|Home | About | Journals | Submit | Contact Us | Français|
Secreted and membrane proteins are translocated across/into cell membranes via a protein-conducting channel (PCC). We present a cryo-EM reconstruction of the E. coli PCC, SecYEG, complexed with the ribosome and a signal anchor containing nascent chain, showing mRNA, three tRNAs, the nascent chain, and detailed features of both a translocating PCC and a second, non-translocating PCC bound to mRNA hairpins. The translocating PCC forms connections with ribosomal RNA hairpins on two sides and ribosomal proteins at the back, leaving a frontal opening. Normal mode-based flexible fitting of the archaeal SecYEβ structure into the PCC EM densities favors a front-to-front arrangement of two SecYEG complexes in the PCC, and supports channel formation by the opening of two linked SecY halves during polypeptide translocation. Based on our observation in the translocating PCC of two segregated pores with different degrees of access to bulk lipid, we propose a model for co-translational protein translocation.
The proper functioning of every living cell requires that secreted, soluble proteins be translocated across, and membrane proteins be inserted into, a cell membrane. Protein translocation at the membrane occurs through a proteinaceous channel 1, the translocon 2, at the core of which lies the protein-conducting channel (PCC). The PCC is a heterotrimeric integral membrane protein complex 3, 4 composed of the α-subunit (SecY in eubacteria/archaea, Sec61α in mammals), the β-subunit (SecG in eubacteria, Secβ in archaea, Sec61β in mammals) and the γ-subunit (SecE in eubacteria/archaea, Sec61γ in mammals). Most membrane proteins and some soluble proteins are translocated co-translationally, a process in which the ribosome binds to the PCC, and nascent chain translocation is concomitant with polypeptide elongation on the ribosome. The PCC is a dynamic complex, which must be able to open and close an aqueous channel 5, 6 relatively isolated from hydrophobic lipids and running perpendicularly to the membrane plane to allow hydrophilic regions of a polypeptide across. The PCC must also be able to open and close laterally 7 to regulate the lipid-mediated partitioning of hydrophobic transmembrane helices (TMHs) into the plane of the lipid bilayer 8.
Low-resolution cryo-electron microscopy (cryo-EM) studies of eukaryotic co-translational translocation complexes 9–12 have demonstrated that the functional PCC is an oligomer of the Sec61αβγ complex. The recent x-ray structure of a non-translocating, monomeric, archaeal SecYEβ heterotrimer 13 – uncomplexed with a ribosome or substrate polypeptide – shows that the α-subunit, SecY, is divided into two independent N-terminal (TMHs 1-5) and C-terminal (TMHs 6-10) halves, forming a ‘clam shell’ 14, which on one side is hinged by the loop between TMHs 5/6 and clamped together by the γ-subunit, SecE, while the other side – forming the lateral gate – is unconstrained. Each SecY half contributes to the formation of a transmembrane funnel-like cavity in the center of the complex, which is blocked by a plug (TMH 2a). It is hypothesized that the signal sequence (SS)/TMH signal anchor (SA) of a translocating nascent chain displaces the plug and wedges itself into the lateral gate, thus opening the two SecY halves in the plane of the membrane, enabling either translocation across of a hydrophilic region of nascent chain or the lateral partitioning of a hydrophobic TMH into the lipid bilayer 13, 14. Additionally, the x-ray structure shows that the long cytoplasmic loops between TMHs 6/7 and 8/9 extend ~20 Å above the membrane plane for interaction with cytosolic factors, such as the large ribosomal subunit 15–17.
We have used a combination of single-particle cryo-EM and computational methods to determine and interpret the structure of a functional, co-translational translocation complex from E. coli, comprising SecYEG and a ribosome-nascent chain (RNC) complex. Our cryo-EM reconstruction is of sufficient resolution to determine the precise number of SecYEG monomers in the PCC and to differentiate, using computational methods, between two principal models of SecYEG monomer arrangements, helping to clarify the structural and mechanistic details of co-translational translocation.
E. coli ribosomes programmed in a cell-free translation system with mRNA encoding for a chimeric nascent chain (Suppl. Fig. 1), which includes the TMH SA of FtsQ and the SecM stalling sequence 18, were complexed with detergent-solubilized SecYEG. These reconstituted RNC-SecYEG complexes were used for structure determination by cryo-EM using single-particle reconstruction. EM data (Fourier shell correlation characteristics ~11 Å at 3σ, ~15 Å at 0.5) extended out to ~1/10 Å−1 in Fourier space, i.e. structural features down to 10 Å in size could be discerned (see Suppl. Fig. 4). In this reconstruction, tRNAs are visible in all three canonical sites, A, P, and E (Fig. 1a). The path of over 100 nucleotides of the mRNA from its entrance to past its exit site in the 30S subunit can be traced, as can most of the path of the nascent chain in the polypeptide exit tunnel. Greatly improving on the globular appearance of the PCC in previous single-particle cryo-EM studies 9–12, detailed rod- and lamella-like features, corresponding to groupings of transmembrane helices in the PCC, are discernible in the current reconstruction (Fig. 1a,b,c). As observed previously there is a large frontal opening between the ribosome and the PCC at the polypeptide exit site, through which the translocating polypeptide is accessible (Fig. 1b, and see Suppl. Discussion 2.1 for a discussion on the discrepancy between existing fluorescence and structural data).
Quite surprising is the observation of a second PCC bound at the exit of the mRNA channel (occupancy ~70%). When viewed perpendicularly to the membrane plane (Fig. 1c,e), the shapes of the two PCCs are different: the PCC at the mRNA channel exit is an elliptic cylinder (110 Å × 70 Å, height ~45 Å) with pseudo-222 symmetry (Suppl. Fig. 2); whereas the PCC at the polypeptide exit site comprises a circular cylinder (~95 Å diameter, height ~45 Å). Since the latter PCC contains the TMH SA of FtsQ (see Methods), it represents a translocating state. The PCC at the mRNA channel exit is non-physiologically bound and so must represent a non-translocating state. Volume calculations on the isolated densities of both PCCs indicate that each PCC contains two copies of SecYEG, as also suggested biochemically 19, 20 (see Suppl. Results 1.1, Suppl. Discussion 2.2 and Suppl. Fig. 5 for a discussion on the effect of low resolution on PCC EM density).
Two copies of a homology-modeled atomic model of E. coli SecYEG (SecYEGEc) (see Methods) were docked rigidly into the isolated densities of both the non-translocating and translocating PCC. While docking into the translocating PCC density was unsuccessful, a good fit (cross-correlation coefficient (CC) 0.78) to the non-translocating PCC density was achieved when two SecYEGEc heterotrimers were placed in a front-to-front arrangement. Thus, the structure of the PCC at the mRNA channel exit may represent a physiologically relevant non-translocating state, providing a starting structure for determining the proper conformation of the translocating PCC at the polypeptide exit site. Since relative motions of the N- and C-terminal halves of SecY are postulated to underlie SecYEG function in a translocating PCC 13, 14, we opted to utilize a computational method, normal mode-based flexible fitting (NMFF) 21. Normal modes provide information on the natural vibrations of a molecule and the preferential directions of collective, many-atom displacements, while maintaining the steric and geometric constraints of the overall structure. Normal mode analysis (NMA) 22 applied to plug-less SecYEM j (see Methods) confirmed the expected opening motion of linked SecY halves 13 (Suppl. Fig. 3). We thus proceeded to use a progression of docking and NMFF steps (see Suppl. Results 1.2 and Suppl. Fig. 4) to obtain a model for the translocating PCC, both in a front-to-front and the recently proposed back-to-back 13 arrangement.
The front-to-front model fit well (CCfinal 0.79) into the non-translocating PCC density, while the back-to-back model fit poorly (CCfinal 0.64). The bulk of each dimeric PCC model could be docked into the translocating PCC density (CCinitial 0.65 (front-to-front), 0.61 (back-to-back)) only when the ribosome-binding cytoplasmic loops between TMHs 6/7 and 8/9 (combined cross-section of Cα backbone ~20 Å) of both SecYEG heterotrimers were docked into the connection densities to the ribosome located on the sides (C1,C2), which have a roughly circular cross-section of ~20 Å (back connection (C3): cross-sectional dimensions ~45×20 Å) (Fig. 2b,c,f,g). Upon application of NMFF to these models, the two SecY halves in each heterotrimer in the front-to-front model (CCfinal 0.79) display significant opening motions within the membrane plane (Fig. 2a,d,i,j), (distance between the centers of mass of linked SecY halves increases by an average of ~5 Å), while no opening motion is seen in the back-to-back model (CCfinal 0.65). Thus, the formation of a pore wide enough to enable polypeptide translocation, a process hypothesized to be necessary for translocation 13, 14, is not observed as in the front-to-front case (Fig. 2a,d,e,h, Suppl. Fig. 4c,d). SecYEG has been shown to be inactivated when SecE is disulfide-crosslinked to a neighboring SecE or SecY molecule, indicating that SecE dynamics, which are restricted when SecE clamps are juxtaposed tightly as in the back-to-back arrangement, are essential for protein translocation 23, 24.
Our experimental and computational data thus provide two major arguments – one structural, the other functional – for why the back-to-back model is less likely. Furthermore, key biochemical observations cannot be explained on the basis of a back-to-back arrangement SecYEG heterotrimers, in which the lateral gates in each heterotrimer face away from each other, towards the bulk lipid in the membrane. During post-translational translocation – a process in which a fully translated preprotein is translocated through the PCC in the absence of a ribosome – disulfide-bonded, partially folded regions of preprotein can cross through the PCC channel 25. Forming an aqueous pore large enough for this process 25, 26 would require the lateral gate in a heterotrimer to open up beyond the point where the wedged SS could sterically block lipid diffusion into the pore, making the translocation of a hydrophilic polypeptide energetically very unfavorable. Similarly, at a certain point during co-translational translocation the SS/SA is no longer associated with SecYEG/Sec61αβγ 27, leaving the lateral gate permeable to lipids and making the continuing translocation of a hydrophilic nascent chain energetically very unfavorable. The front-to-front model, however, is consistent with these biochemical observations and certain observed cross-links in SecY 20, since the lateral gates of each heterotrimer face each other, and thus a consolidated channel – segregated from lipid – could potentially form.
In order to determine the details of interaction between the PCC and ribosomal elements, an E. coli ribosome atomic model was refined into our cryo-EM reconstruction (see Methods). As previously observed 11, 12 both ribosomal proteins and RNA surround the polypeptide exit site at which the PCC is located. Ribosomal proteins L17, L32 and L22 are located in the front above – and not interacting with – the PCC (Fig. 3a); whereas proteins L24, L29, and L23 interact with the PCC in the back (Fig. 3b). The PCC makes three connections with the ribosome (Fig. 2b,c and Fig. 3c). The connections at the sides, C1 and C2, are interactions between ribosomal RNA (rRNA) hairpins (helix 24 in C1 and helix 59 in C2) and the cytosolic factor-associating domains (CFAD) of SecY, the cytoplasmic regions between TMHs 6/7 and 8/9 (Fig. 3a,c), as has been suggested biochemically 15–17. The wall of connection at the back (C3) appears to be mediated mostly through interactions of L29 and L23 with the cytoplasmic region of the SecG subunit and possibly of the N-terminal part of SecE (Fig. 3b,c). Neither the SecG/β/61β subunit nor the N-terminal portion of SecE are essential for viability 28, 29. However, SecG/β/61β has been observed to stabilize the ribosome-PCC complex and facilitate nascent chain translocation 28, 30. It is thus likely that connection C3 lends stability and directionality (to be discussed later) to the PCC-ribosome association, but that connections C1 and C2 are required for PCC attachment to the ribosome 16.
Modeling of 121 nucleotides of RNA into the mRNA density isolated from our reconstruction using sequence-specific secondary structure information indicated the presence of two major RNA hairpins at the channel exit, immediately above the CFAD of both heterotrimers of the non-translocating PCC (Fig. 3d). Biochemical competition experiments have demonstrated that mRNA can compete with rRNA for PCC binding 15. This inferred interaction of the CFADs with immobilized, structured mRNA hairpins is analogous to the interaction observed at the polypeptide exit site with rRNA hairpins and may explain the occurrence of the non-translocating PCC at the mRNA channel exit. It is possible that the structure of this PCC may reflect that of a non-translocating PCC when it is bound to the ribosome at the polypeptide exit site.
Upon NMFF of the dimeric plug-less SecYEG complex into the translocating PCC density, prominent regions of density remain unaccounted for (green and yellow asterisks, Fig. 4a), most notably a long rod-like region (green asterisk) at the front interface of the two heterotrimers. This density most probably corresponds to the nascent chain TMH SA, close to SecY TMH 2 and 7, as suggested by cross-linking experiments 31. Furthermore, the observation that the interface formed by the two heterotrimers at the front changes significantly from the non-translocating to the translocating state may reflect SA binding at this site (Fig. 2i,j). The other empty regions of density (yellow asterisks) in the middle of both heterotrimers could represent either the plug (TMH 2a) or the remaining, hydrophilic region of the nascent chain. An atomic model was fit into the isolated thread of density corresponding to the nascent chain inside the ribosome, which facilitated the identification of putative interactions with ribosomal RNA (Mitra et al., manuscript in preparation) and protein. The real-space refined positions of ribosomal protein domains suggest an interaction of the nascent chain with finger-like projections of proteins L4 and L22 in the polypeptide exit tunnel, as has been demonstrated biochemically 18, 32. Towards the end of the polypeptide exit tunnel the projection of L23 is in proximity to the nascent chain, indicating another potential interaction site (Fig. 4b,d).
The cryo-EM reconstruction of the E. coli RNC-SecYEG presented in this work offers glimpses into the architecture of the ribosome-bound PCC in its translocating and possibly also non-translocating state. Analysis of atomic models obtained by NMFF suggests that the PCC is most likely formed by a front-to-front association of two SecYEG/Sec61αβγ heterotrimers in both states. Our data suggest that during the translocation of a hydrophilic region of the nascent chain, neither of the two heterotrimers opens to form one central, consolidated channel, but both retain two segregated pores. The translocating PCC density displays an asymmetry in the distribution of features, with separated rod-like features dominating at the front and contiguous lamella-like features at the back, possibly reflecting greater lipid accessibility at the front (Fig. 1c). Analysis of the atomic model of the translocating PCC supports this hypothesis. The ‘hook’-shaped C-terminal SecY half in Sec1YEG forces a path to the front, to the bulk lipid, whereas in Sec2YEG the ‘hook’ dictates a path to the back, which is blocked to lipid by a contiguous wall formed by the TMHs of SecG and SecE (Fig. 2i,j and Fig. 3a,c). An SS/TMH exiting laterally from the Sec1YEG/Sec161αβγ pore would interact with SecY/61α TMH 2 and 7 31 at the front interface between both heterotrimers, from where a TMH might then interact with YidC/TRAM 33, 34 and/or diffuse into the lipid bilayer (Fig. 4a,c,d).
Our NMA results support the initial role of the plug to retain tightly juxtaposed SecY halves in the non-translocating state until the plug is displaced by the nascent chain, at which point the SecY halves can open laterally 13 (see methods and Suppl. Results 1.2). We suggest that the partially hydrophobic nature of the plug drives its equilibrium state towards burial between linked SecY halves when not sterically obstructed by the presence of a nascent chain, ensuring the maintenance of the permeability barrier even if linked SecY halves are not tightly juxtaposed. Thus, while the nascent chain is being looped through the pore in one heterotrimer, the pore in the other heterotrimer may remain ‘plugged’. The frontal alignment of the ribosome/PCC opening with the partition site of nascent chain TMHs at the PCC/membrane may ensure reproducible, efficient folding and release of nascent membrane proteins by providing a dedicated site for the association with the nascent chain and the ribosome-PCC complex of cytosolic proteins, such as chaperones and SecA 35–37, through the frontal opening (~20×40 Å, and see Suppl. Discussion 2.1) and membrane proteins at the front PCC/membrane interface. The observation of possible interactions of the projections of ribosomal proteins L4, L22, and L23 with the nascent chain is intriguing and may suggest some form of nascent chain-dependent communication between the ribosome and the PCC.
E. coli ribosomes were programmed in a cell-free translation system with mRNA encoding for a chimeric nascent chain (Suppl. Fig. 1), consisting of an N-terminal Strep-tag followed by the first 74 residues of E. coli FtsQ, which include the TMH signal anchor, and residues 132–170 of E. coli SecM, which include the 17 amino acid SecM stalling sequence 18. A nascent chain consisting of FtsQ residues 1–108 has previously been cross-linked to SecY during cotranslational translocation 35 and the SecM stalling sequence has been shown to interact tightly with the ribosomal polypeptide exit tunnel while efficiently arresting ribosome elongation 18. Thus, stable RNCs were generated with nascent chains of a uniform length and sequence without using truncated mRNA. After in vitro translation, the RNCs were affinity-purified via the N-terminal Strep-tag and concentrated by ribosomal pelleting (Suppl. Fig. 1b,c,d). SecYEG was overexpressed in E. coli, detergent-solubilized, and purified to homogeneity by standard techniques. The RNC-SecYEG complex was reconstituted and analyzed (Suppl. Fig. 1e). The binding affinity for detergent-solubilized SecYEG was estimated to be ten-fold higher for a ribosome containing a nascent chain compared to an empty ribosome. The presence of the nascent chain thus proved to be essential, since in contrast to eukaryotes, empty E. coli ribosomes do not have sufficient affinity (apparent kD ~2 μM) for the detergent-solubilized PCC to form ribosome-SecYEG complexes under the conditions used for cryo-EM (32 nM). The sample used for cryo-EM contained a ten-fold molar excess of SecYEG over RNCs.
Samples were applied to carbon-coated holey grids as published 38. Micrographs were recorded under low-dose conditions on a Tecnai F30 field-emission gun electron microscope at 300 kV using a defocus range of 1.5 to 4.3 μm, and scanned on a Zeiss/Imaging Scanner (Z/I Imaging Corporation, Huntsville, AL), corresponding to a pixel size of 3.59 Å on the object scale. The data were analyzed using the SPIDER software package. After automated particle picking followed by visual inspection, supervised classification 39 was performed on the starting set of 104,257 picked particles. 53,325 particles corresponding to a translocating PCC bound to the RNC (PCC occupancy 100%) were used for the final, CTF-corrected reconstruction. The falloff of Fourier amplitudes toward higher spatial frequencies was corrected using the x-ray solution scattering intensity distribution of 70S ribosomes from E. coli 40 during each round of refinement. Refinement of an E. coli ribosome atomic model into the cryo-EM reconstruction was performed using the program RSRef 41, a real-space refinement module for the TNT program 42. An almost complete atomic E. coli ribosome model obtained by RSRef was adapted from Gao et al. 43 and complemented with the addition of RNA and protein models from the x-ray structure of the D. radiodurans large ribosomal subunit 44.
Normal mode analysis (NMA) with an elastic network 45 was performed on the x-ray structure of the SecYEβ heterotrimer from Methanococcus jannaschii 13. Residues corresponding to the plug (SecY:44-68, M.j. (40-75, E.c.)) and the Secβ subunit were removed for observation of low-frequency, high-amplitude normal modes corresponding to the opening of SecY halves and the motions of loops (Suppl. Fig. 2). Secβ, which lies at the periphery of the SecYEβM j structure and makes few interchain contacts, contributes to high-frequency normal modes and masks the low-frequency normal modes corresponding to inter-domain movement. This chain therefore needed to be removed. For normal mode-based flexible fitting (NMFF) a model of E. coli SecYEG (SecYEGEc) was generated by homology-modeling (similarity/identity between SecY of M. jannaschii and E. coli (40%/20%) and mouse (34%/55%)) based on SecYEβM j. The additional helices of E. coli were modeled according to the placement of helices by van den Berg et al. 13 into the 2-D electron crystal structure of E. coli SecYEG determined by Breyton et al. 46. Only the ten lowest-frequency normal modes were utilized to direct flexible fitting of atomic models into the PCC EM density. In order to observe normal modes corresponding to the opening of SecY halves, models of plug-less SecYEEc were used in which the two linked SecY halves in each complex were separated by 1 Å on the hingeless side. Only Cα coordinates were used for NMFF and NMA, since density for sidechains is not well resolved in the current moderate-resolution EM map. The models obtained upon NMFF were energy-minimized using CNS version 1.1 47. The reliability of the PCC models generated by NMFF extends to the orientation of domains and the general arrangement of helices within these domains, but not to the exact positioning of individual helices/residues or the conformation of loops.
Grid preparation, cryo-electron microscopy, data processing, atomic model generation, fitting and refinement and interpretation were done by K.M. (laboratory of J.F.). E. coli SecYEG-RNC complex was prepared and PCC EM density symmetry analysis performed by C.S. and S.J. (laboratory of N.B.). T.S. (J.F.) assisted in cryo-EM data processing. F.T. (laboratory of C.L.B.III) performed NMA and NMFF.
Coordinates for the translocating and non-translocating PCC have been deposited in the RCSB, with accession codes 2AKI and 2AKH, respectively. The cryo-EM map of the E. coli RNC-SecYEG complex has been deposited in the EBI Macromolecular Structure Database, accession code EMD-1143.
We thank R.A. Grassucci for training on the Tecnai F30 electron microscope; G.S. Allen for assistance with supervised classification, O, and discussion of the manuscript; and M. Watters for assistance with the illustrations. This work was supported by HHMI, NSF DBI 9871347 and NIH grants R37 GM29169 and R01 GM 55440 (to J.F.), Multiscale Modeling Tools for Structural Biology (MMTSB) funded by NIH RR12255 (to C.L.B. III), the NCCR Structural Biology program of the Swiss National Science Foundation (SNSF; to N.B.) and a Young Investigator grant from the Human Frontier Science Program (to N.B.). C.S. was supported by post-doctoral fellowships of the Roche Research Foundation and the Ernst Schering Research Foundation.