LexA Represses CTXΦ Transcription by Blocking Access of the α C-terminal Domain of RNA Polymerase to Promoter DNA*

CTXΦ is a Vibrio cholerae-specific temperate filamentous phage that encodes cholera toxin. CTXΦ lysogens can be induced with DNA damage-inducing agents such as UV light, leading to the release of CTXΦ virions and the rapid dissemination of cholera toxin genes to new V. cholerae hosts. This environmental regulation is directly mediated by LexA, the host-encoded global SOS transcription factor. LexA and a phage-encoded repressor, RstR, both repress transcription from PrstA, the primary CTXΦ promoter. Because the LexA binding site is located upstream of the core PrstA promoter and overlaps with A-tract sequences, we speculated that LexA represses PrstA by occluding a promoter UP element, a binding site for the C-terminal domain of the α subunit of RNA polymerase (RNAP) (αCTD). Using in vitro transcription assays, we have shown that the LexA binding site stimulates maximal rstA transcription in the absence of any added factors. The αCTD of RNAP is required for this stimulation, demonstrating that the LexA site contains, or overlaps with, a promoter UP element. LexA represses rstA transcription by normal RNAP but fails to repress rstA transcription catalyzed by RNAP lacking the αCTD. DNase I footprint analysis mapped the αCTD binding site to the upstream promoter region that includes the LexA binding site. The addition of free α subunits blocked the binding of LexA to rstA promoter DNA, indicating that LexA and the αCTD directly compete for binding to their respective sites. To our knowledge, this is the first report of a repressor blocking transcription initiation by occluding a promoter UP element.

CTX⌽, a lysogenic filamentous phage, has played a critical role in the evolution of toxigenic Vibrio cholerae, the causative agent of the diarrheal disease cholera. The ϳ6.9-kb CTX⌽ genome encodes cholera toxin, the principal virulence factor of this Gram-negative enteric pathogen. Following infection of V. cholerae, the CTX⌽ genome integrates in a site-specific fashion near the terminus of chromosome I (1-3), generating a CTX prophage. Even though CTX⌽ virions are secreted from V. cholerae without cell lysis, the expression of prophage genes required for CTX⌽ virion production is ordinarily repressed, as is also the case for temperate phages that lyse their respective hosts upon prophage induction. Most knowledge of prophage induction has been obtained from studies of phage and some closely related temperate phages that are unrelated to CTX⌽. We are studying the environmental conditions and molecular processes that control CTX⌽ virion production to expand basic understanding of cellular processes that govern phage development and to gain insight into factors that may contribute to the emergence of new pathogenic V. cholerae strains.
Previous studies suggest that the expression of the genes required for CTX⌽ production initiates from a single promoter, P rstA . This promoter is located in ig-2, an intergenic region that separates rstA (a gene required for CTX⌽ replication) from rstR, the gene encoding the phage repressor (see Fig.  1A). RstR binds to three operators within ig-2 and represses transcription from P rstA (4,5). In addition to RstR, we recently found that LexA, a host repressor that regulates the SOS regulon, also represses P rstA (6). LexA represses P rstA by binding to an A ϩ T-rich site positioned at Ϫ41 to Ϫ56 from the start of rstA transcription (Fig. 1A) (6). This 16-bp sequence is nearly identical to the consensus Escherichia coli LexA binding site (7).
The V. cholerae SOS response to DNA damage activates CTX⌽ production (6). DNA-damaging agents, such as mitomycin C or UV light, increase transcription from P rstA and production of CTX⌽ virions from CTX⌽ lysogens in a recA-dependent fashion. No stimulation of transcription from P rstA or CTX production was observed following UV treatment of CTX⌽ lysogens that contained a noncleavable LexA (6). We proposed that the activated form of RecA generated by DNA damage provokes the auto-cleavage of LexA, thereby alleviating its repression of P rstA . RstR levels also decline after treatment of CTX⌽ lysogens with DNA damaging agents. However, unlike the repressor CI, RstR does not appear to undergo RecAstimulated auto-cleavage, and the mechanism that accounts for the decrease in RstR levels in UV-treated CTX⌽ lysogens is unknown.
Maximal expression of rstA requires sequences upstream of the Ϫ10 and Ϫ35 binding sites for the subunit of RNA polymerase (RNAP) 2 in P rstA (6). We hypothesized that these sequences, which are rich in runs of A⅐T or T⅐A base pairs and overlap with the LexA binding site in P rstA , function as a promoter UP element. UP elements are A ϩ T-rich sequences found upstream of the Ϫ35 element in many highly active promoters, where they function to increase promoter strength by binding the C-terminal domain (␣CTD) of RpoA, the ␣ subunit of RNAP (8 -11). UP elements have also been characterized for the rRNA promoters of a related bacterium, Vibrio natriegens (12). Consistent with the idea that P rstA includes a UP element, we previously reported a reduction in the expression of an rstA::lacZ reporter in E. coli producing a mutant RpoA that does not bind DNA (RpoA R265A) (6). Here, we have used in vitro biochemical assays to establish that the LexA binding site in P rstA overlaps with a UP element. Furthermore, we have provided evidence that LexA represses transcription from P rstA by occluding access of the ␣CTD to promoter DNA.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-Strains RLG3538 and RLG3545 carry pT7H6-␣ and pT7H6 R265A, the expression plasmids for the production of His 6 -tagged RpoA and His 6tagged R265A RpoA, respectively (13). E. coli strain BL21(DE3) was used as the host for protein expression. Plasmid p770ig2, the template for in vitro transcription experiments, is a derivative of pRLG770 (14); a PCR-amplified fragment of the CTX ig-2 region, from positions Ϫ160 to ϩ92, was cloned into the EcoRI and HindIII sites of pRLG770 upstream of the rrnB T1 terminator. Plasmid p770SUB is identical to p770ig2, with the exception that positions Ϫ58 to Ϫ40 (GGCTGTTTTTTTG-TACATT) were substituted with the sequence SUB (GACTG-CAGTGGTACCTAGG), which does not function as a UP element (8,15). Plasmid pRLG593 carries the lacUV5 promoter cloned into the same plasmid, pRLG770 (14).
In Vitro Transcription-Briefly, 25-l reaction mixtures contained 0.5 nM supercoiled plasmid DNA in 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 mM MgCl 2 , 1.0 mM dithiothreitol, 100 g/ml bovine serum albumin, 500 M ATP, 100 M CTP and GTP, and 10 M UTP with [␣-32 P]UTP (PerkinElmer Life Sciences) at a specific activity of ϳ30 Ci/mmol. The control plasmid pRLG593 was added to each reaction at a concentration of 0.5 nM. RNAP was added to initiate transcription, and the reaction proceeded for 15 min at room temperature (25°C). In experiments with added repressors, template DNA was preincubated at room temperature for 15 min with LexA and/or RstR in reaction buffer prior to the addition of RNAP. With the exception of experiments utilizing RstR or LexA, reactions were terminated with the addition of an equal volume of stop solution (7 M urea, 10 mM EDTA, 1% SDS, 2ϫ Tris borate-EDTA, 0.5% bromphenol blue, 0.025% xylene cyanol). In experiments that analyze RstR and LexA repression, reactions were terminated by ethanol precipitation. Pellets were air-dried and resuspended in a formamide loading buffer. Samples were heated to 90°C and electrophoresed on 6% urea-denaturing gels in 1ϫ Tris borate-EDTA. Radiolabeled Century RNA markers (Ambion) were used for the estimation of transcript size. Gel images were collected using the PhosphorImaging system (Amersham Biosciences). Quantitation was carried out using ImageQuant TL software (Amersham Biosciences).
DNase I Footprints-DNA probes were 5Ј end-labeled on the rstA nontemplate strand by PCR with one primer labeled at the 5Ј end with T4 polynucleotide kinase (New England Biolabs). PCR products were purified on 6% native polyacrylamide gels. DNA was eluted from crushed gel slices by diffusion and collected by ethanol precipitation. Wild-type RNAP, His-␣, H265R-␣, and/or LexA binding reactions (25 l) contained ϳ250,000 counts/min labeled DNA fragment in 25 mM Tris-HCl, pH 7.5, 20 mM KCl, 25 mM NaCl, 5 mM MgCl 2 , 25 ng/l bovine serum albumin, 1 mM dithiothreitol, and 5% glycerol. RNAP ␣⌬235 binding reactions contain 23 mM Tris-HCl, pH 7.5, 32.5 mM KCl, 5.6 mM MgCl 2 , 0.05 mM EDTA, 25 ng/l bovine serum albumin, 1 mM dithiothreitol, and 5% glycerol. The reactions were incubated for 30 min at room temperature. To compete nonspecific DNA binding by RNAP, heparin sulfate (10 g/ml) was added for 1 min. 0.4 units of DNase I (Ambion) was then added, and incubation was continued for one additional minute. Reactions were terminated with an SDS/EDTA stop solution, and nucleic acids were collected by ethanol precipitation. Dried DNA pellets were counted directly in a scintillation counter and resuspended in loading buffer (99% formamide, 0.1% 1 N NaOH, 0.01% xylene cyanol, and 0.01% bromphenol blue) to yield 20,000 cycles/min/l. G ϩ A sequencing reactions were carried out as previously described (16). Samples were heated to 94°C for 1-2 min prior to loading on prerun 8% sequencing gels. Autoradiography was performed with the PhosphorImaging system (Amersham Biosciences). Quantitation of lane profiles was carried out with ImageQuant TL software (Amersham Biosciences BioSciences) without background correction or data normalization.

RESULTS
The LexA Binding Site in P rstA Is Required for Maximal Transcription-To explore the possibility that the LexA binding site in P rstA overlaps with or contains a UP element, we replaced the LexA operator region (positions Ϫ40 to Ϫ58) ( Fig.  1A) with SUB, a 19-bp sequence that was previously shown not to function as a UP element (8,15). The influence of this substitution on P rstA activity was assessed by in vitro transcription assays using purified E. coli RNA polymerase. The use of E. coli RNAP is justified by our previous findings that 1) The rstA promoter is efficiently transcribed in both E. coli and V. cholerae (5, 6), 2) LexA repression of P rstA transcription is observed in both V. cholerae and E. coli (6), and 3) the ␣CTD of E. coli and V. cholerae RNAP (including all of the determinants important for DNA binding) are highly conserved (12). A plasmid containing the lacUV5 promoter that lacks a UP element (14) was included in the same reaction as a control. The quantity of rstA transcribed in vitro was reduced ϳ4-fold when the DNA template carried the SUB sequence in place of the normal DNA sequence upstream of P rstA (Fig. 1B, lanes 1 and 2). These observations indicate that the LexA binding site in P rstA includes sequences that stimulate transcription in the absence of additional factors, consistent with the presence of a UP element in the substituted region (Ϫ40 to Ϫ58).
To further investigate whether the LexA binding site includes a UP element, we performed in vitro transcription experiments using RNAP reconstituted with ␣⌬235, an ␣ subunit that is deleted for the ␣CTD that directly contacts UP element DNA (18 -20). In contrast to WT RNAP, the relative level of rstA transcripts generated by ␣⌬235 RNAP was reduced ϳ4-fold (Fig. 1B, compare lanes 1 and 5). Furthermore, in reactions with ␣⌬235 RNAP, the relative level of rstA transcripts was not reduced when the LexA binding site was substituted by SUB in the template DNA (Fig. 1B, lanes 5-8). Thus, maximal transcription from P rstA requires both the upstream LexA binding site and the intact ␣CTD of RNAP. Taken together, these observations suggest that P rstA contains a UP element within or overlapping the LexA binding site.
RNAP Binding to P rstA -DNase I protection assays were performed to physically define the sites in ig-2 bound by RNAP. Because specific binding of RNAP to promoter sequences is often obscured by nonspecific DNA binding, RNAP-DNA complexes were treated with heparin sulfate, a polyanion that irreversibly binds free RNAP, prior to DNase I treatment. Protection of the P rstA promoter by WT RNAP was observed from ϳϪ60 to ϳϩ25, including the core promoter elements (Ϫ10, Ϫ35, and transcription start site ϩ1) and the region upstream of the core promoter elements (from ϳϪ45 to ϳϪ60; Fig. 2, see arrows). Similar to WT RNAP, ␣⌬235 RNAP protected the core promoter elements (although higher concentrations were required). However, ␣⌬235 RNAP did not protect sequences upstream of the Ϫ35 element (Fig. 2). Thus, the ␣CTD of RNAP is responsible for the upstream footprint. This region of the promoter overlaps the LexA binding site (Ϫ41 to Ϫ56) ( Fig. 2A).
DNase I protection assays were also performed on a promoter construct containing the SUB sequence in place of the LexA box (SUB RstA). WT RNAP protected the core promoter regions but protected sequences upstream of the Ϫ35 element weakly or not at all (Fig. 2), consistent with the footprints observed at other promoters lacking UP elements (e.g. lacUV5 and P R ; (21,22)). Together, the data suggest that the upstream region of P rstA contains sequence-specific binding sites for the ␣CTDs.
To confirm that the protections observed upstream of the Ϫ35 element resulted from interactions with the RNAP ␣ subunit, as observed in previous studies with the rrnB P1 promoter (8), DNase I protection experiments were also carried out with purified E. coli ␣ subunits (Fig. 3). Free ␣ protected upstream sequences that include those protected by RNAP (Fig. 2), and protection was observed at the ␣ concentrations similar to those required for binding to other UP elements (2 M; Ref. 8). No protection was observed with ␣R265A ( Fig. 3, lanes 7 and 8), a mutant ␣ that does not bind DNA (8,20). The sequences protected by purified ␣ extended further upstream than those observed with RNAP holoenzyme (Fig. 2). Similar extended footprints have been observed previously with purified ␣ at other promoters (8,19). There is evidence in some promoters for binding of ␣CTD to DNA minor grooves upstream of the interactions detected in the rrnB P1 UP element (17,23). However, it is also possible that the extended footprint with free ␣ simply reflects nonphysiologically relevant interactions of additional ␣ subunits with additional A ϩ T-rich upstream sequences. Although small effects on the transcription of ␣ binding to rstA sequences upstream of Ϫ60 cannot be ruled out, the effects of the SUB mutation on P rstA transcription (Fig. 1B) suggests that the Ϫ41 to Ϫ58 region is sufficient to account for the major effects of ␣CTD binding on rstA transcription stimulation.
To assess whether LexA and ␣ compete for binding to the same site, both proteins were combined with the DNA template at concentrations sufficient to protect their respective binding sites (Fig. 3). Lane 9 shows the LexA footprint in the absence of Ϫ35 (TTGAAA) core promoter elements of P rstA . The LexA box (Ϫ41 to Ϫ56) is shown in uppercase type in the expanded sequence above. The Ϫ35 promoter sequence is underlined. The region upstream of P rstA substituted in SUB (Ϫ40 to Ϫ58) is marked above with asterisks. Shown below are the positions of the three RstR operators, O1, O2, and O3 (4). B, P rstA promoter activity analyzed by in vitro transcription assay. The supercoiled DNA templates for in vitro transcription experiments were p770ig2 (W), containing the wild type CTX ig-2 region (Ϫ160 to ϩ92) or p770SUB (S) carrying the same ig-2 region with the SUB replacement. Plasmid pRLG593, which carries the lacUV5 promoter lacking a UP element, served as a control template. Transcription from the rstA and lacUV5 promoters terminated at an rrnBT1 terminator present in the plasmid vector (15). WT and ␣⌬235 RNAP were reconstituted as described under "Experimental Procedures." RNAI is a small RNA transcribed from the ori region of the plasmid vector.
␣. As the ␣ concentration was increased (Fig 3, lanes 11 and 12), protection of the LexA site was reduced, indicating that ␣ inhibits LexA binding. In regions distal to the position Ϫ64, where LexA binding does not obscure the ␣ footprint, ␣ binding was also reduced when both proteins were added to the DNA. The simplest interpretation of these footprint data is that the two proteins directly compete for binding to overlapping DNA sites (see also "Discussion"). Such mutual inhibition likely occurs by a direct steric mechanism.
The ␣CTD Is Required for LexA Repression of P rstA in Vitro-Our discovery of a UP element at P rstA plus our finding that free ␣ and LexA compete for binding to upstream DNA, suggests that the mechanism of LexA repression is to physically occlude ␣CTD from its binding site upstream of P rstA . To further test this model, we investigated whether LexA could repress P rstA transcription catalyzed by reconstituted ␣⌬235 RNAP, which lacks the ␣CTD. The plasmid templates for these experiments were p770ig-2 (Ϫ160 to ϩ92) and pRLG593 carrying the lacUV5 promoter as a control. LexA specifically reduced P rstA transcription catalyzed by WT RNAP (Fig. 4, lanes 1-4). At high LexA concentrations, there was an ϳ5-fold reduction in the level of rstA transcripts, whereas it had little or no effect on transcription from the control lacUV5 template. This partial repression by LexA in vitro is similar to previous in vivo meas-urements of LexA repression (6) and is strikingly similar to the 4-fold enhancement of rstA transcription provided by the UP element (Fig. 1B). RstR was a more potent repressor of rstA transcription than LexA, reducing transcription to nearly undetectable levels (Fig. 4, lane 5). In contrast to WT RNAP, rstA transcription by ␣⌬235 RNAP was not repressed by LexA (Fig. 4, lanes 6 -9). Identical results were obtained when the concentration of ␣⌬235 RNAP was lowered from 40 nM to 20 nM (data not shown). These data indicate that LexA repression of P rstA transcription results from inhibition of ␣CTD binding to the rstA UP element.
RstR was still a potent repressor of rstA transcription catalyzed by ␣⌬235 RNAP (Fig. 4, lane 10), indicating that RstR functions by blocking RNAP binding to the core promoter elements or represses a later step in transcription initiation. These observations demonstrate that LexA and RstR independently repress P rstA in the absence of other cellular factors and that LexA-mediated repression of transcription from P rstA requires the C-terminal domain of ␣.

DISCUSSION
Our previous studies of the regulation of P rstA revealed that a host repressor, LexA, and the CTX⌽-encoded repressor RstR both repress transcription from P rstA . Using in vitro transcription assays, we showed here that each of these repressors acts directly on P rstA to inhibit transcription. The LexA binding site in ig-2 is centered Ϫ48.5 bp from the start of rstA transcription; given its location and high A ϩ T content, we speculated that it overlaps with a binding site for the C-terminal domain (␣CTD) of the ␣ subunit of RNAP. Such ␣ binding sites in other highly active promoters are known as UP elements (8,11). Our experimental observations support this hypothesis. High level transcription from P rstA was dependent on specific sequences between positions Ϫ40 and Ϫ58 overlapping the LexA binding site. Replacement of the LexA box in this region with a sequence (SUB) that does not bind to the ␣CTD significantly reduced transcription from P rstA . The stimulatory effect of this promoter site on P rstA transcription was dependent upon the ␣CTD of RNAP; in vitro transcription with reconstituted RNAP deleted for the ␣CTD yielded a pronounced reduction in the level of transcripts from wild-type P rstA but did not reduce transcript levels from a mutant P rstA containing the SUB sequence in place of the LexA box.
In DNase I protection experiments, RNA polymerase holoenzyme protected several sites clustered around positions Ϫ46 and Ϫ58 that likely represent interactions with the ␣CTD (Fig. 2). ␣ subunit interactions in the context of the RNAP holoenzyme have been observed at a series of positions in the minor groove upstream of the Ϫ35 hexamer (23)(24)(25). Typically, the interactions most significant for function are just upstream of the Ϫ35 hexamer. In the E. coli rrnB P1 UP element, these sites are centered at Ϫ41 and Ϫ52 and are referred to as the proximal and distal subsites, respectively (10,26). Because the SUB mutation, which extends from ϳϪ41 to Ϫ58 in the rstA promoter, eliminated effects of upstream sequences on transcription, it is likely that this region contains the ␣ binding sites corresponding to the proximal and distal subsites. Determining the precise limits of the ␣-DNA interactions in the rstA promoter region will require the use of other reagents (e.g. hydroxyl radical footprinting).
LexA and free ␣ competed for binding to this region. Our findings strongly suggest that LexA represses transcription from P rstA by blocking the binding of the ␣CTD to the promoter UP element. It is likely that ␣CTD binds to successive minor grooves in the P rstA UP element (e.g. from ϳϪ40 to ϳϪ60), similar to its position in other characterized UP elements (8,23,26). The LexA dimer interacts with two successive major grooves using a winged helix-turn-helix DNA binding motif (27). With its binding site centered at position Ϫ48.5, it is likely that LexA would occupy the same face-of-the-helix as the ␣CTD and would overlap with and/or occlude DNA backbone positions along the minor groove that would be required for ␣CTD binding.  Transcription was initiated by the addition of RNAP and nucleotide triphosphates, and the reactions were incubated at 37°C for an additional 15 min. The supercoiled DNA templates were p770ig-2 (carrying the wild-type ig-2 region (Ϫ160 to ϩ92)) and pRLG593 (carrying the lacUV5 promoter). Lanes 1-5, reactions were performed with 10 nM reconstituted wild-type RNAP. Although UP elements sometimes overlap with binding sites for other DNA binding proteins (28) and it has been proposed previously that repressors could act by blocking binding of the ␣ subunits of RNAP to a UP element (26,29), to our knowledge our findings represent the first direct demonstration of this type of transcription regulation. This mechanism could, at least in part, explain SOS control of some other promoters. For example, the ssb gene in E. coli has a LexA box centered at position Ϫ46.5 (30), where LexA binding may also block access of the ␣CTD to a promoter UP element. A rationale for LexAmediated repression of high level transcription of SOS genes could be to enable transient bursts of expression of DNA repair genes whose function is important for the amelioration of cellular stress. LexA cleavage during an SOS response would allow for transient high level expression (promoted by ␣CTD binding to the unoccupied LexA box) of SOS genes; high level expression would be relatively short-lived because resynthesis of LexA after repair of DNA damage would restore expression of these genes to their basal state.
The ␣CTD is a common target for regulatory factors in bacteria. Usually these transcription factor-␣CTD interactions are positive, leading to increased transcription (9), but in some cases, these interactions result in transcription repression. Examples include IclR of E. coli (which has been proposed to repress the aceB promoter by relocation of the ␣CTD to an upstream position less favorable for stimulating transcription (31)), GalR (where ␣CTD-GalR interactions inhibit a step in transcription after closed complex formation (32,33)), and Spx of Bacillus subtilis (which has been described as an "anti-␣" factor that binds to the ␣CTD and inhibits transcription initiation (34,35)). Furthermore, phage T4 encodes factors that ADP-ribosylate the residue in ␣ that is most critical for UP element binding, resulting in inhibition of host transcription and thus an increase in T4 early transcription (36). The data presented here indicate that LexA, the global SOS repressor, can inhibit transcription initiation in yet another manner, by selectively competing for ␣ subunit binding to a UP element in the primary promoter of CTX⌽.