Early intermediates in bacterial RNA polymerase promoter melting visualized by time-resolved cryo-electron microscopy

Abstract

During formation of the transcription-competent open complex (RPo) by bacterial RNA polymerases (RNAP), transient intermediates pile up before overcoming a rate-limiting step. Structural descriptions of these interconversions in real time are unavailable. To address this gap, time-resolved cryo-electron microscopy (cryo-EM) was used to capture four intermediates populated 120 or 500 milliseconds (ms) after mixing Escherichia coli σ⁷⁰-RNAP and the λP_R promoter. Cryo-EM snapshots revealed the upstream edge of the transcription bubble unpairs rapidly, followed by stepwise insertion of two conserved nontemplate strand (nt-strand) bases into RNAP pockets. As nt-strand “read-out” extends, the RNAP clamp closes, expelling an inhibitory σ⁷⁰ domain from the active-site cleft. The template strand is fully unpaired by 120 ms but remains dynamic, indicating yet unknown conformational changes load it in subsequent steps. Because these events likely describe DNA opening at many bacterial promoters, this study provides needed insights into how DNA sequence regulates steps of RPo formation.

Initiation of transcription is a major control point for regulating gene expression in all cells. In bacteria, a single catalytic core RNAP (E, subunit composition α₂ββ’ω) performs all transcription but must combine with a σ factor, such as σ⁷⁰ in Escherichia coli (Eco), to form the holoenzyme (Eσ⁷⁰) for initiation ^1,2. The RNAP structure resembles a crab claw with pincers formed from the large β and β’ subunits ³. The enzyme active site, marked by a bound Mg²⁺-ion, lies deep in the large cleft between the pincers.

In initiation, interactions between Eσ⁷⁰ and specific promoter DNA sequences (most importantly the −35 and −10 elements ⁴, see Extended Data Fig. 1a for the λP_R promoter sequence used here) trigger a series of isomerization steps that separate the DNA strands within the ~13 nucleotide transcription bubble ^5,6. Ultimately, a transcription-competent “open” promoter complex (RPo) forms in which the DNA template-strand (t-strand) is positioned deep (~70 Å) within the RNAP active-site cleft with bases at +1 (start site) and +2 positioned to template incoming NTP substrates ⁷. Correct alignment of the first two NTPs with respect to the active site Mg²⁺-ions catalyzes phosphodiester bond formation ⁸. Thus, RPo formation is required, and is often the rate-limiting step, for the initiation of every RNA chain ⁹.

Promoter opening is a multi-step process during which transient intermediates appear and disappear during progression to the final RPo ^5,6,10,11. Initial recognition of the promoter −35 element by σ⁷⁰ domain 4 (σ⁷⁰₄)¹² outside the RNAP cleft positions the promoter −10 element near the upstream entrance to the cleft. ‘Nucleation’ of the transcription bubble is thought to occur by a ‘flip-and-capture’ mechanism, whereby bases of the −10 element nontemplate-strand [nt-strand; namely A₋₁₁(nt) and T₋₇(nt), Extended Data Fig. 1a] flip-out of the duplex DNA base stack and are captured in cognate pockets of σ^{70 13}. The transcription bubble then propagates in the downstream direction to encompass the transcription start site (+1). Formation of the final RPo involves: 1) Isomerization of an invariant W-dyad of σ⁷⁰ (W433/W434) from an edge-on to a chair-like conformation, which stabilizes the upstream edge of the bubble, and 2) loading of the resulting single-stranded DNA and downstream duplex DNA into the RNAP cleft ^10,14.

In Eco, promoter DNA loading in the RNAP cleft requires the expulsion of the N-terminal domain 1.1 of σ⁷⁰, σ⁷⁰_1.1 ^15,16. Conserved in many bacterial group 1 σ’s ¹⁷, σ⁷⁰_1.1 folds into a negatively-charged four-helix bundle that protects the highly basic RNAP cleft from inappropriately interacting with non-specific nucleic acid. Ejection of σ⁷⁰_1.1 serves as a critical regulatory step in RPo formation, but when and how σ⁷⁰_1.1 is displaced by incoming promoter DNA remains unknown.

Decades of biochemical and biophysical studies have provided mechanistic models for RPo formation, and Eσ⁷⁰ and RPo structures have provided constraints on those models ^5,6,18. Advances in cryo-electron microscopy (cryo-EM) that allow high-resolution information to be extracted from dynamic, heterogeneous samples enabled the first inroads into structural analysis of intermediates in RPo formation ^10,19. Use of standard cryo-EM sample preparation methods in these studies necessitated conditions (e.g., destabilizing factors, promoter mutants) where intermediates were populated at equilibrium. However, formation of RPo is intrinsically a nonequilibrium process. Observing unperturbed RPo intermediates requires high-resolution visualization of RPo formation in real time. To achieve this, we took advantage of two advances in cryo-EM methodology: i) The development of cryo-EM instrumentation that allows mixing and capture of biomolecules on a subsecond time scale (time-resolved-Spotiton, or tr-Spotiton; Extended Data Fig. 1b) ²⁰; and ii) Computational approaches that resolve the conformational heterogeneity in single particle images (3D-variability analysis, 3DVA) ²¹. The results of detailed mechanistic studies of RPo formation at the λP_R promoter ^5,6 allowed the choice of times and conditions where early intermediates are predicted to pile up before the rate-limiting step (Fig. 1a). Together these advances allowed us to capture high-resolution structures of RPo intermediates as they formed in real time.

Fig. 1 |. Promoter melting Intermediates on the λP_R Promoter.

a. (top) Minimal kinetic mechanism for the formation of RPo on λP_R, using the nomenclature of Record & colleagues ^5,6, with two kinetically significant intermediates, I1 and I2. In this scheme RNAP (R) binds the λP_R promoter (P) and forms I1, an ensemble of rapidly equilibrating states ⁵. I1 converts to I2 in a rate-limiting step, which then converts rapidly to RPo. The intermediate RPc is only observed < 7°C.

(bottom) Simulation of the time-course of the reaction under the conditions of the tr-Spotiton experiments (RT, [Eλ⁷⁰] = 15 μM, [λP_R-DNA] = 30 μM). See Methods for the kinetic parameters used to generate the simulation ²⁴. For both 120 and 500 ms mixing times, only I1 (with no RPo) was expected.

b. RNAP clamp conformational changes for intermediates determined in this work. The RPo structure (7MKD) ⁷ was used as a reference to superimpose the intermediate structures via α-carbon atoms of the RNAP structural core, revealing a common RNAP structure (shown as a molecular surface) but with clamp conformational changes characterized as rigid body rotations about a rotation axis perpendicular to the page (denoted by the black dot). The clamp modules for RPc, I1b, and I1d are shown as backbone cartoons with cylindrical helices (λ⁷⁰_NCR is omitted for clarity). The angles of clamp opening for all the intermediates are shown relative to RPo (0°).

c. Structural properties used to order the complexes in the RPo formation pathway.

(top panel) Plotted in black (left scale) is the clamp opening angle [relative to λP_R-RPo (7MKD) ⁷ defined as 0°]. Plotted in orange (right scale) is the DNA-Eσ⁷⁰ interface area (Ų) ⁷⁴.

(bottom panel) Plotted in black (left scale) is the root-mean-square deviation of α-carbon positions (Å) for each complex superimposed with RPo. Plotted in orange (right scale) is the most downstream Eσ⁷⁰-DNA contact observed in each complex.

Also see Extended Data Fig. 1 and Supplementary Video 1.

Results

Direct visualization of DNA melting intermediates by tr-Spotiton.

In the tr-Spotiton robot design, piezo dispensing tips direct two separate streams of ~50 pL droplets onto a nanowire (‘self-wicking’) cryo-EM grid ²² as it traverses towards vitrification, resulting in a stripe of sample across each grid (Extended Data Fig. 1b) ²⁰. Complete mixing of the two samples occurs within ~10 ms of colliding with the grid surface ²³. Reaction times before vitrification can be varied by changing the velocity of the grid (Supplementary Video 1).

To trap RPo formation intermediates, Eco Eσ⁷⁰ (~30 μM) and a λP_R promoter DNA fragment (60 μM; Extended Data Fig. 1a) were deposited onto a self-wicking cryo-EM grid at room temperature (RT) from separate piezo tips (Extended Data Fig. 1b). On-grid mixing occurred at RT under buffer conditions where the kinetics of RPo formation have been well-characterized (Fig. 1a) ^5,24,25 except 8 mM CHAPSO was present (in both samples) to eliminate particle orientation bias at the liquid-air interface ²⁶. Under these conditions, the rate of RPo formation is expected to be largely determined by the rate-limiting step (I1 -> I2, k₂ ~0.04 s⁻¹; Fig. 1a) ²⁵. In vitro mechanistic studies predict an ensemble of early intermediates in rapid equilibrium with each other (I1 ensemble) between about 1 ms to 1 s after mixing (Fig. 1a) ^5,25,27, while relaxation to RPo takes tens of seconds (the conversion from I2 to RPo is extremely fast so I2 is not populated under these conditions; Fig. 1a).

Multiple 120 ms mixing experiments and subsequent cryo-EM data collections were conducted. Of these, three datasets (datasets 1_120ms, 2_120ms, and 3_120ms) were of high quality, yielding consensus structures with nominal resolutions from 3.3 to 3.4 Å (Table 1; Extended Data Fig. 2). A fourth dataset with a longer mixing time was also collected (dataset 4_500ms) and processed, yielding a consensus structure with a nominal resolution of 3.0 Å (Table 1, Extended Data Fig. 3).

Table 1.

Cryo-EM data collection, refinement, and validation statistics.

	1120ms	2120ms	3120ms	4500ms	5°C
Data collection and processing a
Camera	Gatan K3	Gatan K3	Gatan K3	Gatan K3	Gatan K2
Dose rate (e−/Å2/s)	13.81	22.45	33.68	41.86	6.76
Electron exposure (e–/Å2)	34.52	56.13	60.63	62.79	67.73
Defocus range (μm)	−1.0 to −2.3	0 to −2.5	−1.2 to −2.5	−0.8 to −2.5	−1.0 to −2.5
Pixel size (Å)	0.844	0.844	0.844	0.844	1.10
Energy filter slit width (eV)	20	20	20	20	20
CS (mm)	2.7	2.7	2.7	2.7	1x10−3
Symmetry imposed	C1	C1	C1	C1	C1
images collected	2,703	2,614	21,630	14,506	2,764
Initial particle images (no.)	593,407	579,917	5,731,694	2,787,559	581,275
Final particle images (no.)	119,486	122,054	372,303	255,113	235,048
Consensus map res. (Å)b	3.4	3.4	3.3	3.0	3.3
FSC threshold 0.143
	I1a	I1b	I1c	I1d	RPc5°c
	EMD-41433	EMD-41439	EMD-41448	EMD-41437	EMD-41456
	PDB 8TO1	PDB 8TO8	PDB 8TOE	PDB 8TO6	PDB 8TOM
Refinement
Initial models used (PDB code)	7MKD	7MKD	7MKD	7MKD	7MKD/6PSQ
Model resolution (Å)	2.8	2.9	2.9	2.9	3.1
FSC threshold 0.5
Map sharpening B factor (Å2)
Model composition
Non-hydrogen atoms	62,457	61,662	61,897	62,366	63,007
Protein residues	3,822	3,753	3,785	3,738	3,827
Nucleic acid residues	70	74	73	98	80
Ligands	Mg2+: 1 Zn2+: 2 CHAPSO: 4	Mg2+: 1 Zn2+: 2 CHAPSO: 4	Mg2+: 1 Zn2+: 2 CHAPSO: 3	Mg2+: 1 Zn2+: 2 CHAPSO: 3	Mg2+: 1 Zn2+: 2 CHAPSO: 4
B factors (Å2)
Protein	96.51	77.22	58.54	60.27	129.11
Nucleic acids	188.71	185.22	158.09	145.46	231.17
Ligand	84.01	70.81	71.23	64.42	117.25
R.m.s. deviations
Bond lengths (Å)	0.004	0.003	0.004	0.006	0.004
Bond angles (°)	0.591	0.516	0.562	0.628	0.643
Validation
MolProbity score	2.18	1.79	2.07	2.22	2.45
Clashscore	6.10	4.73	5.48	5.85	8.61
Poor rotamers (%)	3.93	1.79	3.39	4.33	4.57
Ramachandran plot
Favored (%)	94.28	94.92	94.54	93.68	92.0
Allowed (%)	5.62	5.08‘	5.4‘	6.26	7.79
Disallowed (%)	0.11	0.0	0.05	0.05	0.21

^aAll micrographs collected using a Titan Krios (ThermoFisher Scientific) microscope with an accelerating voltage of 300 kV.

^bNominal consensus map resolutions are from single particle reconstructions prior to Bayesian polishing in RELION ⁷³.

Examination of the four consensus maps revealed features indicative of structural heterogeneity in the nonconserved insert of σ⁷⁰ (σ⁷⁰_NCR; σ⁷⁰ residues 128-376), the sequence insertion in the trigger-loop (SI3; β’ 942-1131), the DNA upstream of the −35 element, and the α-C-terminal domains (αCTDs). To isolate conformational changes involving the promoter −10 element and the RNAP active site cleft, we performed 3DVA within a mask encompassing the entrance to the cleft (σ⁷⁰₂ and σ⁷⁰₃ but excluding the σ⁷⁰_NCR), the pincers (βprotrusion, βlobe, and clamp), and the downstream end of the channel (including the β’jaw) (Extended Data Fig. 4).

Clamp opening/closing, a well-studied functional characteristic of cellular RNAPs ^28–35 and a target of antibiotics ^36–41, was a major mode of motion observed in all the 3DVAs (Fig. 1b). Analysis of these 3DVA trajectories revealed a correlation between the clamp motions and progression along the RPo formation pathway (Fig. 1c). We therefore focused on this component in the 3DVA analyses (Supplementary Video 2).

Each dataset (1_120ms, 2_120ms, 3_120ms, and 4_500ms) was analyzed independently, as were the combined particles from the 120 ms experiments (123_120ms). Using masked 3DVA cluster analysis of the clamp open/close mode of the 120 ms datasets, we repeatedly found the same three distinct intermediates and sometimes found a fourth, poorly populated intermediate. With the 500 ms dataset, the same four intermediates were observed. We designated the intermediates I1a, I1b, I1c, and I1d to indicate that they are part of the ‘I1 ensemble’ that precedes the conversion to RPo (Fig. 1a) ⁵.

Due to the internal consistency of the independent analyses, we combined all the particles into one large dataset (123_120ms4_500ms) to increase the signal of any low-abundance states. The combined dataset was analyzed for structural heterogeneity using the same masked 3DVA procedure (Extended Data Fig. 4). This yielded the same four intermediates (I1a, I1b, I1c, I1d) observed in the analyses of the individual datasets (1_120ms, 2_120ms, 3_120ms, and 4_500ms). The fractions of particles distributed into each intermediate from each of the three independent 120 ms datasets were nearly the same (standard deviations of average particle fractions < 10%; Supplementary Table 1a), indicating that the tr-Spotiton device and the analysis pipelines were reproducible. The population distribution of dataset 4_500ms was also similar but significantly different according to Jensen-Shannon distances (Supplementary Table 1b)⁴², being skewed towards the most advanced intermediate I1d (Supplementary Table 1, Fig. 1c; Extended Data Fig. 5).

The cryo-EM maps derived from the combined dataset were used for model building and refinement (Table 1, Extended Data Figs. 6 and 7). The intermediate structures were ordered along the RPo formation pathway such that the DNA-Eσ⁷⁰ interface area and the downstream boundary of the DNA-Eσ⁷⁰ contacts increased, while the root-mean-square deviation of α-carbon positions of each complex compared to λP_R-RPo (7MKD)⁷ decreased (Fig. 1c). These metrics of progress along the RPo formation pathway correlated with the RNAP clamp position, with I1a having a relatively open clamp [6.9° compared to RPo (0°)] and the clamp closing between 1° to 2° in each subsequent intermediate (Figs. 1b and 1c). A control dataset to examine whether CHAPSO influenced these results was collected using 1H, 1H, 2H, 2H-perfluorooctyl)phosphocholine (fluorinated Fos-Choline-8, or FC8F, an alternative detergent discovered to mitigate particle orientation bias) and a mixing time of 500 ms. The same structural intermediates were observed (Extended Data Fig. 8).

Duplex DNA rapidly unwinds at λP_R.

By 120 ms at RT, all λP_R-Eσ⁷⁰ complexes converted to intermediates in which DNA melting was nucleated (Fig. 2): the W-dyad was edge-on (forming a W-wedge), the −12 bp was open, and A₋₁₁(nt) was extra-helical (Fig. 3). The single-stranded t-strand within the transcription bubble was dynamic (no interpretable cryo-EM density) in all intermediates. We did not observe upstream DNA wrapping on Eσ⁷⁰ in any I1 intermediate (DNA to −85; Extended Data Fig. 1a).

Fig. 2 |. Transcription bubble nucleation.

a. Top view of λP_R-RPo (7MKD) ⁷. Eσ⁷⁰ is shown as a transparent molecular surface. The DNA is shown as atomic spheres, color-coded as shown on the left.

b. The boxed region in (a) is magnified (the β subunit is removed for clarity), showing the region of transcription bubble nucleation of RP_C5°C. Protein is shown as a backbone worm with a transparent molecular surface. The side chains of the σ⁷⁰ W-dyad (Eco σ⁷⁰ W433/W434) are shown (W-wedge conformation). The duplex (closed) DNA is shown in cartoon format.

c. As in (b) but showing I1a. The DNA is shown in stick format. The −12 bp is open due to steric clash with the W-wedge. A₋₁₁(nt) is flipped but not captured. For reference, the path of the DNA in RPo is shown in chartreuse, with the positions of key nucleotides shown in stick format [−12 bp, A₋₁₁(nt), T₋₇(nt)].

d. Magnified view of I1b. T₋₇(nt) is captured in its cognate σ⁷⁰ pocket.

e. Magnified view of I1c.

f. Magnified view of I1d. A₋₁₁(nt) is captured but the W-dyad remains in the W-wedge conformation and the −12 bp remains open.

g. Magnified view of RPo. The σ⁷⁰ W-dyad has isomerized to the chair conformation, allowing repairing of the −12 bp.

Also see Extended Data Figs. 2–11 and Supplementary Video 2.

Fig. 3 |. Capture of T₋₇(nt) before A₋₁₁(nt).

Each panel shows two views of the λP_R-Eσ⁷⁰ complexes. σ⁷⁰ and DNA (color-coded as in Fig. 3) are shown in stick format; carbon atoms are colored orange but the W-dyad is highlighted in yellow. Transparent cryo-EM density (local-resolution filtered ⁷⁵) is superimposed. For reference, the positions of key RPo elements are shown in stick format and colored chartreuse [A₋₁₁(nt), T₋₇(nt), W-dyad in chair conformation].

(left) The σ⁷⁰_2-A₋₁₁(nt) pocket, viewed from upstream.

(right) The σ⁷⁰-T₋₇(nt) pocket.

a. I1a; (left) A₋₁₁(nt) is flipped but not captured [the σ⁷⁰_2-A₋₁₁(nt) pocket is empty] and the W-dyad is in the wedge (edge-on) conformation, (right) The T₋₇(nt) pocket is empty.

b. I1b; (left) A₋₁₁(nt) is flipped but not captured [the σ^702-A₋₁₁(nt) pocket is empty] and the W-dyad is in the wedge (edge-on) conformation, (right) T₋₇(nt) is captured.

c. I1c; (left) A₋₁₁(nt) is flipped but not captured [the σ⁷⁰_2-A₋₁₁(nt) pocket is empty] and the W-dyad is in the wedge conformation, (right) T₋₇(nt) is captured.

^d. I1d; (left) A₋₁₁(nt) is captured but the W-dyad is still in the wedge conformation, (right) T₋₇(nt) is captured.

e. RPo; (left) A₋₁₁(nt) is completely captured and the W-dyad is in the chair conformation, (right) T₋₇(nt) is captured.

Also see Extended Data Fig. 11.

In I1a, I1b, and I1c, cryo-EM density features for σ⁷⁰_1.1 were present in the RNAP cleft (Fig. 4). In I1b and I1c, the single-stranded nt-strand gradually appears (I1b, −7, −5, −4; I1c, −7, −6, −5, −4; Figs. 2d and 2e), following approximately the same path as in RPo (Figs. 2a and 2g). At I1d, additional large scale changes ocurred: the DNA strands were unambiguously unwound from −12 to +2, A₊₁(nt) stacked on βW183 (as in RPo), and double-stranded DNA (from +3 to +10 visualized) occupied the RNAP downstream channel in place of σ⁷⁰_1.1 (Fig. 4e).

Fig. 4 |. RNAP clamp closure partially unfolds and ejects σ⁷⁰_1.1-.

Each panel shows two views of the λP_R-Eσ⁷⁰ complexes (σ⁷⁰_NCR omitted for clarity), with color-coding shown at the top.

(top) View into the RNAP active-site cleft. Eσ⁷⁰ is shown as a backbone cartoon. Cryo-EM density for σ⁷⁰_1.1 (orange) and the DNA are also shown.

(bottom) View focusing on σ⁷⁰_1.1 [viewed from the direction of the thick orange arrow in (a)]. Eσ⁷⁰ is shown as a transparent molecular surface, but σ⁷⁰_1.1 is shown as a backbone cartoon with cylindrical helices. DNA is shown in stick format. Transparent cryo-EM density (local-resolution filtered ⁷⁵) for σ⁷⁰_1.1 (orange) and the DNA is also shown.

a. RPc_5°c.

b. I1a. Closure of the clamp from the previous intermediate is denoted by the thick green arrow.

c. I1b; σ⁷⁰_1.1-H4 becomes largely disordered.

d. I1c; cryo-EM density for σ⁷⁰_1.1 is present but becomes mostly uninterpretable.

e. I1d; σ⁷⁰_1.1 is replaced by duplex DNA (+3 to +10) in the RNAP cleft.

f. RPo.

Also see Extended Data Figure 12.

Intermediates populated at equilibrium at low temperature at λP_R.

The observation of full transcription bubble formation in the I1 ensemble (I1d; Fig. 4e) was unexpected ^5,25,43,44. To confirm the structural characteristics of the I1 ensemble as observed by tr-Spotiton at RT, we examined λP_R-Eσ⁷⁰ complexes trapped at equilibrium at 5°C using standard cryo-EM grid preparation methods. On λP_R at temperatures below 7°C, the high activation energy (E_a ~34 kcal) to reach the transition state between I1 and I2 blocks conversion to complexes past I1, leaving only I1 and any preceding complexes at equilibrium ^25,44.

λP_R-Eσ⁷⁰ complexes were pre-formed on ice and then spotted and blotted on grids held at 5°C. After collection and processing of 2764 micrographs using the same pipeline as the tr-Spotiton datasets (Table 1; Extended Data Fig. 9), we identified an earlier intermediate not seen in the RT tr-spotiton experiments (RPc_5°C, nominal resolution 3.1 Å, Table 1; Extended Data Figs. 9 and 10a–d), and two complexes that share similarities with I1c and I1d, with population distribution RPc_5°C, 52%; I1c_5°C, 21%; I1d_5°C, 27% (Extended Data Figs. 9 and 10e–m). These data are consistent with our finding that the I1 ensemble contains a significant fraction of DNA-Eσ⁷⁰ complexes with a fully-melted transcription bubble.

Transcription bubble nucleation: −12 bp opening and capture of T₋₇(nt) before A₋₁₁(nt).

In the earliest intermediates detected at λP_R at RT (I1a, I1b, I1c), the transcription bubble was nucleated; A₋₁₁(nt) was flipped but not yet captured in the σ⁷⁰ pocket (Figs. 2c–e, 3a–c). Capture of A₋₁₁(nt) did not occur until I1d (Figs. 2f and 3d). As early as I1b we observed clear cryo-EM density for T₋₇(nt) capture (Fig. 3b). Thus, T₋₇(nt) capture preceded A₋₁₁(nt) capture ⁴⁵, illuminating why T₋₇(nt) is overrepresented in fast melting −10 element sequences ⁴⁶.

Strikingly, through all the intermediates observed here, from initial transcription bubble nucleation (I1a) to full bubble formation (in I1d), the σ⁷⁰ W-dyad remained in its edge-on (wedge) conformation (Figs. 2b–f and 3a–d). As a result, in all the nucleated intermediates (I1a-I1d), the −12 bp was disrupted by steric clash with the edge-on W-wedge (Figs. 2c–f; Extended Data Fig. 11). The un-paired −12 and −11 nt-strand bases were essentially exposed to solution and highly dynamic, although the A₋₁₁(nt) base may stack on σ⁷⁰-Y425 in I1a-c (Fig. 3a–c).

Stepwise closure of the RNAP clamp partially unfolds and ejects σ⁷⁰_1.1.

As the clamp closes through the RPo formation pathway, a striking feature emerges: σ⁷⁰_1.1 becomes progressively disordered. In RPc_5°C and I1a, the four α-helices of σ⁷⁰_1.1 occupy the RNAP cleft (H1-H4, Figs. 4a and 4b) as seen in Eσ^{70 16}. In the conversion from I1a to I1b, cryo-EM density for much of H4 and for the 37-residue “linker” that connects H4 to σ⁷⁰_1.2 (Eco σ⁷⁰ residues 75-91) disappears (Fig. 4c). Upon further closure of the clamp in I1c, the cryo-EM density for σ⁷⁰_1.1 further fragments and becomes largely uninterpretable except for two tube-like densities modeled as α-helices (Fig. 4d). The σ⁷⁰_1.1 finally disappears in I1d (Figs. 4e).

To test the importance of σ⁷⁰_1.1 unfolding for RPo formation, we introduced double-cysteine substitutions into σ⁷⁰_1.1 (in the background of a Cys-less σ⁷⁰ derivative)⁴⁷ designed to form interhelical disulfide bonds under oxidizing conditions to interfere with σ⁷⁰_1.1 unfolding (Fig. 5a). Three Cys-pairs were constructed: Q8C-P32C (crosslinking H1 to H3), Y21C-Q54C (crosslinking the H1-H2 linker with the H3-H4 linker), and I35C-S89C (crosslinking the H3 with H4). Each derivative had a higher mobility under oxidizing conditions than reducing conditions when analyzed by denaturing polyacrylamide gel electrophoresis (Extended Data Fig. 12a), indicating that the expected disulfide bonds formed. Comparison of abortive initiation transcription activity under oxidizing vs. reduced conditions revealed that the I35C-S89C (H3-H4) crosslink was severely defective in producing a transcript (Figs. 5b–d), suggesting that RPo cannot form. This is consistent with the order-to-disorder transition of σ⁷⁰_1.1-H4, begining at the I1a -> I1b transition, being essential for progress through the RPo formation pathway.

Fig. 5 |. σ⁷⁰_1.1-H4 unfolding is necessary for RPo formation.

a. (top) Overall view of I1a. Eσ⁷⁰ is shown as a transparent molecular surface but σ⁷⁰_1.1 (orange) is shown as a backbone cartoon.

(bottom) σ⁷⁰_1.1 from boxed region above, colored as a rainbow ramp from the N-terminus (blue) to the C-terminus of H4 (red). Pairs of residues substituted with Cysteine (in the background of a Cysteine-less σ^{70 47}) are shown as Ca spheres, with engineered disulfide bonds illustrated with thick lines.

b. Synthesis of abortive products (ApUp*G, where * denotes α-[P³²]) from λP_R by Eσ⁷⁰’s containing WT or mutant σ⁷⁰’s as indicated. Oxidizing conditions (to form disulfide bonds, left) are compared with reducing conditions (right).

c. Relative intensity of abortive products (normalized with respect to WT σ⁷⁰) for mutant σ⁷⁰’s under oxidizing and reducing conditions. Error bars denote the standard deviation of n=3 measurements.

d. Ratio of abortive product intensity for oxidized/reduced conditions [error bars were calculated by error propagation from the standard deviations shown in (c)]. The I35C-S89C disulfide, crosslinking σ⁷⁰_1.1-H4 to H3, is severely defective under oxidizing conditions.

Also see Extended Data Fig. 12.

Discussion

We used cryo-EM to determine high-resolution structures of early RPo formation intermediates at λP_R (Fig. 6). Early intermediates were trapped at equilibrium at low temperature (RPc_5°c), or at RT in real time (non-equilibrium) using tr-Spotiton (Extended Data Fig. 1b) ²⁰. The structures delineate conformational changes in both Eσ⁷⁰ and promoter DNA on the pathway to forming RPo and reveal unanticipated features. Analysis of the structures of early RPo intermediates allows direct comparison with extensive mechanistic studies of DNA opening on λP_R ⁵ and provides unprecedented insights into the mechanism of transcription bubble nucleation (Fig. 2) and σ⁷⁰_1.1 ejection from the RNAP cleft (Figs. 4 and 5). This study identifies early intermediates of RPo formation at one promoter, λP_R, but the invariant Eσ⁷⁰ architecture and conserved nature of promoter −10 and −35 elements ⁴ suggests that these structures define key steps of DNA opening at most Eσ⁷⁰ promoters ¹⁰.

Fig. 6 |. Schematic overview of initial steps of promoter opening at λP_R.

The region of transcription bubble nucleation is shown for RPc, each I1 intermediate, and RPo. The RNAP active-site Mg²⁺ is shown as a yellow sphere. The σ⁷⁰2 domain, with its A₋₁₁(nt) and T₋₇(nt) pockets, and σ⁷⁰_1.1, are shown in orange. In RPc to I1c, elements of σ⁷⁰_1.1 are in the RNAP active-site cleft between elements of β (cyan) and P’ (pink). Closure of the clamp is denoted by the black arrows. The DNA is shown as a backbone worm (−10 element colored hot pink). Poorly-resolved regions of the DNA or σ⁷⁰_1.1 are illustrated by dashed lines.

A closed complex observed at 5°C (RPc_5°c).

An RPc was first observed by cryo-EM with Eco Eσ⁷⁰ at the rpsTP2 promoter (6PSQ) where stabilization of early intermediates by the transcription factor TraR likely populated RPc ¹⁰. RPc on λP_R was populated in samples prepared at equilibrium at low temperature (RPc_5°c; Figs. 2b and 4a) but not in any of the tr-Spotiton datasets (prepared at RT). At the relatively high final concentrations of Eσ⁷⁰ and DNA on the tr-Spotiton grids (15-30 μM), we expect RPc to fully occupy the promoter DNA and convert to I1 in <10 ms ^5,27.

The rpsTP2-RPc and λP_R-RPc_5°c have very similar structural features; the promoter DNA is completely duplex and outside the RNAP cleft, the RNAP clamp is relatively open (rpsTP2-RPc, 8.5°; μP_R-RPc_5°c, 6.9°, using μP_R-RPo as 0° reference), and there are no Eσ⁷⁰-DNA contacts downstream of −2 nor base-specific contacts in the promoter −10 element ¹⁰. A superposition of the RNAP structures (excluding the βlobe-Si1, which is repositioned due to TraR binding in rpsTP2-RPc) yielded a root-mean-square-deviation (rmsd) of 1.2 Å over 2,084 RNAP α-carbons. An RPc with Mycobacterium tuberculosis RNAP with the transcription factor WhiB7 on the pWhiB7 promoter shares these same structural characteristics ⁴⁸. Based on these observations, the most straightfoward proposal is that this RPc structure is the first, or one of the first, promoter DNA-RNAP complexes at most, if not all, Eσ⁷⁰ promoters.

The RNAP clamp.

Clamp dynamics play an important role in promoter melting for all cellular RNAPs ^{10,19,32,49–51}. Several antibiotics, including clinically important Fidaxomicin ^39–41, bind RNAP in the switch regions (essentially hinges for the clamp motion) and interfere with initiation, highlighting the importance of clamp dynamics ³⁸. In 3DVAs of the heterogeneous λP_R-Eσ⁷⁰ conformations observed in the tr-Spotiton samples, clamp opening and closing was the major mode of motion that correlated with other structural features expected to be associated with progress on the RPo formation pathway (Supplemental Video 2). As the clamp closed from RPc_5°c (6.9° open) to RPo (0°), the DNA-Eσ⁷⁰ contacts progressed downstream while the DNA-Eσ⁷⁰ interface area increased, and the RNAP conformation became more RPo-like (Fig. 1c).

We propose that entry of the DNA into the RNAP cleft and gradual establishment of in-cleft DNA-RNAP interactions (Fig. 2) drive gradual clamp closure (Fig. 1b). The clamp closure in turn drives many of the conformational changes that propel the complex along the RPo formation pathway. As DNA melting propagates downstream after nucleation at −12/−11, stepwise clamp closing appears to position Eσ⁷⁰ structural elements to establish sequence-specific interactions with the −10 hexamer and discriminator ^52–56 promoter elements. Perhaps most strikingly, for λP_R, interactions with nt-strand discriminator bases (−6 to −4) appear to occur along with capture of T₋₇(nt), followed by A₋₁₁(nt) (Figs. 2d, 2e, and 6). Early establishment of discriminator interactions is consistent with the critical role of the discriminator in transcription start site selection ⁵⁶. We note that DNA sequence differences in these promoter elements (which will affect the lifetimes and relative populations of melting intermediates) may alter the order in which they are stabilized by interactions with Eσ⁷⁰. Promoter sequence effects on the distribution of melting intermediates provides opportunities for promoter-specific regulation by factors that don’t bind DNA, such as DksA ⁵⁷

Clamp closure drives coupled unfolding and extrusion of σ⁷⁰_1.1 out of the RNAP cleft.

Clamp closure appears to drive the coupled unfolding and extrusion of σ⁷⁰_1.1 out of the RNAP cleft (Fig. 4) by a mechanism similar to protein unfolding induced by high hydrostatic pressure ⁵⁸. As the clamp closes, the volume available for σ⁷⁰_1.1, located between the closing RNAP pincers, decreases. In response, elements of σ⁷⁰_1.1 unfold and extrude out the cleft, decreasing the volume occupied by σ⁷⁰_1.1 inside the cleft. σ⁷⁰_1.1-H4, known to be semi-stable and located at the mouth of the cleft ¹⁶, and H3-H4 linker were squeezed out of the cleft by the closing clamp. The unfolding and extrusion of σ⁷⁰_1.1-H4 appears to be essential for further progress on the RPo formation pathway (Fig. 5); crosslinking of H4 to H3, preventing unfolding of H4, may inhibit further closure of the clamp which would be required to completely dislodge σ⁷⁰_1.1 and establish the necessary promoter DNA-RNAP contacts.

Clamp closure drives transcription bubble nucleation.

The σ⁷⁰-W-dyad remained edge-on in all the intermediates observed here (RPc_5°c, I1a, I1b, I1c, I1d), in common with the earliest intermediates at rpsTP2 (RPc and T-RPi1) ¹⁰. Therefore, we propose that this W-dyad conformer may be a shared feature of early intermediates at most Eσ⁷⁰ promoters. The edge-on W-dyad comprises a stable structural unit ⁵⁹ that may act as a wedge (the W-wedge) to nucleate the transcription bubble. Clamp closure may squeeze the DNA between the βprotrusion and σ⁷⁰₂, driving the DNA into the σ⁷⁰-W-wedge and sterically disrupting the −12 and −11 bps, nucleating the transcription bubble (Fig. 2).

Opening of the −12 bp due to steric clash with the W-wedge was first observed in an early intermediate at rpsTP2, T-RPi1 ¹⁰. This was a surprising observation since the - 12 position [immediately upstream of the critical A₋₁₁(nt) base] is base-paired in all known RPo structures, including rpsTP2-RPo and λP_R-RPo ^7,10. Thus, at least at these two promoters, nucleation disrupts the −12 bp early in RPo formation. Our finding of −12 bp opening in all I1 ensemble intermediates (Fig. 2) suggests that unpairing and repairing at −12 may be a conserved feature of RPo formation at all σ⁷⁰-family promoters and could explain the conservation of T₋₁₂(nt) within the −10 element ⁴.

The W-dyad remained in the wedge conformation even in I1d, where full promoter opening was clearly achieved (Fig. 2f). This observation contrasts with intermediates at rpsT P2, where the W-wedge was observed only in the earliest nucleated intermediate, T-RPi1; the W-dyad isomerized to the chair conformation in subsequent intermediates even though full promoter melting was not achieved ¹⁰. The rpsTP2 intermediates were trapped and prepared for cryo-EM at equilibrium (i.e. after long incubation times) while the λP_R I1 intermediates observed here were prepared for cryo-EM in real-time at non-equilibrium conditions by tr-Spotiton after 120 or 500 ms mixing times. We suggest that during RPo formation, W-dyad isomerization from the wedge to the chair conformation occurs at a slow rate compared to full bubble opening; W-dyad isomerization may have occurred in the rpsT P2 intermediates (trapped by TraR and, in some cases, promoter mutations) due to the long incubation times and equilibrium conditions ¹⁰. This highlights a major advantage of time-resolved approaches for cryo-EM sample preparation, critical for trapping important conformations of macromolecular complexes in transient, non-equilibrium states.

Full promoter opening precedes the rate-limiting step at λP_R.

Before recent cryo-EM studies ^10,19, descriptions of RPo intermediates were largely obtained by chemical or enzymatic footprinting of complexes trapped at equilibrium under perturbing solution conditions (low temperature, inhibitors) ^60–63. Permanganate ion (MnO₄⁻) oxidation of unstacked thymines was used to detect DNA melting ⁶⁴. Protection of the DNA backbone from cleavage by DNase I or hydroxyl-radicals was used to map RNAP-DNA interactions ^65,66.

Footprinting of Eσ⁷⁰-λP_R complexes equilibrated at 0°C (so comprising the I1 ensemble) did not detect any unstacked nt-strand thymines despite DNA backbone protection extending to ~+20 ⁴⁴. To rationalize these results, the DNA in I1 was proposed to be sharply bent or kinked at −11/−12, directing the duplex DNA into the RNAP cleft ⁴⁴. Subsequent studies found that the rate constants for the appearance of oxidized thymines were similar to those for the conversion of I1 to I2 ²⁷. These and other studies led to the conclusion that I1 was closed (no bubble nucleation), and that the rate-limiting step for RPo formation at λP_R (the I1 -> I2 conversion) corresponded with the formation of the transcription bubble ⁶⁷.

By the time I1d forms, thymines on the nt-strand are stacked (−4, −3) or bound in a pocket (−7). Lacking stabilizing RNAP interactions (and thus lacking a driving force for unstacking), t-strand bases may remain largely stacked in I1. Although the rates of interconversion between the I1 ensemble intermediates are unknown, they are likely relatively fast: in the absence of protein-DNA interactions, lifetimes of transiently open DNA bubbles (2-10 bp) ⁶⁸ or extrahelical bases ⁶⁹ range from 10⁻² to 10⁻⁶ s. Because the MnO₄₋ oxidation reaction rate constant is relatively slow (~15 M⁻¹ s⁻¹) and decreases with decreasing temperature ⁷⁰, we suggest that the footprinting conditions (low temperature, low MnO₄⁻ dose) used for λP_R may have prevented detection of DNA opening in I1.

In support of this hypothesis, fast MnO₄⁻ and hydroxyl-radical footprinting studies at 37°C on the T7 A1 promoter revealed that DNA melting preceded the rate-limiting step and the formation of stable downstream contacts ⁷¹. Mechanistic studies of RPo formation at a consensus promoter ⁷² and at gal P1 ⁴⁵ found that the rates of unquenching (unstacking) 2-amino purine fluorescence were faster than the rates of forming RPo. These results are consistent with our findings from time-resolved cryo-EM on λP_R that full promoter opening precedes the rate-limiting step for RPo formation.

Rate-limiting steps in RPo formation.

If neither σ⁷⁰_1.1 ejection nor DNA opening are rate-limiting at λP_R, what is? Conversion from I1 to I2 (Fig. 1a) requires overcoming a large positive activation energy ²⁵. Once I2 forms, thymines on the t-strand become solvent-accessible; T₊₁(t) is as MnO4⁻-reactive as in RPo ⁶⁷. These results, along with the observation that the σ⁷⁰-finger in the I1 ensemble clashes with the position of the t-strand in RPo (Extended Data Fig. 12b) suggest that the rate-limiting step at λP_R loads the t-strand in the active site by repositioning the σ⁷⁰-finger and unstacking t-strand bases. Other potentially rate-limiting changes occur as I1 converts to I2 and then RPo: an intricate network of interactions form between σ⁷⁰_1.2, the β “gate loop”, and the nt-strand discriminator bases ⁷, and the W-wedge isomerizes to “chair”, allowing the bp at −12 to re-pair (Fig. 3d). Further kinetic-mechanistic and cryo-EM studies will be required to tease apart these conformational changes and how they are driven by promoter DNA sequence.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment.

Protein expression and purification.

Eco core RNAP [α₂ββ’(His)₁₀ω] was purified largely as described previously ³³. A pET-based plasmid overexpressing each subunit of RNAP (full-length α, β, ω) as well as β’-PPX-His₁₀ (PPX; PreScission protease site, LEVLFQGP, Cytiva, Marlborough, MA) were co-transformed with a pACYCDuet-1 plasmid expressing Eco rpoZ into Eco BL21(DE3). The cells were grown in the presence of 100 μg/mL ampicillin, 34 μg/mL chloramphenicol, and 0.5 mM ZnCl₂ to an OD₆₀₀ of 0.6 in a 37°C shaker. Protein expression was induced with 1 mM IPTG (final concentration) for 4 hours at 30°C. Cells were harvested by centrifugation and resuspended in lysis buffer [50 mM Tris-HCl, pH 8.0, 5% glycerol (v/v), 10 mM DTT, 1 mM PMSF, and 1x protease inhibitor cocktail (Sigma Aldrich)]. After French Press lysis at 4°C, the lysate was centrifuged twice (33,000 x g) for 30 minutes each. Polyethyleneimine [PEI, 10% (w/v), pH 8.0, Acros Organics - ThermoFisher Scientific, Waltham, MA] was slowly added to the supernatant to a final concentration of ~0.6% (w/v) PEI with continuous stirring. The mixture was stirred at 4°C for an additional 25 min, then centrifuged for 1.5 hours (33,000 x g) at 4°C. The pellets were washed three times in lysis buffer + 500 mM NaCl. For each wash, the pellets were resuspended using a Dounce homogenizer, then centrifuged again. RNAP was eluted by washing the pellets three times with lysis buffer + 1 M NaCl. The PEI elutions were combined and precipitated with ammonium sulfate overnight. The mixture was centrifuged and the pellets were resuspended in RNAP buffer [20 mM Tris-HCl, pH 8.0, 5% glycerol (v/v), 5 mM DTT] + 1 M NaCl. The mixture was loaded onto three 5 mL HiTrap IMAC HP columns (Cytiva) for a total column volume (CV) of 15 ml. RNAP(β’-PPX-His₁₀) was eluted with RNAP buffer + 250 mM imidazole. The eluted RNAP fractions were combined and dialyzed against RNAP buffer + 100 mM NaCl. The sample was then loaded onto a 35 mL Biorex-70 column (Bio-Rad, Hercules, CA), washed with 10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 5% glycerol (v/v), 5 mM DTT, in a gradient from 0.2 M to 0.7 M NaCl. The eluted fractions were combined, concentrated by centrifugal filtration, then loaded onto a 320 mL HiLoad 26/600 Superdex 200 column (Cytiva) equilibrated in gel filtration buffer [10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 0.5 M NaCl, 5% glycerol (v/v), 5 mM DTT]. The eluted RNAP was supplemented with glycerol to 23% (v/v), flash frozen in liquid N₂, and stored at −80°C.

Eco His₁₀-SUMO-σ⁷⁰ was expressed and purified as described previously ⁷. Plasmids encoding Eco His₁₀-SUMO-σ⁷⁰ were transformed into Eco BL21(DE3) by heat shock. The cells were grown in the presence of 50 μg/mL kanamycin to an OD₆₀₀ of 0.4 at 37°C, then the temperature was lowered to 30°C. At OD 0.6, protein expression was induced with 1 mM IPTG (final) for 2 hours. Cells were harvested by centrifugation and resuspended in sigma lysis buffer [20 mM Tris-HCl, pH 8.0, 5% glycerol (v/v), 500 mM NaCl, 0.1 mM EDTA, 5 mM imidazole] and flash frozen in liquid N₂. Cells were then thawed on ice, 2-mercaptoethanol (BME) and PMSF were added to 0.5 mM and 1 mM, respectively. After French Press lysis at 4°C, cell debris was removed by centrifugation. The lysate was loaded onto two 5 mL HiTrap IMAC HP columns (Cytiva) for a total CV of 10 ml. His₁₀-SUMO-σ⁷⁰ was eluted at 250 mM imidazole in 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 0.1 mM EDTA, 5% glycerol (v/v), 0.5 mM BME. Peak fractions were combined, cleaved with Ulp1, and dialyzed against 20 mM Tris-HCl, pH 8.0, 500 mM NaCl, 0.1 mM EDTA, 5% glycerol (v/v), 0.5 mM BME, resulting in a final imidazole concentration of 25 mM. The sample was loaded onto one 5 mL HiTrap IMAC HP column (Cytiva) to remove His₁₀-SUMO-tag along with any remaining uncut His₁₀-SUMO-σ⁷⁰. Tagless σ⁷⁰ was collected in the flowthrough and concentrated by centrifugal filtration. Pooled, cleaved samples were diluted to 200 mM NaCl using buffer A [10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 5% glycerol (v/v), 1 mM DTT] and loaded onto three 5 mL HiTrap Heparin columns (Cytiva) equilibrated in buffer A. The σ⁷⁰ was eluted over a NaCl gradient from 200 mM to 1 M. Peak fractions were pooled, concentrated and buffer exchanged into Superdex equilibration buffer [20 mM Tris-HCl, pH 8.0, 0.5 M NaCl, 5% glycerol (v/v), 1 mM DTT] using centrifugal filtration (Amicon Ultra, MilliporeSigma, Burlington, MA). Concentrated sample was then loaded onto a HiLoad 16/60 Superdex 200 size exclusion column (Cytiva); peak fractions of σ⁷⁰ were pooled, supplemented with glycerol to a final concentration of 20% (v/v), flash-frozen in liquid N₂, and stored at −80°C.

Preparation of Eσ⁷⁰ and λP_R DNA for cryo-EM.

Eσ⁷⁰ was formed by mixing core RNAP [α₂ββ’(His)₁₀ω] with a 2-fold molar excess of σ⁷⁰, and then incubating for 20-30 minutes in a heat block at 37°C. Eσ⁷⁰ was separated from free σ⁷⁰ on a Superose 6 Increase 10/300 GL size exclusion column (Cytiva) in gel filtration (GF) buffer (40 mM Tris-HCl, pH 8.0, 120 mM KCl, 10 mM MgCl₂, 10 mM DTT). Peak fractions of the eluted Eσ⁷⁰ were concentrated to ~10-15 mg/mL by centrifugal filtration (Amicon Ultra, MilliporeSigma). Eσ⁷⁰ samples were either taken on ice to the New York Structural Biology Center for same day tr-Spotiton experiments or flash frozen in N₂(l) and stored at −80°C for later experiments.

All λP_R DNA constructs were commercially synthesized (Integrated DNA Technologies, Coralville, IA). To investigate upstream DNA-Eσ⁷⁰ interactions formed at early times (tr-Spotiton experiments), λP_R oligomers from −85 to +20 (Extended Data Fig. 1a) ⁷⁶ were used. For grids prepared using the Mark IV Vitrobot (equilibrium, 5.2 °C; Extended Data Fig. 9), λP_R DNA was −60 to +30 (Extended Data Fig. 9a) ⁷. For either oligomer, nt-strand and t-strand DNA (Extended Data Figs. 1a, 9a) was resuspended in annealing buffer (10 mM Tris-HCl, pH 8.0, 50 mM KCl, 0.1 mM EDTA). Equimolar amounts of the strands were mixed and incubated in a 95°C heat block for 5-10 minutes. The heat block was then removed to the benchtop where annealed strands slow cooled to room temperature. Annealed DNA was stored at −20°C before use.

Cryo-EM grid preparation.

tr-Spotiton.

On-grid mixing experiments were performed using the tr-Spotiton instrument as described previously ^20,77. Eσ⁷⁰ and λP_R DNA were thawed on ice and diluted to 26 or 29.25 μM Eσ⁷⁰ or to 52 or 58 μM λP_R DNA using GF buffer (1 Eσ⁷⁰:2 DNA molar ratio). Before spraying, nanowire self-blotting grids ^22,78 were plasma-treated (Gatan Solarus) at 5 W in H₂ (g) and O₂ (g) for 1 to 2 minutes. Grids were then placed in the tr-Spotiton chamber at RT and held at a relative humidity of 70 to 100%. Prior to loading into the dispensing tips, CHAPSO [(3-([3-cholamidopropyl]dimethylammonio)-2-hydroxy-1-propanesulfonate), Anatrace, Maumee, OH) 8 mM final, all 120 ms experiments and one 500 ms experiment] or fluorinated fos-choline-8 (FC8F; (1H, 1H, 2H, 2H-perfluorooctyl)-phosphocholine, Anatrace, Maumee, OH. 1.5 mM final, one 500 ms experiment) was added to each sample at RT (23 – 24 °C). After loading, equivalent streams of ~50 pL droplets of λP_R DNA and Eσ⁷⁰ were sequentially applied within 10 ms onto the grid (Extended Data Fig. 1b). Grid acceleration, deceleration, and final velocity parameters were chosen to achieve on-grid mixing times of 120 ms and 500 ms prior to the plunge into ethane (I). (See Supplemental Video 1).

5°C equilibrium grids.

After mixing at RT, Eσ⁷⁰ (15 μM final) and λP_R DNA (18 μM final) were equilibrated on ice in GF buffer and 8 mM CHAPSO. C-flat holey carbon grids (CF-1.2/1.3-4Au, Protochips, Morrisville, NC) were glow-discharged using a Solarus Plasma Cleaner (Gatan, Inc., Pleasanton, CA) for 20 seconds in air prior to the application of 3 to 3.5 μL of sample. Grids were blotted and plunge-frozen into liquid ethane with 100% chamber humidity at 5.2 °C using a Vitrobot Mark IV (FEI, Hillsboro, OR) instrument. Eσ⁷⁰ and λP_R DNA were on ice for 45 to 100 minutes during grid preparation.

Cryo-EM data acquisition and processing.

Structural biology software was accessed through the SBGrid consortium ⁷⁹. All datasets were collected at the Simons Electron Microscopy Center (SEMC; New York, NY) and recorded using Leginon ⁸⁰. Grids were imaged using a Titan Krios (300 kV accelerating voltage; ThermoFisher Scientific) equipped with a BioQuantum Imaging filter (slit width 20 eV) and either a K3 (tr-Spotiton datasets) or a K2 (5°C dataset) direct electron detector (Gatan, Inc., Pleasanton, CA). All tr-Spotiton grid images were recorded in counting mode with a physical pixel size of 0.844 Å; defocus ranges are listed in Table 1. 5°C grid images were recorded in super-resolution with a physical pixel size of 1.10 A with a defocus range of −1.0 to −2.3 μM. Dose-fractionated movies were gain-normalized, drift-corrected, summed, and dose-weighted using MotionCor2 ⁸¹. Data were processed using RELION ⁸² and cryoSPARC (CS) ⁸³ (Extended Data Figs. 2–4)

Eσ⁷⁰-λP_R tr-Spotiton - 120 ms mixing time.

Datasets 1_120ms and 2_120ms were independently processed using the pipeline outlined in Extended Data Fig. 2a. After motion correction, the Contrast Transfer Function (CTF) for datasets 1_120ms (2,703 micrographs) and 2_120ms (2,614 micrographs) was estimated using the Patch CTF module in CS ⁸³. CS Blob picker (150-300 Å, local maxima 500) followed by particle extraction (Extract from Micrographs; box size 384 px) yielded an initial set of 593,407 particles (p) (1_120ms) and of 579,917 p (2_120ms). Class averages after two rounds of CS 2D classification (N=100 classes) were used to select a subset of particles for reference-free CS Ab-initio Reconstruction (N=3). The ab-initio models (1=RNAP, 2=decoy, 3=decoy) were used to curate all extracted particles using CS Heterogeneous Refinement (N=6, each ab-initio model supplied twice). After two rounds of heterogeneous refinement, 119,486 p (1_120ms) and 122,054 p (2_120ms) were refined using CS Non-uniform (NU) refinement ⁸⁴. These particles then underwent one round of Bayesian polishing in RELION ⁷³. The polished particles were re-imported into CS where they were NU-refined together (3.0 A, 241,540 p).

To increase the number of particles for later classification strategies, grids from an independent 120 ms tr-Spotiton experiment were imaged and processed (Extended Data Fig. 2b). For dataset 3_120ms, 21,630 movies (Table 1) were motion-corrected and the CTF was estimated using CS Patch CTF. Micrographs with a CTF fit resolution > 10 were eliminated (2,524). CS Blob picker (150-300 Å, local maxima 500) followed by particle extraction (Extract from Micrographs; box size 256 px) from the remaining micrographs (19,106) yielded 5,731,694 p. Two rounds of CS 2D classification (N=100 classes followed by N=50) were used to select a subset of particles for CS Ab-initio Reconstruction (N=3). The ab-initio models (1=decoy, 2=decoy, 3=RNAP) were used to curate all extracted particles using CS Heterogeneous Refinement (N=6, each ab-initio model supplied twice). After four rounds of heterogeneous refinement, 372,303 p were re-extracted with a boxsize of 384 px, NU refined (nominal resolution of 3.6 A) and further processed in RELION where they underwent two rounds of Bayesian polishing ⁷³. Polished particles were NU refined (nominal resolution of 3.3 A). Polished particles from datasets 1_120ms, 2_120ms and 3_120ms were aligned in Class3D (N=1) using the same input volume, 3D auto-refined, and then combined (Join Star) in RELION ⁸² (Extended Data Fig. 4).

Eσ⁷⁰-λP_R tr-Spotiton - 500 ms mixing time.

To examine whether the relative populations of intermediates change at longer mixing times, a 500 ms tr-Spotiton dataset was obtained and imaged (Table 1). After motion correction, CTF for dataset 4500ms (14,506 micrographs) was estimated using the CS Patch CTF module ⁸³. Micrographs with a CTF fit resolution > 10 were eliminated, leaving 13,511 micrographs. CS Blob picker (150-300 Å, local maxima 400) followed by particle extraction (Extract from Micrographs; box size 256 px) yielded an initial set of 2,787,559 p. Two rounds of CS 2D classification (N=100, N=200 classes) were used to select particles for CS Ab-initio Reconstruction (N=3). The ab-initio models (1=decoy, 2=decoy, 3=RNAP) were used to curate the extracted particles using CS Heterogeneous Refinement (N=6, each ab-initio model supplied twice). After three rounds of heterogeneous refinement, 304,074 p were re-extracted with box size of 384 px, refined using CS Non-uniform (NU) refinement ⁸⁴, and then further processed in RELION with two rounds of Bayesian polishing ⁸⁵. The polished particles were NU refined in CS (3.0 A) (Extended Data Fig. 3).

3D classification with subtraction.

Polished particles from the 120 ms or the 500 ms datasets were first independently classified using 3D masked classification with subtraction in RELION ⁸⁶. A soft mask encompassing regions that define or bind in the RNAP active site channel was constructed using Chimera ⁸⁷. The “channel” mask included the following: σ⁷⁰_1.1, σ⁷⁰_1.2, σ⁷⁰₂ and σ⁷⁰₃ (excluding the σ⁷⁰_NCR,), the βprotrusion and βlobe, the clamp and β’jaw (excluding SI3), and a model of DNA from - 17 to +15 based on PDB 6EE8 ¹⁹. Classification of subtracted particles did not yield readily interpretable results for the 120 ms datasets. For the 500 ms dataset, 3D classification within the channel mask (N=3, without alignment) after subtraction yielded a class of low-resolution “junk” particles (16%) and two higher resolution classes (84%; Extended Data Fig. 3). The two high-quality classes were distinguished by differences in the cryo-EM density in the downstream channel. Particles from these two classes were combined and used in further processing (Extended Data Figs. 3 and 4).

3D Variability Analysis (3DVA).

To assess whether discrete states (intermediates) were populated at 120 and 500 ms or whether the conformational landscape at early times is better described as continuous flexibility, the tr-Spotiton datasets were processed using the CS 3DVA algorithm ²¹. Variability components (N=3 components) estimated from the entire consensus 3D maps were dominated by motions in DNA upstream of the −35 hexamer and in the σ⁷⁰_NCR (120 ms or 500 ms). To focus on differences in the active site channel, we first analyzed the 120 and 500 ms datasets independently by locally aligning the particle images in the channel mask, performing 3DVA in the mask, followed by Gaussian mixture model (GMM)-based clustering (N=8). Four discrete classes were detected at each time (detailed below).

For the final processing reported here, the following pipeline was executed. First, we combined the 120 ms and 500 ms particles, reasoning that any conformational differences as a function of time would be robust enough to be sorted during 3DVA classification, and that a larger dataset might amplify low population states. Because the tr-Spotiton sprays do not completely overlap on the grid and because free RNAP may exist in the population in equilibrium with bound DNA, we eliminated free RNAP particles by constructing a soft “DNA” mask around the promoter DNA from ~−40 to −10, the proximal αCTD, and σ⁷⁰_2-4. After signal subtraction using the DNA mask, the combined particle stack underwent masked 3D classification (RELION, N=6 classes, without alignment) followed by NU refinement (CS), yielding a low resolution “junk” class (6%), one free RNAP class (13%), and four DNA bound classes (80%; Extended Data Fig. 4).

DNA-bound particles (635,069 p) were combined, NU refined, and then locally aligned in CS using the channel mask. The resulting particle stack was then subjected to 3DVA (N=3, filter resolution=6 Å). GMM-based clustering of the variability component corresponding to clamp opening and closing movements (N=8; see Main Text, Supplemental Video 2) revealed four distinct conformational states, distinguished primarily by different degrees of clamp closure, extents of resolution of the single stranded nt-strand and changes in the cryo-EM density in the downstream channel. Particles corresponding to each class were combined and NU refined with per-particle defocus refinement and Global CTF refinement enabled (Extended Data Fig. 4).

Eσ⁷⁰-λP_R tr-Spotiton - 500 ms using fluorinated Fos-Choline-8 (FC8F).

Because of its effectiveness in preventing preferred particle orientations, CHAPSO has been used in cryo-EM studies, particularly those of Eco RNAP ²⁶. However, CHAPSO binds multiple sites on Eco RNAP ²⁶, including the σ-finger (σ⁷⁰_3.2) in the active site cleft ⁷. To examine whether CHAPSO influenced the nature of the intermediates captured in tr-Spotiton, we (A.U.M.) found FC8F (1.5 mM) acts as effectively as CHAPSO to mitigate particle orientation bias but does not interact with Eσ⁷⁰. Cryo-EM grids obtained from a 500 ms tr-Spotiton experiment where 1.5 mM FC8F replaced CHAPSO were imaged and processed using the same pipeline used for the 500 ms data in CHAPSO (Extended Data Fig. 8). The same intermediates were found, indicating that the results are independent of CHAPSO.

Eσ⁷⁰-λP_R complexes populated at 5°C.

Cryo-EM grids prepared by vitrifying complexes populated at equilibrium at low temperature were imaged and processed using the pipeline shown in Extended Data Fig. 9.

Local resolution-filtered cryo-EM maps were generated using blocres and blocfilt from the Bsoft package ⁷⁵ with the following parameters: box size 20, sampling using the physical pixel size (Extended Data Figs. 6, 7, 9; Table 1). Directional 3DFSCs were calculated using 3DFSC ⁸⁸ (Extended Data Figs. 6, 7, 9).

Model building and refinement.

To build initial models of the protein-DNA components in the tr-Spotiton cryo-EM density maps, Eσ⁷⁰ bound to λP_R from PDB 7MKD ⁷ was manually fit into the density map using Chimera ⁸⁷ and real-space refined using PHENIX ⁸⁹. The Eσ⁷⁰-rpsT P2 promoter closed complex (RPc) (PDB 6PSQ) ¹⁰ served as the initial model for the RPc intermediate populated at 5°C at λP_R. Real-space refinement, rigid body refinement with sixteen manually-defined mobile domains was followed by all-atom and B-factor refinement with Ramachandran and secondary structure restraints. Models were inspected and modified in Coot ⁹⁰. Alignment shown in Fig. 1b was done using conserved domains of RNAP that exhibit minimal conformational changes in the transcription cycle (the RNAP structural core) ⁹¹. Statistical analyses for model refinement and validation were generated using MolProbity ⁹² and PHENIX ⁸⁹.

Disulfide crosslinking and abortive transcription assays.

Purification of σ⁷⁰ region 1.1 cysteine pair variants.

A cysteine-less variant of Eco σ⁷⁰ (C132S, C291S, C295S) and variants containing cysteine pair mutations at I35C-S89C, Q8C-P32C, or Y21C-Q54C were expressed with a His10-SUMO N-terminal tag from a pET28-based vector in Eco BL21(DE3). 2 L shaking cultures inoculated from overnight cultures of freshly transformed cells were grown at 37°C to OD600 of 0.7 to 1.0, at which point the temperature was reduced to 25°C. Expression was induced after 1 h with a final concentration of 0.5 mM IPTG and continued overnight at 25°C. Cells were harvested by centrifugation and resuspended in about 20 ml/L culture of buffer A [20 mM Tris-HCl pH 8.0/RT, 1 M NaCl, 5% glycerol (v/v), 2 mM DTT] supplemented with 5 mM imidazole, 1 mM phenylmethylsulfonyl flouride (PMSF, Sigma-Aldrich) and 1x protease inhibitor cocktail (c0mplete EDTA-free, Roche). Cells were lysed using a continuous-flow homogenizer (Avestin) at 4°C. After centrifugation of the lysate at 15,000 rpm (JA-18 rotor; ~33,000 x g) for 30 min at 4°C, the soluble fraction was recovered, filtered through 5 μM PVDF membrane (EMD Millipore), and applied to a 5 ml HiTrap IMAC Sepharose FF column (Cytiva) charged with Ni²⁺ ions and equilibrated in buffer B [20 mM Tris-HCl pH 8.0/RT, 1 M NaCl, 5% glycerol (v/v), 2 mM DTT, 500 mM imidazole]. Subsequently, unbound material was eluted from the column by wash steps using buffer A and B of 5 CV 5% B, and 10 CV 10% B. Protein was eluted with 70% B (with 350 mM imidazole) and fractions were analyzed for protein content by SDS-PAGE and Coomassie staining. Protein-containing fractions were pooled and supplemented with 0.5 mM EDTA and 1 mM DTT. To remove the N-terminal tag, His-tagged Ulp1 protease was added at 1:30 to 1:50 molar ratio and the sample was dialyzed in a SpectraPor 12-14 kDa membrane against 2 L buffer A supplemented with 0.1 mM EDTA overnight at 4°C. Ulp1 and uncleaved protein were removed by passing the sample over a 5 ml HiTrap IMAC Sepharose FF column (Cytiva) charged with Ni²⁺ ions. The flow through was concentrated using centrifugal filters with a 30 kDa cutoff membrane (Amicon 30K, EMD Millipore) and applied to a Superdex 200 prep grade 26/600 column (Cytiva) using buffer C [20 mM HEPES-NaOH pH 8.0/RT, 500 mM NaCl, 5% glycerol (v/v), 5 mM DTT, 0.1 mM EDTA]. Peak fractions were collected and pooled according to protein content. Variants were concentrated to 100-250 μM protein using centrifugal filters (Amicon 30K) and the final sample was mixed 1:1 with buffer C containing 50% glycerol (v/v) before freezing aliquots in liquid nitrogen. Frozen protein samples were stored at −80°C until use.

In vitro abortive transcription assay for σ⁷⁰_1.1 cysteine pair variants.

Cysteine pair variants and the cysteine-less variant of Eco σ⁷⁰ were buffer exchanged to buffer R (40 mM Tris-HCl pH 8/RT, 120 mM KCl, 10 mM MgCl2, 20 ug/ml BSA) using Zeba 5K desalting spin columns (ThermoFisher Scientific). The variants were subjected to oxidation by mixing σ⁷⁰ variant at 2.5 μM final concentration with hydrogen peroxide at 5 mM final concentration in buffer R. Reactions (typically 10-20 μl) were incubated for 10 min at 37°C. After the incubation, catalase (1 U/ul) was added to about 0.17 U/ul final concentration to neutralize the hydrogen peroxide and the reactions were incubated at 25°C for at least 5 min. Crosslinking efficiency was assessed by SDS-PAGE using 10% AA/BAA (37.5:1) Bis-Tris gels and Coomassie staining (typically >90%).

Abortive transcription reactions were carried out on a λP_R promoter fragment (−60 to +30) ⁷. First, Eσ⁷⁰ was formed by mixing Eco core RNAP with σ^7- variant (and DTT for the reaction under reducing conditions) in buffer R at 37°C for 10 min. DNA was added to start formation of open complexes. Final concentrations of each component were 40 nM core RNAP, 100 nM σ^7- variant, 10 nM DNA, and 1 mM DTT for the reactions under reducing conditions (DTT was replaced with buffer R for oxidizing conditions). After incubating the reaction at 25°C for 90 seconds, the complexes were challenged with heparin (50 ug/ml final) for 1-2 min (depending on the handling time). NTP mix (250 uM ApU, 50 uM GTP, 130 nCi/μl α-³²P-GTP, 50ug/mL heparin) was added to start abortive transcription. Reactions were incubated for 5 min at 25°C before mixing with an equal volume of 2x STOP buffer [0.5x TBE, 8 M urea, 30 mM EDTA, 0.05 % bromophenol blue (w/v), 0.05 % xylene cyanol (w/v]) and heating to 95°C to stop the reaction. Products were analyzed by loading 4.5 μl sample on a 23% TBE-urea gel run in 1x TBE for 1:30 h at 1000 V. Gels were exposed to a storage phosphor screen overnight at 4°C. Imaging was performed on a Typhoon imager (Cytiva).

Extended Data

Extended Data Fig. 1 |. λP_R promoter fragment and tr-Spotiton.

a. λP_R promoter DNA construct (−85 to +20) used for cryo-EM studies. The sequence from −40 to +10 is magnified below.

b. Schematic diagram illustrating the principle of the tr-Spotiton device. For more details see ref. ²⁰.

Extended Data Fig. 2 |. Cryo-EM processing pipeline for 120 ms datasets.

Cryo-EM processing pipelines for Eco RNAP mixed with λP_R DNA using tr-Spotiton (t = 120 ms, 8 mM CHAPSO)

a. Datasets 1_120ms and 2_120ms.

b. Dataset 3_120ms.

Extended Data Fig. 3 |. Cryo-EM processing pipeline for dataset 4_400ms.

Cryo-EM processing pipeline for Eco RNAP mixed with λP_R DNA using tr-Spotiton (t = 500 ms, 8 mM CHAPSO, dataset 4).

Extended Data Fig. 4 |. Cryo-EM processing pipeline for combined datasets.

Cryo-EM processing pipeline for combining polished particles from Spotiton datasets 1 - 3 (t = 120 ms; see Extended Data Fig. 2) and dataset 4 (t = 500 ms; see Extended Data Fig. 3).

Extended Data Fig. 5 |. Comparison of particle population distributions for 120 ms vs. 500 ms datasets.

Histogram plots showing the fraction of particles that contribute to each intermedate. The open bars show the mean particle fraction for the three 120 ms datasets (<123_120ms>); the error bars denote the standard deviation for n=3. The gray bars denote the particle fraction for the single 500 ms dataset (4_500ms).

Extended Data Fig. 6 |. Cryo-EM of I1a and I1b.

a.-d. Cryo-EM of I1a.

a. Three views of the combined nominal 2.8 Å resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

b. Directional 3D FSC, determined with 3DFSC ⁸⁸.

c. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

d. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

e.-h. Cryo-EM of I1b.

e. Three views of the combined nominal 3.0 A resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; α’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

f. Directional 3D FSC, determined with 3DFSC ⁸⁸.

g. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

h. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

Extended Data Fig. 7 |. Cryo-EM of I1c and I1d.

a.-d. Cryo-EM of I1c.

a. Three views of the combined nominal 3.0 Å resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

b. Directional 3D FSC, determined with 3DFSC ⁸⁸.

c. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

d. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

e.-h. Cryo-EM of I1d.

e. Three views of the combined nominal 2.9 Å resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

f. Directional 3D FSC, determined with 3DFSC ⁸⁸.

g. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

h. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

Extended Data Fig. 8 |. Cryo-EM processing pipeline for dataset 5_500ms,FC8F.

Cryo-EM processing pipeline for Eco RNAP mixed with λP_R DNA using tr-Spotiton (t = 500 ms, 1.5 mM FC8F).

Extended Data Fig. 9 |. Cryo-EM processing pipeline for 5°C dataset.

a. λP_R promoter DNA construct used for 5°C cryo-EM studies.

b. Cryo-EM processing pipeline for Eco RNAP and λP_R DNA (−60 to +30) mixed manually and allowed to come to equilibrium at 5°C (See Methods).

Extended Data Fig. 10 |. Cryo-EM of RP_C5°C, I1_C5°C, and I1D_5°C.

a.-d. Cryo-EM of RP_C5°C.

a. Three views of the combined nominal 3.1 Å resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

b. Directional 3D FSC, determined with 3DFSC ⁸⁸.

c. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

d. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

e.-h. CryoEM of I1_C5°C.

e. Three views of the combined nominal 3.4 A resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

f. Directional 3D FSC, determined with 3DFSC ⁸⁸.

g. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

h. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

j.-m. CryoEM of I1d_5°C.

j. Three views of the combined nominal 3.2 Å resolution cryo-EM map, filtered by local resolution ⁷⁵. The bottom row is colored according to the subunit (αI, αII, light grey; αCTD, limon; β, cyan; β’, pink; σ⁷⁰, orange; DNA t-strand, dark grey; DNA nt-strand, grey). The top row shows the same views but colored by local resolution ⁷⁵. The right view is a cross-section through the middle view.

k. Directional 3D FSC, determined with 3DFSC ⁸⁸.

l. Gold-standard FSC plot ⁹³, calculated by comparing two half maps. The dotted line represents the 0.143 FSC cutoff.

m. Angular distribution plot, calculated in cryoSPARC. Scale depicts number of particles assigned to a specific angular bin.

Extended Data Fig. 11 |. The σ70 W-dyad and the −12 bp in λP_R intermediates.

a. Top view of λP_R-RPo (7MKD) ⁷. Eσ⁷⁰ is shown as a transparent molecular surface. The DNA is shown as atomic spheres, color-coded as in Fig. 2a.

b.-f. The boxed region in (a) is magnified, showing the region of the σ⁷⁰ W-dyad and the −13 to −11 positions of the promoter. σ⁷⁰ and DNA (color-coded as in Fig. 3) are shown in stick format; σ⁷⁰ carbon atoms are colored orange but the W-dyad is highlighted in yellow. Transparent cryo-EM density (local-resolution filtered ⁷⁵) is superimposed. For reference, the positions of key RPo elements are shown in stick format and colored chartreuse (W-dyad in chair conformation and the −12 bp). For I1a, I1b, I1c, and I1d (b.- e.), the W-dyad is in the edge-on (wedge) conformation and the −12 bp is opened. Only in RPo is the W-dyad in the chair conformation and the −12 bp re-paired.

b. I1a.

c. I1b.

d. I1c.

e. I1d.

f. RPo.

Extended Data Fig. 12 |. σ⁷⁰_1.1 disulfide crosslinking, and a conformational change in the σ-finger.

σ⁷⁰ derivatives were analyzed by 10% SDS-polyacrylamide gel electrophoresis and visualized with Coomassie stain. Each σ⁷⁰ derivative was analyzed under reducing (preventing formation of any disulfide bonds) or oxidizing (promoting the formation of disulfide bonds) conditions. Each Cys-pair mutant shows higher mobility under oxidizing conditions indicating formation of the relevant disulfide bond. Moreover, the difference in mobility between the reduced and oxidized condition correlates with the number of residues separating the two engineered Cys substitutions (I35C-S89C, 55 residues; Q8C-P32C, 25 residues; Y21C-Q54C, 34 residues).

b. The λP_R-RPo structure (7MKD) ⁷ in the active-site region is shown; The RNAP is shown as a backbone cartoon (β, light cyan; β’, light pink; σ⁷⁰, orange); t-strand DNA is shown in stick format (carbon atoms dark grey); the RNAP active-site Mg²⁺ is shown as a yellow sphere. The structures of I1a, I1b, I1c, and I1d from the CHAPSO (light green) and FC8F (brown) datasets were superimposed by the RNAP structural core and shown is the σ-finger from each.

Data Availability.

All unique/stable reagents generated in this study are available without restriction from the Lead Contact, Seth A. Darst (darst@rockefeller.edu). The cryo-EM density maps and atomic coordinates have been deposited in the EMDataBank and Protein Data Bank as follows: RP_c5°c (EMD-41456, 8TOM), I1a (EMD-41433, 8TO1), I1b (EMD-41439, 8TO8), I1c (EMD-41448, 8TOE), I1d (EMD-41437, 8TO6).