One-step DNA melting in the RNA polymerase cleft opens the initiation bubble to form an unstable open complex

Abstract

Though opening of the start site (+1) region of promoter DNA is required for transcription by RNA polymerase (RNAP), surprisingly little is known about how and when this occurs in the mechanism. Early events at the λP_R promoter load this region of duplex DNA into the active site cleft of Escherichia coli RNAP, forming the closed, permanganate-unreactive intermediate I₁. Conversion to the subsequent intermediate I₂ overcomes a large enthalpic barrier. Is I₂ open? Here we create a burst of I₂ by rapidly destabilizing open complexes (RP_o) with 1.1 M NaCl. Fast footprinting reveals that thymines at positions from -11 to +2 in I₂ are permanganate-reactive, demonstrating that RNAP opens the entire initiation bubble in the cleft in a single step. Rates of decay of all observed thymine reactivities are the same as the I₂ to I₁ conversion rate determined by filter binding. In I₂, permanganate reactivity of the +1 thymine on the template (t) strand is the same as the RP_o control, whereas nontemplate (nt) thymines are significantly less reactive than in RP_o. We propose that: (i) the +1(t) thymine is in the active site in I₂; (ii) conversion of I₂ to RP_o repositions the nt strand in the cleft; and (iii) movements of the nt strand are coupled to the assembly and DNA binding of the downstream clamp and jaw that occurs after DNA opening and stabilizes RP_o. We hypothesize that unstable open intermediates at the λP_R promoter resemble the unstable, transcriptionally competent open complexes formed at ribosomal promoters.

Interactions between RNA polymerase and specific promoter DNA sequences trigger a precise progression of conformational changes in both biomolecules. Taken together, these steps constitute the mechanism of DNA opening and the start of the transcription cycle. For Escherichia coli RNA polymerase holoenzyme (RNAP, subunit composition: α₂ββ^′ωσ⁷⁰), binding free energy drives opening of the initiation bubble (-11 to +2, numbering relative to the start site base +1) in promoter DNA, placement of +1 template base in the active site of the enzyme, and subsequent conformational changes to form the stable open complex RP_o. Each of these steps provides a checkpoint for regulatory input.

Over the past decade, structural (X-ray, FRET), single-molecule, and rapid mixing kinetic studies have greatly advanced the understanding of this machinery and these steps. Key advances include: (i) elucidation of the RNAP architecture at atomic resolution (1–4); (ii) dissection of composite forward and backward rate constants for RP_o formation into individual rate and/or equilibrium constants for the steps leading to RP_o (5, 6); (iii) single-molecule measurements of DNA topological changes (7); (iv) real-time determination of hydroxyl radial (HO•) protection patterns of DNA during stable open complex (RP_o) formation (8–10); and (v) finding that unstable open complexes are stabilized by binding the initiating nucleoside triphosphate and greatly destabilized by the stress response factors ppGpp and DksA (11–13). These advances make it possible to address the key unresolved questions of initiation. How is the opening of 12–14 base pairs distributed between the steps of RP_o formation? Does RNAP disrupt the DNA duplex in the active site cleft or does DNA melt outside the channel and enter as individual strands?

Evidence for at least two kinetically significant intermediates (generically designated I₁ and I₂) preceding RP_o exists for a variety of promoters recognized by E. coli RNAP (cf. refs. 6, 8, and 14–17). Conversion of I₁ to I₂ is the rate-determining (bottleneck) step in forming the open complex at the λP_R promoter and exhibits a 34-kcal activation enthalpy barrier (17). The reverse direction of this step is the bottleneck step in dissociation of RP_o (6). Because I₁ is a closed complex (9), determining when DNA opens requires trapping I₂.

The minimal mechanism of RP_o formation is formally analogous to minimal mechanisms of solute transport through membranes and enzyme catalysis. All involve three steps in which the initial step in each direction is rapidly reversible and a middle step that is the bottleneck in both directions. In RP_o formation, as in mechanisms of catalysis and transport, ligands and solutes primarily act on the rapidly reversible steps and not on the central bottleneck step (6, 18). Given these analogies, is the central bottleneck step of open complex formation indeed DNA opening, just as transport is the central step in transporter mechanisms and catalysis is in enzyme mechanisms?

We address this fundamental mechanistic question by using a powerful method from physical enzymology, the burst experiment, which forms a transiently high concentration of an otherwise unobservable intermediate (preceding the bottleneck step). In the forward direction, a burst of I₁ is generated by rapid mixing with a sufficiently high [RNAP] (cf. Fig. 1A). HO• and permanganate () footprints of the population of the λP_R promoter DNA in such a forward burst experiment demonstrate that I₁ is a -unreactive complex in which downstream duplex DNA is protected to +20 from HO• attack (9). Structural modeling on the basis of these data and the X-ray structures of the bacterial RNAP (1, 2) indicates that duplex DNA in I₁ is loaded in the active site cleft of RNAP but not yet open (9).

Fig. 1.

A burst dissociation experiment is required to characterize the unstable intermediate I₂. Simulations of changes in the populations of RP_o (Red), I₂ (Green), I₁ (Blue), and free λP_R promoter DNA (P, Gray) at 10 °C, 0.01 M Mg²⁺ as a function of time are shown for: (A) reversible association at high [RNAP] (100 nM) and 0.12 M salt (KCl or NaCl); (B) irreversible (in excess competitor) dissociation at 0.12 M salt; and (C) irreversible dissociation after a rapid upshift to 1.1 M salt. Rate and equilibrium constants for the association simulation (A) are from ref. 17; at no point on the association time course is there a significant population of I₂. Rate and equilibrium constants for the dissociation simulations are from ref. 6. The salt upshift (C) produces a large transient burst of I₂, not observed in the low salt dissociation experiment (B). The species designated RP_o also includes a minor population of the late intermediate I₃ (6), which is not resolved from RP_o in the present studies. Values of the parameters used to generate these simulations are given in Table S1.

Because the rate limiting forward step is the conversion of I₁ to I₂, no subsequent burst of I₂ occurs in the forward direction experiment (Fig. 1A). Hence the dissociation direction must be investigated to obtain a sufficient population of I₂ to characterize. The time course of a standard dissociation experiment in which a competitor such as heparin is added to make dissociation irreversible is shown in Fig. 1B. Because I₂ is an unstable intermediate, rapidly converting back to the stable open complex on the time scale of its conversion to I₁, it never accumulates to a significant level in this experiment. Kontur et al. (6) discovered that rapid destabilization of the stable open complex with moderately high concentrations of urea or salt generates a dramatic transient buildup (burst) of I₂ ∼0.5 s after mixing (Fig. 1C). (A RP_o-destabilizing temperature downshift cannot be performed rapidly enough to detect such bursts.) Because the rate of conversion of I₂ to I₁ is found to be independent of urea or salt concentration, the burst of I₂ persists for a period of approximately 1 s, ample time for characterization of the extent of opening of bases in I₂ and of the decay of I₂ to I₁ by fast footprinting.

In Fig. 1 (all panels) at 10 °C, the stable open complex (labeled RP_o) may be an equilibrium mixture containing some of the intermediate complex I₃ identified previously (6). An increase in the population of I₃ is expected early in the solute upshifts in Fig. 1C, decaying to I₂ in less than 100 ms. Simulations on the basis of kinetic data (6) show that the time resolution of the three-syringe burst/fast footprinting experiments reported in this research is insufficient to investigate I₃, and its population is therefore combined with that of RP_o in Fig. 1.

Results

To determine whether the DNA in I₂ is closed or partially or fully open in the region of the initiation bubble, we footprinted thymines with a constant dose of (19). Complexes were probed as a function of time after rapidly destabilizing open complexes with 1.1 M NaCl or after mixing them with 0.12 M NaCl (control reaction). The sequencing gels in Fig. 2 compare the time-dependent behavior of all -reactive positions on each strand after the upshift with that of the control reaction.

Fig. 2.

Visualizing open thymines on template and nontemplate strands during the burst and subsequent decay of intermediate promoter complex I₂ after an upshift of RP_o to 1.1 M NaCl. Sequencing gels (representative of three independent experiments) show the decay of permanganate reactivity of individual open thymine bases (left lanes) on the template (A) and nontemplate (B) strands as a function of time after upshift, probed with a constant dose of NaMnO₄ (see Material and Methods). Right lanes show corresponding low salt RP_o reactivity controls for both strands as a function of time after mixing with 0.12 M NaCl.

For the template (t) strand (Fig. 2A) the control lanes indicate that thymines at positions +1, -8/-9 (doublet band), and -11 are -reactive in RP_o. After the upshift to 1.1 M NaCl, a monotonic decay of reactivity of these bands is observed in the time range 0.1–10 s. For the nontemplate (nt) strand (Fig. 2B), the control lanes indicate that thymines at positions +2 and -3/-4 (doublet band) are -reactive in RP_o; the kinetics of the decay in reactivity with time after the NaCl upshift is very similar to that of the t strand. Thymines at -7 and -10 on the nt strand are not detected, though the DNA is open in this region (as judged by reactivity at positions -11 and -8/-9 on the t strand). We infer that interactions of these upstream nt strand thymines with σ⁷⁰ region 2 (cf. ref. 20 and references therein) protect them from reacting with in I₂ as well as in RP_o.

In addition to providing visual demonstrations of the positions of reactive thymines in I₂ and of the time course of their decay as I₂ converts to products (I₁ and then free promoter DNA), Fig. 2 allows a visual comparison of the reactivities of these thymines in I₂ [judged by the early time points (0.1–0.25 s) after the upshift], both relative to other thymines in I₂ and to the RP_o control. At 0.25 s, the population distribution of promoter DNA (Fig. 3A) is 80% I₂, 10% RP_o and I₃, and 10% closed I₁ and free promoter DNA. Strikingly, in the early time lanes where I₂ is the major species, thymines at both +2 and -3/-4 on the nt strand appear much less reactive than in the RP_o control, whereas reactive thymines at +1 and other positions on the t strand appear nearly as reactive as in the RP_o control.

Fig. 3.

Kinetics of the open to closed transition of individual thymines as the burst population of I₂ decays to I₁ and promoter DNA. (A) Predicted populations of RP_o and I₂ (see SI Text) are plotted versus time (log scale) after NaCl upshift (0.1–10 s). (B–F) Relative reactivities of each thymine singlet or doublet band (calculated from three independent experiments on each strand; see Fig. 2) as a function of time (log scale). Solid curves are single exponential fits of these data, yielding the rate constants for the decay of I₂ in Table 1. Simulations of the decay using the population distribution in Fig. 1C and varying the reactivity of each thymine singlet or doublet in I₂ relative to the RP_o control are shown as dashed lines. The best fit is shown in red; the limits of the fit (I₂ closed (0% reactive) and I₂ as open as RP_o (100%)) are shown in gray.

To explore the behavior of each -reactive thymine during dissociation, we quantified their individual decay kinetics and reactivities relative to the RP_o control (Fig. 3 and Table 1). In Fig. 3A, the populations of I₂ and of RP_o (including I₃) are plotted as a function of time after the upshift on a logarithmic time scale (two decades, from 0.1 to 10 s) to compare with the analysis of the individual thymines shown in the subsequent panels. For all thymines, the observed change in reactivity is well fit by a single exponential decay (Solid Curve, Fig. 3 B–F). Although I₂ is initially the dominant species after the upshift, the presence of some RP_o, I₃, and closed promoter DNA requires that the data be deconvoluted to quantify the reactivity of each thymine in I₂. A series of simulations of the observed decay kinetics were performed in which these reactivities were systematically varied from 0% (unreactive in I₂) to 100% (as reactive in I₂ as in RP_o). Simulated time courses calculated by using the best-fit reactivities of each thymine in the bubble in I₂ are compared with the 0% and 100% limiting cases and with the experimental data in Fig. 3 B–F. Best-fit reactivities are listed in Table 1 together with the rate constants for decay of each reactive thymine in the conversion of I₂ to I₁.

Table 1.

Permanganate reactivities of thymine bases in the open region of I₂ and rate constants for closing these positions in I₂ to I₁

Strand	Template			Nontemplate
Position	−11	−9/−8	+1	−4/−3	+2
Rate constant (s-1)*	0.50 (±0.04)	0.52 (±0.05)	0.6 (±0.1)	0.7 (±0.2)	0.7 (±0.3)
Predicted reactivity of I2 (relative to RPo control)†	80%	80%	100%	50%	55%

^*Fit parameters from the decay kinetics (solid lines) in Fig. 3. The rate constant for the conversion of I₂ to I₁ (k_-2) determined by nitrocellulose filter-binding experiments at 10 °C is 0.72 ( ± 0.07) s^-1 (6).

^†Uncertainty in these reactivities, estimated from the range of triplicate determinations at a given time point in the range from 0.1 to 0.75 s, is ± 10%. See Material and Methods.

Comparison of the results reveals that:

1. Decay rate constants are the same for all thymines (0.6 ± 0.1 s^-1) and agree within the uncertainty with the rate constant for the conversion of I₂ to I₁ (k_-2 = 0.72 ± 0.07 s^-1) determined by filter binding of quenched samples as a function of time after exposure to a 1.1 M salt upshift (6).

2. The reactivity of the start site (+1) thymine on the t strand in I₂ is the same as in the RP_o control. reactivities of upstream thymines on the t strand in I₂ are about 80% as large as in the RP_o control. reactivities of downstream thymines on the nt strand in I₂ are 50–55% as large as in the RP_o control. (These conclusions are insensitive to assumptions regarding the amount and reactivity of the I₃ present in the residual population labeled RP_o in Fig. 3A) Thymines at -7 and -10 on the nt strand are fully protected (zero reactivity) in both I₂ and RP_o.

Discussion

Opening of the Initiation Bubble Occurs in the Active Site Cleft to Convert the Closed Complex I₁ to I₂.

The architecture of multisubunit RNA polymerases appears to have evolved a series of steric blocks to prevent access of nonpromoter DNA to the active site (2–4, 9, 21, 22). For example, bacterial RNAP recognizes promoter DNA by interactions between the σ subunit and hexameric sequences (-10 and -35 regions) that are upstream of +1 (cf. ref. 23 and references therein). Placement of σ⁷⁰ with respect to the channel requires the DNA to bend sharply at -11/-12 to enter the active site cleft formed by the β and β′ “pincers” (or jaws) (2, 17). Does double-stranded DNA bind in the cleft at this point in the mechanism prior to being opened by RNAP? Or does the DNA open above the cleft, allowing the template strand to then descend down to the active site “floor” (2, 23)?

The relatively “closed” state of the pincers (less than 25 Å apart) observed in crystal structures of the bacterial RNAP (2, 22) and the transcription factor TFIIB bound to the 12-subunit eukaryotic RNAP (24, 25) has motivated proposals that DNA must open outside of the cleft. For the bacterial RNAP, which lacks a helicase cofactor, opening outside the active site cleft is proposed to be nucleated by a thermal breathing mechanism (23). In this proposal, transient opening and closing of the A/T-rich -10 hexamer leads to capture of the nt strand by σ⁷⁰ region 2 at the upstream entrance (23), followed by entry of only the t strand into the cleft (2, 23). Indeed an RNAP subassembly consisting of σ⁷⁰ region 2 and an N-terminal fragment of β′ was observed to form an open (-reactive) complex with a highly A/T-rich promoter set in negatively supercoiled DNA; this polymerase subassembly did not open this promoter on linear DNA (26). Whereas these structure- or equilibrium-based mechanistic hypotheses are appealing and could apply to promoters under highly negative supercoiling stress and/or high temperature, E. coli RNAP readily forms open complexes on linear promoter fragments in vitro.

Kinetic-mechanistic studies of RP_o formation and dissociation at the λP_R promoter combined with footprinting data argue that interactions of regions of E. coli RNAP with promoter DNA bound in the cleft actively distort and open the initiation bubble. HO• and DNase I cleavage of the DNA backbone of the early intermediate I₁ demonstrate that both the t and nt DNA strands are protected without interruption in this region (-15 to +25), whereas does not detect any unstacked thymine bases (9, 27, 28). For the T7A1 promoter, time-resolved HO• footprints of populations of intermediates during formation of stable open complexes have been interpreted in terms of a mechanism involving three classes of intermediates before the rate-determining step (10). Comparison of the kinetics of development of downstream reactivity with the kinetics of downstream HO• protection led the authors to propose that DNA opening occurs outside the cleft and before the rate-determining step (10). Further research is needed to determine the nature of the rate-determining step at T7A1 and to understand the origins of the apparent mechanistic differences between these two promoters.

Experiments presented here demonstrate that -reactive thymines appear in the conversion of I₁ to I₂ at λP_R. We conclude from these results that the pincers of RNAP are sufficiently flexible in solution to allow DNA to enter as a double helix, where it is then opened via binding interactions with elements on RNAP. Additional evidence that DNA opening occurs in the active site cleft and not in solution is provided by the observation that the [salt] dependence of the DNA opening step (conversion of I₁ to I₂) is much smaller in magnitude than that of melting 12–14 base pairs of DNA in solution (18). We proposed that the N-terminal polyanionic domain of σ⁷⁰ (σ region 1.1) must move in the cleft, allowing the duplex DNA to descend (18). These movements would position -2/-1(t) near the highly conserved region on the downstream lobe of β known as fork loop 2. By analogy with base-flipping enzymes, we speculate that fork loop 2 inserts in the minor groove, creating a 90° bend and unwinding the DNA helix to form the I₁-I₂ transition state. Interactions between the DNA phosphate backbone and positive regions in the cleft provide an additional driving force for opening. A bind-bend-open mechanism has also been proposed for RP_o formation by the single subunit T7 phage RNAP and tested by stopped flow kinetic and FRET experiments (29, 30).

Single-molecule DNA magnetic tweezer experiments revealed that unwinding of ∼1 turn of the helix in an E. coli RNAP–promoter complex occurred in a single process at all promoters studied on both negatively and positively supercoiled DNA (7). Because the time resolution of the assay was ∼1 s, these experiments could not resolve whether untwisting occurred in a single kinetic step, or whether the mechanism involved a sequence of unwinding steps. Our results show that opening (unpairing, partial unstacking, and therefore unwinding) occurs in the bottleneck step of RP_o formation (I₁ to I₂) at the λP_R promoter.

Proposed DNA Conformations in the Steps in Open Complex Formation.

Fig. 4 shows a schematic of the proposed series of conformational changes in the downstream DNA in the RNAP active site channel as the reaction proceeds from the closed intermediate I₁ to the unstable open complex I₂ and finally to the stable open complex RP_o. DNA in I₁ is shown as double-stranded with the exception of a distortion at -11(t), which flips out -11A(nt) (20). However, because -11(t) is not -reactive, we infer that it remains stacked with other DNA bases and/or is interacting with RNAP (9). In our model of I₁, the +1 base pair lies ∼40 Å above the active site Mg²⁺ (9, 17). Further descent of duplex DNA in I₁ is blocked by σ region 1.1 and the β′ “bridge” helix that spans the channel.

Fig. 4.

Proposal for DNA conformational changes during the steps of opening and stabilizing the initiation bubble at the λP_R promoter. Proposed location (relative to the active site Mg²⁺, shown as a red dot) and extent of bending and/or opening of DNA strands (template, blue; nontemplate, green; -20 to +20) in the bent and wrapped closed complex I₁, in the relatively unstable initial open complex I₂, and in the stable open complex RP_o. The transcription start site thymine is shown as a red rectangle. -11A (Green Rectangle) on the nontemplate strand is proposed to be flipped out in the 90° bend that directs the duplex into the active site cleft in I₁. Formation of I₁ loads downstream duplex in the active site cleft. RNAP opens the DNA in the cleft and positions the t strand start site base (+1) in the active site, forming I₂. Final loading of the nt strand and assembly of a clamp/jaw (Gray Rectangles) on the downstream duplex DNA greatly stabilize RP_o relative to I₂.

Because the reactivity of +1 appears to be the same in both I₂ and RP_o, we propose that opening of the bubble in I₁ → I₂ places the +1(t) base in the active site (near the catalytic Mg²⁺). However, the reduced reactivities of other thymines, especially those on the nt strand (Table 1), motivate the proposal that the surrounding protein environment and/or degree of stacking of these thymines in I₂ differs from that in RP_o. Large conformational changes in RNAP accompany the conversion of I₂ to RP_o (5, 15). Very large solute effects on the dissociation rate constant k_d and the large activation heat capacity of k_d provide evidence that the downstream clamp/jaw is assembled and tightened on the downstream DNA duplex (+5 to +20) in these late steps of RP_o formation, after the DNA has been opened (6, 18, 31) (see Fig. 4). The twofold increase in permanganate reactivity of nt strand thymines at -4/-3 and +2 in the conversion of I₂ to RP_o indicates that the nt strand is repositioned and/or unstacked in these steps. These local changes in the nt strand in the cleft may be coupled to changes in positioning of σ⁷⁰ region 1.1 and the switch regions (3, 32) and to the large scale downstream conformational changes involved in assembly and DNA binding of the clamp/jaw. Together these events greatly stabilize RP_o relative to I₂.

Opening of the entire bubble (-11 to +2) is rate-determining at the λP_R promoter and has the properties of a single (elementary) kinetic step. At other promoters, opening may be separated into several kinetically distinguishable steps and may or may not be rate-determining (10, 20). For the λP_R promoter, the activation enthalpy barrier is very high for the forward direction of the opening step [34 kcal; (17)]; and the overall enthalpy change for this step is also large [∼24 kcal; (6)]. For comparison, the enthalpy of melting 13 bp of DNA in solution is approximately 75 kcal at 25 °C (33). One interpretation of the 34-kcal activation barrier is that approximately half the bubble is open and unstacked in the transition state and that few enthalpically favorable interactions with RNAP have formed at this stage. Alternatively, the majority of bases in the bubble may be unpaired but not unstacked or, if unstacked, may be engaged in enthalpically favorable interactions with RNAP. Opening in the cleft is likely initiated at the -11 bend by interactions between the -10 region of the nt strand and aromatic residues of σ⁷⁰ region 2 (20). Because these interactions presumably are enthalpically favorable, we infer that the majority of the bubble is open in the transition state. Conversion of this very unstable transition state to I₂ likely establishes additional interactions with RNAP, possibly including those between the t strand and switch 2 of β′ (32).

Proposed Physiological Relevance of the Unstable Open Complex I₂.

Under physiological conditions, I₂ is highly unstable relative to RP_o at the λP_R promoter and converts to RP_o so rapidly that it never accumulates (Fig. 1 A and B). Yet the conversion of I₁ to I₂ opens the entire transcription bubble and may correctly load the start site base in the active site. Do multiple open complexes (RP_o, I₃, and I₂) with large differences in stability play distinct functional roles in the regulation of transcription initiation? We propose that the answer to this question is strongly affirmative.

Extensive studies of transcription initiation at the ribosomal rrnB P1 promoter by Gourse, Ross, and coworkers reveal that the open complex formed at this promoter in the absence of NTPs and negative supercoiling is highly unstable, with a short lifetime, existing in an equilibrium shifted toward closed complexes (cf. ref. 34 and references therein). If open complex formation at λP_R were blocked at I₂ or I₃, unable to progress to the stable RP_o, then a similar situation would be observed. An equilibrium would exist between these unstable open intermediates and the closed complex I₁ shifted toward I₁ at 25 °C and with a lifetime of I₂ of only a few seconds. On the basis of this kinetic/thermodynamic analogy, and because I₂ at λP_R and the functional open complex at rrnB P1 both appear to be the first open complex formed from a closed complex at these promoters, we propose that the unstable intermediate(s) I₂ (and/or I₃) at λP_R behave functionally and structurally like the unstable open complex characterized at rrnB P1.

A key functional property of the open complex at rrnB P1 is the ability to initiate synthesis of a full-length transcript rapidly and efficiently, without any short, abortive product synthesis, upon addition of all four NTPs (35). A key structural property of the rrnB P1 open complex is its shorter downstream boundary of the hydroxyl radical footprint (∼+12 to +15 of the (t) strand for the rrnB P1 (36), compared to +20 to +25 for RP_o at λP_R (9). This difference in downstream boundaries likely reflects a difference in the extent of assembly and DNA binding of downstream mobile domains in β′ including the clamp, jaw, and sequence insertion 3, which we propose stabilize RP_o at λP_R (6, 18, 31). We hypothesize that these hallmarks of the open complex at rrnB P1 will be observed for I₂ (and/or I₃) at λP_R.

The instability (relative to the closed complex) and short lifetime of the open complex at ribosomal promoters makes it a target of regulation by proteins like DksA (12) and by ligands including the stress factor ppGpp (37, 38) and the initiating NTP (11). Clearly, ribosomal expression has been tuned to respond rapidly to changes in conditions, including changes in NTP concentrations. We hypothesize that complete assembly and tightening of the clamp/jaw on downstream DNA characteristic of the stable open complex RP_o at λP_R is disfavored at the rrnB P1 promoter [possibly because of differences in interactions with the nt strand in the cleft (39, 40)], explaining its instability relative to λP_R. If so, tight downstream interactions would not need to be broken for RNAP to escape from the rrnB P1 promoter, allowing highly efficient production of full-length transcripts under exponential growth conditions. In contrast, a fully assembled and tightened clamp at λP_R in RP_o may impede escape and favor abortive initiation, consistent with the observation that downstream DNA from +1 to +20 plays a key role in determining the efficiency of the transition to elongation (41).

For a series of promoters that form stable binary open complexes, Hsu, Chamberlin and colleagues obtained evidence for two ternary initial transcription complexes (ITCs): One ITC makes only abortive products and the other primarily makes full length transcripts. For these promoters, the overall abortive:productive ratio is large and promoter-sequence-dependent [(42); see also ref. 43]. Abortive transcripts are made in vivo (44); their cellular function as small RNAs is not yet known.

What is the relationship between multiple binary open complexes (I₂, I₃, and RP_o) and multiple ITCs? We propose that stable RP_o, with its tightly gripping downstream clamp/jaw, is capable of only abortive synthesis, whereas less-stable I₂ and I₃, in which the clamp/jaw is less assembled and/or less tightly bound to downstream DNA, are capable of productive initiation. In this model, the relative populations of stable (RP_o) and less stable (I₂ and I₃) open complexes determine the observed abortive:productive ratio. The fact that the clamp/jaw assembles only after DNA opening and untwisting is complete indicates that its grip likely impedes the movement of duplex DNA within it. Promoter sequence, sigma factor, concentrations of regulatory ligands, and solution conditions all must affect the relative stability and lifetime of each initiation intermediate and hence are determinants of initiation rate and the abortive:productive ratio, providing potent and relatively unexplored avenues for the regulation of transcription.

Material and Methods

Solutions and Materials.

Standard methods of purifying RNAP and of obtaining ³²P-labeled DNA fragments were used. See SI Text.

High [NaCl] Fast-Kinetic MnO₄ DNA Footprinting.

Labeled DNA promoter fragments in BB were mixed with RNAP in SB at a final concentration of 10 nM and incubated at room temperature (∼20 °C) for 90 min to preform open complexes. These complexes were loaded into the sample A tube of a KinTek Corporation RQF-3 Rapid Chemical Quench-Flow instrument cooled to 10 °C by a circulating water bath. Push syringe A was loaded with BB supplemented to 4% SB. The sample B tube and push syringe B were loaded with BB supplemented to 1.96 M NaCl, 400 μg/mL heparin, and 4% SB. Push syringe C was loaded with a solution containing 200 mM NaMnO₄ and 900 mM NaCl. Collection tubes were filled with 300 μL of a quench solution containing 3.75 M ammonium acetate and 7.1 M β-mercaptoethanol. The quench-flow instrument was operated in a push-pause-push-pause-push mode. The first push rapidly mixed the preformed open complexes with the high salt solution resulting in a final [NaCl] of 1.1 M. This solution was held in reaction loop 7 for the desired perturbation time. The second push mixed the contents of reaction loop with the solution in push syringe C resulting in a final [] of 66.7 mM. This solution was held in the exit tube for 150 ms before the final push expelled the solution into the collection tube containing quench solution. Quenched reactions were immediately ethanol precipitated. Each load–reaction cycle took 250 min. Low salt control reactions were performed as above with the exception of loading solutions that keep [NaCl] at 120 mM. DNA fragments were washed with 70% ethanol, resuspended, reprecipitated, and washed. Modified fragments were cleaved by incubation at 90 °C in 1 M piperidine. Reactions were evaporated and resuspended in TE buffer (10 mM Tris HCl, 1 mM EDTA, pH 8.0) three times. The resulting DNA was resuspended in urea loading buffer and resolved on an 8% acrylamide gel. The gel was dried and exposed to a storage phosphor screen. The screen was scanned on a Typhoon scanner, and the resulting data analyzed with ImageQuant software. Phosphoimager intensities of each reactive thymine band (or doublet) as a function of time after the salt upshift were fit to a first order (single exponential) rate equation in which the long-time plateau value was floated. To obtain normalized θ values [fraction of that base remaining in an open (I₂) condition], this long-time plateau intensity, arising from background/duplex reactivity of the thymine, was subtracted from the observed phosphoimager intensity at each time and divided by the background corrected intensity of that position in a low salt RP_o control. Data fitting was performed by using IgorPro 5 software.

Population Modeling.

See SI Text.