Structural basis of Integrator-dependent RNA polymerase II termination

Molecular cloning and protein expression and purification

The constructs used for expressing Integrator were described previously¹ with some modification. In brief, INTS1 and INTS15 cDNA sequences were codon-optimized for protein expression in T. ni (Hi5 insect cells) and the cDNAs were purchased from Integrated DNA Technologies (IDT). Owing to its size, the cDNA for INTS1 was divided into three fragments for synthesis. The codon-optimized INTS1 cDNA fragments were cloned into vector 438-C (Addgene 55220) and combined with INTS12, which was cloned in vector 438-A (Addgene 55218), to create the INTS1–INTS12 construct. The cDNA for INTS15 was cloned into vector 438-C. Expression constructs for the INTS2–INTS3–INTS5–INTS6–INTS7–INTS8 subcomplex, the cleavage module, the INTS10–INTS13–INTS14 module and the PP2A complex were described previously¹.

Full-length human NABP2 (Q9BQ15-1) and INIP (Q9NRY2-1) cDNAs were codon-optimized for Hi5 insect cells, purchased from IDT and individually cloned into vector 438-B (Addgene 55219) by ligation-independent cloning⁵⁵.

Baculoviruses for protein expression in insect cells were generated in SF9 and SF21 cells (Thermo Fisher Scientific) using a previously described protocol⁵⁶. We expressed the eight-subunit Integrator core by co-infecting Hi5 cells (Thermo Fisher Scientific) with two baculoviruses, one expressing the INTS2–INTS3–INTS5–INTS6–INTS7–INTS8 subcomplex with an N-terminal 6×His-MBP tag on INTS5 and the other containing the INTS1–INTS12 construct. We expressed the tail module (INTS10–INTS13–INTS14–INTS15) by co-infecting Hi5 cells with baculoviruses containing 6×His-MBP–INTS15 and the INTS10–INTS13–INTS14 module. The PP2A complex and the Integrator cleavage module were expressed as previously described¹. NABP2 and INIP were also expressed in Hi5 cells using baculoviruses generated from their respective constructs.

The eight-subunit Integrator core, the cleavage module (and its mutants) and PP2A were purified using the published protocols¹. The tail module was purified essentially as described for the INTS10–INTS13–INTS14 module¹ except that we used amylose instead of Ni affinity.

To prepare NABP2, Hi5 insect cells expressing the protein were collected by centrifugation at 238g for 30 min in a high-speed centrifuge (Beckman Coulter) operated at 4 °C. The supernatant was discarded and the cell pellet from a 1.2-l culture was resuspended in 80 ml lysis buffer (50 mM Tris-HCl pH 7.5, 800 mM NaCl, 20 mM imidazole, 10% (v/v) glycerol, 2 mM EDTA and 5 mM DTT). Cells were lysed by sonication with 30% amplitude for 2 min with a 0.4-s pulse on and a 0.6-s pulse off using a Branson digital sonifier. The lysate was spun at 87,207g in a high-speed centrifuge at 4 °C for 1 h and filtered with a 0.8-μm syringe filter to remove cell debris. The clarified lysate was applied to a pre-equilibrated 5-ml HisTrap HP column (Cytiva) at a flow rate of 1.5 ml per min and the column was washed with 100 ml lysis buffer. The bound protein was eluted from the column using a gradient from 0–100% over 18 column volumes of an Ni elution buffer (50 mM Tris-HCl pH 7.5, 800 mM NaCl, 500 mM imidazole, 10% (v/v) glycerol, 2 mM EDTA and 5 mM DTT). The fractions containing NABP2 were combined and treated with 5 mg 6×His-TEV protease and lambda protein phosphatase, and dialysed overnight against 800 ml low-salt buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 20 mM imidazole, 10% (v/v) glycerol, 2 mM EDTA and 5 mM DTT) at 4 °C in a 7-kDa molecular weight cut-off (MWCO) SnakeSkin dialysis tubing (Thermo Fisher Scientific). The digested sample was applied to a 5-ml HisTrap HP column to remove uncleaved protein and TEV protease. The flow-through fraction was applied to a pre-equilibrated 5-ml HiTrap SP HP column. NABP2 was recovered in the flow-through, concentrated in an Amicon 15-ml centrifugal filter (10 kDa MWCO) (Millipore) to around 1.0 ml and applied to a Superdex 75 10/300 GL column (Cytiva) equilibrated in 50 mM HEPES pH 7.5, 300 mM NaCl, 10% (v/v) glycerol and 2 mM TCEP. Peak fractions were analysed by SDS–PAGE, and the fractions that contain pure NABP2 were concentrated, aliquoted, flash-frozen and stored at −80 °C.

To prepare INIP, Hi5 cells were collected, lysed, filtered and clarified using the protocol for NABP2. Cells were resuspended in a low-salt lysis buffer (20 mM MES pH 6.1, 150 mM NaCl, 20 mM imidazole and 2 mM DTT). Cleared lysate was applied to a pre-equilibrated 5-ml HisTrap HP column (Cytiva) at a flow rate of 1.5 ml per min and the column was washed with 100 ml low-salt lysis buffer. Elution was performed using Ni elution buffer (20 mM MES pH 6.1, 150 mM NaCl, 500 mM imidazole and 2 mM DTT), and peak fractions were analysed by SDS–PAGE. Fractions containing INIP were combined with lambda phosphatase and dialysed overnight at 4 °C in SnakeSkin dialysis tubing (3.5 kDa MWCO) (Thermo Fisher Scientific) against 800 ml low-salt lysis buffer. After dialysis, the protein was applied to a pre-equilibrated 5-ml HiTrap SP HP cation-exchange column. INIP-containing fractions after ion exchange were pooled and concentrated in an Amicon 15-ml centrifugal filter (3 kDa MWCO) (Millipore) to around 1.0 ml and applied to a Superdex 75 10/300 GL column (Cytiva) equilibrated in 50 mM HEPES pH 7.5, 300 mM NaCl, 10% (v/v) glycerol and 2 mM TCEP. Peak fractions were concentrated, aliquoted, flash-frozen in liquid nitrogen and stored at −80 °C.

Preparation of mammalian Pol II (ref. ⁵⁷) human DSIF (ref. ³⁷), NELF (ref. ⁵) and human histones⁵⁸ was done as described in the corresponding references.

Nucleosome preparation

DNA fragments for nucleosome reconstitution were generated by PCR as described previously⁵⁹. In brief, nucleosome DNA was amplified from a vector containing the 145-bp Widom 601 sequence and a 40-bp run-up sequence upstream of the Widom 601. A 50-ml PCR was performed using the following primers: forward: 5′-GCAGTCCAGTTACGCTGGAGTC-3′ and reverse: 5′ATCAGAATCCCGGTGCCG −3′. The sequence of the PCR product is 5′-GCAGTCCAGTTACGCTGGAGTCTGAGGCTCGTCCTGAATGATATGCGGCCTCACGAAGCGTAGCATCACTGTCTTGTGTTTGGTGTGTCTGGGTGGTGGCCGATATCGATGTATATATCTGACACGTGCCTGGAGACTAGGGAGTAATCCCCTTGGCGGTTAAAACGCGGGGGACAGCGCGTACGTGCGTTTAAGCGGTGCTAGAGCTGTCTACGACCAATTGAGCGGCCTCGGCACCGGGATTCTGAT-3′. PCR product purification, TspRI digestion, octamer formation, nucleosome reconstitution and purification of nucleosome with a PrepCell system were performed as previously described⁵⁹. The concentration of the reconstituted and purified nucleosome was determined using the sum of the molar extinction coefficients of DNA and octamer at 280 nm and the absorbance of the nucleosome at this wavelength. Super-helical locations were assigned on the basis of previous publications (see references in ref. ⁵⁹).

RNA extension assays

We performed in vitro RNA extension assays to identify the NELF-dependent Pol II pause site in the nucleosome. A 5′ Cy5-labelled RNA (5′-Cy5/rUrUrArUrCrArCrUrGrUrC-3′) that anneals at the TspRI-generated overhang in the run-up to the nucleosome was used to load Pol II onto the nucleosomal substrate for RNA extension. Assays were performed in a volume of 10 µl in a final buffer containing 100 mM NaCl, 20 mM Na-HEPES, pH 7.4, 5 mM MgCl₂, 1 mM DTT and 4% glycerol. Depending on the reaction, RNA (480 nM) was incubated with either DNA substrate or nucleosomal substrate (240 nM) for 10 min on ice. Sus scrofa Pol II (300 nM) was added to the reaction and incubated for 10 min on ice. DSIF (600 nM), TFIIS (180 nM) and buffer were added to the samples. Transcription was initiated by adding 0.5 mM each of GTP, CTP, UTP and ATP or 3′-dATP with or without NELF (600 nM). After a 60-min incubation at 30 °C, 5 µl transcription reaction was quenched with 5 µl of a 2× Stop buffer (6.4 M urea, 50 mM EDTA, pH 8.0 and 2× TBE buffer). Quenched samples were treated with 1.6 units of proteinase K (NEB) for 30 min at 37 °C and denatured for 10 min at 95 °C before the fluorescent RNA was separated using denaturing PAGE (8 M urea, 1× TBE buffer, 12% acrylamide:bis-acrylamide 19:1 gel; run for 42 min at 300 V in 0.5× TBE buffer). RNA products were visualized by their Cy5 label in a Typhoon 9500 FLA imager.

Reconstitution of complexes for cryo-EM and XL-MS analyses

Pre-termination complex

The RNA extension assay showed that the presence of NELF impaired Pol II transcription to the nucleosomal substrate as compared with Pol II–DSIF alone. We thus used a two-step procedure to form the pre-termination complex for cryo-EM and XL-MS analyses. First, a transcribed Pol II–DSIF–Nuc complex was reconstituted in a buffer containing 100 mM NaCl, 20 mM HEPES, pH 7.4, 5 mM MgCl₂, 1 mM DTT and 4% (v/v) glycerol. The 5′ Cy5-labelled RNA (3.2 µM) and the nucleosomal substrate (1.6 µM) were mixed and incubated for 5 min on ice. S. scrofa Pol II (2 µM) was added to the reaction and incubated for another 5 min on ice. We added DSIF (6 µM) and 3′-dATP (1 mM) and the mixture was incubated for 10 min at 30 °C. Transcription was started by adding TFIIS (1.2 µM), CTP, GTP and UTP (each 1 mM), and it proceeded for 60 min at 30 °C in a final volume of 50 µl. In parallel, we mixed 3.8 µM each of the eight-subunit Integrator core, cleavage module, tail module and PP2A in a final volume of 80 µl and incubated on ice to form the integrator–PP2A complex. We used an INTS11(E203Q) mutant that has reduced catalytic activity in all complexes formed for cryo-EM and XL-MS analyses.

In the second step, we added the preformed Integrator–PP2A complex and NELF (2 µM) to the transcribed Pol II–DSIF–Nuc complex in a final buffer comprising 156 mM NaCl, 28 mM HEPES, pH 7.4, 5 mM MgCl₂, 1 mM DTT and 4% glycerol in a final volume of 200 µl. We incubated the mixture for 30 min at 30 °C. The assembled pre-termination complex was purified using a 4-ml 10–40% glycerol gradient as previously described¹. Samples removed after each step were analysed using denaturing PAGE.

PEC–Integrator–PP2A–SOSS and free Integrator–PP2A–SOSS complexes

The PEC–Integrator–PP2A–SOSS complex from which we obtained the post-termination structure was formed essentially as described for the PEC–Integrator–PP2A complex using the published DNA scaffolds^1,5 and a variant of the HIV TAR RNA that does not form a secondary structure. The RNA has the following sequence, 5′-/6-FAM/rUrUrArArGrGrArArUrUrArArGrUrCrGrUrGrCrGrUrCrUrArArUrArAr CrCrGrGrArGrArGrGrGrArArCrCrCrArCrU-3′. We pre-incubated 3.8 µM each of the eight-subunit Integrator core, Integrator cleavage module, tail module, PP2A, NABP2 and INIP in a final volume of 80 µl on ice to form the Integrator–PP2A–SOSS complex. We formed the PEC using 0.6 µM of Pol II, 1.2 µM each of nucleic acids and 1.8 µM of DSIF and NELF. The preformed Integrator–PP2A–SOSS complex was added to the PEC in a final volume of 163 µl. We incubated the mixture for 30 min at 30 °C and applied it to a 10–40% glycerol.

The free Integrator–PP2A–SOSS complex was formed by mixing 3.8 µM of each of the Integrator–PP2A subcomplexes with 3.8 µM of NABP2 and INIP on ice for 60 min. The complex was purified using a 10–40% glycerol gradient.

Cryo-EM sample preparation

Peak fractions from the glycerol-gradient analyses of the pre-termination, PEC–Integrator–PP2A–SOSS and free Integrator–PP2A–SOSS complexes were separately cross-linked using 0.2% (v/v) glutaraldehyde for 10 min on ice. The cross-linking reaction was quenched using 100 mM Tris-HCl (pH 8) for 10 min on ice. The cross-linked cryo-EM samples were dialysed for 4–6 h against a buffer containing 20 mM HEPES pH 7.4, 150 mM NaCl, 1% (v/v) glycerol, 3 mM MgCl₂, 1 mM DTT and 0.01% (w/v) CHAPS using a 20 kDA MWCO Slide-A-Lyzer MINI Dialysis Unit (Thermo Fisher Scientific). A 2.6–2.8-µm-thin carbon film was floated on the dialysed cryo-EM samples and incubated for 5–15 min depending on the concentration of the sample before cross-linking. The floated carbon film was transferred onto a Quantifoil R3.5/1 copper mesh 200 grid and instantly blotted for 2 s with blot force 5 before being vitrified in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific). The vitrobot was operated at 4 °C and 95–100% humidity.

Cryo-EM data collection and processing

All cryo-EM data were acquired at a nominal magnification of 81,000×, corresponding to a calibrated pixel size of 1.05 Å per pixel, using a K3 direct electron detector (Gatan) on a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV. Images were collected in EFTEM mode using a Quantum LS energy filter (Gatan) with a slit width of 20 eV. A defocus range of −0.5 to −2.0 μm was applied during data collection and images were recorded in electron counting mode. The SerialEM software⁶⁰ was used for automated data acquisition. Motion correction of collected movies, dose weighting, constrast transfer function (CTF) estimation and particle picking were performed using Warp⁶¹.

For the pre-termination complex sample, we collected 59,687 micrographs with a dose rate of 14.86 e⁻ per pixel per s for 3 s, resulting in a total dose of 40.44 e⁻ per Ų that was fractionated into 50 movie frames. Micrographs with bad CTF fits in Warp were excluded from further processing. We extracted 9,107,060 picked particles with a box size of 500 pixels and binned 2× to a pixel size of 2.1 Å per pixel using RELION 3.1 (ref. ⁶²). These particles were subjected to heterogenous refinement in CryoSPARC (ref. ⁶³) using initial models generated from our previous data¹. The selected good particles that had cryo-EM density for the PEC and the Integrator–PP2A complex were further sorted using two-dimensional (2D) classification in CryoSPARC. We identified 1.3 million good particles with this procedure, which were re-extracted in RELION 3.1 (ref. ⁶²) without binning. We performed one round of three-dimensional (3D) classification in RELION to eliminate Integrator–PP2A particles that have only a weak Pol II density, resulting in 278,693 particles. This set of particles were CTF refined and polished in RELION 3.1 to obtain a 3.8-Å reconstruction encompassing the PEC and Integrator–PP2A. We applied soft masks around various parts of this map and performed signal subtraction, 3D classification and refinement in RELION 3.1. This produced good focused refined maps better than 3.5 Å that aided model building. A subset of 80,717 particles was obtained from focused classification with a mask around the PEC. We reverted the signal for these particles and performed global 3D refinement to obtain the overall reconstruction for the pre-termination complex at a resolution of 4.1 Å (map 1). The density for the nucleosome was very weak, showing that it is highly dynamic in this complex.

For the PEC–Integrator–PP2A–SOSS sample that led to the post-termination structure of the Integrator–PP2A–SOSS–CTD complex, 52,976 micrographs were acquired. Each micrograph was acquired with a 2.84-s exposure at a dose rate of 15.50 e⁻ per pixel per s, resulting in a total dose of 39.93 e⁻ per Ų that was divided into 40 movie frames. We excluded micrographs with a bad CTF fit and extracted 9,165,848 particles in RELION 3.1 (ref. ⁶²) using a 480-pixel box size. The extracted particles were binned to a pixel size of 2.1 Å per pixel to speed up initial sorting. Bad particles were removed using iterative 3D and 2D classification in CryoSPARC as described above, resulting in 832,842 good particles. This set of particles was taken through Bayesian polishing, CTF refinement and 3D refinement procedures in RELION 3.1. We identified three main classes when we applied 3D classification without image alignment. The first two classes are similar to the published PEC–Integrator–PP2A complex¹. The third class of 236,382 particles led to the reconstruction of the post-termination complex. We applied a soft mask around this class to subtract out the weak Pol II density that could not be resolved owing to flexibility. Three-dimensional refinement of the signal-subtracted particles led to a 3.7-Å overall reconstruction for the post-termination complex (map 2). We improved the resolution of local regions of the map using signal subtraction, 3D classification and refinement.

We collected 47,268 micrographs using a grid prepared from the free Integrator–PP2A–SOSS complex. The images were collected with 2.82-s exposures with a dose rate of 15.83 e⁻ per pixel per s and a total dose of 40.49 e⁻ per Ų that was split into 40 movie frames. We extracted 7,014,615 particles with a box size of 480 pixels that we binned to 2.1 Å per pixel after preprocessing the data in Warp. We used 3D and 2D classification in CryoSPARC to remove junk and broken particles, resulting in 2,335,349 good particles that we re-extracted without binning and subjected to CTF and 3D refinement procedures in RELION 3.1. We obtained a 3.1-Å reconstruction using the above steps. We performed signal subtraction with recentring of the particles followed by 3D refinement to obtain cryo-EM maps better than 2.9 Å for various regions of Integrator–PP2A. For the tail module, we performed 3D classification on the signal-subtracted particles to identify a subset set of 118,383 particles that refined to 6.1 Å. Further classification of these particles did not improve the resolution of the tail module.

Model building and refinement

To build a model for the pre-termination complex, we first fitted the model of PEC–Integrator–PP2A (PDB ID: 7PKS) into map 1 using ChimeraX (ref. ⁶⁴) and adjusted the fit using focused refined maps. Manual adjustments to the model were made in Coot (ref. ⁶⁵) after initial rounds of ISOLDE flexible fitting⁴⁰. We determined the sequence register of nucleic acids bound in the Pol II cleft using the 3.2-Å PEC focused refinement map. Following this register, we built the DNA–RNA hybrid and extended the downstream DNA. The downstream nucleosome was modelled by rigid-body-docking a structure of the nucleosome (PDB ID: 7OHC) into the low-pass-filtered version of map 1 using the sequence register from the downstream DNA.

For the tail module, we rigid-body-docked AlphaFold2 (ref. ³⁹) models for INTS10 and INTS15 into the focused refined map of this module and adjusted them in Coot and ISOLDE. We predicted the interface between the C terminus of INTS10 and INTS14 using Colabfold (ref. ⁴¹). The predicted model was fitted in the focused refined map and adjusted using ISOLDE. The available crystal structure of INTS13–INTS14 (PDB ID: 6SN1)³⁴ was aligned on our model of INTS10–INTS14 to derive the correct orientation of the INTS13–INTS14 sting. This naturally placed the INTS13–INTS14 sting inside the low-pass-filtered map 1. The INTS13 CMBM, INTS6 inhibitory loop and DSS1 models were copied from the structure of the free Integrator–PP2A and manually adjusted using Coot. Various parts of the model were refined against respective focused refined maps using the phenix.real_space_refine tool in the PHENIX package^42,66. The final model was refined against map 1 with reference model restraints to account for regions with weak cryo-EM density in the consensus refinement.

We built a model for the post-termination complex by copying the Integrator–PP2A structure from the pre-termination complex structure above. The model was first fitted into the overall map of the post-termination complex (map 2) and adjusted in Coot and ISOLDE using the focused refined maps to fit residue side chains. INTS1 residues 1–866 and the INTS13–INTS14 sting were removed from the model because cryo-EM density for these regions was lacking. The crystal structure of the SOSS complex containing INTS3(1–501), NABP2 and INIP (PDB ID: 4OWW)³¹ was docked into the focused refinement map around INTS3 and the SOSS factors. Regions lacking cryo-EM density were removed and the model was further adjusted using ISOLDE. A combined model for the post-termination complex was created in map 2 and subjected to real-space refinement in the PHENIX package^42,66.

The model for free Integrator–PP2A was built by first docking a structure of Integrator–PP2A that was copied from the pre-termination complex structure into map 3. ISOLDE was used to adjust the model to fit the map. The higher-resolution focused refined maps were used to fit side chains. Our XL-MS data suggested that the INTS6 inhibitory loop binds in front of the PP2A-C active site. We used Colabfold to predict possible interfaces between the flexible INTS6 C terminus and PP2A-C. We found one predicted model that perfectly matched our cryo-EM density and used it to model the INTS6 inhibitory loop. The interface between INTS11 and the INTS13 CMBM was also at first predicted using Colabfold and subsequently adjusted using ISOLDE and Coot. To identify the DSS1 peptide in our cryo-EM density and build a model for it, we subjected the 2.7-Å focused refined map around INTS1–INTS2–INTS7 to sequence-free de novo modelling using ModelAngelo (ref. ⁶⁷). The software identified and modelled the conserved C-terminal part of DSS1 into our cryo-EM density. The modelled part of the T. ni DSS1 corresponds to residues 35–60 of the human orthologue. We did not assign a sequence numbering because there is no database with annotated T. ni DSS1. An AlphaFold2 model for the N terminus of INTS1 (residues 143-905) was rigid-body-docked into the overall map to complete the model. The final model was refined in real space using PHENIX^42,66.

The following regions were built as backbone traces where applicable because side-chain information was absent in our cryo-EM density maps. INTS1(143–906), INTS10, INTS13(11-564), INTS14 and INTS15.

Integrator RNA degradation assay

RNA cleavage and degradation activity of Integrator was tested using the RNA cleavage assay preciously described¹. In brief, a completely complementary template and non-template DNA and a single-stranded RNA that anneals to the template DNA were purchased from IDT. The nucleic acids have the following sequences: template DNA, 5′-GCTTTATTGAGGCTTAAGCAGTGGGTTCCAGGTACTAGTGTACAGCTATCGTAAGCTATCGTAGGCAAGGTCCACTGACT/3Bio/-3′; non-template DNA, 5′-AGTCAGTGGACCTTGCCTACGATAGCTTACGATAGCTGTACACTAGTACCTGGAACCCACTGCTTAAGCCTCAATAAAGC-3′; and RNA, 5′-rArGrUrCrGrUrGrCrGrUrCrUrArArUrArArCrCrGrGrArGrArGrGrGrAr ArCrCrCrArCrU/3Cy5Sp/-3′. Note that the RNA has a 3′ Cy5 fluorescent label for visualizing the Integrator cleavage products.

A 6× PEC master mix was prepared using 450 nM Pol II, 90 nM of each nucleic acid and 900 nM each of DSIF and NELF in a final volume of 40 µl. Aliquots of the PEC master mix (6.7 µl) were treated individually with 150 nM preformed Integrator–PP2A complexes containing wild-type INTS11, an INTS11(E203Q) mutant that has reduced enzymatic activity or an INTS11(D72K/H73A) mutant that is catalytically dead, or were not treated with Integrator. Reactions were performed in 40-µl final volumes in buffer R (20 mM HEPES pH 7.4, 150 mM NaCl, 10% (v/v) glycerol, 3 mM MgCl₂, 1 mM DTT, 1 U μl⁻¹ RNAsin plus (Promega)). For the free Integrator control, the preformed wild-type Integrator was mixed with the annealed DNA–RNA scaffold. All reactions were incubated at 30 °C for 1 h, quenched and analysed on a denaturing PAGE as described¹.

The above protocol was used for the time-course RNA cleavage and degradation by wild-type Integrator shown in Fig. 1d and Supplementary Fig. 2f,g, except for the following. For a more efficient PEC formation, the PEC scaffold that contains a mismatch bubble⁵ was used instead of the completely complementary scaffold. We used a 25-nt or 17-nt RNA that has a 3′ Cy5 label to show that Integrator cannot act on RNA that is covered inside Pol II. The RNA sequences are shown in Fig. 1d and Supplementary Fig. 2f,g.

Chemical cross-linking coupled with mass spectrometry

Chemical cross-linking was performed using the peak fractions from the pre-termination complex and the PEC–Integrator–PP2A–SOSS glycerol-gradient ultracentrifugation. For each complex, we ran two gradients using the protocol described above and pooled the peak fractions. This was required to get sufficient material for XL-MS. The pooled peak fractions were cross-linked with 3 mM BS3 for 30 min at 30 °C and quenched with 100 mM Tris-HCl (pH 8.0). Mass spectrometry was performed as previously described¹, except for the following. For PEC–INT–PP2A–SOSS BS3-cross-linked peptides were pre-fractionated by size exclusion or by C18 basic pH reverse-phase chromatography (bRP). For PEC–Nuc–INT–PP2A, only bRP pre-fractionation was performed. Both samples were measured in triplicate in a Thermo Orbitrap Exploris mass spectrometer without FAIMS installed (Thermo Fisher Scientific). BS3-mediated protein-protein cross-links were identified using pLink 2.3.11 (http://pfind.org/software/pLink/).

Visualization

Protein sequence alignments were made with Multalign (ref. ⁶⁸) and visualized with ESPRIPT 3.0 (ref. ⁶⁹). Structural figures were made with UCSF Chimera (ref. ⁷⁰) and ChimeraX (ref. ⁶⁴). The Supplementary Videos were made in ChimeraX. Please note that Supplementary Video 4 was made by interpolating between the various conformations of proteins in our structures and the trajectory of protein domains may not necessarily reflect intermediate conformations.

Reporting summary

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