For transcription initiation, Pol II assembles with the basal transcription factors (TF) IIB, TFIID (or its subunit TBP), TFIIE, TFIIF, and TFIIH^{1, 2, 3, 4} on double-stranded promoter DNA to form the closed complex (CC). Upon DNA opening, the template strand slips into the Pol II active centre, and a DNA ‘bubble’ forms ~20–30 base pairs (bp) downstream of the TATA box⁵, leading to the open complex (OC). Efficient DNA opening requires TFIIE and TFIIH^{6, 7, 8}, but these factors are not required for low levels of transcription^{9, 10} and TFIIE alone can open certain promoters^{11, 12}. Subsequent RNA synthesis results in the initially transcribing complex (ITC), which is then converted into an elongation complex for processive RNA synthesis.

The three-dimensional architecture of Pol II initiation complex intermediates was studied in yeast^{13, 14, 15, 16, 17, 18, 19} and human²⁰ systems. These studies revealed that the promoter assembly containing TBP and TFIIB resides over the Pol II wall and positions DNA above the polymerase cleft and along the clamp. TFIIE and TFIIF bind to opposite sides of Pol II and flank promoter DNA. Owing to the limited resolution of available structural studies, many questions regarding the molecular basis of initiation remain, including the mechanisms of DNA opening and template-strand loading into the active centre. A differing model of the yeast initiation complex¹⁷ was revised recently²¹, and now agrees, on a topological level, with other studies and results reported here.

Here we report cryo-EM structures of CC and OC assemblies from the yeast Saccharomyces cerevisiae at resolutions of 8.8 Å and 3.6 Å, respectively. The structures contain all of the basal transcription factors except TFIIH, which is currently not available at a quality required for high-resolution analysis. We show that DNA opening can occur in the absence of TFIIH, and provide mechanistic insights into DNA opening and template-strand loading. Our results also unveil the exact location and intricate structure of basal factors and their induced folding interactions with each other, with Pol II, and with promoter DNA. The data also demonstrate the high structural conservation between yeast and human initiation systems.

We extended our previous cryo-EM analysis of a Pol II core initiation complex containing TBP, TFIIB, and TFIIF¹⁹ by adding TFIIA and TFIIE (Methods). Formation of a stable and stoichiometric complex required the presence of core Mediator, which showed low occupancy and high flexibility under cryo-EM conditions, as observed previously¹⁹, and was excluded from further analysis (Methods, Extended Data Fig. 1a–c). We acquired 257,259 cryo-EM single particle images using a K2 direct electron detector (Methods, Extended Data Fig. 1b, c). Unsupervised particle sorting led to an OC structure at an overall resolution of 3.6 Å, revealing the Pol II core at up to 3.1 Å (OC1), compared to 7.8 Å for our core ITC structure¹⁹ (Fig. 1a, b, Extended Data Figs 1d–g). Further particle sorting revealed improved density for TFIIB and TFIIF at around 4 Å resolution (OC2 and OC4), and for TFIIE at 4.4 Å resolution (OC3, Methods, Extended Data Figs 1i, k, l). The final structure mainly consists of atomic models (90%) and contains backbone models for parts of the basal factors (Extended Data Fig. 1f–l).

a, Domain organization of yeast basal transcription factors TBP, TFIIA, TFIIB, TFIIE, and TFIIF. Solid and dashed black bars indicate protein regions that are present in the OC structure as atomic and backbone models, respectively. Colour code used throughout. b, Two views⁵⁰ of the yeast OC structure. Pol II is in silver. DNA template and non-template strands are in dark blue and cyan, respectively. On the right, Pol II is shown as a surface representation, all other proteins are shown as ribbon models. c, Protein–DNA contacts. Promoter DNA nucleotides are depicted with solid, shaded, and empty circles when they were included in the structure, excluded owing to weak density, or excluded owing to a lack of density, respectively. Solid and dashed lines indicate observed and putative protein interactions, respectively. A magenta dashed line indicates the contact between closed DNA and the TFIIE E-wing. The register of promoter DNA is given for analogous yeast (black) and human (grey) positions with respect to the transcription start site (TSS, +1). TATA box is indicated by a grey box.

• Full size image (390 KB)
• Download PowerPoint slide (892 KB)

The structure reveals upstream DNA above the Pol II wall and downstream DNA in the active centre cleft¹⁹. We observed only fragmented density for single strands within the DNA bubble (Fig. 1b, c, Extended Data Fig. 1m), and no density for RNA, which was not retained during cryo-EM sample preparation (not shown). The bubble contains 15 mismatches, begins at the natural distance of ~20 bp downstream of the TATA box⁵, and extends further downstream than when it initially forms, resembling the situation during transcription start-site scanning. For the basal factors, regions that are essential for cell viability are generally observed, whereas non-essential, non-conserved regions^{16, 22, 23, 24, 25, 26} are often mobile (Fig. 1a, b, Extended Data Fig. 1 j–l). TFIIA is located near upstream DNA and TBP as expected²³ (Fig. 1b, Extended Data Fig. 1j). For TFIIB, the B-ribbon and B-core domains are well defined, including the newly modelled carboxy-terminal cyclin domain from yeast (Fig. 2), whereas the B-linker shows weak density, and the B-reader^{15, 19} is mobile (Extended Data Fig. 1i).

Details of the upstream DNA assembly viewed from the side⁵⁰. Highlighted are the locations of the Pol II wall (navy blue), protrusion (orange), TBP (red), TFIIB (green), TFIIF arm (purple), Tfg2 linker (dark magenta) and winged helix (light purple), TFIIE Tfa2 WH1 (light salmon), DNA (template, blue; non-template, cyan), and the active site (magenta sphere). TFIIA is transparent (light yellow). Cryo-EM densities for newly modelled regions of TFIIB, TFIIF, and TFIIE are superimposed on their structural models. Interactions with the Pol II protrusion and upstream edge of the DNA bubble are indicated. TBP contacts a density assigned to the Tfg2 C-terminal region, consistent with interaction of their human counterparts²⁹ (Extended Data Fig. 2f).

• Full size image (639 KB)
• Download PowerPoint slide (662 KB)

TFIIF adopts an intricate fold within the OC. Its dimerization module and charged helix are located on the Pol II lobe domain as in the ITC^{18, 19} (Extended Data Fig. 2). The ‘arm’ in the large TFIIF subunit Tfg1 (human RAP74) adds a β-strand to the Pol II protrusion, and projects into the cleft, where it may stabilize the DNA bubble^{18, 19, 20} (Fig. 2). The linker in TFIIF subunit Tfg2 (human RAP30) emanates from the dimerization module, and winds along the base of the protrusion, where it binds a hydrophobic pocket. The Tfg2 linker continues between the protrusion and the TFIIB cyclin domains, and connects to the Tfg2 C-terminal winged-helix (WH) domain on top of the cleft (Fig. 2, Extended Data Fig. 2f). The Tfg2 linker stabilizes TFIIB on the wall of Pol II^{27, 28}. The yeast-specific amino-terminal region of Tfg1 binds near the Pol II external 1 domain, according to a separate crystallographic analysis (Extended Data Fig. 2d, e, Extended Data Fig. 9c).

TFIIE is located between the clamp and the Rpb4–Rpb7 stalk (Fig. 3a, b, Methods, and Extended Data Fig. 3), consistent with previous topological placement of TFIIE^{16, 20, 29, 30, 31} and its archaeal counterpart³². The TFIIE structure differs from a recent model obtained at 6 Å resolution²¹ (see Extended Data Fig. 1e legend for details, Extended Data Fig. 3, and refs 16, 17). The large TFIIE subunit Tfa1 (human TFIIEα) contains an extended winged helix (‘eWH’) domain³³ and a zinc ribbon domain³⁴ (‘E-ribbon’) that are connected by α-helices (called here ‘E-linker’) (Fig. 3b). The eWH domain uses its helix α3 to contact both the tip of the Pol II clamp helices and the DNA backbone at positions −13/−14 upstream of the transcription start site (TSS, position +1) (Fig. 1c, 3a). The E-ribbon binds between the clamp, the Rpb7 oligonucleotide-binding (OB) domain, and the B-ribbon (Fig. 3a). The E-linker^{20, 21} and the mobile C-terminal domain of Tfa1 contact TFIIH^{7, 35}, which may alter TFIIE conformation. The small TFIIE subunit Tfa2 (human TFIIEβ) contains two WH domains (‘WH1’ (ref. 36), ‘WH2’), and two conserved α-helices (called here ‘E-tether’) that bind the E-linker (Fig. 3b). Consistent with the structure, the E-tether is essential for TFIIE function^{16, 37} and subunit dimerization (Extended Data Fig. 3a, refs 16, 38). The structure further indicates that TFIIE must be displaced or at least moved before the elongation factor Spt4/5 (human DSIF) can bind to polymerase^{32, 39}.

a, TFIIE interactions within the OC. Depicted are interactions of the TFIIE E-ribbon with the Pol II clamp, stalk subunit Rpb7, and the TFIIB B-ribbon, and interactions of the TFIIE eWH domain with the Pol II clamp helices and upstream DNA. The eWH E-wing lies close to the upstream DNA edge, similar to WH domains involved in DNA strand separation (Extended Data Fig. 3j). Colours as in Fig. 1, except for the Pol II stalk (Rpb4, dark red; Rpb7, dark blue). b, TFIIE domain architecture. The TFIIE variants used for functional assays are indicated as Cα spheres for point mutations, and with a black bracket for E-wing alterations (compare Extended Data Fig. 3g, h, j). Connectivity of the Tfa2 E-tether helices is uncertain. c, Selected TFIIE variants impair transcription from a HIS4 promoter (Methods, Extended Data Fig. 3g, h). TFIIE-depleted nuclear extract (NE) was reconstituted with recombinant TFIIE or TFIIE variants carrying mutations in the Tfa1 eWH (M1, Tfa1(N50E/K51E/T52E); M2, Tfa1(N50A/K51A/T52A); M3, Tfa1(P56A/A59E/R62E); M4, Tfa1(ΔE-wing); M5, Tfa1(poly-Ala E-wing) and the Tfa1 E-ribbon (M6, Tfa1(L134E/V137E/L140E); M7, Tfa1(L134A/V137A/L140A) (Extended Data Fig. 3g). RNA products were visualized by primer extension and the mean intensity and standard deviation (s.d.) from triplicate experiments are provided, relative (rel.) to the activity of wild-type TFIIE. An asterisk marks RNA products resulting from an alternative upstream transcription start site.

• Full size image (428 KB)
• Download PowerPoint slide (696 KB)

The OC structure thus reveals how TFIIF and TFIIE bind open promoter DNA from opposite sides of the Pol II cleft. First, TFIIF adopts an extended induced structure that allows it to retain the upstream DNA–TBP–TFIIB assembly on the wall and to bind the DNA bubble and downstream duplex in the cleft (Figs 1b, 2). Second, the TFIIF Tfg2 WH domain and the TFIIE Tfa2 WH1 domain contact each other above upstream DNA to encircle and retain DNA. Third, the eWH domain of TFIIE binds DNA in the region of initial DNA opening and its long β1–β2 hairpin³³ (called here the ‘E-wing’) projects to the upstream edge of the bubble (Fig. 3a), suggesting that the eWH domain stabilizes open DNA. Taken together, the highly modular and flexible basal factors TFIIF and TFIIE undergo substantial induced folding transitions to engage in multiple protein–DNA and protein–protein interactions to stabilize the OC.

Modelling of a closed DNA promoter onto the OC structure shows that closed DNA would clash with the TFIIE eWH domain. This suggests that the eWH domain adopts a different position before DNA opening. To investigate this, and to provide insights into the transition from the CC to the OC, we repeated structure determination with closed DNA instead of pre-opened DNA (Methods, Extended Data Fig. 4). Surprisingly, cryo-EM analysis revealed that about 3 out of 4 particles contained open DNA, although closed DNA was used for complex preparation. DNA opening occurred in the absence of TFIIH (Extended Data Fig. 4h, i). From these particles we obtained an independent reconstruction of the spontaneously formed OC at 6.1 Å resolution (OC5, Extended Data Fig. 4h). Weak density for upstream and downstream DNA segments indicates that DNA bubbles of various sizes formed during DNA opening (Extended Data Fig. 4i). The reconstruction resembles the high-resolution OC structure, suggesting that the latter was not perturbed by the use of pre-opened DNA (Extended Data Fig. 4h).

From the remaining particle images we obtained a reconstruction of the CC at 8.8 Å resolution (Fig. 4a). Comparison of this CC reconstruction with the OC structure reveals movements mainly in TFIIE (Fig. 4b, Extended Data Fig. 4e–g). In the CC, the E-wing lies on top of the DNA around position –7, in the region where DNA opening begins⁷ (Extended Data Fig. 4e). Consistent with this contact, Tfa1 (or the human counterpart TFIIEα) crosslinks to DNA near this point in the CC^{30, 40, 41} (Extended Data Fig. 4e). Conversion of the CC to the OC involves movement of upstream DNA and the DNA-associated domains Tfa2 WH1 and Tfg2 WH towards the cleft. DNA opening allows the TFIIE eWH domain to bind the tip of the clamp, and enables the E-wing to move near the upstream edge of the DNA bubble. Consistent with these changes, crosslinks between upstream DNA and the large TFIIE subunit are altered when the CC is converted to the OC⁴⁰. These observations indicate that DNA opening involves TFIIE, and in particular the eWH domain.

a, Details of the closed complex viewed from the front⁵⁰. Highlighted are TFIIE, TFIIF Tfg2 WH, and DNA, superimposed on their density. The promoter DNA displays increased flexibility downstream of the E-wing contact at position −7 upstream of the TSS (+1). b, Different positions of the TFIIE eWH in closed (dark magenta) and open (light magenta) complexes, viewed from the top⁵⁰. Compare Extended Data Fig. 4e–g.

• Full size image (301 KB)
• Download PowerPoint slide (606 KB)

TFIIE may also be involved in loading of the DNA template strand into the active centre during the transition from the CC to the OC. In previous structures, the path for loading the template was obstructed by the TFIIB B-reader and the Pol II fork loop 1 and lid. However, the B-reader is mobile in the OC structure, and fork loop 1 and the lid are moved to provide a path for template-strand loading (Fig. 5b, Extended Data Fig. 5a–c). These movements are apparently triggered by TFIIE because binding of the E-ribbon leads to a shift in the B-ribbon that partially withdraws the B-reader from the cleft (Fig. 5c). Thus allosteric binding of the E-ribbon apparently induces ‘clearance’ of the Pol II cleft that may facilitate template-strand loading into the active centre, and transcription start-site scanning in yeast.

a, OC structure viewed from the top⁵⁰. Highlighted are the Pol II active site (magenta sphere), fork loop 1 (yellow), lid (dark red), rudder (magenta), wall (navy blue), dock (brown), zipper (dark green), TFIIB B-ribbon (green), TFIIE E-ribbon (magenta) and downstream DNA (template, dark blue; non-template, cyan). The template single-strand was modelled using the Pol II–TFIIB ITC¹⁵ crystal structure. b, Fork loop 1 and lid assume new positions in the OC compared to the ITC¹⁵ and this opens a path (arrow) for loading of the template DNA strand (blue) into the active site (magenta sphere). Surface representations of Pol II cleft (silver), and cleft elements fork loop 1, lid, and rudder in the OC (left), and in a Pol II–TFIIB ITC¹⁵ (PDB: 4BBS, right). Movement of the Pol II lid (left, black to dark red) leads to a steric clash with the B-reader (cyan). Compare Extended Data Fig. 5a–c. c, Allosteric binding of the TFIIE E-ribbon may lead to an altered position of the TFIIB B-ribbon. Movements in Pol II wall and flap loop (navy blue), dock (brown), zipper (dark green), and B-ribbon (green) are observed in presence of TFIIE compared to the crystal structures of the binary Pol II–TFIIB complex¹³ (dark grey, PDB: 3K1F) and Pol II–TFIIB ITC¹⁵ (light grey, PDB: 4BBS). The altered B-ribbon position may be stabilized by binding to a short helix formed in loop β12–β13 of the dock domain.

• Full size image (221 KB)
• Download PowerPoint slide (686 KB)

To support the proposed functions of structural elements in TFIIE, we prepared recombinant TFIIE variants and tested them for binding to the CC and for promoter-dependent transcription activity in yeast nuclear extract (Methods, Fig. 3c, Extended Data Fig. 3g–i). Mutation of only three surface residues in the E-ribbon that contact the Pol II subunit Rpb7 strongly impaired both binding to the CC and transcription activity, and led to a severe growth defect in yeast (Fig. 3c, Extended Data Fig. 3h, j). Further, the eWH domain is required for TFIIE function¹⁶, and disruption of the eWH contact with DNA by introducing negatively charged glutamate residues in the eWH helix α3 leads to a transcription defect and impairs binding to the CC (Fig. 3c, Extended Data Fig. 3h). Other mutations in the eWH domain did not show functional defects in our assays. Deletion of the E-wing results in a mild growth phenotype (Extended Data Fig. 3j), and does not impair in vitro transcription (Fig. 3c), maybe because TFIIH compensates for the loss of E-wing function in these assays.

Structural comparisons of the yeast CC, OC, and ITC¹⁹ with each other and with the highly conserved human CC²⁰ reveal differences in the positions of DNA, the clamp, and TFIIE, and lead to an extended model of DNA opening (Fig. 6a, b, Extended Data Fig. 6). In this model, promoter DNA is initially bent away from the active site by ~20° near position −10 at the tip of the clamp helices (yeast CC). The clamp then opens slightly (Extended Data Fig. 6b), allowing promoter DNA to bend in the opposite direction and to enter the upper part of the cleft (‘human CC’). This frees the site at the tip of the clamp helices that can now bind the TFIIE eWH domain. DNA can then no longer swing back to its initial position, because this would result in a steric clash with the repositioned eWH domain. However, when DNA opening occurs around position −10, the upstream DNA can swing back to its original position and this stabilizes the DNA duplex single-strand junction. TFIIH then rotates downstream DNA and pushes the template single strand into the cleft using its ATP-dependent translocase activity, as suggested previously¹⁶. When the bubble extends downstream⁵, the template strand is loaded into the active centre cleft via a path cleared by allosteric binding of TFIIE (yeast OC). Template-strand loading allows the clamp to close again and to trap downstream DNA in the cleft (Extended Data Fig. 6b). The B-reader then covers the template strand, and helps to detect the transcription start site¹³, triggering RNA synthesis (yeast ITC).

a, Gallery of initiation complexes depicting proposed movements (arrows) of DNA and basal factors during the transition from the CC to OC to ITC, from left to right, viewed from the side⁵⁰. Yeast CC and OC structures (this work) were complemented with our previous yeast ITC¹⁹ structure (EMD-2785) and an alternative model of the CC (‘human CC’), which was obtained by replacing the DNA with that in the human CC²⁰ (EMD-2306), and adjusting the clamp to the position observed in the human CC. Shown are cryo-EM densities for DNA, Tfg2 WH, and TFIIE. DNA positions −10, −7 (yeast CC) and +1 (TSS) are labelled. DNA was extended by one turn for the yeast CC (black bracket). The locations of TFIIA and TFIIE in the ITC were inferred from the yeast OC. Obstructing Pol II and TFIIF regions were removed for clarity. b, Schematic representation of a. Key elements for DNA opening are indicated (compare Extended Data Fig. 6).

• Full size image (345 KB)
• Download PowerPoint slide (824 KB)

This model explains how DNA opening can be achieved with the use of binding energy alone, at least at some promoters^{11, 12}. DNA opening allows for new protein interactions at the upstream DNA duplex single-strand junction involving the eWH domain, the TFIIF arm, and the Pol II clamp. DNA loading into the cleft also enables new interactions of the downstream DNA duplex with the TFIIF charged helix and the Pol II cleft and clamp. These additional contacts may compensate for the energy needed for DNA melting and help to trap open DNA and to prevent its re-closure.

A similar mechanism for ATP-independent DNA opening may be used in other transcription systems. Proteins with homologies to TFIIB, TFIIE, and TFIIF, but not TFIIH, are present in the Pol I and Pol III systems⁴², and counterparts of TBP, TFIIB, and TFIIE are found in archaea³⁸. The bacterial transcription initiation system is structurally unrelated, but conceptually similar. DNA opening occurs spontaneously above the cleft, the open DNA is trapped with the use of binding energy, and this requires clamp opening and closure^{43, 44, 45, 46, 47}. DNA opening at a subset of bacterial genes however requires the ATP-dependent initiation factor σ⁵⁴, and this enables further regulation⁴⁸. Similarly, Pol II regulation can occur at the level of DNA opening⁴⁹, and TFIIH is required to keep DNA open at least at selected promoters⁸. This suggests that the Pol II initiation system evolved to depend on the ATP-consuming factor TFIIH for DNA opening, probably in response to an expanded need for gene regulation.

Data reporting

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.

Vectors and sequences

The open reading frame (ORF) encoding full-length TBP was amplified from Saccharomyces cerevisiae (Sc) genomic DNA and cloned into a pOPINE vector, containing a C-terminal 6×histidine tag. Codon-optimized Tfg1, Toa1, and Toa2 were commercially obtained for expression in Escherichia coli (E. coli, Life Technologies). Toa1 (Δ95–209) was obtained by quick-change PCR and cloned with Toa2 into a pOPINE vector, containing a C-terminal 6×histidine-tag on Toa2. Tfa1 and Tfa2 ORFs were amplified from genomic Sc DNA and cloned sequentially into a pET vector. ORFs for Tfg1 and Tfg2 were cloned sequentially into a modified pET-Duet-1 vector, containing an N-terminal 10×histidine-8×arginine-SUMO-tag on Tfg1. Additional ribosomal binding sites were introduced as described⁵¹. For heterologous co-expression of Sc 16-subunit core Mediator (cMed)-Med1, 13-subunit cMed lacking Med4–Med9 was cloned with previously described vectors¹⁹. Three-subunit Med1–Med4–Med9 was prepared by cloning Sc ORFs of Med1, Med4, and Med9 into a modified pET-Duet-1 vector, containing an additional N-terminal 10×histidine-8×arginine-SUMO-tag on Med1. Sequences are available upon request.

Recombinant proteins

All proteins were expressed in E. coli BL21(DE3)RIL cells (Stratagene). The identity of all purified proteins was confirmed by mass spectrometry. All purified proteins and complexes were flash-frozen and stored at −80 °C.

Full-length TBP was expressed in E. coli that were grown in lysogeny broth (LB) medium at 37 °C to an optical density (OD) of 0.5 at 600 nm, and induced with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) for 4 h at 20 °C. Cells were lysed by sonication in buffer A (25 mM HEPES (pH 7.5), 500 mM KCl, 10% glycerol, 2.5 mM dithiothreitol (DTT)) containing protease inhibitors⁵². The soluble fraction was applied to a 5 ml Histrap HP column, washed with buffer A containing 1 M KCl, and eluted with buffer A containing 350 mM imidazole. The sample was diluted 1:4 with buffer A lacking KCl and applied to anion-exchange chromatography using a MonoS 5/50 (GE Healthcare), and eluted with a linear gradient of buffer A from 100–1000 mM KCl. TBP was further purified by size-exclusion chromatography using a Superose 12 10/300 (GE Healthcare) column, equilibrated in buffer B (10 mM HEPES (pH 7), 200 mM NaCl, 5% glycerol, 2 mM DTT). TBP-containing fractions were pooled and concentrated to 8 mg ml⁻¹.

TFIIA was obtained by co-expression of its subunits Toa1 and Toa2 in E. coli. Transformed cells were grown in LB medium at 37 °C to an OD of 0.5 at 600 nm. Expression was induced with 0.5 mM IPTG at 37 °C for 4 h. Cells were lysed in buffer C (25 mM Tris-HCl (pH 8.0), 500 mM NaCl, 10% glycerol, 2 mM DTT) containing protease inhibitors⁵², and cleared by centrifugation. The supernatant was applied to a 5-ml HisTrap HP column, washed with buffer C containing 1 M NaCl, and eluted with buffer C containing 250 mM imidazole. Fractions containing the complex were pooled, diluted 1:4 with buffer C lacking NaCl, loaded on a MonoQ 5/50 anion-exchange column, and eluted with a linear gradient of buffer C from 100–500 mM NaCl. TFIIA containing fractions were pooled and applied to a Superdex 75 10/300 column (GE Healthcare), in buffer B. TFIIA was concentrated to 8 mg ml⁻¹. TFIIB was prepared as described¹⁵.

Recombinant TFIIE was obtained by co-expression of its subunits Tfa1 and Tfa2 in E. coli. Cells were transformed and grown in LB medium at 37 °C to an OD of 0.6 at 600 nm, and expression was induced by 0.5 mM IPTG for 18 h at 18 °C. Cells were lysed by sonication in buffer D (50 mM Tris-HCl (pH 8.0), 300 mM NaCl, 0.02% Tween-20, 5 mM DTT) containing protease inhibitors⁵². The lysate was cleared by centrifugation and applied to a 5 ml Histrap HP column, equilibrated in buffer E (buffer M, lacking Tween-20, containing 10 mM imidazole). The column was washed with 10 column volumes of buffer E and eluted with buffer E containing 250 mM imidazole. TFIIE was then subjected to anion-exchange chromatography using a 5-ml Hi-Trap Heparin column (GE Healthcare), equilibrated in buffer F (50 mM Tris-HCl (pH 8.0), 100 mM NaCl, 2% glycerol, 5 mM DTT). The complex was eluted with a linear gradient of buffer F from 100–2000 mM NaCl. To improve purity, TFIIE was further applied to a Superose 12 10/300 size-exclusion column, in buffer G (5 mM HEPES (pH 7.25), 40 mM ammonium sulphate, 10 μM ZnCl₂, 10 mM DTT). TFIIE containing fractions were pooled and concentrated to 9.6 mg ml⁻¹.

Sc TFIIF subunits Tfg1 and Tfg2 were co-expressed in E. coli and cells were grown in LB medium at 37 °C to an OD of 0.8 at 600 nm. Expression was induced with 0.2 mM IPTG for 3 h at 37 °C. Cells were lysed by sonication in buffer H (50 mM HEPES (pH 7.0), 350 mM KCl, 10% glycerol, 2 mM DTT) supplemented with 50 mM imidazole and protease inhibitors⁵². Cleared lysate was applied to a 5-ml HisTrap HP column equilibrated in buffer H. The column was washed with 8 column volumes of buffer H containing 1000 mM KCl, and eluted with a linear gradient from buffer H to buffer I (50 mM HEPES (pH 7.0), 250 mM KCl, 800 mM imidazole, 10% glycerol, 2 mM DTT). The conductivity of the eluate was adjusted to match that of buffer J (50 mM HEPES (pH 7.0), 150 mM KCl, 10% glycerol, 2 mM DTT) and 3C protease cleavage was carried out for 2 h. The complex was then applied to cation-exchange chromatography using a 1-ml HiTrap SP HP column (GE Healthcare), equilibrated in buffer J, and eluted in a linear gradient from 150–1000 mM KCl. TFIIF was further purified by size-exclusion chromatography using a Superdex 200 10/300 Increase column (GE Healthcare), in buffer K (10 mM MES (pH 6.2), 150 mM KCl, 10% glycerol, 2 mM DTT). Purified TFIIF was concentrated to 3.9 mg ml⁻¹. For previous studies of the initiation complex we used the conserved S. mikatae Tfg1 (refs 18, 19), owing to difficulties in recombinant expression of its Sc homologue. Here, expression of Sc Tfg1 was enabled by codon optimization.

For preparation of recombinant 16-subunit cMed–Med1, 13-subunit cMed lacking Med4–Med9 was prepared essentially as described¹⁹. Separately, a plasmid containing Med1–Med4–Med9 was transformed in E. coli and cells were grown in LB medium at 37 °C to an OD of 0.6 at 600 nm. Protein expression was induced with 0.5 mM IPTG at 18 °C for 24 h. Cells were collected and lysed by sonication in buffer L (25 mM HEPES (pH 7.5), 400 mM potassium acetate, 10% glycerol, 20 mM imidazole (pH 8), 2 mM DTT) containing protease inhibitors⁵². Lysate was cleared by centrifugation and applied to a 5-ml Histrap HP column (GE Healthcare), equilibrated in buffer M (25 mM HEPES (pH 7.5), 400 mM potassium acetate, 10% glycerol, 30 mM imidazole (pH 8), 2 mM DTT). The column was washed with 5 column volumes of buffer M and a linear gradient over 10 column volumes from 30–100 mM imidazole. The heterotrimer was eluted with 300 mM imidazole over 10 column volumes. Fractions containing the complex were diluted 1:3 with buffer N (25 mM HEPES (pH 7.5), 100 mM KCl, 10% glycerol, 1 mM EDTA, 5 mM DTT) and incubated with 3C protease for 2 h on ice, to remove the affinity tag. The complex was further purified by anion-exchange chromatography using a MonoQ 5/50 GL column (GE Healthcare), equilibrated in buffer N, and was eluted with a linear gradient from 100–600 mM KCl over 150 column volumes. Fractions containing the complex were pooled and applied to size-exclusion chromatography using a Superose 6 10/600 column (GE Healthcare), equilibrated in buffer O (25 mM HEPES (pH 7.5), 200 mM KCl, 5 mM DTT). Purified Med1–Med4–Med9 complex was concentrated to 2.7 mg ml^–1. To prepare 16-subunit cMed–Med1, 13-subunit cMed was incubated with a twofold molar excess of Med1–Med4–Med9 for 30 min at 25 °C and 30 min on ice, and purified by size-exclusion using a Superose 6 10/600 column, in buffer P (25 mM HEPES (pH 7.5), 400 mM KCl, 5% glycerol, 5 mM DTT). Fractions containing cMed–Med1 were pooled and concentrated to 2 mg ml⁻¹.

Preparation of initiation complexes

Yeast 12-subunit Pol II was prepared as described⁵³. Open and closed initiation complexes were prepared with several modifications of the previous cITC–cMed assembly scheme¹⁹ and with a different nucleic acid scaffold for the closed complex. The 72 nucleotide nucleic acid scaffold previously used to prepare the Pol II core ITC (cITC)¹⁹ contains a 15 nucleotide mismatch transcription bubble and six nucleotide RNA, and was used for assembly of the open initiation complex (OC). The closed complex (CC) contained a nucleic acid scaffold with 13 nucleotide longer downstream DNA, based on the HIS4 promoter (template, 5′-TGATATTTTTATGTATGTACAACACACATCGGAGGTGAATCGAACGTTCCATAGCTATTATATACACAGCGTGCTACTGTTCTCG-3′; non-template, 5′-CGAGAACAGTAGCACGCTGTGTATATAATAGCTATGGAACGTTCGATTCACCTCCGATGTGTGTTGTACATACATAAAAATATCA-3′). The initiation complex was prepared as follows. Pol II (200 μg at 3 mg ml⁻¹) was incubated with a fourfold molar excess of TFIIF. A twofold molar excess of nucleic acid scaffold over Pol II, tenfold molar excess of TFIIA, fourfold molar excess of TBP and TFIIB were added to buffer Q (25 mM HEPES-KOH (pH 7.5), 150 mM potassium acetate, 5% glycerol, 2 mM MgCl₂, 5 mM DTT) and incubated with pre-formed Pol II–TFIIF complex for 8 min at 25 °C. TFIIE was added in a tenfold molar excess over Pol II and incubated for 5 min at 25 °C. cMed–Med1 was added in a 1.2-fold molar excess over Pol II and incubated for 50 min at 25 °C. Open and closed complexes were purified using a Superose 6 3.2/300 size exclusion column (GE Healthcare), equilibrated in buffer Q. Fractions containing the complex were pooled (0.4–0.8 mg ml⁻¹) and additionally incubated with equimolar amounts of nucleic acid scaffold. The sample was then crosslinked for 30 min on ice using 0.1% glutaraldehyde (Electron Microscopy Sciences), and the reaction was quenched with 50 mM lysine (Sigma). The crosslinked sample was re-purified in a second size-exclusion step using a Superose 6 3.2/300 column, equilibrated in buffer Q lacking glycerol. Fractions containing initiation complexes were pooled (0.2–0.6 mg ml⁻¹) and used for EM grid preparation.

Electron microscopy

Initiation complex samples were applied to R3.5/1 holey carbon grids (Quantifoil). Grids were glow-discharged for 15 s before deposition of 4.5 μl complex, and subsequently blotted and vitrified by plunging into liquid ethane with a Vitrobot Mark IV (FEI) operated at 4 °C and 100% humidity. Cryo-EM data was acquired on a FEI Titan Krios operated in EFTEM mode at 300 keV, and equipped with a K2 Summit direct detector (Gatan). Automated data collection was carried out using the TOM toolbox⁵⁴ to acquire 1756 movies of the OC with a range of defocus values (from −0.7 μm to −4.2 μm) at a nominal magnification of 37,000× (1.35 Å per pixel). The camera was operated in ‘super-resolution’ mode (0.675 Å per pixel), with a total exposure time of 10 s fractionated into 25 frames, a dose rate of ~5 e⁻ per pixel per second, and total dose of 33 e⁻ Å⁻² per movie. Cryo-EM data of the CC was collected in the same manner, except that 959 movies were acquired with a defocus range from −0.8 μm to −5 μm, a total exposure time of 6 s with 20 frames, a dose rate of ~8 e⁻ per pixel per second, and total dose of 40 e⁻ Å⁻². Movies were aligned as described^{19, 55}, except that images were not partitioned into quadrants.

Image processing

For single particle analysis of the OC, an initial set of 10,225 particles was selected semi-automatically using e2boxer.py from EMAN2 (ref. 56). CTF parameters were estimated using CTFFIND4 (ref. 57). CTF correction and subsequent image processing was performed with RELION 1.3 (ref. 58), unless otherwise noted. Resolution was reported on the basis of the gold-standard Fourier shell correlation (FSC) (0.143 criterion) as described previously⁵⁹ and temperature factors were determined and applied automatically in RELION⁵⁸. Selected particles were extracted with a 300² pixel box and pre-processed to normalize images. Reference-free two-dimensional (2D) class averages were calculated, and twelve representative classes were low-pass filtered to 25 Å resolution and used as templates for automated picking⁶⁰ of all micrographs. The resulting 415,030 particle images were screened manually and by reference-free 2D classification, yielding 257,259 particle images that were used for subsequent processing. The 7.8 Å cryo-EM map of the yeast cITC¹⁹ (EMD-2785) was low-pass filtered to 50 Å and used as initial model for 3D refinement of the 10,225 particle set. This revealed an OC density to an estimated resolution of 10 Å. This density was low-pass filtered to 50 Å and used for processing of the complete OC cryo-EM data set (Extended Data Fig. 7a–d). A 3D reconstruction of all particles was calculated to 3.75 Å, and subjected to particle polishing using RELION 1.4 beta⁵⁸. This lead to an improved density at 3.58 Å resolution.

Hierarchical 3D classification was carried out without image alignment, to reduce computational requirements and identify homogeneous single particle groups (Extended Data Fig. 7a–d). Soft masks encompassing the complete OC or smaller regions of Pol II and basal factors were generated using the volume eraser in UCSF Chimera⁶¹ and RELION⁵⁸. This included masks for Pol II stalk–TFIIE E-ribbon (Extended Data Fig. 7a, c), upstream DNA–TBP–TFIIA–TFIIB–Tfg2 linker (Extended Data Fig. 7b), TFIIE–Tfg2 WH (Extended Data Fig. 7c), TFIIF (Extended Data Fig. 7d), and TFIIF dimerization domain (Extended Data Fig. 7d). Each class was refined using the 3D auto-refine procedure against the respective particles within that class with a soft reference mask in the shape of the OC (maximum diameter of 270 Å), generated in RELION⁵⁸. The OC1 reconstruction (improved Pol II core, B-ribbon, and E-ribbon density) was determined from 102,876 particles to a resolution of 3.58 Å and with a temperature factor of −111 Ų (Extended Data Fig. 8a). The OC2 reconstruction (improved upstream DNA, TBP, TFIIA, TFIIB, and Tfg2 linker density) was determined from 17,282 particles to a resolution of 3.97 Å and with a temperature factor of −95 Ų (Extended Data Fig. 8b). The OC3 reconstruction (improved TFIIE density) was determined from 11,231 particles to a resolution of 4.35 Å and with a temperature factor of −125 Ų (Extended Data Fig. 8c). The OC4 reconstruction (improved TFIIF dimerization and Tfg1 arm density) was determined from 29,455 particles to a resolution of 3.89 Å and with a temperature factor of −79 Ų (Extended Data Fig. 8d). Focused refinement of the upstream DNA assembly (OC2-focused) was achieved by continuing the auto-refinement (round 1, class 4; Extended Data Fig. 7c) from the first round of local searches using a corresponding soft mask. Focused refinement of TFIIE–Tfg2 WH (OC3-focused) was achieved by continuing the OC3 auto-refinement from the first round of local searches using a soft mask encompassing TFIIE, Tfg2 WH, and the interacting segment of upstream DNA. Focused refinement of TFIIF dimerization domain (OC4-focused) was achieved by continuing the auto-refinement (round 2, class 2; Extended Data Fig. 7d) from the first round of local searches using a soft mask encompassing the TFIIF dimerization domain and the Pol II lobe. The resolution of focused refinements was determined using a soft mask with a 30-pixel soft edge⁶², to 4.7 Å, 7.5 Å, and 4.09 Å for OC2-, OC3-, and OC4-focused refinements respectively (Extended Data Fig. 8b, c, d).

Single particle cryo-EM analysis of the closed complex was carried out essentially as for the open complex, with the following differences. Reference-free 2D classification of 10,591 particles, picked semi-automatically using e2boxer.py from EMAN2⁵⁶, gave three representative 2D class averages that were low-pass filtered to 30 Å resolution and used as templates for automated particle picking in RELION of all micrographs⁶⁰. The resultant 155,079 particles were screened manually and by reference-free 2D classification, yielding 111,625 particles that were used for subsequent processing. The cITC (EMD-2785), low-pass filtered to 50 Å, was used as the reference model for the initial 10,591 particle set, and the resultant 3D reconstruction (14 Å resolution), again low-pass filtered to 50 Å, was used as initial reference model for 3D refinement using the full cryo-EM data set. This revealed the CC at 7.5 Å, and was improved by particle polishing to 6.5 Å resolution using RELION 1.4 beta⁵⁸. Hierarchical 3D classification without image alignment using a soft mask encompassing the complete CC resulted in two populations, CC and OC5, respectively (Extended Data Fig. 7e). Both classes were refined using the 3D auto-refine procedure against the respective particles within that class with a soft mask in the shape of the CC. The CC soft mask was also suitable for the OC as both complexes have a similar shape. The CC reconstruction was determined from 7,527 particles to a resolution of 8.2 Å. The OC5 reconstruction was determined from 79,797 particles to a nominal resolution of 6.1 Å and a temperature factor of −176 Ų was applied (Extended Data Fig. 8g). To improve the densities TFIIE, Tfg2 WH and downstream DNA in the CC, we carried out focused classification of the Pol II stalk and subsequently TFIIE, Tfg2 WH and downstream DNA using soft masks encompassing these regions. The individual classes were refined as before and yielded a CC at 8.8 Å comprising 5,690 particles and a temperature factor of −300 Ų was applied (Extended Data Fig. 8f). A similar focused classification scheme for OC5 revealed moderately improved density for TFIIE (not shown).

Local resolution estimates were determined using a sliding window of 40³ voxels as previously described except that a single pair two half-maps was used and resolution estimates were not capped at the nominal resolution¹⁹ as no local filters were applied.

Structural modelling

A composite model of the OC was obtained using cryo-EM densities OC1–OC4 and the focused refinements of OC2, OC3, and OC4 particles. Structural models were built in COOT⁶³ unless indicated otherwise. Models were refined using the real space refinement routine in Phenix⁶⁴ into the respective OC density, as indicated, with secondary structure and rotamer restraints. First, structural models of Pol II (lacking Rpb4–Rpb7), TFIIB B-ribbon (residues 22–59), and downstream DNA (PDB: 4V1N¹⁹) were placed into the OC1 map using UCSF Chimera⁶¹, followed by rigid-body group refinement in Phenix⁶⁴. Models of the Pol II core and TFIIB B-ribbon were adjusted and extended manually, and refined in Phenix⁶⁴ into the OC1 density. The Rpb4–Rpb7 structure (PDB: 4V1N) was fitted, and residues at the base of the stalk (Rpb7 1–81, 150–159) were modified to fit the density and refined into the OC1 map. TFIIE was modelled using structural information on eWH³³, E-ribbon³⁴, and WH1³⁶, and a homology model for WH2 (Fig. 3a, b). A homology model of the human TFIIEα E-ribbon (PDB: 1VD4 (ref. 34)) was generated using MODELLER⁶⁵ (Tfa1 residues 121–158) and rigid-body fitted into the unsharpened OC1 density. Upstream DNA (PDB: 4V1N), TFIIB N-terminal cyclin (PDB: 4BBR, chain M and residues 122–213), the homology model of TFIIB C-terminal cyclin domain (PDB: 4V1N, chain M and residues 233–343) were fitted into the OC2 cryo-EM density. Models of Pol II protrusion, wall and the TFIIB N-terminal cyclin were adjusted and extended manually, and the TFIIB C-terminal cyclin (residues 236–328) and Tfg2 linker regions (residues 249–280) were built into OC2 and OC2-focused maps and refined into the OC2 map in Phenix⁶⁴. X-ray structures of TBP (PDB: 1YTB⁶⁶, chain A), and TFIIA (PDB: 1YTF²⁵, chains B–D) were individually fitted into the OC2 map obtained from focused refinement (OC2-focused). The fit of upstream DNA, from the upstream edge of the DNA bubble to TBP, was improved in UCSF Chimera⁶¹ using the OC3 map. Protein models of Tfa1 eWH (residues 1–90) and Tfa2 WH2 (residues 187–249), generated with the I-TASSER prediction server⁶⁷, were fit into the OC3 density and adjusted. The TFIIE linker helices were built de novo and modified to match the density, and together with Tfa1 eWH and Tfa2 WH2 models were refined into the unsharpened OC3 map in Phenix⁶⁴ and subjected to the phenix.model_idealization routine to optimize geometry. Published NMR data of the Tfg2 WH domain⁶⁸ (BMRB accession number 17916) was used to calculate a low-energy model with the BMRB CS-ROSETTA server⁶⁹. The Tfg2 WH and an I-TASSER model of Tfa1 WH1 (residues 123–181) were placed into the OC3 density obtained from focused refinement (OC3-focused). The Tfg2 WH positioning is also consistent with NMR data for the Tfg2 WH–DNA interface⁶⁸. A part of the Tfg1 arm domain (residues 327–349) was built into OC3 and OC4 densities and refined into the OC4 density. A previous homology model of the TFIIF dimerization domain¹⁸ was used as a starting template for modelling into OC4 and OC4-focused density maps, and was refined into the OC4 density. The crystal structure of the Tfg1 N-terminal peptide (residues 21–35, see below) was fitted into an OC4 density (round 2 class 2, Extended Data Fig. 7d). The individually refined models showed good stereochemistry and were validated with Molprobity⁷⁰, and the FSC of map versus model (Extended Data Figs 8 and 9).

To generate the CC model, the OC model of Pol II, TFIIA, TFIIB, TBP, TFIIE E-ribbon and TFIIF dimerization domain was rigid-body fitted into the CC density using an automated global 6D correlation search in Situs⁷¹. The remaining protein part was divided into two rigid bodies containing (i) Tfa1 and Tfa2 WH2, and (ii) Tfa2 WH1 and Tfg2 WH, which were independently fitted into the CC density using Situs⁷¹. To model closed promoter DNA, upstream DNA was extended with canonical duplex B-form DNA in COOT⁶³ and rigid-body fitted in UCSF Chimera⁵⁷ to reflect the density.

All figures were generated using UCSF Chimera⁶¹.

Crystallographic analysis of Pol II–TFIIF complex

The structure of the Tfg1 N-terminal region (residues 21–35) bound to Pol II was determined by X-ray crystallographic analysis. 12-subunit Pol II was prepared and crystallized⁵³, and TFIIF or selenomethionine (SeMet)-substituted Tfg1 peptide (residues 19–41, Peptide Speciality Laboratories GmBH, Heidelberg) was added to the cryo-protectant as described⁷². Diffraction data were collected on a PILATUS 6M detector at the X06SA beamline (SLS, Villigen, Switzerland) at 100 K. Data for the Pol II–Tfg1 SeMet peptide crystal was collected at a wavelength of 0.97972 Å and for the Pol II–TFIIF crystal at 0.91889 Å. Data were processed with XDS and XSCALE⁷³. The structure was phased with the crystal structure of the 12-subunit Pol II (PDB: 3PO2) lacking nucleic acid and refined using BUSTER⁷⁴. Model building of the TFIIF N-terminal residues in COOT⁶³ was guided by a selenium anomalous difference Fourier peak (M27). Subsequent refinement in BUSTER used secondary structure restraints for the Tfg1 peptide helix. The final structures had an R_free factor of 18.0% and 19.4% for the Pol II–Tfg1 SeMet peptide crystal and the Pol II–TFIIF crystal, respectively (Extended Data Fig. 9c), and showed good stereochemistry⁷⁰. In the Pol II–Tfg1 SeMet peptide structure 90% of the residues fall in favoured regions of the Ramachandran plot and 3% in disallowed regions. For the Pol II–TFIIF structure 89% of the residues fall in favoured regions of the Ramachandran plot and 3% in disallowed regions.

Yeast and functional assays

TFIIE interfaces with Pol II and DNA were probed by in vivo mutagenesis. The yeast strain used for TFA1 genetic assays¹⁶ was provided by S. Hahn (Fred Hutchinson Cancer Research Center). Mutations were introduced into a plasmid encoding TFA1 (pSH810, ARS CEN LEU2 3×Flag), as indicated in Extended Data Fig. 3g. The tfa1(ΔE-wing) mutant was generated by deletion of TFA1 residues 71–84, and their replacement by the residues GSG. The TFA1 constructs were transformed into the shuffle strain, and were streaked twice on −Ura −Leu plates, and subsequently onto yeast extract-peptone-dextrose (YEPD) plates. Yeast were freshly grown in YEPD medium and resuspended in water to an OD of 1 at 600 nm, and tenfold dilutions were spotted on 5-fluoroorotic acid (5-FOA) and YEPD plates, and incubated at 30 °C.

To further characterize the interfaces between TFIIE and the initiation complex, recombinant TFIIE mutants were purified according to the same protocol as wild-type TFIIE (Recombinant proteins, Extended Data Fig. 3g, h). The Tfa1(poly-Ala E-wing) mutant was generated by replacement of TFA1 residues 71–84 with poly-Alanine.

To assess the interaction of wild-type TFIIE and the TFIIE mutants with the CC by protein pulldown, 3 μg purified Pol II was first biotinylated on the Rpb3 subunit as described¹⁹. The CC, containing biotinylated Pol II, was subsequently prepared as above (‘preparation of initiation complexes’), but without TFIIE. This preparation was immobilized on 15 μL Dynabeads M280 streptavidin resin (Life Technologies), equilibrated in buffer Q. 10 μg TFIIE or TFIIE mutant (tenfold molar excess over Pol II) was incubated with the immobilized CC or control beads for 1 h at 4 °C. The beads were washed four times and bound proteins were analysed by SDS–PAGE (Extended Data Fig. 3h).

Promoter-dependent in vitro transcription was used to determine the activity of TFIIE and the TFIIE mutants. The yeast strain used for TFA1 genetic assays¹⁶ was transformed with a plasmid containing 3×Flag-tagged TFA1 (pSH810, ARS CEN LEU2 3×Flag). Transformants were streaked once onto −Ura −Leu plates, once onto −Leu plates, twice onto 5-FOA plates, and subsequently onto YEPD plates. Nuclear extract from 3 l yeast culture was prepared as described¹⁹. The nuclear extract was immunodepleted of 3×Flag–Tfa1 as described¹⁹ with the following modifications. Before nuclear extract was incubated with anti-Flag M2 agarose beads (Sigma), beads were incubated with 1 mg ml⁻¹ bovine serum albumin (BSA) protein (Sigma) for 1 h at 4 °C on a turning wheel followed by three wash steps. Immunodepleted nuclear extract was separated from beads by Micro Bio-Spin chromatography columns (Biorad). Specificity of the depletion was confirmed by western blot carried out as described previously¹⁹ for 3×Flag-tagged Tfa1 (Macs Miltenyi Biotec, 130-101-572), Rpb3 (Neoclone, WP012), TFIIB (Abcam, sc-274) and Histone H3 (Abcam, ab21054) (Extended Data Fig. 3i). The secondary antibodies anti-rabbit IgG horse-radish peroxidase (HRP; GE Healthcare, NA934) and anti-mouse IgG HRP (Abcam, ab5870) were used (Extended Data Fig. 3i). Activator- and promoter-dependent in vitro transcription and primer extension were carried out as described previously¹⁹. Recombinant TFIIE (5 pmol) and TFIIE mutants (5 pmol) were added to the depleted nuclear extract as indicated in Fig. 3c. Transcripts were visualized on a denaturing 8% polyacrylamide TBE gel with a Typhoon 9500 scanner (GE Healthcare) and quantified with ImageQuant (GE Healthcare). For quantification, the relative activity of each variant compared to TFIIE was calculated for each replicate. The mean intensity and standard deviation of three replicates was calculated from their relative activities. A second RNA product from the HIS4 promoter was observed, apparently resulting from an alternative upstream transcription start site. Although some differences in the relative use of the two TSSs are observed in the assay, we refrained from interpreting these.

1. Buratowski, S., Hahn, S., Guarente, L. & Sharp, P. A. Five intermediate complexes in transcription initiation by RNA polymerase II. Cell 56, 549–561 (1989)
   • CAS
   • OpenURL
   • ISI
   • PubMed
   • Article
2. Roeder, R. G. The role of general initiation factors in transcription by RNA polymerase II. Trends Biochem. Sci. 21, 327–335 (1996)
   • CAS
   • OpenURL
   • ISI
   • PubMed
   • Article
3. Grünberg, S. & Hahn, S. Structural insights into transcription initiation by RNA polymerase II. Trends Biochem. Sci. 38, 603–611 (2013)
   • CAS
   • OpenURL
   • PubMed
   • Article
4. Sainsbury, S., Bernecky, C. & Cramer, P. Structural basis of transcription initiation by RNA polymerase II. Nat. Rev. Mol. Cell Biol. 16, 129–143 (2015)
   • CAS
   • OpenURL
   • PubMed
   • Article
5. Giardina, C. & Lis, J. T. DNA melting on yeast RNA polymerase II promoters. Science 261, 759–762 (1993)
   • CAS
   • OpenURL
   • ISI
   • PubMed
   • Article
6. Ohkuma, Y. & Roeder, R. G. Regulation of TFIIH ATPase and kinase activities by TFIIE during active initiation complex formation. Nature 368, 160–163 (1994)
   • CAS
   • OpenURL
   • ISI
   • PubMed
   • Article
7. Maxon, M. E., Goodrich, J. A. & Tjian, R. Transcription factor IIE binds preferentially to RNA polymerase IIa and recruits TFIIH: a model for promoter clearance. Genes Dev. 8, 515–524 (1994)
   • CAS
   • OpenURL
   • ISI
   • PubMed
   • Article
8. Holstege, F. C., van der Vliet, P. C. & Timmers, H. T. Opening of an RNA polymerase II promoter occurs in two distinct steps and requires the basal transcription factors IIE and IIH. EMBO J. 15, 1666–1677 (1996)
   • CAS
   • ISI
   • PubMed
9. Parvin, J. D., Timmers, H. T. & Sharp, P. A. Promoter specificity of basal transcription factors. Cell 68, 1135–1144 (1992)
   • CAS
   • OpenURL
   • PubMed
   • Article
10. Pan, G. & Greenblatt, J. Initiation of transcription by RNA polymerase II is limited by melting of the promoter DNA in the region immediately upstream of the initiation site. J. Biol. Chem. 269, 30101–30104 (1994)
    • CAS
    • ISI
    • PubMed
11. Buratowski, S., Sopta, M., Greenblatt, J. & Sharp, P. A. RNA polymerase II-associated proteins are required for a DNA conformation change in the transcription initiation complex. Proc. Natl Acad. Sci. USA 88, 7509–7513 (1991)
    • CAS
    • OpenURL
    • PubMed
    • Article
12. Holstege, F. C., Tantin, D., Carey, M., van der Vliet, P. C. & Timmers, H. T. The requirement for the basal transcription factor IIE is determined by the helical stability of promoter DNA. EMBO J. 14, 810–819 (1995)
    • CAS
    • ISI
    • PubMed
13. Kostrewa, D. et al. RNA polymerase II-TFIIB structure and mechanism of transcription initiation. Nature 462, 323–330 (2009)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
14. Liu, X., Bushnell, D. A., Wang, D., Calero, G. & Kornberg, R. D. Structure of an RNA polymerase II-TFIIB complex and the transcription initiation mechanism. Science 327, 206–209 (2010)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
15. Sainsbury, S., Niesser, J. & Cramer, P. Structure and function of the initially transcribing RNA polymerase II–TFIIB complex. Nature 493, 437–440 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
16. Grünberg, S., Warfield, L. & Hahn, S. Architecture of the RNA polymerase II preinitiation complex and mechanism of ATP-dependent promoter opening. Nat. Struct. Mol. Biol. 19, 788–796 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
17. Murakami, K. et al. Architecture of an RNA polymerase II transcription pre-initiation complex. Science 342, 1238724 (2013)
    • CAS
    • OpenURL
    • PubMed
    • Article
18. Mühlbacher, W. et al. Conserved architecture of the core RNA polymerase II initiation complex. Nat. Commun. 5, 4310 (2014)
    • OpenURL
    • PubMed
    • Article
19. Plaschka, C. et al. Architecture of the RNA polymerase II-Mediator core initiation complex. Nature 518, 376–380 (2015)
    • CAS
    • OpenURL
    • PubMed
    • Article
20. He, Y., Fang, J., Taatjes, D. J. & Nogales, E. Structural visualization of key steps in human transcription initiation. Nature 495, 481–486 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
21. Murakami, K. et al. Structure of an RNA polymerase II preinitiation complex. Proc. Natl Acad. of Sci. USA 112, 13543–13548 (2015)
    • OpenURL
    • Article
22. Nikolov, D. B. et al. Crystal structure of a TFIIB–TBP–TATA-element ternary complex. Nature 377, 119–128 (1995)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
23. Geiger, J. H., Hahn, S., Lee, S. & Sigler, P. B. Crystal structure of the yeast TFIIA/TBP/DNA complex. Science 272, 830–836 (1996)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
24. Tan, S., Hunziker, Y., Sargent, D. F. & Richmond, T. J. Crystal structure of a yeast TFIIA/TBP/DNA complex. Nature 381, 127–134 (1996)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
25. Eichner, J., Chen, H.-T. T., Warfield, L. & Hahn, S. Position of the general transcription factor TFIIF within the RNA polymerase II transcription preinitiation complex. EMBO J. 29, 706–716 (2010)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
26. Deng, W. & Roberts, S. G. E. TFIIB and the regulation of transcription by RNA polymerase II. Chromosoma 116, 417–429 (2007)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
27. Fishburn, J. & Hahn, S. Architecture of the yeast RNA polymerase II open complex and regulation of activity by TFIIF. Mol. Cell. Biol. 32, 12–25 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
28. Čabart, P., Újvári, A., Pal, M. & Luse, D. S. Transcription factor TFIIF is not required for initiation by RNA polymerase II, but it is essential to stabilize transcription factor TFIIB in early elongation complexes. Proc. Natl Acad. Sci. USA 108, 15786–15791 (2011)
    • OpenURL
    • PubMed
    • Article
29. Robert, F., Forget, D., Li, J., Greenblatt, J. & Coulombe, B. Localization of subunits of transcription factors IIE and IIF immediately upstream of the transcriptional initiation site of the adenovirus major late promoter. J. Biol. Chem. 271, 8517–8520 (1996)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
30. Forget, D., Langelier, M.-F. F., Thérien, C., Trinh, V. & Coulombe, B. Photo-cross-linking of a purified preinitiation complex reveals central roles for the RNA polymerase II mobile clamp and TFIIE in initiation mechanisms. Mol. Cell. Biol. 24, 1122–1131 (2004)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
31. Chen, H.-T. T., Warfield, L. & Hahn, S. The positions of TFIIF and TFIIE in the RNA polymerase II transcription preinitiation complex. Nat. Struct. Mol. Biol. 14, 696–703 (2007)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
32. Grohmann, D. et al. The initiation factor TFE and the elongation factor Spt4/5 compete for the RNAP clamp during transcription initiation and elongation. Mol. Cell 43, 263–274 (2011)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
33. Meinhart, A., Blobel, J. & Cramer, P. An extended winged helix domain in general transcription factor E/IIE alpha. J. Biol. Chem. 278, 48267–48274 (2003)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
34. Okuda, M. et al. A novel zinc finger structure in the large subunit of human general transcription factor TFIIE. J. Biol. Chem. 279, 51395–51403 (2004)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
35. Okuda, M. et al. Structural insight into the TFIIE-TFIIH interaction: TFIIE and p53 share the binding region on TFIIH. EMBO J. 27, 1161–1171 (2008)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
36. Okuda, M. et al. Structure of the central core domain of TFIIEβ with a novel double-stranded DNA-binding surface. EMBO J. 19, 1346–1356 (2000)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
37. Okamoto, T. et al. Analysis of the role of TFIIE in transcriptional regulation through structure-function studies of the TFIIEβ subunit. J. Biol. Chem. 273, 19866–19876 (1998)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
38. Blombach, F. et al. Archaeal TFEα/β is a hybrid of TFIIE and the RNA polymerase III subcomplex hRPC62/39. eLife 4, e08378 (2015)
    • OpenURL
    • PubMed
    • Article
39. Martinez-Rucobo, F. W., Sainsbury, S., Cheung, A. C. M. & Cramer, P. Architecture of the RNA polymerase–Spt4/5 complex and basis of universal transcription processivity. EMBO J. 30, 1302–1310 (2011)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
40. Kim, T. K., Ebright, R. H. & Reinberg, D. Mechanism of ATP-dependent promoter melting by transcription factor IIH. Science 288, 1418–1421 (2000)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
41. Miller, G. & Hahn, S. A DNA-tethered cleavage probe reveals the path for promoter DNA in the yeast preinitiation complex. Nat. Struct. Mol. Biol. 13, 603–610 (2006)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
42. Vannini, A. & Cramer, P. Conservation between the RNA polymerase I, II, and III transcription initiation machineries. Mol. Cell 45, 439–446 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
43. Chakraborty, A. et al. Opening and closing of the bacterial RNA polymerase clamp. Science 337, 591–595 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
44. Feklistov, A. & Darst, S. A. Structural basis for promoter-10 element recognition by the bacterial RNA polymerase σ subunit. Cell 147, 1257–1269 (2011)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
45. Zhang, Y. et al. Structural basis of transcription initiation. Science 338, 1076–1080 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
46. Zuo, Y. & Steitz, T. A. Crystal structures of the E. coli transcription initiation complexes with a complete bubble. Mol. Cell 58, 534–540 (2015)
    • OpenURL
    • PubMed
    • Article
47. Bae, B., Feklistov, A., Lass-Napiorkowska, A., Landick, R. & Darst, S. A. Structure of a bacterial RNA polymerase holoenzyme open promoter complex. eLife 4, (2015)
48. Yang, Y. et al. Structures of the RNA polymerase-σ⁵⁴ reveal new and conserved regulatory strategies. Science 349, 882–885 (2015)
    • OpenURL
    • PubMed
    • Article
49. Kouzine, F. et al. Global regulation of promoter melting in naive lymphocytes. Cell 153, 988–999 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
50. Cramer, P., Bushnell, D. A. & Kornberg, R. D. Structural basis of transcription: RNA polymerase II at 2.8 angstrom resolution. Science 292, 1863–1876 (2001)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
51. Baumli, S., Hoeppner, S. & Cramer, P. A conserved mediator hinge revealed in the structure of the MED7.MED21 (Med7.Srb7) heterodimer. J. Biol. Chem. 280, 18171–18178 (2005)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
52. Seizl, M. et al. A conserved GA element in TATA-less RNA polymerase II promoters. PLoS One 6, e27595 (2011)
    • CAS
    • OpenURL
    • PubMed
    • Article
53. Sydow, J. F. et al. Structural basis of transcription: mismatch-specific fidelity mechanisms and paused RNA polymerase II with frayed RNA. Mol. Cell 34, 710–721 (2009)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
54. Korinek, A., Beck, F., Baumeister, W., Nickell, S. & Plitzko, J. M. Computer controlled cryo-electron microscopy – TOM² a software package for high-throughput applications. J. Struct. Biol. 175, 394–405 (2011)
    • OpenURL
    • ISI
    • PubMed
    • Article
55. Li, X., Zheng, S. Q., Egami, K., Agard, D. A. & Cheng, Y. Influence of electron dose rate on electron counting images recorded with the K2 camera. J. Struct. Biol. 184, 251–260 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
56. Tang, G. et al. EMAN2: an extensible image processing suite for electron microscopy. J. Struct. Biol. 157, 38–46 (2007)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
57. Rohou, A. & Grigorieff, N. CTFFIND4: Fast and accurate defocus estimation from electron micrographs. J. Struct. Biol. 192, 216–221 (2015)
    • OpenURL
    • PubMed
    • Article
58. Scheres, S. H. RELION: implementation of a Bayesian approach to cryo-EM structure determination. J. Struct. Biol. 180, 519–530 (2012)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
59. Chen, S. et al. High-resolution noise substitution to measure overfitting and validate resolution in 3D structure determination by single particle electron cryomicroscopy. Ultramicroscopy 135, 24–35 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
60. Scheres, S. H. Semi-automated selection of cryo-EM particles in RELION-1.3. J. Struct. Biol. 189, 114–122 (2015)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
61. Pettersen, E. F. et al. UCSF Chimera--a visualization system for exploratory research and analysis. J. Comput. Chem. 25, 1605–1612 (2004)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
62. Nguyen, T. H. et al. The architecture of the spliceosomal U4/U6.U5 tri-snRNP. Nature 523, 47–52 (2015)
    • CAS
    • OpenURL
    • PubMed
    • Article
63. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132 (2004)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
64. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 (2010)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
65. Eswar, N. et al. Comparative protein structure modeling using Modeller. Current Protoc. Bioinformatics http://dx.doi.org/10.1002/0471250953.bi0506s15 (2006)
66. Kim, Y., Geiger, J. H., Hahn, S. & Sigler, P. B. Crystal structure of a yeast TBP/TATA-box complex. Nature 365, 512–520 (1993)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
67. Yang, J. et al. The I-TASSER Suite: protein structure and function prediction. Nat. Methods 12, 7–8 (2015)
    • CAS
    • OpenURL
    • PubMed
    • Article
68. Kilpatrick, A. M., Koharudin, L. M., Calero, G. A. & Gronenborn, A. M. Structural and binding studies of the C-terminal domains of yeast TFIIF subunits Tfg1 and Tfg2. Proteins 80, 519–529 (2012)
    • OpenURL
    • PubMed
    • Article
69. Shen, Y. et al. Consistent blind protein structure generation from NMR chemical shift data. Proc. Natl Acad. Sci. USA 105, 4685–4690 (2008)
    • CAS
    • OpenURL
    • PubMed
    • Article
70. Chen, V. B. et al. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D 66, 12–21 (2010)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
71. Wriggers, W. Using Situs for the integration of multi-resolution structures. Biophys. Rev. 2, 21–27 (2010)
    • OpenURL
    • PubMed
    • Article
72. Kinkelin, K. et al. Structures of RNA polymerase II complexes with Bye1, a chromatin-binding PHF3/DIDO homologue. Proc. Natl Acad. Sci. USA 110, 15277–15282 (2013)
    • OpenURL
    • PubMed
    • Article
73. Kabsch, W. XDS. Acta Crystallogr. D 66, 125–132 (2010)
    • OpenURL
    • PubMed
    • Article
74. Blanc, E. et al. Refinement of severely incomplete structures with maximum likelihood in BUSTER-TNT. Acta Crystallogr. D 60, 2210–2221 (2004)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
75. Gaiser, F., Tan, S. & Richmond, T. J. Novel dimerization fold of RAP30/RAP74 in human TFIIF at 1.7 A resolution. J. Mol. Biol. 302, 1119–1127 (2000)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
76. Ha, I. et al. Multiple functional domains of human transcription factor IIB: distinct interactions with two general transcription factors and RNA polymerase II. Genes Dev. 7, 1021–1032 (1993)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
77. Harami, G. M., Gyimesi, M. & Kovács, M. From keys to bulldozers: expanding roles for winged helix domains in nucleic-acid-binding proteins. Trends Biochem. Sci. 38, 364–371 (2013)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
78. Esnault, C. et al. Mediator-dependent recruitment of TFIIH modules in preinitiation complex. Mol. Cell 31, 337–346 (2008)
    • CAS
    • OpenURL
    • ISI
    • PubMed
    • Article
79. Čabart, P. & Luse, D. S. Inactivated RNA polymerase II open complexes can be reactivated with TFIIE. J. Biol. Chem. 287, 961–967 (2012)
    • OpenURL
    • PubMed
    • Article

Download references

We thank C. Bernecky, W. Mühlbacher, S. Neyer, S. Sainsbury, and D. Tegunov for help and discussions; L. Larivière and L. Wenzeck for cloning and initial purification of TFIIE; W. Mühlbacher for initial cloning of TFIIA; S. Bilakovic for the modified pET-DUET-1 vector; J. Mahamid for help with data collection for the CC; K. Maier for help with yeast growth assays; M. Raabe and H. Urlaub for protein identification; K. Kinkelin for initial Pol II–TFIIF co-crystallization; and S. Hahn for providing the TFA1 yeast strain and the shuffle plasmid pSH810. C.P. (SFB860), M.H. (GRK1721), and P.C. were supported by the Deutsche Forschungsgemeinschaft, the Advanced Grant TRANSIT of the European Research Council, and the Volkswagen Foundation.

Three-dimensional cryo-EM density maps of OC1, OC2, OC2-focused, OC3, OC3-focused, OC4, OC4-focused, OC5, and CC have been deposited in the Electron Microscopy Data Bank under the accession numbers EMD-3375, EMD-3376, EMD-3377, EMD-3378, EMD-3379, EMD-3380, EMD-3381, EMD-3382, and EMD-3383, respectively. Coordinate files of the OC and CC have been deposited in the Protein Data Bank under accession numbers 5FYW and 5FZ5. Coordinates and structure factors of the Pol II-Tfg1 peptide and the Pol II-TFIIF crystals have been deposited at the Protein Data Bank under the accession numbers 5IP7 and 5IP9.