Nucleotides from –16 to –12 Determine Specific Promoter Recognition by Bacterial σS-RNA Polymerase*

The alternative sigma factor σS, mainly active in stationary phase of growth, recognizes in vitro a –10 promoter sequence almost identical to the one for the main sigma factor, σ70, thus raising the problem of how specific promoter recognition by σS-RNA polymerase (EσS) is achieved in vivo. We investigated the promoter features involved in selective recognition by EσS at the strictly σS-dependent aidB promoter. We show that the presence of a C nucleotide as first residue of the aidB –10 sequence (–12C), instead of the T nucleotide canonical for σ70-dependent promoters, is the major determinant for selective recognition by EσS. The presence of the –12C does not allow formation of an open complex fully proficient in transcription initiation by Eσ70. The role of –12C as specific determinant for promoter recognition by EσS was confirmed by sequence analysis of known EσS-dependent promoters as well as site-directed mutagenesis at the promoters of the csgB and sprE genes. We propose that EσS, unlike Eσ70, can recognize both C and T as the first nucleotide in the –10 sequence. Additional promoter features such as the presence of a C nucleotide at position –13, contributing to open complex formation by EσS, and a TG motif found at the unusual –16/–15 location, possibly contributing to initial binding to the promoter, also represent important factors for σS-dependent transcription. We propose a new sequence, TG(N)0–2CCATA(c/a)T, as consensus –10 sequence for promoters exclusively recognized by EσS.

Bacterial cells adapt to changing environmental and physiological conditions by modulating gene expression. Sigma () factors of RNA polymerase, as the subunits responsible for promoter recognition, play a major role in programming gene expression. At least seven different subunits have been identified in Escherichia coli; 70 is the main subunit and can carry out transcription from the majority of E. coli promoters. The alternative subunits can direct transcription toward specific sets of genes (i.e. heat-shock, extracellular proteins, etc.) whose transcription is directed by -specific promoter sequences. A partial exception to the typical role for alternative subunits is represented by S , the product of the rpoS gene, mainly expressed in the stationary phase of growth (1)(2)(3). Unlike the other subunits, S -RNA polymerase (E S ) 1 can initiate transcription from several promoters also recognized by E 70 , suggesting that they recognize similar promoter sequences (4). The recognition of similar promoter sequences by S and 70 is reflected by their strong similarity in the DNA binding domains (5). The alignment of E S -dependent promoters and the search for an optimal promoter for E S in vitro using the systematic evolution of ligands by exponential enrichment (SELEX) procedure have pointed to a Ϫ10 consensus sequence for E S , CTATA(c/a)T that is very similar to the canonical TATAAT sequence for 70 (6 -10). These results suggest that promoter selectivity between 70 and S might be determined by factors other than promoter DNA sequence. Indeed, studies of different S -dependent promoters suggest that other parameters such as increased intracellular salt concentration, degree of DNA supercoiling, the ppGpp alarmone, and modulation by trans-acting regulators all contribute to S selectivity (reviewed in Refs. [11][12][13][14][15]. The main sequence feature specific for S -dependent promoters would be a C nucleotide immediately upstream of the Ϫ10 promoter element (CTATA(c/a)T at the Ϫ13 position relative to the transcription start (Ϫ13C)); the Ϫ35 element does not appear to be important for promoter recognition by S , although conflicting data on its role have been reported (10, 16 -19). Mutagenesis studies on S -dependent promoters have shown that deviations from the consensus CTATACT affect promoter recognition by S in vivo and have substantiated the importance of the Ϫ13C element (8,20,21). Suppression genetics data suggest that the Ϫ13C is directly contacted by the residue Lys-173 of region 2.5 of S (8). Interestingly, the amino acid located at the equivalent position in region 2.5 of 70 is a glutamate residue (Glu-458) that has been proposed to interact with another promoter element located upstream of the Ϫ10, the TG motif (22). At the so-called "extended Ϫ10" promoters, this TG dinucleotide can be found two nucleotides upstream from the Ϫ10 sequence, where it allows transcription even in the absence of a conserved Ϫ35 sequence (23)(24)(25)(26)(27)(28). However, in vitro selection experiments (10) suggest that a TG motif might also be recognized by S and have proposed TGTGCTATA(c/a)T as the optimal extended Ϫ10 sequence for S binding.
The aidB gene belongs to the adaptive response to DNA alkylating agents and is activated by the Ada regulatory protein (29) (for reviews, see Refs. 30 and 31). The Ada protein, in its methylated form ( me Ada), can activate transcription at the aidB promoter (PaidB) by both the E 70 and E S forms of RNA polymerase (32,33). However, in the absence of activation by Ada, aidB expression is strictly S -dependent in vivo, and only the E S form of RNA polymerase can efficiently carry out transcription from PaidB in vitro (34,35). Although E 70 can bind PaidB in vitro in the absence of any activator, binding by E 70 results in an unusual complex that is partially resistant to heparin challenge but unable to carry out full promoter opening and transcription initiation (21).
In this work, we investigate the importance of nucleotides around the Ϫ10 region of the aidB promoter for S selectivity. We confirm the role of Ϫ13C for S activity and show that a TG motif at the unusual Ϫ15/Ϫ16 position plays a specific role in S -dependent transcription. We show that, unlike 70 , S can recognize a C residue instead of the canonical T as first nucleotide of the Ϫ10 box (Ϫ12C). These features are not unique to the aidB promoter, but they are conserved in a subset of Sdependent promoters. Thus, based on the presence of either a C or a T as the first nucleotide in the Ϫ10 box, we can distinguish between specific E S -dependent promoters and promoters that can be recognized by both E S and E 70 , at which additional regulatory factors are likely to play a role in selective promoter recognition by E S . We propose TG(N)CCATA(c/a)T as the new consensus sequence for the factor-independent class of S -dependent promoters.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The strains used in this study are the E. coli K12 strains MV1161 and its rpoS derivative MV2792 (34). These strains were transformed with derivatives of the reporter plasmid pRS1274 (36) carrying lacZ fusions under the control of either the wild type aidB promoter or its mutant derivatives. To produce the reporter plasmids, we cloned the aidB promoter region into the multiple cloning site of the pRS1274 vector as 238-bp BamHI-EcoRI fragments. The PaidB wild type fragment was generated by PCR using the aidBbam primer (5Ј-TATAGCAAGCTTCGTGCGGAATGGGGATCC-3Ј), annealing at Ϫ140 to Ϫ123 of PaidB, and aidBeco (introducing an EcoRI site, 5Ј-CGGAAAGAATTCGCAGAGCGCGCCATCAGA-3Ј), annealing at positions 93-116. The aidB mutant promoters shown in Fig. 1 were generated by PCR using the primers aidBeco together with the appropriated mutagenic primer MaidBnco (5Ј-AATCCATGGCAGTgaccA-TACTaATGG-3Ј; the small italic letters indicate the position of the various substitutions), annealing at Ϫ29 to Ϫ2 and encompassing the NcoI restriction site of PaidB. The PCR products were cloned into the plasmid pMV120 (pUC18, carrying the aidB promoter region; Ref. 36) using the NcoI and EcoRI restriction sites, and introduction of the desired mutations was checked by sequencing. Finally, the 238-bp BamHI-EcoRI fragments encompassing either the wild type or the mutant aidB promoters were subcloned into either the pRS1274 plasmid (for in vivo transcription assays) or the pJCD01 plasmid (37) for in vitro transcription experiments (see below).
In Vivo Transcription Assays-Bacterial cultures grown overnight in Luria broth (LB) medium at 37°C were diluted 1:100 in LB, grown to an OD 600 nm ϭ 0.2, then re-diluted 1:100 in pre-warmed LB. The second dilution was performed to reduce the ␤-galactosidase activity carried over from the overnight cultures. Samples were collected when the re-diluted cultures reached OD 600 nm ϭ 0.1 and then after 4 h of growth; at this time, OD 600 nm was ϳ2.5 for both MV1161 and MV2792 strains. ␤-galactosidase activity from the aidB::lacZ fusions was determined as described previously (38) and was comparable with the activity obtained by a promoterless lacZ gene in samples taken at OD 600 nm ϭ 0.1 (data not shown).
Protein Purification and Holoenzyme Reconstitution-Core enzyme and factors were purified as described previously (Refs. 39 and 16, respectively); proteins appeared to be pure from contaminants as determined from denaturing protein gel electrophoresis. Reconstitution of active holoenzymes for the different experiments was achieved by incubating the core enzyme and either S or 70 at a 1:4 ratio (to ensure saturation of the core enzyme by factors) for 30 min at 37°C. For the competition experiment in a bandshift assay, the ratio S : 70 :core was reduced to 1, and reconstitution time was reduced to 10 min. The reconstituted holoenzymes were diluted at room temperature in K-glu200 buffer (40 mM HEPES, pH 8.0, 10 mM magnesium chloride, 200 mM potassium glutamate, 4 mM dithiothreitol, and 500 g/ml bovine serum albumin) prior to their use for in vitro experiments. Dithiothre-itol was omitted from the buffer for permanganate (KMnO 4 ) reactivity experiments.
In Vitro Transcription Assays-Single-round transcription assays were performed in K-glu200 buffer on derivatives of the supercoiled plasmid pJCDO1 (37) carrying either wild type or mutant PaidB. Plasmids (3 nM) and reconstituted RNA polymerase holoenzymes (50 nM for E S or 100 nM for E 70 ) were incubated for 15 min at 37°C to allow complex formation. Elongation was started by the addition of a prewarmed mixture containing nucleotides and heparin (final concentrations were 500 M ATP, GTP, and CTP; 30 M UTP; 0.5 Ci of [␣-32 P]UTP; and 500 g/ml heparin) to the template-polymerase mix and allowed to proceed for 10 min at 37°C. Reactions were stopped by the addition of 10 l of loading buffer (formamide containing 20 mM EDTA, xylene cyanol, and bromphenol blue). After heating to 65°C, samples were loaded on 7% polyacrylamide sequencing gels. Reaction products from PaidB were quantified using a PhosphorImager (Molecular Dynamics) and normalized to the standard RNAI product after background subtraction.
For gel mobility shift assays, the reconstituted holoenzyme (5-50 or 6 -60 nM for the competition experiment) and the 180-bp promoter DNA fragments (1 nM) were incubated for 16 min at 37°C in K-glu200 buffer in a final reaction volume of 10 l. The reaction mixture was then loaded onto a native 5% polyacrylamide gel after the addition of 2.5 l of loading buffer (50% sucrose, 0.025% xylene cyanol, 0.025% bromphenol blue, and 150 g/ml heparin).
For DNase I footprinting, reconstituted RNA polymerase (100 nM either E S or E 70 ) was incubated with either PaidB(WT) or PaidB(12C3 T) (4 nM) for 30 min at 37°C in K-glu200 buffer. Protein-free DNA samples were treated with 1 g/ml DNase I for 20 s, whereas the incubation was prolonged to 30 s in the presence of RNA polymerase. After the addition of loading buffer containing 150 g/ml heparin, the samples were separated on a 5% native polyacrylamide gel, and the bands corresponding to the RNA polymerase-promoter complexes were eluted from the gel, precipitated, and re-suspended in EDTA (20 mM)-formamide buffer before being loaded on a 7% polyacrylamide sequencing gel.
KMnO 4 reactivity experiments were performed as described (41). Briefly, 50 nM RNA polymerase and promoters were incubated in K-glu200 (without dithiothreitol) for 15 min at 37°C; KMnO 4 was added to a final concentration of 10 mM, and the reaction was stopped after 30 s by adding 2-mercaptoethanol to a final concentration of 330 mM. For the kinetic reactivity experiment, samples were taken after 1.2, 2, 4, 6, 8, and 19 min of incubation. The KMnO 4 -reactive bands were expressed as a percentage of the total labeled DNA loaded onto the gel.

Promoter Elements Determinant for Selective Recognition of the aidB Promoter by E S -
The aidB promoter (PaidB) is strongly S -dependent in the absence of Ada activation, i.e. in physiological growth conditions (36). To identify which features determine specific recognition of PaidB by E S , we performed site-directed mutagenesis in and around the Ϫ10 sequence. We chose to target four putative promoter elements; as shown in Fig. 1A, a TG dinucleotide, typically a feature of 70 -dependent extended Ϫ10 promoters, is placed at an unusual location in PaidB (three and four nucleotides upstream of the Ϫ10 box instead of two and three). At this location, a TG motif cannot stimulate 70 -dependent transcription (28). We also targeted for mutagenesis the Ϫ13C residue, proposed to be the main feature for S -specificity (7-9), and the Ϫ12C, i.e. the first nucleotide in the Ϫ10 hexamer. Finally, we changed to T the A nucleotide at position Ϫ6 immediately downstream of the Ϫ10 sequence, because T appears to be conserved at this position in S -dependent promoters (8,9). We examined the effect of the mutations on promoter activity in vivo using a low copy plasmid carrying the PaidB::lacZ transcriptional fusion (Fig. 1B). We compared promoter activity in the MV1161 strain (wild type relative to rpoS) and its rpoS derivative, MV2792, upon entry into stationary phase (OD 600 nm ϳ 2.5). Substitutions that either disrupt the TG motif (15G3 C; designated G15C in Figs. [1][2][3][4] or eliminate the Ϫ13C (13C3 G; designated C13G in Figs. 1-4) show a rather small (1.5-to 2-fold) albeit significant and reproducible reduction in promoter activity, suggesting a possible role for these two promoter elements in modulation of affinity by E S for PaidB. These effects are confirmed in a 15G3 C/13C3 G double mutant ( Fig. 1B; this mutant is designated G15C/C13G in Figs. [1][2][3][4]. Substitution of the Ϫ6A nucleotide had no effect on transcription from PaidB. Displacement of the TG motif (15G3 T/14A3 G double mutant) to the optimal location for 70 (Ϫ15/Ϫ14) slightly improves transcription by both E S and E 70 . Substitution of the Ϫ12C nucleotide to a T (12C3 T; designated C12T in Figs. 1-7) results in an almost perfect Ϫ10 sequence for 70 (TATACT); the 12C3 T mutation strongly increases aidB transcription both in the wild type strain (roughly 4-fold) and the rpoS strain (20-fold). Although an increase of 70 -dependent transcription was expected for this promoter mutation, its extent is surprising; transcription from PaidBC12T reaches levels similar to the wild type aidB promoter in the wild type MV1161 strain. This observation suggests that the Ϫ12C plays a major role in selective recognition of the aidB promoter by E S in vivo.
In Vitro Transcription Experiments-To confirm that the effects observed in vivo were indeed due to interaction between the promoter and the two forms of RNA polymerase and not to indirect effects, we performed an in vitro transcription assay with purified RNA polymerases. In vitro transcription experiments confirmed that PaidB displays a clear preference for E S (Fig. 2); however, the ratio between E S -and E 70 -dependent transcription is reduced to ϳ5-fold compared with 20-fold in vivo (Fig. 1B), probably due to the excess RNA polymerase used for in vitro assays and the lack of competition for RNA polymerase by other promoters. The effects of the mutations in the aidB promoter were, in most cases, consistent with the ␤-galactosidase assays, even for those mutations that only have moderate effects on promoter activity. Substitution of G at Ϫ15 to C (15G3 C) reduced E S -dependent transcription by ϳ50% but did not affect transcription by E 70 , thus confirming the importance of a TG motif located at Ϫ16/Ϫ15 for E S -dependent transcription. In contrast with the results of in vivo transcription, substitution of C at Ϫ13 to G (13C3 G) did not impair transcription by E S in vitro, whereas it resulted in a 3-to 4-fold stimulation for E 70 . Stimulation to a similar extent of E 70 -dependent transcription by the 13C3 G mutation can also be observed in vivo (Fig. 1B), strongly suggesting that a G nucleotide at position Ϫ13 favors E 70 , which is in agreement with previous observations (8,9). The effect of the 15G3 C/ 13C3 G double mutation clearly confirms the preference of E 70 for a G at the Ϫ13 position as well as the importance of the TG for optimal transcription by S -RNA polymerase. Replacement of the TG motif (15G3 T/14A3 G; designated G15T/A14G in Figs. 1 and 2) to the optimal location for 70 (Ϫ15/Ϫ14) showed no effect on E S -dependent transcription, strongly suggesting that E S , unlike E 70 (28), can interact with the TG motif even when placed at alternative locations. As expected from the in vivo results, the 6A3 T substitution (designated different extent, i.e. by 2-fold for E S opposed to Ͼ5-fold for E 70 . Thus, for the 12C3 T mutant of PaidB, the difference between levels of E S -versus E 70 -dependent transcription is reduced to Ͻ2-fold. Gel Retardation Assays-Transcription initiation takes place in at least three distinct steps, i.e. binding of RNA polymerase to the promoter sequence, formation of the so-called "open complex," and escape. Although transcription from PaidB in the absence of additional proteins is strictly E S -dependent, both E S and E 70 can bind the promoter (32,35). Binding by E 70 results in the formation of a complex partially resistant to heparin challenge and thus akin to the open complex but unable to carry out transcription initiation (21). We investigated the role of the different promoter elements important for interaction with either form of RNA polymerase in the various steps of transcription initiation. To analyze the effects of the mutations on initial binding and formation of a heparin-resistant RNA polymerase-promoter complex, we performed gel mobility shift assays in the presence of heparin (Fig. 3). Mutations affecting E S -dependent transcription also inhibited formation of heparin-resistant E S -PaidB complexes, although to different extents. Surprisingly, disruption of the TG motif (15G3 C) resulting in the strongest inhibition of E S -dependent transcription in vitro (Fig. 3) only moderately affected formation of heparin-resistant complexes by E S . In contrast, PaidB mutants in which the Ϫ13C was targeted (both the 13C3 G and the 15G3 C/13C3 G mutants) displayed impaired formation of heparin-resistant complexes by E S by a 2-fold factor. None of these mutations showed any significant effect on the formation of heparin-resistant complexes by E 70 (Fig. 3B). However, formation of the E 70 -PaidB complex was clearly stimulated by substitution of the Ϫ12C nucleotide, which, in contrast, did not strongly affect interaction with E S .
KMnO 4 Reactivity Experiments-To investigate whether PaidB mutations affect open complex formation by either form of RNA polymerase, we performed permanganate (KMnO 4 ) reactivity experiments. This assay takes advantage of KMnO 4 reactivity with thymine residues in single-stranded DNA, thus allowing us to probe open complex formation by RNA polymerase (42). As expected, E S is favored in promoter opening at the wild type aidB promoter; E S -induced KMnO 4 reactivity extends to a T nucleotide at position ϩ2 in the template strand. In contrast, E 70 only induces partial promoter opening, extending between Ϫ11 and Ϫ5, suggesting that the extension of the reactivity to ϩ2 is necessary for transcription initiation (Fig. 4; Ref. 21). Thus, to quantify the amount of initiation-proficient complex, we measured KMnO 4 reactivity of the ϩ2T residue at the different mutant promoters. The results of these experiments are summarized in Fig. 4 and mirrored our observations in gel retardation experiments. Disruption of the TG motif at Ϫ16/Ϫ15 as well as substitution of the first nucleotide of the Ϫ10 hexamer (12C3 T) did not affect E S -dependent KMnO 4 reactivity at ϩ2T. In contrast, mutations targeting the C nucleotide at Ϫ13 (13C3 G and the 13C3 G/15G3 C double mutant) clearly impaired open complex formation.
Although disruption of the TG motif did not affect E 70 -PaidB interaction, mutations resulting in C to G substitutions at Ϫ13 appear to stimulate promoter opening by E 70 , consistent with the results of ␤-galactosidase assays and in vitro transcription experiments (Figs. 1 and 2). However, the 13C3 G substitution appears to only improve promoter open-ing by E 70 in the Ϫ11 to Ϫ5 region of the promoter without affecting KMnO 4 reactivity at the ϩ2 nucleotide (Fig. 4A). In contrast, the C12T mutation results in a strong increase of E 70 -dependent KMnO 4 reactivity at ϩ2, strongly suggesting that the presence of a T as first nucleotide in the Ϫ10 sequence allows efficient promoter opening and transcription initiation by E 70 . At the 12C3 T mutant promoter, efficiency of the E 70 -dependent promoter opening is roughly 40% compared with E S (Fig. 4B), which is consistent with the 2-fold difference observed in the in vitro transcription experiments (Fig. 2).

Interaction of the Wild Type and the C12T Mutant aidB Promoter with E S and E 70 -The results of both in vivo and in vitro experiments clearly point to the C nucleotide at position
Ϫ12 as an important determinant for E S specificity at the aidB promoter and demonstrate that its substitution to a T allows E 70 to carry out transcription initiation (Figs. 2 and 4). We further investigated the nature of the interaction between both forms of RNA polymerase and PaidB(12C3 T) in comparison to the wild type aidB promoter. We assessed the role of the first nucleotide of the Ϫ10 hexamer in open complex formation by measuring the kinetics of KMnO 4 reactivity (Fig. 5). Although the C to T substitution at position Ϫ12 does not increase E S -dependent KMnO 4 reactivity, it clearly results in faster promoter opening by E S . Faster promoter opening might explain the higher levels of E S -dependent transcription from PaidB(12C3 T) compared with the wild type promoter both in vivo and in vitro (Figs. 1-2). In contrast, the 12C3 T substitution is absolutely necessary for full promoter opening by E 70 (Fig. 4). Interestingly, the presence of a T as first nucleotide of the Ϫ10 hexamer dramatically improves formation of the "partially open" E 70 -PaidB(12C3 T) complex, i.e. the formation of a region of single-stranded DNA between Ϫ11 and Ϫ5. However, KMnO 4 reactivity of the ϩ2T remains weaker than in the E S -PaidB complexes and follows a slower kinetic (Fig. 5B).
The KMnO 4 reactivity experiments point to clear differences in the interaction between PaidB(12C3 T) and the two forms of RNA polymerase. We performed Dnase I protection assays to probe the heparin resistant complexes formed by either E S or E 70 at PaidB(WT) and PaidB(12C3 T) (Fig. 6). The heparinresistant complexes were separated on native polyacrylamide gel as described under "Experimental Procedures." As expected from its weak affinity for PaidB(WT), E 70 was only able to promote formation of a limited amount of the heparin-resistant complex (see the amount of uncut DNA at the top of the gel), and its interaction with the wild type aidB promoter resulted in very limited protection from DNase I attack (Fig. 6A). Consistent with the band shift assay, the 12C3 T mutation enhances formation of heparin-resistant complexes by E 70 to similar levels as E S , as indicated by the increased amount of uncut DNA at the top of the denaturing gel. No differences could be detected between the complexes formed by E S at either the wild type or 12C3 T mutant aidB promoter (compare densitometric scans in Fig. 6, B and C). In contrast, differences between the E S -and the E 70 -PaidB(12C3 T) complexes are detectable in the region immediately upstream of the Ϫ35 sequence (Fig. 6A, shown by the arrows). Both forms of RNA polymerase induce the formation of DNase I-hypersensitive bands between positions Ϫ36 and Ϫ38 and between Ϫ46 and Ϫ49. However, the relative intensity of the hypersensitive bands is clearly dependent on the form of RNA polymerase; whereas E S induces stronger hypersensitive bands at Ϫ36 and Ϫ46, binding by E 70 results in a stronger hyper-sensitive band at Ϫ38 (Fig. 6A; also compare the densitometric scan in Fig. 6C). This would suggest that the two forms of RNA polymerase interact differently with the Ϫ35 region of PaidB(12C3 T).
The results of both in vitro and in vivo experiments strongly suggest that, despite the C to T change at Ϫ12, PaidB(12C3 T) still favors transcription by E S . We measured the relative affinity of the two forms of RNA polymerase for PaidB(12C3 T) as their ability to compete for the promoter in a gel retardation assay performed in presence of heparin. As shown in Fig. 7, it is possible to distinguish between the two different complexes because the E 70 -promoter complex migrates slower in the gel retardation assay. In the E S -E 70 competition experiment, equal concentrations of 70 , S , and core RNA polymerase were mixed prior to the addition of the promoter DNA. At lower concentrations of RNA polymerase, binding to PaidB(12C3 T) by E 70 appears to be favored. Surprisingly, binding by E 70 also appeared to be at least equally as efficient as that by E S for the wild type aidB promoter, which is in contrast with our previous observations (Fig. 3). It is likely that this apparently higher binding affinity by E 70 might depend on a higher affinity of 70 for the core RNA polymerase (43,44), which would result in more efficient assembly for E 70 than for E S . However, at a higher RNA polymerase concentration the E S form is clearly favored at both PaidB and PaidB(12C3 T), confirming that the C12T mutant of the aidB promoter also retains preferential binding by E S .

A C Nucleotide at the12 Position Is a Determinant for Specific Transcription Initiation by E S at a Subclass of E S -dependent
Promoters-To understand whether a C nucleotide at the first position of the Ϫ10 promoter element (Ϫ12C) acts as a specific determinant for E S -dependent transcription at promoters other than PaidB, we analyzed the occurrence of either C or T as the first Ϫ10 sequence nucleotide in the known E S -dependent promoters. The list of 57 E S -dependent promoters used in our analysis is an revised version of the one presented in Ref. 9 and can be found in supplemental Table A (available in the on-line version of this article). The results summarized in Table  I indicate that T is by far the most represented nucleotide (77% of the total); however, C appears in 16% of the promoters at a much higher frequency than either G or A, suggesting a specific function for C as a promoter determinant. Interestingly, among E S -dependent promoters, a TG motif occurs at a significantly high frequency (higher than the random frequency of 6.25% expected for any dinucleotide) at positions between Ϫ15/Ϫ14 and Ϫ17/Ϫ16, i.e. between two and four nucleotides upstream of the Ϫ10 promoter element, consistent with role of the TG motif as promoter element for E S at these different locations (See Table I legend).
We targeted for mutagenesis the E S -dependent sprE(P2) and csgB promoters, both carrying a C as first nucleotide of the Ϫ10 sequence (Ϫ12C). Wild type and 12C3 T mutants of both sprE(P2) and csgB were tested in KMnO 4 reactivity assays in the presence of either E S or E 70 and compared with the aidB promoter (Table II). In this set of experiments, reactivity of all bands in the Ϫ13 to ϩ5 region was considered. Although to different extents, 12C3 T substitutions in the sprE(P2) and the csgB promoters stimulated open complex formation by E 70 while resulting in smaller or no increase in open complex formation by E S . Thus, the 12C3 T substitution resulted in a loss of selective E S -dependent transcription at both sprE(P2) and csgB promoters, similar to the results observed for PaidB.

DISCUSSION
In this report, we have investigated the question of which promoter features confer strong S dependence at the aidB promoter (PaidB). The aidB promoter possesses a recognizable Ϫ35 sequence (CTGTCA, 4 of 6 conserved nucleotides) and a Ϫ10 sequence, TGACCATACT, very similar to the proposed consensus for both E S and E 70 (Fig. 1A). However, the aidB promoter displays extreme dependence on E S both in vivo and in vitro (Refs. 21, 34, and 35; Figs. 1-2). We propose that dependence of PaidB on the E S depends on the presence of a C nucleotide instead of a canonical T at position Ϫ12. Additional features (location of the TG motif and the presence of a C nucleotide at the Ϫ13 position) further increase specific recognition by E S .
Ϫ12C Nucleotide-Substitution to a T of the Ϫ12C (12C3 T mutation) strongly improves transcription from PaidB by E 70 (Figs. 1 and 2). The 12C3 T mutation reduces the ratio of in vivo transcription in the MV1161 (wild type) strain to transcription in the MV2792 (rpoS) strain from 15-to 2.5-fold (Fig.  1) and the ratio of E S -to E 70 -dependent transcription in vitro from 5-to 1.5-fold (Fig. 2). E S specificity mediated by the Ϫ12C nucleotide does not appear to be a unique feature of PaidB. Indeed, a C nucleotide at the Ϫ12 position can be found in 16% (9 of 56 promoters, Table I) of E S -dependent genes in E. coli. A C to T mutation results in a loss of selective recognition by E S at the osmE promoter as measured by both in vivo and in vitro transcription experiments (20) as well as at the sprE(P2) and csgB promoters as determined by KMnO 4 reactivity assays (Table II). Thus, we propose the existence of two subclasses of E S -dependent promoters. A first subclass displays preferential interaction with E S but can also be recognized by E 70 . These promoters would be more likely to have a T as first nucleotide of the Ϫ10 promoter element, and their selective recognition by E S would depend on additional factors. A second subclass of promoters, displaying a C as first nucleotide of the Ϫ10 promoter sequence, would be exclusively recognized by E S in the absence of any additional factors.
The presence of a T at Ϫ12 improves formation of a heparinresistant E 70 -PaidB complex (Fig. 3), similar to what was observed previously (45), and allows promoter opening by E 70 (Fig. 4). The kinetics of open complex formation is stimulated by the 12C3 T mutation for E S also (Fig. 5), consistent with increased levels of E S -dependent transcription at the mutant aidB(12C3 T) promoter; however, S clearly displays more tolerance than 70 for a C nucleotide at the first position of the Ϫ10 sequence.
Despite the presence of a T at position Ϫ12 of PaidB(12C3 T), this promoter retains preferential interaction with E S as shown by direct competition assays (Fig. 7). The two forms of RNA polymerase form complexes of a different nature with PaidB(12C3 T) as determined by both the kinetics of KMnO 4 reactivity (Fig. 5) and DNase I protection assays (Fig. 6). Binding by E 70 results in strong KMnO 4 reactivity in the Ϫ11 to Ϫ5 region of PaidB(12C3 T) but a more limited reactivity at ϩ2, suggesting that, despite being favored by the presence of a T nucleotide at position Ϫ12, the E 70 -PaidB(12C3 T) complex is not fully proficient in promoter opening (Fig. 5). DNase I protection assays suggest different interactions between the two forms of RNA polymerase and PaidB(12C3 T) in and immediately upstream of the Ϫ35 region (Fig. 6), confirming previous observations at the synthetic "Ϫ35 con" promoter and at osmY (10,41). This preferential interaction of PaidB(12C3 T) with E S is likely to depend on two promoter elements, i.e. the TG motif at the unusual Ϫ16/ Ϫ15 location and the C at Ϫ13.
TG Motif-A TG dinucleotide placed at Ϫ15/Ϫ14, i.e. two and three nucleotides upstream of the Ϫ10 hexamer, is important for initial binding of E 70 to the so-called extended Ϫ10 promoters (23,24,26). Although the TG motif is present in the aidB promoter, it is located at the unusual Ϫ16/Ϫ15 location, i.e. three and four nucleotides upstream of the Ϫ10 box. At this position, the TG motif provides little or no contribution to E 70 -dependent transcription (28). In contrast, disruption of the TG motif at PaidB results in a Ͼ2-fold reduction in E Sdependent transcription, whereas no effect on E 70 -dependent transcription was detectable (Figs. 1 and 2). Repositioning of the TG motif to the canonical Ϫ15/Ϫ14 location (15G3 T/ 14A3 G mutation) results in no significant increase of E S -dependent transcription in vitro (Fig. 2). This suggests that E S , unlike E 70 , can recognize a TG motif even when placed 1-2  Ϫ12 and Ϫ13 positions We determined the number of times each nucleotide appeared in the Ϫ12 or Ϫ13 position in 56 E S -dependent promoters. The results are presented below as a percentage (e.g. T appeared in the Ϫ12 position in 77% of the promoters analyzed). We also analyzed the occurrence of a TG motif at certain positions in these same promoters. A TG motif appeared at positions Ϫ13/Ϫ14 in 3.5% of the promoters; at positions Ϫ14/Ϫ15 in 15.8% of the promoters; at positions Ϫ15/Ϫ16 in 12.3% of the promoters; at positions Ϫ16/Ϫ17 in 14% of the promoters; and at positions Ϫ17/Ϫ18 in 1.8% of the promoters.  nucleotides upstream of its usual location, possibly due to higher flexibility of region 2.5 in S . This observation is confirmed by the relative occurrence of the TG motif at alternative locations upstream of the Ϫ10 element in S -dependent promoters (Table I). Disruption of the TG motif (15G3 C mutation) did not prevent promoter opening by E S (Fig. 4), suggesting that the TG motif might affect another step of transcription initiation, possibly initial binding to PaidB.
Ϫ13C-The presence of a C nucleotide at position Ϫ13 is widely conserved among S -dependent promoters and has already been proposed to be a specific determinant for S (8,9). The aidB promoter displays this feature, which is indeed necessary for optimal E S -dependent transcription in vivo (Fig. 1). Deletion of the Ϫ13C or its substitution to a G nucleotide results in an overall reduction of PaidB activity and a loss of promoter specificity for E S (Figs. 1-2; Ref. 21). In particular, the 13C3 G change strongly favors E 70 -dependent transcription in vitro, confirming previous observations at the csiD and the fic "con" promoters (8,9). This mutation did not significantly affect E S -dependent transcription in vitro at PaidB (Fig. 2); it did, however, inhibit formation of E S -PaidB heparin-resistant complexes (Fig. 3) and open complex formation as determined by KMnO 4 reactivity (Fig. 4). Thus, a C nucleotide is needed for optimal transcription initiation and promoter opening by E S ; however, conditions used in standard in vitro transcription assays (excess of RNA polymerase, highly supercoiled plasmid as DNA template, and high nucleotide concentrations) might bypass the requirement for the interaction between E S and this promoter element.
Introduction of mutations disrupting E S -specific promoter elements such as the TG motif and the Ϫ13C element only results in a moderate, although remarkably reproducible, effect on promoter activity (ϳ2-fold; Figs. 1-4). This small effect would be consistent with the concomitant presence of many different promoter features (TG motif, Ϫ13C, Ϫ10 and Ϫ35 sequences, and the UP element; Fig. 1A), all contributing to promoter recognition and transcription initiation by E S .
Based on our observations, we propose TG(N) 0 -2 CCATA(a/ c)T as the consensus sequence for strictly E S -dependent promoters, where (N) represents one or two possible additional nucleotide(s) interposed between the TG motif and the Ϫ10 sequence. This sequence would be opposed to TGGTATAAT, which is optimal for E 70 -dependent transcription. A search for this consensus sequence using the Colibri data base (genolist. pasteur.fr/Colibri/genome.cgi) reveals that this sequence can be found within 200 base pairs upstream of 14 open reading frames or operons, in addition to aidB, in the E. coli chromosome (supplemental Table B, available in the on-line version of this article). Two of these genes encode flagellar biosynthetic proteins (flk and flgJ); the glf gene encodes UDP-galactopyranose mutase, which is involved in core lipopolysaccharide biosynthesis. The rplV gene encodes a ribosomal protein whose expression is dramatically reduced in the rpoS mutant MV2792 strain during stationary phase 2 ; the melR gene encodes a reg-ulatory protein that controls catabolism of the sugar melibiose. Finally, the remaining nine identified genes have unknown functions, suggesting that the role of a significant part of the E S regulon might yet have to be investigated.