Structural basis of regulated N-glycosylation at the secretory translocon

Antibodies

The following primary antibodies were used: mouse anti-CCDC134 (Santa Cruz Biotechnology, #sc-393390, RRID: AB_3662100, 1:500); mouse anti-GRP94 (R&D Systems, #MAB7606, RRID: AB_3644153, 1:2,000); mouse anti-GRP94 (Santa Cruz Biotechnology, #sc-393402, RRID: AB_2892568, 1:2,000); rabbit anti-LRP6 (Cell Signaling Technology, #2560, RRID: AB_2139329, 1:1,000); mouse anti-α-tubulin (MilliporeSigma, #T6199, RRID: AB_477583, 1:10,000); mouse anti-α-tubulin (Abcam, #ab11304, RRID: AB_297909, 1:1,000); mouse anti-STT3A (Abnova, #H00003703-M02, RRID: AB_530104, 1:1,000); rabbit anti-STT3A (Proteintech, #12034-1-AP, RRID: AB_2877818, 1:1,000); rabbit anti-IGF1Rβ (Cell Signaling Technology, #9750, RRID: AB_10950969, 1:1,000); mouse anti-GAPDH (Proteintech, #60004-1-Ig, RRID: AB_2107436, 1:10,000); rabbit anti-PSAP (GeneTex, #GTX101064, RRID: AB_2037779, 1:1,000); mouse anti-HA (GenScript, #A01244, RRID: AB_1289306, 1:1,000); rabbit anti-HA (Bethyl, #A191-102, RRID: AB_2891412, 1:2,000); rabbit anti-uL22 (Abcepta, #AP9892b, RRID: AB_10613776, 1:1,000); rabbit anti-uL2 (Abcam, #ab169538, RRID: AB_2714187, 1:1,000); rabbit anti-SEC61β (Cell Signaling Technology, #14648, RRID: AB_2798555, 1:1,000); rabbit anti-TRAPα (MilliporeSigma, #HPA011276, RRID: AB_1857503, 1:1,000); rabbit anti-STT3B (Proteintech, #15323-1-AP, RRID: AB_2198046, 1:1,000); rabbit anti-FKBP11 (Atlas Antibodies, #HPA041709, RRID: AB_10794487, 1:1,000); mouse anti-FLAG (MilliporeSigma, #F1804, RRID: AB_262044, 1:1,000); mouse anti-BiP (BD Transduction Laboratories, #610978, RRID: AB_398291, 1:1000).

The following secondary antibodies conjugated to horseradish peroxidase were used: Peroxidase AffiniPure Donkey Anti-Mouse IgG (H+L) (Jackson ImmunoResearch Laboratories, #715-035-150, RRID: AB_2340770, 1:10,000); Peroxidase AffiniPure Donkey Anti-Rabbit IgG (H+L) (Jackson ImmunoResearch Laboratories, #111-035-144, RRID: AB_2307391, 1:10,000); Peroxidase Donkey Anti-Rabbit IgG (Fc specific) (Sigma-Aldrich, #SAB3700863, RRID: AB_3675584, 1:10,000); Peroxidase Rabbit Anti-Mouse IgG H&L (Abcam, #ab6728, RRID: AB_955440, 1:10,000). Secondary antibodies conjugated to IRDye 800CW were obtained from LI-COR (IRDye 800CW Donkey anti-Mouse IgG Secondary Antibody, 1:10,000).

Constructs

CCDC134 constructs

Doxycycline-inducible 3xFLAG–CCDC134 (tag inserted after the ER signal sequence) was cloned by polymerase chain reaction (PCR) amplification followed by Gibson assembly (New England Biolabs) into pcDNA5-FRT/TO. Doxycycline-inducible 3xHA–CCDC134 (tag inserted after the ER signal sequence) mutants were cloned by PCR amplification followed by Gibson assembly (New England Biolabs) into pLenti-TRE-rtta3G-BLAST⁴.

GRP94 constructs

pENTR2B-3xFLAG-GRP94 (wild type) and pENTR2B-3xFLAG-GRP94^5N (includes mutations in all five facultative sequons: S64A, S109A, S447A, T483I, T504I)⁴ were used as templates to clone all mutant constructs used to generate RKO addback cell lines. All mutant constructs were generated using PCR amplification followed by Gibson assembly and cloned into pLenti CMV Puro DEST⁶³ using gateway methods. Doxycycline-inducible 3xFLAG–GRP94 (tag inserted after the ER signal sequence) constructs for transient expression in HEK293 cells were cloned by PCR amplification followed by Gibson assembly into pcDNA5/FRT/TO. These were used as templates to generate mutant constructs by QuikChange Site-Directed Mutagenesis (Agilent, 210519, 200514).

STT3A constructs

pENTR2B-STT3A-FLAG⁴ was used as a template to generate STT3A mutants using PCR amplification followed by Gibson assembly and cloned into pLenti CMV Puro DEST⁶³ using Gateway methods. All constructs were fully sequenced to confirm accuracy.

Cell culture

Flp-In T-REx 293 cells (Invitrogen) were maintained at 37 °C and 5% CO₂ in DMEM (Corning, MT10013CV) and supplemented with 10% foetal bovine serum (GeminiBio, 900-108) and 100 U ml⁻¹ penicillin plus 100 µg ml⁻¹ streptomycin (GeminiBio, 400-109). Cells were checked approximately every 6 months for mycoplasma contamination using the Universal Mycoplasma Detection Kit (ATCC, 30-1012K) and were found to be negative.

HEK293 cell lines

FKBP11 (ref. ²⁶) and STT3A⁶⁰ knockout Flp-In T-REx 293 cell lines were generated by CRISPR–Cas9 as previously described. A CCDC134 knockout cell line was generated similarly. In brief, an sgRNA (5′-AGAAGATGTTTGAGGTGAAG-3′) targeting exon 3 of human CCDC134 was designed using the Synthego CRISPR design tool (https://design.synthego.com) and cloned into pSpCas9(BB)-2A-Puro plasmid (PX459; Addgene #62988). Twenty-four hours after transfection, cells were selected under 1 µg ml⁻¹ of puromycin (InvivoGen, ant-pr-1) for 72 h. Single cells were isolated by sorting in 96-well plates and allowed to grow clonally. Clonal cells were screened for gene knockout by immunoblotting for CCDC134. Further validation was performed by PCR amplification of ± 200 base pairs of PAM sequence and tracking INDEL by decomposition of Sanger sequencing data using the TIDE web tool⁶⁴ and Synthego ICE analysis tool. An STT3B knockout cell line was generated similarly with sgRNA (5′-ACACATCATCTTGCATCTCA-3′) targeting exon 2 of human STT3B and validated by immunoblotting and tracking INDEL.

A stable, doxycycline-inducible 3xFLAG-FKBP11 Flp-In T-Rex 293 cell line was generated in FKBP11 knockout cells as previously described²⁶. A stable, doxycycline-inducible 3xFLAG-CCDC134 Flp-In T-REx 293 cell line was generated similarly. Briefly, pOG44 (encoding Flp- recombinase) and pcDNA5 SS_ER-3xFLAG-CCDC134 at a ratio of 9:1 with 3 µl of TransIT-293 transfection reagent (Mirus Bio, MIR2700) were co-transfected into CCDC134 knockout Flp-In T-REx 293 cells. After 24 h, cells were selected with 100 µg ml⁻¹ of hygromycin for 12 days to obtain the stably integrated cell lines. For doxycycline-inducible expression of 3xFLAG–CCDC134, cells were tested in a range of doxycycline concentrations; overexpression was observed without further doxycycline in the culture media.

RKO cell lines

Stable addback of tagged GRP94, STT3A or CCDC134 were introduced into clonally derived knockout RKO cells using the lentiviral expression system as described previously⁴. Briefly, virus was generated by transfecting HEK293T cells in six-well plates with 200 ng pMD2.G (Addgene), 400 ng psPAX2 (Addgene) and 800 ng of the desired pLenti CMV Puro DEST or pLenti-TRE-rtta3G-BLAST construct using 7 μl of 1 mg ml⁻¹ polyethylenimine (Polysciences) per well. Lentivirus-conditioned media was collected after 48 h, filtered through a 0.45-μm filter and 0.5 ml of filtered lentivirus was mixed with 1.5 ml of complete media containing 8 μg ml⁻¹ polybrene (MilliporeSigma). The diluted virus was then added to the indicated cells seeded on six-well plates. 24 h post-infection, cells were split and selected with puromycin (2 μg ml⁻¹) or blasticidin (10 µg ml⁻¹) for 3–7 days or until all of the cells on the control plate are dead. For doxycycline-inducible expression of 3xHA–CCDC134, cells were grown for 24 h in a range of doxycycline concentrations with 5 nM inducing low, near-endogenous expression levels.

Preparation of rough microsomes

Typically, ten 15-cm dishes of the desired HEK293 cell line were grown to approximately 80% confluency, washed once with ice-cold PBS and collected by scraping in ice-cold PBS. Cells were collected by centrifugation at 1,000 × g for 5 min at 4 °C, flash-frozen in liquid nitrogen and stored at −80 °C for future use. For microsome preparation, the frozen cell pellet was thawed, resuspended in hypotonic homogenization buffer (10 mM HEPES pH 7.4, 10 mM KOAc, 1 mM MgCl₂) and incubated on ice for 30 min. Cells were then mechanically lysed with 50 strokes in a glass Dounce homogenizer chilled in ice and sucrose was added to a final concentration of 200 mM. Whole cells and nuclear fragments were removed by two rounds of centrifugation at 2,000 × g for 10 min at 4 °C and the ER-enriched membrane fraction was collected from the supernatant by centrifugation at 10,000 × g for 15 min at 4 °C. Pelleted membrane fractions were resuspended to a total volume of 1 ml in microsome buffer (50 M HEPES pH 7.4, 250 mM KOAc, 10 mM MgCl₂, 250 mM sucrose) supplemented with 0.5× cOmplete EDTA-free protease inhibitor cocktail (Roche, 11836170001) and 1 mM CaCl₂. The microsome suspension was treated with 10,000 U ml⁻¹ micrococcal nuclease (MNase) (NEB, M0247S) at 37 °C for 25 min followed by 5 U ml⁻¹ RQ1 RNase-Free DNase (Promega, M610A) for 5 min at room temperature. Nuclease digestion was quenched with 2 mM EGTA and microsomes were collected by centrifugation at 10,000 × g for 15 min at 4 °C. Pelleted microsomes were resuspended in microsome buffer supplemented with 50 U ml⁻¹ SUPERaseIn (Invitrogen, AM2696) to the desired concentration and flash-frozen in liquid nitrogen and stored at −80 °C.

Sample preparation for selective ribosome profiling

Microsomes were prepared as described above, with the following modifications. All buffers were supplemented with 100 ng ml⁻¹ cycloheximide. Samples were kept on ice or at 4 °C throughout, unless otherwise noted. Homogenization buffer was supplemented with 0.5× cOmplete EDTA-free protease inhibitor cocktail (Roche). Nuclear fragments were removed by pelleting twice at 2,800 rpm (1,578 × g) for 10 min. Before MNase treatment, microsomes were resuspended with 1.5 ml of 100 mM KOAc microsome buffer.

Micrococcal nuclease-treated 3xFLAG–CCDC134 and 3xFLAG–FKBP11 microsome pellets (from nine 15-cm dishes) were resuspended with 1 ml microsome buffer supplemented with 2.5% digitonin and 1× cOmplete EDTA-free protease inhibitor cocktail (Roche). The sample was rotated to solubilize for 30 min and insoluble material was removed by pelleting at 13,500 × g for 15 min. The soluble fraction was layered over a 1-ml sucrose cushion (50 mM HEPES pH 7.4, 150 mM KOAc, 1 M sucrose, 5 mM MgCl₂, 0.25% digitonin) and centrifuged at 250,000 × g for 2 h in a TLA100.3 rotor. The resulting ribosomal pellet was resuspended in 1 ml microsome buffer supplemented with 1.25% digitonin. A portion of this was reserved for sequencing (which served as the ‘input’ sample for ribosome profiling) and the remainder was immunoprecipitated in batch using 60 μl anti-FLAG M2 affinity gel (MilliporeSigma, A2220) and end-over-end mixing overnight at 4 °C. The sample was subsequently washed four times with 12 column volumes of microsome buffer supplemented with 0.4% digitonin. The sample was eluted twice with 1.5 column volumes of microsome buffer (with 200 mM KOAc, 0.4% digitonin, 0.5 mg ml⁻¹ FLAG peptide). The eluate (which served as the ‘IP’ sample for ribosome profiling) was collected using a pre-equilibrated spin filter column, frozen in liquid nitrogen and stored at −80 °C (‘IP’ sample for sequencing).

Selective ribosome profiling library preparation using Rfoot-seq

To prepare the ribosome profiling library, the RNA concentration in both input and IP samples was determined using the Qubit RNA high-sensitivity assay. To generate RNase footprints, samples with 800 ng of RNA were digested with RNase 1 (LGC Biosearch Technologies, N6901K) at 0.5 U µl⁻¹ in a 90-µl reaction at room temperature for 1.5 h. The reactions were stopped by adding 400 µl of TRIzol (Ambion, 15596026), vortexing thoroughly and then adding 100 µl of chloroform. RNA in the aqueous layer was separated by centrifugation at 12,000 × g for 15 min at 4 °C and precipitated overnight by adding 0.1 volumes of 3 M sodium acetate (Invitrogen, AM9740), 10 mg of GlycoBlue (Invitrogen, AM9515) and 1.2 volumes of isopropanol (Fisher Scientific, BP2618). Purified RNase footprints were subjected to Rfoot-seq library preparation following a previously published method^27,28.

Rfoot-seq data processing and analyses

Ribosome profiling data were processed by trimming adaptors with Cutadapt v4.1, removing rRNA reads with Bowtie v2.2.6 (ref. ⁶⁵) and aligning remaining reads to the human hg38 genome and RefSeq-defined transcriptome with TopHat v2.1.0 (ref. ⁶⁶). Uniquely mapped reads were used for downstream analyses. High-quality reads showing 3-nt periodicity were adjusted to the ribosomal A-site using RibORF (refs. ^67,68,69).

Transcript abundance was quantified as transcripts per million using HTSeq v2.0.3 (ref. ⁷⁰). Transcript enrichment was calculated as the log₂ ratio of IP to input samples, median-centred to all genes and tested for significance using chi-squared tests with Benjamini–Hochberg correction. Positional interaction profiles were generated from codon-level enrichment, computed as the log₂ ratio of IP to input at each codon position.

Sample preparation for cryo-EM

3xFLAG-FKBP11 HEK293 cells were seeded in ten 15-cm dishes and induced with 0.1 ng ml⁻¹ doxycycline for 24 h. Microsomes (A260 of 35) were prepared as described above and solubilized in microsome buffer supplemented with 1.5% digitonin and 1× cOmplete EDTA-free protease inhibitor cocktail (Roche). After 1 h at 4 °C on a rotating wheel, the solubilized microsomes were diluted 2× with microsome buffer (150 mM KOAc) and cleared by centrifugation at 15,000 × g for 15 min at 4 °C. The cleared supernatant was incubated overnight at 4 °C with approximately 70 µl of anti-FLAG M2 affinity gel (MilliporeSigma, A2220) pre-equilibrated with wash buffer (50 mM HEPES pH 7.4, 200 mM KOAc, 2 mM MgCl₂, 1 mM glutathione, 0.15% digitonin). Resin was washed twice with five column volumes of wash buffer and bound material was eluted with 200 µl of the same buffer supplemented with 0.5 mg ml⁻¹ FLAG peptide (APExBIO, A6001) for 10 min at room temperature. The eluate was passed through a pre-equilibrated approximately 30 µm Pierce spin column (Thermo Fisher, 69725), layered over a 300-µl sucrose cushion (50 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl₂, 1 mM glutathione, 500 mM sucrose, 0.15% digitonin) and centrifuged at 355,000 × g for 1 h at 4 °C in a TLA120.1 rotor (Beckman Coulter; Optima MAX-XP). The ribosome pellet was resuspended in 20 µl of wash buffer supplemented with 0.5 mg ml⁻¹ FLAG peptide to a final A260 of 0.57 and immediately used to freeze grids for single-particle cryo-EM.

Cryo-EM grid preparation and electron microscopy

Quantifoil R1.2/1.3 400 mesh gold grids with 2 nm ultrathin carbon were glow-discharged for 25 s immediately before use. Using a Thermo Fisher Vitrobot Mark IV, 3 μl of freshly prepared sample was applied to the grid, incubated at 22 °C and 100% humidity for 60 s, blotted for 7 s, drained for 0.5 s and frozen in liquid-nitrogen-cooled ethane. Three datasets (811, 1,684 and 4,074 videos) were collected on an FEI Titan Krios at 300 KV with EPU software, using a defocus range from −1.9 to −0.9 μm. Super-resolution videos (pixel size = 0.84 Å) were recorded at a nominal magnification of 53,000× using a K3 BioQuantum direct electron detector (Gatan) and a total electron exposure of 60 e⁻ Å⁻² over 52 frames.

Image processing

All data processing was performed using RELION 5.0 (ref. ⁷¹). Videos were motion-corrected using MotionCor2 (ref. ⁷²) with 7 × 5 patches and dose weighting. The contrast transfer function (CTF) was estimated with CTFFIND4.1 (ref. ⁷³). An initial particle set was generated by automated picking using the Laplacian-of-Gaussian method with a diameter range of 200 to 300 Å and a threshold of 1, applied to 4,074 videos binned by 2 during motion correction. Particles were extracted using a 128-pixel box (5.04 Å per pixel) and subjected to 2D classification (k = 100, T = 2) to select ribosome-like classes. Selected particles were re-extracted with a 384-pixel box (1.68 Å per pixel) and used for 3D refinement. The resulting map was used to train a Topaz⁷⁴ model for autopicking across the full combined dataset (6,569 unbinned videos), yielding 741,983 particles. Topaz-picked particles were extracted with a 256-pixel box (2.52 Å per pixel) and subjected to 2D classification (k = 100, T = 2) using a 550 Å spherical mask and 100 classes. From these, 26 ribosome-like classes were selected (676,520 particles). 3D refinement was performed using a low-pass-filtered (60 Å) mammalian ribosome reference, followed by 3D classification (k = 9, T = 4). This yielded four classes corresponding to the secretory translocon (73.8%), one class with weak 40S subunit density (4.5%) and four classes containing poorly aligned particles (21.7%). The 498,812 particles from secretory translocon-containing classes were re-extracted with a 768-pixel box (0.84 Å per pixel) and refined in 3D. Two rounds of 3D refinement and CTF refinement were performed before each dataset was individually polished. The polished particles were then combined for a third round of 3D refinement. To further resolve compositional heterogeneity, focused classification with signal subtraction (k = 9, T = 4, 416-pixel box, 0.84 Å per pixel) was performed using a mask around the translocon region. This yielded three classes with density for the secretory translocon, GRP94, CCDC134 and FKBP11. These were combined (135,132 particles) and refined with local angular searches. An extra round of focused classification with signal subtraction (k = 4, T = 4) targeting GRP94, CCDC134, FKBP11 and STT3A yielded two classes: one with (41.2%) and one without (58.8%) density for the three accessory factors. The particle subset containing GRP94, CCDC134 and FKBP11 (55,750 particles) was subjected to local refinement, yielding Map 1 (translocon-only map). These particles were then re-extracted using a 480-pixel box (1.4 Å per pixel) and further refined. A subsequent 3D refinement focused on the 60S ribosomal subunit yielded Map 2 (RTC map). Maps used for model building and display were uniformly low-pass-filtered to different resolutions, low-pass-filtered by local resolution (in RELION) or post-processed using DeepEMhancer⁷⁵.

Model building, refinement and validation

The human 60S subunit from PDB ID 6ZMI and the A/P- and P/E-site tRNAs from PDB ID 6W6L were used as starting models for the ribosome. The TRAP complex from PDB ID 8RJC³⁰ (mutated to human) and the OST-A complex from PDB ID 8B6L²⁹ were used as starting models for the secretory translocon.

Initial models for the remaining translocon components (FKBP11, CCDC134 and KCP2) and the GRP94 nascent chain were generated using the ColabFold2 (ref. ⁷⁶) implementation of AlphaFold2 multimer v3 (ref. ⁷⁷), as described below (Extended Data Fig. 4). Protein sequences were obtained from UniProt⁷⁸.

The quality of each AlphaFold2 multimer prediction was initially assessed by the predicted local-distance difference test (pLDDT), which provides a per-residue confidence score for each subunit, the predicted aligned error (PAE), which provides a confidence measure of the predicted protein–protein interface(s), and by the overall (pTM) and interface (ipTM) predicted TM scores. The AlphaFold2 complexes were also validated by their fit to the experimental density, comparison with previously determined structures and mutational analysis. Low-confidence regions not supported by the cryo-EM density were removed from the model.

All maps were used for model building. The 60S subunit, A/P and P/E tRNAs were placed as rigid bodies into Map 2 and a 43-residue poly-Ala segment of the nascent chain was modelled into exit tunnel density extending into the open SEC61 channel. The model was fit using tightly restrained real-space refinement in Coot⁷⁹, including planar and trans peptide restraints, Ramachandran restraints and Geman-McClure local distance restraints.

The OST-A complex from 8B6L was placed into Map 1 as a rigid body and adjusted as a single unit using tightly restrained real-space refinement in Coot, as above. The AlphaFold2 model of KCP2 (bound to the N-terminal region of DC2) was placed into the map and adjusted as a single unit (to maintain predicted inter-chain contacts) using tightly restrained real-space refinement in Coot. Weak density in the STT3A catalytic site was visible for a co-purifying LLO ligand, including the pyrophosphate group and portions of the isoprenyl tail and glycan moiety, and was built using PDB ID 8AGC as a guide. Several ordered N-glycans were also built for STT3A and RPN1. Further weak density was visible for what are probably digitonin and lipid molecules, but these were not modelled.

The AlphaFold2 model for SEC61–RAMP4 complex was placed as a rigid body into Map 1 and the complex adjusted as a single unit using tightly restrained real-space refinement in Coot. The displaced plug region of the fully opened SEC61 channel was rebuilt using PDB ID 8RJB as a guide.

Density for the TRAP complex, which was strongest in the lumenal regions, showed substantial displacement from its canonical position adjacent to SEC61. To model this, we divided the model into three separate regions: (1) the TRAPβ,γ,δ membrane and cytosolic segments; (2) TRAPα,β,δ lumenal domains; and (3) the TRAPα TMD and cytosolic tail. The membrane and lumenal regions were fit using tightly restrained real-space refinement in Coot. The TRAPα cytosolic tail was adjusted manually and fit using real-space refinement; no density was observed for the TRAPα TMD, which was not modelled.

The FKBP11–CCDC134 AlphaFold2 model was docked into Map 1 as a rigid body. The C-terminal, positively charged helix of FKBP11 was modelled into weak density adjacent to the ribosome. The flexible C-terminus of CCDC134 was placed into helical density at the OST48–RPN2 interface, guided by an additional AlphaFold2 prediction. The resulting model was adjusted using tightly restrained real-space refinement in Coot.

The AlphaFold2 model of GRP94 (residues 22–97) bound to DC2 and STT3A was placed into Map 1 by superimposing onto the previously placed STT3A subunit. GRP94 residues 73–97 showed poor fit to the helical cryo-EM density and were removed and the three-subunit complex (including GRP94 residues 22–72) was adjusted using tightly restrained real-space refinement in Coot.

Next, the AlphaFold2 model of GRP94 (residues 22–97) bound to CCDC134 was docked into Map 1 by superimposing onto the previously placed model of CCDC134–FKBP11. GRP94 residues 63–92 are predicted by AlphaFold2 to form a single extended helix (‘bridge helix’), in excellent agreement with the experimental cryo-EM density. GRP94 residues 22–72 were removed and the two-subunit complex was adjusted using tightly restrained real-space refinement in Coot.

Finally, the globular N-domain of GRP94 was docked into low-resolution density adjacent to CCDC134. Placement was guided by the central helical element of the N-domain and its β-sheet. No density was visible in the GRP94 nucleotide binding site or for the dynamic active-site ‘lid’, which was removed from the model. Similarly, no density was visible for the GRP94 ‘charged linker’ motif or for the M- and C-domains. The final GRP94 model (residues 22–283) was adjusted as a single complex with CCDC134, FKBP11, DC2 and STT3A using tightly restrained real-space refinement in Coot.

The translocon model was subjected to real-space refinement in PHENIX⁸⁰ alone (versus Map 1) or after combining with the 60S model (versus Map 2). Five rounds of global minimization and group B-factor refinement were carried out with Ramachandran and rotamer restraints (rotamer outliers were fixed using the fit option ‘outliers’ and the target was set to ‘fix_outliers’), reference model restraints (starting model) and hydrogen-bonding, base-pair and stacking parallelity restraints applied to the rRNA. Secondary structure restraints were turned off. Final model statistics for the translocon and the 60S-translocon models are provided in Extended Data Table 1. Structure figures were generated with ChimeraX⁸¹.

Functional analysis in RKO and HEK293 cell lines

RKO whole-cell lysates were prepared in lysis buffer: 50 mM Tris at pH 8.0, 150 mM NaCl, 2% NP-40, 0.25% deoxycholate, 0.1% SDS, 0.5 mM TCEP, 10% glycerol, 1× SIGMAFAST protease inhibitor cocktail (MilliporeSigma) and 1× PhosSTOP phosphatase inhibitor cocktail (Roche). Samples were placed on a shaker for 30 min at 4 °C, centrifuged for 30 min at 20,000 × g at 4 °C and the supernatant was collected and measured by BCA (Thermo Fisher Scientific). To resolve LRP6 and full-length GRP94 glycoforms, samples were run on Novex Tris-Glycine 4 to 12% gels (Thermo Fisher Scientific). For immunoblot analysis of IGF1R, cells were collected by scraping, as IGF1R is sensitive to trypsinization. The resolved proteins were transferred onto nitrocellulose membrane (Bio-Rad Laboratories) using a wet electroblotting system (Bio-Rad Laboratories) followed by immunoblotting.

For EndoH analysis of wild-type and knockout (ΔFKBP11, ΔCCDC134, ΔSTT3A, ΔSTT3B) HEK293 cells, microsomes were prepared from two 15-cm dishes as described above. A 20-μl aliquot (A260 of 60) was thawed and centrifuged at 10,000g for 10 min at 4 °C. The samples were treated with 2,500 U EndoH (New England Biolabs, P0702L), according to the manufacturer’s protocol. Samples were run on home-made tris-glycine 5.5% gels at 4 °C and constant voltage (150 V) or on 4–20% Bio-Rad TGX precast gels, followed by immunoblotting.

For analysis of GRP94 mutants in HEK293 cells, wild-type Flp-In T-REx 293 cells were seeded onto a six-well plate at a density of 200,000 cells per well one day before transfection. A transfection mixture was prepared containing 1 μg pcDNA5-SS-3xFLAG-GRP94, 150 μl Opti-MEM and 3 μl TransIT293, which was incubated at room temperature for 20 min before adding dropwise to each well. After 24 h, cells were collected using cold 1× PBS and centrifuged at 500g for 5 min at 4 °C. The supernatant was discarded and cells were resuspended in 100 μl lysis buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 0.5 mM PMSF). Samples were incubated on ice for 15 min (with vortexing every 5 min) and the supernatant was collected after centrifugation at 20,000g for 10 min at 4 °C. Samples were analysed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting.

Reporting summary

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