Key Roles of the Downstream Mobile Jaw of Escherichia coli RNA Polymerase in Transcription Initiation

Buffers

RNAP storage buffer contains 50% glycerol (v/v), 0.01 M Tris (pH 7.5 at 4°C), 0.1 M NaCl, 0.1 mM DTT, and 0.1 mM Na₂EDTA. Wash buffer for nitrocellulose filter binding experiments contains 0.01 M Tris (pH 8.0 at room temperature), 0.1 M NaCl, and 0.1 mM Na₂EDTA. Phosphate buffer (PB) for dissociation kinetic experiments in the absence of Mg²⁺ is 0.01 M Na₂HPO₄⁻, 0.16 M NaCl (0.14 – 0.30 M NaCl in salt dependence experiments), 1 mM dithiothreitol (DTT), and 100 µg/ml BSA. Tris buffer (TB), a common transcription buffer used here for kinetics experiments in the presence of Mg²⁺, is 0.04 M Tris (adjusted to pH 8.0 at the experimental temperature with HCl), 0.12 M KCl, 0.01 M MgCl₂, 1 mM DTT, and 100 µg/ml BSA. Low salt Tris buffer (LSTB) is the same as TB, but with 0.06 M KCl and 0.005 M MgCl₂. RNAP storage buffer was added to keep the glycerol concentration constant in series of association experiments (where the RNAP concentration was varied) and in dissociation experiments (6% total RNAP storage buffer in TB and LSTB; 13% in PB). For MnO₄⁻ footprinting experiments, DTT was omitted; for HO• footprinting experiments, the Tris concentration was reduced to 0.01 M.

RNAP Preparation

E. coli ΔJAW RNAP core was prepared as described previously.¹⁵ E. coli WT RNAP core, σ⁷⁰, and holoenzyme were endogenously expressed and purified from E. coli MG1655¹⁶ as described previously.¹ ΔJAW holoenzyme was reconstituted by incubating ΔJAW RNAP core (500 nM) and WT σ⁷⁰ (2 SM) for 90 minutes at 37 °C.

λP_R Promoter DNA Preparation

Full-length (FL) λP_R DNA duplex fragments (~180 bp long; approximately −120 to +60) were PCR-amplified from plasmid pPR59 (λP_R −59 to +34, a gift from Dr. Wilma Ross)¹⁷ using downstream primer 5’-CAGGACCCGGGGCGCGCTTAATTAACACTCTTATACATTATTCC-3’ and upstream primer 5’-GTACGAATTCGATATCCAGCTATGACCATGATTACGCCAGC…3’. Fragments were isolated using the Qiagen QIAquick PCR Purification Kit (Valencia, CA). To end label the template strand downstream, duplex fragments were cleaved with BssHII, and the 5’ phosphates were removed with Antarctic Phosphatase and replaced with [γ-³²P] ATP using T4 polynucleotide kinase. The 3’ ends were filled in using Sequenase with 1 mM dGTP + dCTP, and the upstream end was cut with EcoRV to remove the upstream label. For nontemplate strand downstream fill-in label, duplex fragments were cleaved with EcoRV and BssHII, and filled in with [α-³²P] dCTP (along with 1 mM cold dGTP + dCTP) using Sequenase.

Downstream-truncated (at +12; DT+12) λP_R DNA duplex fragments (~140 bp long; approximately −120 to +12) were obtained by PCR as above, using the same upstream primer as for full-length λP_R (see above) and downstream primer 5’-GAGGACCCGGGTTACTACATGCAACCATTATCACCGCCAGAG-3’. To fill-in label the upstream end of the template strand, the PCR product was cleaved with EcoRI and SmaI, and labeled with [α-³²P] dATP (added with 1 mM cold dATP + dTTP) using Sequenase.

Enzymes used in λP_R promoter DNA preparation and labeling were obtained from New England Biolabs (Ipswitch, MA), except for Sequenase, which was obtained from Affymetrix (Santa Clara, CA). All labeled fragments were purified on a 5% acrylamide gel and isolated using an Elutip-d Purification Minicolumn (Whatman (GE Healthcare, Pittsburgh, PA)).

Dissociation Assays

For dissociation kinetic assays, an initial population of 90 ± 10% open complexes was obtained by incubation of excess WT or ΔJAW RNAP holoenzyme (5–20 nM active) with ³²P-labeled λP_R promoter (~150 pM) at 25 or 37 °C. Irreversible dissociation was initiated by addition of 50–100 µg/ml heparin, a non-displacing competitor for both WT¹⁸ and ΔJAW RNAP (Figure S2). For studies of effects of urea on dissociation, urea was added at the same time as heparin. As a function of time, 100 µl aliquots (~3000 cpm) were removed and applied to nitrocellulose filters on a vacuum manifold.⁴ Filters were dried under heat lamps and counted with a Hewlett-Packard Tri-Carb 2100 TR scintillation counter. In TB, about 20% of open complexes formed by ΔJAW RNAP are found to be non-dissociating. Addition of urea reduces the initial population of ΔJAW open complexes, making it difficult to quantify the dissociation rate constant k_d accurately as a function of urea in TB. Experiments with WT and ΔJAW RNAP to determine effects of urea on k_d were therefore performed at a convenient NaCl concentration (0.16 M) in PB, the buffer used to determine the effects of univalent salt concentration on k_d.

Analysis of Dissociation Data

Filter-retained counts were corrected by subtracting background counts (determined by filtering solutions in the absence of RNAP) to determine cpm_t, the filter-retained counts due to open complexes. For each data set, the dissociation rate constant k_d was determined by fitting cpm_t vs. time to the first order rate equation

where A = cpm_t=∞ (which can differ from background because of a subpopulation of non-dissociating complexes) and B = cpm_t=0 − cpm_t=∞. The fraction of promoter DNA in the form of open complexes at time t (θ_t) is calculated as θ_t = (1/E)(cpm_t=0/cpm_tot) where cpm_tot is the total cpm in the aliquot filtered and E is the efficiency of retention of open complexes on the nitrocellulose filter membrane, typically in the range 0.6 to 0.9 (see below) depending on the details of how the filters are prepared and how the filter bound complexes are treated with wash buffer.³ In Figures 5 and 6, the quantity plotted is Δθ = θ_t − θ_t=∞, where θ_t=∞ = (1/E)(cpm_t=∞/cpm_tot). All data fitting, analysis and normalization were performed using Igor Pro Version 5.03.

Determination of the Equilibrium Constant K₃ for Stabilization of the Initial Open Complex by Assembly of Downstream Mobile Elements

Values of K₃ (defined in Figure 1) were determined from experimental values of dissociation rate constants k_d using the relationship³

The value of the DNA closing rate constant k₋₂ for WT RNAP and FL λP_R promoter DNA at 25 °C was interpolated from results at 10 and 37 °C,³ using the experimentally determined activation energy (9.9 kcal/mol). This DNA closing rate for WT RNAP and FL λP_R is independent of urea concentration; the entire effect of urea on k_d is on K₃.³ A preliminary dissection of the dissociation kinetics for ΔJAW RNAP - FL λP_R complexes (E. Ruff, unpublished) indicate that the difference in k_d between ΔJAW and WT RNAP is largely in K₃, and that the effect of urea in both cases is primarily on K₃. We assume that this situation also applies for DT+12 λP_R. For the dissociation of WT RNAP from FL λP_R the uncertainty in k₋₂ is approximately 20%,³ and that in K₃ is approximately 30%.

Association Assays

To initiate association, excess WT or ΔJAW RNAP (1–50 nM active) was added to ³²P-labeled λP_R promoter (~150 pM) in TB at 37 °C. At selected times after mixing, 100 µl aliquots (~3000 counts) were withdrawn. Heparin (50 µg/ml final) was added for 10 s and each aliquot was filtered and counted as described above.

Analysis of Association Data

Association experiments for ΔJAW RNAP at low concentrations (e.g. in Figure 8 in Results) do not go to completion under the conditions investigated, and kinetic data were therefore analyzed as reversible associations⁴ according to Equation 3:

where β is the decay to equilibrium rate constant, cpm_t is the background-corrected counts retained on the filter at time t, and cpm_eq and cpm_t=0 are the values of cpm_t at equilibrium and extrapolated to t=0, respectively. The decay rate constant β is related to the rate constants k_d and α of dissociation and association, respectively, by:

where θ_eq is the fraction of promoter DNA in the form of competitor-resistant (open) complexes at equilibrium, and

where K₁ is the equilibrium constant for forming the I₁ intermediate closed complex and k₂ is the isomerization (DNA opening) rate constant.₄ The composite association rate constant k_a (Table 1) is given by:⁴

If association is irreversible, then θ_eq = 1, k_d = 0, and β = α. More generally, determination of α from β requires knowledge of either θ_eq or k_d. Determination of θ_eq from the experimental quantities cpm_eq and cpm_tot (the background-corrected total cpm in each filtered aliquot) requires knowledge of the filter efficiency E, the fraction of competitor resistant (open) complexes in solution that are retained on the filter:⁴

Therefore, from Equations 4 and 6,

Analysis of the variation in experimental values of βθ_eq^obs vs. β for a series of experiments at different [RNAP] yields best-fit values of E and k_d, allowing calculation of θ_eq and α from each determination of β using Equation 4. For the sets of experiments analyzed here, efficiencies E of 0.62, 0.77 and 0.81 were obtained for WT-DT+12 complexes, ΔJAW-FL complexes, and ΔJAW-DT+12 complexes, respectively. Values of E and θ_eq were used to calculate θ_t for Figure 8. For WT RNAP and DT+12 promoter DNA, values of α obtained from Equation 4 were analyzed as a hyperbolic function of [RNAP] using Equation 5 to obtain k_a, K₁, and k₂. For ΔJAW RNAP, where α is observed to be proportional to [RNAP] over the accessible concentration range, only k_a is obtained from the analysis. Igor Pro Version 5.03 was used for data fitting and analysis.

KMnO₄ Footprinting

Open complexes (5 nM active RNAP + ~1 nM ³²P-λP_R promoter DNA) were formed in TB (-DTT) or LSTB (-DTT) by incubation for 30 minutes at 37 °C. After incubation, heparin (100 µg/ml) was added to some samples for 10 s before addition of KMnO₄ to eliminate short-lived complexes. KMnO₄ (1.25 mM final concentration) was added for 10 s (determined to be a single-hit dose) and reactions were quenched with 1 M β-mercaptoethanol. Samples were then cleaved with 1 M piperidine, purified, and run on an 8% acrylamide gel together with control (-RNAP) lanes.^{19, 20}

HO• Footprinting

Open complexes were formed in LSTB at 37 °C as described for KMnO₄ footprinting, except that buffers contained DTT and were prepared with a reduced Tris concentration (10 mM) because Tris scavenges HO•.²¹ After a brief (10 s) challenge with 100 µg/ml heparin, [Fe-EDTA]⁻² (1.8 mM Fe²⁺, 4.2 mM EDTA) and sodium ascorbate (1.2 mM) were added, followed by addition of H₂O₂ (0.1 %) to generate HO•. After 10 s, a short enough exposure time to be in the single-hit regime, reactions were quenched (11.7 mM thiourea + 2.0 SM EDTA), and worked up by phenol extraction and ethanol precipitation. Samples were run on an 8% acrylamide gel.^{6, 10}

Gel Imaging and Analysis

Footprinting and transcription gels were imaged using either a Typhoon 9410 Variable Mode Imager or a Typhoon FLA-9000 (both from GE Healthcare), and analyzed using Image Quant TL. HO• line scans were normalized to a short region of the footprint upstream of −60. This upstream normalization provides a better match of downstream as well as upstream baselines for the different line scans on each gel than is achieved by normalization to any group of downstream bands or to total cpm. While small drifts in the downstream baselines are observed with this normalization, (visible in the difference line scans in Figure 4E,F), they are not large enough to affect the semiquantitative conclusions of our analysis. To reduce noise, a moving average of 10 positions (approximately 0.7% of the length of the gel) was applied to each line scan (Excel). Difference line scans for each DNA strand were determined by measuring the distance between the free DNA backbone and RNAP-complexed DNA backbone for each DNA base position +1 to +22. For each strand, distances were normalized to the maximum distance, averaged with the independent (duplicate) experiment, and plotted in a bar graph to more explicitly portray the relative protection of WT and ΔJAW RNAP open complexes.

Prediction of RNAP Disordered Regions

The E. coli RNAP β and β’ subunits²² were analyzed using VL-XT, the most stringent program in PONDR⁸ (Predictions of Naturally Disordered Regions). Only sequences in which at least 15 sequential residues are predicted to be disordered were considered, and predicted disordered sequences known to be in subunit-subunit interfaces were excluded.

KMnO₄ Footprints of λP_R Promoter DNA in ΔJAW Complexes

How does the highly conserved bacterial jaw domain affect the distribution and structure of RNAP-promoter complexes? Ederth et al. observed that ΔJAW RNAP forms unstable complexes at the λP_R promoter at 30 °C at moderate salt concentration (0.01 M MgCl₂, 0.15 M KCl).⁹ To eliminate the possibility that a mixed population of complexes and free promoter DNA exists under these conditions, we sought conditions to increase the stability and lifetime of ΔJAW RNAP complexes at the λP_R promoter (Figure 3A) for footprinting experiments. At 37 °C in lower-[salt] buffer (LSTB: 0.005 M MgCl₂, 0.06 M KCl), ΔJAW RNAP-λP_R open complexes are long-lived (Figure 3B,C; LSTB lanes), suggesting that all bound DNA is in the form of an open complex.

To compare extents of DNA opening, we exposed ΔJAW and WT RNAP complexes to KMnO₄ in either LSTB or TB (0.01 M MgCl₂, 0.12 M KCl). KMnO₄ preferentially oxidizes unstacked thymines, providing a method of detecting and quantifying DNA opening in RNAP-promoter complexes.^{1, 20, 23} We find that ΔJAW-λP_R complexes are open: thymines at −11, −9, −8 and +1 on the template strand (Figure 3B) and −4, −3 and +2 (but not −7 and −10) on the nontemplate strand (Figure 3C)) exhibit the same KMnO₄ reactivity, and therefore are exposed to the same extents in the open regions of WT-λP_R and ΔJAW-λP_R open complexes. In TB, detected thymines on both strands in ΔJAW open complexes are significantly less KMnO₄-reactive than in WT open complexes (Figure 3B,C; TB lanes). Though ΔJAW open complexes are less stable and shorter-lived in TB than in LSTB, this cannot explain the large reductions in KMnO₄ reactivity (see below). For both WT and ΔJAW RNAP, λP_R promoter DNA opening and closing rate constants are only weakly (if at all) dependent on salt concentration³ (E. Ruff, unpublished), so this is not a salt effect on DNA opening. Nor can it be a salt effect on KMnO₄ reactivity, since the KMnO₄ reactivities of these thymines in WT open complexes are the same in TB as in LSTB.

HO• Footprints of λP_R Promoter DNA in ΔJAW Open Complexes

The jaw domain is located near the downstream DNA (Figure 2A), and can interact with DNA nonspecifically.^{13, 24, 25} Furthermore, a large β’ region that includes the jaw crosslinks to the DNA backbone.²⁶ However, direct evidence demonstrating jaw contacts to DNA in any RNAP-DNA complex (e.g. closed, open, initiation, or elongation states) is lacking. Here, we ask whether the jaw deletion alters protection of the DNA backbone in open promoter complexes. Figure 4 compares normalized line scans from quantitative single-hit HO• footprints of both template (panels A,C) and nontemplate (panels B,D) strands of FL λP_R in stable open complexes formed by either WT or ΔJAW RNAP in LSTB at 37 °C. (Sequencing gels for the experiments in panels A and B are shown in Figure S1.) Bar graphs of average differences in HO• protection between open complexes and free promoter DNA at each downstream position from +1 to +22 are shown in Figure 4E,F.

Two very significant differences are observed in HO• footprints of ΔJAW and WT open complexes. First, the downstream boundaries of HO• protection of both strands in the ΔJAW RNAP open complex are shorter than for WT RNAP. The boundary on the template strand (Figure 4E) is approximately 1 helical turn shorter for the ΔJAW open complex than for that formed by WT RNAP, ending near +10 instead of +20. For the nontemplate strand (Figure 4F), the downstream boundary of the ΔJAW RNAP open complex is near +15, about 0.5 helical turn shorter than that observed for WT RNAP (+20). (The difficulty in quantitatively normalizing the entire length of these >100 nucleotide line scans is revealed in the drift in the region downstream of the boundary of the WT footprint at +20, observed in the difference plots in Figure 4E,F. This drift is not large enough to affect the semiquantitative results presented above, though it does increase the uncertainty in assignment of the downstream boundary of the ΔJAW footprints (sequence position +10 ± 2 bases on the template strand and position +15 ± 3 bases on the nontemplate strand).) The reduction in HO• protection of the ΔJAW RNAP open complex provides direct evidence that the jaw, and/or neighboring regions of RNAP perturbed by deleting the jaw, likely contacts one side of the duplex DNA from +10 to +20 in WT RP_o. Direct contact between the jaw and the distal downstream duplex (+10 to +20) was proposed previously.^{9, 24, 26} However, structural studies of bacterial RNAP suggested that the interaction between the jaw and the DNA downstream duplex is proximal to the transcription start site, forming an interface only to about +12.^{22, 25} In this case, the presence of the jaw might act indirectly by influencing the concurrent folding/assembly of one or more additional DMEs onto the distal downstream duplex. At this time, we are not able to distinguish between these two possibilities.

A second very striking difference between HO• footprints of ΔJAW and WT open complexes, shown for the downstream region of the promoter DNA analyzed in Figure 4E,F (but also clearly observable in the upstream promoter regions in Figure 4A–D) is that all regions of both DNA strands in the ΔJAW open complex are significantly less protected from HO• than is the case for WT RP_o. This effect extends at least from −45 to the downstream boundary of the footprint, and is especially pronounced in the open region of the promoter (initiation bubble; −11 to +2) in the active site cleft. This global reduction in protection cannot be attributed simply to the shorter lifetime of the ΔJAW open complex, which is more than one hour in LSTB and therefore much longer than the time scale of the footprinting experiments (10 s). In TB, the ΔJAW open complex exhibits no significant HO• protection of either template or nontemplate strand (data not shown), even though the lifetime of the ΔJAW complex in TB (30 min) is much greater than the time required to footprint (10 s; see below), and KMnO₄ footprints (Figure 2) reveal that this complex is an open complex.

A plausible explanation of the reduced protection observed for the ΔJAW RNAP open complex in both buffers is that more water remains in the ΔJAW RNAP-promoter interface than in the WT RNAP-promoter interface, and that HO• gains access to the DNA in the interface by facilitated diffusion on a network of residual water of hydration. Alternatively, or in addition, the ΔJAW RNAP-promoter interface may be more dynamic than that of the WT enzyme, exhibiting normal modes of vibration that result in local separations of RNAP and promoter DNA that allow HO• access to all portions of the interface in a 10 s footprinting experiment without any global dissociation.

Large Effects of Deletion of the Jaw or of Downstream Promoter DNA on the Rate of Dissociation of Open Complexes

To investigate when, where, and how strongly the jaw and neighboring DMEs assemble on the downstream duplex in the process of forming the stable open complex at λP_R, we compared the kinetics of dissociation of open complexes formed by WT and ΔJAW RNAP on FL and truncated (at +12) λP_R promoter DNA (DT+12). Dissociation time courses in PB (0.16 M NaCl) at 25 °C are shown in Figure 5A. Dissociation rate constants k_d calculated from these data (see Table 1) are compared with one another, and with the corresponding quantities determined at 37 °C in TB (0.12 M KCl, 0.01 Mg²⁺), in Figure 5B. Values of k_d in TB, obtained by analysis of reversible association kinetic experiments (shown in Figure 8 below) are listed in Table 2. In both PB (25 °C) and TB (at 37 °C), WT-FL open complexes are long-lived. Dissociation rate constants k_d for WT-FL open complexes are small and of comparable magnitude (k_d = 2.7 × 10⁻⁵ s⁻¹ in PB at 25 °C; k_d = 2.2 × 10⁻⁵ s⁻¹ in TB at 37 °C¹⁸), permitting convenient study of the destabilizing effects of urea on k_d of the less stable variant complexes.

For both sets of solution conditions, effects of deletion of the jaw or of the downstream DNA on k_d are large (~20–60 fold increase in k_d from that observed for dissociation of WT RNAP from FL λP_R). Details differ somewhat in the two conditions investigated. In TB, each single variant increases k_d approximately 25 fold; in PB, deleting the jaw increases k_d even more dramatically (67 fold), while the effect of deleting the downstream duplex is similar to that observed in TB. The additional effect of the second deletion (comparing k_d for ΔJAW-DT+12 with k_d for ΔJAW-FL or WT-DT+12) is small in PB at 25 °C, but larger in TB at 37 °C. In PB at 25 °C, deletion of the downstream DNA in ΔJAW open complexes does not affect the dissociation rate, and deletion of the jaw in open complexes formed with DT+12 DNA increases the dissociation rate by only a small amount (~ 2 fold) compared to WT RNAP. In TB at 37 °C, however, deletion of the downstream DNA in ΔJAW open complexes increases the dissociation rate by approximately 8 fold. These differences between the variants in PB and TB probably result from specific effects of Mg²⁺; previous dissociation studies with WT RNAP and FL λP_R as a function of Mg²⁺ concentration provided evidence for binding of 2–3 Mg²⁺ ions in the conversion of I₂ to RP_o.²⁷

Because k_d = k₋₂/(1+K₃), where k₋₂ is the DNA closing rate constant (the bottleneck step in dissociation) and K₃ is the composite equilibrium constant for forming the stable open complex (RP_o, I₃, or an equilibrium mixture of the two) from I₂ (Figure 1), these large effects on k_d of deleting either the jaw or the downstream DNA could result from a combination of increases in k₋₂ and/or reductions in K₃. For WT RNAP, the large effect of urea on k_d arises from K₃, while k₋₂ is independent of urea concentration.³ Because deletion of either the jaw or the downstream duplex significantly reduces the urea dependence of k_d (as shown in Figure 6D below), we infer that the primary effect of these deletions must be on K₃. Consistent with this deduction, preliminary measurements of k₋₂ for dissociation of ΔJAW RNAP from FL λP_R promoter DNA (E. Ruff, unpublished) revel that it is similar to (indeed somewhat smaller than) k₋₂ for WT RNAP.

Values of the equilibrium constant K₃ and the standard free energy change ΔG₃⁰ for conversion of I₂ to the stable open complex obtained from these data using Equation 2 are listed in Tables 1 and 2 for the two conditions investigated. For WT RNAP and FL λP_R, the full stabilization of the open complex is −7 to −8 kcal/mol (K₃ of order 10⁵), as determined previously.³ A similar conclusion was obtained from dissociation kinetic studies with DT heteroduplexes.¹² Tables 1 and 2 show that deletion of either the far downstream duplex (at +12) or of the jaw eliminates approximately one-third of the stabilization provided by the DME assembly. In PB, deletion of both the downstream duplex (at +12) and the jaw has no additional effect, as expected if the main effect of deleting the jaw is to eliminate interactions of DMEs with region +12 to +20 of the downstream duplex. Effects observed in both buffers indicate that removal of the jaw disrupts a salt- and Mg²⁺-dependent network of contacts of DMEs with each other and with the distal downstream duplex, greatly reducing K₃ and the lifetime of the open complex and greatly increasing its HO• reactivity.

Effects of Urea on Dissociation Rate Constants k_d for WT vs. Variant Open Complexes

The effect of urea on dissociation rate constants for WT RNAP-FL λP_R open complexes in PB is as large as that observed for unfolding a small globular protein (100–150 residues),²⁸ which we interpreted in terms of the folding and assembly of DMEs on the downstream duplex DNA (+3 to +20) in the process of converting I₂ to RP_o.^{3, 7} If deletion of the jaw and/or the downstream duplex reduces the folding and assembly of DME on the downstream DNA in these late steps, then dissociation of these variant open complexes should exhibit a smaller dependence on urea concentration than for the WT-FL RP_o. Panels A–C of Figure 6 compare the kinetics of dissociation of WT-FL and variant open complexes in 0.94 M urea with that observed in the absence of urea (from Fig 5A). These panels illustrate the general observation that the effect of urea on the rate of dissociation of WT-FL RP_o is much larger than the effect of urea on the rates of dissociation of the ΔJAW RNAP and DT+12 downstream deletion variants. Dissociation rate constants k_d obtained from these data and analogous experiments at intermediate urea concentrations are plotted on a log scale (lnk_d) vs. urea concentration for these four situations in Figure 6D. Linear fits represent the data well in all cases, with intercepts (compared with WT RNAP-FL λP_R in Table 1) that agree within uncertainty with the experimentally determined dissociation rate constants in the absence of urea (Figure 5A).

Slopes obtained from linear fits (dlnk_d/ d[urea]) are listed in Table 1. For dissociation of WT-FL RP_o in PB, dlnk_d/ d[urea] = 3.1 M⁻¹, the same within uncertainty as that reported previously in TB at 17 °C (3.0 M⁻¹).⁷ For the three single and double variant open complexes formed with ΔJAW RNAP and/or DT+12 λP_R, dlnk_d/ d[urea] is only 2/3 as large (dlnk_d/ d[urea] = 1.9 to 2.1 M⁻¹). From Equation 2, these urea derivatives of k_d are equal in magnitude and opposite in sign to the corresponding urea derivatives of K₃, because DNA closing rate constants k₋₂ are independent of urea concentration for WT³ and ΔJAW (E, Ruff, unpublished) RNAP. Deletion of the jaw, of the downstream duplex at +12, or of both, eliminates 1/3 of this urea effect.

Dissection of urea effects on the dissociation rate constant k_d for WT RNAP-FL λP_R open complexes reveals that 2/3 of it is in conversion of I₂ to the (I₂ – I₃)^‡ transition state, with the other 1/3 in the conversion of this transition state to I₃ and then to RP_o.³ Both the urea results and the reductions in open complex lifetime observed for the jaw deletion and/or downstream DNA truncation indicate that by these criteria the ΔJAW-FL λP_R final open complex is more similar to (I₂ – I₃)^‡ or I₃ than to RP_o. Comparison of the effects of urea on dissociation of open complexes formed by WT and ΔJAW RNAP at FL and DT+12 λP_R with these kinetic-mechanistic results argues strongly that the jaw assembles with other DMEs on the downstream duplex promoter DNA late in the mechanism of RP_o formation, most likely in the conversion of I₃ to RP_o, and that the final open complex formed by ΔJAW RNAP is a plausible model of the intermediate I₃ formed by WT RNAP at FL λP_R. Relative dissociation rate constants (Tables 1 and 2) are consistent with this conclusion, as are the different effects of salt concentration on dissociation rate constants of WT and ΔJAW open complexes, presented next.

Effects of Salt Concentration on Dissociation Kinetics of the ΔJAW RNAP - FL λP_R Open Complex are Consistent with an I₃-like Open Complex

Figure 7 compares the effects of NaCl concentration on dissociation rate constants for WT-FL open complexes at 25 °C² and ΔJAW-FL open complexes at 25 °C in PB. Paralleling the reduced effect of urea on k_d for ΔJAW open complex dissociation (33% reduction compared to WT RNAP), dissociation of the ΔJAW open complex in PB is much less salt concentration dependent (50% reduction in dlnk_d/dln[Na⁺]; see Table 3) than dissociation of RP_o formed by WT RNAP at FL λP_R. While the previous kinetic data for WT RNAP did not allow the salt effect on k_d to be uniquely dissected into contributions from the individual steps of dissociation, as was done for the urea effect,³ the greatly reduced salt effect on dissociation of the ΔJAW open complex is completely consistent with the proposal that this complex is I₃-like, with a salt dependence which is halfway between that of RP_o (dlnk_d/dln[Na⁺] = 9) and that of the initial open complex I₂ (dlnk_d/dln[Na⁺] = 0).

Early Steps of ΔJAW – λP_R Open Complex Formation

We previously proposed that DMEs have large effects on early as well as late steps of open complex formation.^{5, 6} Here, we use association kinetic studies to test our hypothesis that the jaw participates in the steps leading up to and/or including opening (R + P ⇔ I₁→I₂; Mechanism 1). For a series of concentrations of ΔJAW RNAP, all in excess over promoter DNA, time courses of open complex formation at FL and DT+12 λP_R at 37 °C in TB are plotted in Figure 8A,C. Analogous results for association of WT RNAP at DT+12 λP_R are plotted in Figure 8B. In all cases the kinetics are single-exponential, demonstrating that the initial binding step(s) (R + P ⇔ I₁) are in rapid equilibrium on the time scale of the subsequent slow step (I₁→I₂) involving DNA opening.²⁹

At all but the highest RNAP concentrations investigated, the kinetics of association of ΔJAW RNAP with both FL and DT+12 (Figure 8A,C) are reversible, forming equilibrium mixtures of open complexes and unbound promoter DNA. By contrast, time courses of association of WT RNAP with FL λP_R⁴ and with DT+12 λP_R (Figure 8B) are irreversible. Values of RNAP-concentration-dependent composite association rate constants α (Equation 5) for Δ JAW RNAP at FL and DT+12 λP_R, obtained as described in Experimental Procedures from the decay-to-equilibrium data in Figure 8A,C, are plotted vs. concentration of ΔJAW RNAP in Figure 8D. Also plotted in Figure 8D are values of α for the association of WT RNAP with DT+12 λP_R (Figure 8B), and fitted curves summarizing the RNAP-concentration-dependences of α previously reported for association of WT RNAP with FL λP_R⁴ and with upstream-truncated (UT-47) λP_R.¹⁹

From Figure 8D, rates of open complex formation from free reactants at all but the lowest RNAP concentrations are slower for all variants than for the WT-FL case. In particular, values of the rate constant α for the ΔJAW variant and either FL or DT+12 λP_R are as small as those observed previously for association of WT RNAP with UT-47 λP_R.¹⁹ Removing either the jaw or the upstream DNA reduces the rate of open complex formation to a similarly large extent. This finding supports our previous proposal that the upstream clamp, jaw, and other DMEs function as a relay system, in which the upstream clamp interacts with the far upstream DNA in the wrapped I₁ intermediate and this interaction shifts the positions of the jaw and other DMEs to open a gate in the downstream cleft and allow entry of the downstream duplex prior to opening.^{5, 6} Removing only the downstream DNA (i.e. open complex formation by WT RNAP with DT+12 λP_R) reduces α to a much smaller though still significant extent at higher RNAP concentrations.

Initial slopes of plots of α vs. RNAP concentration (Figure 8D) are overall second order rate constant for open complex formation k_a (Equation 6). Values of k_a for WT RNAP and either FL or DT+12 λP_R are similar, and are approximately 10 fold larger than those for ΔJAW RNAP and either FL or DT+12 λP_R, or for WT RNAP and UT-47 λP_R. This large effect on k_a of deleting the jaw cannot be explained entirely by an interaction of RNAP with the downstream duplex in I₁ or the subsequent transition state, since the same behavior is observed with both FL and DT+12 λP_R. This is consistent with the deduction from the k_d data that the jaw deletion has large effects on the interaction of DMEs with the downstream duplex late in RP_o formation, in the conversion of I₃ to RP_o.

From Equation 6, k_a is the product of the equilibrium constant K₁ for formation of the I₁ closed complex and the rate constant k₂ for the isomerization step, including DNA opening, that converts I₁ to I₂. For situations where K₁ and the range of RNAP concentrations investigated are large enough, individual values of K₁ and k₂ are obtained from analysis of the dependence of α on [RNAP] using Equation 5. For open complex formation by WT RNAP, truncation of the downstream duplex (DT+12), like upstream truncation (UT-47), increases K₁ and reduces k₂. For DT+12, these effects compensate, so the overall k_a is unaffected by downstream truncation; for UT-47, Davis et al. found that the reduction in k₂ was the dominant effect, greatly reducing k_a even though K₁ increased.¹⁹ For open complex formation by ΔJAW RNAP with either FL or DT+12 λP_R, the data do not permit dissection of k_a into K₁ and k₂.

The finding that association rate constants k_a are the same for open complex formation by ΔJAW RNAP and FL or DT+12 λP_R promoter (Table 2) appears to contradict what is observed in Figure 8, where decay to equilibrium rates at a specified [RNAP] are faster for ΔJAW-DT+12 complexes (Figure 8C) than for ΔJAW-FL complexes (Figure 8A). This is not a contradiction; the faster decay to equilibrium is entirely because k_d for dissociation of ΔJAW-DT+12 open complexes is larger than for ΔJAW-FL open complexes in TB (Table 2).

The Role of the Jaw in Early (Pre-Opening) Steps

The profound effects of deleting the jaw on the early steps of open complex formation, observed for both FL and DT+12 promoter DNA, indicate that the jaw and/or neighboring regions affected by its deletion are a key part of the proposed relay system^{5, 6} by which the upstream promoter sequence, transcription factors, ligands and solution variables work together to regulate the rate of forming the initial open complex (I₂), and thereby regulate the rate of transcription initiation. The limited evidence available to date suggests that much of this regulation occurs in the steps that put the downstream DNA duplex in the cleft. The subsequent opening step is relatively independent of solute concentration^{2, 7} suggesting that, like catalytic steps of enzymes, it occurs within a somewhat sealed cleft that is not readily accessible to solutes and regulatory ligands.

We have proposed that these early-acting upstream regulatory factors have a common basis of action: together they set or modify the trajectory of far-upstream DNA in early, relatively unstable closed intermediate complexes (designated RP_c) in which the promoter DNA lies outside the cleft.⁵ The trajectory of far-upstream DNA determines its ability to interact with the upstream clamp, which in turn affects the positioning of the jaw and other DMEs.⁶ More specifically, we proposed that when −35 and UP elements are recognized by σ⁷⁰ region 4 and the αCTDs (respectively), far-upstream DNA is bent and wrapped around the back of RNAP, in close proximity to the downstream end of the active site cleft and the upstream clamp. These interactions trigger release of DMEs from their blocking positions, opening the downstream gate to allow entry of the downstream duplex.

Interestingly, yeast RNAP II may also use a molecular relay to mediate loading of the duplex DNA into the active site cleft, which is obstructed in the free enzyme in all multi-subunit RNAPs. The Pol II subunit Rpb4/7 (absent in bacterial RNAP) binds to the RNAP core, facilitating clamp closure.³² Recent studies postulate that subsequent binding of the multi-subunit mediator complex attracts Rpb4/7, resulting in formation of the open clamp state and thus allowing for the entry of duplex DNA into the RNAP active site cleft.³³

The Role of the Jaw in Late (Post-Opening) Steps

The observed dissociation rate constant k_d of stable open complexes (mainly RP_o) to free promoter and RNAP provides information about the steps that convert RP_o back to I₂ and the rate of DNA closing (I₂ → I₁) (Figure 1). The DNA closing rate (k₋₂) is not a function of salt or urea concentration.³ Therefore, solute and salt effects on k_d arise from their effect on the equilibrium constant for conversion of I₂ to RP_o (K₃). For the conditions of our experiments, steps by which RP_o converts to I₂ rapidly equilibrate on the time scale of DNA closing (I₂ → I₁).²⁹ We proposed that the conversion of I₂ to RP_o involves assembly of DMEs on the downstream duplex, and movement of the downstream portion of the nontemplate strand in the cleft.^{1, 2, 3} Our current studies test these proposals by deleting the jaw and/or the downstream duplex (DT+12).

Kontur et al. concluded from the urea effect on k_d that approximately 100–150 residues of RNAP are buried by folding/assembly in the steps converting the initial open complex I₂ to the final open complex RP_o (Figure 2).³ Roughly 2/3 of this urea effect occurs in the conversion of I₂ to the subsequent (I₂–I₃)^‡ transition state, and the remaining 40 residues are buried in the conversion of this transition state to RP_o. In the current study, we find that, in contrast to WT enzyme, the ΔJAW variant does not protect the distal downstream DNA from HO• attack. The effect of urea on dissociation of ΔJAW RNAP from either FL or DT+12 λP_R and of WT from DT+12 λP_R indicates that ~40 fewer residues are buried in the open complex formed by these variants. PONDR predicts that 33 residues near the jaw are unstructured in RNAP (Figure 2B). The simplest explanation of these observations is that the jaw folds and assembles with other DMEs on the distal promoter DNA in the final step of open complex formation (Figure 2D), and that the open complex formed by ΔJAW may therefore be a stable mimic of the relatively unstable I₃ intermediate open complex formed by WT RNAP on the path to RP_o at λP_R.

What events (involving burial of approximately 80 residues by folding/assembly) occur in the conversion of I₂ to I₃? We propose that folding and assembly of regions of the β’ clamp and/or the β lobe, DMEs proposed to interact with the proximal downstream DNA duplex (+3 to +9),⁹ occur in this step (Figure 2C). PONDR predicts that 99 residues of the β’ clamp and 57 residues of the β lobe are unstructured in RNAP. Formation of the proximal downstream contacts would logically precede assembly of the jaw and/or other DMEs to the distal downstream duplex, as stability increases in the conversion of I₂ to RP_o. Experiments are in progress to test these predictions for interactions of the DMEs with downstream DNA in I₂ and I₃, and to determine the effect of promoter sequence and interactions in the cleft on the extent of assembly and tightening of the jaw and other DMEs.

Our studies reveal the existence of three types of "open" promoter complexes that differ in interactions between the downstream duplex DNA with DMEs. We hypothesize that the distribution among these complexes differs at different cellular promoters, underlying their regulation by transcription factors such as DksA, which allosterically alters RNAP contacts with DNA downstream from the active site.³⁴ We propose that this mode of regulation may be utilized by eukaryotic RNAPs, which have structural elements that are homologous (β' clamp, β lobe) or functionally analogous (β' jaw) to bacterial DMEs.³⁵ Furthermore, similar to our footprinting results obtained at λP_R, recent single-molecule studies of yeast RNAP II open complexes revealed very different positions of the downstream DNA,³⁶ supporting our idea that promoter DNA/RNAP interactions in this region are highly dynamic and suggesting that they may contribute to regulation of eukaryotic gene expression.