Abstract
The protein translocon at the endoplasmic reticulum comprises the Sec61 translocation channel and numerous accessory factors that collectively facilitate the biogenesis of secretory and membrane proteins. Here, we leveraged recent advances in cryo-EM and structure prediction to derive insights into several novel configurations of the ribosome-translocon complex. We show how a transmembrane domain (TMD) in a looped configuration passes through the Sec61 lateral gate during membrane insertion; how a nascent chain can bind and constrain the conformation of ribosomal protein uL22; and how the translocon-associated protein (TRAP) complex can adjust its position during different stages of protein biogenesis. Most unexpectedly, we find that a large proportion of translocon complexes contains RAMP4 intercalated into Sec61’s lateral gate, widening Sec61’s central pore and contributing to its hydrophilic interior. These structures lead to mechanistic hypotheses for translocon function and highlight a remarkably plastic machinery whose conformations and composition adjust dynamically to its diverse range of substrates.
Introduction
Most eukaryotic secretory and membrane proteins are translocated co-translationally across the endoplasmic reticulum (ER) membrane at a ribosome-translocon complex (RTC). The central component of this translocon is the Sec61 complex, a heterotrimer containing a channel-forming α subunit and peripheral β and γ subunits1,2. In prokaryotes, the homologous SecYEG complex mediates translocation across the plasma membrane. In all organisms, this central channel associates dynamically with a variety of partners and accessory factors3–6. The structure and function of some accessory factors, such as the oligosaccharyl transferase complex (OST), are well established7, whereas many others are poorly understood3–6. Although these various translocon components have long been speculated to engage in a context-and substrate-dependent manner8,9, the rules governing their coordination are largely unclear and only now beginning to emerge10,11.
Sec61α is a pseudosymmetric channel that can open axially across the membrane and is laterally gated toward the lipid bilayer1,2. In its closed state, the axial channel’s pore is constricted, blocked by a short helix known as the plug, and the lateral gate is closed. Transport of hydrophilic polypeptide segments through Sec61 can be initiated by a flanking hydrophobic α-helix. These hydrophobic helices act as signal sequences that bind to Sec61’s lateral gate12,13. This binding is thought to widen the central pore, destabilise the plug, and thread one of the signal’s hydrophilic flanking regions into the channel to initiate translocation. A set of hydrophobic residues, known as the pore ring, lines the narrowest part of the channel and form a gasket-like seal that maintains the permeability barrier during translocation14–16.
This model of channel opening is derived from structures of a cleavable signal peptide (SP) bound to the mammalian Sec61 complex12 or bacterial SecYEG13,16. Most membrane proteins lack an N-terminal SP and instead initiate translocation using their first TMD, often termed a signal anchor (SA). Although it has long been thought an SA initiates flanking domain translocation the same way as an SP17–20, direct evidence for this idea is sparse and somewhat contradictory. Furthermore, the route an SP or SA takes to the lateral gate remains speculative. One model posits that they all access the lateral gate from the channel interior, displacing the plug en route. Alternatively, a hydrophobic helix could slide along the lipid-facing side of the lateral gate21, or use a member of the Oxa1 family for insertion22–24. Structures of TMD insertion intermediates would help resolve these and other crucial mechanistic issues in membrane protein biogenesis.
Beyond the Sec61 channel, a number of accessory factors co-translationally modify the polypeptide, facilitate membrane protein biogenesis, or facilitate secretion through Sec61. The two major modification factors are OST7 and the signal peptidase complex (SPC)25. Based on recent structural and functional studies10,11,23,24,26–29, and consistent with genetic co-dependency analysis across more than a thousand cancer cell lines30–32, four protein complexes are now assigned to co-translational membrane protein biogenesis: EMC, PAT, GEL, and BOS. EMC, whose core EMC3 subunit is a member of the Oxa1 family of insertases22, mediates insertion of SAs in the Nexo topology (N-terminus facing the exoplasmic side of the membrane)23,33. The other three complexes are recruited to the ribosome-Sec61 complex at a later stage of multipass biogenesis to form the multipass translocon (MPT)10,11,26,27. The MPT is thought to facilitate insertion of pairs of downstream TMDs connected by a short translocated loop5,6. This insertion might be facilitated by the GEL complex, whose TMCO1 subunit is another member of the Oxa1 family22, and does not seem to rely on the lateral gate of the Sec61 complex10.
Three factors are linked to co-translational Sec61-mediated secretion: TRAM34–36, TRAP37, and RAMP4 (also called SERP1, known as Ysy6 in yeast)18,38–40. Both TRAP and RAMP4 are tightly associated near-stoichiometrically with ER-localized ribosome-Sec61 complexes from pancreas18, an exceptionally secretory tissue. TRAM, although not tightly associated with ribosomes, can be crosslinked to SPs and TMDs at a point when they are at Sec61’s lateral gate34,41,42. Depletion of TRAM or TRAP reduces the capability of some SPs to successfully initiate translocation through Sec61 in vitro, placing their function at an early gating step34,35,37. The role of RAMP4 is largely unclear, but it can be crosslinked to nascent chains translocating through Sec6138, shows secretion defects in knockout cells and mice43 and it is induced by ER stress39. The mechanisms by which any of these factors impacts secretion are unclear, but would be aided by structural information on how they engage the ribosome-Sec61 translocon.
In this study, we have taken advantage of the heterogeneity of most stalled protein biogenesis intermediates assembled in vitro. Whereas substrates at certain key steps might have uniform and stable interactions with the biogenesis machinery12, others probably sample multiple states dynamically11. For example, both OST and MPT factors dynamically and heterogeneously associate at their overlapping sites behind Sec61. TMDs sample multiple environments such as the lateral gate, intramembrane chaperones, and the surrounding lipid. Hence, a substrate stalled at a single point during elongation can form multiple ribosome-translocon complexes. Improvements in electron cryo-microscopy (cryo-EM) imaging and particle classification now allow sample heterogeneity to be resolved into multiple discrete density maps44,45. When combined with rapid advances in structure prediction46–48, these maps can be fitted with reliable models. Using this approach, we present structures of a TMD insertion intermediate at Sec61’s lateral gate, RAMP4 and TRAP within ribosome-translocon complexes, and a new configuration of ribosomal protein uL22 that forms contacts with Sec61 and the nascent substrate.
Results and Discussion
We performed a thorough single-particle cryo-EM analysis of a previous dataset of RTCs engaged in biogenesis of the multipass membrane protein rhodopsin (Rho)10. The construct is termed Rhoext because it contains the first two TMDs of Rho and is extended at its N-terminus by fusion to an SP and an epitope tag (Figure 1A). The Rhoext intermediate in this sample has elongated to the point where the SP has been removed after directing translocation of the N-terminal domain across the membrane, RhoTM1 has inserted into the membrane, and RhoTM2 has emerged from the ribosome10. At this critical chain length, the nascent chain is poised at a point where it can potentially form several different RTCs (Figure 1B-E; Figure S1), only one of which was analyzed previously10.
Crosslinking and co-association experiments10,11,27 show that the nascent chain is just long enough for RhoTM1 to begin recruiting the PAT complex and initiate assembly of the MPT. Analysis of this subset of RTCs previously showed how the PAT complex latches Sec61 shut and redirects RhoTM2 toward the just-assembling MPT (Figure 1E)10. Because PAT complex recruitment and MPT assembly are just beginning, the dataset also contains multiple PAT-free complexes. We now analyze these particles via extensive classification (Figure S1) and find that without PAT, Sec61 is either closed, opened by RhoTM2, or opened by an unexpected factor, RAMP4 (Figure 1B-D). The closed structure is similar to the previously reported PAT-bound structure, so we focus on RhoTM2-and RAMP4-bound structures. We then compare various available structures, together with structure predictions and modelling, to reveal new insights into the structural dynamic of RTCs.
The structure of Sec61 inserting a membrane protein
In the absence of the PAT complex, the Sec61 complex is free to open. In the intermediate being analyzed, RhoTM2, whose eventual topology in the final protein would be Ncyt (i.e., the N-terminal flanking domain facing the cytosol), has just fully emerged from the ribosome with a 33 amino acid downstream tether to the P site tRNA. This is long enough for the first half of RhoTM2 to begin engaging the Sec61 complex, and a subset of RTCs displayed this event (Figure 2A,B). We find that the N-terminal half of RhoTM2 binds to Sec61 very similarly to the previously characterized hydrophobic helix (h-region) of a signal peptide (Figure 2C)12,13. Thus, both kinds of Ncyt substrate, SPs and TMDs, can bind Sec61 in the same way, despite TMDs having much longer hydrophobic helices than SPs do. The extra length of this TMD is passing through the channel pore, as will be discussed below.
Bound in this position, the N-terminal end of the hydrophobic region of the SP or TMD is only 11 Å (∼3 aa) from the tip of 28S rRNA helix 59 (H59) and the adjacent polyacidic tail of eL22 (Figure 2B). The backbone phosphates and the exposed aromatic rings of U2707-8 in H59, and the acidic residues of eL22, are potential binding sites for basic side chains. Importantly, basic resides are sharply enriched ∼3-5 aa from the hydrophobic helix49. Thus, H59 and eL22 are well positioned to engage cationic residues flanking a signal and retain them in the cytosol, a phenomenon called the positive-inside rule50. Notably, the next nearest rRNA, H47, is too far to serve this role (32 Å, ∼9 aa). Earlier lower resolution maps of a bacterial RTC similarly noted the proximity of H59 to the membrane and SecY’s lateral gate51. Given that H59 is also near the N-terminus of a SP bound to bacterial and mammalian signal recognition particle (SRP)52,53, H59 may contribute to the positive-inside rule throughout both targeting by SRP and insertion by Sec61.
To accommodate the TMD, the N-half of Sec61 has rotated out of the membrane plane toward the ribosome (Figure 2D). It is striking that the open conformation induced by the TMD is indistinguishable from that induced by an SP (Cα RMSD 0.691 Å; Figure 2C), despite their dissimilar sequences and despite Sec61’s continuous flexibility54,55. As an explanation for this bistable behaviour, we observe that the open conformation is stabilised by contacts between the N-half of Sec61α and the ribosome (Figure 2D). The N-half and ribosome have been thought to be isolated from one another, but we find that Sec61 residues 21-27 contact the 28S rRNA helix 24 and uL24, including a particularly well-resolved cation-π interaction between Sec61α R24 and 28S A389. Bacterial SecY lacks this entire loop, perhaps because bacterial secretion is driven by SecA, which competes with ribosomes2,56,57. Archaeal SecY, however, does conserve this loop’s structure (Cα RMSD 0.67 Å with M. jannaschii) and consensus sequence (euk. KPERKIQ vs arc. KPERKVSL, with the cation-π arginine underlined). Stabilising interactions with this widely conserved motif may help Sec61 respond to its diverse substrates with a consistent open state.
Whereas the N-terminal half of RhoTM2 is helical, its C-terminal half loops back through the Sec61 pore in an unfolded conformation similar to the segment of polypeptide downstream of a SP during secretion12,16. This is accommodated by the plug helix moving toward the lumenal tip of Sec61γ, effectively becoming part of the channel’s lumenal funnel (Figure 2E), as previously seen with the bacterial plug13,16. Our observation that different segments of a TMD can simultaneously occupy the lateral gate and central channel is consistent with the through-pore model of insertion, rather than the sliding model21, at least for this substrate. The through-pore translocation may be favoured because the ribosome exit tunnel’s mouth holds an emerging TMD closer to the Sec61 pore than to the lateral gate. Moreover, the free energy cost of passing a hydrophobic TMD through the hydrophilic pore could be mitigated by the hydrophobic pore ring and partly hydrophilic TMD.
Prior structures of ribosome-bound, open Sec61 disagreed on the structure of the channel pore due to differences in how the rotations of core helices were modelled (Figure S4)7,12. With ∼3.7 Å resolution (Figure S2A) and 1.8-3.5 Å predicted aligned error (PAE) restraints from the AlphaFold2 (AF2) prediction of Sec61, we find that the pore is ringed by four aliphatic residues: V85, I183, I292, and L449. Nearby residues F42, I81, I84, L89, I179, I187, I289, and I453 extend hydrophobic patches outward from the pore ring (Figure 2F). The four pore ring residues identified here are homologous to the four pore ring residues of bacterial SecY14,16,58, and adopt similar positions as in the open SecY structure (Cα RMSD 1.1 Å)16. These four residues are pseudosymmetric and likely inherited from SecY’s dimeric ancestor58. Thus this structure shows that the likely ancestral pore architecture is universally conserved across the SecY family.
The structure of RhoTM2 trapped during insertion via the central channel of Sec61 represents a paradigm for insertion of Ncyt signal anchors and the subset of Ncyt TMDs in multipass membrane proteins followed by a long (>100 aa) translocated loop. Both types of substrates are potently inhibited by Sec61 inhibitors10,33,59–64 whose binding site at the lateral gate is mutually exclusive with the position of the TMD observed in our structure65,66. In the case of Ncyt TMDs followed by a short (<50 aa) translocated loop and another TMD, insertion is thought to occur as a TMD-pair via an Oxa1 family member such as the GEL complex within the MPT5,6,10,11. Given that Rho biogenesis is not sensitive to Sec61 lateral gate inhibitors, RhoTM2 seems to normally be inserted together with RhoTM3 via the GEL complex10. Although the structure seen here probably represents an alternative route captured due to ribosome stalling, it nonetheless proved to be an illuminating paradigm.
RAMP4 occupies the Sec61 gate
Alongside a RhoTM2-bound map, image classification yielded a map in which Sec61 is bound to a different and unexpected density. This density consists of a kinked TMD bound to the Sec61 lateral gate and a ribosome-binding domain (RBD; Figure 3A). The resolution of the RBD density was sufficient for de novo modelling, identifying the sequence as RAMP4. The location, structure, and function of RAMP4 has long been unclear.
RAMP4’s RBD is hook-shaped (Figure 3B). Its N-terminus is comprised of an α-helix (residues 5-15) flanked by 310-helices (3-7,13-20). Subsequent residues then loop back along the helix. The hook is stabilised by an intramolecular hydrogen bond (K13-S28) and a dense network of intermolecular contacts with the ribosome. Specifically, the RBD binds the 28S rRNA’s helices 47, 57 and 59 and ribosomal proteins eL19, 22, and 31, via electrostatic interactions and via hydrophobic interactions with pockets on each of the three ribosomal proteins. The only other factor known to bind in this region is the nascent polypeptide-associated complex subunit β (NACβ)67, whose anchor domain has a very different structure but would partly clash with the RAMP4 RBD.
RAMP4’s RBD is connected by a flexible linker to the kinked TMD. This region is less well resolved, so to inform modelling we used AF2 to predict the structure of the Sec61•RAMP4 complex. The resulting prediction is confident (Figure S3A) and agrees with the density map. To check that this prediction was specific to RAMP4, we used AF2 to screen a diverse panel of other TMDs and SPs, and found that RAMP4 was indeed the only protein predicted to bind Sec61 (Figure S3B). Thus our model of the RAMP4 TMD is supported by both the density map and a confident, specific structure prediction.
The cytoplasmic half of the RAMP4 TMD is hydrophobic and binds the open Sec61 gate as if it were a TMD or SP (Figure 3C). Bound here, RAMP4 holds the Sec61 pore ring wide, just as SPs and TMDs do (Figure 2C). Unlike an SP or TMD, however, RAMP4 does not displace the plug helix. Instead, the plug moves together with the widening pore ring, keeping it plugged. This observation contrasts with prior speculation that pore widening is sufficient to trigger plug displacement12. Pore widening and unplugging do indeed occur together when the gate is opened by an SP or TMD (Figure 2E), but RAMP4 shows that widening can occur without unplugging. This implies that unplugging occurs when an SP or TMD pulls a flanking segment of nascent chain through the channel pore, and this necessarily displaces the plug. RAMP4, being a tail-anchored protein, threads nothing through the pore, and thus does not clash with the plug.
At its midpoint, the TMD of RAMP4 is kinked 40° at a conserved glycine, and the lumenal half of the TMD is amphipathic. The TMD’s hydrophilic face is oriented toward the channel interior and completes the hydrophilic lumenal funnel of Sec61α (Figure 3D). RAMP4’s integral contribution to forming the Sec61 channel explains why it, and not the more peripheral Sec61β or γ subunits, crosslinks with nascent chain-encoded photoprobes in the Sec61 channel38.
Comparing different classes of particles, we see that RAMP4 competes with other factors. RAMP4 is partially or completely depleted from classes in which the gate is closed, occupied by RhoTM2, or latched shut by the PAT complex (Figure S5). By competing with RAMP4’s TMD, RhoTM2 and PAT also reduce the occupancy of its RBD, which indicates that the RBD alone binds too weakly to be retained during extraction, purification and grid preparation. This is consistent with biochemical evidence that RAMP4’s association is detergent-sensitive18. Together, these factors explain why RAMP4’s occupancy in prior cryo-EM maps was low enough to be overlooked, although in hindsight seems to be visible in several7,68,69.
To assess the abundance of Sec61•RAMP4 complexes in native membranes, we examined the best native maps currently available, from the subtomogram averages of ribosomes from human cell-derived microsomes29. This dataset was separated into four classes of RTCs: 70% Sec61•TRAP•OSTA, 12% Sec61•MPT, 10% Sec61•TRAP, and 9% Sec61•TRAP•MPT. Examining each class average for the RAMP4 RBD, we find that it is present in the classes without MPT and absent from the classes with MPT, consistent with our single-particle analysis. We then estimated the occupancy of RAMP4 using OccuPy70, and found that it is present in ∼85% of Sec61•TRAP•OSTA RTCs and ∼53% of Sec61•TRAP RTCs (Figure S5), which equals ∼81% of the non-MPT RTCs. Thus, RAMP4 is absent from MPT-containing RTCs and seems to be present in almost all non-MPT RTCs.
Alongside the RAMP4 RBD, the subtomogram averages also show density for RAMP4’s kinked TMD (Figure 3E). Previously, this TMD density had been speculated to represent SPs7,69,71. But the helical density observed is ∼25 aa long, like RAMP4’s TMD, whereas the helices of SPs are only 7-15 aa long72. Moreover, the occupancy of this TMD is high, even after hours of translation inhibition (Figure S5)29, when SP occupancy should be low due to their co-translational dissociation and subsequent processing and degradation25,42,73–75. We therefore assign the kinked TMD density in these earlier maps to RAMP4, implying that most co-translational translocation normally occurs through Sec61•RAMP4 channels.
Having described the structure of mammalian RAMP4, we briefly consider how widely conserved this structure may be. Examining representative model organisms (S. cerevisiae and C. reinhardtii), we find that in both cases RAMP4 is confidently predicted to bind Sec61 like animal RAMP4 does (Figure S3A). Especially conserved is the region surrounding RAMP4’s glycine kink and Sec61 N300 (Figure 3F,G), which is part of Sec61’s polar cluster that is important for gating76. To assess when RAMP4 arose in evolution, we searched for homologs in several early-branching eukaryotic taxa (Discoba, Metamonada, Malawimonada) and found it to be largely absent from those groups, suggesting an origin more recent than the last eukaryotic common ancestor. Thus RAMP4, although not ubiquitous, is present across the major eukaryotic kingdoms and forms a dynamically associating part of the Sec61 channel.
uL22’s C-tail switches to contact Sec61 and the nascent chain
In our non-MPT maps, we were surprised to observe density for the C-tail of ribosomal protein uL22 (Figure 4A). This tail has not been described in any prior structural studies. Here, we find it stretched across the ribosome’s membrane-facing surface toward Sec61 and the nascent chain, contacting eL31 and several RNA helices along the way (Figure 4A). Density for the C-terminal helix (CTH) is moderately strong (Figure S2A) and similar in each class of particles, indicating that it is not strongly correlated with TRAP or RAMP4 occupancy, nor Sec61 conformation or nascent chain binding to Sec61. The sole exception is the absence of the CTH from the MPT map10, where it would clash with the gate latch helices of the PAT complex (Figure 4A). It would also clash with SRP (via SRP54’s M-domain)53,77,78, NAC (via its ribosome-binding helices)67, RAC (via Zuo1)79, NatA (via Nat1)80 (Knorr et al. 2019), and NatB (via MDM20)81.
Having seen that uL22 can adopt an engaged or displaced state in different ribosome-translocon complexes, we then compared these states to a map of cytoplasmic ribosomes without bound translocons (EMD-40205, the best-resolved mammalian map available)82. The cytoplasmic ribosome displays clear density for the uL22 tail up to the point where it binds H24/47, but the following CTH is diffuse, indicating that it is flexibly anchored to H24/47. Anchored here, the CTH would be well-positioned to scan emerging nascent chains and sense when Sec61 binds.
To judge how common uL22 engagement is, we re-examined publicly available RTC maps for previously overlooked uL22 density. We found clear uL22 CTH density in maps from several recent studies (Figure S6)29,83,84. Most importantly, it is present in subtomogram averages of RTCs transporting diverse endogenous nascent chains through intact microsomal membranes29, and the uL22 CTH density in those maps is just as strong as in the present dataset (∼40% occupancy; Figure S6A). This indicates that uL22 engagement is common in native translocons. It is unclear why it is not visible in earlier maps; it may be present and not well-resolved, or absent due to a preference for specific nascent chain features.
Comparing the uL22 tail across species, we find that it is conserved by the earliest branches of the animal tree, but not by the nearest non-animal branches (Figure 4B). Although fungi have an elongated uL22 tail, it is unlike the animal tail in sequence, and appears to have arisen independently. Indeed structures show that the fungal uL22 tail binds to an entirely different site (Figure 4C)85, and does so constitutively instead of dynamically. Taken together, this suggests that the animal uL22 tail structure was acquired during the evolution of the first animals. Among animals, the uL22 tail’s conservation suggests that it is functionally important. Particularly well-conserved is the SXKK motif that initiates the CTH and binds H24/47 (Figure 4D). The tip of the CTH that would contact nascent chains typically contains a mix of basic, acidic, and hydrophobic residues, and this complexity makes it difficult to predict what if any nascent chain features it may recognise.
It is noteworthy that when engaged, the uL22 CTH blocks a gap between the ribosome and Sec61 that would otherwise allow the nascent chain to exit the channel vestibule and enter the cytoplasm (Figure 4E). This gate-side exit is one of two such exits, with the other being on the Sec61 hinge side, where multipass proteins exit for insertion by the MPT10. While uL22 blocks the gate-side exit, the hinge-side exit would remain open. The functional consequences of this change are currently unclear.
Structure of the TRAP-calnexin complex
The occupancy of the hetero-tetrameric TRAP complex (comprised of α, β, γ, and δ subunits) in the all-particle map was low, so we performed focused classification on this region and obtained two well-resolved TRAP classes (Figure S1C). TRAP in these two classes adopts slightly different tilts with respect to the ribosome, which we call conformations 1 and 2, although the true range of TRAP conformations is presumably continuous rather than discrete. The TRAP maps were well fit by a predicted model (Figure 5A) after minor adjustments. One additional protein density was observed in TRAP class 1, consisting of a TMD and cytoplasmic helix bound to TRAPγ. We provisionally assign this density to Calnexin because it is by far the major TRAP-binding protein37,86 and it is known to bind TRAP via its TMD. The TRAP-calnexin interaction was not detected by AF2, perhaps because Calnexin di-palmitoylation, which AF2 ignores, is crucial for TRAP interaction with Calnexin87. Anchoring of Calnexin at this position would explain how its elongated lumenal domain interacts co-translationally with nascent chains as they are translocated by RTCs88,89.
The TMDs of TRAPβγδ and calnexin all bundle together, but not the TMD of TRAPα. Instead, the TRAPα TMD binds the lumenal hinge loop of Sec61α, as does the TRAPα lumenal domain (Figure 5B.3). While making those lumenal contacts, the TRAPα TMD tilts its cytoplasmic end away from Sec61α, allowing conserved basic residues flanking its TMD to contact the tip of 5.8S rRNA helix 7 (Figure 5B.4). Strikingly, this rRNA segment has undergone a dramatic rearrangement from its cytoplasmic state, which puts C81, U85, and U86 in contact with the membrane and the basic residues in TRAP (Figure 5B.5). From there, the C-tail of TRAPα continues along the ribosome surface and binds tightly at a site contacting uL23, uL29, and 5.8S rRNA helix 9 (Figure 5B.6). Thus TRAPα makes extensive contacts with both the ribosome and Sec61.
Besides TRAPα, TRAPβ, TRAPγ, and Calnexin also contact the ribosome and Sec61: TRAPβ’s basic C-terminal residues contact the 5.8S rRNA at H9 ES3 (Figure 5B.7); TRAPγ’s helical hairpin residues R110, K111, and especially R114 contact the 28S rRNA at H54 and ES26 (Figure 5B.8); TRAPγ’s flexible N-tail helix binds eL38 (Figure 5B.9); Calnexin’s flexible C-tail contacts the 28S rRNA at H63 ES27a (Figure 5B.10); TRAPγ’s C-terminus contacts Sec61γ’s N-terminus (Figure 5B.1); and TRAPγ’s lumenal loop 3/4 contacts Sec61α’s lumenal loop 8/9 (Figure 5B.2). Altogether, TRAP’s multivalent site-specific and non-specific interactions across many sites via both rigid and flexible elements explains how it remains associated with the translocon while adopting a range of positions and orientations, compared to other translocation factors like OST or the MPT which bind at relatively fixed sites.
TRAP competes and cooperates with different translocon subunits
The TRAPδ Ig-like domain contains a helical hairpin between β-strands 6 and 7. No similar hairpin appears in any other domain in the AFDB50 database queried by Foldseek90 (Figure S7A). This unique hairpin is isolated in our Sec61-TRAP structure, but in Sec61-TRAP-OSTA structures29, it extends toward OST-A subunit RPN2 (Figure 6A). A basic patch on TRAPδ is separated by less than 5 Å from an acidic patch on RPN2, indicating that they would share an electrostatic attraction (Figure 6A, inset). Each partner presents three conserved charges, and the closest pair (K117 on TRAPδ and D386 on RPN2) is the most conserved, indicating that their attraction is functionally important. OSTA’s attraction to TRAPδ is weak compared to its binding to the ribosome, but TRAPδ may nonetheless help recruit OSTA, since TRAPδ would attract OSTA from most possible angles of approach, whereas OSTA’s ribosome contacts are stereospecific. The TRAPδ-OSTA interaction may explain why TRAPδ defects cause congenital disorders of glycosylation91–93.
Comparing the TRAP structure to the MPT structure10, we find that TRAP would clash with part of the MPT, namely the BOS complex comprised of TMEM147, Nicalin and NOMO. The prior BOS model omitted NOMO, so to fully characterise this clash we ran a prediction of the full BOS complex structure, obtaining a high-confidence prediction that fits a previously reported subtomogram average7 (Figure 6B). As an aside, it is noteworthy that the first pre-albumin-like domain (PLD1) in NOMO is predicted to bind PLD10 and PLD11, suggesting that its reported ability to regulate the spacing between ER sheets94 results from it forming antiparallel homodimers across the ER lumen (Figure S7B).
Comparing the TRAP and BOS structures, TRAPα competes with BOS subunit TMEM147 for the same binding sites on the hinge loop of Sec61α and the 5.8S rRNA’s helix 7, while PLD12 of NOMO would clash to a more limited degree with the lumenal domain of TRAPβ. When TRAP is displaced by BOS, its interactions with Sec61 are disrupted and its transmembrane bundle (which is no longer visible in the cryo-EM map) evidently pivots toward the still-bound TRAPα C-tail RBD, causing a pronounced shift in the shape of the detergent micelle (Figure 6C). This extensive competition explains why prior studies found TRAP in only 40% of MPT complexes, but at high occupancy at all other RTCs29
Additional functionally relevant TRAP features
The above analysis shows that TRAP binds together ribosomes, OSTA, and Sec61, but competes with the MPT, whose presence inhibits Sec61. In principle, these activities could suffice to explain TRAP’s observed functions as a stimulator of Sec61-dependent secretion37 and OSTA-dependent glycosylation93. However, we observe four additional TRAP features that appear functionally relevant and are therefore noteworthy.
First, TRAPα presents a conserved patch to the nascent chain where it first emerges from Sec61 (Figure 7A). This patch could potentially bind the nascent chain or redirect it toward the active site of OSTA. Second, by binding the C-half of Sec61, TRAP can potentially influence the opening of the lateral gate (Figure 7B). If TRAP caused the C-half to favour opening, this would explain in part why it stimulates the recognition of weakly-gating signals37. The combined effect could explain its overall stimulatory role in translocation.
Third, TRAP prefers a membrane plane tilted 20° relative to Sec61’s, and imposes this curvature on the surrounding micelle (Figure 7C). The same tilt is observed in Sec61-TRAP maps from intact membranes, but it is unclear whether TRAP imposed this curvature on the membrane, since it is the same degree of curvature found in native ER tubes and sheet edges95. Thus while it is possible that TRAP modulates Sec61 activity by disrupting the local membrane, it may instead have adapted to reside in and sense pre-existing membrane deformations.
Fourth, the lumenal domain of TRAPα near the Sec61 channel has an extraordinarily long and acidic N-tail anchored tipped by hydrophobic residues (Figure 7D). This could interact with any lumenal part of the translocon or nascent chain. No data is available on this tail’s function, aside from the fact that it and calnexin each bind far more calcium than any other protein in ER membrane extracts86 as expected for its exceptional charge and abundance.
TRAP conservation across eukaryotes and archaea
Having described the features of metazoan TRAP, here we briefly contrast it with TRAP in representative model organisms from plantae (C. reinhardtii), fungi (S. cerevisiae), and excavates (Trypanosoma brucei), a group of early-branching eukaryotes. C. reinhardtii has already been shown by subtomogram averaging to contain TRAPαβ-like density at its translocons96. Indeed we find that its genome contains TRAPαβ but not γδ, and the predicted structure of Cr TRAPαβ-Sec61 fits the tomographic density with minor adjustments (Figure 8A). Cr TRAP conserves animal TRAP’s predicted binding to the Sec61 hinge and its C-tail RBD. However, whereas animal TRAP binds to the ribosome at eL38 via TRAPγ, Cr TRAP is predicted to bind this same ribosomal protein via TRAPβ (Figure 8A, inset). Thus the Cr and animal complexes share the similar ribosome-binding sites despite their differences in composition. As an aside, it is noteworthy that the Cr TRAPβ TMD is too hydrophilic to insert on its own (ΔGpred = 1.824, i.e. 5% insertion), suggesting that it is bound and stabilised by an unidentified protein functionally analogous to TRAPγ.
S. cerevisiae has no annotated TRAP genes, and indeed it has been suggested that most fungi lack TRAP96. However, we find that S. cerevisiae contains a previously unidentified gene for TRAPα (Irc22, HHpred E = 10-35). This suggests TRAP’s prevalence in fungi has been underestimated. Sc TRAPα conserves animal TRAPα’s predicted binding to the Sec61 hinge and its C-tail RBD (Figure 8B). No other TRAP subunits were detected in cerevisiae by sequence-or structure-based searches (HHpred, Foldseek). Thus TRAP’s links to co-translational translocation may be conserved in fungi, but its complexity is dramatically reduced.
The early-branching excavate T. brucei has TRAPα, β, and γ, but no δ (Figure 8C). Tb TRAPαγ conserve the same predicted interactions as their homologs in fungi, plants, and animals (where found). Overall, the observed distribution and conservation of features suggest that the original eukaryotic TRAP consisted of TRAPαβγ, bound the Sec61 hinge loop via TRAPα, bound the ribosome flexibly via the TRAPα C-tail RBD and the TRAPγ N-tail, and also contacted the ribosome via the TRAPβγ bundle.
If TRAPαβγ were present in the last eukaryotic common ancestor, they may have been inherited from archaea, which are the ancestors of eukaryotes. Structural queries of eukaryotes’ archaeal sister group (Heimdallarchaeota) identified candidate homologs of TRAPα (E = 9.95×10-8), β (E = 5.15×10-10), possibly TRAPγ (E = 1.84), and not TRAPδ, as expected. To test whether these candidates are also similar to TRAPαβγ in sequence, we used them to perform reciprocal HHpred queries of the human proteome, and in each case the corresponding human TRAP protein was the top hit (E = 0.031 for TRAPα, 9.4×10-14 for TRAP β, and 110 for TRAPγ).
We then queried AF2 to see whether the top-scoring archaeal TRAPαβγ hits were predicted to form a complex with each other or with SecY. No contacts among the top hits were detected, but this could be an artefact of how few sequences AF2 could collect to constrain the prediction (as few as 5). By contrast, contacts were confidently predicted between archaeal TRAPα and SecY, forming a dimer similar to the eukaryotic TRAPα-Sec61 dimer (Figure 8D). This similarity was predicted even though no eukaryotic TRAPα were included in the input alignment or training set, and thus it was based on archaeal TRAPα alone.
While performing sequence searches with TRAPα and β, we were surprised to find that they are much more similar in sequence than we had expected (HHpred p = 2×10-7), suggesting that they share a common ancestor. In fact they are more similar to one another than to any other human proteins (HHpred), including other Ig-like domains. By contrast, TRAPδ is quite dissimilar (HHpred α p = 0.64, β 0.66). Thus although TRAP’s three lumenal domains all have Ig-like folds, it appears that two of them, TRAPα and β, originated when an archaeal proto-TRAP protein duplicated, whereas TRAPδ is evolutionarily unrelated (Figure 8E). If proto-TRAP formed dimers, it would presumably resemble a homomeric version of the heterodimeric TRAPαβ complex found in plants96.
Toward understanding TRAP’s mechanism, it is noteworthy that although its affinity for the ribosome and Sec61 are conserved, at least two of the four other functionally relevant features highlighted above are not conserved: yeast TRAP’s single TMD is unlikely to deform membranes, and TRAP from many organisms does not have a polyacidic lumenal domain. The other two features, allosteric effects on gating and a putative substrate-binding activity, could plausibly be conserved together with the TRAPα-Sec61 structure.
Conclusions and perspective
The most striking finding of this work is that the Sec61 protein secretion channel essentially has a fourth subunit, RAMP4, which is present in about 80% of non-MPT ribosome-translocon complexes. Our structure shows that RAMP4 binds to Sec61 like a SP, thereby blocking the channel’s lateral gate, holding wide its central pore, and completing its lumenal funnel. For these reasons, we speculate that RAMP4 acts as a surrogate SP. It is possible that once a secretory protein’s SP dissociates from the lateral gate, the central channel would narrow, thereby impeding translocation speed or efficiency. By binding the lateral gate during these later stages of translocation, RAMP4 could smooth transport through the channel for certain types of sequences. Such a role would explain why RAMP4 seems to be present during the vast majority of co-translational translocation through Sec61. Although RAMP4 appears to be eukaryote-specific, other proteins may serve similar functions in some prokaryotes. For example, the first TMD of E. coli YidC invades the lateral gate of SecYEG97,98, as does the TMD of PpiD99, a periplasmic chaperone.
A second noteworthy advance in this work is a relatively clear view of how an Ncyt TMD initially engages Sec61, revealing a mode of interaction very similar to SPs. In both cases, the positive-inside rule may be enforced in part by interactions with rRNA helix 59 and the anionic C-tail of eL22. The difference is that the TMD is more than twice as long as the hydrophobic region of an SP. Hence, while the N-terminal part of the TMD binds like an SP, its C-terminal part loops back through the channel pore similar to the mature domain downstream of an SP. This finding supports the through-pore model for TMD insertion rather than the sliding model, in which TMDs would instead translocate through the lipid phase at the lateral gate21. It remains to be seen whether this model also applies to Nexo TMDs, which unlike Ncyt TMDs, are highly refractory to inhibitors that block Sec61’s lateral gate10,33,59.
The third major area of insight concerns the TRAP-Calnexin complex, adding to and refining recent structural models of the TRAP complex29,83,84,100. We describe the extensive network of contacts with the ribosome and Sec61, potentially responsible for TRAP’s effects on secretion. We show how a unique hairpin on the lumenal domain of TRAPδ forms a bridge to RPN2 in OSTA, potentially explaining why TRAP deficiency causes glycosylation defects93. Unlike this TRAP-OSTA cooperation, we observe competition between TRAP and the BOS complex for the same binding site on Sec61 and the ribosome, explaining why TRAP is depleted and disordered in MPT-containing RTCs. Our sequence analysis and structure predictions for TRAP complexes in taxa beyond mammals reveal conserved features indicating that the original eukaryotic TRAP complex was similar to mammalian TRAPαβγ, which probably evolved from a predecessor in archaea.
These three major areas of insight, together with a number of additional findings regarding Sec61-ribosome interactions, the Sec61 pore ring, the dynamic tail of uL22, and a marked conformational change in 5.8S rRNA on translocon binding collectively lead to a wealth of new hypotheses regarding protein translocation at the ER. A major theme emerging from this and other studies in recent years is that the translocon is not static in either its composition or conformation. Instead, various factors and domains are often mutually exclusive, requiring dynamic reorganisation in ways small and large. Understanding how different types of substrates drive such reorganisation to facilitate their biogenesis remains a major challenge. This problem is analogous to the cytosol where multiple factors must dynamically access the ribosome exit tunnel to triage the nascent protein toward different fates101. Our analysis of various new translocon configurations sheds light on how this complex machinery facilitates secretory and membrane protein biogenesis.
Methods
Sample preparation and electron microscopy
Sample preparation and electron microscopy was described previously10. In brief, the in vitro transcription reaction used a PCR-generated template containing the SP6 promoter102,103. The transcription reactions were for 1 hour at 37°C. The resulting transcript was used without further purification and was diluted 1:20 in the IVT reaction, which was carried out in rabbit reticulocyte lysate (RRL) as described previously102,103. The reaction was supplemented with canine rough microsomes (cRMs) prepared and used as described previously104. The translation reaction was incubated for 30 min at 32°C, then halted by transferring the samples to ice. All further steps were performed at 0-4°C, unless stated otherwise.
A 2 ml translation reaction was divided in four, and each aliquot layered on a 500 µl cushion of 20% sucrose in 1x RNC buffer (50 mM HEPES-KOH, pH 7.5, 200 mM KOAc, 5 mM Mg(OAc)2). The microsomes were sedimented by centrifugation at 4°C in the TLA-55 rotor (Beckman) at 55,000 rpm for 20 min. The cRM pellets were each resuspended in 25 μl of RNC buffer and pooled. The sample was incubated with 250 μM BMH on ice for 15 min and quenched with 5 mM 2-mercaptoethanol. The microsomes were diluted with 400 μl of solubilisation buffer (RNC buffer containing 1.5% digitonin) and incubated for 10 min on ice. The digitonin was obtained from Calbiochem and further purified as described previously18. The sample was centrifuged at 20,000 × g and 4°C for 15 min. The supernatant was transferred to a tube containing 20 μl of StrepTactin High Performance Sepharose beads (GE Healthcare) and incubated for 1.5 h at 4°C. The resin was then washed five times with 0.5 ml RNC buffer containing 0.25% digitonin and eluted by incubation for 1 h on ice with 40 μl of RNC buffer containing 0.25% digitonin and 50 mM biotin. The absorbance of the eluate for both samples was 3.4 at 260 nm.
The affinity-purified RNCs were vitrified on UltrAuFoil R 1.2/1.3 300-mesh grids (Quantifoil) coated with graphene oxide (Sigma-Aldrich). In a Vitrobot Mark IV (Thermo Fisher Scientific) at 4°C and 100% ambient humidity, each grid was loaded with 3 μl of sample, blotted 4 sec with Whatman filter papers at a blot force of -15, and plunge frozen in liquid ethane at 92 K. Automated data collection was performed on a Titan Krios microscope (Thermo Fisher Scientific) equipped with an XFEG source operating at an accelerating voltage of 300 kV. Defocus was programmed to range between 2.7 and 1.9 μm. Movies were captured using a K3 Bioquantum direct electron detector (Gatan) operating in super-resolution mode. Movies were dose-fractionated into 54 frames covering a total dose of 54 e-/Å2. 17,540 images were collected.
Image processing
Movie frames were motion-corrected using MotionCor2105 with 7×5 patches and dose-weighting, and their contrast transfer functions (CTFs) were fit using CTFFIND 4.1106. After manual curation, 285 of 12,540 micrographs (2.3%) were discarded due to poor CTF fits or thick ice. Subsequent steps were performed in RELION-4.0107. Manual picking on 20 randomly selected micrographs yielded 2,745 particles, which were used to train the automatic particle-picker Topaz108. Topaz was run on all micrographs and picks assigned a figure of merit (FOM) above -3 were retained. This cutoff was chosen based on a histogram of pick FOMs, which was approximately normal above -3 but displayed a long lower-FOM tail, and subsequently checked against the micrographs to verify sensible results. Picked particles were extracted in 412 px boxes (1.34 Å/px), binned to 128 px (4.31 Å/px), and classified in 2-D with a 300-Å diameter mask, 200 classes k and a regularisation parameter T of 2. 80 2-D classes showing clear molecular features were retained, encompassing 1,188,459 particles. These particles’ coordinates were used to retrain Topaz. The retrained Topaz model picked 1,389,410 particles at FOM ≥ -1, a cutoff selected by the same criterion as above.
The picked particles were extracted in 412 px boxes, binned to 128 px, and refined in 3-D against a mammalian ribosome reference map lowpass-filtered at 70 Å, yielding a Nyquist-limited (8.62 Å) map. With fixed alignments from this refinement, particles were classified in 3-D (k = 20, T = 4), yielding 80S (38%), 80S ratcheted (15%), 60S (10%), poorly resolved (36%), and ten noise (1%) classes, as well as 6 empty classes. The 875,065 well-resolved ribosomal particles were combined, re-extracted in 420 px boxes without binning, and refined in 3-D to obtain a 2.86 Å map, which improved to 2.68 Å after CTF refinement and Bayesian polishing. This is the all-particle map.
The aligned particles were then subclassified without realignment using three different masks for residual signal subtraction: a tight mask surrounding the mobile parts of Sec61 and RAMP4, a tight mask surrounding the TRAP complex, and a loose mask encompassing both Sec61•RAMP4 and TRAP. The loose masking did not clearly separate most classes, but did provide the clearest view of TRAPα’s TMD, which crosses the boundary between the tighter masks. The resulting classes are shown in the processing flowchart (Figure S1C). The well-resolved TRAP classes were further subclassified to obtain maps where Sec61 was also more homogeneous. This second subclassification yielded noisier classes because the particle numbers are much smaller, but the TRAP and Rho-bound Sec61 classes yielded clear density. After reconstruction, density maps’ local scale and occupancy were estimated using OccuPy70. Maps were postprocessed using Relion’s MTF-correction and automated B-factor sharpening, and were also lowpass-filtered where indicated. Some maps were in parallel postprocessed using DeepEMhancer109 for use in rendering figures.
Molecular modelling
The 60S subunit and P-site tRNA from PDB 7tm3 were used as an initial model for the rabbit ribosome. An initial model for uL22 was fetched from the ΑlphaFold DB110. For the Sec61-RAMP4 complex and TRAP complex, initial models were generated using colabfold2111 with ΑlphaFold2-Multimer v347. Default options were used, except that the top-scoring models were refined by AMBER relaxation to avoid clashes that would interfere with subsequent simulations. Separate multimer predictions were also generated for the interactions of TRAP with Sec61α, TRAPγ with eL38, and TRAPα with uL23/29, for use as references for modelling restraints. All sequences were fetched from UniProt112.
The initial models for these complexes were fit to density using ISOLDE113 molecular dynamics flexible fitting (MDFF) with adaptive pLDDT-and PAE-dependent restraints114. Segments that were not confidently predicted, as indicated by high PAE or low pLDDT scores, were omitted unless they could be built based on the cryo-EM density alone. Final real-space refinements were performed using PHENIX115. Three rounds of global minimisation and group B-factor refinement were performed, with tight secondary structure, reference model, rotamer, and Ramachandran restraints applied. Secondary structure-and reference model restraints were determined from the starting models. Hydrogen-bonding and base-pair and stacking parallelity restraints were applied to the rRNA. Final model statistics are provided in table S1. Models were rendered using ChimeraX116.
No refinements were performed for the BOS complex model, nor the four non-mammalian TRAP•Sec61 models. The C. reinhardtii TRAP model predicted by AF2 was fitted to the tomographic density map EMD-4145 in ISOLDE using its ribosome interactions as anchor points (the Sec61-ribosome binding site and predicted TRAPβ-eL38 binding site). For each non-mammalian TRAP•Sec61 model, separate predictions were run to probe for interactions with eL38 or uL29. For the panel of multimeric predictions used to test whether RAMP4’s predicted interaction with Sec61 was a positive outlier, a separate prediction was run for Sec61 with each of the other sequences in the panel. The resulting predictions were checked visually to confirm that none besides RAMP4 engaged the lateral gate, and a Python script was used to gather the intersubunit PAEs and generate a violin plot of their distribution.
Data availability
The EM maps have been deposited in EMDB with the accession numbers EMD-19195, EMD-19196, EMD-19197, EMD-19198, 19199, EMD-19200, EMD-19201, EMD-19202, EMD-19203, and EMD-19204. The molecular models have been deposited in the PDB with accession numbers 8RJB, 8RJC, and 8RJD. The AlphaFold2-predicted structures reported in this study have been deposited in ModelArchive as follows:
ma-jjnuw - Chlamydomonas reinhardtii Sec61-RAMP4 complex
ma-hbsof - Saccharomyces cerevisiae Sec61-RAMP4 complex (aka Ysy6p)
ma-hknzh - Canis lupus familiaris Sec61-RAMP4 complex
ma-x3uvj - Heimdallarchaean TRAPα-SecY complex
ma-j8wag - Trypanosoma brucei TRAP-Sec61 complex
ma-9gsmk - Chlamydomonas reinhardtii TRAP-Sec61-RAMP4-eL38 complex
ma-l9qqq - Saccharomyces cerevisiae Sec61-TRAP complex (aka Irc22p)
ma-ise4t - Canis lupus familiaris BOS complex (TMEM147/Nicalin/NOMO)
Funding
This work was supported by the Medical Research Council as part of United Kingdom Research and Innovation (MC_UP_A022_1007 to R.S.H.).
Declaration of interest
R.S.H. is a scientific advisor and equity holder of Gate Bioscience.
Acknowledgements
We are grateful to Robert Keenan for helpful discussions; Min Kyung Kim and the staff of LMB’s EM facility for collecting the dataset on which this study’s analysis is based; J. Grimmett and T. Darling of LMB’s Scientific Computing for support.
References
- 1.X-ray structure of a protein-conducting channelNature 427:36–44
- 2.Structural and Mechanistic Insights into Protein TranslocationAnnu. Rev. Cell Dev. Biol 33:369–390
- 3.A clearer picture of the ER translocon complexJournal of Cell Science 133
- 4.Protein biosynthesis at the ER: Finding the right accessoriesMol. Biol. Cell 34
- 5.The mechanisms of integral membrane protein biogenesisNature Reviews Molecular Cell Biology 23:107–124
- 6.The Biogenesis of Multipass Membrane ProteinsCold Spring Harb. Perspect. Biol 15
- 7.Structural basis for coupling protein transport and N-glycosylation at the mammalian endoplasmic reticulumScience 360:215–219
- 8.The translocon: A dynamic gateway at the ER membraneAnnual Review of Cell and Developmental Biology 15:799–842
- 9.The concept of translocational regulationJ. Cell Biol 182:225–32
- 10.Mechanism of an intramembrane chaperone for multipass membrane proteinsNature 611:161–166
- 11.Substrate-driven assembly of a translocon for multipass membrane proteinsNature 611:167–172
- 12.Structure of the Sec61 channel opened by a signal sequenceScience 351:88–91
- 13.Crystal structure of a substrate-engaged SecY protein-translocation channelNature 531:395–399
- 14.The SecY complex forms a channel capable of ionic discriminationEMBO Rep 10:762–768
- 15.Preserving the membrane barrier for small molecules during bacterial protein translocationNature 473:239–242
- 16.Structure of the substrate-engaged SecA-SecY protein translocation machineNat. Commun 10
- 17.Sec61p is adjacent to nascent type I and type II signal-anchor proteins during their membrane insertionJ. Cell Biol 121:743–50
- 18.Protein translocation into proteoliposomes reconstituted from purified components of the endoplasmic reticulum membraneCell 75:615–30
- 19.Signal sequence recognition in cotranslational translocation by protein components of the endoplasmic reticulum membraneJ. Cell Biol 142:355–64
- 20.The Sec61p complex mediates the integration of a membrane protein by allowing lipid partitioning of the transmembrane domainCell 102:233–44
- 21.Mechanisms of integral membrane protein insertion and foldingJ. Mol. Biol 427:999–1022
- 22.Identification of Oxa1 Homologs Operating in the Eukaryotic Endoplasmic ReticulumCell Rep 21:3708–3716
- 23.EMC Is Required to Initiate Accurate Membrane Protein TopogenesisCell 175:1507–1519
- 24.Mechanism of signal-anchor triage during early steps of membrane protein insertionMol. Cell 83:961–973
- 25.Structure of the human signal peptidase complex reveals the determinants for signal peptide cleavageMol. Cell 81:3934–3948
- 26.An ER translocon for multi-pass membrane protein biogenesisElife 9
- 27.An intramembrane chaperone complex facilitates membrane protein biogenesisNature 584:630–634
- 28.The ER membrane protein complex interacts cotranslationally to enable biogenesis of multipass membrane proteinsElife https://doi.org/10.7554/eLife.37018
- 29.Visualization of translation and protein biogenesis at the ER membraneNature 614:160–167
- 30.Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in cancer cellsNat. Genet 49:1779–1784
- 31.A genome-wide atlas of co-essential modules assigns function to uncharacterized genesNat. Genet 53:638–649
- 32.Agreement between two large pan-cancer CRISPR-Cas9 gene dependency data setsNat. Commun 10
- 33.An alternative pathway for membrane protein biogenesis at the endoplasmic reticulum. CommunBiol 4
- 34.A protein of the endoplasmic reticulum involved early in polypeptide translocationNature 357:47–52
- 35.Signal sequence-dependent function of the TRAM protein during early phases of protein transport across the endoplasmic reticulum membraneJ. Cell Biol 134:25–35
- 36.TRAM regulates the exposure of nascent secretory proteins to the cytosol during translocation into the endoplasmic reticulumCell 92:621–631
- 37.Substrate-specific function of the translocon-associated protein complex during translocation across the ER membraneJ. Cell Biol 160:529–39
- 38.Control of glycosylation of MHC class II-associated invariant chain by translocon-associated RAMP4EMBO J 18:4804–4815
- 39.Stress-associated endoplasmic reticulum protein 1 (SERP1)/ribosome-associated membrane protein 4 (RAMP4) stabilizes membrane proteins during stress and facilitates subsequent glycosylationJ. Cell Biol 147:1195–1204
- 40.Yeast gene which suppresses the defect in protein export of a secY mutant of E.coliJ. Biochem 109:799–802
- 41.Site-specific photocross-linking reveals that Sec61p and TRAM contact different regions of a membrane-inserted signal sequenceJ. Biol. Chem 268:26745–51
- 42.A posttargeting signal sequence recognition event in the endoplasmic reticulum membraneCell 82:261–70
- 43.Deletion of SERP1/RAMP4, a Component of the Endoplasmic Reticulum (ER) Translocation Sites, Leads to ER StressMol. Cell. Biol 26:4257–4267
- 44.Processing of Structurally Heterogeneous Cryo-EM Data in RELIONMethods in Enzymology Academic Press Inc :125–157
- 45.Focused classification and refinement in high-resolution cryo-EM structural analysis of ribosome complexesCurrent Opinion in Structural Biology 46:140–148
- 46.Highly accurate protein structure prediction with AlphaFoldNature 596:583–589
- 47.Protein complex prediction with AlphaFold-MultimerbioRxiv 2021.10.04.463034 https://doi.org/10.1101/2021.10.04.463034
- 48.ColabFold: making protein folding accessible to allNat. Methods 19:679–682
- 49.Charged residues next to transmembrane regions revisited: “Positive-inside rule” is complemented by the “negative inside depletion/outside enrichment rule.”BMC Biol 15
- 50.Comparative analysis of amino acid distributions in integral membrane proteins from 107 genomesProteins Struct. Funct. Genet 60:606–616
- 51.Cryo-EM structure of the ribosome-SecYE complex in the membrane environmentNat. Struct. Mol. Biol 18:614–621
- 52.Structures of the E. coli translating ribosome with SRP and its receptor and with the transloconNat. Commun 7
- 53.Structure of a prehandover mammalian ribosomal SRP•SRP receptor targeting complexScience (80-.) 360
- 54.Lateral gate dynamics of the bacterial translocon during cotranslational membrane protein insertionProc. Natl. Acad. Sci. U. S. A 118
- 55.Stepwise gating of the Sec61 protein-conducting channel by Sec63 and Sec62Nat. Struct. Mol. Biol 28:162–172
- 56.Competitive binding of the SecA ATPase and ribosomes to the SecYEG transloconJ. Biol. Chem 287:7885–7895
- 57.Structure of a complex of the ATPase SecA and the protein-translocation channelNature 455:936–943
- 58.A unified evolutionary origin for the ubiquitous protein transporters SecY and YidCBMC Biol 19
- 59.Ipomoeassin F Binds Sec61α to Inhibit Protein TranslocationJ. Am. Chem. Soc 141:8450–8461
- 60.Mycolactone reveals the substrate-driven complexity of Sec61-dependent transmembrane protein biogenesisJ. Cell Sci 130:1307–1320
- 61.Proteomics reveals scope of mycolactone-mediated Sec61 blockade and distinctive stress signatureMol. Cell. Proteomics mcp.R :A118–824https://doi.org/10.1074/mcp.RA118.000824
- 62.Secretory Protein Profiling Reveals TNF-α Inactivation by Selective and Promiscuous Sec61 ModulatorsChem. Biol 18:1082–1088
- 63.Apratoxin Kills Cells by Direct Blockade of the Sec61 Protein Translocation ChannelCell Chem. Biol 23:561–566
- 64.Coibamide A Targets Sec61 to Prevent Biogenesis of Secretory and Membrane ProteinsACS Chem. Biol 15:2125–2136
- 65.A common mechanism of Sec61 translocon inhibition by small moleculesNat. Chem. Biol 19:1063–1071
- 66.Signal peptide mimicry primes Sec61 for client-selective inhibitionNat. Chem. Biol 19:1054–1062
- 67.Mechanism of signal sequence handover from NAC to SRP on ribosomes during ER-protein targetingScience (80-.) 375:839–844
- 68.Structure of the Mammalian Ribosome-Sec61 Complex to 3.4 Å ResolutionCell 157:1632–1643
- 69.Structure of the native Sec61 protein-conducting channelNat. Commun 6
- 70.A robust normalized local filter to estimate compositional heterogeneity directly from cryo-EM mapsNat. Commun 14
- 71.Toward a structural understanding of co-translational protein translocationCurr. Opin. Cell Biol 41:91–99
- 72.The limits of variationJ. Mol. Biol 184:99–105
- 73.Systematic probing of the environment of a translocating secretory protein during translocation through the ER membraneEMBO J 13:3973–82
- 74.Signal sequence processing in rough microsomesJ. Biol. Chem 270:19873–19878
- 75.Requirements for signal peptide peptidase-catalyzed intramembrane proteolysisMol. Cell 10:735–744
- 76.A gating motif in the translocation channel sets the hydrophobicity threshold for signal sequence functionJ. Cell Biol 199:907–18
- 77.Structures of the scanning and engaged states of the mammalian srp-ribosome complexElife 4:1–21
- 78.Molecular mechanism of cargo recognition and handover by the mammalian signal recognition particleCell Rep 36
- 79.Structural inventory of cotranslational protein folding by the eukaryotic RAC complexNat. Struct. Mol. Biol 30:670–677
- 80.Ribosome–NatA architecture reveals that rRNA expansion segments coordinate N-terminal acetylationNat. Struct. Mol. Biol 26:35–39
- 81.The dynamic architecture of Map1-and NatB-ribosome complexes coordinates the sequential modifications of nascent polypeptide chainsPLoS Biol 21
- 82.mRNA decoding in human is kinetically and structurally distinct from bacteriaNature 617:200–207
- 83.Molecular basis of the TRAP complex function in ER protein biogenesisNat. Struct. Mol. Biol 30:770–777
- 84.Structural insights into TRAP association with ribosome-Sec61 complex and translocon inhibition by a CADA derivativeSci. Adv 9
- 85.Structural basis for clearing of ribosome collisions by the RQT complexNat. Commun 14
- 86.SSRα and associated calnexin are major calcium binding proteins of the endoplasmic reticulum membraneJ. Biol. Chem 266:19599–19610
- 87.Palmitoylated calnexin is a key component of the ribosome-translocon complexEMBO J 31:1823–1835
- 88.Chaperone selection during glycoprotein translocation into the endoplasmic reticulumScience 288:331–3
- 89.N-linked glycans direct the cotranslational folding pathway of Influenza hemagglutininMol. Cell 11:79–90
- 90.Fast and accurate protein structure search with FoldseekNat. Biotechnol https://doi.org/10.1038/s41587-023-01773-0
- 91.A new congenital disorder of glycosylation caused by a mutation in SSR4, the signal sequence receptor 4 protein of the TRAP complexHum. Mol. Genet 23:1602–1605
- 92.Mutations in the translocon-associated protein complex subunit SSR3 cause a novel congenital disorder of glycosylationJ. Inherit. Metab. Dis 42:993–997
- 93.The translocon-associated protein (TRAP) complex regulates quality control of N-linked glycosylation during ER stressSci. Adv 7
- 94.Nodal modulator (NOMO) is required to sustain endoplasmic reticulum morphologyJ. Biol. Chem 297
- 95.Mechanisms determining the morphology of the peripheral ERCell 143:774–788
- 96.Dissecting the molecular organization of the translocon-associated protein complexNat. Commun 8
- 97.YidC occupies the lateral gate of the SecYEG translocon and is sequentially displaced by a nascent membrane proteinJ. Biol. Chem 288:16295–16307
- 98.YidC and SecYEG form a heterotetrameric protein translocation channelSci. Rep 7
- 99.Inner membrane YfgM–PpiD heterodimer acts as a functional unit that associates with the SecY/E/G translocon and promotes protein translocationJ. Biol. Chem 298
- 100.Molecular view of ER membrane remodeling by the Sec61/TRAP transloconEMBO Rep 24
- 101.Cotranslational sorting and processing of newly synthesized proteins in eukaryotesTrends in Biochemical Sciences https://doi.org/10.1016/j.tibs.2023.10.003
- 102.In vitro reconstitution of translational arrest pathwaysMethods 137:20–36
- 103.In vitro dissection of protein translocation into the mammalian endoplasmic reticulumMethods Mol. Biol 619:339–63
- 104.Preparation of microsomal membranes for cotranslational protein translocationMethods Enzymol 96:84–93
- 105.MotionCor2: anisotropic correction of beam-induced motion for improved cryo-electron microscopyNat. Methods 14:331–332
- 106.CTFFIND4: Fast and accurate defocus estimation from electron micrographsJ. Struct. Biol 192:216–21
- 107.New tools for automated cryo-EM single-particle analysis in RELION-4.0Biochem. J 478:4169–4185
- 108.Positive-unlabeled convolutional neural networks for particle picking in cryo-electron micrographsNat. Methods 16:1153–1160
- 109.DeepEMhancer: a deep learning solution for cryo-EM volume post-processing. CommunBiol 4
- 110.Highly accurate protein structure prediction for the human proteomeNature 596:590–596
- 111.ColabFold - Making protein folding accessible to allbioRxiv 2021.08.15.456425 https://doi.org/10.1101/2021.08.15.456425
- 112.UniProt: the universal protein knowledgebaseNucleic Acids Res 46:2699–2699
- 113.ISOLDE: A physically realistic environment for model building into low-resolution electron-density mapsActa Crystallogr. Sect. D Struct. Biol 74:519–530
- 114.Adaptive Cartesian and torsional restraints for interactive model rebuildingActa Crystallogr. Sect. D Struct. Biol 77:438–446
- 115.Macromolecular structure determination using X-rays, neutrons and electrons: Recent developments in PhenixActa Crystallogr. Sect. D Struct. Biol 75:861–877
- 116.UCSF ChimeraX: Structure visualization for researchers, educators, and developersProtein Sci 30:70–82
Article and author information
Author information
Version history
- Preprint posted:
- Sent for peer review:
- Reviewed Preprint version 1:
- Reviewed Preprint version 2:
- Version of Record published:
Copyright
© 2024, Aaron J. O. Lewis & Ramanujan S. Hegde
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
- views
- 1,748
- downloads
- 189
- citations
- 4
Views, downloads and citations are aggregated across all versions of this paper published by eLife.