Structure and function of the bacterial Sec translocon (Review)

Abstract

Bacteria and archaea possess a protein complex in the plasma membrane that governs protein secretion and membrane protein insertion. Eukaryotes carry homologues in the endoplasmic reticulum (ER) where they direct the same reaction. A combination of experiments conducted on the systems found in all three domains of life has revealed a great deal about protein translocation. The channel provides a route for proteins to pass through the hydrophobic barrier of the membrane, assisted by various partner proteins which maintain an unfolded state of the substrate, target it to the channel and provide the energy and mechanical drive required for transport. In bacteria, the post-translational reaction utilizes an ATPase that couples the free energy of ATP binding and hydrolysis to move the substrate through the protein pore. This review will draw on genetic, biochemical and structural findings in an account of our current understanding of this mechanism.

• Protein translocation
• SecY complex
• structure & dynamics
• energy transduction

Introduction

Proteins required in extra-cytosolic locations rely on specific targeting and transport apparatus for their localization. This process directs protein secretion and membrane protein insertion, and is therefore essential for cellular biogenesis. There exists a ubiquitous membrane protein complex found in the plasma membrane of bacteria and archaea (SecY), and in the ER of eukaryotic cells (Sec61). Two modes of translocation (co- and post-translational) converge at the Sec complex, which is a versatile and dynamic structure capable of conducting large substrates through the membrane without the loss of small molecules.

Some of the proteins synthesized in the cytosol carry N-terminal signal sequences, which are recognized by factors required for targeting [1]. In the event of co-translational translocation, the ribosome-nascent chain complex is targeted to the membrane courtesy of the signal recognition particle (SRP) and its receptor [2], [3]. Subsequent reactions deliver the translating ribosome to the SecY/Sec61 complex and in this state the ribosomal polypeptide exit site is located in close vicinity to the Sec complex [4], enabling protein translocation to be driven by the concomitant chain elongation [5].

Post-translational translocation in eukaryotes and bacteria operate by different mechanisms. In eukaryotes this process relies on BiP, an Hsp70 homologue in the ER lumen which utilises ATP to effectively pull the polypeptide chain in a ratchet-like mechanism [6], [7]. Bacteria have adopted a mechanism that pushes proteins across the channel by employing a motor protein ATPase SecA [8], [9]. In addition, the reaction requires a cytosolic component SecB to aid targeting to the membrane and to prevent premature protein folding and aggregation, to maintain a translocation competent conformation of the pre-protein [8], [10–12].

Elements of the Sec pathway were identified through different genetic screens conducted with Escherichia coli. Conditional-lethal mutations associated with a generalized protein-secretion defect facilitated the identification of the genes secA, secB, secE and secY[13–15]. The second strategy selected for mutations restoring the periplasmic localization of proteins with secretion-defective signal sequences, where prlD, prlG and prlA were identified as allelic forms of secA, secE and secY, respectively [16–18]. A landmark advance illustrated the requirement for ATP in post-translational translocation [19] and subsequent analysis of the gene products identified three membrane protein components SecY, SecE and SecG of the protein channel, and SecA a soluble ATPase [9], [20–22]. Seminal experiments reconstituted the entire translocation reaction in vitro, when a Sec dependent substrate could be transported to the interior of vesicles containing SecYEG in the presence of SecA and ATP [9].

The SecY protein channel

Images recorded by electron cryo-microscopy provided the first pictures of the Sec complex and revealed an oligomeric assembly of an estimated 3–4 complexes with an area of low density formed at the interface [23]. This pore aligned with the polypeptide exit channel of the ribosome [24–26]. The first crystal structure determined to a significantly higher resolution was of the membrane bound E. coli SecYEG complex. Electron microscopy and image reconstruction revealed all 15 of the predicted trans-membrane (TM) α-helices, [27], [28]. The arrangement of the SecYEG protomers seen in the membrane bound state was that of a dimer in a so-called ‘back-to-back’ arrangement.

The breakthrough came with the solution of a high-resolution X-ray structure of a monomeric and closed archaeal SecY complex, resolved in detergent solution (Figure 1) [29]. The most surprising observation was a hydrophilic restriction point in the centre of the monomer reminiscent of a closed channel or pore. This narrow pore is held closed by a short reinserted loop (the ‘plug’) and a girdle of hydrophobic side chains. The more conserved residues in SecY line the channel, which is also the site of several prl mutations [29]. This putative channel is conveniently located next to the signal sequence binding pocket between TM2 & 7 [30] and has also been shown to form cross-links with translocating pre-protein [31]. The periphery of the monomeric complex is too hydrophobic to form part of a protein channel at the interface of an oligomeric complex.

Figure 1.  Ribbon and schematic representation of the detergent solubilized monomeric Methanococcus jannaschii SecYEβ complex viewed from the cytosolic face [29]. Transmembrane helices 1–5 are coloured in light green, helices 6–10 dark green, the ‘plug’ orange, SecE red and Secβ (SecG) is shown in sea green. The equivalent E. coli cysteine cross-link (L106C) is indicated in blue, close to the blue arrow. The green arrow denotes the position of the lateral gate.

The structure and oligomeric state of the active channel

The channel pore is formed between two distinct halves of the SecY protein (Figure 1). SecE has been proposed to form a clamp around these domains, in order to maintain the channel closed in the resting state [29]. When perturbed, SecE may relax this grip, enabling the channel to open. This would require the concomitant displacement of the ‘plug’ domain to a location at the periphery of the complex, in addition to a widening of the central pore to accommodate the substrate protein. These conformational changes are presumably directed by the association of the partner protein and substrate, and the subsequent forces brought to bear by protein synthesis or the hydrolytic cycle of SecA. The separation of the two halves of SecY, comparable to a ‘clam-shell’ also offers an escape route to the membrane. Therefore, the structure is compatible with the ability of the complex to facilitate the passage of proteins both through and into the membrane.

There are a few clues with respect to how the channel might open. The ‘plug’ has been shown to cross-link to the C-terminus of SecE, approximately 20 Å away from its position in the structure [32], [33], identifying a potential pocket which it can inhabit during protein translocation. The formation of this cross-link, as well as the deletion of the plug resulted in enhanced protein translocation [32], indicating that the complex may be held in a partially open state. The removal of the plug domain was however not lethal [34] indicating that this domain is not the only determinant for channel regulation. The Hsp70 homologue BiP involved in the eukaryotic post-translational translocation reaction is thought to contribute to the seal formed by the channel [35]. However, prokaryotes lack this component, indicating that other mechanisms can be adopted to provide a barrier to small molecules. A careful comparison of the membrane bound dimer [28] and the detergent solubilized monomer [29], indicates that in the former the plug has moved about 6 Å towards the outside, and the lateral gate for membrane protein insertion has opened slightly (Figure 2) [36]. In this context, although closed, the dimer of SecYEG appears to be primed for protein translocation.

Figure 2.  (A) Homology model of E. coli SecYEG based on the detergent solubilized SecYEβ monomer. (B) Model fitted according to the membrane bound dimer [36]. Both are side views with the cytosolic face uppermost; the latter shows only the membrane sector, as the loops were not resolved sufficiently. TM2b (olive) and TM7 (green) form the signal sequence-binding site and have parted slightly in the dimer. The ‘plug’ (red) is obscured in (A) and is about 6.5 Å higher. For purposes of clarity schematic representations have been drawn on the right hand sides.

One puzzling aspect of the structure is the fact that the monomer appears to provide all of the components for translocation: the signal sequence binding site, the protein-pore, exposed cytosolic loops for partner protein interaction and a lateral and traversal exit site. Several studies from independent laboratories have observed the bacterial channel in a dimeric form [28], [33], [37–40]. The reasons for the presence of high oligomeric states are not entirely clear. The complex may exist as an oligomer with multiple independent active sites as many other soluble and membrane proteins do. Alternatively, the larger assemblies might be required to provide a large enough platform for the association of the more sizable partners. Specific inter-subunit interactions might also be required for activation or to support a dynamic process required for the reaction. Finally, the mechanism might rely on the partner proteins or substrate to introduce an asymmetric element important for the reaction, to generate a singly active protomer. A recent study has shown that protein translocation occurs through only one copy of the two that are present in the active complex (and not through a consolidated channel) [Osborne & Rapoport (2007) Cell, 129, 97–110].

The ‘back-to-back’ orientation observed of the membrane bound form has also been detected biochemically. In this particular form TM3 of SecE is located at the interface and the lateral gate for membrane protein insertion (between TM2 & TM7 of SecY; Figure 3) point in opposite directions and toward the lipid bilayer [28], [29], [36]. Cysteine mutagenesis experiments identified residues in SecE that were close to the equivalent helix in the neighbouring monomer [41], [42]. These observations were enhanced by conditions that promote a productive association of the partner protein SecA [41]. Moreover, the cross-linked dimer retained the capacity to tightly bind and activate SecA, but had lost the ability to couple ATP hydrolysis to the work of translocation [41]. A careful inspection of the position of the respective side chains reveal that, although close, they point away from each another (Figure 3). The association of SecA may have perturbed them and brought them closer together, and thus more amenable to cross-linking. The formation of a disulphide in this position would bring about a considerable distortion of the two helices. It is conceivable that this imposed distortion and restriction on the motility at the dimer interface may have had a minor affect on the association with SecA, but a disastrous one for the energy coupling process.

Figure 3.  Space filling representation of the dimeric membrane-bound E. coli atomic model [36]. SecY and SecE are coloured green and red, respectively. The green arrows denote the position of the lateral gate. The E. coli cysteine cross-link (L106C) is shown in blue. The distance between the two cysteines is 18.5 Å, contrasting to the 2.05Å disulphide bond length. (A) View from the periplasmic face of the membrane. (B) View from the side of the membrane. Simplified views have been drawn underneath; the open blue circle (B) denotes that the cysteine is on the other side of TM3 of SecE.

Another study by electron cryo-microscopy has revealed two SecYEG dimers with a ribosome-nascent chain complex (one bound close to the exit channel and the other to mRNA) [39]. The map was of insufficient detail to resolve clear and individual TM domains, but could be used to fit the high-resolution structure into the density. On the basis of normal mode-based flexible fitting the ‘back-to-back’ SecYEG structure was rejected in favour of a new ‘front-to-front’ model. This arrangement places the same TM domain of SecE on the opposite side of the membrane complex, and the two lateral gates close to each other at the dimer interface. This proposal has been used as a foundation for a model of translocation incorporating both monomers of SecYEG in a single translocation cycle; in this context it has been suggested that the channels open at the dimer interface to form a consolidated environment for protein transport [39], [43], [44]. However, disulfide cross-linking shows that the contacts the translocating pre-protein makes with SecY are within the strict confides of the channel, and not, as might be expected from such an arrangement, in the teeth of the ‘clam-shell’ [31].

The structure and domain organization of SecA

Like other molecular motors SecA must undergo large ATP-dependent conformational changes and couple them to the domain movements required for function. The key step for energy transduction in SecA must be the conformational change regulated by ATP binding and product release in the nucleotide-binding fold (NBF). It is how these movements are relayed through the enzyme and to their partners, which provides the power stroke, and holds the key to understanding the reaction mechanism.

SecA is a soluble protein of 102 kDa and exists in solution in a monomer-dimer equilibrium, predominantly dimeric [45], [46]. Of the six published SecA crystal structures [47–52], five of them are dimeric and one monomeric (Figure 4). The protomers of each dimer all have a similar structure, the major differences between them being at the dimer interface. Interestingly, the structure from Thermus thermophilus is in a parallel conformation [47], in contrast to all of the other structures which are packed in an anti-parallel manner [48], [50–52]. It is not clear which of these structures is the correct physiological state, as it is possible that the oligomeric arrangement observed in X-ray structures is a consequence of the extreme crystallization conditions or the lattice contacts. The parallel and two anti-parallel forms have however been shown to exist in solution by directed cysteine mutagenesis and cross-linking [47], [50] suggesting that in the absence of other translocation partners and pre-protein there may be different conformational states of the enzyme.

Figure 4.  Ribbon representation of the monomeric SecA structure [49]. NBF1 is shown in dark green, NBF2 pale blue, HSD purple, HWD bright green, PPXD in dark blue and Mg^2 +-ADP in orange/red. The arrow denotes the position of the putative polypeptide-binding site.

Within each protomer there are five major domains, from N- to C-terminus these are: nucleotide binding fold 1 (NBF1), pre-protein cross-linking domain (PPXD) [53], [54], nucleotide binding fold 2 (NBF2), helical scaffold domain (HSD) and helical wing domain (HWD) [51]. In the monomeric crystal, movements of the HWD, HSD and PPXD domains results in opening of a groove that may be the polypeptide binding site [49], [55] (Figure 4). In all of the apo- and nucleotide-bound structures, there are essentially no major changes in the structure of the NBFs [48], [51], [56]. This means that there are missing active states of the complex. The constraints of the crystallization process or the transient nature of these states may mean that they are difficult to characterise structurally.

Interaction sites and stoichiometry of SecA and SecYEG

SecA has been shown to associate with membranes containing SecYEG to a high affinity [8]. The structure of the SecYEG complex indicates that there are loops exposed to the cytosol available for partner protein interaction [28], [29], [36], but the precise locations are ill defined. Both the N-terminal domain (comprising the NBF) and C-terminal domains of SecA may be important for the interaction with SecYEG [57–59]. Second site suppressors, cross-linking and peptide scanning experiments map the interaction sites in SecY to TM4, the fourth and fifth cytoplasmic loops (C4 and C5) and the C-terminal tail [60–63]. There is evidence to support the fact that the dimer of SecA is an active form. The inactivation of a single SecA protomer has a dominant effect on the activity of dimer and resonance energy transfer experiments designed to monitor the appearance of monomers proved negative [64]. However, recent reports have challenged this view. The fact that one of the structures of SecA is a monomer led to suggestion that the dimer might dissociate when activated [49], [50]. The observed loss of specific inter-molecular cross-links and fluorescence resonance energy transfer between dimeric SecA on exposure to translocation specific ligands is an indication of monomerization [49], [65], [66]. These observations may also be the result of a rearrangement to form one of the other dimeric structures, or a hitherto uncharacterized and active dimeric form of the protein. During translocation, a SecA monomer remained associated with a SecYEG protomer and the pre-protein [67]. Two further studies identified that both monomers and dimers of SecA were able to associate with SecYEG dimers in solution, and that the specific oligomeric association was dependent on nucleotide [37], [67]. Perhaps both monomers and dimers of SecA play important roles in the translocation reaction, and association and dissociation are required to complete the reaction cycle.

The hydrolytic cycle of SecA

The ATPase activity of SecA is sensitive to ligands, including: mature proteins [68], signal peptides [68–71], SecB [72], [73], ADP [46], magnesium [46], lipids [8], [46], [65], [66], [68], [74–76] and SecYEG [8], [68]. Two specific regulatory domains for ATPase activity have been proposed. The first Intramolecular Regulator of ATPase activity (IRA1) resides at the far C-terminus and controls ATP hydrolysis in both soluble and SecY-associated SecA [77]. The second, IRA2 (NBF2) controls the nucleotide binding and release occurring at NBF1 [78].

The ATP bound state of SecA has been shown to be in an extended conformation [79], whereas the ADP bound form is more compact [51], [77–81], the release of which is thought to be the rate limiting step of the steady-state reaction cycle [80]. The non-hydrolysable ATP analogue AMP-PNP stabilises the association of SecA and SecYEG and promotes the ‘inserted’ state of a substrate protein [37], [82]. Experiments employing surface plasmon resonance and cross-linking show that ATP induces the release of SecA from SecYEG [67], [83].

The divalent cation magnesium exerts a strong inhibitory effect upon SecA, affecting both the dimeric conformation (but not oligomeric state) and the affinity for ATP; it does this by acting on an allosteric binding site distinct from the catalytic one [46]. This inhibition could be overcome by the acidic phospholipid cardiolipin, which had a stimulatory role only in the presence of magnesium. Lipid binding sites have been found on SecA [84], [85], and acidic phospholipids are required for translocation [86]. This suggests that the ‘lipid-ATPase’ activity reported may in fact be related to a counteraction of magnesium inhibition by cardiolipin-SecA interactions [46]. This inhibition presumably prevents the futile hydrolysis of ATP during moments of inactivity, and is released as a result of intimate exposure to the membrane.

The catalytic cycle has been proposed to proceed by a multi-step mechanism for translocation via SecA insertion and de-insertion events [82]. The evidence is based upon the protease protection of a large domain of SecA while bound by SecY, which is dependent on ATP and pre-protein [82], [87]. This protection is the result of a necessary and uncharacterized large conformational change of the peripherally associated protein and not necessarily the result of a deep penetration of the enzyme. The thermodynamics of this reaction have also not been well characterized; one study proposes that one ATP drives the passage of 40 amino acids, and another determines that each amino acid requires 5 molecules of ATP [88], [89].

Concluding comments

This is an extremely complicated field of membrane biology and our comprehension of it is thanks to a generation of research on the related eukaryotic, bacterial and archaeal systems. In spite of the recent progress there are several disputed and unanswered aspects of the reaction mechanism. These outstanding problems will need to be addressed by detailed structural determination of an active state of the translocation complex with pre-protein, and by further biophysical studies aimed to comprehend the dynamics of the reaction. Understanding details both on the nature and timing of the conformational changes that derive from the release of chemical energy, and of specific protein-protein interactions holds the key to understanding how they are coupled to protein transport.