite the fact that the ComK and LexA recognition motifs partially overlap. Although ComK did not replace LexA from the recA promoter, in vitro transcription analyses indicated that the presence of ComK is sufficient to alleviate LexA repression ofrecA.

Bacillus subtilis differentiates into cells competent for genetic transformation by synthesizing a complex DNA binding and uptake system and by activating recombination genes. Paramount among the recombination gene products that enable cells to incorporate newly acquired genetic material is the RecA protein, which plays a crucial role in recombination by promoting homologous pairing and DNA strand exchange (1,2). RecA is also essential for the regulation of the SOS DNA repair system, which is activated when DNA is damaged or when B. subtilis cells differentiate to a competent state (3).
The SOS system operates in many bacteria and has been the subject of extensive studies in Escherichia coli (for review, see Ref. 4). Following exposure to DNA damaging treatments, a set of damage inducible (din) genes becomes transcriptionally activated. Expression of B. subtilis din genes is regulated by the products of the lexA (formerly called dinR) and recA genes (5,6). LexA acts as the repressor of all B. subtilis din genes, including recA and lexA, by binding specifically to DNA sequences located within the putative promoter regions (5,(7)(8)(9)(10). Comparison of the DNA sequences of more than 20 din promoter regions that bind LexA revealed a consensus sequence for binding of a LexA dimer, CGAACATATGTTC. 1 Analogous to the situation in E. coli, RecA is required for SOS induction in B. subtilis (6,11,12). B. subtilis RecA is activated to promote the autocleavage of LexA repressor in vitro when it binds single-stranded DNA and nucleoside triphosphate (10). Correspondingly, B. subtilis RecA is activated in vivo by binding single-stranded DNA exposed by discontinuous replication past UV-induced lesions and to a lesser extent by the processing of gaps formed during excision repair (13). Although the SOS system is induced during competence development by a similar RecA/LexA-dependent mechanism, B. subtilis recA expression is additionally stimulated by a competence-specific mechanism (8,14). Competence is a starvationinduced differentiation process that develops optimally at sufficiently high cell densities and in minimal growth medium with glucose as the main carbon source (for review, see Ref. 15). The various environmental signals are interpreted by a complex signal transduction cascade and ultimately lead to the activation of comK, which encodes the competence transcription factor (16,17). ComK is essential for: (i) the expression of all late competence gene products that assemble the DNAbinding and -uptake system; (ii) the competence-related expression of the recombination genes recA and addAB (the homologue of E. coli recBCD); and (iii) its own expression (8,16,18). Purified ComK has been shown to bind to the promoter regions of all these genes, and its transcription stimulating activity has been demonstrated in vitro with the late competence gene, comG (19). ComK footprinting analyses with a number of ComK-regulated genes established a conserved ATrich palindromic sequence (called the AT-box) as the ComKrecognition sequence (19).
Since competence-dependent recA induction occurs in RecAminus cells, deficient in LexA cleavage, and before LexA is cleaved in wild-type cells, 2 the LexA-imposed transcriptional repression of recA is presumably alleviated by the activity of ComK (14). To examine whether ComK is able to prevent the association of LexA with the recA promoter region, we analyzed the binding of purified ComK in the absence and presence of LexA. Hydroxyl radical footprinting analysis revealed four possible ComK dimer-binding sites, which was substantiated by the results of gel mobility shift experiments. Surprisingly, the occupation of all ComK-binding sites did not interfere with LexA binding, yet our in vitro transcription analysis showed that ComK is sufficient to overcome LexA repression.

EXPERIMENTAL PROCEDURES
General Methods and Materials-All molecular cloning and PCR 3 procedures were carried out using standard techniques (20,21). Labeled nucleotides were from Amersham Pharmacia Biotech. Media for growth of B. subtilis and E. coli have been described by Sambrook et al. (21) and Venema et al. (22). B. subtilis strain 8G5 chromosomal DNA used as template for PCR, was purified as described by Venema et al. (22).
Purification of ComK-ComK was purified as an MBP-ComK fusion protein on an amylose resin (New England Biolabs) column and separated from MBP by cleavage with protease Factor Xa, as previously described (16). After cleavage was complete, Factor Xa was inactivated by the addition of 1 mM phenylmethylsulfonyl fluoride. To separate ComK from MBP and DNA, the protein mixture was loaded onto a DEAE column (Amersham Pharmacia Biotech) equilibrated with 20 mM Tris-HCl, pH 8, 1 mM EDTA, and 0.5 mM dithiotreitol. MBP and ComK were sequentially eluted with a 0 to 50 mM Na 2 SO 4 gradient and a 0 to 1 M KCl gradient (containing 50 mM Na 2 SO 4 ). Fractions were collected and the Na 2 SO 4 concentration increased to 100 mM to prevent precipitation of ComK. The ComK containing fractions were checked for the absence of contaminating DNA by ethidium bromide-stained agarose gel electrophoresis, aliquoted, and stored at Ϫ70°C. Purification and cleavage of MBP-ComK were followed by SDS-polyacrylamide gel electrophoresis.
Purification of LexA-B. subtilis LexA was purified as described previously (10). E. coli strain BL21(DE3) containing pET21a-dinR was grown in 1 liter of LB broth containing carbenicillin (50 g/ml) with shaking until an A 600 of 0.6. The culture was induced with 10 ml of 100 mM isopropyl-1-thio-␤-D-galactopyranoside and grown for an additional 3 h. Cells were harvested by centrifugation at 4°C, 5000 ϫ g for 20 min, and resuspended in 5 ml of 20 mM Tris, pH 7.5, 10% (w/v) sucrose, 1 mM EDTA. Cells were lysed with lysozyme (0.2 mg/ml) by incubation on ice for 30 min followed by a 15-min incubation at 37°C. After 3 times freeze-thawing and sonication, debris was removed by centrifugation at 18,000 ϫ g for 15 min, and the supernatant was used for further purification. The supernatant (0.5-2 ml) was filtered and applied to a 5-ml heparin-agarose column, equilibrated in 20 mM Tris, pH 7.5, 10 mM NaCl, and subjected to fast protein liquid chromatography. Following elution with a 40-ml linear NaCl gradient (10 mM to 1 M), fractions containing LexA protein were pooled, concentrated 10-fold by centrifugation in a Centricon 10 unit (Amicon), and diluted to the original volume in 20 mM Tris, pH 7.5, 10 mM NaCl; concentration and dilution was then repeated twice. Sample was applied to a Mono-S column and chromatographed as described for the heparin-agarose column, and fractions containing LexA were pooled and dialyzed against 20 mM sodium phosphate buffer, pH 7.5, 200 mM NaCl, and concentrated by centrifugation in a Centricon 10 unit.
Hydroxyl Radical Protection Analyses-Hydroxyl radical protection analyses were performed as described by Tullius and Dombroski (23) with the modifications described by O'Halloran et al. (24). The DNA probes were obtained by PCR amplification using primers R1 (5Ј-TACG-GCTGCCATTTAATG-3Ј) and R2 (5Ј-CTGCCTGACGATCACTC-3Ј), complementary to the sequence at nucleotide position Ϫ184 and ϩ32, relative to the transcriptional start of recA. Primers were end-labeled with T4-polynucleotide kinase using [␥-32 P]ATP. Binding reaction conditions were as for the gel mobility shift experiments described below, except that glycerol was omitted. DNA fragments, containing ϳ30,000 cpm, were added to each 50-l reaction mixture. After 20 min at room temperature, 1 l of 5.6 mM (NH 4 ) 2 Fe(SO 2 )2⅐6H 2 O was mixed with 1 l of 11.2 mM EDTA, and added to the incubation mixture, followed by the addition of 2 l of 3.36% H 2 O 2 and 2 l of 112 mM Na-ascorbate. Reactions were incubated for 1 min at room temperature and terminated with 44 l of stop solution (10 l of 3 M Na-acetate, 6 g of yeast tRNA, and 10 l of 320 mM thiourea) and the subsequent addition of 300 l of ethanol. Samples were extracted with phenol-chloroform, and ethanol precipitated. The precipitates were resuspended in 3 l of loading buffer. Analysis of DNA products was carried out by electrophoresis on a 6% polyacrylamide urea gel. Maxam-Gilbert G ϩ A reactions were run with each experiment to locate sequence positions and protected regions (21).
Gel Mobility Shift Analyses-Gel mobility shift experiments were carried out essentially as described (16). The recA promoter fragment was isolated by PCR, using the primer set as described for the hydroxyl radical protection analyses, and end-labeled with T4-polynucleotide kinase using [␥-32 P]ATP. The purified proteins and probes were premixed on ice in binding buffer (20 mM Tris-HCl, pH 8, 100 mM KCl, 5 mM MgCl 2 , 0.5 mM dithiotreitol, 10% (v/v) glycerol, 0.05 mg/ml poly(dI-dC), and 0.05 mg/ml bovine serum albumin). After 15 min incubation at 37°C, the samples were loaded on a nondenaturing 4% polyacrylamide gel. The following modifications were applied to increase the resolution of retardation. Electrophoresis was performed with the Bio-Rad Miniprotein cell system, using a spacer thickness of 0.75 mm. A voltage gradient stacking was established by using 1 ϫ TAE buffer (40 mM Tris acetate, pH 8, 2 mM EDTA) for the nondenaturing 4% polyacrylamide gel, 0.5 ϫ TAE buffer for the cathode compartment, and 2 ϫ TAE buffer for the anode compartment. Gels were run at 50 V, dried, and autoradiographed in the absence of intensifying screens.
Autoradiograms of the dried gels were digitized and analyzed by densitometry using an Alpha Innotech IS-1000 imaging system. To determine the fraction of DNA molecules with exactly i dimers bound, Ø i (i ϭ 0, 2, 3, or 4), the total OD of all bands in each lane was quantified. Ø i was calculated for each band from Ø i ϭ OD tot,i /⌺OD tot,i , summed over all of the bands in a given lane. The equations for the fractions of DNA molecules with i ligands bound in a three-site system have been derived previously using a general statistical mechanical approach (25). In our analyses we assumed that the three sites correspond to the binding of a tetramer first and then subsequently by two dimers such that the equations representing the number of dimers bound are given by, Experimental data shown in Fig. 7 were analyzed according to Equations 1-4 with Mathematica and KaleidaGraph using nonlinear least squares methods to determine and apply the best curve fits.
In Vitro Transcription Assays-In vitro transcription experiments were performed essentially as described by Hamoen et al. (19). The DNA template, pAN-recA, was constructed by cloning a PCR fragment containing the recA promoter, between the SmaI and XbaI sites of the pAN583 vector (26). The recA promoter fragment was isolated by PCR using the primers R1 and R3 (5Ј-TGTTCTAGAGCCATATCTAAGG-3Ј), and subsequent digestion with HindIII (position ϩ49 relative to the transcriptional start of recA) and XbaI (underlined in R3). The HindIII site was converted to a blunt-end by using a Klenow fill-in reaction, prior to XbaI digestion. The transcription reactions were performed in the binding buffer described for the gel mobility shift experiments (poly(dI-dC) included). DNA templates, purified B. subtilis RNA polymerase, and purified ComK were incubated for 15 min at 37°C in a final volume of 20 l, before the addition of 3 l of a nucleotide mixture (1 mM ATP, 1 mM UTP, 1 mM GTP, 0.5 l of [␣-32 P]CTP). After 1 min incubation at room temperature, 2 l of 0.3% heparin was added to the mixture and incubation resumed for another 10 min at 37°C. After the addition of 2 l of 1 mM CTP, incubation was continued for 10 min before terminating the reaction by the addition of 18 l of formamide containing 0.05% bromphenol blue and 0.05% xylene blue. After heating for 3 min at 90°C, the samples were loaded on a 8% polyacrylamideurea gel and run at 300 V. Gels were subjected to autoradiography immediately after electrophoresis without prior drying.

The recA Promoter Contains Four ComK Dimer-binding
Sites-In a previous study Haijema et al. (8) determined the ComK-binding site at the recA promoter using DNase I footprinting analysis (8). They obtained a clear footprint extending from approximately positions Ϫ150 to Ϫ50, which is the longest ComK-protected region identified thus far. By analyzing ComK footprints of several ComK-dependent promoters, Hamoen et al. (19) concluded that ComK binds as a tetramer composed of two dimers, where each dimer recognizes the dyad symmetrical sequence AAAAN 5 TTTT, the so called AT-box (19). The two putative AT-boxes in the recA promoter appeared to be located at the center of the DNase I protections, yet the region confined by these AT-boxes covers only half the sequence protected by ComK (19).
To examine which nucleotide sequences could be responsible for this peculiarly extended ComK-binding region, we improved the resolution of the ComK footprint analyses by using a hydroxyl radical protection assay (Fig. 1). As shown in Fig. 2, the hydroxyl radical footprint fits well within the borders of the DNase I footprint. The central hydroxyl radical protections mark the two previously assigned AT-boxes in a pattern comparable with that found in a hydroxyl radical ComK footprint of the addAB promoter (19). A closer inspection of the hydroxyl radical protections at the extremities of the ComK-binding region revealed two additional AT-boxes: a perfect AT-box around position Ϫ140, and one AT-box, with two replacements in the thymine tract, around position Ϫ55.
The presence of four ComK-protected AT-boxes in the recA promoter suggests that this promoter is able to accommodate a total of four ComK dimers. Since it has been shown that ComK binds as a single tetramer to the promoter regions of the ComKregulated genes studied so far (19), we were interested in determining the oligomerization state of ComK on the recA promoter with its two additional dimer-binding sites. Specifically, we wondered if the detection of partial saturation of the recA promoter at low ComK concentrations might provide clues regarding the nature of ComK binding. Although previous gel mobility shift assays with the recA promoter and purified ComK yielded a single retarded band, the retarded band displayed in the autoradiograms was diffuse and may have obscured the presence of multiple bands (8).
To increase the resolution of the gel mobility shifts we omitted the use of intensifying screens, reduced the thickness of the nondenaturing polyacrylamide gels, and applied a voltage gradient using buffers with various ionic strengths (for details, see "Experimental Procedures"). These alterations resolved the single diffuse band into three separate bands (Fig. 3). Repeating these high resolution gel mobility shift assays with other ComK-regulated promoters, such as addAB and comK, yielded only a single retarded band (data not shown) consistent with the binding of a single ComK tetramer (19). Gel mobility shift experiments at low ComK concentrations with these ComKregulated promoter fragments never revealed intermediate retarded bands that would correspond to the binding of discrete ComK dimers. Based on its relative mobility, we assume that the least retarded recA fragment is bound by a single ComK tetramer, and that the two further retarded bands are due to successive binding of two additional ComK dimers.
ComK Binding Does Not Interfere with LexA Binding-The LexA-binding site of the recA promoter partially overlaps with the downstream AT-box (AT-box 4) suggesting that ComK binding to this AT-box could reduce the affinity of LexA for the recA SOS-box. Thus, a plausible mechanism for competencedependent recA induction could be that ComK precludes stable binding of LexA to its site. Arguing against this simple mechanism, Haijema et al. (8) showed that both proteins can bind simultaneously to the recA promoter. However, it is possible that in the presence of LexA the downstream AT-box was not occupied (i.e. only three ComK dimers bound) since the unresolved mobility shift experiment could not distinguish between partial and complete occupancy of al ComK-binding sites.
To examine whether ComK is capable of displacing LexA from the recA promoter, we repeated the high-resolution gel mobility shift assays with ComK in the presence of purified LexA. Prior to testing retardation with both proteins, we tested the binding of LexA alone to determine the saturating LexA concentration. The recA promoter contains a single SOS-box, and only a single shifted band was observed in a gel mobility shift assay (Fig. 4). Graphical analysis of these data gave a K d of 5 nM for LexA binding to the recA SOS-box (data not shown). Using a LexA concentration (13 nM) sufficient to ensure binding to all recA promoter molecules, we incubated the recA promoter with increasing concentrations of ComK in the absence and presence of LexA. As indicated in Fig. 5, the presence of LexA added to the electrophoretic mobility shift of all three ComK retarded bands. Apparently, despite partially overlapping recognition sites, ComK does not exclude LexA from binding to the recA promoter. The alternative possibility, that LexA-induced supershift resulted from a specific interaction between ComK and LexA, could be refuted. In a gel mobility shift assay with a ComK-dependent promoter which does not contain a SOS-box, the presence of LexA did not result in an additional retardation in electrophoretic mobility of ComK-bound promoter fragments (data not shown).
LexA Does Not Affect the Affinity of ComK for the recA Promoter-We quantified our mobility shift data by densitometry and analyzed the data graphically to determine equilibrium binding constants. Our objective was to assess the extent to which LexA alters the ability of ComK to bind the recA promoter. Any changes in ComK binding constants at saturating LexA concentration would suggest that saturating concentrations of ComK could alter the binding of LexA to the recA SOS-box. Fig. 6 shows graphical analyses of our ComK titration data in the absence and presence of LexA at a saturating concentration of 13 nM. Assuming the stoichiometry indicated by our footprinting data (i.e. a maximum occupancy of four ComK dimers per recA promoter), we analyzed the average number of ComK dimers bound per recA promoter as a function of free ComK concentration. The data for the binding of ComK in the presence and absence of LexA fit a binding isotherm with a dissociation constant, K d , of 84 nM. The value of 84 nM was determined from a Scatchard plot of the same data (Fig. 6, middle graph). It is noteworthy that the Scatchard plot is linear suggesting that the thermodynamics of binding is not cooperative despite the existence of multiple contiguous binding sites. Consistent with the linear Scatchard plots, Hill plots of the data (Fig. 6, lower graph) had slopes of exactly 1.0. These results do not rule out the possibility that the kinetics of ComK binding is cooperative, which we suspect is the case for binding of the ComK tetramer since we never detect binding of a single dimer. It is also possible that the kinetics of dimer binding is cooperative. In any case, the significant result is that the presence of LexA does not appear to affect the affinity of ComK for the recA promoter.
Our ability to resolve different ligation states for ComK binding makes it possible to analyze individual binding constants to better assess cooperative interactions that might occur between sites. Senear and Brenowitz (25) demonstrated the validity of this approach for the Lac, Gal, and cI repressors (25). As described above, our gel mobility shift results can best be explained by a three-site system where one site is bound by a ComK tetramer and the other two sites are bound by ComK dimers. The fraction of DNA molecules with exactly i dimers bound, Ø i (where i ϭ 0, 2, 3, or 4), was calculated from digitized autoradiogram images and the data were curve fitted as described under "Experimental Procedures." Fig. 7 shows a plot of the fraction of DNA with ComK bound to 0, 1, 2, or 3 sites, the first site bound (i.e. the site bound at low ComK concentration) by a tetramer, and the other sites by dimers. The curves in Fig.  7 result from fitting the data from gel mobility shift assays, in the presence and absence of LexA, according to Equations 1-4 to give K 1 ϭ 1.21 ϫ 10 7 , K 2 ϭ 8.94 ϫ 10 13 , and K 3 ϭ 7.25 ϫ 10 20 for the respective equilibrium association constants. It is noteworthy that the dissociation constant corresponding to K 1 ϭ 1.21 ϫ 10 7 (i.e. the reciprocal of this value) is 83 nM, is in very good agreement with the dissociation constant determined from the Scatchard plot in Fig. 6. Although the curve fit is not perfect for our 1-and 2-site data, there is no significant difference between the data in the presence or absence of LexA.
As described for the binding of the cI repressor to its operator, cooperative binding of the second ligand in a 3-site system can be inferred if K 2 Ͼ K 1 2 /3 and cooperative binding of the third ligand can be inferred if K 3 Ͼ K 1 K 2 /3 (25). For ComK binding to the recA promoter the ratio of K 2 :K 1 2 /3 is 1.8 and the ratio of K 3 :K 1 K 2 /3 is 2.0 suggesting moderate cooperativity in each case. Although the ComK-binding sites are not identical, Senear and Brenowitz (25) showed that the ratios are greater than one when cooperativity equals or exceeds site heterogeneity. A possible explanation for the apparent discrepancy between these results and our Hill analysis is that the cooperativity is about equal to the site heterogeneity. That is, the free energy corresponding to cooperative interactions between the tetramer and the adjacent dimers is about equal to the respective differences in the free energy of binding.
Figs. 6 and 7 provide strong evidence that the binding constants for ComK are unaffected by the presence of LexA. According to equilibrium laws, it follows that if the binding constants for ComK are unaffected by the presence of LexA, the binding constant for LexA is unaffected by the presence of ComK. Thus, binding of ComK alone does not contribute to the displacement of LexA from the recA promoter.
ComK Is Sufficient to Activate recA Transcription-Since ComK does not prevent LexA from binding to the recA promoter, is ComK alone capable of counteracting the LexA-imposed repression or are additional factors required? We performed in vitro transcription experiments with purified RNA polymerase to address this question. The recA promoter region was cloned into the multiple cloning site of pAN583, followed by the strong T7 terminator for use in an in vitro transcription assay (26). The resulting pAN-recA construct was incubated in a transcription buffer containing purified RNA polymerase and combinations of purified LexA and ComK. Fig. 8 shows that the inhibition of recA transcription by LexA is overridden by the addition of ComK, although the level of transcription is not as high as with ComK alone. We therefore conclude that the presence of ComK is sufficient for competence induced recA expression.

DISCUSSION
Using hydroxyl radical protection analysis we identified a total of four ComK dimer-binding sites, two more than are present in other ComK-regulated promoters. Our mobility shift results further suggest that the recA promoter is bound by a ComK tetramer at low ComK concentration, and then successively by two additional ComK dimers as ComK concentration increases. Complete retardation of most recA promoter molecules does not occur until the ComK concentration reaches about 1 M. Assuming a cell volume of 1 ϫ 10 Ϫ9 l (calculated from the size of a typical cell), and about 90,000 ComK molecules per cell (27), the physiological concentration of ComK would be well over 100 M, high enough to ensure that all four ComK-binding sites are occupied in vivo.
On the basis of the separation of the ComK dimer-binding sites (AT-boxes), three classes of ComK-regulated promoters are distinguished (19). In the first class (addAB, dinA, and nucA) the interval between dimer-binding sites is 21 nucleotides, placing both AT-boxes at the same side of the DNA helix, assuming 10.5 nucleotides per helical turn (28). The second class comprises the late competence genes (comC, -G, -E, and -F). These promoters contain AT-boxes separated by an interval of 31 nucleotides, corresponding to 3 complete helical turns. The third class contains a single representative, the comK promoter. In this case the repetition of the two AT-boxes occurs at an interval of 44 nucleotides, ϳ4 helical turns. Although the distances between the ComK dimer-binding sites show a remarkable variability, in all classes the two AT-boxes are located at the same face of the DNA helix. This enables the ComK dimers to interact to form a tetramer, which has been shown to be essential for efficient binding of ComK (19). In the recA promoter the two centrally located AT-boxes (AT-boxes 2 and 3) are separated by an interval of 21 nucleotides, corresponding to the first class of ComK-regulated promoters. The interval between AT-box 1 and AT-box 2 is 33 nucleotides, equivalent to the second class of promoters. The same interval separates AT-box 3 and AT-box 4. Thus, all of the AT-boxes face the same side of the DNA helix suggesting that the ComK dimers could interact with the ComK tetramer. In support of such interactions, our analysis of the individual binding constants for the three distinct ligation states revealed cooperativity for binding of the dimers. In order for class 2 and class 3 promoters to form tetramers, the two ComK dimers must span a considerable distance suggesting significant distortion of the DNA. In fact, ComK has been shown to induce bending, estimated at about 75 degrees, in the promoters of comG and comF (19). It is therefore likely that binding of the 4 ComK dimers would induce a substantial bending of the recA promoter.
We have provided strong evidence that ComK alone cannot displace the LexA repressor from the recA operator. The SOSbox partially overlaps with the downstream located AT-box, yet when all ComK-binding sites are sequestered, our results indicate that LexA still binds to the recA promoter with equal affinity. Thus, it is reasonable to assume that ComK and LexA neither interact nor interfere sterically with each other and that ComK-induced bending does not affect the LexA-binding site. According to the hydroxyl radical protection pattern, the protected bases in the upper and lower strand nearest to the centers of the AT-box dyad symmetries are offset toward the 3Ј direction, relative to the AT-box centers, which is indicative of two ComK monomers contacting each other across the major groove of DNA (23). This implies that ComK recognizes the AT-box consensus sequence in the minor groove as previously documented for the addAB promoter (19). In contrast, hydroxyl radical protection studies showed that the B. subtilis LexA protein recognizes the SOS consensus sequence in the major groove (10,29). Comparison of the protection patterns of LexA (29) and ComK on the recA promoter suggests that the center of LexA binding at the operator is rotationally separated from ComK, bound at AT-box 4, by about 120 degrees relative to the helical axis. Thus, it is plausible that ComK does not interfere with LexA binding.
Miller et al. (10) have shown that binding of B. subtilis LexA to the recA promoter inhibits both binding of RNA polymerase and transcription in vitro. The location of the recA SOS-box suggests that the nature of LexA repression is unusual among the known din genes. Unlike many of the B. subtilis SOS-boxes, which are centered between position Ϫ10 and position Ϫ35, the SOS-box of recA is centered at position Ϫ52. This region corresponds to the distal subsite of E. coli UP elements thought to interact with the C-terminal domain of an ␣-subunit (␣CTD) of RNA polymerase (30). Although not as extensively characterized as those in E. coli, B. subtilis UP elements are required for high level expression of certain genes and interact with the B. subtilis ␣CTD (31-33). Indeed, the sequence of the recA promoter between position Ϫ46 and position Ϫ59 is AT-rich and resembles the E. coli UP element consensus sequence. We suppose that high-level transcription of the recA gene requires the interaction of the ␣CTD with the SOS-box region, and that LexA binding blocks this interaction.
Our results suggest that ComK overrides LexA repression without directly displacing LexA. Thus, ComK either enables RNA polymerase to displace LexA and/or provides an alternative binding interaction such that RNA polymerase can bind the recA promoter in the presence of LexA. In either case, we propose that the bending of DNA by ComK is an essential feature. DNA bending is commonly observed with transcriptional activators and is thought to facilitate wrapping of DNA around RNA polymerase (34). For the recA promoter, binding of ComK to all four AT-boxes may provide the appropriate bending for such wrapping. In support of a requirement for ComK binding at the most upstream site, which would contribute to such bending, Cheo et al. (35) have shown that deleting the region upstream of position Ϫ120, containing AT-box 1, causes a 70% reduction in competence-related recA expression. In addition to DNA wrapping, binding of RNA polymerase to the recA promoter could also be facilitated by direct interactions between ComK and ␣CTD. Considering evidence for the interactions of E. coli ␣CTDs with the proximal UP element subsite, centered at Ϫ42, and the distal UP element subsite, centered at Ϫ52 (30), it seems significant that the downstream AT-boxes of the comC, comE, comF, and comG promoters are all centered at about position Ϫ48, whereas the downstream AT-boxes of the recA and addAB promoters are centered at about position Ϫ58 (19). The precise location of all these downstream AT-boxes relative to the ␣CTD-binding sites suggests a mechanism in which ComK affects the binding of ␣CTD. In the case of the recA promoter, ComK may provide an alternative binding site for the ␣CTD of RNA polymerase when the UP element is masked by LexA. On the other hand, ComK may affect ␣CTD binding indirectly by inducing a conformation change in the DNA. Indeed, recent biochemical evidence suggests that some class I transcription factors, thought to interact with ␣CTD, may stimulate the DNA binding activity of ␣CTD by inducing a conformation change in the DNA (36). Further research will be necessary to distinguish between these possibilities.