MarA-mediated Transcriptional Repression of the rob Promoter*

The Escherichia coli transcriptional regulator MarA affects functions that include antibiotic resistance, persistence, and survival. MarA functions as an activator or repressor of transcription utilizing similar degenerate DNA sequences (marboxes) with three different binding site configurations with respect to the RNA polymerase-binding sites. We demonstrate that MarA down-regulates rob transcripts both in vivo and in vitro via a MarA-binding site within the rob promoter that is positioned between the –10 and –35 hexamers. As for the hdeA and purA promoters, which are repressed by MarA, the rob marbox is also in the “backward” orientation. Protein-DNA interactions show that SoxS and Rob, like MarA, bind the same marbox in the rob promoter. Electrophoretic mobility shift analyses with a MarA-specific antibody demonstrate that MarA and RNA polymerase form a ternary complex with the rob promoter DNA. Transcription experiments in vitro and potassium permanganate footprinting analysis show that MarA affects the RNA polymerase-mediated closed to open complex formation at the rob promoter.

The Escherichia coli transcriptional regulator MarA affects functions that include antibiotic resistance, persistence, and survival. MarA functions as an activator or repressor of transcription utilizing similar degenerate DNA sequences (marboxes) with three different binding site configurations with respect to the RNA polymerase-binding sites. We demonstrate that MarA down-regulates rob transcripts both in vivo and in vitro via a MarA-binding site within the rob promoter that is positioned between the ؊10 and ؊35 hexamers. As for the hdeA and purA promoters, which are repressed by MarA, the rob marbox is also in the "backward" orientation. Protein-DNA interactions show that SoxS and Rob, like MarA, bind the same marbox in the rob promoter. Electrophoretic mobility shift analyses with a MarA-specific antibody demonstrate that MarA and RNA polymerase form a ternary complex with the rob promoter DNA. Transcription experiments in vitro and potassium permanganate footprinting analysis show that MarA affects the RNA polymerase-mediated closed to open complex formation at the rob promoter.
The Rob protein is an abundant nucleoid protein that was initially discovered bound to the right origin of replication of the Escherichia coli chromosome (1). Despite its association with the origin of replication, no experimental evidence supports the role of Rob in chromosome replication, chromatin structure, and superhelicity (2).
Rob is, however, a member of a subgroup within the AraC/XylS family of transcriptional regulators, which includes the MarA and SoxS proteins that are involved in a wide range of regulatory functions (3). The N-terminal domain of Rob shares similarity with MarA and SoxS, which is in contrast to the other members of the subgroup that share homology within the C-terminal domain (3). When induced by bile salts and dipyridyl, the C-terminal domain undergoes post-transcriptional modification that results in the conversion of Rob from a low to a high activity state in the cell (4). However, in the absence of these compounds, overexpression of either the full or C-terminal domain deleted protein is sufficient for activity and promoter binding in vitro (5,6) and in vivo (5,7). This observation is consistent with structural data derived from the Rob-micF complex, where only the N-terminal domain makes DNA-specific contacts (8). Overexpression of rob confers multidrug, organic solvent and heavy metal resistance (6,7); in accord, some experimental data indicate that rob null mutants are hypersensitive to antibiotics, organic solvents, and heavy metals (5,7,9).
Transposon mutagenesis experiments aimed at defining the Rob regulon revealed eight Rob-regulated targets, namely inaA, marRAB, aslB, ybaO, mdlA, yfhD, ybiS, and galT (10), some of which are also known members of the mar and sox regulons (11). In addition, studies demonstrate that Rob induced by decanoate or bile salts results in the increased expression of acrAB in the absence of both the mar and sox loci (12).
The expression levels of rob do not vary dramatically during different growth phases, unlike other nucleoid-associated proteins (e.g. Fis) (2,13,14). There is, however, some evidence that increased rob expression occurs in glucose-and phosphate-limited media and in the stationary phase of cell growth, attributable to activation by the factor rpoS (13). Moreover, recent studies have shown that rob is subject to down-regulation by MarA and SoxS (11,15).
It has been shown previously that all three transcriptional factors (MarA, SoxS, and Rob) regulate the expression of themselves (16,17) and each other (15)(16)(17)(18). This cross-and auto-regulation strongly suggest the presence of putative binding sites within the mar, sox, and rob promoters. Thus far, only the MarA/SoxS/Rob-binding site within the mar promoter has been identified and shown to be responsible for activation by all three transcription factors (17). In contrast, the repression of both soxS and rob by SoxS, Rob, or MarA occurs via a yet to be identified binding site(s) within the sox and rob promoter regions (15,16).
Extensive genetic and crystallographic analyses have shown that the interaction of MarA/SoxS/Rob with activated promoters involves a degenerate DNA sequence known as the "marbox/soxbox" (3). These binding sites are asymmetric and exist in two possible orientations (class I and II promoters) (3) with respect to the RNA polymerase-binding sites. Generally at class I promoters, the marbox lies upstream of the Ϫ35 hexamer in the backward orientation with the exception of the zwf promoter (19), and at the class II promoters the marbox overlaps the RNA polymerase-binding site and lies in the forward orientation (19). Direct transcriptional repression by MarA has been linked more recently to a similar degenerate marbox that lies in the "backward" orientation with partial or complete overlap of the Ϫ35 hexamer (20). This configuration is unlike that of either the class I or class II marboxes necessary for activation (19).
In this study we have characterized the MarA-binding site within the rob promoter in relation to the RNA polymerase-binding sites and have studied the mechanism of MarA-mediated transcriptional repression. Down-regulation of rob mediated by MarA (and also SoxS and Rob) adds to the regulatory cross-talk already reported (15)(16)(17) and shows that the rob promoter and protein are controlled at both pre-and posttranscriptional levels in vivo.

MATERIALS AND METHODS
Bacterial Strains and Plasmids Used-E. coli strains (with or without plasmids) ( Table 1) were grown in LB broth supplemented with either ampicillin (100 g/ml) or kanamycin (30 g/ml) where required.

RNA Preparation and Northern
Blot Analysis-Total RNA was extracted from the bacterial strains using the Qiagen RNeasy kit (Qiagen, Germany). Membranes used for the transfer of RNA were stained with methylene blue to confirm the transfer and were also used for quantitation. The rob DNA probe was generated by PCR amplification with DNA from E. coli strain AG100 (20) using the following primers: RNF, 5Ј-ATGGATCAGGCCGGCATTA-3Ј; RNR, 5Ј-CTGATCCT-GGGCTAACGC-3Ј (T m ϭ 58°C). The subsequent 618-bp probe was random prime-labeled with [␣-32 P]dCTP (PerkinElmer Life Sciences). Hybridizations and subsequent ImageQuant (Amersham Biosciences) analyses were performed as described previously (20).
5Ј-RACE 2 -The transcriptional start site of the rob gene was mapped using the 5Ј-RACE system (Invitrogen). Total RNA was extracted from AG100 and AG100R (constructed by P1 transduction of rob::kan from RA4468 (5) into AG100 (20)). The RNA was digested twice with DNase I to eliminate the contaminating genomic DNA prior to use in the 5Ј-RACE reaction. The cDNA was produced by extension with primer RR1, 5Ј-GCGGGTAAATGTCTGTTG-3Ј, and Moloney murine leukemia virus-reverse transcriptase (Invitrogen). The cDNA was dC-tailed and amplified by PCR with a primer hybridizing to the poly(C)-tail (Abridged Anchor Primer) and another nested gene-specific primer RR2 5Ј-CGG-GAAGCTTTCTGTTGAGAGTCGAAGCGGT-3Ј (T m ϭ 55°C). Another round of PCR amplifications using the Abridged Universal Anchor Primer and the nested primer RobR 5Ј-ACAGGGGCTGATCCAGAT-GACCTTCC-3Ј (T m ϭ 50°C) was performed before a PCR product sized at ϳ180 bp was obtained. No product was found for the rob deletion strain AG100R. This PCR product was purified using the Qiagen QIAquick PCR purification kit (Qiagen, Germany) prior to cloning into the pGEMT vector (Promega, Madison, WI). The presence of inserts was verified by digesting with the restriction enzymes NdeI and NcoI (New England Biolabs). Plasmids containing the inserts were subsequently sequenced at the Tufts University Core Facility, and the junction between the C-tail and the start of the rob gene promoter sequence was taken to be the transcriptional start site.
Transcription Studies in Vitro with MarA, SoxS, and Rob-Transcription experiments in vitro were carried out as described previously (20,21). DNA fragments were gel-purified using the QIAquick gel extraction kit (Qiagen, Germany). The location of the marbox was determined using DNA fragments that consisted of sequential truncations of the rob promoter region and/or selective mutations (see Fig. 2). The mechanism of repression was examined by preincubation experiments with either purified MarA (Paratek Pharmaceuticals, Boston; see Fig. 3A) or RNA polymerase (RNAP) holoenzyme (Epicenter, Madison, WI). Briefly, each protein (RNAP or MarA) was preincubated for 5 min at 0 or 37°C before the addition of the second protein. Following a further 10 min of incubation, the initiating nucleotide mixture with heparin was added and allowed to proceed for another 5 min. The effect of MarA on blocking reinitiation by RNA polymerase in vitro at the rob promoter was assessed by multiple-round transcription experiments. The DNA fragment (2 nM) was initially incubated for 5 min at 37°C in the presence of RNA polymerase (40 nM) to allow for open complex formation. This was followed by the addition of 200 nM MarA and incubation at 37°C for a further 5 min. The levels of transcription were assessed by adding initiating nucleotide mixture without heparin and allowed to proceed for different times (shown in Fig. 6B). The reaction was stopped by adding 30 l of Gel Loading Buffer II (Ambion). The products were separated by electrophoresis on a 7.5% polyacrylamide, 8 M urea denaturing gel and visualized after exposure to a phosphorimaging screen. The quantification of the levels of repression was determined as described previously (20).
Electrophoretic Mobility Assays (EMSA)-EMSA experiments were performed with the wild type, truncated, and mutated rob promoter fragments. To define the marbox sequence, annealed oligomers representing four sequential nonoverlapping promoter sequences of 20 -21 bps each were incubated with MarA, SoxS, or Rob (Paratek Pharmaceuticals, Boston; see Fig. 3A). The promoter fragments (only the coding strands of the oligomer sequences are shown here) used were as follows: RobO1, 5Ј-CTAAAACATACTCTACTAAG-3Ј; RobO2, 5Ј-GAAAAAAACACT-GAATGCTAA-3Ј; RobO3, 5Ј-AACAGCAAAAAATGCTATTAT-3Ј; and RobO4, 5Ј-CCAATTACCTGATGTCAGGT-3Ј (see Fig. 2). The orientation of the marbox was examined using amplified PCR fragments harboring mutations within recognition elements (RE1 and RE2) of the MarAbinding site in both the forward (RbO3F6) and backward (RbF7M6, RbF7M5) marboxes (see Fig. 3 legend for mutation descriptions). The annealed oligomers and DNA fragments were end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP (PerkinElmer Life Sciences). All binding reactions were performed with 32 P-end-labeled DNA oligomers (2 nM), purified MarA (200 -400 nM), and competitor DNA (100 mg/ml, poly[d(I-C)]). The binding reactions were performed at 25°C for 15 min before the addition of loading buffer (0.25ϫ TBE, 60%; glycerol, 40%). The samples were subjected to electrophoresis at room temperature for 2 h on a 6% polyacrylamide gel. The formation of closed complexes was determined as described by Nechaev and Severinov (22). Briefly, binding reactions with MarA (200 -1000 nM) and the rob promoter fragment (pRobF4) were performed at 0°C for 10 min prior to the addition of RNAP (40 nM). Both proteins were incubated on ice for a further 10 min. For the supershift experiments, 1 l of neat penta-His antibody (Qiagen, Germany) was added directly to the reactions, which were performed as described above. Following the addition of loading buffer, products were loaded immediately onto a 3.5% native polyacrylamide gel. The samples were subjected to electrophoresis at 4°C for 4 h. The gel was then dried and exposed for Phos-phorImager analysis.
Potassium Permanganate Footprinting-The footprinting experiments were performed as described previously (23). Briefly, the promoter fragment (pRobF4 at 1 ϫ 10 6 dpm) and MarA (400 nM) or RNAP (40 nM) only or MarA and RNAP together were incubated in 5ϫ binding buffer (25% glycerol, 0.5 M NaCl, 25 mM MgCl 2 , 0.5 M EDTA, 5 mM dithiothreitol, 0.25 mg/ml bovine serum albumin, 100 mM Tris-HCl, pH 8) for 15 min at 37°C. One microliter of 200 mM potassium permanganate was added to each reaction, mixed, and incubated for a further 4 min. The reaction was stopped with the addition of 50 l of Stop solution (3 M ammonium acetate, 0.1 mM EDTA, 1.5 M 2-mercaptoethanol). After extraction with an equal volume of phenol/chloroform/isoamyl alcohol (25:24:1), the aqueous phase was precipitated with ethanol at Ϫ80°C for 15 min. The pellet was then resuspended in 10% piperidine and incubated at 90°C for 30 min. The reaction was stopped with the addition of 50 l of 1 M LiCl and cold absolute ethanol. The pellet was washed in ethanol before resuspension in 5 l of sequencing gel loading buffer (Epicenter, Madison, WI). The resulting products were then analyzed by electrophoresis in an 8% denaturing urea gel.

Analysis of rob Expression Under Different
Conditions-Microarray analyses have demonstrated that either salicylate induction of MarA or overexpression of MarA or SoxS results in decreased expression of rob (11). To confirm that the decrease in rob expression was a result of MarA expression, we analyzed the levels of the rob transcript by Northern blot analysis and reverse transcription-PCR under different conditions. Levels of the rob transcript were determined in a strain deleted for the marRAB locus (JHC1096 ((24)), but which was transformed with an IPTG-inducible plasmid either with (pMB102) or without (pJPBH) marA. As expected, the levels of rob remained unaffected in the vector only control (JHC1096/pJPBH), but a decrease was observed under IPTG induction of MarA (JHC1096/pMB102) (Fig. 1). The levels of repression at 0.25, 0.5, and 1 mM IPTG were 0.43, 0.28, and 0.21, respectively (all ratios were normalized to the amounts of rRNA present on the membrane; see Fig. 1). Using both Northern blot and reverse transcription-PCR analyses, the salicylate induction of marA decreased levels of rob expression in AG100 to 0.67 relative to the uninduced sample (data not shown). Thus both plasmid-mediated expression and salicylate induction of MarA, directly or indirectly, reduced the levels of rob expression 2-4-fold in vivo.
Identification of the Transcriptional Start Site of rob-It was necessary to map the transcriptional start site of the rob gene to determine the positions of the RNA polymerase-binding sites and subsequently the location of the MarA-binding site (marbox) relative to these signals. PCR analysis using the anchor primer (see "Materials and Methods") and a rob gene-specific primer yielded a 180-bp product that was subsequently cloned into the pGEMT vector. Four clones were selected, sequenced, and found to harbor the same rob and dC-tailed primer junction sequence. The transcriptional start site was mapped to 43 nucleotides upstream of the open reading frame with the 5Ј-RACE method (Fig. 2). The experimentally defined position of the transcriptional start site is consistent with the size of the transcripts obtained in both the Northern blot and in vitro transcription experiments (see below). Based on the position of the transcriptional start site, potential Ϫ10 and Ϫ35 sequences within the rob promoter sequence could also be deduced (Fig. 2).
Binding of MarA, SoxS, and Rob to the rob Promoter-MarA, SoxS, and Rob bound to the full-length rob promoter (spanning Ϫ234 to ϩ110) (data not shown). Because previous studies have shown that the MarA-binding sites of two other repressed promoters (hdeA and purA) lie in a backward orientation (20) and either partially or completely overlap the Ϫ35 hexamer, we examined whether the putative rob marbox might be in a similar configuration and position. The promoter fragment pRobF5 (Ϫ55 to ϩ110), which contains the putative marbox region, and pRobF6 (Ϫ26 to ϩ110), which does not, were used in the initial binding experiments with MarA (see Fig. 2). Most unexpectedly, MarA bound to both fragments, indicating that the rob marbox was in a different position from those described previously (20). Therefore, MarA binding was tested with four pairs of nonoverlapping 20 -21-bp annealed oligomers that sequentially span Ϫ67 to ϩ15 (see Fig. 2) within the rob promoter region (Fig. 3B). The only oligomer pair (RobO3) to which MarA specifically bound corresponded to the DNA sequence between the Ϫ35 hexamer with a partial overlap of the Ϫ10 hexamer (Fig. 3B). As expected, this promoter region (spanning Ϫ27 to Ϫ7) was also bound by SoxS and Rob (Fig. 3C).
We next attempted to determine the orientation of the binding site. Close analysis of the marbox sequence to which all three regulatory proteins bound showed that there were two putative marboxes, one in the forward and one in the backward orientation. The forward and the backward marbox had 17/20 (14/17) and 15/20 (12/ 17) matches to the revised consensus marbox sequence, respectively (A GRGCACRWWNNRYYAAA GN where indicates any base except G) (25). The crucial determinants of a marbox include two 4-bp sequences known as recognition elements (RE). Mutations in RE1 (GCAC) or RE2 (CAAA) can either reduce affinity or abolish binding to MarA and SoxS (19,20,26,27), although mutations within RE1 affect MarA or SoxS binding more than those within RE2 (20,26). The RE2 sequence in the backward marbox is a poor match (TGTT) to the consensus sequence (YAAA (25)); the change TGTT3 GGTG had no effect on MarA binding (RBF7M6; Fig. 3D). Mutations within RE1 for the forward marbox sequence (GCAA3 ATGG) (RB03F6; Fig. 3D) did not hinder MarA binding to this sequence; however, the RE1 mutations (GCAT3 TATC) (RBF7M5, Fig. 3D) within the backward marbox  destroyed MarA binding. These results suggest that the repressed rob promoter marbox also lies in a backward orientation, as was reported previously for the hdeA and purA promoters (20).
Transcription Studies in Vitro-To determine whether the binding of MarA/SoxS and Rob to the rob promoter reflected direct repression of the rob promoter, we performed transcription in vitro experiments. As described previously (20), the test transcripts (regulated by MarA) in in vitro studies were compared against a control (gnd) transcript (not regulated by MarA) after background subtraction. The average ratios among these transcripts in the absence and presence of MarA represented the level of repression. (We find that this approach corrects for the inter-and intra-experimental variability of both the test and control transcripts.) Initially a large fragment of the rob promoter region (Ϫ234 to ϩ110; see Fig. 2), amplified with the primers reported by Michan et al. (15), was used. With this full-length rob promoter fragment, MarA (200 nM) reduced transcription to 0.29 Ϯ 0.024 (Fig. 4A). The promoter regions upstream of the Ϫ35 hexamer (see Fig. 2) were sequentially removed to define the region of the marbox involved in MarA repression. Transcription from all the 5Ј-deleted promoter fragments tested was repressed by MarA (data not shown for all fragments except for representative truncated fragments, pRobF5 and pRobF4; see Fig. 4 and Fig. 6A, respectively).
Further truncations of the rob promoter were restricted by the presence of the putative Ϫ35 hexamer (atgcta). In an attempt to confirm whether the Ϫ35 hexamer was indeed where predicted and important for transcription, we introduced point mutations within the putative region (Fig. 4A). First, a C (atgCta) to G (atgGta) change in the Ϫ35 region resulted in the loss of gene transcription completely (Fig. 4A). In addition, the change of the potential Ϫ35 hexamer (atgcta) in the rob promoter to the E. coli consensus sequence (ttgaca) did not alter transcription, dramatically contrasting the effect noted previously at the hdeA promoter (20). This finding could be attributed to the original Ϫ35 sequence already being a constitutive promoter given the relative abundance of Rob molecules within the cell (1,14). Thus the effects on gene transcription observed in vitro support our hypothetical Ϫ10 and Ϫ35 predictions within the rob gene promoter. Unexpectedly, the consensus promoter appeared much less responsive to MarA repression (ratio, 0.8) in comparison to the original sequence (ratio, 0.29), especially because changes introduced were not within the marbox sequence. The introduction of the consensus E. coli Ϫ35 hexamer may have altered RNAP binding at the rob promoter making it less responsive to MarA-mediated repression. In vitro transcription experiments involving the full-length promoter fragment and purified Rob protein also resulted in repression (ratio of 0.33 compared with no Rob protein control) (Fig. 4B). Therefore, both MarA and Rob are able to repress the rob promoter directly in vitro, and this repression occurs at the level of transcription.
Defining the MarA-mediated Mechanism of Repression-Because the marbox in the rob promoter lies between the Ϫ10 and Ϫ35 hexamers, it is possible that MarA mediates repression by excluding the access of RNA polymerase to the promoter. To test this hypothesis, we performed gel shift analyses of the rob promoter with MarA preincubation at 0°C prior to the addition of RNAP. MarA binding to the rob promoter produced, as expected, a shifted complex (Fig. 5, A1); RNAP binding produced three complexes (Fig. 5, R1-R3). The preincubation of increasing concentrations of MarA (200 -1000 nM) with the rob promoter followed by the addition of RNAP (40 nM) produced complexes (Fig. 5, see complexes in lanes 6, 8, and 10) similar to those formed by RNAP alone (Fig. 5, R1-R3, lane 4). Because MarA binds as a monomer (28) and has a molecular mass of 15 versus 450 kDa for RNA polymerase, the differences in shifts of a heavier ternary complex (MarA-RNAP-DNA), for the complexes (Fig. 5, RA), would not be easy to differentiate from complexes formed with RNAP-DNA alone (Fig. 5, R1-R3). However, the heavier complexes (RA) formed in Fig. 5, lanes 12 and 14, suggest a ternary complex of DNA-RNAP-MarA.
To confirm that MarA was indeed a part of the RA complexes (Fig. 5,  lanes 6, 8, 10, 12, and 14), supershift assays with a penta-his tag antibody   Fig. 2; rob fragment 5 is pRobF5, which spans from Ϫ55 to ϩ110; rob mutated represents a point mutation within the Ϫ35 hexamer, which converts the wild type atgCta to atgGta; rob consensus fragment contains a change in the wild type Ϫ35 hexamer (AtgCTa) to the E. coli consensus sequence (TtgACa). B shows the transcription of the rob full-length in the absence or presence of purified Rob. The rob and gnd transcripts appear as doublets in the gels.
(directed against the purified MarA) were performed under identical reaction conditions. The addition of the anti-His 5 tag antibody formed supershifted complexes, RAH1-RAH3 in place of the RA complex (Fig.  5). More importantly the simultaneous addition of both MarA and RNAP (Fig. 5, lanes 6 and 7) and preincubation reactions involving RNAP before the addition of MarA produced similar supershifted complexes (RAH1-3) as that seen with the MarA preincubation supershift experiment (data not shown). These observations confirm that MarA was indeed present within the RA complexes. Of note, the RNAP-MarA-DNA shifted complexes were disassociated when challenged with heparin, indicative of the closed RNAP-promoter complex (data not shown). These results demonstrate that steric hindrance is not the mechanism of repression as excess MarA did not interfere with the binding of RNAP to the promoter and that MarA and RNAP bind together at the rob promoter.
Transcription studies in vitro were also performed to understand the mechanism of repression (see "Material and Methods"). When MarA was preincubated with the truncated rob promoters (pRobF4 or pRobF5) for 5 min at 37°C before the addition of RNAP, repression was detected (average repression ratio, 0.37). In contrast, when RNAP was preincubated with the DNA for 5 min at 37°C before the addition of MarA, no repression (average repression ratio, 1.2-1.4) was observed (Fig. 6A). Of note, the order of protein addition did not alter the transcriptional outcome at the activated class I (mar) and class II (nfnB) promoters, although the levels of activation were greatest with MarA preincubation. 3 When RNA polymerase was preincubated with the rob promoter at 0°C for 5 min followed by the addition of MarA and subsequent transfer to 37°C, repression was observed (average levels, 40 -90% for MarA concentrations tested; Fig. 6B). As expected, when RNA polymerase was preincubated with the promoter at 37°C for 5 min, no repression was observed with the addition of MarA (200 nM) (Fig. 6B); only at higher MarA concentrations (400 and 800 nM) was some repression noted under the 37°C reaction conditions (Fig. 6B). The transcription in vitro results together with the gel shift data indicate that MarA-mediated repression at the rob promoter occurs before open complex formation but subsequent to RNA polymerase binding; once the open complexes were formed, MarA had little effect.
To determine whether MarA can repress the rob promoter after transcription elongation, thereby interfering with reinitiation by RNA polymerase, we performed multiple-round transcription experiments. The promoter (pRobF4) was preincubated with RNA polymerase for 5 min at 37°C (previously shown to be sufficient for open complex formation at the rob promoter) before the addition of MarA followed by initiating nucleotides without heparin. The multiple-round transcription experiments demonstrated that in the absence of MarA, the mRNA transcripts for both the rob and gnd promoters accumulated during the course of the assay (Fig. 6C). However, in the presence of MarA the rob transcript levels decreased over time, but as expected the levels of the gnd transcript (the control) remained unaffected (Fig. 6C). Thus MarA is able to repress the rob promoter after the preformed open complex clears the promoter.
Potassium Permanganate Footprinting-Potassium permanganate has been shown to preferentially modify single-stranded thymines (T) and cytosines (C) to some extent, which permits the identification of melted DNA in RNAP-mediated open complexes (29). Preincubation with either MarA or RNAP followed by the second protein resulted in cleavage at the following sites: Ϫ11 (T), Ϫ10 (T), Ϫ4 (A), Ϫ3 (A), Ϫ2 (T), Ϫ1 (T) (Fig. 7). Preincubation of the rob promoter with MarA resulted in a dramatic decrease in cleavage, and thus of open complex formation after RNAP addition (Fig. 7). In contrast, preincubation with RNAP and the subsequent open complex formation were not affected by MarA. Moreover, the same cleavage patterns were observed in the presence or absence of MarA (Fig. 7). This finding suggests that MarA

DISCUSSION
This study demonstrates that MarA can repress the expression of the rob gene both in vivo and in vitro via a marbox sequence located within its promoter. Like the previously characterized repression marboxes for hdeA and purA (20), the rob marbox is also in the backward orientation, but lies between the Ϫ10 and Ϫ35 hexamers. This region has been called the "exclusive zone of repression" (30), as transcription factorbinding sites that lie within this position almost always lead to repression, e.g. IclR at the iclR promoter (31), Fur at the aerobactin promoter (32), and the 434 repressor at 434 bacteriophage P r promoter (33). Exceptions to this rule do exist such as positive regulation by MerR at the merTPAD promoter (34), Arc at the P ant variant promoter of bacteriophage P22 (35), and SoxR at the soxS promoter (36).
The two marboxes (type I and type II) (see Fig. 8) involved in repression utilize the same degenerate MarA-binding sequence (this work and see Ref. 20) as those involved in activation (19). In addition the DNA and MarA contacts found to be crucial in activation (19,26,28,37) are also important for repression (20) as mutations within RE1 in both the purA and rob marboxes were sufficient to abolish MarA binding and, in the case of purA, to eliminate MarA-mediated repression in vitro (20).
Transcription initiation is a multistep process and is represented by the formation of several intermediary transcription factor-RNAP-DNA complexes (38). Given the location of the rob marbox, several possible scenarios for the mechanism of MarA-mediated repression emerge. First, MarA could simply sterically hinder the access of RNAP to the rob promoter. Second, MarA could inhibit the transition from closed to open complex formation or from open to initiated complex formation. Finally, MarA could hinder promoter clearance.
Preincubation of the rob promoter with saturating concentrations of MarA before the addition of RNAP did not prevent the formation of RNAP-DNA complexes (see Fig. 5). The prior presence of either MarA or RNAP did not inhibit binding of the other protein to the rob promoter. When RNAP was added to the rob promoter at 0°C prior to MarA addition and then transfer to 37°C, MarA inhibited rob transcription (see Fig. 6B). These results argue against steric hindrance as the MarA-mediated mechanism of repression at the rob promoter. Instead MarA, RNAP, and DNA form a ternary complex as demonstrated by the supershift assays performed with penta-his antibody (directed against the purified MarA).
At the rob promoter, MarA represses by inhibiting the transition from the closed to open complex (see Fig. 6B), but it has no effect after the formation of open complexes (see Fig. 6, A and B, and Fig. 4A). These findings are further supported by the potassium permanganate footprinting results that show that MarA preincubation decreases the formation of open complexes at the rob promoter; once formed, the open complexes are not affected by MarA (see Fig. 7). Because open complex formation is a transient stage in vivo, the multiple-round transcription experiments showed that MarA was able to repress once the bound RNAP had cleared the rob promoter (see Fig. 6C).
Residues 265 and 294 within the ␣-CTD of RNAP contact MarA when interacting at the activated promoters (39). Based on helical spacing experiments between the marbox and the RNAP binding signals at the activated genes, it is predicted that MarA binds the same face of the DNA relative to RNAP (3,19). Our data demonstrate that MarA and RNAP co-bind the rob promoter (see Fig. 5); however, the conformational aspects of this ternary complex (MarA-rob-RNAP) are unclear. The crystal structures of MarA with the mar promoter (28) and Thermus aquaticus RNAP holoenzyme with a fork junction promoter DNA (40) suggest that both MarA and RNAP might bind the same face of the DNA. However, this predictive model does not take into account the DNA bends introduced by each protein. Therefore, other conformations may exist at the rob promoter. Because helical dependence is a critical factor in activation by MarA (19), the transcriptional outcome, repression versus activation, may be dependent on which DNA face the protein binds as is the case for GalR (41). The phage 434 protein represses the P r promoter via a binding site that lies between the Ϫ10 and Ϫ35 hexamers (33,42). Ethylation interference experiments at this promoter demonstrate that repression occurs via the binding of the 434 repressor and RNAP on opposite faces of the promoter (33). Alternatively, specific interactions between MarA and the rob promoter may be different from that described for the activated promoters (28) and occur through only one of the helix-turn-helix motifs as predicted in the Rob-  micF crystal structure, thereby allowing both proteins to bind the same face of the DNA (8).
The interaction of MarA via the ␣-CTD of RNAP in the absence of DNA strengthens the hypothesis that a MarA-RNAP complex scans the chromosome for cognate binding sites for both proteins (43,44). At the activated promoters, the relative positions of both the MarA and RNAPbinding sites suggest that the disassociation of the MarA-RNAP complex may not be necessary before the complex contacts DNA (39). The position of the rob promoter marbox does not exclude the possibility of the MarA-RNAP complex binding opposite faces of the helix simultaneously but in a nonproductive conformation, or that RNAP or MarA disassociate from the complex and bind the promoter independently. At other promoters with repressor binding sites similar to that in the rob promoter, different strategies to attract either the repressor or RNAP are proposed. At the merOP promoter, MerR has been shown to stabilize the binding of RNAP that supports the recruitment theory (45), and at the early A2c promoter, RNAP has been shown to recruit the repressor p4 (46).
This study identifies a uniquely positioned marbox, unlike those described previously (19,20), and demonstrates the mechanism of MarA-mediated repression at the rob promoter. Examples of MarAactivated genes involve marboxes that can be in the "forward" or backward orientation and in different positions relative to the RNAP-binding sites (19) (see Fig. 8). In contrast, in all three known examples of MarA-repressed genes (hdeA, purA, and rob), the marbox is in the backward orientation and is proximal to or overlapping the RNAP-binding site (see Fig. 8).