Research Article

Structure of the yeast oligosaccharyltransferase complex gives insight into eukaryotic N-glycosylation

See allHide authors and affiliations

Science  02 Feb 2018:
Vol. 359, Issue 6375, pp. 545-550
DOI: 10.1126/science.aar5140

Remember the sugar when making proteins

Eukaryotes have an elaborate trafficking and quality-control system for secreted glycoproteins. The glycosylation pathway begins in the endoplasmic reticulum with the enzyme oligosaccharyltransferase (OST), which attaches a long chain of sugars to asparagine residues of target proteins. Wild et al. report a cryo-electron microscopy structure of yeast OST, which includes eight separate membrane proteins. The central catalytic subunit contains binding sites for substrates and is flanked by accessory subunits that may facilitate delivery of newly translocated proteins for glycosylation.

Science, this issue p. 545

Abstract

Oligosaccharyltransferase (OST) is an essential membrane protein complex in the endoplasmic reticulum, where it transfers an oligosaccharide from a dolichol-pyrophosphate–activated donor to glycosylation sites of secretory proteins. Here we describe the atomic structure of yeast OST determined by cryo–electron microscopy, revealing a conserved subunit arrangement. The active site of the catalytic STT3 subunit points away from the center of the complex, allowing unhindered access to substrates. The dolichol-pyrophosphate moiety binds to a lipid-exposed groove of STT3, whereas two noncatalytic subunits and an ordered N-glycan form a membrane-proximal pocket for the oligosaccharide. The acceptor polypeptide site faces an oxidoreductase domain in stand-alone OST complexes or is immediately adjacent to the translocon, suggesting how eukaryotic OSTs efficiently glycosylate a large number of polypeptides before their folding.

N-linked glycosylation is a posttranslational modification of asparagine residues found in all domains of life (1). The covalently attached glycans are essential for correct protein folding, sorting, and secretion, or for modulating specific cell surface interactions (24). The central enzyme in the N-glycosylation pathway is the oligosaccharyltransferase (OST), which catalyzes the initial transfer of a defined glycan (Glc3Man9GlcNAc2 in higher eukaryotes) from the lipid carrier dolichol-pyrophosphate to polypeptide chains entering the secretory pathway in the endoplasmic reticulum (ER) (5, 6). OST specifically recognizes the Asn-Xaa-Ser/Thr motif, where Xaa can be any amino acid except proline (7). Whereas in bacteria and some lower eukaryotes this process is carried out by a single-subunit oligosaccharyltransferase (ssOST) (810), most eukaryotes encode large, membrane-embedded OST complexes that contain multiple subunits: eight in yeast and possibly more in multicellular organisms (11, 12). In Saccharomyces cerevisiae, five subunits (STT3, SWP1, WBP1, OST1, and OST2) were shown to be essential for cell survival, whereas deletion of the remaining three subunits was found to reduce complex stability and glycosylation activity (11, 13).

Eukaryotic OST enzymes have been visualized by using single-particle cryo–electron microscopy (cryo-EM) or cryo–electron tomography (1416), but the resolution of these studies, ranging from 20 to 9 Å, did not allow unambiguous assignment of OST subunits. X-ray structures of ssOSTs from bacteria and archaea (PglB and AglB proteins, homologous to the catalytic STT3 protein of multisubunit OSTs) have provided insight into the substrate recognition and glycan transfer mechanisms of prokaryotic enzymes (8, 9, 17, 18). X-ray structures of the isolated luminal domains of yeast OST6 and of the human Tusc3 (also named N33) have revealed the disulfide formation and cleavage mechanism of this redox chaperone (19, 20). However, in the absence of a high-resolution structure of a eukaryotic OST complex, it is unclear how the noncatalytic subunits extend the range of acceptor polypeptides over that of ssOST enzymes. To reveal the architecture of eukaryotic OSTs and to understand how they recognize and process a large number of acceptor proteins (21), we used single-particle cryo-EM to determine a high-resolution structure of the yeast OST complex.

Preparation of OST complex for cryo-EM analysis

Yeast OST exists in two isoforms, containing either OST3 or the homologous OST6 subunit. To avoid potential heterogeneity in our purified samples, we generated a S. cerevisiae strain lacking the OST6 gene and overexpressed OST3 from a plasmid to compensate for the absence of OST6. A 1D4 tag fused to the C terminus of the OST4 subunit facilitated efficient affinity purification and sufficient yields of the OST complex despite low endogenous expression levels. The resulting strain (Δost6 pOST3 OST4-1D4) showed almost complete N-glycosylation of the OST1 and WBP1 proteins, whereas a strain expressing only an OST6-containing complex (Δost3 pOST6 OST4-1D4) exhibited hypoglycosylation (Fig. 1A). Mass spectrometric analyses based on SILAC (stable isotope labeling by amino acids in cell culture) demonstrated that for six out of eight subunits, the detergent-purified, OST3-containing complex has the same subunit composition and stoichiometry as OST in wild-type yeast cells (Fig. 1B and table S1). OST4-derived peptides could not be detected, and the observed ratios for OST3 and OST6 showed that the two paralogs are mutually exclusive in the assembled OST complex. The OST3-containing complex reconstituted into lipidic nanodiscs was fully functional, as shown by an in vitro glycosylation assay developed earlier (10, 22), using a synthetic lipid-linked oligosaccharide (LLO) analog and a fluorescently labeled acceptor peptide (Fig. 1C). The measured glycosylation activity of 3.5 peptides per minute per OST matches previously reported rates of a eukaryotic ssOST (10) (Fig. 1D).

Fig. 1 Purification, stoichiometry, and functional characterization of yeast OST.

(A) Silver-stained SDS-PAGE analysis of detergent-purified OST complexes containing either OST6 or OST3. A 1D4 affinity tag was fused to the C terminus of OST4. Multiple bands for OST1 and WBP1 indicate heterogeneous N-glycosylation. MW, molecular weight marker. (B) Mass spectrometry–based quantification of subunit abundance in purified OST samples containing either OST3 (green bars) or OST6 (blue bars) relative to OST complexes from wild-type cell extract (n = 3 technical replicates, error bars indicate SD). (C) In vitro glycosylation of a fluorescently labeled peptide [TAMRA-DANYTK (TAMRA, tetramethylrhodamine fluorophore; D, Asp; A, Ala; N, Asn; Y, Tyr; T, Thr; K, Lys)] by a nanodisc-reconstituted, OST3-containing yeast OST complex, using a synthetic C20-LLO (NerylCitronellyl-PP-GlcNAc2) as a donor substrate. Glycosylated and nonglycosylated peptides were separated using a Tricine gel. (D) Following quantification of band intensities in (C), the ratio of glycosylated to unreacted peptide was plotted against the reaction time and fitted using a Michaelis-Menten saturation curve (n = 3 biological replicates, error bars indicate SD).

The resolution of the 3D reconstruction of nanodisc-reconstituted OST was 3.3 Å, on the basis of the Fourier shell correlation = 0.143 criterion (figs. S1 and S2). The EM map was of excellent quality in the transmembrane (TM) region and in most of the luminal regions. The luminal domain of SWP1, most distant to the membrane, featured lower map quality, probably indicating higher domain flexibility, but still displayed secondary structure features (Fig. 2A and figs. S1, D to F, and S3). Missing parts in the model include the luminal domain and TM1 of OST3 and the external loop EL5 of STT3, for which no density was observed. Because we visualized OST in an apo state, these segments are likely mobile (see below).

Fig. 2 Structure of yeast octa-subunit OST.

(A) EM map of the OST complex, with density covering individual subunits colored as in (D). (B) Cartoon representation of the yeast OST structure. Ordered glycans are shown in stick representation. (C) Cytosolic view onto the TM region of OST. (D) Structures of the single subunits in cartoon representation. EL5 and TM9 of STT3, as well as TM1 of OST3, are shown schematically. The crystal structure of the luminal domain of the homologous OST6 subunit (PDB ID: 3G9B) was used to illustrate the OST3 luminal domain (gray ribbon in dashed red box). The names of the corresponding human OST subunits are indicated in parentheses.

Architecture and subunit structure

The membrane-embedded part of yeast OST contains a total of 28 TM helices, with each subunit contributing at least one TM segment (Fig. 2, B to D). The membrane topologies of the subunits agree with previous predictions (11), except for OST5, whose N and C termini are located in the ER lumen. Previous in vivo experiments suggested that OST assembly occurs through the formation of three subcomplexes (13). Our structure revealed that the spatial arrangement of the OST subunits agrees with this subdivision: Subcomplex 1 contains OST1 and OST5; subcomplex 2 contains STT3, OST3, and OST4; and subcomplex 3 contains OST2, WBP1, and SWP1 (Fig. 2D). At the membrane-embedded interfaces between subcomplexes, several ordered phospholipids could be identified, particularly in the lipid layer facing the ER lumen (fig. S4, F to I).

In subcomplex I, the N-terminal helix of OST5 is in close contact with the OST1 subunit. OST1 has a single TM helix and a luminal domain containing two subdomains that have similar folds and superimpose well on each other (root mean square deviation = 2.2 Å for 153 out of 191 residues) (fig. S5A). The fold features two stacked β sheets and no α helices and was previously observed in other multidomain proteins—for example, in aminopeptidases and leukotriene hydrolases (23, 24) (fig. S5B). Because it does not contain the catalytic residues in these proteins, its function within OST1 cannot be deduced.

Subcomplex II includes the catalytic subunit STT3, which is homologous to the functionally and structurally related ssOST enzymes PglB from Campylobacter lari and AglB from Archaeoglobus fulgidus (8, 9, 18). Our structure confirms that yeast STT3 contains a similar TM topology with 13 TM helices (Fig. 2C) (25, 26), of which TM9 is poorly resolved in the EM map and is probably flexible in the absence of bound substrate (fig. S6). The luminal domain of STT3 also resembles the PglB and AglB proteins, although the folds are not identical (fig. S7). We found the external loop EL5 of STT3 to be disordered, in line with previous observations that EL5 of PglB only becomes fully ordered when substrates are bound (18). The TM helices of STT3 are tightly interacting with OST4 (which consists of a single TM helix) and with TM2 to TM4 of OST3 (Fig. 2C). Although the presence of full-length OST3 in the OST complex was confirmed by SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and mass spectrometric analysis (Fig. 1A and table S1), no clear density for TM1 and the N-terminal luminal domain of OST3 was visible in the EM map. We conclude that in the absence of a peptide substrate, the luminal domain of OST3 is highly flexible. Both TM1 of OST3 and TM9 of STT3 are in close proximity to the likely LLO-binding site, which suggests that their flexibility might be associated with the absence of bound LLO.

Subcomplex III contains the OST2 subunit, whose TM helices mediate contacts between STT3 and the TM helices of WBP1 and SWP1. OST2 contains an N-terminal α helix [amino acid (aa) 21 to aa 38] located at the cytosolic membrane boundary and parallel to the membrane plane, where it forms contacts with TM8 and TM9 of STT3. The WBP1 subunit contains two luminal domains. The N-terminal domain shares structural homology with various proteins of distinct functions, including an intraflagellar transport protein or glutamine amidotransferase (27, 28) (fig. S5C), whereas the central domain of WBP1 features a fold found in proteins of the complement system (29, 30) (fig. S5D). However, none of these structural homologs unambiguously identify a potential function of WBP1. The SWP1 subunit contains a single luminal domain, which is most distant to the membrane of all OST domains and connected to a single TM helix by a long linker.

OST1, WBP1, and SWP1 were previously suggested to be involved in substrate recognition and to act as chaperones, which would coordinate protein folding and glycosylation (3133). Our structural data show that none of their luminal domains adopt a chaperone-like fold. However, these domains might serve as docking platforms for interaction partners, including chaperones or enzymes acting on nascent glycoproteins. For example, the human OST1 homolog ribophorin I was shown to interact with the carbohydrate-binding protein malectin (34, 35).

To analyze whether eukaryotic OST complexes share common structural features, we plotted the degree of sequence conservation between different species onto our yeast OST structure. We found that the active site groove in STT3 is highly conserved (Fig. 3A and fig. S7D). In addition, regions of high sequence conservation were found at the interfaces between subunits OST1 and STT3 (1129 Å2 of buried surface area), as well as between WBP1 and SWP1 (2369 Å2 of buried surface area) (Fig. 3B). This suggests that both the arrangement of the luminal domains and the active site are conserved in eukaryotic, multisubunit OST.

Fig. 3 Sequence conservation in eukaryotic OST complexes.

(A) Surface representation of the yeast OST complex, colored according to sequence conservation. The predicted peptide and LLO binding sites are indicated. (B) Conserved subunit interfaces between OST1 and STT3 (left) and between WBP1 and SWP1 (right) are marked by dotted circles and arrows.

Previous genetic and mass spectrometric analyses suggested that yeast OST contains seven N-glycosylation sites (36). We observed EM density for five of these glycans (at Asn539 of STT3, Asn336 and Asn400 of OST1, and Asn60 and Asn332 of WBP1), whereas the remaining two glycans (at Asn99 and Asn217 of OST1) are located in flexible loop regions and are disordered. The best-ordered N-glycan is attached to the strictly conserved STT3 residue Asn539, where we observed density covering eight glycan moieties (Man6GlcNAc2). This N-glycan forms interactions with WBP1 and SWP1 (fig. S4, A to E) and is in the immediate vicinity of the proposed binding site for the 14-unit saccharide of bound LLO (see below).

Active site and substrate-binding pockets

The STT3 subunit contains the active site and the acceptor peptide and donor LLO binding pockets. Although we visualized yeast OST in the apo state with no substrates bound, key mechanistic insight could be deduced by comparing our structure to bacterial and archaeal ssOSTs (8, 9, 37). A superposition of yeast STT3 and substrate-bound PglB (18) (fig. S7, A and B) revealed that functionally important conserved residues and motifs have a similar spatial arrangement (Fig. 4A). These include the Asp-X-Asp/Glu motif (X, any amino acid) involved in coordinating the catalytic metal ion, an aspartate residue (Asp47 in yeast STT3 corresponding to Asp56 in PglB) that binds both the metal ion and the carboxamide group of the acceptor asparagine, the WWD motif providing hydrogen bond contacts to the β-hydroxyl group of the +2 serine/threonine of the acceptor sequon, and the Lys586 residue of the so-called “DK motif” that contributes additional contacts to the +2 serine/threonine of the acceptor peptide (Fig. 4A). Residues shown to be involved in LLO binding to PglB (18) are also conserved. These include an essential arginine (Arg404 in yeast STT3) that interacts with the pyrophosphate group of LLO and a tyrosine residue (Tyr521 in yeast STT3) that forms a hydrogen bond with the N-acetyl group of the reducing-end GlcNAc moiety (Fig. 4B).

Fig. 4 Active center of STT3.

(A) Ribbon representation of peptide-bound PglB (PDB ID: 5OGL) and yeast STT3. Residues involved in substrate recognition or metal ion binding are shown as sticks. The numbering in the bound peptide (DQNATF sequence) is relative to the acceptor Asn residue. Q, Gln; F, Phe; W, Trp; I, Ile; E, Glu. (B) Superposition of yeast STT3 (green ribbon) and LLO-bound PglB (gray ribbon). Functionally important residues (pink for PglB residues; blue for yeast STT3 residues) are shown as sticks, with interactions observed in PglB indicated by dashed lines. (C) Electrostatic surface representations of PglB and yeast STT3. The peptide bound to PglB was modeled into the yeast STT3 structure using the WWD motif and residues D47/D56 (PglB/STT3) as anchors. EL5 (aa 294 to aa 322) of PglB was removed for clarity. Corresponding regions revealing the structural basis of distinct peptide specificity are indicated with dashed ovals or boxes. (D) In vivo activity assay of STT3 point mutants.

Comparison of the electrostatic surface potential map of the peptide-binding pocket of PglB with that of the predicted binding pocket of yeast STT3 provides a molecular explanation for some of the differences in acceptor peptide specificities of bacterial and eukaryotic OSTs (Fig. 4C) (38, 39). In PglB, an arginine residue (Arg311) was found to interact with a negatively charged side chain (Asp or Glu) at the −2 position of the sequon. This arginine is conserved in bacterial ssOST enzymes and correlates with an extended sequon requirement (DxNxS/T) for bacterial N-glycosylation (8, 38) (Fig. 4C). In contrast, no such requirement is present at the −2 position of acceptor sequons in eukaryotes, and no positively charged residue is present in yeast STT3 where the −2 side chain is expected to bind (Fig. 4C). Instead, a larger cavity providing space for more voluminous side chains at the −2 position of sequons is observed, which is in line with previous findings that human OSTs display an increased glycosylation efficiency for substrates with aromatic residues at the −2 position (39). At the expected binding site of the −1 residue of the sequon, we found a “knob” in the yeast STT3 structure formed by the side chain of Glu45 (Fig. 4C). The reduced space may result in a more efficient binding of sequons with smaller side chains at the −1 position.

A groove between TM6 and TM11 of STT3 indicates where the dolichol tail of the LLO will likely bind (Fig. 3A). A similar groove was demonstrated to represent the polyprenyl-binding site in PglB, with four prenyl units ordered in the x-ray structure (18). Given their locations, it is conceivable that in yeast OST, the partially ordered TM9 of STT3 and the disordered TM1 of OST3 might interact with bound LLO. Our results suggest that PglB and yeast STT3 share a common substrate recognition and glycan transfer mechanism despite differing in their substrate specificities.

Because glycosylation is an essential process in yeast, abolishing OST function prevents cell growth (40). To validate the importance of the residues identified in the yeast OST structure, we therefore used a tester strain that expressed the ssOST enzyme LmSTT3D from Leishmania major (41) and generated six chromosomal mutations in the STT3 locus (Asp47→Ala47, Asp166→Ala166, Glu168→Gln168, Glu350→Ala350, Arg404→Ala404, and Lys586→Ala586). These mutations were predicted to affect metal binding, peptide binding, or LLO binding, but they did not affect OST complex stability (fig. S7C). With the exception of Lys586→Ala586, which resulted in a temperature-sensitive phenotype, all generated mutations prevented growth in the absence of LmSTT3D (Fig. 4D). We conclude that these mutations impaired the catalytic activity of the STT3 subunit and thus of OST function in vivo (40).

Acceptor polypeptide delivery by the redox chaperone OST3 or the translocon

Higher eukaryotes express two paralogs of the catalytic STT3 subunit, termed STT3A and STT3B, which are part of distinct OST complexes. STT3A-containing OST associates with the translocon, thereby bringing native peptide chains entering the ER into close proximity to the glycosylation machinery (14, 15, 4246). In contrast, STT3B-containing OST complexes (including the yeast enzyme) are stand-alone units that contain either an OST3 or OST6 subunit (homologous to the mammalian Tusc3 and MagT1, respectively), neither of which is present in translocon-associated OST. The luminal domains of OST6 and Tusc3 have oxidoreductase activity and feature a thioredoxin fold. Both can form disulfide bonds with acceptor proteins, but OST3- or OST6-containing complexes process different subsets of disulfide-forming acceptor polypeptides in vivo (19, 47, 48). In our apo structure of yeast OST, the oxidoreductase domain of OST3 appears disordered and is likely flexible, which may allow diverse and transiently bound OST substrates to be efficiently glycosylated.

Recent studies reported a mammalian ribosome-translocon-OST supercomplex visualized by cryo–electron tomography (14, 15). We docked our high-resolution OST structure into this tomography map and found a good fit with the portion of the density previously assigned to OST (14, 15) (Fig. 5A). This provides additional evidence that yeast and mammalian OST complexes share a common architecture. Additional density in the tomography map near the N terminus of SWP1 (Fig. 5A) probably corresponds to an N-terminal extension of ~300 residues of the mammalian SWP1 homolog ribophorin II, as compared with yeast SWP1 (11).

Fig. 5 OST-translocon interactions.

(A) Yeast OST structure docked into a cryo–electron tomography map of the mammalian ribosome-translocon-OST complex (EMDB ID: 3068). A close-up of the dashed region is shown in the left panel. Additional EM density above SWP1 and WBP1 (dashed oval) probably corresponds to a ~300-aa N-terminal extension present in mammalian SWP1 homolog ribophorin II. (B) View from the ER lumen onto the TM regions of OST and of the Sec61 translocon (PDB ID: 5A6U) (14) after the docking shown in (A). Although the OST3 subunit provides all contacts to Sec61 in this docking, it is replaced by the DC2 subunit in translocon-associated OST complexes. (C) Model of translocon-associated OST complex architecture and function. The orange, curved line depicts a nascent polypeptide entering the ER through the translocon and binding to the active site of STT3. An LLO molecule was manually placed in its likely binding pocket, with a red line representing dolichol, circled “P” denoting phosphate moieties, and the Glc3Man9GlcNAc2 moiety depicted by blue and green symbols according to standard glycan nomenclature. The black arrow depicts the proposed direction of the nucleophilic attack during glycan transfer.

Our docking analysis suggests that the active site of STT3A faces the heterotrimeric translocon complex (49) (Fig. 5, A and B). The contact point between OST and the translocon comprises the region corresponding to TM2 to TM4 of the OST3 subunit in yeast OST (Fig. 5B). However, OST3 is not present in the mammalian translocon-associated OST (43). Instead, recent studies suggest that the OST-translocon interaction is mediated by the DC2 protein (50). A sequence comparison reveals that the three TM helices of DC2 share a marked similarity with TM2 to TM4 of OST3 (fig. S8), lending further weight to the docking model shown in Fig. 5. Multisubunit OSTs may thus be described as a modular assembly of an OST core complex formed by seven subunits (STT3, OST1, OST2, OST4, OST5, WBP1, and SWP1), which associates with either oxidoreductases (OST3/6 or Tusc3/MagT1) or the translocon.

The arrangement shown in Fig. 5C allows unhindered access of the polypeptide substrate from the luminal exit of the translocon to the catalytic center of OST. The observed separation of ~40 Å is in agreement with earlier distance estimates based on the minimal polypeptide length of 65 residues between the peptidyltransferase center of the ribosome and the first possible glycosylation site (51). It is also in line with the previously reported minimal distance of 30 to 40 Å between the OST active site and the ER membrane deduced from evaluating the glycosylation of integral membrane proteins (52).

Conclusions: Functions of OST subunits

Organisms encoding multisubunit OSTs are found to glycosylate a substantially expanded range of protein substrates compared with single-subunit OST enzymes (53). This implies that the auxiliary subunits increase the efficiency of the catalytic STT3 core by contributing to substrate acquisition or by affecting the folding of acceptor proteins. Our yeast OST structure suggests that OST2, OST4, and OST5, which contain mostly TM helices, have a scaffolding function and are thus important for complex stability, but without directly contacting the substrates (11, 54). The luminal domains of OST1, SWP1, and WBP1 may also have structural roles by stabilizing the STT3 subunit conformation. In addition, some of them may directly interact with the substrates. Our structure revealed a cavity ranging from the active site of STT3 to the WBP1 and SWP1 subunits, just above the membrane boundary (Fig. 5C). The cavity is lined by the highly ordered N-glycan attached to Asn539 of STT3 and is sufficiently large to accommodate the glycan moiety of bound LLO substrate. It is conceivable that WBP1, SWP1, and possibly even the ordered N-glycan contribute to the recognition of the lipid-linked Glc3Man9GlcNAc2 moiety and thus help to define the preference for an LLO substrate containing terminal α–1,2-linked glucose (55). For OST1, it is worth speculating that, given its proximity to the peptide-binding pocket of STT3, it might interact with acceptor proteins and influence their folding. The active site of the OST complex faces the peptide-binding OST3 (or OST6) subunit or the translocon, both of which present polypeptide substrates in an unfolded state. This arrangement favors N-glycosylation over the competing folding reactions and thus extends the substrate range of OST.

Supplementary Materials

www.sciencemag.org/content/359/6375/545/suppl/DC1

Materials and Methods

Figs. S1 to S8

Tables S1 and S2

References (5684)

References and Notes

Acknowledgments: We thank J. Boilevin, T. Darbre, and J.-L. Reymond for providing the LLO analog and P. Tittmann for technical support. Funding: This work was supported by the Swiss National Science Foundation Sinergia programs TransGlyco (CRSII3_147632) and GlycoStart (CRSII5_173709) and grant 310030_162636 to M.A. R.W. acknowledges support from the ETH postdoctoral fellowship program. Author contributions: K.P.L. and M.A. designed the project. J.E. generated the yeast strain, developed initial purification protocols, and purified OST for mass spectrometry analyses. R.W. purified OST for structural studies, reconstituted OST in nanodiscs, and performed in vitro glycosylation assays. E.M.N. carried out in vivo mutational analysis in yeast. J.K. and R.W. performed negative-stain EM experiments and prepared cryo-EM grids. J.K. collected cryo-EM data and performed data analysis. R.W. built the OST model and performed model refinement. K.P.L. revised the model. R.W., J.K., and K.P.L. analyzed the structure. R.W. and K.P.L. wrote the manuscript with the help of J.K. and J.E.; all authors contributed to its revision. Competing interests: None declared. Data and materials availability: Cryo-EM data were collected at the electron microscopy facility of ETH Zurich (ScopeM). Atomic coordinates of the de novo built yeast OST model have been deposited in the Protein Data Bank (PDB) under ID 6EZN. The three-dimensional cryo-EM density maps have been deposited in the Electron Microscopy Data Bank (EMDB) under accession numbers EMD-4161 and EMD-4257. All data needed to evaluate the conclusions of this paper are provided either in the paper or in the supplementary materials.
View Abstract

Navigate This Article