Main

In eukaryotes, the removal of noncoding introns from pre-mRNAs is catalyzed by the large and dynamic spliceosome complex1. The spliceosome assembles de novo on each intron from small nuclear ribonucleoprotein particles (snRNPs) and numerous protein factors. The 39-subunit U4/U6.U5 tri-snRNP is the largest preassembled building block of the spliceosome, which joins the splicing pathway at the precatalytic (pre-B) stage and delivers two components of the RNA catalytic core, the U5 and U6 snRNAs, in their inactive configurations requiring further remodeling2,3. Despite recent progress in the mechanistic understanding of spliceosome assembly, the biogenesis and recycling of its building blocks remain elusive.

Several factors associate with the tri-snRNP components but are absent in mature particles; hence, they are believed to play roles in tri-snRNP biogenesis and/or recycling. These include the U5 snRNP binding proteins: AAR2 (refs. 4,5), CD2BP2 (U5-52K; Saccharomyces cerevisiae Lin1)6,7, TSSC4 (refs. 8,9) and ZNHIT2 (refs. 10,11,12), as well as the U4/U6 annealing factor SART3 (S.cerevisiae Prp24)13,14,15. Mechanistically, the exact roles and the interplay of these assembly factors remain poorly understood.

The 20S U5 snRNP isolated from HeLa cells contains at least 16 subunits, including its hallmark protein CD2BP2 (refs. 16,17), which also plays a role in the binding of the CD2 receptor18. Conditional knockout (KO) of CD2BP2 in mice leads to growth defects and premature death during embryonic development19. In its role as a splicing factor, CD2BP2 binds to DIM1 (U5-15K) with its GYF domain, forming a protein–protein interface that differs from the canonical binding mode to sequence motifs in the CD2 antigens16,18. Lin1, the yeast homolog of CD2BP2, was reported to bind PRP8, suggesting a possible mode of its recruitment to the U5 snRNP complex20.

Although 20S U5 snRNP was first isolated several decades ago17, its molecular structure and the function of CD2BP2 remain unknown. In this Brief Communication, we investigate how CD2BP2 interacts with other components of the U5 snRNP and how it facilitates tri-snRNP formation.

First, we analyzed the steady-state composition of the spliceosomal snRNPs in the absence of CD2BP2. We purified snRNP using anti-2,2,7-trimethylguanosine (TMG) antibody-coupled resin from nuclear extracts (NE) prepared either from wild-type (WT) HEK293T cells or a homozygous CRISPR–Cas9 CD2BP2 KO cell line (CD2BP2KO; Extended Data Fig. 1a–c). The composition of both samples was compared by quantitative mass spectrometry (Fig. 1a). As expected, we observe a clear depletion of the CD2BP2 in the KO sample as well as a subtle, but consistent, underrepresentation of nearly all U5 snRNP subunits in the KO condition. U4/U6 and tri-snRNP specific components are also affected, yet, to a lesser extent, while the U2 snRNP proteins remained virtually unchanged, as their assembly into snRNPs is not expected to depend on CD2BP2. As such, these results point to a subtle defect in the U5 and U4/U6.U5 tri-snRNP assembly in the absence of CD2BP2. Yet, we could not observe a noteworthy impact on the cell viability and in vitro splicing efficiency under the conditions tested (Extended Data Fig. 1e). Interestingly, another U5 snRNP assembly factor, AAR2, is significantly enriched in snRNPs isolated from the CD2BP2KO cells. This could be due to the upregulation of the AAR2 in the absence of CD2BP2 or due to a failure in the CD2BP2-dependent conversion of AAR2-containing U5 snRNP into 20S U5 snRNP during the biogenesis. The latter seems more probable, as the expression levels of AAR2 in NE of WT and KO cell lines are largely unchanged (Extended Data Fig. 1g). This data provide evidence that CD2BP2 is indeed involved in the U5 snRNP assembly and establish a functional link to another U5 snRNP assembly factor AAR2.

Fig. 1: Cryo-EM structure of the 20S U5 snRNP.
figure 1

a, Quantitative mass spectrometry analysis of snRNPs isolated via TMG agarose from WT or CD2BP2-KO HEK293T cells. A moderated two-sided t-test was applied for statistical analysis. b, Experimental cryo-EM map of the 20S U5 snRNP colored by the subunit identity fitted into the low-pass filtered map at the lower contour level. c, Atomic model of the 20S U5 snRNP shown in the same orientation as in b. d, Orthogonal view of the atomic model. e, Zoomed-in view of the PRP8–CD2BP2 interaction highlighting the extended hook-like domain. f, Domain architecture of CD2BP2.

Source data

Full size image

Next, to gain insights into the structure of the 20S U5 snRNP, we engineered HEK293F cells to express a 3xFLAG_TEV_SBP-tagged CD2BP2 and used it to purify a 17-subunit complex containing the U5 snRNA, seven Sm core proteins and nine other factors (Extended Data Table 1 and Extended Data Fig. 2). The composition of the complex is in good agreement with previous reports7,17 and, interestingly, includes an additional assembly factor, TSCC4 (not resolved in the structure)8,9. We determined a cryogenic electron microscopy (cryo-EM) structure of the CD2BP2-bound 20S U5 snRNP complex at the 3.1 Å resolution (Fig. 1, Table 1, Methods and Extended Data Figs. 3–5). The architecture of the 20S U5 snRNP closely resembles that of the U5 snRNP captured as a part of the tri-snRNP2,21 and in low-resolution U5 snRNP studies22. At least three major states are present in our 20S U5 snRNP reconstruction (Extended Data Fig. 6). State I contains most of the components and is referred to as the 20S U5 snRNP hereafter. State II is missing two helicases, BRR2 and DDX23, and may represent an earlier stage of the U5 snRNP assembly (Extended Data Fig. 6). State III contains particles missing the Sm ring, most probably damaged during the vitrification process. In all reconstructions, PRP8 provides the scaffold for the entire complex and interacts with multiple other subunits (Fig. 1). PRP6 is present in the sample, but only its N-terminal helices are visible, and the tetratricopeptide (TPR) repeat remains disordered. We observed two well-defined densities located near PRP8RT/En and PRP8Nterm domains, which were assigned to CD2BP2 domains D1 (62-130) and D2 (150-233), respectively (Fig. 1). The CD2BP2GYF domain (280-341) and its binding partner DIM1 are not visible in our structure. A hook-shaped extension of the CD2BP2D1 bridges the PRP8RT/En and PRP8Nterm domains and probably stabilizes their relative orientation, which differs from the one observed in the tri-snRNP (Fig. 1 and Extended Data Fig. 7a,b). CD2BP2D1 occupies the surface of PRP8 that accommodates several different factors during the splicing cycle, including AAR2 in the U5 snRNP precursor5,23, DIM1 in tri-snRNP and the precatalytic spliceosome2,21, as well as RNF113 in Bact24 and CWF19L2 in the postsplicing ILS complexes25 (Fig. 2 and Extended Data Fig. 7c).

Table 1 Cryo-EM data collection, refinement and validation statistics
Full size table
Fig. 2: Mutually exclusive interactions of assembly factors with PRP8 during tri-snRNP formation.
figure 2

a, AAR2 binding mode to PRP8RT/EN domain23. b, CD2BP2D1 occupies an overlapping surface of PRP8 in 20S U5 snRNP. c, DIM1 binding site in tri-snRNP21 is mutually exclusive with CD2BP2D1. d, PRP8 binding surface on DIM1 is occupied by the CD2BP2GYF domain in the CD2BP2–DIM1 binary complex16 and clashes with PRP8 when superimposed on DIM1 in tri-snRNP. e, A structural model of the tri-snRNP assembly. Left: an atomic model of the 20S U5 snRNP including the disordered DIM1, CD2BP2GYF and PRP6TPR domains. Right: an atomic model of the fully assembled tri-snRNP21. Recruitment of the U4/U6 di-snRNP to 20S U5 snRNP triggers the relocation of PRP6, allowing DIM1 to compete with CD2BP2 and displace it from its binding site on PRP8.

Full size image

Interestingly, the CD2BP2GYF domain delivers DIM1 to the U5 snRNP, which then competes with CD2BP2D1 for the same binding site on PRP8. The interface of DIM1 that contacts PRP8 in tri-snRNP is most probably occupied by the CD2BP2GYF domain, as shown in the crystal structure16. Therefore, CD2BP2 constitutes a two-layered buffer blocking the DIM1–PRP8 interaction by simultaneously binding to the interfaces on both PRP8 and DIM1 (Fig. 2). Based on our structural data, we believe that at least two functions of CD2BP2 should be considered. First, it facilitates the recruitment of PRP6 and DIM1, both of which are critical for the tri-snRNP formation. PRP6, together with PRP8, forms an interface that is necessary for PRP31 (and U4/U6 di-snRNP) anchoring, which is then further stabilized by the PRP6TPR domain, forming a bridge between U4/U6 and U5 snRNPs (Fig. 2). PRP6 and CD2BP2 interact directly7 (Extended Data Figs. 8 and 9). Therefore, CD2BP2-mediated prerecruitment of PRP6 to the U5 snRNP would probably enhance the efficiency of the tri-snRNP formation. Second, CD2BP2 acts as a placeholder preventing PRP8RT/En association with its numerous binding partners in a wrong spatiotemporal context. A similar mechanism is utilized by some factors involved in ribosome biogenesis26. Both of the above functions could serve to ensure the correct order of events during the formation of complex intersubunit interfaces within tri-snRNP.

The remaining question is how CD2BP2 displacement is regulated. Our data shows that DIM1 competes with CD2BP2 for the same binding site on PRP8. We could not locate DIM1 in our map, but we observed some additional, low-resolution density near U5 snRNA, which most probably belongs to DIM1/CD2BP2GYF (Extended Data Fig. 6 and 9). Our cross-linking mass spectrometry (XL-MS) data detect DIM1–CD2BP2GYF and 40K–CD2BP2GYF interactions consistent with this putative location (Extended Data Fig. 9). Therefore, it is possible that DIM1 remains constrained in this position and cannot engage in the competition with CD2BP2D1. Recruitment of the U4/U6 di-snRNP would trigger a large-scale movement of PRP6 (Fig. 2). Since PRP6 and CD2BP2 interact with one another (Extended Data Figs. 8 and 9), such movement of PRP6 could exert a force on CD2BP2, displacing it from PRP8 and liberating DIM1, allowing it to adopt its final location.

It has been previously shown that CD2BP2 undergoes phosphorylation, which in principle, could also regulate its displacement27,28. One of the putative phosphorylation sites lies at the interface between CD2BP2D1 and PRP8Nterm (Extended Data Fig. 8) and could potentially modulate their affinity. However, phosphorylation of CD2BP2 does not appear necessary for the in vitro reconstitution of the tri-snRNP, and its function remains unclear27.

Although our data indicate that CD2BP2 is required for the tri-snRNP formation and acts downstream of AAR2, we cannot discriminate whether its role concerns predominantly the initial biogenesis of the U5 snRNP or its potential recycling from postsplicing complexes. As such, more work is required to shed light on the interplay between these two factors and their roles in respective pathways.

Methods

CD2BP2 KO cell line generation

Two guide RNAs were designed to delete 420 bp of the genomic locus near the translation start site of the CD2BP2 gene and cloned into the PX458 vector (pSpCas9(BB)-2A-GFP; a gift from Feng Zhang; Addgene plasmid no. 48138) using pairs of annealed oligonucleotides as follows:

SG1_FW:CACCGaaagtgaccttccaaggcgt+SG1_REV:aaacacgccttggaaggtcactttC; SG2_FW:CACCGACACTCTTTGGATAGCGATG+SG2_REV:aaacCATCGCTATCCAAAGAGTGTC.

HEK293T cells (ATCC CRL-3216) were seeded in a 6-well plate at a density of 0.3 × 106 cells per well and incubated for 24 h in Dulbecco’s modified Eagle medium (Gibco) supplemented with 5% fetal bovine serum and penicillin/streptomycin (Thermo Fisher Scientific). Then, 1 µg of each guide RNA-containing PX458 plasmid was transfected into the cells using Lipofectamine 2000 (Thermo Fisher Scientific), following the manufacturer recommendations. After 5 days of growth, cells were trypsinized with trypsin-EDTA 0.05% (Thermo Fisher Scientific), and single cells showing green fluorescent protein (GFP) signal were sorted into 96-well plates using a BD FACSAria IIu (BD Biosciences) sorter. Clonal cell lines were expanded over the period of 2 weeks and analyzed for the presence of the desired deletion using polymerase chain reaction (PCR; 52K_FW: GATCCAGAGGGTCCGCTCC; 52K_REV: CCTTCCTCCATCTCCTCCTGC) and western blotting with anti-CD2BP2 antibodies (Thermo Fisher Scientific; PA5-59603; RRID:AB_2639539).

TMG agarose immunoaffinity chromatography and quantitative mass spectrometry

NEs from HEK293T WT and CD2BP2KO were prepared following the original Dignam protocol29. TMG agarose beads (TMG mouse antibodies, K121, agarose conjugate, Merck NA02A) were preblocked by incubation with 0.1% BSA in phosphate-buffered saline (PBS) buffer for 1 h at 8 °C, then washed with two bead volumes (CV) of the immunoprecipitation (IP) buffer (20 mM Tris-HCl pH 7.9; 150 mM KCl) supplemented with protease inhibitor cocktail (Roche cOmplete). TMG-beads were added to NEs to the final volume of 10% and incubated ON at 8 °C, with shaking. TMG-beads were collected in Mini Bio-Spin chromatography columns, Bio-Rad, by centrifugation for 30 s at 2,500g at 8 °C. After two wash steps with the IP buffer, beads were eluted by boiling for 10 min with the buffer containing 50 mM Tris-HCl pH 7.9, 150 mM KCl and 0.5% sodium dodecyl sulfate (SDS). The eluates from the experiment performed in triplicates were analyzed by a TMT-plex quantitative mass spectrometry as previously described30.

Western Blotting

For the TMG pull-down eluates and NEs, equal amounts of total protein or fractions after glycerol gradient were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis using WedgeWell 4–20% Tris-glycine system, Invitrogen. The transfer to the polyvinylidene difluoride (PVDF) membrane was done in the Trans-Blot Turbo system, Bio-Rad, using a Turbo-transfer buffer. The following primary rabbit polyclonal antibodies were used: CD2BP2 (Sigma, HPA061309), DIM1 (Proteintech, 27646-1-AP), PRP6 (Invitrogen, PA5-61428) and SNU114/EFTUD2 (Invitrogen, PA5-96559). The secondary antibody was goat anti-rabbit IgG HRP-conjugate (Abcam, ab205718). Mouse monoclonal conjugated antibodies were anti-FLAG M2-peroxidase (Sigma, A8592), anti-GAPDH (Invitrogen MA515738HRP) and anti-HA-Tag F-7 HRP-conjugate (SantaCruz, sc-7392). The blots were visualized using Pierce ECL Western Blotting Substrate, Thermo Fisher Scientific, and documented on the ChemiDoc MP imaging system and ImageLab, Bio-Rad.

For the PRP6-CD2BP2 pull-down experiment, HEK293T cells were seeded into 6-well plates 24 h before transfection at a density of 500,000 cells per well in 1.5 ml Dulbecco’s modified Eagle medium medium with 10% fetal bovine serum. Plasmids containing 3xHA_PRP6270-941 and 3xFLAG_CD2BP2, both under CMV promoters, were mixed 1:1 and a total of 1 μg of DNA was diluted into in 50 μl of opti-MEM and mixed with 3 μg of polyethylenimine (PEI) MAX 40K in 50 μl of opti-MEM and incubated at room temperature for 10 min. Transfection solutions were added drop by drop to each well. The cells were collected by centrifugation 48 h after transfection, lysed in 400 μl of lysis buffer (150 mM KCl, 20 mM K-HEPES pH 7.8 and 0.1% Triton X-100) and sonicated for 10 s at 30% amplitude. The lysates were cleared by centrifugation in a table-top centrifuge at 20,000g at 4 °C for 30 min. The supernatant was incubated for 2 h with 5% (v/v) of FLAG-agarose to capture the bait protein. Affinity resin was washed three times with ten resin volumes of buffer 3 (150 mM KCl and 20 mM K-HEPES pH 7.8) and subsequently resuspended in SDS sample buffer and heated up to 95 °C for 5 min to release bound proteins. Input and elution fractions were analyzed by western blotting.

A PVDF membrane (Merck) was activated for 5 min in 100% EtOH and incubated for 5 min in the transfer buffer (1× Tris-glycine, 20% EtOH). A wet transfer was performed for 60–90 min at 30 V in an Invitrogen XCell II Blot Module. The membrane was blocked with 5% milk in PBS supplemented with 0.2% Tween 20 (PBST) for 1 h at room temperature. Primary antibodies were added in the following dilutions: anti-HA 1:5,000 ((HA-7) HRP ab49969, Abcam); anti-FLAG 1:5,000 (HRP sigma A8592-.2MG). The membrane was washed three times for 5 min with 20 ml of PBST, and chemiluminescence was detected with an HRP substrate kit (Pierce ECL Western Blotting Substrate) in a ChemiDoc imager (Bio-Rad).

In vitro splicing assay

AdML-M3 pre-mRNA substrate was obtained by run-off in vitro T7-transcription31, capped by VCE, NEB and labeled with fluorescein-5-thiosemicarbazyde at the 3′ end as previously described32. NEs prepared from WT or CD2BP2KO cells were used. The typical reaction contained 30 mM KCl, 3 mM MgCl2, 2 mM ATP, 20 mM creatine phosphate, 20 nM RNA AdML_M3 RNA substrate and 40% NE. Splicing reactions were assembled in 20 µl volume and incubated for 2 h at 30 °C. RNA was then isolated by phenol/chloroform extraction and ethanol precipitation and analyzed by denaturing 6% polyacrylamide gel electrophoresis in 7 M urea. Fluorescence of the RNA substrate and splicing product was visualized on ChemiDoc MP.

3xFLAG_TEV_SBP_CD2BP2 cell line generation

Open Reading Frame of CD2BP2 was cloned into a modified pFLAG_CMV10 vector containing an N-terminal 3xFLAG_TEV_SBP affinity tag. FreeStyle 293-F cells were transfected with this plasmid, and a stable, polyclonal cell line was derived through G418 antibiotic selection. Expression of the target protein was confirmed by western blot analysis.

Purification of the 20S U5 snRNP for cryo-EM analysis

Suspension culture of FreeStyle 293-F cells was grown in the FreeStyle medium (Thermo Fisher Scientific) to the density of ~2 × 106 cells ml−1 in an orbital shaker (Infors) at 37 °C, 8% CO2 and 90 rpm. The cell culture was collected by centrifugation, and NE was prepared following the original Dignam protocol29. After the final dialysis step, samples were aliquoted and flash-frozen in liquid nitrogen. For each preparation, an aliquot of NE was thawed on ice, and the salt concentration was adjusted to the final 500 mM KCl. EZview Red anti-FLAG M2 Affinity Gel (Sigma) was added to 10% (v/v) of the reaction volume and incubated overnight at 8 °C with shaking. The resin was washed three times with 10 column volumes (CV) of the wash buffer 1 (20 mM K-HEPES pH 7.9, 500 mM KCl, 2 mM MgCl2, 0.1% Igepal CA-630 and 5% glycerol), and complexes were eluted by incubation in 1 CV of wash buffer 1 supplemented with 10% (v/v) of TEV protease (1 mg ml−1), for 3 h at room temperature. A second-step purification was performed by incubation FLAG eluate with 5% total volume of Pierce high-capacity streptavidin agarose for 3 h at 8 °C with shaking. Beads were washed three times with 10 CV of the wash buffer 2 (20 mM K-HEPES pH 7.9, 500 mM KCl, 2 mM MgCl2). Samples were eluted from the resin by several incubations with 0.5 CV of wash buffer 2 supplemented with 10 mM biotin for 15 min on ice. The eluate was loaded onto a 4 ml 10–30% glycerol gradient containing 20 mM K-HEPES pH 7.9, 500 mM KCl, 2 mM MgCl2, 0.1% Igepal CA-630 and 0–0.1% glutaraldehyde33 and centrifuged for 16 h at 35,000 rpm at 4 °C (Beckman Coulter Ultracentrifuge Optima L-90K). The peak fraction of the glycerol gradient was analyzed by negative staining EM, and the fractions containing most homogeneous particles were dialyzed against a buffer containing 20 mM K-HEPES pH 7.9, 150 mM KCl and 2 mM MgCl2 and used directly for grid preparation without further manipulations.

Cryo-EM data collection and processing

The sample was applied to 300 mesh Quantifoil R 1.2/1.3 grids covered with 3 nm continuous carbon, which had been glow-discharged for 30 s at 15 mA at 0.4 mbar using the Pelco EasiGlow. The grids were plunge frozen in liquid ethane after applying 2 µl at 4 °C, 100 % humidity and blotting for 2 s at blot force −5 in a Vitrobot Mark IV. The grids were screened on a Glacios 200 kV microscope equipped with a Falcon III detector and transferred to a Titan Krios microscope operating at 300 kV equipped with a Gatan Energy filter34. A total of 8,506 micrographs were recorded using SerialEM35 and a K2 direct electron detector at a magnification of 130,000×, a defocus between −1.5 and −3.5 μm with a dose rate of 4.6 e− per pixel per second and inserted energy slit at 20 eV, as well as the 70 μm objective aperture. The total dose was 40.5 e− Å−2, accumulated in 40 frames at a final pixel size of 1.045 Å. All image processing was done using cryoSPARC v3.3 (ref. 36). For preprocessing, we used patch motion correction and determined the contrast transfer function (CTF) parameters using patch CTF estimation. Using the blob picker functionality, 503,581 particles were picked and extracted in a 504-pixel box. After binning two times, the particles were subjected to two-dimensional classification to create templates for template picking, which resulted in 490,503 picked particles. These particles were subjected to two-dimensional classification, ab initio reconstruction, followed by three-dimensional structure heterogeneous refinement until a homogeneous subpopulation of 76,918 particles was identified. Nonuniform refinement resulted in a final 3.1 Å resolution map based on the 0.143 Fourier Shell Correlation (FSC) criterion37,38. The obtained map was sharpened by applying a B-factor of −55 Å2.

Model building and structural analysis

Atomic coordinates of the U5 snRNP components extracted from the structure of the human tri-snRNP21 (PDB ID: 6qw6) were used as templates for modeling. Individual chains were fitted into the cryo-EM density as rigid bodies using UCSF Chimera39, the components with well-resolved density were manually adjusted and rebuilt in Coot v0.9.8.5 (ref. 40). Other components with poorly resolved densities (that is, BRR2, PRP8RNaseH, PRP8Jab1/MPN, DDX23, Sm ring, 40K) were docked into the map as rigid bodies and left in their original form. CD2BP2 binding sites were initially identified by an exhaustive in silico AlphaFold2-based search41,42 for all possible interactions with other U5 snRNP components, using a previously described approach43.

Atomic models were initially refined with Refmac Servalcat v5.8.0267 (ref. 44) with secondary structure restraints generated with ProSMART45 via the CCP-EM software suite46. Final models were refined in real space in Phenix47 and validated in Molprobity48. Structural representations for figures were prepared with Pymol (Schrödinger) and ChimeraX49.

Cross-linking and mass spectrometry analysis

CD2BP2 complex at 3 mg ml−1 was incubated with 0.25 mM or 1 mM BS3 for 30 min at 30 °C with shaking at 600 rpm (ThermoMixer, Eppendorf), and the cross-linking reaction was quenched by the addition of Tris-Cl pH 7.5 to the final concentration of 50 mM and incubated for 10 min at 35 °C at 600 rpm. Then, samples were mixed with 0.05 (v/v) of RapiGest and, after the addition of 10 mM DTT, incubated at 50 °C for 30 min, with shaking at 600 rpm. Subsequently, 2-chloroacetamide was added to 50 mM final concentration, and samples were incubated at 25 °C for 30 min at 600 rpm, protected from direct light. Proteins were digested with 1:50 (m/m ratio) of trypsin and 1:100 of LysC for 16 h at 37 °C. Digestion was stopped by adding 0.5% (v/v) of trifluoroacetic acid. Further analysis was performed by EMBL Proteomics Core Facility in Heidelberg.

Digested peptides were concentrated and desalted using an OASIS HLB µElution Plate (Waters), according to manufacturer instructions. Crosslinked peptides were enriched using size exclusion chromatography50. In brief, desalted peptides were reconstituted with size exclusion chromatograph buffer (30% (v/v) acetonitrile (ACN) in 0.1% (v/v) trifluoroacetic acid (TFA)) and fractionated using a Superdex Peptide PC 3.2/30 column (GE) on a 1200 Infinity high-performance liquid chromatography system (Agilent) at a flow rate of 0.05 ml min−1. Fractions eluting between 50–70 μl were evaporated to dryness and reconstituted in 30 μl 4% (v/v) ACN in 1% (v/v) FA.

Collected fractions were analyzed by liquid chromatography‐coupled tandem mass spectrometry using an UltiMate 3000 RSLC nano liquid chromatography system (Dionex) fitted with a trapping cartridge (µ-Precolumn C18 PepMap 100, 5 µm, 300 µm × 5 mm, 100 Å) and an analytical column (nanoEase M/Z HSS T3 column 75 µm × 250 mm C18, 1.8 µm, 100 Å, Waters). Trapping was carried out with a constant flow of trapping solvent (0.05% trifluoroacetic acid in water) at 30 µl min−1 onto the trapping column for 6 min. Subsequently, peptides were eluted and separated on the analytical column using a gradient composed of solvent A (3% dimethyl sulfoxide and 0.1% formic acid in water) and solvent B (3% dimethyl sulfoxide and 0.1% formic acid in acetonitrile) with a constant flow of 0.3 µl min−1. The outlet of the analytical column was coupled directly to an Orbitrap Fusion Lumos (Thermo Scientific) mass spectrometer using the nanoFlex source.

The peptides were introduced into the Orbitrap Fusion Lumos via a Pico-Tip Emitter 360 µm × 20 µm; 10 µm tip (CoAnn Technologies) and an applied spray voltage of 2.1 kV, and the instrument was operated in positive mode. The capillary temperature was set at 275 °C. Only charge states of 4–8 were included. The dynamic exclusion was set to 30 s and the intensity threshold was 5 × 104. Full mass scans were acquired for a mass range 350–1,700 m/z in profile mode in the orbitrap with resolution of 120,000. The AGC target was set to standard and the injection time mode was set to auto. The instrument was operated in data-dependent acquisition mode with a cycle time of 3 s between master scans and tandem mass spectrometry (MS/MS) scans were acquired in the Orbitrap with a resolution of 30,000, with a fill time of up to 100 ms and a limitation of 2 × 105 ions (AGC target). A normalized collision energy of 32 was applied. MS2 data were acquired in profile mode.

All data were analyzed using the cross-linking module in Mass Spec Studio v2.4.0.3524 (www.msstudio.ca, ref. 51). Parameters were set as follows: trypsin (K/R only), charge states 3–8, peptide length 7–50, percent E-value threshold of 50, mass spectrometry (MS) mass tolerance of 10 ppm, tandem mass spectrometry mass tolerance of 10 and elution width of 0.5 min. BS3 cross-links residue pairs were constrained to KSTY on one end and one of KSTY on the other. Identifications were manually validated, and cross-links with an E-value corresponding to <0.05% false discovery rate (FDR) were rejected. The data export from the Studio was filtered to retain only cross-links with a unique pair of peptide sequences and a unique set of potential residue sites.

Structural and functional analysis of the XL-MS data were performed with XiView52.

Reporting summary

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