E. coli RNA Polymerase Determinants of Open Complex Lifetime and Structure

Abstract

In transcription initiation by Escherichia coli RNA polymerase (RNAP), initial binding to promoter DNA triggers large conformational changes, bending downstream duplex DNA into the RNAP cleft and opening 13 base pairs to form a short-lived open intermediate (I₂). Subsequent conformational changes increase lifetimes of λP_R and T7A1 open complexes (OC) by >10⁵-fold and >10²-fold, respectively. OC lifetime is a target for regulation. To characterize late conformational changes, we determine effects on OC dissociation kinetics of deletions in RNAP mobile elements σ⁷⁰ region 1.1 (σ_1.1), β’ jaw and β’ sequence insertion 3 (SI3). In very stable OC formed by WT RNAP with λP_R (RP_O) and by △σ_1.1 RNAP with λP_R or T7A1 we conclude that downstream duplex DNA is bound to the jaw in an assembly with SI3, and bases −4 to +2 of the nontemplate strand discriminator region are stably bound in a positively-charged track in the cleft. We deduce that polyanionic σ_1.1 destabilizes OC by competing for binding sites in the cleft and on the jaw with the polyanionic discriminator strand and downstream duplex, respectively. Examples of σ_1.1-destabilized OC are the final T7A1 OC and the λP_R I₃ intermediate OC. Deleting σ_1.1 and either β’ jaw or SI3 equalizes OC lifetimes for λP_R and T7A1. DNA closing rates are similar for both promoters and all RNAP variants. We conclude that late conformational changes that stabilize OC, like early ones that bend the duplex into the cleft, are primary targets of regulation, while the intrinsic DNA opening-closing step is not.

INTRODUCTION

Multisubunit RNA polymerases (RNAPs) are molecular machines with many moving parts which carry out the physical and chemical steps of transcription [1, 2]. In E. coli, the focus of most studies of bacterial transcription, RNAP holoenzyme, composed of core enzyme (α₂ ββ'ω) and an initiation factor (here σ⁷⁰), uses binding free energy to open promoter DNA and orient the +1 template (T) strand base in the active site for initiation. A series of conformational changes, triggered by initial specific binding, form the initial open complex (OC). Promoter DNA is bent at or near the upstream ends of both the −35 and −10 regions, forming an ensemble of closed intermediates (designated I₁ ; see Figure 1) involving wrapping of far-upstream DNA on RNAP and insertion of the downstream duplex (−10 to +20) into the active site cleft. Then, in the rate-determining step, RNAP opens the entire 13-base pair transcription bubble, placing the +1 base in the active site and forming an initial, unstable OC (designated I₂) [3].

Figure 1.

Minimal mechanism of formation and dissociation of stable OC (RP_O) at λP_R, showing steps that contribute to the observed kinetics (k_obs, k_d). Where the kinetics are single exponential [48], the initial steps (with composite equilibrium constants K₁ or K₃) rapidly equilibrate on the time scale of the subsequent rate-determining (k₂ or k₋₂) step. For this situation, the dissociation rate constant k_d is determined by the rate constant k₋₂ for DNA closing (I₂ → I₁) and the equilibrium constant K₃ for stabilization of the open complex I₂ to form longer-lived I₃ and/or RP_O complexes [3, 5]. Likewise, in RNAP excess, the rate constant for OC formation k_obs is determined by the equilibrium constant K₁ for formation of the ensemble of closed (I₁) intermediates and the forward rate constant k₂ of the isomerization step that includes DNA opening and converts this I₁ ensemble to OCs [42, 43, 49, 50][49][50][50]. For these situations, k_d = k₋₂/(1 + K₃) (Equation 2) and k_obs = k₂ K₁ [R]/(1 + K₁ [R]).

After opening, at many promoters including λP_R, additional conformational changes occur to stabilize the OC and increase its lifetime. These late steps of the mechanism (see Figure 1) are best investigated by dissociation kinetic experiments [3-11]. Kinetic and fast-footprinting studies with λP_R, reveal that the discriminator region of the nontemplate (NT) strand is repositioned in the cleft [3] and downstream mobile elements (DME) of RNAP (including the jaw and sequence insertion 3 (SI3) domains of β’) assemble on the DNA downstream of the transcription bubble (+3 to ~+20) [7]. The mobile in-cleft element σ region 1.1 (σ_1.1) is also repositioned in these late stabilization steps [12], affecting the structure and stability of the open complex [13, 14]. These conformational changes form an exceedingly stable OC (λP_R RP_O), 10⁵-fold more stable and longer-lived than I.

σ_1.1 and other mobile elements are targeted by bacteriophages in order to take over the cell for viral replication. For example, β SI1, located at the top of the active site cleft (also known as the β downstream lobe or β insert 4, Figure 2), is a target for phage T4-encoded inhibitor Alc [15]. The T7 phage-encoded Gp2 inhibits initiation by binding to σ_1.1 and β’ jaw [16, 17]. Region σ_1.1 is highly negatively charged, consisting of a N-terminal three-helix bundle (residues 1-56; net charge −8) and a more flexible linker region (residues 57-98; net charge −12) [16, 18] (see Figure 2 and Supplemental Figure 1). Both of these regions are highly mobile. In the process of formation of a highly stable OC, σ_1.1 moves out of the active site cleft [12] and this mobility appears key to its function [19]. Gp2-σ_1.1 interactions prevent RNAP from binding to the downstream duplex in the stable OC [16, 17]. With Gp2 bound to RNAP, only an unstable (short-lived) open complex is formed [9, 10, 20, 21] with a shorter downstream DNase I footprint boundary and a smaller open region [10].

Figure 2.

E. coli RNAP holoenzyme (α₂ ββ’ωσ⁷⁰). A. View from above the active site cleft, showing RNAP as a van der Waals surface. Subunits: α₂ : dark blue; β: yellow; β’: green; ω: light orange; σ residues 7-55: purple (note: 7 is the observed N-terminus); σ residues 56-98: red; σ residues 99-106: cyan; σ residues 107-613: light pink; β SI1 (residues 225-343): magenta; β’ SI3 (residues 943-1130): bright orange; β’ jaw deletion from [27] (residues 1149-1190): gray; active site Mg²⁺ position is indicated by an arrow. Structure adapted from PDB 4LK1 [16]. B. Same view as panel A, highlighting peptide backbone residues that contact NT strand (blue spheres) and T strand (black spheres) in a dinucleotide (GpA)-stabilized complex of Thermus thermophilus RNAP with a heteroduplex promoter fragment in which −12 to +2 (NT) and −4 to +2 (T) are ss DNA, with a 13 bp downstream duplex [34]. C. Orientation of σ⁷⁰ region 1 (residues 7-106) in E. coli RNAP holoenzyme. Subunits are colored as in panel A, expanded 120%. Gaps in electron density observed within the flexible linker are labelled and represented as red dots. D. The sequences of λP_R, T7A1, and rrnB P1 from the −35 element to base +2 are shown for comparison. The −10 and −35 elements are underlined; the transcription bubble bases (λP_R : −11 to +2; T7A1: −12 to +2) are shown in blue; base +1A is shown in red.

Sequence insertions at the position of β SI1 (and that of β’ SI3, also known as β’i6) are common but by no means universal in bacteria [22]. Deletion of SI1 in E. coli is nonlethal, though it does lead to temperature-sensitive growth [15, 23]. By contrast, a deletion mutant of SI3 is not viable [23, 24].

The details of how mobile regions like σ_1.1, β’ jaw and SI3 contribute to the basal mechanism of transcription initiation currently are not well understood. Here, we investigate the effects of single and double RNAP deletion variants that remove in-cleft (σ_1.1) and/or downstream mobile elements (DME; β’ jaw, SI3) on the lifetimes of the initial (I₂) and final (RP_O) OCs at two well-characterized promoters, λP_R and T7A1. Variants investigated include partial (residues 1-55) or total (1-98) deletion of σ_1.1 (designated △55 and △98, respectively) and deletions of either β’ jaw (residues 1149-1190 (△JAW)) or β’ SI3 (residues 943-1130 (△SI3)) (Figure 2). Double variants with part or all of σ_1.1 as well as either β’ jaw or β’ SI3 deleted were studied to determine the extent of coupling between these regions.

From these results, we determine the contributions of these regions to the late steps of the transcription initiation mechanism, and deduce the network of conformational changes involving these mobile elements and the discriminator DNA sequence that forms the long-lived OC at λP_R . We obtain compelling evidence that σ_1.1, functioning as a nucleic acid mimic, competes with both the discriminator region of the NT strand and the downstream duplex to destabilize the OC (see Discussion). These deletion variant studies also explain why the T7A1 OC is relatively short-lived, and allow us to relate the final OC at T7A1 to the intermediate (I₃) OC at λP_R, both in terms of lifetime and structure.

RESULTS

Effects of deleting the three-helix bundle and linker regions of σ_1.1 on the lifetime of the stable open complex at the λP_R and T7A1 promoters

How do σ_1.1 and its mobility affect OC stability? Deletion of RNAP σ_1.1 increases OC lifetime (1/k_d, where k_d is the dissociation rate constant) at λP_R [25] and the ribosomal promoter rrnB P1 [13]. To extend this data set, and as a basis for comparison with lifetimes of OCs formed by double deletion variants of RNAP, we quantified effects of two σ_1.1 deletions (△55 and △98) on lifetimes of λP_R and T7A1 OCs in TB at 37 °C. At this condition, we find that the lifetime of the WT RNAP-λP_R OC is 17 hours (6.3 × 10⁴s), 170-fold longer than that of the corresponding WT RNAP-T7A1 OC (370 s). Dissociation kinetic data were obtained by the nitrocellulose filter binding assay using as competitor an excess of λP_R +UP, a faster, stronger binding variant of λP_R (see Materials and Methods). Representative kinetic assay results are shown in Figure 3; values of k_d are listed in Table 1 and compared to other variants in Figures 5 and 6.

Figure 3.

Dissociation kinetic data for stable OCs formed by WT (white circles, black fit), △55 (purple circles and fit) and △98 (red circles and fit) RNAP, obtained by nitrocellulose filter binding. Nonlinear fits to Equation 1 are obtained from Igor Pro version 5.03 as described in Methods. A. Dissociation of λP_R OCs at 37 °C. B. Data from A. replotted on a logarithmic time scale. C., D. Dissociation of T7A1 OCs at 37 °C, plotted and analyzed as in A. and B.

Table 1.

Dissection of dissociation rate constants k_d into DNA closing (k_-2) and open complex stabilization (K₃) for various RNAP variant-λP_R and -T7A1 promoter complexes

	kd (s−1)	k−2 (s−1)	K3
RNAP variant	λPR
WT	(1.6 ± 0.3) × 10−5 a	3.3 ± 0.2 b	(2.1 ± 0.4) × 105
Δ55	(4.8 ± 0.6) × 10−6
Δ98	(2.9±0.4) × 10−6	1.8 ± 0.6	(6.1 ± 2.2) × 105
ΔJAW	(5.0±0.7) × 10−4 d	0.58 ± 0.12	(1.2 ± 0.3) × 103
ΔSI3	(1.6 ± 0.3) × 10−4	3.5 ± 0.4 d	(2.2 ± 0.5) × 104
	T7A1
WT	(2.7 ± 0.4) × 10−3	0.63 ± 0.21	(2.4 ± 0.8) × 102
Δ55	(1.1 ±0.1) × 10−3
Δ98	(1.5 ± 0.2) × 10−5 d	0.81 ± 0.11	(5.6 ± 0.7) × 104
ΔJAW	(3.1 ± 0.9) × 10−3	0.47 ± 0.05 d	(1.5 ± 0.5) × 102
ΔSI3	(7.9 ± 0.8) × 10−3 d	0.89 ± 0.10	(1.1 ± 0.1) × 102

^aPrevious determinations of k_d for WT RNAP-λP_R OC: k_d = (1.9 ± 0.6) × 10⁻⁵ s⁻¹ (DNA fragment extending from −115 to +76 was cut out of a plasmid and labelled) [5]; k_d = (2.2 ± 0.3) × 10⁻⁵ s⁻¹ (890 bp fragment containing both λP_R and λP_RM was cut out of a plasmid and labelled) [47].

^bData from [5].

c Data from [7].

^dUncertainty estimated as 10%.

Figure 5.

Lifetimes and stabilizations of λP_R and T7A1 OCs formed with RNAP variants. Comparison of lifetimes of stable OCs (1/k_d; panel A) and initial OC I₂ (1/k₋₂; panel B), and OC stabilization equilibrium constants (K₃; panel C) for λP_R (unshaded, left bar of each pair) and T7A1 (shaded, right bar of each pair) promoters and for WT RNAP, σ_1.1 deletion △98, and β’ deletions △JAW and △SI3. All bar graphs use same log scale vertical axis, starting at the same point, so heights can be visually compared.

Figure 6.

Comparison of lifetimes of stable OC (1/k_d) and initial unstable OC (1/k₋₂) formed by λP_R (blue) and T7A1 (red) promoters. The lifetime axis is on a logarithmic scale. Values for WT RNAP (first column) are compared with single deletions in σ_1.1 (second column) and β’ DMEs (third column) and double deletions △55/△DME (fourth column) and △98/△DME (fifth column).

Figure 3 shows that the effects of deleting both the three helix bundle and the linker region of σ_1.1 (△98) are very large, especially for T7A1, requiring a log time scale to compare dissociation kinetic data for △98 and WT RNAP visually. For T7A1, Figure 3 reveals that the lifetime of the △98 OC exceeds that of the WT by more than 100-fold, while for λP_R the increase in OC lifetime with △98 is about 6-fold. By contrast, increases in lifetime upon deletion of only the three-helix bundle of σ_1.1 (△55) are more modest (3-fold for λP_R, 2.5-fold for T7A1), and slightly larger for λP_R than for T7A1. The simplest interpretation of these data is that both regions of σ_1.1 act to destabilize the stable OC, with the linker region (56-98) exerting a much larger effect than the three-helix bundle, especially for T7A1. These effects presumably arise from competition between these polyanionic regions of σ_1.1 and polyanionic promoter DNA.

Effects of σ_1.1 △98 deletion on lifetimes of T7A1 and λP_R open complexes result primarily from differences in stabilization of the initial open complex (K₃), not in the DNA closing rate (k₋₂)

What is the origin of the large differences in lifetime observed for T7A1 and λP_R OCs with WT RNAP and σ_1.1 variants? Kinetic-mechanistic studies of dissociation reveal that the lifetime of the stable OC is determined by the equilibrium constant K₃ for converting the initial open intermediate I₂ to the stable OC (RP_O at λP_R) and by the lifetime of the initial open complex, I₂ (1/k₋₂; see Equation 2) [3, 5, 7]. Does the much greater lifetime of the stable OC at λP_R than at T7A1 result from much greater OC stabilization (K₃) and/or a much greater lifetime of I₂? Does the very large increase in lifetime of the stable T7A1 OC observed upon deletion of σ_1.1 (△98) result from a dramatic increase in K₃, or a reduction in k₋₂, or both?

To determine DNA closing rate constants k₋₂ and thereby obtain OC stabilization equilibrium constants K₃ from k_d, we take advantage of the finding that K₃ decreases strongly with increasing KCl concentration, while k₋₂ is salt-independent [5]. For the λP_R promoter, k_d increases with increasing salt concentration to attain a [salt]-independent plateau above ~1 M KCl where K₃ is negligible (K₃ << 1) and k_d = k₋₂ = 3.3 ± 0.2 s⁻¹ at 37 °C [5]. (Published results are replotted in Figure 4B.) Hence when RP_O complexes are rapidly shifted to high [KCl], RP_O immediately converts to I₂, forming a high transient concentration (“burst”) of this key open intermediate which dissociates in a single exponential decay with rate constant k₋₂ .

Figure 4.

Kinetics of dissociation of T7A1-WT RNAP OCs. A. Nitrocellulose filter binding dissociation kinetic data for T7A1-WT RNAP complexes in TB (0.12 M KCl; white triangles with black fit) and after rapid mixing upshifts to 0.48 M (red triangles and line), 0.6 M (green triangles and line), 0.8 M (blue triangles and line), and 1.1 M KCl (purple triangles and line). B. Log-log dependences of dissociation rate constants k_d for T7A1-WT RNAP (black triangles) OCs on KCl concentration at 37 °C. Data for λP_R (white hourglasses), shown for comparison, are from [5]. At high salt concentration where K₃ << 1, k_d = k₋₂ which is found to be independent of salt concentration; these high-salt plateaus are shown by horizontal lines (T7A1: solid, average of four KCl points; λP_R : dotted, from [5]).

Determinations of k_d for dissociation of T7A1-WT RNAP OC after upshifts to moderate to high [KCl] (≥ 0.48 M) are compared with the corresponding k_d determinations for λP_R in Figure 4A. In TB at 0.12 M KCl, T7A1 is more than 100-fold faster dissociating than λP_R. For upshifts to 0.48 M KCl or higher, k_d for T7A1 exhibits a KCl-independent plateau value k_d = 0.6 ± 0.2 s⁻¹ (Figure 4B). As observed for λP_R, this DNA closing rate is independent of salt concentration. For T7A1, I₂ lifetime is approximately 5-fold longer than for λP_R, while the stabilization equilibrium constant (K₃ = 240 ± 80), though very significant, is almost three orders of magnitude less than for λP_R (K₃ = (2.1 ± 0.4) × 10⁵) (see Table 1 and Figure 5). Hence the shorter lifetime of the T7A1-WT RNAP stable OC is entirely the result of differences in the OC stabilization step. The much smaller K₃ for T7A1 explains the onset of the high salt plateau in k_d at a much lower salt concentration (~0.5 vs. ~1 M KCl).

Lifetimes of initial OC (I₂) formed by △98 RNAP at λP_R and T7A1 promoters were also determined by this burst method, using upshifts to 1.1 M KCl. (Results are tabulated in Table 1 and compared to other variants in Figure 5 and Supplemental Figure 2). Effects of the △98 deletion on I₂ lifetimes at both promoters are small: a two-fold increase for λP_R and no significant effect for T7A1. Hence the effects of the △98 deletion are primarily on the OC stabilization equilibrium constant K₃, especially for T7A1 where the △98 deletion increases K₃ by ~250-fold (to (5.6 ± 0.7) × 10⁴), almost to the value observed for the WT RNAP-λP_R OC. By contrast, the △98 deletion results in a more modest ~3-fold increase in K₃ at λP_R.

Promoter-specific effects of β’ jaw DME deletion on initial and stable open complex lifetimes

Footprinting and dissociation kinetic studies revealed that the β’ jaw DME stabilizes λP_R RP_O by interacting with the downstream DNA duplex from +10 to +20 [7]. Recent crystal structures show that this jaw-downstream duplex interaction (centered at +15) is also present in initiation complexes [26]. Partial deletion of the jaw (β’△1149-1190; Figure 2) reduces the lifetime of the λP_R stable OC to ~3% of the WT value [7, 9, 27]. Here we find that the origin of this effect is the massive destabilization of λP_R RP_O relative to I₂ by deletion of the jaw, which reduces K₃ to ~0.6% of its WT value (see Table 1, Figure 5 and 6 and Supplemental Figures 3 and 4). On the other hand, deletion of the jaw increases the lifetime of λP_R I ₂ (1/k₋₂) ~6-fold, making it comparable to that of the WT RNAP-T7A1 OC (Table 1, Figure 5, and Supplemental Figure 2).

In striking contrast to the stabilization at λP_R, we find that deleting the jaw has no significant effect on the lifetime of either the stable or the initial OC at T7A1; values of k_d, k₋₂, and K₃ (Table 1, Figure 5) for OC formed with △JAW RNAP are the same within uncertainty as for WT RNAP. The lack of effect of the jaw deletion on K₃ indicates that the jaw does not participate in the conformational changes and interactions with the downstream duplex that stabilize the T7A1 OC, and can explain much of the 100-fold difference in lifetimes of the stable OC with WT RNAP at these promoters.

Similar effects of SI3 DME deletion on lifetimes of λP_R and T7A1 stable open complexes

The DME designated β’ SI3 (Figure 2) consists of two sandwich-barrel hybrid motifs extending from the middle of the trigger loop. Located high in the active site cleft at the downstream end, SI3 forms a large interface with the β’ jaw in the holoenzyme structure [28]. SI3 is required for proper regulation by transcription factors DksA [29] and GreB [30]. The SI3 deletion is lethal and has strong effects on initiation, elongation, pausing, and termination [23, 24]. The lethal effect of SI3 deletion is thought to result from its connection to the trigger loop. Modelling indicates that SI3 does not interact with the downstream duplex DNA in the stable OC and suggests that SI3 moves when the trigger loop folds during NTP addition cycle [28, 31]. Here we study the effects on OC lifetime and stabilization of a β’ SI3 deletion mutant (△SI3, β’ △943-1130).

Dissociation kinetic studies with stable λP_R-△SI3 and T7A1-△SI3 OC (see Table 1, Figures 5 and 6, and Supplemental Figures 3 and 4) reveal a three-fold reduction in lifetime (1/k_d) for T7A1 and a ten-fold decrease for λP_R, relative to WT RNAP, similar to previously published results for T7A1 [23] and λP_R [5]. Rapid salt upshift dissociation experiments were used to determined k₋₂ and the lifetime of the initial OC. As summarized in Figure 5 and Table 1 (kinetic data in Supplemental Figure 2) deletion of SI3 has at most a minor effect, possibly increasing k₋₂ slightly for both λP_R and T7A1.

The main effect of △SI3 is therefore to reduce K₃ for the I₂ → RP_O conversion, to 10% of the WT value for λP_R and 50% for T7A1. These are significant but smaller effects relative to those of the deleting the jaw on λP_R OC lifetime, or of deleting σ_1.1 on T7A1 OC lifetime. This finding is consistent with the previous observation that for λP_R, deleting SI3 did not significantly reduce the urea dependence of the dissociation rate constant at λP_R [6], indicating no large effect of △SI3 on the folding/assembly interactions of the jaw and other regions involved in converting I₂ to RP_O at λP_R. Like △JAW, the △SI3 mutation has a somewhat larger effect on k_d and K₃ for λP_R than for T7A1, probably reflecting a greater degree of assembly of these DME in the stable λP_R open complex than for T7A1. However, deletion of SI3 has small but detectable effects on both T7A1 k_d and K₃, while deletion of the jaw does not.

Striking Promoter-Specific Differences in Effects of Double Deletions of σ_1.1 and a DME

Effects of double deletions (one in σ_1.1, one in a β’DME) on lifetimes of stable λP_R and T7A1 OC are reported in Table 2 and compared with single deletions in Figure 6. (Kinetic data are given in Supplemental Figures 3 and 4.) These data and Figure 6 reveal large context-depending differences (λP_R vs T7A1, WT vs σ_1.1 variant core RNAP). Fundamentally these differences appear to originate in a close coupling between effects of σ_1.1 (especially linker region 56-98) and the β’ DME on OC lifetime, which we propose are present for both promoters but differ in detail for λP_R and T7A1.

Table 2.

Dissociation rate constants k_d for λP_R and T7A1 stable open complexes with Δσ_1.1/ΔDME deletion variants

RNAP variant	λPR kd (s−1)	T7A1 kd (s−1)
Δ55ΔJAW	(2.6 ± 0.3) × 10−4 a	(8.4 ± 4.2) × 10−4
Δ98ΔJAW	(2.1 ± 0.9) × 10−4	(2.2 ± 0.2) × 10−4 a
Δ55ΔSI3	(2.8 ± 0.3) × 10−5 a	(1.4 ± 0.1) × 10−3
Δ98ΔSI3	(1.9 ± 0.5) × 10−4	(1.3 ± 0.5) × 10−4

^aUncertainty estimated as 10%.

Comparison of the lifetimes of △98/△JAW RNAP OC at T7A1 and λP_R with those of single deletion variants and WT RNAP (Figure 6) provides clear example of promoter-dependent linkages between σ_1.1 and the jaw. Comparisons with the analogous set of lifetime data for △55 reveal large differences that implicate the chain region as the principal origin of the linked σ_1.1-jaw effects on lifetime. Additional comparisons with sets of lifetime data for both single and double variants involving either △98 or △55 and the △SI3 instead of △JAW reveal quite parallel linkages between the SI3 and the σ_1.1 chain region. Here, we summarize the key results.

Although deleting the jaw does not significantly affect the lifetime of the T7A1 stable OC compared to WT RNAP, deleting the jaw greatly (15-fold) reduces the lifetime of the △98 RNAP OC. This suggests that the stabilization of the T7A1 OC by the △98 deletion, relative to WT RNAP, is in large part the result of an interaction of the jaw that is absent in the WT OC.

At λP_R, the effect of deleting the jaw is also greater in the context of the △98 deletion (70-fold) than in the context of WT RNAP (20-fold). For T7A1, deleting the jaw in the context of △55 has no significant effect, as for WT, while for λP_R the effect of deleting the jaw in the context of △55 (50-fold) is intermediate between the △98 (70-fold) and WT (20-fold) contexts. For both promoters, context dependences of effects of SI3 deletion are broadly similar to those summarized above for deletions of the jaw, with some differences in the details. Deletion of SI3 causes as much as a 70-fold reduction in OC lifetime (in the context of △98 and λP_R), but has no effect in the context of T7A1 and the △55 deletion variant.

Deletion of σ_1.1 (△98) together with either DME largely eliminates the original 170-fold difference in lifetimes of WT RNAP-λP_R and -T7A1 OC. Lifetimes of △98/△DME OC at both promoters are similar (4 × 10³s to 8 × 10³s), differing less than two-fold between λP_R and T7A1 for either DME deletion. These lifetimes are in between those of OC with WT RNAP for the two promoters studied. △55/△JAW RNAP is almost as effective at equalizing OC lifetimes at these two promoters (3-fold difference), but △55/△SI3 RNAP is ineffective in this regard because of smaller, compensating effects of the two deletions that preserve the WT RNAP lifetimes.

In addition, deletions in σ_1.1 trump DME deletions for T7A1, while the reverse is true for λP_R. For T7A1, lifetimes of double variant OCs with a σ_1.1 deletion and a DME deletion exceed that of WT RNAP, indicating that the increases in T7A1 lifetime expected as a result of these σ_1.1 deletions (especially △98) outweigh the modest (△SI3) or minimal (△JAW) reductions in lifetime expected from the DME deletions. For λP_R, the situation is reversed. Lifetimes of the double variant OCs are shorter than for WT RNAP-λP_R, indicating that the reduction in λP_R lifetime expected from the DME deletion outweighs the increase in lifetime expected from the σ_1.1 deletion.

The only deletion investigated that approaches context-independence is △55. Deletion of this N-terminal three-helix bundle increases lifetime of stable OCs formed by WT, △JAW or △SI3 RNAP by 2-6 fold at both λP_R and T7A1.

KMnO₄ reactivity of T7A1 and λP_R promoter DNA in stable △98 RNAP open complexes

To determine whether the increased lifetimes of OC formed by △98 RNAP at T7A1 and λP_R promoters are accompanied by changes in the open region or distortions in other regions of the promoter DNA, KMnO₄ footprints of both strands of WT and △98 stable OC at these promoters were obtained in TB at 37 °C (λP_R: Figure 7A, B; T7A1: Figure 7C, D).

Figure 7.

Permanganate footprinting of △98 and WT RNAP OCs at T7A1 and λP_R. Positions of KMnO₄-reactive thymines are labelled, as determined from A+G sequencing lanes (lanes A8, B7, C1, and D1). λP_R NT strand footprints (panel A) of △98 RNAP open complexes are shown in the absence (lanes 4,6) and presence (lanes 5,7) of 20 μg/ml heparin; that of WT RNAP (+ 20 μg/ml heparin) is shown in lane 3. Free DNA controls are shown in the absence (lane 1) andpresence (lane 2) of 20 μg/ml heparin.λP_R T strand footprints (panel B) of △98 RNAP OC are shown in the absence (lane 5) and presence (lane 6) of 20 μg/ml heparin; that of WT RNAP (+ 20 μg/ml heparin) is shown in lane 4. Free DNA controls are shown in the absence (lanes 1,3) and presence (lane 2) of 20 μg/ml heparin. T7A1 NT strand (panel C) and T strand (panel D) footprints of △98 RNAP OC (lanes C5, D5) and WT RNAP (lanes C4, D4) are compared to free DNA controls, exposed to KMnO₄ (lanes C3, D3) or not exposed to KMnO₄ (lanes C2, D2). Conditions: TB, 37 °C; 0.2 mM KMnO₄, 10 s [1].

These footprints show no differences in the size of the initiation bubble in △98 vs. WT OC. At λP_R, T strand thymines at −11, −9, −8 and +1 are observed in both △98 and WT OCs, as are thymines at −4, −3 and +2 on the NT strand, defining a 13 base open region. MnO₄⁻ reactivities of λP_R NT strand thymines are higher (for the same dose of MnO₄⁻) for △98 than for WT RNAP on the NT strand, but appear similar for the T strand. Thymines at −10 and −7 on the NT strand are not MnO₄⁻ reactive in △98 OC or WT RNAP OC. Similar NT strand footprints were previously reported for WT and △100 RNAP OCs at λP_R [14]. The lack of MnO₄⁻ reactivity of NT strand thymines at −7 and −10 is most simply explained if these bases are buried in pockets of σ region 2. From the crystal structures, evidence exists for a −7 pocket but not a −10 pocket. The −10 base is stacked on −9, which should reduce its MnO₄⁻ reactivity, but the upstream face of −10 appears accessible to MnO₄⁻ [32-34].

On the λP_R T strand, for △98 but not WT OC, MnO₄⁻ hypersensitivity is observable immediately upstream of major promoter interaction sites including position −38 (at the upstream end of the −35 element), position −63 (at the upstream end of the UP element [35]), and position −81 (near the upstream end of the proposed upstream wrap [36] and near the upstream limit of the range of contact by the αCTD, according to crosslinking [12]). These MnO₄⁻ hypersensitivities indicate sites of DNA distortion/unstacking, presumably where the upstream DNA in the △98 RNAP open complex is strongly bent in wrapping it on RNAP. The precise origin of these hypersensitivities is unclear, but previous results have indicated an effect of deletion of σ_1.1 on interactions with regions upstream of the active site cleft. DNase footprints with a variant RNAP lacking σ_1.1 at the promoter Pmin show upstream enhancements around −40 and −50 [37]. At some promoters, the effects of σ_1.1 deletion on the kinetics of OC formation and the structure of OC also depend on the length and sequence of the promoter spacer [37, 38].

At T7A1, T strand thymines at −12, −10, −6, −4 and +1 are observed in both △98 and WT OC, as are thymines at −7 and +2 on the NT strand, defining a 14 base open region. T strand MnO₄⁻ reactivity decreases from −12 to +1, with the intensity at +1T barely above background. Reactivity at each position is considerably stronger for △98 than for WT RNAP. On the NT strand, MnO₄⁻ reactivity at +2 is somewhat higher for △98 than for WT RNAP, but the reactivity difference on this strand between △98 and WT RNAP OC is not as great as for the T strand of T7A1 or for the NT strand of λP_R. Thymines at −11 and −8, the analogs of −10 and −7 for λP_R, are not MnO₄⁻ reactive for either △98 OC or WT RNAP OC.

In both WT-and △98-T7A1 OC, permanganate reactivity is observed in the UP element region in tracts of thymines at approximately −50 and −58 on the NT strand and −45, −55, −60, and −65 on the T strand. Presumably this reactivity is the result of DNA bending/distortion by tight binding of the αCTDs to the strong, near-consensus T7A1 UP element.

The promoter-specific increases in MnO₄⁻ reactivity of thymines in the open region of the NT strand (λP_R) and the T strand (T7A1) in △98 OC are a clue to the structural differences in these complexes. Gries et al. [3] observed relatively large increases in MnO₄ ⁻ reactivity in the conversion of the initial to the stable OC at λP_R, as well as small increases in reactivity for −11, −9 and −8 (but not +1) thymines on the T strand. One interpretation of this result is that either the rearrangement of the discriminator region in this process and/or the tighter binding of the strand backbone in the stable OC increases MnO₄⁻ reactivity. Strong binding of the DNA phosphate-sugar backbone is observed to unstack the bases in the SSB-ss DNA complex [39] and a similar effect could explain the greater MnO₄⁻ reactivity of thymines in the stable vs. the initial λP_R OC, and of one or the other strand in the λP_R-△98 and T7A1 - △98 OC. Alternatively the elimination of the negatively charged σ_1.1 may increase local concentrations of MnO⁻ in the cleft of the △98 OC, leading to these reactivity increases. Neither of these possibilities explains why the increased reactivity is on the λP_R NT strand and the T7A1 T strand.

DISCUSSION

Hypothesis to explain the large differences in lifetime of open complexes at different promoters and for different RNAP variants

We propose that σ_1.1 and the jaw and SI3 DME participate in a sophisticated network of interactions and conformational changes, directed by the discriminator element, to determine the extent of OC stabilization. For λP_R, with its 6 base discriminator and G/G bases at positions −6/−5 at the upstream end, this stabilization exceeds 10⁵ [5]. For T7A1, with a 7 base discriminator and T/A bases at positions −7/−6, the stabilization is a very significant but more modest factor of 240, a difference which we find is due in part to the lack of involvement of the jaw.

Previously we deduced that the jaw engaged on the far downstream DNA (+10 to +20) late in the process of RP_O formation at λP_R, and that the complexes formed by △JAW RNAP at full-length λP_R and by WT RNAP on λP_R truncated downstream at +12 were analogues of the I₃ intermediate OC (see Figure 8) at this promoter [7]. Based on our current findings, we propose that the stable T7A1 OC is also analogous in lifetime and structure to the late I₃ intermediate at λP_R.

Figure 8.

Interactions between RNAP elements and promoter DNA that define properties of the various OCs, compared with free RNAP (see Discussion). Note that only the DNA strands from ~ −10 to +20 are shown.

rrnB P1, with a 8 base discriminator and G/C bases at positions −8 and −7 (see Figure 2), exhibits little if any stabilization of the initial OC. On linear DNA in the absence of NTP the rrnB P1 OC is short-lived and thermodynamically unstable relative to closed complexes. We therefore propose that the rrnB P1 OC is an analogue in lifetime and structure of the initial OC (I₂) at λP_R and T7A1 promoters. In support of this proposal, I₂ lifetimes are found to be similar for λP_R and T7A1 for all RNAP variants studied here. Mutation of −7 C to G increases the lifetime of the rrnB P1 OC 40-fold [40], which may result entirely from the interaction with −7G, or may also involve some in-cleft and downstream elements responsible for stabilization of OC at T7A1 and λP_R promoters.

Roles of the discriminator and RNAP in the conformational changes and interactions thatstabilize the open complexes at λP_R and T7A1

To investigate the network of conformational changes and interactions that stabilize λP_R and T7A1 OC, we have determined the effects of deletion of σ_1.1, β’ jaw and β’ SI3, on OC lifetime (1/k_d) and in many cases OC stabilization (K₃ = k₋₂/k_d). For purposes of this discussion, we assume that the effect of deleting a region is a direct indication of the effect of that region, though we recognize that longer range effects from changes in location and folding of nearby regions may accompany the deletion.

Interactions in highly stable open complexes

The maximum stabilization observed is K₃ = 6.1 × 10⁵ for the △98-λP_R OC (−8 kcal in △G°). Both WT-λP_R OC (K₃ = 2.1 × 10⁵) and △98-T7A1 OC (K₃ = 5.6 × 10⁴) are also highly stabilized. We propose that all these long-lived OC are stabilized by similar interactions (Figure 8).

a)On the NT strand, bases −5 and −6 interact with σ_1.2 while bases −4 to +2 (the CRE) are bound in a positively charged “CRE track” [34] on β in the cleft. The smaller stabilization of △98-T7A1 OC than of WT λP_R and △98-λP_R OCs is probably because interactions of its discriminator with σ_1.2 and/or the CRE track are less favorable than for λP_R.

b)The jaw is bound to the downstream duplex (in the region +10 to +20) as part of an assembly with SI3 and other DME, probably including the β’ clamp [2, 7]. This interaction is proposed to involve residues T1169, R1174, and M1189 [17]. Nonspecific coulombic interactions with the phosphate backbone are also involved, as evidenced by the greatly reduced power dependence (from 9.5 to 4.7 power) of the salt effect on k_d for λP_R-△JAW OC [7].

c)σ_1.1 is absent or is ineffective at destabilizing this OC, probably because it is interacting with β SI1 [12] (see Figure 8).

d)The cleft is closed.

These interactions are tightly coupled, as indicated by the non-additive effects of the RNAP deletion variants. For example, deletion of σ_1.1 and the jaw together result in a net stabilization of the T7A1 OC, but a net destabilization of λP_R, eliminating the large difference in OC stabilization for these two promoters. Effects of β’ SI3 also appear coupled to those of σ_1.1 in OC stabilization.

The ineffectiveness of σ_1.1 in destabilizing WT RNAP-λP_R RP_O indicates that it is bound elsewhere. The N-terminal bundle of σ_1.1 moves to interact with β SI1, located at the top of the cleft (Figure 2), during stable OC formation at the consensus lac promoter, which has a 6 base discriminator with upstream sequence −5T and −6G, similar to λP_R [12]. The σ_1.1 linker region is presumably high in the cleft, interacting with β and unable to compete with NT DNA for binding to the CRE track.

The above proposals are consistent with the structure of a stable OC between T. thermophilus RNAP (lacking σ_1.1 and the jaw DME) and a short (ending at +13) DNA heteroduplex with a λP_R - like discriminator (upstream GG, 6 bases long) [34]. The −5G and −6G bases interact with σ_1.2 whereas bases −4 to +2 are bound to β, making interactions with β residues R371, D199, R394, and W183 [34] (E. coli residue numbering is used throughout). This β surface is net positively charged, indicating the likelihood of coulombic interactions with the discriminator DNA backbone or the negatively charged σ_1.1 linker with this surface of β. In E. coli holoenzyme, helices 1 and 2 of the σ_1.1 three-helix bundle are bound to a surface of β immediately above this CRE track, including H165, S166, R197, R202, and K203 [16]. This three-helix bundle almost certainly moves earlier in OC formation, perhaps to allow entry of the downstream duplex to the cleft prior to opening. The positive patch on β exposed upon this movement may be a binding site for the σ_1.1 linker or the discriminator strand in one or another OC.

Interactions in moderately stable open complexes

The final WT-T7A1 OC (K₃ = 240) and λP_R I₃ are only moderately stabilized relative to I₂, indicating that key interactions that stabilize the WT-λP_R OC are missing or compromised. We propose that σ_1.1 (linker and bundle) exchange places and interactions with the discriminator NT strand and downstream duplex DNA in the conversion of the late λP_R I₃ to RP_O and that the stable T7A1 OC is structurally similar to I₃. In T7A1 OC (or λP_R I₃) we propose that:

a)The discriminator region of the NT strand is not bound to σ_1.2 or to the CRE track, and presumably is higher in the cleft, possibly interacting with the same cluster of positively charged groups that bind the three-helix bundle in holoenzyme (see above), directly above the CRE track.

b)The jaw does not interact with downstream duplex DNA, and may instead interact with the 3-helix bundle of σ_1.1.

c)A region of the σ_1.1 linker is bound in the CRE track; and

d)The cleft may be less closed.

These proposals are consistent with the observations that deleting the jaw or SI3 has only a small effect on the lifetime and stability of the WT RNAP-T7A1 final OC, whereas deleting the entire σ_1.1 dramatically stabilizes this OC, presumably by allowing the discriminator region of the NT strand to bind to σ_1.2 and the CRE track, and allowing the jaw to bind to the downstream duplex. This dramatic stabilization is substantially reversed by deleting the jaw as well. We propose that a negatively charged region of the σ_1.1 linker interacts with the CRE track. The highly negatively charged three-helix bundle may mimic ds DNA in interacting with the jaw or SI1.

What regions of σ_1.1 might be involved in these steps? On the three-helix bundle, the eight highly conserved [14] negatively-charged (4D, 4E) residues out of 17 amino acids between residues 26-42 (Supplemental Figure 1) plausibly are a duplex DNA mimic, competing with the downstream duplex for binding to the jaw DME. A simplistic interpretation of the large effect of the jaw deletion on the power law exponent Sk_d = dln(k_d)/dln[salt] [7] is that the jaw-duplex interaction involves +/− charge interactions with at least 6 DNA phosphates. We concluded that the jaw binds to one face of the DNA helix in the +10 to +20 region, making up to 10 phosphate interactions (similar to the 8 negative charges from residues 26-42).

There are three additional clusters of negative charge in σ_1.1. The largest cluster (3D, 3E) is in the structurally-undefined region (residues 57-69). A second (2D, 2E; residues 73-78) and a third cluster (1D, 2E; residues 88-96) are separated by the uncharged α-helical region of the linker. All three clusters are moderately to highly conserved [14]. Based on charge distribution alone, the first cluster is the most logical candidate for the proposed exchange with the 6 base CRE region. However, proximity in the crystal structure suggests that the third region, which is contiguous to σ_1.2, may also or instead be involved in competition with the CRE.

We propose that the key structural difference between the final OC at λP_R and T7A1 is similar to that between λP_R RP_O and λP_R I₃. These differences are in the positions and interactions of the discriminator DNA and linker region of σ_1.1 in the cleft, and of the downstream duplex DNA and the three helix bundle of σ_1.1 relative to the jaw and SI3 DME. In converting λP_R I₃ to λP_R RP_O, the polyanionic σ_1.1 N-terminal three-helix bundle changes places with duplex DNA and the polyanionic σ_1.1 linker region changes places with the discriminator strand. The competition between duplex DNA and duplex DNA mimic, and between ss DNA and ss DNA mimic, determine whether the structure and lifetime of the OC are like the highly stable (λP_R RP_O-like) or moderately stable (T7A1-like or λP_R-I₃ like) OC. Since (at the level of k_d) △55 deletions have much smaller effects than △98 deletions, the key element of σ_1.1 is its flexible, polyanionic chain region (residues 56-98).

Deletion of a β’ DME reduces K₃ for stabilization of the WT RNAP-T7A1 OC by approximately 50%, and hence β’ DME do not contribute significantly to the interactions involved in converting I₂ to the T7A1 final OC or to the analogous λP_R I₃. What elements stabilize λP_R I₃ relative to I₂, and stabilize T7A1 relative to rrnB P1? Various other candidates have been identified as potentially important for OC stabilization, including the β’ clamp (residues 131-347, 1260-1368). Deletion of β’ residues 215-220 reduces the OC lifetime ~40-fold at λP_R and ~15-fold at rrnB P1 [41]. A comparison of effects of these deletions on k₋₂ and K₃ for λP_R and T7A1 promoters should be very informative.

Possible interactions in unstable open complexes

In the conversion of λP_R I₃ to RP_O (or in comparison of T7A1 OC to λP_R RP_O) the discriminator region of the NT strand descends in the cleft and the σ_1.1 linker region ascends. Possibly this pattern extends to initial (unstable) OC like I₂ at λP_R or T7A1, or to the unstable rrnB P1 OC. In these unstable OCs, we propose that:

a)The downstream duplex interacts with SI1 and the bubble NT strand is located high in the cleft, occupying positions that interact with the three-helix bundle and the flexible chain of σ_1.1 in the most stable OCs

b)The jaw interacts with the σ_1.1 three-helix bundle.

c)The region of the σ_1.1 linker adjacent to σ_1.2 is bound in the CRE track. A second negatively charged patch of the σ_1.1 chain may interact with the positively charged region above this track, preventing the discriminator from binding there, as it may do in T7A1 or λP_R I₃ OC.

d)The cleft may be open, and binding interactions of the strand backbones in the cleft may be weak because the phosphates are not dehydrated.

Gries et al [3] obtained information about the condition of the open DNA in I₂ at λP_R and the changes in converting I₂ to RP_O from MnO₄ footprinting. The entire initiation bubble (−11 to +2) opens in the step forming I₂ from the advanced I₁ intermediate. In I_2, the −10 region of the NT strand is already tightly bound to σ₂; thymines at −10 and −7 are not MnO₄⁻ –reactive, as is the case for RP_O. Reactivities of T strand thymines in I₂ and RP_O are similar in the upstream region (25% less reactive in I₂) and identical for the start site (+1) thymine, consistent with the proposal that in I₂ this strand has descended to put the +1 base in or near the active site. But thymines in the discriminator region of the NT strand (−4, −3; also +2) are only half as reactive in I₂ as in RP_O. Therefore, Gries et al concluded that conversion of I₂ to RP_O involves a significant conformational change in the discriminator region of the NT strand, and a lesser conformational change in the upstream region of the T strand, but without a change in the start site region.

One interpretation of the difference in MnO₄⁻ reactivity of the discriminator is that the DNA backbone of this region is not as tightly held in I₂, allowing these bases to stack and hence reducing their MnO₄⁻ reactivity. Likewise, the backbone of the upstream region of the T strand may be somewhat less tightly bound in I₂ than in RP_O. These proposals are based on the observation, discussed above, that substantial base unstacking occurs in backbone-binding of single-stranded DNA to SSB [39]. Significant base unstacking in the conversion of I₂ to RP_O is a likely explanation of the large positive enthalpy change (△H^o) of this conversion, which is responsible for the observed negative activation energy of the dissociation rate constant [42,43].

The DNA closing step (k₋₂)

DNA closing rate constants (k₋₂) for all RNAP variants and both promoters cluster in a small range from about 0.6 s⁻¹(T7A1) to 3 s⁻¹ for λP_R (see Figure 6). Deletions of σ_1.1 or the jaw reduce the λP_R closing rate somewhat, but have no effect on the T7A1 closing rate. With △JAW RNAP, λP_R and T7A1 closing rates are the same. For λP_R, deletion of σ_1.1 or the jaw also reduces the forward isomerization rate which includes DNA opening [7, 14], but the effect on the intrinsic opening rate is not known [1].

Because nucleic acids are so highly charged, DNA helix formation kinetics in solutionare a strong function of salt concentration [44]. The observation that the rate of the DNA closing step on RNAP is independent of salt concentration is explained by the fact that the DNA is in the cleft, and the phosphates are interacting with cationic side chains of RNAP rather than with salt ions. Likewise the kinetics of isomerization of the advanced closed intermediate at λP_R, and consequently the thermodynamics of the DNA opening-closing step (K₂) are only weakly salt concentration dependent [4], very different from the situation for DNA opening-closing in solution.

The lack of effect of the jaw deletion on the DNA closing rate constant k₋₂ for T7A1 OC, together with the similarity of k₋₂ values for λP_R and T7A1 OC with △JAW RNAP, indicates that the DNA closing step for the WT RNAP-λP_R OC with the jaw present and originally engaged is facilitated relative to the situation where the jaw is missing (△JAW) or disengaged (T7A1). For T7A1, the jaw plays no role in DNA closing and in stabilizing OC, while for λP_R the jaw affects many steps, if not every step, of the mechanism, including DNA opening/closing [7].

Elsewhere we have discussed the analogy between the mechanism of OC formation and stabilization and that of enzyme catalysis [1]. The rate-determining DNA opening-closing steps are the analogues of the forward and reverse catalytic steps, which also are generally rate-determining. Regulation of the kinetics of enzyme catalysis is largely from effects of inhibitors, activators, solutes or cooperativity of multisubunit enzymes on the equilibrium constants of the initial steps of substrate binding and subsequent conformational changes, and we have proposed that the same is true of the kinetics of OC formation and dissociation. Changes in RNAP elements like σ_1.1 or binding of ligands or factors in the active site cleft are likely to affect the intrinsic opening-closing process. Changes in promoter sequence, solute and salt concentration, and in concentrations of ligands and protein factors that do not bind in the active site cleft are most likely to affect the equilibrium constants of the reversible steps that bind,bend, and wrap the promoter duplex to put the downstream duplex in the cleft prior to opening, and the equilibrium constant (K₃) that determines the stabilization of the OC.

Roles of these mobile elements earlier in transcription initiation

Each of the regions of RNAP investigated in this work has a large effect on OC lifetime for at least one of the two promoters studied. The presence of the β’ jaw and SI3, which form part of the DME assembly on the DNA downstream of the active site cleft, strongly increase OC lifetime at λP_R while σ_1.1 greatly reduces OC lifetime, especially at T7A1, and all of these effects are elaborately coupled. As discussed above, σ_1.1 and the β’ jaw also influence the rate of OC formation. In particular, deletion of σ_1.1 greatly reduces the rate of isomerization, including DNA opening [14]. By analogy with the behaviour of UT-47 vs. FL λP_R, we propose that deletion of σ_1.1 shifts the ensemble of closed complexes (I₁;see Figure 1) back to less advanced species in which the downstream DNA is not fully bent into the cleft, thereby greatly reducing the isomerization rate. Deletion of the β’ jaw reduces k₂ but, intriguingly, also appears to reduce K₁ [7]. Deletion of SI3 increases K₁, but has no significant effect on k₂ (E. Ruff, unpublished data). We hypothesize that by affecting multiple steps of OC formation, these various regions make RNAP exquisitely tuneable to regulation during initiation. This is an important property of group 1 σ factors and core RNAP, which orchestrate coordinated expression of thousands of promoters that differ greatly in their sequence signatures, strength, and regulatory inputs.

MATERIALS AND METHODS

Buffers

All buffers are the same as those used previously [7]. RNAP storage buffer contains 50% v/v glycerol, 0.01 M Tris (pH 7.5 at 4 °C), 0.1 M NaCl, 0.1 mM DTT, and 0.1 mM Na₂ EDTA. Tris buffer (TB) for kinetics experiments contains 0.04 M Tris (adjusted with HCl to pH 8.0 at the temperature of the experiment), 0.12 M KCl, 0.01 M MgCl₂, 1 mM DTT, and 100 μg/mL BSA. DTT was omitted for permanganate footprinting experiments.

Protein preparation

WT RNAP core and holoenzyme were endogenously expressed and purified from E. coli MG1655 as described [3], or were reconstituted using core expressed from pVS10 and purified as previously described [45]. No significant difference was observed between reconstituted and endogenously expressed WT holoenzyme. △JAW RNAP core (△β’1149-1190) was prepared from pIA1024 and purified as previously described [20]. △SI3 RNAP core (△β’ 943-1130) was prepared from pIA329 and purified as previously described [23] with minor modifications. σ⁷⁰ with residues 1-98 deleted (△98) was prepared from pIA389 and purified as previously described [45]. σ⁷⁰ with residues 1-55 deleted (△55) was prepared from pIA1129. All protein stocks were dialyzed into RNAP storage buffer and stored at −80°C. Working stocks were stored in RNAP storage buffer at −20 °C.

RNAP holoenzymes with σ_1.1 deletions (△55, △98) were reconstituted by incubating 5 μM WT, ?JAW, or △SI3 core RNAP with 5 μM σ△98 or △55 for 1 hour at 37 °C. Holoenzymes containing WT σ were reconstituted by incubating WT, △JAW, or △SI3 core RNAP with σ⁷⁰ in a 1:2 ratio at either 5 or 6 μM core for 1 h at 37 °C. Essentially 100% of RNAP reconstituted from His-tagged WT core and WT σ is active in OC formation, as judged by filter binding titrations at high (X? nM) concentrations of λP_R promoter DNA [42]. About 50% of RNAP reconstituted with △JAW core and WT σ is active in OC formation. However only about 10% of RNAP prepared from WT σ and △SI3 core, or from σ_1.1 variants (△55, △98) and WT or variant core, is active in OC formation. The inactive RNAP in these samples should not affect dissociation kinetic assays with excess competitor, since only the decay of the population of OC is being observed.

PCR synthesis and purification of promoter DNA

All DNA oligomers were obtained from IDT (Coralville, IA) and purified using standard desalting or PAGE. Radioactive nucleotides were ordered from Perkin-Elmer (Waltham, MA). All enzymes used in DNA preparation were ordered from NEB (Ipswitch, MA), except for Sequenase, which was obtained from Affymetrix (Santa Clara, CA).

λP_R DNA duplex fragments were prepared, radiolabeled and purified either as described [7] or by PCR amplification of a 120 base pair template with the WT λP_R sequence from base −60 to +20. For preparation of the 120 base pair fragment, two oligonucleotides with a central overlap of 13 bp were designed to generate λP_R fragment (−70 to +31): 5’ end: 5’-CCA CGAATT CGG ATA AAT ATC TAA CAC CGT GCG TGT TGA CTA TTT TAC CTC TGG CGG TG-3’; and 3’ end: 5’-ACA AAA GCT TCA TAC AAC CTC CTT AGT ACA TGC AAC CAT TAT CAC CGC CAG AGG T-3’. These oligos were mixed with an excess of short primers: HTOP: 5’-CCA GCA TTC CTC CAC GAA TTC-3’; and HBOT: 5’-CAC CTG CAC CGA CAA AAG CTT-3’, extended using Taq or Vent DNAP, and purified using QIAquick PCR purification kits (Qiagen). ³²P-labeled λP_R DNA was generated by PCR using γ-³²P-labeled HTOP primer. Kinetic data obtained with these fragments for WT and △JAW RNAP agree with previously published results on promoter fragments prepared by other methods [5, 7, 23, 43].

T7A1 DNA duplex fragments were prepared, radiolabeled, and purified by PCR amplification of a 120 base pair template with the WT T7A1 sequence from base −60 to +1. Two oligonucleotides with a central overlap of 13 bp were designed to generate T7A1 fragment (−70 to +31): 5’ end: 5’-CCA CGA ATT CAA TTT AAA ATT TAT CAA AAA GAG TAT TGA CTT AAA GTC TAA CCT ATA G-3’; and 3’ end: 5’-ACA AAA GCT TAT TCG CCG TGT CCC TCT CGA TGG CTG TAA GTA TCC TAT AGG TTA GAC-3’. These oligos were mixed with an excess of short primers, HTOP and HBOT, extended using Taq or Vent DNAP, and purified using QIAquick PCR purification kits (Qiagen).³²P-labeled T7A1 DNA was generated from PCR using γ-³²P-labeled HTOP primer.

T7A1 footprinting data were obtained using a longer fragment with the T7A1 sequence from base −60 to base +2: 5’-AGT CAC TCG AGG GCC CAG GAA ACA GCT ATG ACC ATG ATT ACG CCA AGC TTG CAT GCC TGC AGG CTT GAA CAG TAT TTA AAA TTT ATC AAA AAG AGT ATT GAC TTA AAG TCT AAC CTA TAG GAT ACT TAC AGC CAT CGA GAG GGA CAC GGG GAA GGA ACA ACG CAT AAC CCT GAG ATA TCG ATC CAT GCG CG-3’ This fragment was purchased as two single strands, annealed, and then amplified by PCR using primers “T7A1_downstream_XhoI,” 5’-CGC TCG AGG ATC GAT ATC TCA GGG TTA TGC G-3’ and “T7A1_upstream_EcoRI,” 5’-AGT CAG AAT TCG GCC CAG GAA ACA G-3’. To end-label the NT strand of the PCR-amplified fragment, after cutting with XhoI (New England Biolabs), the end was filled in using Sequenase with α-³²P-labeled dTTP and unlabelled dATP, dGTP and dCTP.³²P-labeled T strand DNA was generated by PCR using γ-³²P-labeled “T7A1_downstream_XhoI” primer.

Unlabelled +UP-λP_R promoter fragment was made by PCR from the pCAD5 plasmid, prepared using a previously-described protocol [46] with minor modifications [7]. For use as competitor in the dissociation assays reported here, 172 base pair λP_R +UP fragments (extending from −130 to +42) were prepared by PCR amplification of pCAD5 using the following primers: 5’ end: 5’-A GGA CCC GGG AGT CAC GAC GTT GTA AAA CGA CGG CCA GTG AAT TCG AGG-3’; 3’ end: 5’-GA GTA CTC GAG CGG GAT CCG ATA TCT AGA TAC AAC CTC CTT AGT-3’.

PCR-labelled promoter DNA for most dissociation experiments was purified using spin columns (Qiagen). PCR-labelled promoter DNA for footprinting and some dissociation experiments was purified further, either by excising the promoter band after electrophoresis on a 5% acrylamide gel and eluting from an Elutip-d Purification Minicolumn (Whatman; GE Healthcare, Pittsburgh, PA), or by electrophoresis on a 1.3% low-melt agarose gel and subsequent purification with a Wizard SV Gel and PCR Clean-Up Kit (Promega; Madison, WI).

Nitrocellulose filter binding dissociation kinetic assays

Dissociation kinetic assays were performed as previously described [7, 47]. OCs were formed by incubating 2-5 nM RNAP (total concentration) with ~0.5 nM ³²P-labeled promoter DNA fragment for 0.5-3 hours at 37 °C. Irreversible dissociation was initiated by addition of heparin (at specified concentration) or 20-50 nM +UP-λP_R unlabelled promoter (in ~10-fold excess over the total concentration of RNAP). For dissociation assays lasting longer than ~2 days, DTT concentration was increased to 6 mM and reactions were supplemented with DTT throughout the time course. Salt-upshift burst dissociation assays at 1.1 M KCl, used to determine rate constant k₋₂, were performed as previously described [5].

Analysis of kinetic data

Analysis of dissociation kinetic data

Dissociation of RNAP-promoter OC is single-exponential. Observed dissociation rate constants k_d were determined by fitting data to the equation

$${\theta }_1={\theta }_0+({\theta }_{\mathrm{plateau}}-{\theta }_0)\exp (-{\mathrm{k}}_{\mathrm{d}}\mathrm{t})$$ Equation 1

where θ_t, θ₀, and θ_plateau are the fractions of promoter DNA present as open complexes at time t, at time t = 0, and at t → ∞, respectively, determined by filter binding as in [7]. For OC formed with WT RNAP and PCR-labelled, gel-purified promoter DNA, θ₀ was typically ~90-100% (after correction for filter efficiency), and θ_plateau was ~5-10%. For OC formed with WT RNAP and PCR-labelled promoter DNA purified by spin columns (Qiagen), θ₀ was typically ~60-70% (probably due to residual labelled primers), and θ_plateau (stable over ~5 days) was about 30%. For RNAP variants, θ₀ was often lower, and in the worst case, for the weaker binding, less active RNAP deletion variant △98/△SI3, values of θ₀ were as low as 15%, but always at least 50% larger than θ_plateau. Control experiments were performed comparing DNA purified by gel and by spin columns, and the k_d values were found to be the same within uncertainty.

All data were fit nonlinearly to Equation 1 using the programs Sigmaplot or Igor 3.0. No significant difference was observed in fitting parameters obtained with these two programs. Determination of the DNA closing rate constant k₋₂ and the OC stabilization equilibrium constant K₃ from salt upshift dissociation kinetics (burst experiments) Although the mechanism of dissociation is multi-step (Figure 1), the kinetics of dissociation are single exponential for both promoters and all RNAP variants studied here. Single-exponential kinetics indicates that the unstable open intermediate I₂ rapidly reverses to the stable OC (RP_O at λP_R), on the time scale of its conversion of I₂ to I₁ (with DNA-closing rate constant k₋₂) [5]. Hence the observed dissociation rate constant k_d is interpreted as:

$${\mathrm{k}}_{\mathrm{d}}=\frac{{\mathrm{k}}_{-2}}{1+{\mathrm{K}}_3}$$ Equation 2

where K₃ is the equilibrium constant for conversion of I₂ to the stable OC (RP_O at λP_R) [5].

Salt upshift “burst” experiments with the rapid quench mixer are used to determine k₋₂. These take advantage of the observation that k₋₂ is independent of salt concentration while K₃ decreases strongly with increasing salt concentration. At or above 1.1 M KCl, K₃ << 1 and k_d = k₋₂, whereas at the low salt conditions of the standard dissociation assay for k_d typically K₃ >> 1 and k_d = k₋₂/K₃. Since k₋₂ is salt-independent, K₃ at the low-salt dissociation condition is obtained from K₃ = k₋₂/k_d. Uncertainties in reported values of K₃ are determined by propagation of the independent uncertainties in k₋₂ and k_d.

Determination of experimental uncertainty

Reported values of k_d and k₋₂ for dissociation of final (stable) and initial open complexes are averages of at least two, and in some cases as many as five, independent experiments. For two determinations, experimental uncertainties are estimated from differences between the average and the individual k_d values; for larger data sets, the uncertainty represents the standard error of the mean. Where the uncertainty determined by these methods was less than 10%, we have assigned a 10% uncertainty as noted in Tables 1 and 2.

KMnO₄ footprinting

OC were formed by incubating 20 or 40 nM RNAP (active concentration) with ~1 nM promoter DNA in modified TB (without DTT) for 1 hour at 37 °C. When indicated, samples were challenged for 10 s with 20-50 μg/ml heparin, after which 0.2 mM KMnO₄ was added for 10 s, followed by a quench with 1 M β-mercaptoethanol. Subsequent workup and analysis were as described [7].

HIGHLIGHTS.

• Open complex (OC) lifetime is a transcription regulation target in E. coli .

• In stable λP_R OC, discriminator binds to a track in cleft and duplex binds to β’jaw.

• In less stable T7A1 and λP_R I₃ OC, σ_1.1 binds to same discriminator track and β’ jaw.

• Deleting both σ_1.1 and β’ jaw or β’SI3 equalizes OC lifetimes for λP_R and T7A1.

• Understanding roles of these RNAP mobile elements will facilitate design of antibiotics.

Abbreviations

OC

open complex

DME

downstream mobile element

NT

nontemplate

T

template

CRE

core recognition element