Structures of transcription preinitiation complex engaged with the +1 nucleosome

The preinitiation complex (PIC) assembles on promoters of protein-coding genes to position RNA polymerase II (Pol II) for transcription initiation. Previous structural studies revealed the PIC on different promoters, but did not address how the PIC assembles within chromatin. In the yeast Saccharomyces cerevisiae, PIC assembly occurs adjacent to the +1 nucleosome that is located downstream of the core promoter. Here we present cryo-EM structures of the yeast PIC bound to promoter DNA and the +1 nucleosome located at three different positions. The general transcription factor TFIIH engages with the incoming downstream nucleosome and its translocase subunit Ssl2 (XPB in human TFIIH) drives the rotation of the +1 nucleosome leading to partial detachment of nucleosomal DNA and intimate interactions between TFIIH and the nucleosome. The structures provide insights into how transcription initiation can be influenced by the +1 nucleosome and may explain why the transcription start site is often located roughly 60 base pairs upstream of the dyad of the +1 nucleosome in yeast.

The preinitiation complex (PIC) assembles on promoters of protein-coding genes to position RNA polymerase II (Pol II) for transcription initiation. Previous structural studies revealed the PIC on different promoters, but did not address how the PIC assembles within chromatin. In the yeast Saccharomyces cerevisiae, PIC assembly occurs adjacent to the +1 nucleosome that is located downstream of the core promoter. Here we present cryo-EM structures of the yeast PIC bound to promoter DNA and the +1 nucleosome located at three different positions. The general transcription factor TFIIH engages with the incoming downstream nucleosome and its translocase subunit Ssl2 (XPB in human TFIIH) drives the rotation of the +1 nucleosome leading to partial detachment of nucleosomal DNA and intimate interactions between TFIIH and the nucleosome. The structures provide insights into how transcription initiation can be influenced by the +1 nucleosome and may explain why the transcription start site is often located roughly 60 base pairs upstream of the dyad of the +1 nucleosome in yeast.
Previous studies provided structures of yeast and human preinitiation complexes (PICs) on various promoters [1][2][3][4][5][6][7][8][9] . However, promoters are flanked by nucleosomes within the chromatin environment in vivo 10,11 and thus PIC structures must also be determined in the presence of nucleosomes. In the yeast Saccharomyces cerevisiae, PIC assembly occurs adjacent to the +1 nucleosome 12,13 , which resides at the downstream end of the core promoter. The +1 nucleosome is often well-positioned 14,15 and thought to be involved in PIC assembly 13 , although the underlying mechanisms are unclear 16 . The +1 nucleosome is associated with the PIC at most Pol II promoters 17 . Correct positioning of the +1 nucleosome positively influences binding of the TATA box-binding protein (TBP), selection of the transcription start site (TSS) and transcription activity [18][19][20][21] in vivo.
Here we use a combination of biochemistry and cryo-electron microscopy (cryo-EM) to provide structural insights into the PIC in the context of the +1 nucleosome. We show that TFIIH can engage with the +1 nucleosome in different ways and provide evidence that there is a preferred mode of TFIIH-nucleosome interaction that relies on multiple contacts. Finally, we use the PIC-nucleosome structures and modeling to provide a molecular explanation for long-standing observations on the preferred relative position of the TSS and the location of the +1 nucleosome. Our work thus provides the basis for a detailed analysis of structural and functional PIC-nucleosome interactions at gene promoters.

Formation of PIC-nucleosome complex
To investigate PIC assembly in the presence of the +1 nucleosome, we formed a PIC from the yeast S. cerevisiae on a promoter flanked by a +1 nucleosome (Fig. 1a,b). At most yeast promoters, the +1 nucleosome covers the TSS, which is typically located roughly 10-15 base pairs (bp) downstream of the proximal border of the nucleosome 15,20,22 . To mimic this natural arrangement, we prepared a His4 promoter template with a Widom-601-derived nucleosome positioning sequence that places

PIC-nucleosome contacts
The rotation of the incoming nucleosome leads to a more intimate association of the PIC and the +1 nucleosome. In the structure of complex A, TFIIH contacts the +1 nucleosome only via its Tfb2-Tfb5 dimerization domain ( Fig. 4a and Supplementary Video 1). In complex B, however, TFIIH forms four contact sites with the nucleosome (Fig. 4b and Supplementary Video 1). TFIIH subunits Ssl2 (XPB in human), Tfb2 (p52), Ssl1 (p44) and Tfb4 (p34) all contain charged loop residues that protrude toward the nucleosome in complex B (Fig. 4b). The Ssl2 ATPase domain engages with downstream DNA, whereas the Ssl2 N-terminal extension and clutch domains form a wedge between DNA and the nucleosome (Fig. 4b). This Ssl2 wedge stabilizes the two detached turns of nucleosomal DNA. Tfb2 possesses a lysine-containing loop in its HTH-3 domain (residues 258-270) that approaches the acidic patch of the nucleosome (Fig. 4b). Ssl1 uses a lysine-rich insertion in its RING domain (residues 414-421) to contact nucleosomal DNA around the dyad (Fig. 4b). Finally, Tfb4 uses an extension in its vWA domain (residues 90-105) to reach near the N-terminal region of histone H4 (Fig. 4b). These PIC-nucleosome interactions may counteract further rotation of the +1 nucleosome and impair TFIIH translocase action beyond this state.

Evidence for a preferred nucleosome orientation
We next asked whether the position of the nucleosome observed in complex B, and its interactions with TFIIH are specific to the DNA sequence used here or whether they may be of more generic nature. Yeast promoters vary with respect to the distance between their TATA the +1 nucleosome at a position in which the TSS is 10 bp downstream of the proximal border of the nucleosome (Fig. 1a). We reconstituted a nucleosome on this DNA and used the obtained nucleosomal template for in vitro assembly of the PIC 4 (Extended Data Fig. 1a and Methods).

The +1 nucleosome represses transcription
To evaluate the effect of the +1 nucleosome on transcription activity, we performed promoter-dependent in vitro transcription assays 9 ( Fig. 1b and Methods). We found that the presence of the +1 nucleosome strongly reduced RNA synthesis in this assay (Fig. 1c). The amount of full-length RNA product was reduced to 20% and shorter RNA transcripts were produced, as expected from Pol II stalling within the nucleosome 23 (Fig. 1c). To test whether the reduction of RNA synthesis is caused by the high stability of a nucleosome obtained on a Widom-601 derived positioning sequence, we repeated the assay with a template containing the natural His4 promoter sequence. We observed that the level of RNA synthesis was again repressed, albeit to a lesser extent than for the original template containing the nucleosome positioning sequence (Extended Data Fig. 1b). These results indicate that the degree of transcription reduction is related to the stability of the nucleosome while any type of nucleosome causes a decrease in RNA production.

Cryo-EM structure determination
We then determined cryo-EM structures of the reconstituted PIC-nucleosome complex (Methods) in the absence (complex A) or presence (complex B) of nucleoside triphosphates (NTPs) under the conditions of our transcription assay. Classification of the data identified a subset of particles that contained the complete complex (Extended Data Figs. 2 and 3). Cryo-EM densities for TFIIH and the nucleosome were further improved by focused refinement. For complex A we obtained a reconstruction at an overall resolution of 3.3 Å, with local resolutions of 2.9 Å for Pol II, 3.2 Å for the nucleosome and 3.7 Å for TFIIH (Extended Data Figs. 2 and 4 and Supplementary Video 1). Complex B was resolved at an overall resolution of 4.0 Å, with local resolutions of 3.4 Å for Pol II, 3.6 Å for the nucleosome and 3.9 Å for TFIIH (Extended Data Figs. 3 and 4 and Supplementary Video 1). The structures were obtained with the use of atomic models of the PIC 9 and the nucleosome 24 and subsequent manual modeling, leading to good stereochemistry (Table 1).

PIC structures are largely unchanged
In both cryo-EM structures, the overall conformation of the PIC resembles that in the absence of the nucleosome 4,9 , except for a minor rotation of TFIIH with respect to the rest of the PIC ( Fig. 2 and Extended Data Fig. 5a). The complex adopts the previously described closed promoter state with distorted DNA 25 and shows the TFIIH ATPase Ssl2 in the pretranslocation state. The DNA is located above the active center cleft and the initially melting DNA region is flanked by the Rpb1 clamp head loop and the TFIIF charged region as observed before 9 (Fig. 2a,b). In contrast to the closed promoter state observed here, previous cryo-EM studies of the yeast PIC revealed a large portion of PIC particles in the open promoter state 4,9 , suggesting that the +1 nucleosome counteracts DNA opening and that impaired DNA opening is responsible for the observed suppression of transcription in the presence of the +1 nucleosome.

Rotation of the +1 nucleosome
Comparison of the structures of complexes A and B shows that the +1 nucleosome is rotated by roughly 75° in the presence of NTPs (Fig. 3a). This rotation is apparently caused by the translocase activity of TFIIH subunit Ssl2 that hydrolyzes ATP to propel downstream DNA into the PIC. Such ATPase action is predicted to cause a rotation of the nucleosome by 30-40° with respect to the PIC for each translocated DNA base pair. The observed roughly 75° rotation would thus correspond to a DNA translocation of 2 bp toward the active center of Pol II, and this is reflected by an observed bending of the DNA duplex into the cleft of Pol II around the initially melted region (Extended Data Fig. 5b) Article https://doi.org/10.1038/s41594-022-00865-w box and their TSS in a range of roughly 40-120 bp (ref. 26 ), and thus the distance between their TATA box and the dyad of the +1 nucleosome can vary in a range of roughly 100-180 bp. To investigate any such distance effects, we prepared an additional PIC-nucleosome complex on an altered DNA template where the nucleosome positioning sequence was shifted downstream by 10 bp (complex C), resulting in a 130 bp distance between the TATA box and the nucleosome dyad, as compared to a 120 bp distance in complexes A and B (Fig. 5a).
We incubated complex C with NTPs and could obtain a cryo-EM structure at an overall resolution of 6.6 Å (Extended Data Fig. 6). Although the resolution prevented us from observing molecular details within the cryo-EM map, the overall orientation of the nucleosome with respect to the PIC in complex C resembled that in complex B (Fig. 5b). This observation indicates that TFIIH uses a common surface to contact the nucleosome in the state of complexes B and C even though the detailed PIC-nucleosome interaction is partially different, suggesting a preferred orientation between TFIIH and the nucleosome may exist when they collide even with different initial PIC-nucleosome distances and on different promoters. We speculate that the preferred orientation of the nucleosome with respect to the PIC observed in complexes B and C occurs within a common intermediate of the transcription initiation process at yeast promoters.

Rpb6 N-terminal tail (NTT) in the Pol II active center
In all three PIC-nucleosome structures, we obtained an ordered conformation of the NTT of Pol II subunit Rpb6 ( Fig. 6a and Supplementary Video 1) as confirmed by crosslinking mass spectrometry (Extended Data Fig. 7a). Whereas the NTT is mobile in all previous structures of Pol II complexes, Rpb6 residues 12-35 are observed here in the active center cleft of the polymerase (Fig. 6a). The Rpb6 NTT contains several negatively charged residues that interact with positively charged residues in the cleft that are often conserved ( Fig. 6a and Extended Data Fig. 7b).
Structural superposition shows that the NTT would clash with DNA and RNA in an initially transcribing complex (ITC) 27 (Fig. 6b), indicating its position is incompatible with transcription. The NTT is dispensable for growth of budding yeast 28 but its mutation causes temperature sensitivity in the fission yeast S. pombe 29 . We note that Pol I and Pol III also contain elements that can transiently occupy the active center (Extended Data Fig. 8). These elements are referred to as the expander

Coactivators may be accommodated
Finally, we asked whether the coactivators Mediator and TFIID may be accommodated in our PIC-nucleosome structures. Superimposition of the yeast core Mediator-PIC structure 4 indeed showed that Mediator could be appended to our structures without clashes (Extended Data Fig. 9a). In the resulting model, the Mediator hook domain approaches the +1 nucleosome up to a distance of roughly 40 Å (Extended Data Fig. 9a). We also superimposed our PIC-nucleosome structures onto two human PIC structures containing TFIID 6,7 since no yeast TFIID-containing PIC structure is available. This showed that TFIID could in principle be accommodated in the PIC-nucleosome complex surface without clashes ( Fig. 7 and Extended Data Fig. 9b). The putative TFIID position is consistent with reports that the double bromodomain of TFIID subunit TAF1 and the PHD finger domain of TAF3 can contribute to promoter recognition by binding modified histone tails 33,34 (Extended Data Fig. 9b). Histone modifications might also be involved in regulating the assembly of PIC.

Discussion
Here, we report structures of Pol II PIC-nucleosome transcription complexes. The structures show that TFIIH directly interacts with the nucleosome in distinct conformations. They also indicate that a preferred orientation of the nucleosome with respect to the PIC is adopted that is characterized by multiple TFIIH-nucleosome contacts. Although the nucleosome may initially be found in different rotational states relative to the PIC, action of the Ssl2 translocase may lead to a preferred nucleosome orientation that allows for multiple TFIIH contacts as observed in complex B and C of this study. As TFIIH apparently has ceased to unravel nucleosomal DNA beyond SHL-5 in complex B, the translocase activity of TFIIH may not be sufficient to enable the PIC to pass through nucleosomal  Article https://doi.org/10.1038/s41594-022-00865-w sites with strong histone-DNA interactions (such as SHL-5 and -1). Passage of PIC through the +1 nucleosome may thus rely on assistance from chromatin remodelers and chromatin modifying complexes. Together with modeling, the observed structures with the stalled PIC on the nucleosome also provide a possible explanation for why the TSS is often located roughly 60 bp upstream of the dyad of the +1 nucleosome in yeast 15 . Before initiation of RNA chain synthesis, the yeast PIC scans downstream DNA for the TSS 35 . Modeling based on complex B indicates that scanning requires further progression of the PIC into the +1 nucleosome and detachment of three additional turns of DNA (Extended Data Fig. 10). This may be achieved by the ATP-dependent Ssl2 translocase whose activity is required for scanning 36,37 . We speculate that scanning may be impaired at the major histone-DNA interacting sites (for example, SHL-1) just upstream of the nucleosome dyad 23,38 , which then may trigger TSS usage and RNA chain initiation as suggested 13 . This model of nucleosome-defined TSS usage may explain why TSSs in yeast are generally located at a distance of roughly 60 bp from the dyad of the +1 nucleosome, even though the distance from the TSS to the TATA box varies in the range of roughly 40-120 bp (ref. 26 ). In our experimental system, the high stability of the nucleosome on the Widom-601 positioning sequence may have prevented scanning and led to a stable intermediate amenable to structure determination.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-022-00865-w.   12-subunit Pol II, TBP, TFIIA, TFIIB, TFIIE, TFIIF and TFIIH  were purified as described previously 2,4,39 . The DNA scaffolds containing the modified His4 promoter and Widom-601 sequence were synthesized by Integrated DNA Technologies, amplified and purified as described 40 . Xenopus laevis and S. cerevisiae histones were prepared and assembled into nucleosomes as described 41,42 . The sequences for the scaffold used in this study is listed with the Widom-601 sequence underlined. The TATA box and TSS are indicated in bold. The PIC-nucleosome complex was assembled according to a previously reported protocol 4 , with minor modifications. Briefly, scaffold containing a reconstituted nucleosome was incubated with TBP, TFIIA and TFIIB, whereas Pol II and TFIIF were incubated for 10 min at 25 °C. These two preparations were then combined and incubated for another 5 min. TFIIE and preassembled 10-subunit TFIIH were added to the mixture simultaneously and incubated for 5 min. NTPs at a concentration of 400 μM were added and the assembly was incubated for 1 h at 25 °C. The PIC-nucleosome samples were subjected to GraFix ultracentrifugation 43 at 137,600g (32,000 r.p.m. for SW60 rotor) for 16 h at 4 °C in a 15-40% (w/v) sucrose gradient with 0-0.1% glutaraldehyde crosslinker. Subsequently, the gradient solutions were fractionated and quenched with a mixture of 40 mM aspartate and 10 mM lysine for 10 min. Fractions were analyzed by native PAGE. Gels were stained with SYBR Gold (Invitrogen) and Coomassie brilliant blue. Peak fractions containing crosslinked PIC-nucleosome complex were dialyzed for 16 h in dialysis buffer (20 mM HEPES-Na pH 7.5, 50 mM KCl, 2 mM MgCl 2 , 1 mM TCEP) to remove sucrose. The dialyzed samples were concentrated to 0.2 mg ml −1 and used for grid preparation.

In vitro promoter-dependent transcription assay
Assays were performed as described 9 , with minor alterations. DNA scaffolds with or without nucleosome were prepared as described above. All scaffolds contained identical DNA sequences irrespective of the nucleosome component. Assembled scaffolds were stored in a low salt buffer (50 mM KCl, 5 mM K-HEPES pH 7.5, 0.025 mM EDTA). PIC was reconstituted on scaffold DNA essentially as reported 9 . All incubation steps were performed at 25 °C unless indicated otherwise. Per sample, 1.6 pM of TBP, 1.8 pM Pol II, 2.7 pM TFIIE and TFIIH, 9 pM TFIIF, 9 pM TFIIB and 18 pM TFIIA were used. Reactions were prepared in a sample volume of 23.8 μl with final assay conditions of 3 mM HEPES-K pH 7.9, 20 mM Tris-HCl pH 7.9, 60 mM KCl, 8 mM MgCl 2 , 2% (w/v) PVA, 3% (v/v) glycerol, 0.5 mM DTT, 0.5 mg ml −1 BSA and 20 units of RNase inhibitor. Samples were incubated for 45 min and transcription was started by adding 1.2 μl of 10 mM NTP solution and permitted to proceed for 60 min. Reactions were quenched with 100 μl of Stop buffer (300 mM NaCl, 10 mM Tris-HCl pH 7.5, 0.5 mM EDTA) and 14 μl of 10% SDS, followed by treatment with 4 μg of proteinase K (New England Biolabs) for 30 min at 37 °C. RNA products were isolated from the samples as described 9 , applied to urea gels (7 M urea, 1× TBE, 6% acrylamide:bis-acrylamide 19:1) and separated by denaturing gel electrophoresis (urea-PAGE) in 1× TBE buffer for 45 min at 180 V. Gels were stained for 30 min with SYBR Gold (Invitrogen) and RNA was visualized with a Typhoon 9500 FLA imager (GE Healthcare Life Sciences). The densities of the bands on the gels were quantitated with ImageJ.

Cryo-EM analysis and data processing
Four microliters of PIC-nucleosome samples were applied to glow-discharged UltrAuFoil 2/2 grids (Quantifoil). After incubation on grids for 10 s, samples were blotted for 4 s and vitrified by plunging into liquid ethane via a Vitrobot Mark IV (FEI) operated at 4 °C and 100% humidity. Cryo-EM data were collected on a Titan Krios G2 transmission electron microscope (FEI) operated at 300 keV, equipped with a K3 summit direct detector and a GIF quantum energy filter (Gatan). Automated data acquisition was performed using SerialEM software at a nominal magnification of ×81,000, corresponding to a physical 1.05 Å per pixel. Image stacks of 40 frames were collected in counting mode over 1.5 s at a defocus range from 0.8-2.0 μm. The dose rate was 27 e − /Å 2 per second resulting in 1.02 e − /Å 2 per frame. Totals of 26,764, 31,286 and 15,515 videos were collected for complexes A, B and C, respectively.
Image stacks were motion-corrected, contrast-transfer function corrected, dose-weighted and auto-picked using Warp 44 . Image processing was performed with RELION v.3.0.5 (ref. 45 ). Particles were extracted using a box size of 400 2 or 360 2 pixels, and normalized. Reference-free 2D classification was performed to remove poorly aligned particles. An ab initio model generated with cryoSPARC 46 was used for subsequent 3D classification. All classes containing PIC-nucleosome density were combined and used for a global 3D refinement. To obtain an improved density map for cPIC, TFIIH and the nucleosome, particles were subjected to focused 3D classification without image alignment. All classes containing good cPIC density were subjected to contrast-transfer function refinement, Bayesian polishing and 3D refinement. Postprocessing of refined models was performed using automated B factor determination in RELION and reported resolutions were based on the gold-standard Fourier shell correlation 0.143 criterion. The density of TFIIH was further improved by applying signal subtraction and focused refinement. Local resolution estimates were obtained using the built-in local resolution estimation tool of RELION and previously estimated B factors.

Model building
The structural models were built into the density of the final reconstructions with the best local resolutions for PIC or the TFIIH-nucleosome complex. A nucleosome structure with 145 bp Widom-601 DNA (Protein Data Bank (PDB) 7OHC) 24 and the structure of yeast PIC (PDB 7O73) 9 were placed into the density maps by rigid-body fitting in Chimera 47 , followed by manually adjustment and connection of linker DNA. The assignment of Rpb6 NTT was guided by the densities of bulky side chains, the crosslinking mass spectrometry and secondary structure prediction. The models were subjected to alternating real-space refinement and manual adjustment using PHENIX 48 and COOT 49 , resulting in very good stereochemistry as assessed by Molprobity 50 .

Crosslinking and mass spectrometry
Samples for mass spectrometry were prepared by sucrose gradient centrifugation as described above for cryo-EM sample preparation without glutaraldehyde. Fractions containing fully assembled complexes were pooled, subjected to chemical crosslinking using zero-length crosslinker EDC (100 mM) and NHS (100 mM) for 1 h at room temperature, and quench with 100 mM ammonium bicarbonate. The samples were adjusted to 8 M urea, 50 mM NH 4 HCO 3 , 10 mM DTT followed by an incubation for 30 min at 37 °C. Proteins were

Nature Structural & Molecular Biology
Article https://doi.org/10.1038/s41594-022-00865-w alkylated in the presence of 40 mM iodoacetamide for another 30 min at 37 °C in the dark and the reaction was quenched by 10 mM DTT for 5 min at 37 °C. The reaction volume was adjusted to reach a final concentration of 1 M urea and 50 mM NH 4 HCO 3 . Nucleic acids fragments within the PIC-nucleosome complex were digested for 30 min at 37 °C by the addition of 0.1 M MgCl 2 to a final concentration of 1 mM in the reaction and 500 U of universal nuclease (Pierce, catalog no. 88702, 250 U μl −1 ). Trypsin digest was performed overnight at 37 °C with 5 μg of trypsin (Promega, V5111). Peptides were acidified with 4 μl of 100% formic acid, desalted on MicroSpin columns (Harvard Apparatus) following the manufacturer's instructions and vacuum dried. Dried peptides were dissolved in 50 μl of 30% acetonitrile/0.1% TFA and peptide size exclusion (pSEC, Superdex Peptide 3.2/300 column) on an ÄKTA micro system (GE Healthcare) was performed to enrich for crosslinked peptides at a flow rate of 50 μl min −1 . Fractions of 50 μl were collected. The first 21 fractions enriched in crosslinked peptides were vacuum dried and dissolved in 5% acetonitrile/0.05% TFA (v/v) for analysis by liquid chromatography with tandem mass spectrometry.
Crosslinked peptides derived from pSEC were analyzed as technical duplicates on Q Exactive HF-X hybrid quadrupole-orbitrap mass spectrometer (Thermo Scientific), coupled to a Dionex UltiMate 3000 UHPLC system (Thermo Scientific). The sample was separated on an in-house-packed C18 column (ReproSil-Pur 120 C18-AQ, 1.9 μm pore size, 75 μm inner diameter, 30 cm length, Dr. Maisch GmbH) at a flow rate of 300 nl min −1 . Sample separation was performed over 60 min using a buffer system consisting of 0.1% (v/v) formic acid (buffer A) and 80% (v/v) acetonitrile, 0.08% (v/v) formic acid (buffer B). The main column was equilibrated with 5% B, followed by sample application and a wash with 5% B. Peptides were eluted by a linear gradient from 15-48% B or 20-50% B. The gradient was followed by a wash step at 95% B and re-equilibration at 5% B. Eluting peptides were analyzed in positive mode using a data-dependent top-30 acquisition methods. MS1 and MS2 resolution were set to 120,000 and 30,000 full width at half maximum, respectively. Precursors selected for MS2 were fragmented using 30% normalized, higher-energy collision induced dissociation fragmentation. Allowed charge states of selected precursors were +3 to +7. Further tandem mass spectrometry parameters were set as follows: isolation width, 1.4 m/z; dynamic exclusion, 10 s and maximum injection time (MS1/MS2), 60/200 ms.
For identification of crosslinked peptides, raw files were analyzed by pLink (v.2.3.5), pFind group 51 using EDC as crosslinker and trypsin/P as digestion enzyme with maximal three missed cleavage sites. The search was conducted against a customized protein database containing all proteins within the complex (Supplementary Table 1). Carbamidomethylation of cysteines was set as a fixed modification, oxidation of methionines and acetylation at protein N termini were set as a variable modification. Searches were conducted in combinatorial mode with a precursor mass tolerance of 10 ppm and a fragment ion mass tolerance of 20 ppm. The false discovery rate was set to 0.05 (separate mode). Spectra of both technical duplicates were combined and evaluated manually.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The electron density reconstructions and final models were deposited with the EM Data Bank (accession codes EMD-14927, 14928 and 14929) and with the PDB (accession codes PDB 7ZS9, 7ZSA and 7ZSB). All mass spectrometry raw files were deposited to the ProteomeXchange Consortium (https://www.proteomexchange.org/) via the PRIDE 52 partner repository with the dataset identifier PRIDE: PXD029840. Source data are provided with this paper.