The DNA-dependent, multisubunit RNAP of E. coli is an evolutionarily-conserved protein which shares functional and structural relatedness with RNAPs of eubacteria, archaebacteria, yeast, and mammals [12–14]. The catalytically competent core has a conserved subunit composition of α2ββ′ω with a molecular mass of ∼380kDa, and is capable of catalyzing DNA-dependent RNA synthesis, RNA hydrolysis and pyrophosphorolysis. Binding of the bacterial-specific initiation factor σ converts core to holoenzyme, which is capable of specific promoter recognition and efficient transcription initiation [15,16]. All prokaryotic organisms express one or more σ-like factors. Use of alternative σ's allows RNAP to recognize different classes of promoters, thus affording organisms the specificity and selectivity in transcription process required for optimal cell growth [16,17]. In E. coli, which expresses seven σ factors, the σ70-associated holoenzyme (Eσ70) transcribes the bulk of its housekeeping genes, including those of the lac operon.

A transcription cycle carried out by RNAP proceeds through three stages: initiation, elongation, and termination, all of which are targets of regulation. During initiation, RNAP holoenzyme binds specifically to two conserved hexamers in the promoter at nucleotide (nt) positions −35 and −10 relative to the transcription start site (+1) to form a closed promoter complex (RP_c). In a process involving several intermediates, RP_c converts to a stable open promoter complex (RP_o) in which DNA duplex becomes unwound around −10 region (from −12 to +3). In the presence of rNTPs, RP_o begins to synthesize and release short (2–12 nts) RNA products (‘abortive initiation’) without leaving the promoter [18]. After several rounds of abortive initiation, the initiation complex (RP_i) enters elongation stage. This transition (‘promoter escape’) is marked by a significant conformational change [18–20], leading simultaneously to loss of RNAP-promoter contacts, possible σ-dissociation [21–23], and formation of a highly processive ternary elongation complex (TC) [18,20]. Elongation by TC continues until it encounters a termination signal encoded within the DNA sequences resulting in irreversible dissociation of core, DNA and RNA.

Eσ70 recognizes two types of promoters, the so-called −35 and extended −10 promoters [24]. Many promoters, including Plac, belong to the former, having both the −10 and −35 hexamers. The consensus sequence and positions of −35 and −10 hexamers are ⁻³⁵TTGACA⁻³⁰ and ⁻¹²TATAAT⁻⁷ [24–26]; the hexamers are separate by a 16–18 base-pair-long spacer region of nonspecific sequence [27]. The extended −10 promoters do not have any discernable −35 element, a defect that is functionally compensated by the presence of a dinucleotide 5′-TG-3′ (‘extended −10 element’) at positions −15 and −14 [28]. If present, the extended −10 element can also improve the initiation efficiency at −35 promoters. The −35, −10 and extended −10 elements are involved in direct and sequence-specific interactions with Eσ70 during RP_c formation, and are therefore major determining factors in establishing the equilibrium binding constant for RNAP-promoter interaction and the rate of RP_c formation [29]. Some E. coli promoters possess additional cis-element located immediately upstream to the −35 element (nt positions −40 to −60) called the ‘UP-element’ [30,31]. The UP-element, which can be recognized by the presence of an AT-rich sequence, facilitates RNAP binding through its interaction with αCTD of RNAP, and stimulates the intrinsic transcription by up to two orders of magnitude [30,31]. Additional DNA sequences located in and around the promoter can compensate for weak −10 and −35 elements and affect the overall promoter strength. These auxiliary promoter elements, which include the −15 enhancer, discriminator region (DSR), and initial transcribed sequences (ITS), were shown to affect the rate of RP_c formation and the efficiency of DNA melting, abortive initiation, and promoter escape [29,32–35].

In the absence of external regulatory input, many naturally occurring promoters, including Plac, are relatively weak due to their non-consensus sequence elements and/or suboptimal spacer lengths. However, many prokaryotic promoters are programmed to respond to a variety of regulatory signals that modulate their activities by either increasing or decreasing the rate of productive initiation. In most cases, the signal entails a sequence specific communication between regulatory protein and its cognate binding site located within, near, or at some distance from the target promoter [2,36]. Plac is a prime example of the promoters that respond to both negative and positive regulatory inputs. In the presence of glucose, lac repressor (LR) binds to the operator sites in the lac promoter (Plac) region and prevents RNA polymerase (RNAP) from binding to Plac [37]. The repression is removed by lactose, which binds to LR and causes its dissociation from DNA, thus allowing RNAP to bind and initiate transcription from Plac. During glucose starvation, Plac is positively regulated in response to elevated intracellular levels of cAMP by catabolite activator protein, CAP. The CAP-cAMP complex binds to CAP binding sites upstream of Plac, recruits RNAP, and facilitates transcription initiation from Plac [38].

2.2.1 Escherichia coli

RNAP is the most extensively characterized bacterial RNAP, both genetically and biochemically. However, the structure of this enzyme determined by cryo-electron microscopy (EM) has a relatively low resolution of ∼15Å [39]. The atomic-resolution (high-resolution) crystal structures have been obtained for Thermus aquaticus (Taq) core and Thermus thermophilus (Tth) holoenzyme, at 3.3 Å and 2.6 Å, respectively [3,4]. The subunits of E. coli and Taq/Tth enzymes share substantial sequence homology and are functionally similar [3,4,14,40,41]. Therefore, the structural data obtained for Taq/Tth RNAP can be readily applied to E. coli enzyme. According to available structural data, Taq and Tth RNAPs share the similar crab claw-like shape, of which the top and bottom pincers are made up of the two largest subunits, β and β′ (Fig. 1a and b). The pincers are joined at the back by the N-terminal domains of asymmetrically placed α-subunit dimer (αI- and αII-NTD). The ω subunit is located near the bottom pincer, wrapped around the β′ C-terminus. In all structures, the internal space of the protein between the pincers is intersected by three channels: the main channel with a diameter of 20–27 Å, which accommodates the double-stranded DNA and the DNA/RNA hybrid, and two minor channels that branch off from the major channel to form the upstream-facing ‘RNA exit channel’ and the downstream-facing substrate-accessible ‘secondary channel’. The minor channels are ∼10–12Å in diameter and serve as exit pathways for the single stranded 5′-terminal RNA and the 3′-terminal backtracked RNA, respectively [3–6,14,42]. The active center of the enzyme with a catalytic triad of β′ Asp residues holding two essential Mg²⁺ ions is located on the back wall of the primary channel [3–6,42]. The two pincers near the active center are connected by the β′ F-bridge α-helix, which joins the flexible β′ G-loop element to form the wall of the secondary channel (Fig. 1a and b). The RNA exit channel walls are made of the upstream portions of β and β′ pincers (the clamp) including the β′ ‘rudder’, ‘lid’ and the N-terminal ‘Zn-finger’ elements, and the β ‘fork loop’ and flexible ‘flap’.

2.2.2 Non-conserved domains

E. coliσ70 and σ70-like factors of other bacteria share four regions of sequence homology designated 1 to 4, which are further divided into subregions [16,17] (Fig. 2). All conserved regions have been implicated in either σ-core or σ-DNA interactions ([45,48] and references therein). In Taq and Tth holoenzyme structures, the σ-subunit is visible as a V-shaped structure partially wedged between the upper and lower pincers (i.e. β/β′ subunits) of core on the upstream face of the enzyme [4,5]. The crystallographically resolved portion of σ comprises four structural domains, σ2,σ3, linker domain LD and σ4, connected by short flexible linkers. These four domains contain conserved regions 1.2, 2.1–2.4; 3.0–3.1; 3.2; and 4.1–4.2, respectively. The three α-helical domains σ2,σ3 and σ4 are located on the enzyme's surface (Figs. 1b and 2), stretched over the upstream opening of the primary channel, while σLD is buried inside the primary channel (Fig. 2). σLD forms a hairpin loop that approaches the catalytic pocket, and emerges underneath the β-flap via RNA exit channel. In the available holoenzyme structures, σ2 and σ4 domains are located ∼70Å apart, which is an appropriate distance for these domains to contact, respectively, the −10 and −35 elements of the promoter DNA in RP_c. σ3 is positioned to interact with the extended −10 region of the promoter and the −15 enhancer element.

The extreme N-terminal portion of σ polypeptide (σ1–73), which includes poorly conserved region 1.1, is not resolved in either holoenzyme, or RNAP-DNA binary complex, or free σ structures. σ1.1 possesses an autoinhibitory function: it obscures the DNA binding regions of free σ before it binds core ([49] and references therein). It also facilitates the RP_o formation and transcription initiation at some promoters, while inhibiting initiation at others [50]. Additionally, σ1.1 may be involved in the initial σ binding to core by interacting with β flap domain [51].

Recent biochemical and biophysical evidence suggest a multistep and cooperative process of σ-core binding [45,48,49,51,52], which is characterized by a KD in the range of 10⁻⁹ M. Such high binding affinity derives from multiple independent interactions between discrete domains of σ and different parts of the core. However, most of the potential contacts in σ-core interface, including electrostatic (salt bridges), polar (hydrogen bonds) and non-polar (hydrophobic and van der Waals) interactions are relatively weak and distributed over a large area [45]. For the most part, these contacts are limited to the β and β′-subunits of core. The strongest interaction is observed between σ2 and β′ coiled-coil domain (β′540–585), which serves as the major σ docking site. Less strong interaction is observed between β-flap and σ4 [53], and between σ3 and β region I [45] (Figs. 1b and 2). In the presence of specific activators, σ4 also interacts with α-CTD [54].

2.3.1 Conformational flexibility

Structural information on how RNAP recognizes and binds promoter DNA was gleaned from two crystallographic studies: the 2.4-Å-resolution structure of Taq σ4 in complex with −35 element DNA (from position −26 to −37) [9], and the 6.5-Å-resolution structure of Taq holoenzyme binary complex with fork-junction promoter DNA [8], which partially mimics the RP_o. The latter complex contained ds DNA from position −12 to −45, and the ss nt-DNA from −11 to −7. Complemented with vast biochemical, biophysical and genetic data accumulated in the last 20 years, these studies led to construction of structural models of binary RNAP-DNA complexes RP_c and RP_o [8,49].

2.4.1 RP_c

In RP_c, the ds promoter DNA lies on the surface of holoenzyme, outside the RNAP active-site channel (Fig. 3a). The RNAP-bound ds DNA appears to be bent at three places: at position around −25, where DNA may bend or kink by ∼8° to accommodate variable spacer length [9], at the −35 element region, where a ∼36° bending is induced by insertion of σ4 helix-turn-helix motif into the major groove [9], and at further upstream −45 region, where αCTD-DNA interaction may take place [55]. The DNA bending at −35 region may be important for a proper orientation of DNA towards αCTD and for binding upstream transcription activators [10,11].

2.4.2 RP o

Unlike the ideal/consensus promoter DNA used in the structural studies and modeling of RP_c and RP_o, the P_lac of the lac operon in E. coli deviates significantly from canonical promoter [24–26]. These deviations include substitutions of consensus G for T at −34 in the −35 element (TTTACA), and AA for GT at −9/−8 in the −10 element (TATGTT). Plac also has an 18 bp-long spacer, which is one bp longer than the optimal 17 bp-length. Inspection of structural data reveals how these changes might affect the interactions between promoter elements and σ subunit that are essential for transcription initiation. Specifically, the loss of −34G/C base pair recognition by E410 of σ4 together with suboptimal spacer would cause substantial decrease in initial promoter binding by RNAP and the rate of RP_c formation [29]. Additionally, the non-consensus −10 element might destabilize the interactions of NT strand bases with aromatic residues (F248, and Y253) of σ region 2.3 resulting in decreased efficiency of DNA melting and RP_o formation. The promoter mutations that increase the activity of Plac include compensatory substitutions in −35 and −10 elements, insertions that alter the spacer length, and mutations in the −15 enhancer region [29,32].

In vivo, transcription initiation from Plac is stimulated under conditions of positive regulation by transcription activator protein CAP. Although activation by CAP principally affects the RNAP-binding step at Plac, [55] it may also exert a stimulatory effect on the rates of RNAP isomerisation, RP_o formation, and even promoter escape ([32,35] and references therein).

During transcription activation, CAP-homodimer complexed to its effector, cAMP, specifically binds to its cognate 22 bp binding site centered at position −62 of Plac and bends the DNA 80° (Fig. 3c). CAP then recruits RNAP by interacting with one of its two αCTDs, which can be either αCTDI (CTD of the α subunit that interacts with β) or αCTDII (CTD of the α subunit that interacts with β′) ([38] and references within). The CAP-αCTD interaction is mediated by ‘activating region 1’ (residues 156–164, 209 of E. coli CAP) of the downstream subunit of CAP dimer and the ‘287 determinant’ of αCTD (residues 285–290, 315, 317, and 318 of E. coli αCTD). αCTD, in turn, interacts nonspecifically with the ribose-phosphate backbone of the DNA minor groove immediately downstream of CAP-binding site on DNA centered at position −43 [10,55]. These interactions, which are mediated by the ‘265 determinant’ of αCTD (residues 265, 294, 296, 298, 299, and 302), are relatively weak in the native Plac. However, if nonspecific sequences at −43 region are replaced by A/T-rich UP element, a stronger binding by αCTD is observed and more efficient transcription activation is elicited. Lastly, transcription activation by CAP requires specific interactions between αCTD ‘261 determinant’ (residues 257, 258, 259, and 261) and the α-helical segment 593–604 of E. coli σ4, specifically residues K593, R596, K597, H600, P601, R603 and S604 [54]. Thus, recruitment of αCTD by CAP appears to merely tether RNAP at the promoter site, while subsequent specific αCTD-σ4 interaction positions the holoenzyme at the promoter (likely at −35 element) leading to stable RP_c formation and efficient initiation.

A structural model of the ternary initiating complex containing CAP, RNAP and Plac DNA (Fig. 3c) was constructed [55] by combining the crystal structures of CAP-αCTD-DNA complex [10], σ4-(−35 element) complex [9], and RNAP-DNA complex [8]. The model provides new insight into the central role played by αCTD in the CAP-mediated transcription activation at Plac. αCTD, by providing three discrete interaction interfaces, serves as a thee-way bridge connecting CAP, DNA, and RNAP. The model also supports the view that simple recruitment through protein–protein adhesion, with minimal number of contacts, is sufficient for transcription activation at intrinsically weak promoters such as Plac.