The Whole Set of Constitutive Promoters Recognized by RNA Polymerase RpoD Holoenzyme of Escherichia coli

The promoter selectivity of Escherichia coli RNA polymerase is determined by the sigma subunit with promoter recognition activity. The model prokaryote Escherichia coli contains seven species of the sigma subunit, each recognizing a specific set of promoters. The major sigma subunit, sigma-70 encoded by rpoD, plays a major role in transcription of growth-related genes. Concomitant with the increase in detection of promoters functioning in vivo under various stressful conditions, the variation is expanding in the consensus sequence of RpoD promoters. In order to identify the canonical sequence of “constitutive promoters” that are recognized by the RNA polymerase holoenzyme containing RpoD sigma in the absence of supporting transcription factors, an in vitro mixed transcription assay was carried out using a whole set of variant promoters, each harboring one base replacement, within the model promoter with the conserved -35 and -10 sequences of RpoD promoters. The consensus sequences, TTGACA(-35) and TATAAT(-10), were identified to be ideal for the maximum level of open complex formation and the highest rate of promoter opening, respectively. For identification of the full range of constitutive promoters on the E. coli genome, a total of 2,701 RpoD holoenzyme-binding sites were identified by Genomic SELEX screening, and using the reconfirmed consensus promoter sequence, a total of maximum 669 constitutive promoters were identified, implying that the majority of hitherto identified promoters represents the TF-dependent “inducible promoters”. One unique feature of the constitutive promoters is the high level of promoter sequence conservation, about 85% carrying five-out-of-six agreements with -35 or -10 consensus sequence. The list of constitutive promoters provides the community resource toward estimation of the inducible promoters that operate under various stressful conditions in nature.


Introduction
The bacterial RNA polymerase core enzyme with the subunit structure a 2 bb9v is fully active in catalysis of RNA polymerization but is unable to initiate transcription from promoters. Transcription initiation from gene promoters requires an additional dissociable sigma subunit, which reversibly associates with the core enzyme to form the holoenzyme, and directs the core enzyme to recognize promoters for transcription initiation. Most bacteria encode multiple species of the sigma factor [1,2]. In Escherichia coli, seven species of the sigma subunit exist, each recognizing a specific set of promoters [1,3]. The intracellular levels of seven sigma subunits vary depending on cell growth conditions [3,4]. Sigma replacement is a simple mechanism of switching of the pattern of genome transcription [4,5], and the intracellular concentration of seven sigma factors is therefore a primary determinant of the pattern of genome transcription.
The sigma-70, encoded by the rpoD gene, is the primary and major sigma, which is responsible for transcription of most of the genes that are expressed in exponentially growing phase of E. coli [1,2,6,7]. The holoenzyme containing RpoD sigma recognizes in vitro a pair of hexanucleotide sequence elements, TTGACA (-35) and TATAAT (-10) which are situated at 10 and 35 bp upstream, respectively, of transcription initiation sites [2, 6, 8. 9], and a spacer DNA of approximately 17 bp in length separates these two hexanucleotide sequences. This consensus sequence of RpoD promoters was originally proposed based on in vitro transcription assays of some model templates by purified RNA polymerase [10,11]. Later the promoter sequences were determined for a variety of functioning promoters in vivo, one by one, by using ordinary molecular genetic approaches [12,13]. More recently high-throughput experimental systems such as ChIP-chip analysis of RNA polymerase-bound DNA sequences [14,15] and RNA-Seq analysis of whole sets of transcripts [16][17][18] have been employed for searching functioning promoters in vivo. In parallel, computational approaches have been employed to identify promoters relying on the consensus sequences predicted based on the known promoters. Up to the present time, more than 2,000 experimentally defined or computational predicted promoters have been identified, which are under the direct control of RpoD sigma (see the promoter databases such as RegulonDB [19] and EcoCyc [20]). As a result, the striking diversity appeared in the canonical promoter sequence, mainly arising from the detection of weak promoters, which are activated only in the presence of positive transcription factors (TFs).
The constitutive promoters have been defined as those that are active in vivo in all circumstances, but it is practically impossible to identify the whole set of constitutive promoters under various environmental conditions. Instead we propose in this study to define the ''constitutive promoter'' as the promoters that are recognized in vitro by RNA polymerase RpoD holoenzyme alone in the absence of additional supporting proteins with regulatory functions. For the identification of the whole set of constitutive promoters on the E. coli genome, we performed the Genomic SELEX screening system of genome DNA sequences recognized by the RNA polymerase holoenzyme containing RpoD sigma. The Genomic SELEX system was developed to identify DNA sequences recognized in vitro by DNA-binding transcription factors [21] and successfully applied for the identification of regulation targets of many TFs [22]. After SELEX-chip screening, a total of 2,071 sites were identified on the E. coli genome, which are recognized by the RpoD holoenzyme alone. The location of constitutive promoters within these RpoD holoenzyme-binding regions was then computationally identified using the consensus sequence, which was experimentally determined in this study using the in vitro mixed transcription assay [23][24][25]. The total number of constitutive promoters on the entire E. coli genome was thus estimated to be between minimum 492 and maximum 669. This number of constitutive promoters represents only about one-forth of the hitherto identified promoters on the E. coli genome, indicating that the rest of promoters listed in the promoter databases represents the ''inducible promoters'', which are activated in the presence of supporting TFs. One unique feature of the constitutive promoters is the high-level conservation of consensus sequences, TTGACA (-35) and TATAAT(-10), each being separated by a 17-bp spacer.
In the absence of regulatory proteins with repression activity, the constitutive promoters must be always expressed. Based on the Genomic SELEX screening of the binding sites of nucleoid protein H-NS along the E. coli genome, we also propose that the H-NS plays a major role in silencing of the unnecessary expression of constitutive promoters.

Results
Genomic SELEX screening of RpoD holoenzyme-binding sequences on the E. coli genome The constitutive promoters are transcribed in vitro by the RNA polymerase RpoD holoenzyme alone in the absence of supporting TFs. In order to identify the whole set of constitutive promoters on the entire genome of E. coli K-12 M3110, we first performed a mass-screening in vitro of the whole set of sequences that are recognized by the reconstituted RpoD holoenzyme. For this purpose, we prepared sigma-free core enzyme by passing the purified RNA polymerase three times through phosphocellulose column chromatography in the presence of 5% glycerol (note that sigma-core interaction becomes stronger in the presence of increasing glycerol concentration as used for prolonged storage of the holoenzyme [26]). The level of remaining sigma subunits was less than 0.1%, if any, as detected by both protein staining and immuno-staining with antibodies against each of all seven species of E. coli sigma subunits (RpoD, RpoN, RpoS, RpoH, RpoF, RpoE and FecI) (data not shown). The stoichiometry between core enzyme subunits was also checked by immuno-staining with antibodies against the core subunits, RpoA, RpoB, RpoB and RpoZ. The RpoD holoenzyme fully saturated with RpoD sigma was reconstituted by mixing this sigma-free core enzyme and 4fold molar excess of purified RpoD sigma, which alone does not bind to DNA.
For the identification of DNA sequences that are recognized by RpoD holoenzyme, we employed the Genomic SELEX screening system [21], in which a library of E. coli genome DNA fragments of 200-300 bp in length was used instead of synthetic oligonucleotides with all possible sequences used in the original SELEX method [27][28][29]. The multi-copy plasmid library of 200-300 bplong random DNA fragments was constructed from the E. coli K-12 W3110 genome [21]. The library used in this study contained 7-fold molar excess of the entire genome, and thus a single and the same sequence might be included in 7 different overlapping segments on average, thereby increasing the resolution of mapping of SELEX fragments. In each experiment of Genomic SELEX screening, the mixture of genome DNA fragments, which was regenerated by PCR from the genome DNA library, was mixed with 2-fold molar excess of the reconstituted RpoD RNA polymerase holoenzyme, and subjected to Genomic SELEX screening. DNA-holoenzyme complexes formed were recovered using the anti-RpoC antibody, which gave the highest level of RNA polymerase recovery among all the anti-core subunits. RNA polymerase-associated DNA was isolated from the antibody precipitates, amplified by PCR, and subjected to next cycles of SELEX. After two-cycles of SELEX screening, the final products of RpoD holoenzyme-bound DNA fragments were subjected to mapping on the genome using a DNA tilling microarray (Oxford Gene Technology, Oxford, UK) [30][31][32]. On the DNA tilling array used, the 60 b-long DNA probes are aligned at 105 bpintervals in the order of E. coli genome sequence, and therefore approximately 300 bp-long SELEX fragments should bind to two or more consecutive probes. This criterion was employed to avoid the background noise of non-specific binding of RpoD holoenzyme-bound DNA fragments to the tilling array.
The sequences with binding affinity to the RpoD holoenzme formed a number of peaks along the entire E. coli genome. By setting the cut-off level of 2.0% relative to the highest peak located within a spacer upstream of ssrA (SsrA smRNA) and downstream of smpB (SmpB trans-translation factor), a total of 2,701 RpoD homolenzyme-binding peaks were identified, of which 1,075 (40%) are located within 543 intergenic spacers (average 1.98 sites for each spacer) (Fig. 1A). On the other hand, a total of 1,626 (60%) peaks are located inside of 777 open reading frames (average 2.09 peaks per gene) (Fig. 1A). Since the majority of hitherto identified promoters are located within spacers, detailed search for the constitutive promoters was focused on the total of 1,075 peaks within 543 spacers.

Location of the constitutive promoters within the E. coli genome
The spacers containing RpoD holoenzyme-binding sites can be classified into three types (Fig. 1A): 404 peaks are located within 177 type-A spacer between bidirectional transcription units ( Table 1 for the whole list; see also Table S1); 583 peaks are located within 315 type-B spacers located upstream of one transcription unit but downstream of another transcription unit (Tables 2 and 3 for the whole list; see also Table S2); and 87 peaks are located within type-C spacers downstream of both transcription units of type-C spacers. Based on the transcription direction of flanking genes, the total number of constitutive promoters was predicted to range between minimum 492 (177 A-type plus 315 Btype) and maximum 669 (177x2 A-type plus 315 B-type).
Type-A spacers should contain at least two promoters for bidirectional transcription. The RpoD holoenzyme-binding sites identified in a total of 177 type-A spacers should represent promoters for one or both of bidirectional transcription. Close observation of the SELEX-chip pattern indicates that two RpoD holoenzyme-binding sites can be identified if the spacer is longer than 500 bp in length (Fig. 2). For instance, two peaks of RpoD holoenzyme binding were identified within a single and the same Type-A spacer between 755 bp-long csgD-csgB and between 1458 bp-long nanC-fimB (Fig. 2). Generally this-group promoters associated with stress-response genes are located within long spacers including the binding sites of a number of TFs such as in the spacers of 755 bp-long csgD-csgB, 920 bp-long lrhA-alaA and 1456 bp-long nanC-fimB. One typical example is the promoter for the csgD gene encoding the master regulator of biofilm formation, which is under the control of more than 20 TFs [23,33].
The binding of RpoD holoenzyme was identified in a total of 315 type-B spacers (Fig. 1A). The binding of RpoD holoenzyme alone to type-B spacers represents the presence of at least one constitutive promoter for one direction transcription. Even for this group of constitutive promoters, more than two RpoD holoen-zyme-binding peaks were identified for some spacers (Fig. 3), indicating the presence of multiple promoters for one and the same transcription units such as the cydA promoters within 847 bp spacer and the yobF promoters within 670 bp spacer. In good agreement with the presence of multiple peaks for the mngB-cydA type-B spacer, five promoters have been identified for the cydAB operon encoding cytochrome bd-1 terminal oxidase [34][35][36], of which at least two may be the constitutive promoters that function in the absence of activator TF. The collection of constitutive promoters within type-B spacers also includes a total of 40 internal promoters located within intergenic spacers of single operons (indicted by symbol ''Int'' in P column of Tables 2 and 3; and  Table S2). These internal promoters might play physiological roles under as yet unidentified circumstances. In fact a constitutive internal promoter within the rplKAJL-rpoBC operon has been identified [37], which should contribute the expression level control between four ribosomal proteins (L11, L1, L10 and L12) and two RNA polymerase subunits (RpoB and RpoC).
The total number of RpoD promoters (or the transcription initiation sites) listed in the current databases (RegulonDB and EcoCyc) are as many as 2,082, indicating that the majority of known promoters represent TF-dependent ''inducible promoters'' that are expressed only under the support of positive regulatory proteins (Fig. 1B). Among the total of 2,082 RpoD promoters (or transcription initiation sites) listed in the current version of promoter databases, the promoter sequences are described for a total of 582, of which 434 (75%) are located within the same spacers that contain the constitutive promoters (255 in type-A spacers and 179 in type-B spacers) (shown under black background of P columns in Tables 1, 2 and 3).

Identification of the consensus sequence of constitutive promoters recognized by the RpoD holoenzyme
For identification of the constitutive promoters within type-A and type-B spacers with the binding sites of RpoD holoenzyme, we performed in silico search using the consensus sequence of constitutive promoters. The current databases of E. coli promoters include both experimentally identified and computationally predicted promoters. The prediction of promoters has been performed using the canonical promoter, TTGACA-17 bp-TATAAT, which was originally identified using the in vitro transcription studies [10,11,38]. We then tried to identify the consensus sequence of constitutive promoters recognized in vitro by RpoD holoenzyme alone.
In order to experimentally confirm the consensus sequence recognized by RNA polymerase RpoD holoenzyme, we first constructed an ideal promoter with the complete set of consensus TTGACA (-35) and TATAAT (-10) sequences, separated by a spacer of 17 bp in length, starting from the lacUV5 promoter [23]. To identify the best RpoD promoter giving the highest activity and to examine the role of individual bases within two hexanucleotide sequences, we then constructed a total of 48 variant consensus promoters, each carrying one base replacement at each position of both -35 and -10 signals. For accurate measurement of the RNA product directed by each variant promoter, we employed the in vitro mixed transcription system [24,25], in which transcription of each variant promoter was carried out in the simultaneous presence of the ideal promoter with the complete consensus sequence added as an internal reference. The test promoter directed the synthesis of 42 b-long run-off transcript while the ideal promoter directed the synthesis of 22 b-long run-off transcript [23]. The same amounts of two promoter fragments were mixed and incubated with 10-fold molar excess of RpoD holoenzyme for various time periods to allow the formation of open complexes, Genomic SELEX search of RpoD holoenzyme-binding sequences was performed using the standard procedure [21]. RpoD holoenzyme-bound DNA fragments were recovered by immunoprecipitation using anti-RpoC antibody. SELEX fragments were isolated from the immuno-precipitates and subjected to mapping on the E. coli genome by using tilling DNA microarray as described previously [32,33].
[A] Location of the constitutive promoters. A total of 2,701 RpoD holoenzyme-binding sites were identified (see Fig. 1), of which 1,075 (40%) are located within intergenic spacers. On the basis of transcription direction of flanking genes, the spacers were classified into three types: type-A between bidirectional transcription units; type-B upstream of one transcription unit but downstream of another transcription unit; and type-C, downstream of both transcription units.
[B] Classification of the constitutive promoters. A total of 2,082 promoters have been identified and listed in the current versions of RegulonDB and EcoCyc databases, whereas the total number of constitutive promoters identified by Genomix SELEX screening ranges between minimum 492 and maximum 669, indicating that the majority of E. coli promoters listed in promoter database are TF-dependent inducible promoters. doi:10.1371/journal.pone.0090447.g001    Fig. 4. In each panel, the promoter activity is compared between four templates with different bases at the same position. Among the collection of -35 variants, the best promoter giving the highest activity of open complex formation (parameter-I) was identified for the consensus TTGACA sequence (Fig. 4A). This indicates that the consensus sequence of promoter -35 influences the binding affinity of RNA polymerase to the promoter in agreement with the previous estimation [23]. As to the promoter -10 signal, the best sequence giving the highest rate of open complex formation (parameter-II) was identified for the consensus TATAAT (Fig. 4B), indicating that promoter -10 influences the rate of promoter opening.
Using this experimentally confirmed consensus sequence, TTGACA-17 bp-TATAAT, of the constitutive promoter recognized by RpoD holoenzyme alone, we search for the location of constitutive promoters within both type-A and type-B spacers including RpoD holoenzyme-binding sites.

Unique features of the RpoD constitutive promoters
After sequence analysis of the entire genome of E. coli K-12, we realized that there is no ideal sequence of RpoD promoter with perfect matching to this consensus sequence. We then analyzed whether the constitutive RpoD promoters harbor unique sequences. By setting a rather severe screening condition of the sequence matching of more than 4 out of 6 bases for both -35 and -10 signals (total score, higher than 8; the highest score, 12) and with a spacer length of 17 plus/minus 2 (score 3 for 17 bp spacer, score 2 for 16 and 18 bp spacers, and score 1 for 15 and 19 bp spacers), a total of as many as 316 promoter sequences (89%) were identified among 354 predicted constitutive promoters within type-A spacer, and a total of 226 promoter sequences (82%) were identified among 276 predicted promoters excluding the internal promoters within type-B spacer. Overall the total amount of constitutive promoters with high-level (higher than 4/6) matching with the consensus RpoD promoter for both -36 and -10 signals is more than 85% (Fig. 5A-1). This is in sharp contrast with the collection of 582 experimentally defined promoters, in which the amount of promoters containing a high-level (4/6) agreement to both -35 or -10 signals are less than 40% (Fig. 5B-1). On average, only less than half of the 12 canonical bases of the -35 and -10 boxes are conserved among the experimentally identified promoters [39]. The length of spacers between -35 and -10 signals ranges from minimum 14 to maximum 21 bp. Only 10-20% efficient promoters have been proposed to include either -35 or -10 box that resembles the consensus with five-out-of-six bases (5/6 agreement).
Using a total of 669 sequences of the constitutive promoter with high-level of sequence conservation, the consensus sequence of RpoD promoter was reevaluated. The Logo plot, shown in Fig. 5A-2, indicates that: 1) the patterns of conserved bases of both -35 and -10 sequences are similar for both left and right genes with both type-A and type-B spacers (data not shown); and 2) a significant difference exists in the consensus sequence of both -35 and -10 signals between a total of 543 newly identified constitutive promoters and a total of 582 experimentally identified promoters. The Logo pattern generated from the alignment of a total of 582 experimentally identified E. coli promoters, shown in Fig. 5B-2, indicates that the level of sequence conservation is significantly higher for -10 than -35, and the conserved bases are limited, TT (positions 1 and 2) of -35 signal and TA (positions 1 and 2) and T (position 6) of -10 signal. This Logo pattern agrees well with that analyzed by Mitchell et al. [40] using the promoter set of RegulonDB [41]. In the case of constitutive promoters, -35 and -10 sequences are equally well-conserved ( Fig. 5A-2). Within -35 sequence, TT (position 1 and 2) and AA (positions 4 and 6) are well-conserved but G at position 3 and C at position 5 appear less important even for the constitutive promoters. On the other hand, all six bases of TATAAT within -10 signal are equally wellconserved, indicating their equal contribution to the promoter activity. The marked difference of the consensus sequence of experimentally identified RpoD promoters from that of constitutive promoters indicates that the promoter databases include a number of inducible promoters that were active under the support of positive TFs. A total of 1,075 RpoD holoenzyme-binding sites were identified within spacers on the entire E. coli K-12 W3110 genome. The constitutive promoters were predicted within type-A and type-B intergenic spacers (see Fig. 1A for classficiation). A total of 178 RNA polymerase RpoD holoenzyme-binding sites were identified within type-A spacers, which direct bidirectional transcription. Based on the gene orientation around these promoters, the genes and operons under the control of these promoters were estimated, that are

Constitutive promoters carrying the binding site of silencer H-NS
The constitutive promoters direct transcription by the RpoD holoezyme alone in the absence of positive transcription factors. Expression in vivo of the constitutive promoters must be repressed under conditions where the target genes are not needed. One possible transcription factor with the silencing activity is H-NS, one of the major core nucleoid proteins with functional dichotomy [42]. H-NS is known as a universal silencer for repression of a number of xenogeneic genes [43,44]. In parallel with the mapping of RpoD promoters, we identified a total of 987 H-NS-binding on the E. coli genome using the Genomic SELEX system [45]. Within a total of 492 spacers (177 type-A and 315 type-B) with RpoD holoenzyme-binding sites, H-NS binding sites were identified within 63 type-A (36%) (indicated as H-NS in column P of Table 1) and 77 type-B spacers (24%) (Tables 2 and 3).
The genome-silencing function of H-NS is attributable to its unique mode of DNA binding. H-NS recognizes and binds to intrinsically curved (or bent) DNA with AT-rich sequences, and such curved sequences are often located near promoters and also within transported DNA segments such as phage genomes [46,47]. Promoter-associated DNA curvature provides H-NS with the initial contact site, leading to form DNA-H-NS filaments via cooperative protein-protein interaction [48]. Since once H-NS binds within one spacer, it could be fully covered with H-NS by spreading from the initial binding sites through high level of protein-protein cooperativity. Thus, if H-NS binds within one type-A spacer, it influences the silencing effect to promoters for both of the bidirectional transcription. Taken together we estimated that a total of 203 promoters (63x2 + 77) or 31% among a total of 669 constitutive promoters could be under the control of genome silencer H-NS, indicating that one third of the constitutive promoters are subject to silencing by H-NS.
Among the total of 203 possible targets of H-NS silencing, a total of 20 promoters are known to be under the control of H-NS. Up to the present time, however, the involvement of H-NS in promoter regulation has not analyzed seriously because this bifunctional nucleoid protein is abundant, playing both architectural and regularatory roles. The involvement of H-NS in transcription regulation of a specific gene has only been identified Among the total of 1,075 RpoD holoenzyme-binding sites, 181 are located within type-B spacers upstream of left-side genes and downstream of right-side genes, indicating that these promoters direct leftward transcription. The genes and operons under the control of these 181 promoters were estimated, of which 16 represent putative internal promoters. Descriptions and symbols are as in Table 1 during the systematic search for transcription factors. For instance, the csgD (the master regulator of biofilm formation), the ndh gene (NADH:ubiqinone oxidoreductase), and the gadX gene (the stressresponse regulator for acid resistance) are all under the control of more than 10 transcription factors including H-NS [4,22]. By using the newly developed PS-TF (promoter-specific transcription factor) screening system, we have identified the involvement of H-NS for regulation of a set of genes that are all under the control of multiple transcription factors [49]. These findings altogether suggest that the involvement of H-NS in transcription silencing of a number of constitutive promoters will be established once PS-TF screening is performed.

Definition of the constitutive promoters
The pattern of genome transcription in E. coli is determined by controlling the utilization of a limited number of RNA polymerase [4,22]. High-throughput systems have been developed for identification of the RNA polymerase distribution within the genome. For instance, large-scale mapping of the functioning promoters in vivo on the E. coli genome has been performed by the chromatin immuno-precipitation and microarray (ChIP-chip) analysis with specific antibodies against RNA polymerase subunits [14,15]. To prevent moving RNA polymerase along DNA, E. coli cells were treated with rifampicin, which binds to the RpoB subunit of RNA polymerase and blocks transcription initiation, thereby fixing the initiated RNA polymerase on promoters [50]. For identification of the distribution pattern of functioning RNA polymerase, RNA-Seq analysis of high throughput sequencing of transcriptomes is becoming the method of choice [16][17][18]. Concomitant with the increase in the number of promoters detected under various stressful culture conditions, the variation of promoter sequence is expanding. The fluctuation of functioning promoters is attributable to the modulation of promoter selectivity of RNA polymerase by sigma factors and transcription factors with positive and negative regulatory functions [3,4].
The constitutive promoter has been considered to represent a set of promoters associated with the genes constitutively expressed in vivo in all circumstances. Based on this definition, however, it is practically impossible to identify the whole set of constitutive promoters of E. coli under various environmental conditions. At present, the whole sets of promoters determined by high-through put procedures (see above) are accumulating, but the experiments have been performed using different E. coli strains, under different culture conditions and using different experimental systems. In addition, even in laboratory culture conditions, it now turned clear that the steady state of cell growth does not exist and moreover, the genome expression pattern varies between individual cells within the same culture. In contrast, the whole set of constitutive promoters can be identified in vitro because in the case of E. coli, the faithful transcription can be established in vitro using purified RNA polymerase and pure DNA template under defined conditions. To avoid the complexity arisen from in vivo determination of the functioning promoters, an attempt was then made in this study to identify the whole set of the constitutive promoters using an in vitro system. Based on the results, we propose to revise the definition of the constitutive promoter as the promoter that is recognized and transcribed by the RNA polymerase holoenzyme alone in the absence of supporting TFs.

SELEX-chip search for the constitutive promoters
For this purpose, we used the improved method of Genetic SELEX screening [21], which has been successfully employed for identification of binding sites of a number of TFs on the E. coli genome [22]. As a result, the number of regulation targets markedly increased even for the TFs with known regulatory functions. For instance, the number of regulation targets increased more than 2.5 fold from 150 to 350 even for the best characterized transcription factor, CRP (cAMP receptor protein) or CAP (catabolite activator protein) [51]. This experimental system is particularly useful for short-cut estimation of the regulation targets of uncharacterized TFs including YbjK (renamed to RcdA), YcdC (renamed to RutR), YcjZ (renamed to PgrR), YdhM (renamed to NemR) and YgiP (renamed to Dan) [22]. The Genomic SELEX screening system has also been successfully employed for detection of the alteration of promoter-recognition properties of transcription factors after phosphorylation (in the case of two-component systems) or interaction with effector ligands. For instance, the selection of regulation targets of SdiA, a regulator of genes for cell division and differentiation, was found to alter differently in the presence of each of three homoserine lactone analogs, the QS signals [52].
As an extension of screening of the whole set of RpoD promoters, we have successfully performed the screening of promoters recognized by RNA holoenzymes containing minor sigma factors, RpoS, RpoH, RpoF, RpoE and FecI. In the case of RpoN sigma factor, it requires enhancer-binding proteins such as NtrC and for formation of stable RNA polymerase-promoter complexes. Results will be described elsewhere.

Physiological roles of the constitutive promoters
The genes under the control of constitutive promoters are supposed to be expressed constitutively although unnecessary genes are subject to repression by silencers (see the H-NS chapter). One short-cut interpretation is that the essential genes are under the control of constitutive promoters. After systematic deletion of the genome, the total number of essential genes in the E. coli genome has been minimized to generate the minimal genome consisting of 302 essential genes ( [53]; also see PEC database). Among 302 essential genes, 70 (23%) were identified to be under the control of constitutive promoters. Noteworthy is that the expression of some essential genes should be under the control of constitutively expressed positive transcription factors. In fact, we have identified the presence of approximately 100 species of E. coli transcription factor throughout cell growth.
Marked variation of the promoter sequences listed in databases must have been arisen from several different factors: 1) the list of RpoD-dependent promoters include those recognized by RpoS (and possibly other minor sigma factors), of which the promoter recognition properties overlaps with RpoD [54,55]; 2) the list also includes promoters that are functional in vivo only under the support of positive TFs [56][57][58]; and 3) a variety of promoter-like sequences have been identified in silico to be promoters. In fact approximately 60% of the RpoD holoenzyme-binding sites are located inside open reading frames (see Fig. 1A). Binding of RNA polymerase on some open reading frames has been recognized [59]. The unexpected high number of RpoD holoenzyme-binding sites inside open reading frames raises a possibility of an as yet unidentified functional role(s) for the RpoD holoenzyme. These promoter-like sequences may contribute transcription initiation from internal promoters and/or blocking the migration of elongation complexes. Clustering of promoter-like sequences within the E. coli genome were predicted by Collado-Vides and colleagues [60] while Ozoline and colleagues proposed the presence of as many as 78 'Promoter Islands' [59,61]. These promoter-like sequences could form transcriptionally inactive   complexes with RNA polymerase but might contribute to increase local concentrations of RNA polymerase on the genome. Some of the constitutive promoters with high-level of conservation of the consensus sequence are located within long spacers with long UTR sequences but lacking protein-coding sequences of reasonable sizes. In these cases, it would be worthwhile to test as yet unidentified regulatory sRNAs [62]. Noteworthy is that such long spacers including the constitutive promoters often correspond to the Promoter Islands.

The consensus sequence of constitutive promoters
The canonical model of RpoD promoters consisting of two hexanucleotide sequences, TTGACA -35 signal and TATAAT -10 signal, each being separated by 17 bp linker was originally identified using in vitro transcription of some model templates by purified RNA polymerase [10,11]. Here we identified the consensus sequence of as many as 669 constitutive promoters. The most significant feature of constitutive promoters is the highlevel conservation of canonical TTGACA(-35)-17bp-TATAAT(-10) sequence. We also identified the roles and -35 and -10 signals and the conservation level of each base within these two hexanucleotide sequences. The promoter -35 TTGACA signal plays a key role in binding the RNA polymerase (see Fig. 4) but G at position 3 and C at position 5 are relatively less important for the constitutive promoters, suggesting that both play roles in TFdepending inducible promoters. The promoter -10 TATAAT signal plays a major role in promoter opening (see Fig. 4) in agreement with the previous proposals [63,64], but the novel finding is that all six bases of TATAAT are equally important for this -10 signal function, supporting the hypothesis that the cooperativity of energy threshold, but not interaction of individual bases with RpoD, are important factors guiding the dynamics and selectivity of promoter open complex formation [65].

Regulation of the conserved promoter
The constitutive promoters must be repressed in cases when the genes under their control are not necessary. A group of silencing proteins play roles in preventing the potentially harmful effects of uncontrolled expression of the constitutive promoters. In E. coli, the H-NS family proteins are the major players of anti-silencing [44,45,66]. The high-level of overlapping was observed in the distribution between the constitutive promoters (see Tables 1, 2 and 3) and the binding sites of silencer H-NS [45]. A total of 203 (30%) of the constitutive promoters were predicted to be under the control of H-NS silencer, but silencing proteins for other 70% promoters remain unidentified. The spectrum of silencing targets by H-NS should be modulated by interaction with the members of Hha/YdgT family of small-sized co-regulators [67,68]. Possible involvement of other growth condition-specific nucleoid proteins in silencing such as Dps (DNA-binding protein in starved cells) [69] and Dan (DNA-binding protein under anaerobic conditions) [32] awaits further studies.
It should be noted that the constitutive promoters are also subject to activation by positive TFs for enhanced expression of the target genes. For instance, the csgD gene encoding the master regulator of biofilm formation carries one of the Type-A constitutive promoters, which is controlled by as many as 20 positive and negative TFs [22,33]. The level of constitutive csgD promoter is higher than that of csgB promoters directing transcription toward opposite orientation (see Fig. 2A[b]), but Among the total of 1,075 RpoD holoenzyme-binding sites, 154 are located within type-B spacers upstream of right-side genes and downstream of lest-side genes, indicating that these promoters direct rightward transcription. The genes and operons under the control of these 154 promoters were estimated, of which 18 represent putative internal promoters. A total of 9 genes encoding transcription factors are indicated by star symbols (*). The essential genes listed in the PEC database are underlined within the operons. The promoter sequence with complete match with the canonical promoter (see Fig. 4) is shown in bold and italic while the promoter sequence with 5-out of-6 match is shown in bold. once the regulator CsgD is produced, it antagonizes H-NS silencer and the expression of the csgBAC operon is markedly enhanced, leading to production of curli fimbriae for biofilm formation.
In conclusion, we classified E. coli promoters into the constitutive promoters recognized by RNA polymerase holoenzyme alone and the transcription factor-assisted inducible promoters. Using the information of RpoD holoenzyme-binding sites identified by Genomic SELEX screening system, we predicted a total of 669 constitutive promoters with high-level conservation of the promoter consensus sequences. This finding indicates that the majority of hitherto identified promoters represent the TF-dependent inducible promoters.

Materials and Methods
Bacterial strains and plasmids E. coli K12 W3350 type-A [70] was used for purification of RNA polymerase and the template DNA for Genomic SELEX screening of RpoD promoters. E. coli BL21(DE3) was used for the expression and purification of sigma and core enzyme subunit proteins. Expression plasmids for the core enzyme subunits (pRpoD, pRpoA, pRpoB and pRpoC) and all seven sigma subunits (pRpoD, pRpoS, pRpoN, pRpoH, pRpoF, pRpoE and pFec) were constructed by ligating the respective coding sequences, which were prepared by PCR amplification of the E. coli K12 W3350 type-A genome DNA as template, into pET21 expression vector essentially according to the standard procedure used for expression of all seven sigma subunits and all 300 transcription factors in this laboratory [71,72].
Purification of core RNA polymerase RNA polymerase was purified from log-phase cells of E. coli K-12 W3350 by the standard procedure [26]. Separation of the core and holoenzymes was performed by passing the purified RNA polymerase through P11 phosphocellulose column in the presence of 50% glycerol [26]. To remove trace amounts of the core enzyme-associated sigma factors, the purified RNA polymerase in the storage buffer containing 50% glycerol was dialyzed against the same buffer but containing 5% glycerol and fractionated by phosphocellulose column chromatography in the presence of 5% glycerol [26]. The level of remaining sigma factors was less than 0.1%, if any, as checked by immuno-staining with antibodies against each of seven sigma factors.

Purification of core and sigma subunits
The core enzyme subunits (RpoA, RpoB, RpoC and RpoZ) were expressed using the respective expression plasmids and purified by two cycles of column chromatography through DEAE (DE52) and P11 phosphocellulose [26]. Sigma subunits were expressed and purified by ion-exchange column chromatography through DE52 and P11 followed by Sephacryl S-300 gel filtration column [26]. The purified sigma and core subunit proteins were more than 99% pure as judged by protein staining of SDS-PAGE gels.

Preparation of antibodies
Antibodies against RpoD sigma and core enzyme subunits were produced in rabbits by injecting purified sigma proteins. The protocol for antibody production was raised following the Ethical Guidelines proposed by the Science Council of Japan and the Japanese Government, and approved by the Committee on the Ethics of Animal Experiments in the Animal Laboratory of Mitsubishi Chemical Medience Co. (Uto, Kumamoto, Japan). Antibodies against each RNA polymerase proteins were produced in two rabbits, and after examination of antibody activity using immune-blot analysis, the batch of higher activity was used in this study. Anti-RpoD, anti-RpoC used in this study did not cross-react with each other.

Genomic SELEX screening of RpoD holoenzyme-binding sequences
The genomic SELEX method was carried out as previously described [21]. A mixture of DNA fragments of the E. coli K-12 W3110 genome was prepared after sonication of purified genome DNA, and cloned into a multi-copy plasmid pBR322. In each SELEX screening, the DNA mixture was regenerated by PCR. For SELEX screening, 5 pmol of the mixture of DNA fragments and 10 pmol RNA polymerase RpoD holoenzyme were mixed in a binding buffer (10 mM Tris-HCl, pH 7.8 at 4uC, 3 mM magnesium acetate, 150 mM NaCl, and 1.25 mg/ml bovine serum albumin) and incubated for 30 min at 37uC. The DNA-RNA polymerase mixture was treated with anti-RpoC antibody and DNA fragments recovered from the complexes were PCRamplified and subjected to next cycle of SELEX for enrichment of RNA polymerase-bound DNA fragments.
For SELEX-chip analysis, DNA samples were isolated from the DNA-protein complexes at the final state of SELEX, PCRamplified and labeled with Cy5 while the original DNA library was labeled with Cy3. The fluorescent labeled DNA mixtures were hybridized to a DNA microarray consisting of 43,450 species of 60 b-long DNA probe, which are designed to cover the entire E. coli genome at 105 bp interval (Oxford Gene Technology, Oxford, UK) [27][28][29]. The fluorescent intensity of test sample at each probe was normalized with that of the corresponding peak of original library. After normalization of each pattern, the Cy5/Cy3 ratio was measured and plotted along the E. coli genome.

Mixed transcription assay in vitro
The promoter sequence of 195 bp-long lacUV5 template and its 39 truncated 175 bp-long DNA was modified to the consensus sequence, TTGACA(-35)-17 bp-TATAAT(-10), each producing run-off transcripts of 42 and 22 nucleotides in length, respectively. Starting from these DNA fragments with the ideal promoter, the complete set of variant consensus promoters, each containing one base substitution at all positions of the hexanucleotide sequences of promoter -35 and -10, was prepared as described previously [23]. The test promoters of variant consensus promoters directed the synthesis of 42 b-long RNA while the reference template containing the original consensus promoter directed the synthesis  [23][24][25]. In brief, a mixture of these two templates was preincubated with RpoD holoenzyme for various times (0.5, 1.0, 2.5, 5.0, 10, 15 and 30 min) at 37uC in the standard transcription assay buffer, and then a mixture of substrates and heparin was added to allow a single-round transcription for 15 min. Heparin inactivates RNA polymerase that was not involved in open complex formation. RNA products were separated by electrophoresis on 10% PAGE containing 8.3 M urea. The amount of RNA was determined by measuring the intensity of 32 P radioactivity and corrected for the U content of each transcript.

Promoter sequence analysis
The complete genome sequence of E. coli K-12 MG1655 (U00096.2) was used for the promoter sequence analysis. The perl scripts for finding the candidate sequences of promoters include the following functions: 1) extraction of every 6 bases in the genome sequences with sliding 1 bp; 2) comparison of the 6 base sequences with TTGACA (-35) and calculation of the score by setting one match scoring 1 point, one mis-match scoring 0 point, and without gap; 3) comparison of the 6 bases with TATAAT (-10) under the same conditions; and 4) extraction of a pair of -35 and -10 hexanucleotide sequences by setting the best-match score of 12 (6 at -35 and at -10 signal); 6) extraction of a pair of -35 and -10 signals with a spacer length of 17 plus/minus 2, giving the score of +3 for 17 bp, +2 for 16 and 18 bp; and +1 for 15 and 19 bp, respectively. Thus, the maximum score for the best-match promoter in both -35 and -10 sequences separated by 17 bp spacer is 15.
The Logo pattern analysis of promoter sequences was performed using the established sequence logo generator [73].

Supporting Information
Table S1 RNA polymerase RpoD holoenzyme-binding sites on the E. coli genome. A total of 1,075 RpoD holoenzyme-binding sites were identified within spacers on the entire E. coli K-12 W3110 genome. The binding sites identified within intergenic spacers were classified into Type-A, Type-B and Type-C (see Fig. 2 and Table 1). The constitutive promoters were predicted based on the location of the RpoD holoenzyme-binding sites.
[S1A] Promoters within type-A spacers. The genes under the constitutive promoters on left-and right-sides of type-A intergenic spacers and their functions are described. The genes shown under filled background encodes DNA-binding transcription factors (TFs), each regulating a set of target genes. The spacers under grey background contain two distinct peaks of holoenzyme binding. The positions of these genes on the genome are described on the left-end column while the levels of SELEX-peaks relative to the highest peak are described on the right-end column.
[S1B] Promoters within type-B spacers. The genes under the constitutive promoters on either left-or right-sides of type-B intergenic spacers and their functions are described. The positions of these genes on