Complete architecture of the archaeal RNA polymerase open complex including DNA, TBP, TFB and TFE

Abstract

The molecular architecture of RNAPII-like transcription initiation complexes has been studied for years but its structure has remained opaque due to its conformational flexibility and size. We determined the three-dimensional architecture of the complete open complex (OC) composed of the promoter DNA, TATA box-binding protein (TBP), transcription factors TFB and TFE, and the 12-subunit RNA polymerase (RNAP) from M. jannaschii. By combining single-molecule Förster resonance energy transfer (smFRET) and the Bayesian parameter estimation based Nano-Positioning System (NPS) analysis, we modelled the entire archaeal OC, which elucidates the path of the ntDNA strand and interaction sites of the transcription factors with the RNAP. Compared to models of the eukaryotic OC, the position of the TATA DNA region with TBP and TFB is positioned closer to the surface of the RNAP, likely providing the mechanism by which DNA melting can occur in a minimal factor configuration, without the dedicated translocase/helicase encoding factor TFIIH.

Introduction

Transcription of all cellular genomes is carried out by evolutionary related multisubunit RNA polymerases (RNAPs). In contrast to eukaryotes, where different types of RNAPs exist, archaea utilise only one RNAP to transcribe their genes, but its subunit composition, structure and utilisation of general transcription factors is strikingly similar to the eukaryotic RNA polymerase II (Pol II) system^1,2.

Eukaryotic Pol II involves the interplay of a large set of transcription factors to initiate transcription, most importantly general transcription factors TFIIA, -B, -D, -E, -F, and -H. However, not all factors are strictly required; in particular TFIIA is only necessary to alleviate the repressive effects of negative regulators such as NC1. Moreover, using strong promoters and negatively supercoiled DNA templates only two factors TBP and TFIIB suffice to direct start-site specific transcription initiation by Pol II in vitro³. TBP and TFIIB assemble at the promoter⁴ and recruit Pol II as well as other factors to form the preinitiation complex (PIC). This complex is referred to as the closed complex (CC), which subsequently undergoes large conformational rearrangements during which the DNA strands are separated and the template strand is loaded into the RNAP active site to form the open complex (OC). In eukaryotes, this process is greatly enhanced by the helicase activities encoded by TFIIH.

The large size, heterogeneous composition and conformationally dynamic nature of eukaryotic PICs have made their structural and functional analysis problematic. Recent advances in the field have improved our understanding of the overall structural organisation of the eukaryotic PIC. X-ray structures of PIC sub-complexes (encompassing Pol II and TFIIB fragments) at high resolution have provided hints of mechanistic aspects of the closed to open complex transition during transcription initiation. Thus the crystal structure of the Pol II-TFIIB complex allowed for modelling of the closed and open complex^5–7. Several cross-linking studies have yielded information about the location of transcription factors TFIIB⁸, TFIIF, TFIIE^9,10 and TFIIH¹¹. Recent cryo-EM studies have provided overall structures of both the eukaryotic and archaeal PICs at low to intermediate resolution^12–14. However, all of these studies have failed to fully resolve the course of the DNA within the OC, likely due to the flexibility of the transcription bubble.

The archaeal transcription apparatus is an excellent model system for the eukaryotic Pol II system¹⁵ as its RNAP and associated basal transcription factors are homologous, and because the entire system from hyperthermophilic archaea can be reconstituted from recombinant proteins¹⁶. This enables us to site-specifically introduce mutations or molecular probes such as fluorescent dyes for single-molecule fluorescence analysis^17,18. The factors TBP and TFB (homologous to TFIIB) are necessary and sufficient for promoter-directed start site-specific transcription initiation of the archaeal RNAP, which mirrors the minimal factor requirements for Pol II^3,16,19. A third factor, TFE (homologous to TFIIEα) interacts with RNAP and stimulates open complex formation, also comparable to the Pol II system^18,20–22. The pivotal difference between the two systems is the apparent ease at which the open complex is formed in archaea, whereas in eukaryotes the additional helicase/translocase activity of the general factor TFIIH is required in vivo, which is not conserved in any archaeal species.

Single-molecule techniques have shown great potential to resolve the dynamics of transcription processes because they allow for the direct and real-time observation of transcription, one molecule at a time²³.

In order to obtain quantitative structural and dynamic information about transcription complexes during various phases of transcription, the Nano-Positioning System (NPS) was developed²⁴. The NPS combines data from smFRET measurements with existing structural information and a rigorous analysis using Bayesian parameter estimation. As a result, three-dimensional probability density functions for dye molecules (“antennas”) attached to positions in unknown, flexible regions of the complex of interest can be calculated. For visualisation, the smallest volume enclosing a certain probability of the computed density, the credible volume, can be displayed together with the known structure. The recorded smFRET data are measured between the antennas and dye molecule (“satellites”) attached to known positions from the crystal structure ²⁴. This method has been used in Pol II transcription elongation complexes to study the position of the exiting RNA²⁵, the influence of transcription factor TFIIB on the position of the nascent RNA²⁴ and the position of non-template and upstream DNA²⁶. Moreover, the architecture of a minimal Pol II open complex²⁷ and the position of transcription factor TFE in the archaeal PIC¹⁸ have been determined. NPS has been further extended to a global analysis where a complete data set of all measured information about all the antenna-satellite pairs is used as a network to simultaneously infer the position of all the antennas²⁸.

Here, we used smFRET, global NPS analysis and fluorescently labelled components of the transcription machinery from the hyperthermophilic archaeon M. jannaschii to determine the molecular architecture of the complete archaeal open complex consisting of RNAP, promoter DNA, TBP, TFB and TFE. We determined the smFRET efficiencies between unknown "antenna" dye molecules and several known "satellite" dye molecules incorporated at one of five reference sites in the RNAP, whose position can be inferred from crystallographic structures. The “antenna” dyes were attached to either one of several positions on the upstream and downstream ntDNA strand, to TBP, or to TFB. The probability densities generated from the NPS calculation allowed us to build a model of the complete archaeal OC, which provides valuable insights into the mechanism of transcription initiation. We find the two factors TBP and TFB to be located closer to the RNAP surface in archaeal complexes compared to the Pol II system. This can provide an answer to the question why the closed-to-open complex transition readily occurs in archaea but necessitates TFIIH in the Pol II system, thus illustrating how during evolution of the eukaryotic domain of life subtle changes in the architecture of the initiation complex renders DNA melting largely dependent on TFIIH.

Results

Assembly of well defined OCs for smFRET experiments

Complete archaeal OCs were assembled using M. jannaschii TBP, TFB, TFE and RNAP (Methods) on the strong SSV T6 promoter as DNA template. In order to ensure that the complexes were in the open state, a non-complementary 4 nucleotide ‘mismatch’ was introduced in the promoter^16,29. To perform smFRET experiments, the complexes were labelled with a fluorescent donor and acceptor at desired locations. Fluorescently labelled DNA oligonucleotides, TBP, TFB, TFE and RNAP were combined to yield a large network of around 70 differently labelled complexes, each with a single smFRET pair at a desired location (Figure 1A). Previous biochemical studies had established the formation of stable open complexes from these components, capable of promoter specific transcription initiation in vitro, forming RNA templates from a precise starting point (+1)²². Complex formation of this in vitro OC was also verified by Electrophoretic Mobility Shift Assays (EMSA)¹⁸.

Figure 1. Schematic representation of the global FRET network used to determine the complete architecture of the archaeal open complex with NPS.

(A) All unknown antenna positions (green circles) and the five known satellite sites on the RNAP (dark red circles) are shown together with the corresponding attached dyes (A647 = Alexa 647; A555 = Alexa555; Dl550 = DyLight550; Dl650 = DyLight650). FRET efficien-cies were measured between pairs of satellite (acceptor) and antenna (donor) dyes (dotted black lines) and in between antennas (dotted red lines). In case of the latter measurements, one of the antenna positions had to be labelled with an acceptor, as indicated. (B) Cartoon depicting the mismatched (-3 to +1) viral SSV T6 promoter (tDNA in blue, ntDNA in cyan, TATA box in red), which is used throughout this study⁵¹. Labelling sites on the DNA and on the transcription factors TBP, TFB and TFE are marked with a green star. Labelling sites on the RNAP are marked with a dark red star. Proteins contained within the archaeal OC are shown schematically. See also Supplementary Figures 1, 2 and 4.

While smFRET experiments can reveal both the structure and dynamics of macromolecular complexes, they only focus on one smFRET value of one dye pair at a time. Therefore, it is important to ensure that complexes are formed properly and that the information obtained from the measurement is indicative of the desired complex. To this end, we performed a number of control experiments (Supplementary Methods). We found that stable open complexes were formed in a factor-dependent manner (Supplementary Fig. 1A-B) but that transcription factor TFE did not exert an influence on the architecture of the open complex (Supplementary Fig. 2A-D). Also, we found that the choice of the dye on the DNA strand had no effect on the distance information obtained from smFRET experiments (Supplementary Fig. 2E).

Open complexes with the template strand at the active center

In contrast to the non-template strand, the position of the template strand of the melted region in the open complex could be inferred from the crystal structure of yeast RNA polymerase II using a tailed DNA template³⁰. The position of the template strand resembles that of the respective elongation complex, even in the absence of RNA³¹. In order to build the model of the eukaryotic open complex, DNA opening was assumed to commence 20 bp downstream of TATA³² yielding a DNA melted in a region between positions (+2) to (-13)⁵.

In order to ascertain that the extent of the melted region in the M. jannaschii OC was comparable to published data of the Pyrococcus system^33,34, we used KMnO₄ footprinting (Supplementary Fig. 3 and Methods). This method detects thymidine nucleobases in single stranded DNA regions. The SSV T6 promoter template contains a mismatch region from (-3) to (+1) and thus contains an obligate single stranded T at register (-1). This residue serves as positive control and is detectable in the free promoter probe. Addition of TBP, TFB and RNAP leads to novel signals at register (-5) and (-7) reflecting the opening of the promoter by the transcription complex. Since the next T residue occurs at (-12) we conclude that the transcription bubble starts at (-1) extends to at least (-7) and importantly not beyond (-12), showing that the size of the melted region in our OC is in good agreement with the published results for the Pyrococcus system³³. Since pre-opened promoter templates were used the transcription factor TFE did not significantly alter DNA-melting and did not exert an influence on the architecture of the open complex (Supplementary Fig. 2 and 3).

Furthermore, we tested whether the template strand was properly loaded into the active centre cleft. Therefore, we used NPS to localise two dye positions on the template DNA (tDNA) strand, namely tDNA(+3) and tDNA(-9) (Methods, Figure 2A). For each position we performed a set of smFRET measurements with a second dye molecule attached to one of five reference sites on the RNAP: residue 257 of Rpo1’, residue 373 of Rpo2”, residue 11 of Rpo5, residues 49 and 65 of Rpo7 (Figure 2A, Supplementary Fig. 4 and Methods). Exemplary histograms are shown for position tDNA(-9) (Figure 2B-F) and the extracted data are summarised in Supplementary Table 1.

Figure 2. Localisation of two positions on the tDNA strand in the archaeal OC.

(A) Cartoon depicting the labelling positions on the tDNA strand which were localised using the NPS (colour coding according to Figure 1).

(B) - (F) Framewise smFRET histograms used in the NPS localisation of tDNA(-9). Shown is the smFRET data from measurements between tDNA(-9) and Rpo7-V49 (B), Rpo7-S65 (C), Rpo2"-Q373 (D), Rpo1'-G257 (E) and Rpo5-K11 (F), respectively. The histograms were fitted with a double (B-C, F) or single (D-E) Gaussian distribution indicated by the black lines. In (B-C, F) the gray line represents the sum of the two Gaussian distributions. Results of the fits together with those obtained during localisation of tDNA(+3) are summarized in Table 1.

(G) NPS results for the fluorescent probes attached to tDNA(+3) (pink) and tDNA(-9) (yellow). The X-ray structure of the archaeal polymerase of S. shibatae³⁷ (PDB: 2WAQ) is represented as dark grey ribbon. Note, that at this confidence level, the credible volume of tDNA(-9) is divided into two areas, if the credible volume is drawn at higher confidence these two areas merge (see text for details).

(H) Comparison of the NPS results to the eukaryotic open complex model⁵. The X-ray structure of the yeast polymerase is represented as light grey ribbon, the tDNA is shown in blue and the ntDNA is shown in cyan. The corresponding eukaryotic bases for tDNA(+3) (green) and tDNA(-9) are encircled. The NPS credible volumes are in good agreement to the model. See also Supplementary Fig. 3.

Many of the observed smFRET histograms showed a secondary peak with a relative intensity varying between 5-30 %. However, there was no evidence for dynamic interconversion between the two peaks. Instead, the side peak is likely caused by a different static population. Comparison of NPS analysis of side peaks and main peaks showed only minor changes in credible volume position (Supplementary Fig. 5). In the following we restrict ourselves to the discussion of the main peaks; however one should note that, while the side peaks would lead to small Angstrom level alterations of the model, the general conclusions of this work are not affected.

The use of Bayesian parameter estimation allows the computation of the most likely position and the three-dimensional uncertainty of the position of the fluorescent dye attached to the unknown position²⁴. For this, the uncertainties due to the presence of flexible linkers between the dye and the known positions on the RNAP were computed first (Supplementary Fig. 4 and Methods). Moreover, for each dye pair, we experimentally determined the fluorescence anisotropies and the isotropic Förster radii (Methods and Supplementary Table 2). Three-dimensional probability densities were then calculated using the respective linker lengths and the sizes of the dye molecules (Supplementary Table 3) and credible volumes were calculated and displayed in comparison with the crystal structure of the RNAP (Figure 2G and H). The size of all credible volumes presented in this study corresponds to 68 % credibility, representing the smallest volume, which encloses a probability of 68 %. The credible volume of the dye attached to tDNA (+3) is located inside the cleft, in good agreement to the eukaryotic OC models^5,27. Also tDNA(-9) localises at a position consistent to the eukaryotic OC models and the position is distinct from the one it would adopt in a closed complex conformation⁵ (Figure 2H). In the case of tDNA(-9) the displayed volume is split into two distinct sub-volumes. One should note that this is not originating from dynamic movement between these positions (our model is a static model) but represents the positioning uncertainty of this DNA position in the calculation at 68 % confidence level. In fact, if drawn at higher confidence the two volumes merge.

The path of the ntDNA strand within the archaeal OC

To determine the path of the ntDNA strand, we assembled a variety of open complexes, where a fluorescent donor dye was attached to ntDNA(+7), (-1), (-5), (-7), (-10), (-12), or (-14) (Methods). For each of these ntDNA fluorescent donor positions, a fluorescent acceptor was attached to one of five different reference sites on the RNAP, generating 5 unique complexes per labelled donor position. Exemplary histograms are shown (Supplementary Fig. 5B-F) and the extracted data are summarised in Supplementary Table 4. In the global NPS analysis²⁸, the complete data set consisting of mean FRET efficiencies, dye attachment information (position, length of linker and size of dye molecule), steady state fluorescence anisotropies and isotropic Förster distances of all antenna-satellite pairs and the uncertainty in position of the satellite dyes (due to linker length and dye molecule size) was used to simultaneously infer the positions of all antennas within the RNAP coordinate system (Figure 1 and Methods). As a result of Bayesian parameter estimation, we obtained the three dimensional probability density of each antenna, which represents the position of the dye attached to the DNA base (Figure 3A-C).

Figure 3. Localisation of the ntDNA strand in the archaeal OC.

(A) - (B) Side and Front view of the archaeal RNA polymerase including the NPS results for the fluorescent probes attached to ntDNA(+7) (brown), ntDNA(-1) (dark red), ntDNA(-5) (red), ntDNA(-7) (orange), ntDNA(-10) (yellow), ntDNA(-12) (green) and ntDNA(-14) (dark green, hardly visible).

(C) Top view (slightly rotated for a clearer view) of archaeal RNA polymerase with selective volumes shown at a time for clarity. See also Supplementary Figures 2 and 6.

The furthest downstream position was ntDNA(+7) in the downstream duplex region. The NPS credible volume is located close to the clamp head region of subunit Rpo1' but outside of the cleft (Figure 3C, black). NtDNA(-1) lies within the single-stranded region of the ntDNA strand and its position (Figure 3C, brown) is split into two distinct sub-volumes located at the edge of the cleft, proximal to the lobe domain of subunit Rpo2". As for tDNA(-9), these sub-volumes are not originating from dynamic movement but represent the uncertainty in position at 68 % confidence level; if drawn at higher confidence the two volumes would merge. The positions of the dye molecules attached to the next upstream bases ntDNA(-5) (Figure 3C, red) and ntDNA(-7) (Figure 3C, orange) occupy a similar region within the cleft, between the clamp core of subunit Rpo1' and the lobe domain of Rpo2". The position of ntDNA(-10) (Figure 3C, yellow) is located closer to the clamp coiled coil region of subunit Rpo1'. The volumes for ntDNA(-12) (Figure 3C, green) and ntDNA (-14) (Figure 3C, dark green, hardly visible) are largely overlapping and remain at the same side of the clamp coiled coil region as ntDNA(-10).

We repeated the NPS calculation with a slightly modified crystal structure of the archaeal RNAP, where we moved the clamp core region by 8 Å to mimic an open clamp polymerase structure such as observed in EM studies of the eukaryotic OC¹³. The position of the calculated credible volumes for all our antenna dyes remained largely unchanged by this alteration, i.e. changes were small compared to the size of the credible volumes and therefore all further discussion is based upon the closed clamp state of the polymerase in accordance with single-molecule experiments on the bacterial OC³⁵.

Position of TBP, TFB, TFE and TATA DNA in the archaeal OC

In order to determine the position of TBP, TFB and the upstream TATA DNA in the OC, we assembled complexes where a fluorescent donor was attached to positions on or around the predicted binding region of TBP, namely ntDNA(-18), (-24), (-30) or (-37), to residue S71 of TBP, and to residue G262 of TFB (Figure 1, Methods). Fluorescent acceptors were attached to one of four reference sites on the RNAP, namely residues Rpo1'-G257, Rpo2"-Q373, Rpo7-V49 or Rpo7-S65, as before (Supplementary Table 5). Residue K11 of Rpo5 was situated too far away to yield information for the localisation process and was therefore left out from the analysis. The position of residues G44 in the winged helix domain and G133 in the zinc ribbon domain of TFE had previously been determined by NPS¹⁸ and we included this smFRET data into the global NPS calculation to yield an accurate model of the complete archaeal OC (Figure 1).

The obtained NPS probability densities were relatively large, due to the small number of different satellites for each antenna (data not shown). To increase our resolution, we used a valuable feature of the global NPS analysis, which allows inclusion of FRET measurements between two unknown positions. We therefore included smFRET measurements from all the TATA DNA positions to both TBP and TFB and also smFRET measurements in between TBP and TFB (Figure 1A, dotted red lines and Supplementary Table 5). This procedure greatly increased the accuracy of all the determined dye positions. Corresponding histograms are shown (Supplementary Figures 6H-M and 7). As a result of Bayesian parameter estimation, we obtained the three dimensional probability density for the position of each antenna dye (Figure 4A-D).

Figure 4. Localisation of the TATA DNA, TBP, TFB and TFE in the archaeal OC.

(A) - (B) Side and front view of the archaeal RNA polymerase together with the NPS results for the fluorescent probes attached to ntDNA(-18) (dark cyan), ntDNA(-24) (dark blue), ntDNA(-30) (magenta), ntDNA(-37) (gold), TBP-S71 (purple), TFB-G262 (olive), TFE-G44 (yellow) and TFE-G133 (green).

(C) Top view of archaeal RNA polymerase with only the NPS credible volumes TFE-G44 (yellow) and TFE-G133 (green) shown at a time to illustrate their proximity to the clamp coiled coil region of subunit Rpo1' (red) and to the stalk of the RNAP (black), respectively.

(D) Alternative view obtained by a 90° rotation of the top view, showing the proximity of the NPS credible volume ntDNA(-18) (dark cyan) to the protrusion (brown) of subunit Rpo2". The NPS credible volume of ntDNA(-24) (dark blue) is situated further away from the RNAP surface compared to the position of ntDNA(-18). Also, the proximity of the NPS credible volumes TBP-S71 (purple) and TFB-G262 (olive) to the wall domain (beige) of subunit Rpo2" and to subunit Rpo12 (light blue) can be seen. The volume of TFB-G262 is made transparent for clarity. See also Supplementary Figures 6 and 7.

The credible volume for ntDNA(-18) (Figure 4D, dark cyan) is adjacent to the protrusion domain, and defines the path of the double-stranded ntDNA strand when compared to the position of the more downstream ntDNA(-14) and ntDNA(-12) (Figure 3C), which are located further away toward the clamp domain. The first credible volume describing the position of the TATA box, ntDNA(-24) (Figure 4D, dark blue) is situated closer to the RNAP wall but further away from the RNAP surface compared to the position of ntDNA(-18). Together with the credible volume of the second TATA box position, ntDNA(-30) (Figure 4A-B, magenta), the bend in the DNA caused by TBP (centred at positions -26/-27) can be visualised. The credible volume of the last localised position on the ntDNA strand, ntDNA(-37) (Figure 4A-B, gold) is located adjacent to ntDNA(-30), indicating the upstream path of double-stranded DNA leading away from TBP. The position of residue S71 of TBP (Figure 4D, purple) is located between the credible volumes of ntDNA positions (-24) and (-30) and in proximity to the RNAP wall and subunit Rpo12 and is consistent with crystal structures of TBP in complex with DNA. The credible volume of TFB-G262 (Figure 4D, olive) is located further away from the protrusion domain than TBP and positioned closer to RNAP subunit Rpo12. For the localisation of TFE we used the previously published smFRET data in our global analysis. The global NPS localisation for the two residues of TFE yields very similar positions than those determined previously¹⁸ but the credible volumes are smaller due to the increased accuracy of the global NPS calculation (Figure 4C, yellow and green).

This location of the TATA box in the archaeal OC is distinct from that previously determined in our group using NPS for a minimal eukaryotic open complex²⁷. In these studies a different promoter DNA sequence had been used (together with endogenous yeast Pol II and recombinant yeast transcription factors) and thus the question arises whether the particular conformation in an open complex depends on the respective sequence³⁶. We performed control measurements with a different DNA scaffold and concluded that the observed conformation is independent of the underlying promoter sequence, and as such our structural conclusions about the archaeal OC have general value (Supplementary Fig. 8A and Supplementary Methods), and that the differences compared to the earlier studies are due to the difference in the OC structure between yeast and archaea.

Model of the complete archaeal open promoter complex

To build a model of the complete archaeal RNAP OC, we started with the RNAP structure from S. shibatae (PDB 2WAQ^37,38) and used the calculated probability densities of the antenna dye attachment points on the ntDNA strand, TBP, TFB and TFE to position these elements and the template DNA onto this RNAP structure (Figure 5). In order to arrive at a unique structural model, we made some structural assumptions such as the size of the melted region, or the point of melting and re-annealing, all based on published data (see Methods for details).

Figure 5. Model of the complete archaeal OC complex.

Side, Front and Top view an alternate view obtained from the top view by a 90° rotation of the model of the open complex. DNA template and non-template strand are in blue and cyan, respectively. The transcription factors TBP, TFB and TFE are in purple, green and yellow, respectively. The ntDNA positions used for building the model are colour coded according to the NPS densities in Figures 3 and 4, namely ntDNA(-1), ntDNA(+7), ntDNA(-1), ntDNA(-5), ntDNA(-7), ntDNA(-10), ntDNA(-12), ntDNA(-14), ntDNA(-18), ntDNA(-24), ntDNA(-30) and ntDNA(-37). The X-ray structure of the archaeal polymerase of S. shibatae (PDB: 2WAQ³⁷) is represented as dark grey ribbon. See also Supplementary Fig. 8.

To estimate how well our new model fits the NPS probability densities we calculated the accessible volume of the antenna dyes using the coordinates from our model of the archaeal OC. We found that in all but one case, the accessible volumes overlap with the corresponding NPS probability densities, showing that the built model is accurate (for more details see Methods).

Discussion

The smFRET and global NPS data presented here reveal the complete architecture of the open promoter complex in archaea including the paths of the non-template and template DNA strands, and the location of the three transcription initiation factors TBP, TFB and TFE. The resulting model provides a framework for understanding the molecular mechanisms of transcription initiation in the archaea, as well as allowing a comparison to the mechanism in the eukaryotic OC, and providing insights into the evolution of the transcription machinery following the divergence of the archaeal and eukaryotic lineages.

The formation of complete archaeal OCs was strictly dependent on TBP and TFB, and the overall architecture and in particular the path of the DNA was independent of its sequence. The DNA strands are melted and the template strand has been loaded into the active centre cleft directly comparable to eukaryotic OCs^13,30.

In our model of the archaeal OC, the double-stranded downstream DNA enters the archaeal polymerase at a similar angle to that previously shown for structures in yeast, bacteria and archaea containing short duplex DNA^30,38,39. In this position the downstream DNA can be stabilised by the proximal lysine-rich region of the jaw domain of Rpo1" (residues 189-239) whereas the corresponding eukaryotic jaw domain would need a rotation inwards to superimpose with the archaeal counterpart, a movement hindered by the eukaryotic subunit Rpb9 and the N-terminal domain of the eukaryotic subunit Rpb5 that are not conserved in the archaeal RNAP³⁸. The archaeal subunit Rpo5, which lacks the N-terminal domain of Rpb5, is required for the formation of stable open complexes⁴⁰ and has been shown to photo-crosslink to the downstream DNA⁴¹ which is perfectly consistent with our model.

Previously, we observed a dynamic switching of the downstream DNA into and out of the cleft in single molecule studies of a minimal eukaryotic OC²⁷. Cryo-EM data shows that TFIIF appears to facilitate this transition in eukaryotes¹³. In contrast, in the archaeal OC, the smFRET data for ntDNA(+7) in the downstream double-stranded region showed no evidence of a dynamic movement of the DNA. Presumably, the transcription factor TFE, which is known to stabilise the DNA in the OC, renders the downstream DNA in a stable conformation. Note that the studies on the minimal eukaryotic OC were performed in absence of TFIIE. Thus it would be interesting to see whether the eukaryotic TFIIE has a similar function and would lead to a stabilisation of the loaded state. Functional transcription assays using the Pol II system demonstrate that TFIIE stimulates open complex formation and transcription on negatively supercoiled templates independently of TFIIH, which is in good agreement with our data and validates the use of archaeal transcription systems as bona fide model systems for eukaryotic Pol II^42,43.

In our model the downstream DNA strands are separated at register ntDNA(+2) in proximity to fork loop 2 (subunit Rpo2", residues 436-445), and close to the highly conserved residue R446, which corresponds to Rpb2 residue R504 in Pol II. A point mutation of this arginine in the Pyrococcus RNAP (R445) to alanine leads to elongation deficiency in vitro²¹. Our model is also in agreement with the additional function of fork loop 2 of sterically blocking the duplex binding of the DNA and thus preventing re-association of the separated strands⁴⁴. Therefore, at these positions, both template and non-template strand conformations are very similar to those observed in the EC²⁶.

Further upstream, between registers ntDNA(-1) to (-7), the non-template strand runs adjacent to the fork loop 1 element (residues 404-410) and lobe domain of Rpo2", the rudder of the Rpo1' clamp (residues 278-297) and the linker region of TFB. Here, the path of the ntDNA of the archaeal PIC diverges from that in the eukaryotic EC, as the TFB linker region is situated at a position where it would clash with the ntDNA strand of the EC. NtDNA registers (-1) to (-3) pass close to the rudder, fork loop 1 and the TFB linker, whereas registers (-4) to (-7) are closer to the lobe. These protein interactions with the middle of the transcription bubble are highly likely to play role in bubble melting and/or maintenance. Previous studies have shown the essential role of the Rpo1' rudder in DNA strand separation where mutants lacking this loop could not separate or maintain melted DNA²¹. Additionally, yeast nuclear extracts containing temperature sensitive TFIIB were transcriptionally inactive in vitro and rescued only by adding recombinant wtTFIIB and not with TFIIB containing mutations in the linker region⁵. Similarly, in vitro transcription assays with P. furiosus RNAP and its initiation factors showed that point mutations or deletions in the TFB linker region allowed the formation of PICs but were inhibited for transcription⁵ and subsequent footprinting studies showed that these PICs were incapable of opening the promoter DNA. Thus, the B-linker region and Rpo1' rudder are essential for promoter opening and open complex stabilisation. Our new model of the archaeal OC gives a mechanistic reason for these observations since the single-stranded ntDNA is positioned adjacent to all of these elements (Figure 6 A-B) and their interaction is likely to influence the stability and formation of the transcription bubble.

Figure 6. Open complex model has implications for the melting of DNA in the CC to OC transition.

(A) Side view (slightly rotated for a clearer view) of the archaeal polymerase of S. shibatae (PDB: 2WAQ)³⁷ (dark grey) together with components of the open complex model, namely TFB, TBP as well as the template and non-template strand.

(B) Detail of the yellow rectangular region in (A) showing the clamp coiled coil domain of subunit Rpb1' (light grey) and the point mutation introduced into the B-linker helix in a previous study (red)⁵. The rudder loop of Rpb1' is shown in pink.

(C) - (D) Detailed views of the archaeal model compared to the eukaryotic open complex model. The eukaryotic open complex model⁵ is displayed as superposition to the archaeal model and the archaeal RNAP. The eukaryotic transcription factors TBP and TFIIB are shown in light pink and light green, respectively. The DNA is shown in light blue and the polymerase is represented as semi-transparent surface. For the OC model the same colour coding is used as in Figure 5.

As a consequence of the displacement of the ntDNA to the outside of the cleft relative to its path in the EC, its path comes close to the edge of the clamp core region and in particular to the clamp helix-coil-helix motif, at register nt(-12) where the upstream end of the bubble lies. Previously, we have localised the binding position of the winged-helix domain of TFE¹⁸ to the tip of the helix-coil-helix motif and the global NPS analysis presented in this paper also confirms this observation. Moreover, cryo-EM as well as crosslinking data show that eukaryotic TFIIE contacts the RNAP at a similar binding site^9,13,14. Our model describes an interaction at this point between ntDNA at the upstream end of the bubble, TFE and the RNAP helix-coil-helix motif. These interactions are likely stabilising the upstream end of the transcription bubble to prevent its collapse.

The point of DNA re-annealing at register ntDNA(-12) lies above a tunnel formed by the N-terminal domain of the TFB-core, the TFB-linker helix region, the rudder, the protrusion and TFE. A comparison to the position of the upstream DNA in the elongation complex²⁶ shows that a rearrangement of the complete upstream double-stranded region, including a release of TFB core from the RNAP surface and movement of the upstream DNA to a position in between Rpo1' helix α8 (residues 235-251) and Rpo2" helix α11 (residues 349-373), is required during the initiation to elongation transition, presumably leading to a release of the transcription initiation factors and bubble collapse.

Interestingly, the probability density for position ntDNA(-24), together with those for ntDNA(-30), ntDNA(-37), TBP-S71 and TFB-G262 define the pathway of the DNA strand around the TATA box in close proximity to the surface of the polymerase. Previous Far-western blotting studies showed the strongest protein-protein interactions of TBP and TFB with subunits Rpo12, Rpo10 and Rpo2"⁴⁵, which are all very close to the positions of the transcription factors in this OC model. Compared to the eukaryotic system, we find the position of the TFB core domain in the archaeal OC has shifted and the position of TBP has changed substantially by ≈ 45 Å (Figure 6 C-D and supplementary movie 1)^5,13. Whereas the N-terminal cyclin fold of the TFB-core is only slightly tilted, the C-terminal cyclin fold of the TFB-core is shifted and is localised closer to the DNA strand but still remains in proximity to the wall of the polymerase. Thereby, the helix-turn-helix motif consisting of helices TFB-H 4' and TFB-H 5' (residues K1265 to K1292) is facing the non-template strand at registers ntDNA(-31) to ntDNA(-36) which form the purine-rich B-recognition element BRE.

Mechanistically, transcription initiation in archaea is ancestral and streamlined compared to the eukaryotic Pol II system. Archaeal genomes do not encode homologues of TFIIA, TFIIF, TBP-associated factors (TAFs), and TFIIH. In particular the latter two could make important contributions to the open complex formation since the TAFs make contact with the promoter DNA around the transcription start site, and TFIIH because of the ATP-dependent helicase/translocase activity is crucial for DNA melting on the majority of promoters tested in vitro, and probably all transcription initiation in vivo¹¹. However, using negatively supercoiled DNA and strong promoter templates TBP and TFIIB suffice for initiation of eukaryotic Pol II³, which demonstrates that the same ancestral mechanisms are able to facilitate open complex formation in eukaryotes and archaea. Why are additional factors required by Pol II provided that the basic mechanisms are conserved? Our model of the complete archaeal open complex provides a structural hypothesis for this apparent ease of DNA melting in archaea (Figure 7). Since the archaeal RNAP pulls the promoter-bound factors TBP and TFB much closer to its surface than Pol II (Figure 6C-D) and the downstream promoter DNA is bound between the RNAP jaws, this topology likely induces a torsional strain in the DNA that lowers the local melting temperature of the promoter DNA. Interactions between the tDNA and residues on the inside of the DNA binding channel subsequently facilitate a swift loading of the template strand DNA into the RNAP active site. But why has this process evolved to become ATP energy dependent in the Pol II system while remaining spontaneous in archaea? Neither Pol I, nor Pol III, nor the bacterial sigma70 holo-RNAP requires energy for open complex formation, which indicates that Pol II could be exceptional in this regard. Since the complexity of the Pol II transcriptome is higher than of any other RNAP system mentioned above, the energy dependence could reflect an additional layer of regulation of Pol II transcription. Support for this concept is provided by a recent report about the global regulation of open complex formation in naïve lymphocytes⁴⁶, which upon activation undergo a transcriptome amplification that is regulated by TFIIH.

Figure 7. Mechanisms of the closed (CC) to open complex (OC) transition in archaea and eukaryotes.

During open complex formation the double-stranded promoter DNA is melted and the template DNA strand (tDNA) is loaded into the active site while the nontemplate strand (ntDNA) interacts with the RNAP clamp, and with TFE and TFIIE in archaea and eukaryotes, respectively (highlighted in orange). Concomitantly the entire complex – RNAP and initiation factors – undergoes large scale conformational changes. In archaea OC formation occurs spontaneously and is possibly driven by the torsional strain in the promoter DNA induced by the interaction network between initiation factors, RNAP and the promoter DNA elements. While the upstream BRE and TATA promoter elements are anchored to the PIC by TFIIB (green) and TBP (magenta), the downstream DNA interacts with the RNAP jaws. In the Pol II system OC formation is largely driven by the ATP hydrolysis-dependent activities of the TFIIH subunit ssl2 (red) which also induces a torsional strain by translocating the downstream promoter DNA in the upstream direction into the active site of RNAP.

In conclusion, the presented data provides a structural model for the organisation of the archaeal OC. Given this model, a mechanism by which DNA melting could occur without transcription factor TFIIH becomes apparent.

Methods

Recombinant Protein preparation and labelling

RNAP subunits from the hyperthermophilic archaeal model system M. jannaschii were expressed in recombinant form in E.coli and purified¹⁶. For the smFRET experiments five differently labelled RNAPs were prepared. Therefore, either single cysteine residues were introduced into the RNAP at position K11 of Rpo5 or positions V49 and S65 of Rpo7 and the subunits were purified and labelled with the dye Alexa647¹⁷. Or an unnatural amino acid (p-Azido-L-phenylalanine) was introduced at position G257 of Rpo1' and position Q373 of Rpo2" and labelled with the dye DyLight650 by Staudinger ligation⁴⁷. The fluorescently labelled subunits were directly introduced into RNAP reconstitution reactions following known protocols¹⁶.

Unlabelled transcription factors TBP, TFB, and TFE were produced as described previously^22,48. Labelled TFE was prepared as described previously with either the dye Cy3B attached to position G44 or the dye DyLight550 attached to position G133¹⁸.

Preparation of fluorescently labelled TBP derivative

TBP was labelled with an Alexa647 or Alexa555 fluorophore via a cysteine-maleimide coupling strategy. In order to introduce a unique cysteine residue the native cysteines at positions C48 and C67 that are buried inside the protein have been substituted by serine residues and a single cysteine residue has been introduced at position S71. The mutations have been introduced into the TBP gene using either the QuikChange II site-directed mutagenesis kit (Agilent) or the SOE (splice by overlap extension) PCR strategy. Recombinant TBP-S71C was expressed from a pET21a(+) vector in BL21(DE3)/Rosetta cells and expression of TBP was induced in exponentially growing cultures with 1 mM IPTG at an optical density of ~0.6 to 0.8 in rich medium for 4 hr at 37°C. Bacterial cells were harvested, resuspended and extracted in P300 buffer (200 mM Tris/Acetate pH 7.9, 100 mM MgAc, 0.1 mM ZnSO₄, 300 mM KAc, 10 % Glycerine). Cells were lysed using sonification. Recombinant and heat stable MjTBP could further be pre-purified using a heat denaturation step (65 °C for 20 min). The heat stable fraction contained MjTBP and the protein was precipitated with saturating amounts of ammonium sulfate. After pelleting the precipitated protein fraction, the pellet was resuspended in 5 mL P300 with 0.05 % beta-ME and further purified by size exclusion chromatography (HiPrep-Sephacryl, S100 16/60, GE Healthcare). MjTBP containing fractions were combined, subsequently further purified and β-mercaptoethanol was removed by ion exchange chromatography (MonoQ 4.6/100 PE, GE Healthcare) using a gradient from 100 to 1000 mM potassium acetate. The labelling reaction was carried out using a 10-fold molar excess of dye at 4 °C for 16 hours. Labelled protein was separated from excessive free dye using a NAP-5-column and P100 buffer (200 mM Tris/Acetate pH 7.9, 100 mM MgAc, 0.1 mM ZnSO₄, 100 mM KAc, 10 % Glycerine) with 0.05 % beta-ME.

Preparation of fluorescently labelled TFB derivative

For the production of fluorescently labelled TFB variants a nonsense-suppressor strategy was chosen, which allows the specific labelling of the protein via a unique unnatural amino acid (p-Azido-L-phenylalanine)⁴⁹. An amber mutation (TAG) was introduced at position G262 into the TFB gene using the QuikChange II site-directed mutagenesis kit (Agilent) The mutated protein was expressed from a pET21a(+) plasmid which allowed purification of the full-length protein via a C-terminal His-tag. The recombinant protein was produced in BL21/DE3 cells that additionally carried the arabinose-inducible pEvolv-pAzF plasmid encoding multiple copies of an amber-suppressor tRNA (tRNA_CUA) and an engineered tyrosyl-tRNA synthetase⁵⁰. Bacterial cultures were grown in rich medium containing 100 μg/ mL ampicillin and 25 μg/mL chloramphenicol. 1 mM p-Azido-L-phenylalanine (Chem-Impex International Inc.) and 0.02 % arabinose were added to the culture at an optical density of 0.3-0.4. TFB expression was induced with 1 mM IPTG at an optical density of 0.5-0.6 and cells were harvested after 3h. After harvesting the cells by centrifugation (5000xg, 15 min) the cells were resuspended in N500 buffer (200 mM Tris/Acetate pH 7.9, 100 mM MgAc, 0.1 mM ZnSO₄, 500 mM NaCl, 10 % Glycerine) containing 0.5 % Triton. Cells were lysed by sonification and the soluble protein fraction was extracted. The cell lysate was removed from cell debris and unsoluble fractions by a centrifugation step (15000xg, 30 min) and the supernatant was further purified by affinity chromatography (HisTrap FF 1 mL, GE Healthcare). Following labelling with 10-fold molar excess of either DyLight550 or DyLight650 via Staudinger ligation⁴⁷ overnight at 4 °C, the excess of free dye was removed by affinity chromatography (HisTrap FF 1 mL, GE Healthcare).

KMnO₄ Footprinting

The DNA template encoding the SSV T6 promoter was prepared by annealing 5’, ³²P-labelled non-template strand (5’- GATTGATAGAGTAAAGTTTAAATACTTATATAGAT AGAGTATAGATAGAGGGTTCAAAAAATGGTT-3’) and unlabelled template strand (5’-AACCATTTTTTGAACCCTCCGCTTATACTCTATCTATATAAGTATTTAAACTTTACTCTATCTATC-3’). For the footprinting reactions the components were combined in 23 μL reactions containing 1 x HNME buffer (40 mM HEPES [pH 7.3], 250 mM NaCl, 2.5 mM MgCl, 0.02 mM EDTA, 1 % glycerol and 2 mM DTT), 8.3 nM template DNA, 0.6 μM RNAP, 8.7 μM TBP, 0.5 μM TFB and 740 nM TFE. The reaction was incubated at 65 °C for 15 min followed by 2 min incubation with 2 μL KMnO₄ (at 4, 8 or 16 mM) and stopped with 1.5 μL 14 M β-ME. Protein was digested by addition of 0.25 % SDS and 1 mg/mL proteinase K and incubation at 65 °C for 1 hr. DNA was ethanol precipitated prior to treatment with 5 % piperidine at 90 °C for 30 min, followed by one round of chloroform extraction and subsequent ethanol precipitation. To prepare the A+G ladder the DNA was treated with formic acid for 5 min prior to DNA precipitation and piperidine treatment as described above. DNA separated on 10 % urea PAGE, exposed to a phosphor storage screen and visualised on a Typhoon FLA 9500 bioimager.

Archaeal OC preparation for single-molecule FRET experiments

The OCs were assembled freshly before each smFRET experiment by adding 1μL each of nucleic acid scaffold (2 μM), TBP (10 μM), TFB (10 μM), RNAP ΔRpo4/7 (2 μM) and Rpo4/7 (10 μM) to 10 μL HNME buffer. The mixture was incubated at 60 °C for 10 min. Heparin (final concentration 0.5 mg/mL) was added to reduce non-specific binding of the RNAP to nucleic acids. Unbound transcription factors and nucleic acids were removed using Amicon Ultra centrifugal filters (Millipore) by washing two times with 450μL HNME buffer. All smFRET experiments were done in presence of TFE (12 μM), which was added to the purified complexes and incubated for 10 min at 60 °C. The complexes were then diluted 1000-fold in HNME buffer and loaded into the sample chamber of the TIRF microscope. For surface immobilisation of the complexes the ntDNA strand had Biotin attached at the 5’-end via a C6-amino linker.

The DNA single-strands were purchased from IBA (Göttingen, Germany) and annealed as described before²⁵. The viral SSV T6 promoter DNA⁵¹ was used for all the smFRET experiments as it is known to form very stable PICs in promoter-directed transcription in vitro¹⁶. Our promoter DNA constructs consists of a 66 nucleotides long double-stranded DNA with template and non-template DNA strands containing a 4 nucleotide heteroduplex region around the transcription start site (-3 to +1) that stabilised the PIC by forming the open complex (Figure 1).

For the determination of the course of the ntDNA within the OC, the non-template strand was purchased with either Cy3B at position (+7), (-1), (-5), (-7), (-10), (-12), (-14), (-18), (-30) and (-37) or 6-TAMRA at position (-24) (Figure 1B). For the question of the conformation of the tDNA in the OC, the template strand was purchased with 6-TAMRA at position (+3) and Cy3B at position (-9).

Experimental setup for smFRET, data collection and analysis

All smFRET experiments were performed on a custom-built prism-based total internal reflection fluorescence microscope (TIRFM) described previously¹⁸. Briefly, a frequency-doubled Nd:YAG laser (532 nm, Spectra-Physics) was used for the excitation of donor molecules and a diode laser (643 nm, Toptica) for the direct excitation of the acceptor molecules. Fluorescence intensity was collected through a water immersion objective (Plan Apo 60X, NA 1,2, Nikon) and directed to an EMCCD camera (iXon, Andor). OCs were immobilised onto the surface of a microfluidic chamber surface via PEGBiotin-Neutravidin-Biotin as described previously²⁵. The acquired data was analysed using custom-written MATLAB software. We used a fully automated routine to find FRET pairs, calculating and subtracting the local background and computing fluorescence trajectories²⁵. The correction factors were determined individually for every FRET pair. The resulting histograms were computed for every time point (frame-wise histogram). Data from at least three individual smFRET measurements were used for each pair of labelling sites. The FRET efficiencies from all molecules of all measurements were plotted in histograms. The peaks were fitted with one (or two Gaussian) function(s) to extract the mean FRET efficiencies (Tables S1, S4 and S5). A standard deviation of 2 % for the FRET efficiencies was included into the calculation. These results were then used for further analysis with NPS²⁴ or global NPS²⁸, respectively as indicated.

Determination of the probability densities in the archaeal OC using NPS

The X-ray structure of the archaeal RNA polymerase of S. shibatae (pdb file: 2WAQ³⁷) was used as a reference frame for the position calculation. Moreover, the volume occupied in the crystal structure was used as a restriction for the possible positions of the dye molecules. We assumed zero probability density within an already occupied volume, which was the volume of the protein shrunk by 5 Å to account for uncertainties in the x-ray structure, and equal probability density elsewhere in order to calculate the ADM prior.

The global NPS method²⁸, software freely available at http://www.uni-ulm.de/nawi/nawi-biophys/software.html) was then applied using the available X-ray structures, the measured FRET efficiencies and Bayesian parameter estimation. As a result we obtained the three dimensional probability density function for the positions (+7), (-1), (-5), (-7), (-10), (-12), (-14), (-18), (-24), (-30) and (-37) on the ntDNA strand, as well as the positions of residue S71 of TBP and residue G262 of TFB. From this we calculated the smallest volumes that enclose a certain probability, so-called credible volumes. The surface of the credible volumes was displayed by using the interactive visualisation program UCSF Chimera which was also used for displaying all structural data⁵². All credible volumes shown in the paper are calculated at 68 % probability. For more details see Methods.

Determination of the isotropic Förster radius and anisotropies

For each donor-acceptor pair the isotropic Förster radius R₀^iso was determined using standard procedures ⁵³. First, the quantum yield of the donor sample was determined using Rhodamine 101 dissolved in ethanol as a standard (QY = 91,5 %)⁵⁴ (Supplementary Table 2).

The ntDNA positions (+7), (-1), (-5), (-7), (-10), (-12), (-14), (-18), (-24), (-30) and (-37) were labelled either with the donor dye Cy3B or 6-TAMRA, residue S71 of TBP was labelled with the dye Alexa555 and residue G262 of TFB was labelled with the dye DyLight550.

Second, overlap integrals were calculated from recorded donor emission spectra (528 to 700 nm with an excitation wavelength of 523 nm) and acceptor absorption spectra (400 to 700 nm). Together with the refractive index (n = 1.35) and the orientation factor (κ² = 2/3) the isotropic Förster radii R₀^iso were determined for all the different donors and Alexa647 as acceptor (Supplementary Table 2).

In order to account for uncertainties in the Förster distance due to orientation effects we then measured the steady state fluorescence anisotropies of the donor and acceptor dyes for all attachment sites using a steady state fluorescence spectrometer (Edinburgh Instruments F900) (Supplementary Table 2). Both, isotropic Förster distances as well as fluorescence anisotropies were used as prior information in the global NPS analysis²⁸.

Uncertainty in the position of dye molecules attached to known positions

Satellite dye molecules (SDMs) were attached to known positions within the archaeal polymerase using flexible linkers. While the attachment point is known from the x-ray structure of the archaeal polymerase of S. shibatae (pdb file: 2WAQ³⁷), the precise location of the dye molecule is not. For the NPS analysis we therefore calculated the volume that is sterically accessible to the dye molecules, given the point of attachment, size of the dye molecule and the linker length²⁴. To this end, the SDMs were approximated by a sphere of diameter d_dye and linked to the protein complexes by flexible linkers of dimensions L_linker and d_linker (Supplementary Table 3). We assume each SDM position within this accessible volume equally probable (Supplementary Figure 4).

Calculation of model based prior volumes

For the dyes attached to the double-stranded DNA region, the coordinates of the C7 atom of the base were used as attachment point and the linker length corresponded to 12 C-atoms. In case of the single-stranded ntDNA region, the base orientations were left out of the model, since in a single strand of nucleic acids base stacking energies are small and as a result any base is relatively free to rotate about the backbone. Therefore, a 17 C-atom linker and an attachment point on the backbone C1' atom of the DNA was used for single-stranded regions. The sequence alignment of the proteins TBP and TFB from M. jannaschii that were used in all our experiments with the corresponding proteins from P. woesei contained in the crystal structure of the TBP/TFB/DNA sub-complex used for the modelling (pdb file: 1D3U⁵¹) resulted in the definition of the analogous residues S72 for TBP and E1223 for TFB in the model. Exemplary Figures of the comparisons can be found in Supplementary Figure 8B-E.

Modelling

A number of structural assumptions were made to arrive at a unique model: (1) The extent of the single stranded transcription bubble is between positions (-11) and (+1), corresponding to the permanganate footprinting results and published literature^5,32,33. (2) The template strand position (+1) would be positioned at the active site for base pairing with the first NTP of the RNA transcript. (3) The DNA conforms to a B-form duplex outside of the melted region, and the downstream duplex occupies a similar position to that of the eukaryotic/bacterial EC^44,55, the OC-mimic of Pol II³⁰ and the archaeal RNAP-DNA binary complex³⁸. (4) The structure of archaeal TBP/TATA/TFB from P. woesei (PDB 1D3U⁵¹) containing the C-terminal cyclin core of TFB and a bent TATA box DNA fragment was used as the template for TBP-TATA-TFB in this OC model and would not change in conformation when bound to RNAP. (5) The path of the TFB N-terminal regions within the RNAP cleft would follow the same approximate path as observed in the structure of eukaryotic Pol II in complex with TFIIB (PDB 4BBR⁷). (6) The position of the N- and C-terminal domains of TFE would also be consistent with the cryo-EM density observed for the eukaryotic Pol II PIC complex containing TFIIE¹³.

Probability densities were visualised in Coot⁵⁶ and USCF Chimera⁵². Template models assembled into the complete OC model were based on PDB entries 2WAQ, 1D3U, 4BBR, 1Q1H and 1VD4. Models were manipulated to fit the probability densities using the same programs as for visualisation, and geometry regularised using phenix.refine⁵⁷. Model coordinates are given in a supplementary file.

To estimate how well the model fits the NPS densities, we calculated the accessible volume priors of the dyes attached to the respective positions in the model (Methods) and compared them to our resulting probability densities. It should be noted that a perfect overlap is not expected given the nature of the model based prior and the NPS posterior. The prior volume encompasses the complete volume the dye molecule could be sitting, given its size and the length of the linker and the position of the anchor point defined by the model of the OC. Therefore, its size is simply a measure of the uncertainty before the measurement. The posterior volume, in contrast, represents the probability for the dye position and its size is a measure of the uncertainty after the measurement. As long as there is overlap between prior and posterior the model is in accordance with the data.

The model and the derived accessible volumes fit the obtained credible volumes from the NPS calculation drawn at 68 % credibility in all but one case (Supplementary Fig. 8B-C), the exception being the ntDNA(-14) position, where the calculated accessible volume is not overlapping with the computed NPS volume drawn at 68 % credibility. The clamp coiled-coil region of the RNA polymerase is situated exactly in between the accessible volume and the NPS credible volume of ntDNA(-14). If the NPS credible volume of ntDNA(-14) is displayed at 90 % confidence level, the prior and posterior overlap (Supplementary Fig. 8D-E). One should note that TFE also binds to the clamp coiled-coil region as determined by previous cryo-EM, cross-linking and NPS studies^9,13,18. However, we did not assign a particular volume for TFE in the NPS analysis. Thus, it is quite likely that large parts of the NPS determined credible volumes for the dye attached to ntDNA(-14) are in fact excluded by TFE, preventing overlap between the model and the NPS result for ntDNA(-14).