Targeting and Insertion of Membrane Proteins

Abstract

The insertion and assembly of proteins into the inner membrane of bacteria are crucial for many cellular processes, including cellular respiration, signal transduction, and ion and pH homeostasis. This process requires efficient membrane targeting and insertion of proteins into the lipid bilayer in their correct orientation and proper conformation. Playing center stage in these events are the targeting components, signal recognition particle (SRP) and the SRP receptor FtsY, as well as the insertion components, the Sec translocon and the YidC insertase. Here, we will discuss new insights provided from the recent high-resolution structures of these proteins. In addition, we will review the mechanism by which a variety of proteins with different topologies are inserted into the inner membrane of Gram-negative bacteria. Finally, we report on the energetics of this process and provide information on how membrane insertion occurs in Gram-positive bacteria and Archaea. It should be noted that most of what we know about membrane protein assembly in bacteria is based on studies conducted in Escherichia coli.

INTRODUCTION

Integral membrane proteins are ubiquitous in cells and come in two classes: one in which the membrane-spanning regions are in the form of α-helices and the other in the form of β-sheets (1, 2). Those that contain β-sheets are found in the outer membrane in Gram-negative bacteria, while the helical membrane-spanning membrane proteins are predominantly found in the inner (cytoplasmic) membrane. α-Helical membrane proteins come in a variety of membrane topologies. They can span the membrane once, twice, or multiple times with different locations of their N and C termini (Fig. 1).

Figure 1.

Membrane topology of proteins. Single-spanning membrane proteins are classified into three groups. Type 1 proteins span the membrane with the N terminus out and the C terminus in. They can be synthesized as precursor proteins with a cleavable signal sequence (red arrow). Type 2 and 3 both span the membrane with the N terminus in and the C terminus out, but type 3 proteins have their TM region at the C terminus and are specified as tail-anchored proteins. Multispanning proteins span the membrane more than twice.

β-Barrel proteins that possess β-sheets as membrane-spanning regions are generally less hydrophobic than membrane proteins containing α-helical membrane regions and therefore less prone to aggregation (3, 4). For helical (bundle) proteins, the hydrophobic membrane-spanning regions can cause problems during their synthesis because exposing these regions to the aqueous environment of the cytosol can cause aggregation. To prevent the aggregation problem, α-helical membrane proteins typically insert their hydrophobic domain(s) into the Sec translocation channel cotranslationally as they are being synthesized on the ribosome (4) (Fig. 2A).

Figure 2.

Membrane targeting and insertion components. (A) Cotranslational targeting to the membrane is initiated by binding of the signal recognition particle (SRP) to the hydrophobic sequence protruding from the exit tunnel of the ribosome. The SRP-ribosome nascent chain complex (SRP-RNC) binds to its receptor FtsY at the membrane surface where the nascent chain engages the insertion site (SecYEG, YidC, SecYEG/YidC, or holotranslocon). (B) Bacterial insertion complexes: The holotranslocon with its components SecYEGDF, YidC, and SecA is required for multispanning proteins with large periplasmic domains, whereas SecYEGDF and YidC are sufficient for proteins with small periplasmic regions. YidC alone is capable of inserting small single- and double-spanning proteins. Some proteins can autonomously insert. For all these translocation pathways, most membrane proteins are targeted by SRP and its receptor FtsY.

In cotranslational membrane insertion, the signal recognition particle (SRP) interacts with the hydrophobic region of the nascent protein chain, as it emerges from the ribosomal tunnel exit. SRP binds to the ribosome in proximity to the universally conserved ribosomal protein uL23 (u stands for universally conserved) near the tunnel exit (5, 6) (Fig. 2A). The SRP-ribosome nascent chain (RNC) complex is then directed to the SRP receptor FtsY at the cytoplasmic membrane surface via the SRP-FtsY interaction. The RNC is subsequently released and transferred to either the Sec translocon or YidC (Fig. 2B) for membrane insertion as protein synthesis continues (7, 8). The interaction of the GTP-charged domains of Ffh, the proteinaceous SRP component, and FtsY facilitates the coordinated GTP hydrolysis resulting in the dissociation of SRP and FtsY, allowing the next round of targeting (9, 10).

After delivery of RNCs to the membrane, membrane insertion likely occurs at the lateral gate of SecY, the channel-forming subunit of the bacterial Sec translocon (Fig. 3) (11, 12). The hydrophobic segments of the inserting membrane protein contact the transmembrane helices (TM) TM2b, TM3, TM7, and TM8 of the lateral gate of SecY (13, 14), whereas the periplasmic portions of the substrate protein chain are translocated along the central hydrophilic cavity of SecY in an unfolded state similar to the way secretory preproteins are translocated by the translocon. As additional hydrophobic segments enter and exit the Sec translocon, it cooperates with YidC (15) (Fig. 2B), which probably collects the inserted TM segments in contact with the lateral gate (16–19), facilitating the 3-dimensional folding of the nascent protein. Finally, the fully membrane inserted and folded protein is released laterally into the lipid bilayer of the cytoplasmic membrane.

Figure 3.

Membrane translocase SecYEG. The SecYEG of Thermotoga maritima is shown (Protein Data Bank [pdb]: 3DIN). Highlighted in the structure are the TM segments of the lateral gate (dark blue) comprising TM2b, TM3, TM7, and TM8 and the plug domain (red), which is thought to seal the channel on the periplasmic side. SecG is displayed in light blue and SecE in green.

In addition to cotranslational integration, some small membrane proteins can be inserted posttranslationally. Typically, this is mediated by YidC alone (Fig. 2B) for proteins of less than about 50 amino acid residues. YidC has also been suggested to insert C-tail-anchored proteins (20). Since the ribosomal peptide tunnel covers about 40 amino acid residues (21), small proteins expose their transmembrane segment at the ribosomal tunnel exit only after the release of the protein chain from the peptidyltransferase center (PTC) of the ribosome. C-tailed proteins also expose their hydrophobic segment to the cytoplasm only after release from the ribosome. Therefore, these proteins likely contact the membrane surface first and are then posttranslationally inserted by YidC. Most small membrane proteins that are posttranslationally inserted interact directly with the membrane surface and subsequently enter into the hydrophilic groove of YidC (Fig. 4). Since monomeric YidC has no transmembrane channel, the hydrophilic periplasmic portions of the substrate are released from the groove prior to reaching the periplasm. How this occurs is currently unknown.

Figure 4.

Membrane insertase YidC. The crystal structure of YidC of Escherichia coli (adapted from reference 170; pdb: 3WVF) has a large periplasmic β-sandwich domain (gold) and a cytoplasmic coiled-coil domain (green). The indicated membrane-spanning helices form a hydrophilic groove with TM3 (purple) and TM5 (red) functioning as a hydrophobic clamp that binds the substrate hydrophobic sequence during its insertion. The substrate contact sites of TM3 and TM5 are highlighted (yellow dots in right panel).

In this review we will discuss the events by which membrane proteins are targeted and inserted by the SecYEG and YidC components of Gram-negative bacteria. We also will discuss the occurrence of homologous mechanisms in Archaea, Gram-positive bacteria, and Eukarya as well as discuss what is known about autonomous membrane insertion (Fig. 2B).

MEMBRANE TARGETING PATHWAYS

Cotranslational protein targeting offers a significant conceptual advantage over posttranslational protein targeting because it reduces the risk of protein misfolding and aggregation by coupling protein synthesis to membrane targeting and insertion. Cotranslational protein targeting in both eukaryotes and prokaryotes is mediated by the universally conserved SRP pathway and appears to be essential in all organisms with the notable exception of the yeast Saccharomyces cerevisiae (22), some Streptococcus species (23), and Leptospira sp. The latter organism appears to lack the SRP pathway completely, and it remains to be analyzed how membrane proteins are recognized and membrane targeted in this bacterium (24).

Substrates of the Bacterial SRP Pathway

In bacteria, targeting of most secretory and outer membrane proteins occurs posttranslationally via the SecA pathway (25, 26), and the SRP pathway is primarily used for the insertion of cytoplasmic membrane proteins (7, 8, 27). This is in contrast to most eukaryotes, where the majority of integral membrane proteins and secreted proteins engage the SRP pathway (28). The bacterial SRP recognizes only a few secretory proteins by virtue of their unusually hydrophobic signal sequences (29). The hydrophobicity of N-terminal signal anchor sequences or cleavable signal sequences, in the case of secretory proteins, appears to be a major determinant for substrate recognition by SRP (10, 30). Additional factors like translational slowdown by nonoptimal codons (31, 32) and the absence of helix-breaking amino acids like proline within the signal anchor sequence (33, 34) further contribute to SRP recognition. Finally, amphipathic helices can also be recognized by SRP; this was shown for the potassium sensor KdpD (35) and the heat shock transcription factor σ³² (36, 37).

SRP and Its Receptor

Usually, the bacterial SRP is composed of a single GTPase protein component, termed Ffh (38), which is bound to either the 4.5S RNA in Gram-negative bacteria or the 6S RNA in Gram-positive bacteria (39). The bacterial SRP receptor (SR) also consists of a single GTPase subunit, termed FtsY, which is peripherally attached to the cytoplasmic membrane via lipid-binding helices (40–43). The structure of SRP and the structure of the NG domain of its receptor are shown in Fig. 5. Despite their simpler architecture in comparison with the eukaryotic SRP components, Escherichia coli SRP and FtsY are sufficient to mediate cotranslational protein targeting to the endoplasmic reticulum membrane in vitro (44).

Figure 5.

The structure of SRP and it receptor. (A) Crystal structure of the Escherichia coli SRP (adapted from reference 112; pdb: 2XXA). The conserved NG domain of the protein subunit Ffh (54 homologue) is shown in blue and the signal-sequence binding M domain in yellow. The 4.5S RNA is shown in red and the conserved tetraloop and loops A and B are indicated. (B) Crystal structure of the E. coli SRP receptor FtsY (pdb: 2QY9). Only the structure of the conserved NG domain is known, while the less conserved N-terminal A domain has not been crystalized so far.

The 4.5S RNA is encoded by the ffs gene in E. coli and consists of 115 nucleotides. Processing of both the 5′ end by RNaseP (45) and the 3′ end by RNaseT and RNasePH has been documented (46), but its physiological significance is not clear. 4.5S RNA is homologous to domain IV of the much longer eukaryotic 7S RNA and forms an elongated hairpin structure with a conserved GGAA tetraloop located at the closed end (Fig. 5A). Five additional internal loops are present of which the loops A and B mediate binding to a helix-turn-helix motif of SRP’s M domain (47) (Fig. 5). The available data demonstrate that the RNA plays an active role during the SRP cycle by regulating GTP hydrolysis (see below). The affinity of the 4.5S RNA for Ffh is in the picomolar range (48, 49). Assuming a cellular Ffh concentration of 0.3 to 0.5 μM and a cellular 4.5S concentration of 1.3 to 2 μM in E. coli (25), all Ffh should have 4.5S RNA bound. The 2- to 4-fold excess of 4.5S RNA over Ffh led to the idea of an SRP-independent function of 4.5S RNA (3).

Ffh and FtsY are composed of three domains each, and their respective N and G domains are remarkably similar in sequence and structure (Fig. 5). The N domains form a four-helix bundle, while the G domains contain a Ras-like GTPase domain. A particular feature of the GTPase domains of Ffh and FtsY is the presence of an insertion box sequence that forms a β-α-β-α secondary structure (50, 51). This insertion box probably favors nucleotide exchange, i.e., GTP exchanges for GDP, and increases the stability of both proteins in the absence of nucleotides (52). Different from classical signaling GTPases, both Ffh and FtsY display only small structural differences in the apo-form compared with the GTP- or GDP-bound state (51, 53, 54), and both exhibit only a low basal GTPase activity (55).

The N-terminal NG domains in SRP are connected by a 30-amino-acid-long linker to the methionine-rich C-terminal M domain that allows a certain flexibility of these domains relative to each other (56) (Fig. 5). The M domain of SRP is enriched in methionine residues and contains five amphipathic α-helices, denoted αM1 to αM5. These helices form a flexible and hydrophobic groove for signal sequence binding, which is sealed by a fingerloop between αM1 and αM2 (57–59) and the flexible C-terminal helix αM5 (13). The M domain also contains the binding site for the SRP-RNA.

In many bacteria, the conserved NG domain of FtsY is N-terminally preceded by the highly acidic A domain, which in E. coli consists of 197 amino acids. However, the length of this A domain is variable in different bacteria and sometimes completely absent (42, 60, 61). The A domain is probably very flexible and its three-dimensional structure has not been solved so far. The deletion of almost the entire A domain does not drastically interfere with FtsY function, suggesting that it is not essential for the targeting reaction (42, 61). Nevertheless, the A domain stabilizes the FtsY-membrane contact because it contains binding sites for both phospholipids and the SecYEG translocon (41–43, 62–66) (see below). The A domain furthermore increases the fidelity of the targeting reaction (67; see below).

The SRP Cycle

Cotranslational targeting is initiated by binding of SRP to ribosomes/RNCs (Fig. 6). SRP forms three contacts with the ribosomal surface in close vicinity to the ribosomal tunnel exit where the signal anchor sequence (SAS) emerges (13, 68, 69). The NG domain binds close to the ribosomal proteins uL29 and uL23, and the M domain forms extensive contacts with the 23S rRNA (13) and is also in contact with uL23 (13, 68, 69). Finally, the 4.5S RNA is in contact with the bacteria-specific ribosomal protein bL32 (13) (Fig. 6). Two important aspects of the SRP-ribosome interaction are still controversial.

Figure 6.

Structure of SRP bound to RNCs. Atomic model of the RNC-bound SRP M and NG domains (adapted from reference 13; pdb: 5GAF). The NG domain is shown in blue, the M domain in green, and the 4.5S RNA in gold. The signal anchor sequence (SAS) is displayed in magenta, and the ribosomal protein uL23 is shown in red. The ribosomal surface is shown in grey.

1. How can the small number of SRP complexes scan the large excess of ribosomes? The cellular concentration of SRP in both eukaryotic (70) and bacterial cells (25, 71, 72) accounts for only about 1% of the ribosome concentration. How the small number of SRPs is able to scan the large excess of translating ribosomes for correct substrates has been extensively discussed in the field. Biochemical data demonstrate that SRP and the chaperone Trigger factor (TF) bind to uL23 in an anticooperative fashion (73–75). Because TF is much more abundant than SRP, TF could prevent stable binding of SRP to most ribosomes, unless they present an SRP-specific signal sequence (21). However, in vivo TF appears to bind preferentially to substrates longer than 100 amino acids (76), while SRP is cross-linked to nascent chains as short as 30 amino acids (77) and, on average, binds an emerging substrate when the N terminus reaches a distance of 42 amino acids from the peptidyltransferase site of the ribosome (27).

2. The timing of SRP binding to ribosomes/RNCs is also still unclear. Initially, the exposure of the entire signal sequence was proposed to be a prerequisite for SRP binding (78), which is supported by a single-molecule fluorescence resonance energy transfer (FRET) study indicating that SRP binding occurs mainly after signal sequence emergence (79). However, it was also shown that SRP can be cross-linked to RNCs even before the signal sequence is fully exposed (77) and that membrane targeting of RNCs does not require the complete exposure of the signal sequence (80, 81). A similar observation is made in eukaryotic cells (82), and once as few as 12 to 14 residues of a transmembrane domain enter the ribosomal peptide tunnel, SRP is recruited to RNCs (83). This led to the idea that the presence of a signal sequence within the tunnel is communicated by an unknown mechanism to the ribosomal surface for early SRP recruitment (80, 82). This mechanism likely involves uL23, because uL23 is part of the SRP binding site and in bacteria contains a 19-amino-acid loop formed by two β-strands that reaches into the ribosomal peptide tunnel. Deletion of this loop reduced the affinity of SRP for RNCs (80). A recent high-resolution cryo-electron microscopy study indicated that the C-terminal helix αM5 of Ffh inserts into the ribosomal peptide tunnel (13). In vivo and in vitro cross-linking experiments further validated that helix αM5 of Ffh contacts the tip of the uL23 hairpin loop in nontranslating ribosomes (84). In translating ribosomes, this contact is lost and the uL23 hairpin loop then contacts the nascent protein (Fig. 7) (84). Thus, SRP already surveys translating ribosomes before the signal sequence is surface exposed. This early interaction probably enables the small number of SRP molecules to scan many ribosomes and to initiate efficient targeting of proper substrates.

Figure 7.

The SRP cycle in Escherichia coli. Cartoon showing the individual steps of substrate recognition, targeting, and insertion of membrane proteins by the SRP/SecYEG pathway (step 1). SRP binds to nontranslating ribosomes (grey) and contacts the ribosomal protein uL23 via both NG domain (blue) and M domain (yellow). The C-terminal helix of the M domain (αM5) inserts into the ribosomal tunnel and binds to the intratunnel loop of uL23 (step 1, lower panel) (step 2). When a nascent chain (green) is approaching, the M domain retracts into the distal part of the ribosomal tunnel and the intratunnel loop of uL23 now contacts the nascent chain (step 3). Further chain elongation and full exposure of the signal sequence to the outside of the ribosomal tunnel results in stable SRP binding to the substrate and (step 4) its subsequent targeting to the membrane-bound SRP receptor. The SRP receptor FtsY binds to both lipids and the SecYEG translocon, but only the translocon-bound conformation allows stable contact between the respective NG domains of SRP and FtsY (step 5). Upon interaction of the NG domains of SRP and FtsY, a transient quaternary complex is formed consisting of the SRP-RNC and the SecYEG-bound FtsY. This quaternary complex is primed for nascent chain transfer from the targeting machinery to the SecYEG translocon (step 6). For stable docking of the RNCs onto the SecYEG translocon, SRP dissociates from uL23, which is now free to bind to SecY. Likewise, FtsY dissociates from the cytosolic loops of SecY, which now can contact uL23. These conformational changes allow, on the one hand, the insertion of the nascent chain into the SecY channel, but, on the other hand, also activate the SRP-FtsY complex for GTP hydrolysis. This activation is dependent on the 4.5S RNA (step 7). GTP hydrolysis by the FtsY-SRP complex leads to their dissociation and allows for the next targeting reaction.

The combined structural and biochemical evidence indicates that SRP already binds to vacant ribosomes (K_d: 50 to 60 nM, K_off: 10 s⁻¹) (80, 81) with the N domain of Ffh contacting the globular domain of uL23 and the C terminus inserting into the ribosomal peptide tunnel. This probably reflects a scanning mode in which SRP scans ribosomes for potential substrates (Fig. 7, step 1). In the presence of a nascent chain, SRP is displaced from the tip of the uL23 hairpin loop but remains bound to the globular domain of uL23 (Fig. 7, step 2). The C terminus of SRP retracts into the distal part of the ribosomal tunnel (13, 84), while the loop tip of uL23 is now in contact with the nascent chain. Further nascent chain elongation subsequently induces the proposed stand-by or anticipatory mode (80, 83), which likely prepares SRP to accept the possible substrate (Fig. 7, step 3). This state is characterized by a high-affinity binding of SRP to ribosomes (K_d: 1 nM, K_off: 1 s⁻¹) (80, 81). In the presence of exposed signal sequence, SRP forms a stable complex with the RNC (K_d: ≤1 nM, K_off: 0.08 s⁻¹) (80, 81) (Fig. 7, step 4) that primes it for the subsequent membrane targeting reaction. In this model, the hairpin loop of uL23 plays a crucial role in early SRP recruitment. However, it is important to note that this loop is not present in eukaryotic or archaeal uL23 and that it is also not essential in E. coli under standard laboratory conditions, although mutants with a shortened loop exhibit a cell division defect (84). Thus, the early contact of SRP to the uL23 loop is probably not mandatory for the targeting reaction, but rather improves the fidelity of targeting in fast-growing bacterial cells.

In eukaryotic cells, the SRP receptor is composed of the membrane-integral SRβ subunit, which tethers the SRα-subunit to the endoplasmic reticulum (ER) membrane. Both GTPases are required for distinct steps during the targeting reaction: while SRα is required for SRP binding, SRβ is involved in coordinating the RNC transfer to the Sec61 translocon (85, 86). The bacterial SR consists of the single GTPase subunit FtsY, which is homologous to SRα. Cell fractionation studies have indicated that FtsY exists in both a membrane-bound pool and a soluble pool (87) and it was therefore assumed that FtsY could bind to SRP-RNC complexes already in the cytosol (88). However, in vivo FtsY is almost exclusively membrane localized (43), owing to the presence of two lipid-binding helices (40–42, 64, 89). One of these amphipathic helices is located at the N terminus of the FtsY (residues 1 to 14, E. coli numbering) and a second resides at the beginning of the N domain (residues 195 to 207) (40–42). The second lipid binding helix is absolutely essential for FtsY function and conserved between different bacterial species (40, 89–92). In particular, anionic phospholipids are required for FtsY binding (41, 63) and the conformational changes associated with lipid binding activate FtsY for the subsequent interaction with SRP (89, 90, 92). Strikingly, anionic phospholipids are enriched in close vicinity to the SecYEG translocon (93, 94), ensuring that FtsY is located close to SecYEG.

The observation that FtsY binding to E. coli membranes is protease sensitive (95) led to the hypothesis that FtsY also binds to an integral membrane protein, which was later identified as SecY, the channel-forming subunit of the SecYEG translocon (65, 66). Further studies revealed that FtsY binds to the cytosolic loops C4 and C5 of SecY (64, 96). Interestingly, the same SecY binding residues are used to contact the ribosome and SecA, the motor protein in the posttranslational translocation pathway (64), suggesting competitive binding of these proteins/complexes to the SecYEG translocon (97). In vivo and in vitro cross-linking studies in combination with mass spectrometry demonstrated that each domain of FtsY contains a binding site for SecY (96) (Fig. 7). The FtsY-SecY interaction, which has in vitro a K_d of 0.18 μM, is nucleotide independent but requires the presence of phospholipids (96). Upon binding of FtsY to the phospholipid-embedded SecYEG complex, the A domain of FtsY moves away from the NG domain, which exposes the SRP binding site of FtsY (67) (Fig. 7). This ensures that SRP delivers RNCs directly to translocon-bound FtsY and probably prevents futile SRP-FtsY interactions (Fig. 7, step 5).

SRP and FtsY are representatives of a growing class of GTPases, which are regulated by GTP-dependent dimerization (98). SRP and FtsY interact via their respective NG domains through a series of conformational changes that ultimately stimulate their GTPase activities leading to their dissociation at the end of the targeting cycle (Fig. 7). The GTPase activity of the SRP-FtsY complex requires the presence of the 4.5S RNA and is defective in tetraloop mutants of the 4.5S RNA (99). RNCs induce a conformational change in SRP, leading to a more open structure in which the NG domain is exposed for accelerated interaction with FtsY (81, 88, 100) (Fig. 7, step 5). Likewise, FtsY is primed for stable interaction with SRP-RNCs by binding to phospholipids (89, 90, 92) and SecY (64–67, 96) (Fig. 7, step 5). The SRP-RNC-FtsY interaction includes a rapidly formed, but unstable early intermediate, which involves electrostatic contacts between their respective N domains (101–104) and contact of FtsY to the tetraloop of 4.5S RNA (105, 106). Further rearrangements induce the formation of a closed complex between the respective G domains of SRP and FtsY (107–109) and a reorientation of the NG heterodimer. As a result, the SecY binding site on the ribosome and the ribosome binding site on SecY are exposed (96, 110), which allows for docking of the RNC onto SecY. Conformational changes within the formed quaternary SRP-RNC-FtsY-SecY complex (96, 106, 111) likely provide the final GTPase activation step (Fig. 7, step 6). The dimerized NG domains of SRP and FtsY move to the distal end of the 4.5S RNA (106, 112, 113) and this favors the release of the signal sequence and its concurrent binding by SecYEG. The simultaneous GTP hydrolysis then renders the SRP-FtsY complex unstable, leading to its disassembly (111, 114) (Fig. 7, step 7).

SRP-Dependent and Independent Targeting to YidC

YidC constitutes a SecYEG-independent insertion site for membrane proteins and this was initially demonstrated for small phage proteins like Pf3 and M13 procoat (115–118). Because of their short length, SRP probably fails to make stable contact to their hydrophobic N-terminal region, and, therefore, these proteins do not engage the SRP pathway for targeting, although SRP can bind if their length is extended (115). Electrostatic interactions between the phospholipid head groups and the phage proteins are thought to ensure their membrane targeting (119), but whether any additional chaperones or targeting factors are involved in vivo is currently unknown. For other YidC substrates, like the 139-amino-acid-long subunit c of the F₁F_o ATP synthase, the SRP dependency is controversial (120, 121), although its length should allow for a stable SRP contact. Indeed, a recent ribosome-profiling study identified subunit c as clear SRP substrate in vivo (27). Other YidC substrates also clearly depend on SRP for targeting, like MscL (122, 123). Finally, some membrane proteins, like the multispanning membrane proteins mannitol permease or TatC, can be targeted by SRP to either SecY or YidC (124), indicating that SRP does not necessarily differentiate between SecY and YidC substrates. This is not unexpected, because SRP binds cotranslationally to N-terminal signal anchor sequences and, thus, before any downstream topogenic information about the nascent protein is available. This is also in line with data showing that substrates can be routed from the YidC pathway into the Sec pathway by increasing the charge and polarity of the translocated region (122, 125, 126).

Because of the small number of substrates that have been analyzed so far, it is currently unknown whether SRP-dependent targeting to YidC operates by the same mode as the targeting to SecYEG (Fig. 7). Several observations indicate that there might be some differences after the initial binding of SRP to YidC substrates. One important difference reflects ribosome binding. While SecY has a high affinity even to nontranslating ribosomes (127, 128), the affinity of E. coli YidC for nontranslating ribosomes is low (124, 129). This is related to the lack of an extended C-terminal helix, which in the mitochondrial homologue Oxa1 mediates ribosome binding (130, 131). Different from nontranslating ribosomes, RNCs exposing a YidC substrate show strong binding to E. coli YidC (129), suggesting that stable ribosome binding to YidC requires the insertion of the signal anchor sequence. Some marine bacteria contain YidC homologues with extended C termini, and they have been shown to bind ribosomes (132). Strikingly, appending these C-terminal extensions to E. coli YidC reduces the requirements for FtsY but not for SRP (132). In vitro transport assays using reconstituted YidC proteoliposomes also indicate that the insertion of membrane proteins via YidC is less dependent on FtsY than on SRP (124). Both FtsY and Ffh have been cross-linked to the C terminus of YidC (124), but contacts to Ffh appear to be much more pronounced than contacts to FtsY, while the opposite is observed for contacts of Ffh and FtsY to SecY (64). Whether this indeed reflects differences in the targeting modes to either SecY or YidC requires further analyses. However, it should be noted that deviations of the canonical SRP targeting mode have also been observed for the targeting to SecY (90, 133, 134).

MEMBRANE INSERTION OF PROTEINS IN BACTERIA

Although the trimeric SecYEG translocon is sufficient for RNC binding and insertion of many nascent membrane proteins (90, 135), it associates at least transiently with additional proteins. This results in different SecYEG translocon assemblies, which are dedicated to inserting a particular type of membrane protein substrate (Fig. 2B). A second, SecYEG-independent insertion site is formed by YidC, and we will begin here with the classical SecYEG pathway, then move to the slightly more complex SecYEGA system, continue with discussing YidC-mediated insertion, and end with the more complex SecYEG/YidC system.

Classical SecYEG pathway

In all three domains of life, the SecYEG complex, which provides the channel for insertion of membrane proteins and protein export, is essential. It is composed of the 10-spanning SecY protein together with SecE and SecG (Fig. 3). The SecYEG channel forms an hourglass-like helical bundle in the membrane (12). The channel narrows towards the center with an isoleucine-rich ring that opens around a translocating protein chain to form a gasket-like seal. At the periplasmic face of the channel, a movable plug helix (TM2a, Fig. 3) tightens the pore in the resting state. Upon activation by a substrate, the plug moves sidewise and thereby opens the channel.

In addition to the pore ring and plug, SecYEG possesses a lateral gate composed of TM3, TM2b, TM7, and TM8 helices of SecY. The hydrophobic segments of inserting membrane proteins can open the lateral gate to release the transmembrane segments into the lipid bilayer (Fig. 3). The opening of the gate is controlled by a hinge region between TM5 and TM6 of the SecY protein and is stabilized by the tilted helix TM3 of SecE. Early cross-linking experiments had shown that a signal sequence of a preprotein binds in the lateral gate (136) and this was verified by cryo-electron microscopy of a nascent “in transloco” complex (137). There, an amino-terminal hydrophobic segment of a nascent membrane protein was found in the lateral gate as documented by cryo-electron microscopy, underlying the fact that the lateral gate has a central function to initiate membrane insertion.

Exported proteins are predicted to insert as a loop with the signal peptide binding to the lateral gate region and the hydrophilic region in the mature region sliding through the central pore ring of SecY. Similarly, for single spanning membrane proteins, the TM segment would bind to the lateral gate, the hydrophilic region translocated through the central pore region, and the cytoplasmic region of the membrane protein released from the Sec channel. For multispanning membrane proteins, the mechanism is more complicated, although it is proposed that TM segments bind to the lateral gate, translocated regions go through the central pore region (if sufficiently long), and cytoplasmic regions of the membrane protein are released from the Sec translocon (138). It is known from photo-cross-linking studies that several TM regions interact with the lateral gate (136, 139). While it is still unknown how these are bound to the lateral gate and exit the translocase, for some membrane proteins it was shown that TMs exit the channel sequentially (17), while others are probably released from the lateral gate as pairs (140–143). Consequently, the assembly and folding of the inserting substrate then occurs outside the translocon.

Recently, several high-resolution structures of the prokaryotic as well as eukaryotic Sec translocases have been reported. One structure with a nascent chain of an exported protein bound to the SecYEG complex showed that the hydrophilic chain is translocated through the Sec pore while the signal sequence is bound to the open lateral gate (144). A structure determined by cryo-electron microscopy of the eukaryotic Sec61 complex with a nascent membrane protein chain revealed an open lateral gate and some rod-like density within it, suggesting that the hydrophobic segment resided in the lateral gate (145). Another structure of a ribosome-Sec61 complex by cryo-electron microscopy showed the signal sequence of a nascent chain results in opening of the Sec61 pore and the signal peptide is localized within the lateral gate (146).

To understand in detail how TM segments exit the SecYEG channel, the features of the nascent chain that trigger opening of the lateral gate and allow the membrane protein to exit the channel have been investigated. Von Heijne and coworkers have discovered that there is a minimum hydrophobicity of a protein segment that is required for exiting the Sec channel in bacteria and eukaryotes (147–149). Using a leader peptidase-based assay, they discovered that a stretch of 16 alanines and 3 leucines is sufficient for integration into the membrane. Further studies examined the contribution of each amino acid for membrane insertion (147). In general, the more apolar the side chain of the amino acid, the greater the tendency to insert into the lipid bilayer during exiting the channel (150). Polar and charged side chains inhibit insertion into the lipid bilayer from the channel. Notably, there is a good correlation between the mammalian Sec61 system in the ER membrane and the SecYEG system in the bacterial inner membrane. This is intriguing considering that the ER membrane lacks cardiolipin and does not generate a proton motive force (pmf) (151), which are both important for SecYEG function (93, 152). However, the hydrophobic threshold is lower for SecYEG-mediated insertion than for Sec61-promoted insertion (147).

Typical substrates for the SecYEG-dependent insertion are membrane proteins with short periplasmic domains, like mannitol permease (MtlA) (Table 1) or TatC (8, 124). These proteins contact YidC during their transfer from the SecYEG channel into the lipid phase (16) and the interaction with YidC facilitates insertion in vivo but is not required in vitro (90, 124, 153).

Table 1.

Distinct requirements of Sec-dependent membrane proteins

Protein	SRP-dependent targeting	SecYEG-dependent insertion
SecA-dependent translocation	YidC-dependent insertion	YidC-dependent folding
MtlA	+	–	–	–
FtsQ	+	+	–	–
LPase	+	+	–	–
MalF	+	+	–	+
LacY	+	–	–	+
Foa	+	–	+	–
CyoA	+	+	+	–

Note that MtlA, FtsQ, and LPase interact with YidC, but this interaction is not essential for insertion or folding.

SecYEGA

The slightly more complex SecYEGA system is responsible for inserting a number of membrane proteins with long periplasmic regions and some single-spanning membrane proteins with short loops (154–157). The ATPase SecA is found as a dimer, with high affinity to SecY. A substrate protein binds to the dimeric SecA, and SecA is eventually converted to a monomer after binding to SecYEG (158). ATP hydrolysis is required to drive the translocation of the hydrophilic protein chain across the SecYEG translocon. During this process, a cascade of conformational events occurs in SecA, e.g., the movement of the long two-helix finger, which is thought to catalyze the movement of the substrate chain within SecY (Fig. 3).

The binding of SecA to SecYEG is known to prime and open the translocation channel. Zimmer et al. (159) showed by determination of the X-ray structure of the SecAYEG complex that the lateral gate is open when SecA binds to SecYEG and the plug that seals the pore ring is displaced. The structure of the complex suggests a possible mechanism by which the preprotein chain could be inserted into the SecYEG channel. The two-helix finger domain of SecA is located at the SecYEG channel entrance where the protein chain is expected to be. They proposed that the two-helix finger could bind to preprotein and then drag the preprotein into the channel upon binding of ATP to SecA. After ATP hydrolysis, the helix finger would dissociate from the polypeptide and reset to its original position at the cytoplasmic entrance. This mechanism could explain how SecA uses ATP hydrolysis to push the preprotein across the membrane (160). However, protein translocation was not inhibited by cross-linking the two-helix finger to the SecY protein, which suggests that a large movement of the helix finger is not required (161). When SecA is in the ADP-bound state, the preprotein was found to slide back and forth suggesting that cycles of sliding, in addition to pushing, move the protein chain (162). In a more recent model, it was proposed that the polypeptide chain moves passively through the channel, and the two-helix finger plays a role in sensing the polypeptide chain, allowing translocation of the protein chain by a Brownian ratchet mechanism (163). Backsliding of the protein chain is prevented ultimately by closing the translocation pore within SecY, which is thought to be controlled by ADP binding to SecA.

Recently, the structure of a substrate-engaged SecY protein translocation intermediate was solved by X-ray crystallography (164). In this study, SecYEG, SecA, and a preprotein segment fused to SecA were used to form the structure. The SecA/SecYEG/preprotein complex revealed the signal peptide intercalated into the lateral gate with the C-terminal hydrophilic region passing through the pore ring. The signal peptide cleavage site of the nascent chain was positioned on the periplasmic side of the SecYEG channel.

For a SRP-dependent nascent membrane protein with a long hydrophilic chain following a hydrophobic domain, how might SecA access the SecYEG channel to translocate the protein cotranslationally? One possibility is that the ribosome would have to dissociate at least partly from the SecYEG channel in order to let SecA bind to the C-terminal hydrophilic loop of the nascent chain that is to be translocated. If the SecA-dependent loop is a periplasmic loop of the membrane protein, most likely the ribosome would then have to give space for SecA to perform its function.

A well-known membrane protein that is targeted by SRP (165) and goes by the SecYEGA pathway is FtsQ, which spans the membrane once with a long periplasmic tail. FtsQ requires the SecYEG translocon and the SecA motor ATPase for translocation of its large carboxyl-terminal domain to the periplasm (Fig. 8, bottom panel; Table 1) (166). The accessory component YidC is not required for translocation. YidC depletion has little effect on FtsQ insertion, even though cross-linking studies have shown YidC contacts the TM segment of FtsQ during membrane insertion (167).

Figure 8.

YidC only and SecYEG substrates. (Top) The YidC-only substrates include the simple phage coat proteins of Pf3 (50 amino acids [aa]) and M13 (73 aa), the single-spanning protein TssL (SciP; 217 aa), and the double-spanning proteins mechanosensor protein MscL (138 aa) and subunit c of F₁F_o ATP synthase (F_oc) (79 aa). (Bottom) The SecYEG substrates include the single-spanning protein FtsQ (276 aa), leader peptidase (LPase) (323 aa), CyoA of the respiration complex (315 aa), subunit a of F₁F_o ATP synthase (F_oa, 271 aa), and the maltose transporter subunit MalF (514 aa). F_oa and MalF require YidC for insertion and folding, respectively.

Another SecYEGA protein is leader peptidase (LPase) that spans the membrane twice with a large C-terminal extended periplasmic region. It follows a more complicated scenario and involves SecYEGA (Fig. 8, bottom panel; Table 1). After interaction with SRP, the short nascent LPase chain is targeted to FtsY, the SRP receptor, at the membrane surface. The TM1 segment is then inserted in the head-first direction at the interface of SecYEG and YidC (18). This insertion at the interface of SecY/YidC was demonstrated by cross-links between TM1 and YidC as well to SecY. Soon after, the TM2 segment of LPase is membrane inserted (17) and the C-terminal P2 region is pushed by the action of the SecA motor ATPase through the SecYEG channel (168).

YidC Insertase

Many small proteins that span the membrane once or twice are Sec independent and involve only the YidC insertase. Also, some multispanning membrane proteins with short loop(s) may use this “YidC-only” insertion pathway (124).

In E. coli, the YidC insertase is a six-spanning membrane protein of 548 amino acid residues. It possesses a large periplasmic β-sandwich domain before TM2 and a cytoplasmic α-helical coiled-coil domain located after TM2. The transmembrane segments 2 to 6 form a globular bundle with a hydrophilic groove in the inner leaflet of the membrane bilayer (Fig. 4). It is closed at the periplasmic side by helix EH that is oriented parallel to the membrane surface (169). No structural information is available for TM1, but it is likely a flexible helix adjacent to the folded globular structure of YidC.

As it appears in the X-ray structure, the functional unit of YidC is a monomer (170). However, YidC dimers have also been observed (171). It has been speculated that YidC is in a dynamic equilibrium between a monomeric and dimeric state, in addition to being in equilibrium as a SecYEG-associated component in the membrane.

In YidC, the coiled-coil domain located between TM2 and TM3 sprouts out into the cytoplasm (Fig. 4) and shows a different angle to the membrane when the YidC X-ray structure was compared with a structure obtained by cryo-electron microscopy, indicating structural mobility (172). Deletion of the C-terminal helix of this coiled-coil domain (Δ399-415, E. coli numbering) has a deleterious effect on the insertase activity (173). However, it is presently unknown whether this mutation affects the binding or the insertion of the substrate proteins.

Even before the structure of the full-length YidC was determined, the X-ray structure of the large periplasmic P1 domain had been solved (174, 175), which showed a β-sandwich region with a bound molecule of polyethylene glycol suggesting this domain could bind an extended polypeptide chain or lipid moiety (175). Despite its size, the exact function of the P1 domain of YidC is unclear. A motif in the P1 domain (residues 215 to 254) has been found to interact with SecF (176), and, thus, the P1 domain might be involved in the formation of the holotranslocon. It is also possible that P1 supports the folding of multispanning membrane proteins. It is important to note that many YidC homologues do not contain a large P1 domain which is in line with it having no essential function. Also consistent with this is the finding that deletions within the P1 domain do not impact the activity of the E. coli YidC for inserting the M13 procoat, and ATPase subunits a and c (176, 177).

The interaction of YidC with its substrates has been analyzed by disulfide cross-linking, which monitors only close contacts within about 4 Å apart. These studies have shown that substrates interact with TM3, TM4, and TM5 (Fig. 4) (178–180). The substrate contact sites are located at one face of the TM3 and TM5 helices across the entire membrane segments and include only hydrophobic amino acid residues. This underscores the fact that YidC binds its substrate TM segments via hydrophobic interactions. No substrate contact has been observed within TM2. This long hydrophobic segment wraps around several of the helices forming the YidC helical bundle at its outside surface and is most likely important for its stability.

Remarkably, YidC possesses a hydrophilic groove within the inner leaflet of the membrane that faces both the cytosol and lipid bilayer. The hydrophilic groove, which is mainly built by TM2 to TM6, has been suggested to contain water molecules (170). The amino acid residues exposed at the inner surface of the groove show a distribution of hydrophilic and hydrophobic side chains (170). Most intriguingly, an arginine residue is located in the center of the groove and the presence of this positively charged residue is essential in Gram-positive YidC homologues (169), however, not in E. coli (173). It has been speculated that the arginyl residue in Bacillus halodurans YidC interacts with negatively charged residues in the periplasmic domains of substrates like MifM (169). Indeed, an interaction of MifM with the hydrophilic groove has been observed using photo-cross-linkers, supporting the view that the hydrophilic groove hosts the substrate region prior to its translocation.

Another conserved feature for most YidC members is that they have a C-terminal region that contains an array of positively charged residues. This region apparently is not essential for YidC to function in E. coli, because YidC mutants lacking the C-tail can fully complement the YidC depletion strain and can insert the M13 procoat protein. Nevertheless, the C-terminal region has been shown to support the binding of YidC to translating ribosomes (181). Direct ribosome binding has been clearly shown with the YidC homologue of the marine bacterium Rhodopirellula baltica, particularly with the ribosomal proteins uL23 and uL24 (132). In addition, the mitochondrial homologue Oxa1 shows binding of its C-terminal domain to mitochondrial ribosomes in proximity to the MrpL40 ribosomal protein, the mitochondrial homologue of uL24 (182). The C-terminal region of both Oxa1 and R. baltica YidCs are extended and rich in arginines and lysines, which favor the interaction with rRNAs.

The prototypic YidC-only substrate is Pf3 coat protein. Pf3 coat is a short membrane protein that spans the membrane once with the amino terminus in the periplasm (type I membrane protein) (Fig. 8, top panel). Pf3 is targeted to the membrane without the assistance of a known protein factor and is inserted into the membrane by the YidC insertase (115). The hydrophilic groove may recruit the N-tail region to be translocated (Fig. 9). Its insertion into the membrane occurs via a hydrophobic slide mechanism (183). Support for the slide model comes from disulfide cross-linking studies that show the contacts between the Pf3-TM residues and the TM3 and TM5 helices of YidC at similar depths in the membrane (180) (Fig. 4 and Fig. 9). After translocation of the amino-terminal tail across the membrane, Pf3 coat is released from YidC. Single-molecule studies with reconstituted YidC in proteoliposomes have shown that the entire insertion process of Pf3 binding, translocation, and release from YidC takes about 20 ms (184).

Figure 9.

Predicted steps in the membrane insertion of the Pf3 coat protein. The single-spanning Pf3 protein binds to the membrane surface where it can then engage the YidC protein. The amino-terminal hydrophilic region of Pf3 coat binds into the groove of YidC, whereas the hydrophobic region is clamped by TM3 and TM5 of YidC and then slides into the membrane interior (greasy slide). During this insertion step, the hydrophilic region dissociates from the groove and is translocated to the periplasm to complete the membrane insertion process. Finally, the fully inserted Pf3 protein is released from YidC into the lipid bilayer.

Another type 1 membrane protein that requires YidC for insertion is the M13 procoat protein, which is synthesized in a precursor form of 73 amino acids with an amino-terminal signal peptide (Fig. 8, top panel). M13 procoat is targeted to the membrane by an electrostatic mechanism with the positively charged lysines at the amino- and carboxyl-terminal regions interacting with the negatively charged surface of the membrane (119). Next, YidC inserts procoat into the membrane (185) by a helical hairpin mechanism (186), most likely via the hydrophobic slide similar to the Pf3 coat protein with both hydrophobic segments inserting at the interface of TM3 and TM5. After membrane translocation of the periplasmic loop, the protein is released from YidC. In order for signal peptidase to cleave the M13 procoat, the signal sequence region most likely must first be released from YidC.

For these two proteins, YidC ideally combines the entry of the hydrophobic segment with the greasy slide mechanism and the incorporation of the hydrophilic periplasmic region of the substrate into the groove cavity (Fig. 9) (169, 183). In addition, a negatively charged residue in the substrate N-tail region is further attracted by the conserved positively charged residue in the groove of YidC (169). For translocation, the positioning of the substrate periplasmic region into the groove of YidC during insertion reduces the energy costs by about half because only the outer leaflet of the bilayer remains to be traversed by the periplasmic substrate region. It has also been speculated, based on a computational model of YidC, that the bilayer is compressed by YidC and reduces the membrane thickness around YidC (172). If this is the case, it could also contribute to the lowering of the energetic costs for the translocation of a hydrophilic region.

YidC also mediates insertion into the membrane of the tail-anchored protein TssL (SciP), a type 3 membrane protein (Fig. 8, top panel) (20, 187). Notably, this mechanism is different than what is observed in eukaryotes for C-tail-anchored proteins, which are inserted into the ER by a different machinery, the GET translocase (188). While current studies show that YidC does catalyze membrane insertion, exactly how membrane targeting is achieved is being worked out. Aschtgen et al. (20) had shown that the heat shock protein DnaK is important for facilitating membrane localization, and the SRP pathway does not seem to be involved. However, Pross et al. (187) showed in a more detailed study that the tail-anchored protein is targeted by the SRP pathway early on in its synthesis, even though the TM segment is localized at the very C terminus of the protein. They found two hydrophobic stretches in the N-terminal part of SciP that direct membrane binding of a fused green fluorescent protein (GFP). Further work is needed to fully understand the membrane biogenesis of these unconventional proteins, which make up only a small proportion of E. coli membrane proteins with only 12 members reported so far (189).

In addition to the single-spanning membrane proteins, YidC catalyzes the membrane insertion of the double-spanning membrane protein MscL that has a short translocated loop and spans the membrane twice with the topology shown in Fig. 8 (top panel). MscL is targeted to the membrane by the SRP pathway (122). Once at the membrane surface, it interacts with YidC which catalyzes the translocation of the short periplasmic region between TM1 and TM2 across the membrane (122). Most likely, the insertion mechanism is similar to M13 procoat with both transmembrane segments contacting the hydrophobic slide residues.

Finally, subunit c (of F₁F_o ATP synthase) with its short amino and carboxyl terminus in the periplasmic space is inserted by YidC (Fig. 8, top panel) (120, 121, 190, 191). Along with the results with MscL, these findings reinforce that YidC can insert proteins with a variety of membrane topologies. The common denominator is that the translocated region is short. Subunit c can also be inserted via the holotranslocon but not by just YidC (192).

SecYEG/YidC/SecDFYajC Holocomplex

The most complex system for inserting membrane proteins is SecYEG/YidC/SecDFYajC. SecYEG can associate with SecDF, YajC, and YidC to generate the so-called holotranslocon (193). In the membrane, the holotranslocon and subcomplexes of different composition are likely to exist in a dynamic equilibrium (171, 194). Owing to the considerably lower concentration of SecDFYajC in E. coli compared with SecYEG and YidC (25), only a small number of holotranslocons may exist in the E. coli membrane.

The holotranslocon is suggested to be more effective in cotranslational insertion of membrane proteins (193). The association of SecYEG with YidC and SecDFYajC does not seem to influence binding of FtsY or SecA to SecY, because identical cross-link patterns are observed in vivo and in reconstituted SecYEG proteoliposomes lacking YidC and SecDFYajC (64, 96). Thus, the interactions of the receptors of the co- and posttranslational targeting with SecY are not significantly influenced by the presence of SecDFYajC-YidC. On the other hand, the holotranslocon displays a higher affinity for nontranslating ribosomes or RNCs than SecYEG (193). This is attributed to the presence of YidC, although YidC on its own has only a very weak affinity to nontranslating ribosomes (124, 129). It is likely that ribosomes engage a shared binding site located on both SecYEG and YidC and that the increased stability of RNCs on the holotranslocon increases insertion efficiency.

Within the holotranslocon, YidC interacts with SecYEG and with SecDFYajC to cooperatively support membrane protein insertion (195). For most membrane proteins, both SecYEG and YidC are directly involved in the insertion process, whereas, for others, YidC plays a crucial role in folding of membrane proteins (196). In addition, SecDF within the holocomplex has been proposed to promote translocation of Sec-dependent substrates by binding to the hydrophilic domain of the substrate (197).

Clues to how SecDF might catalyze the translocation of hydrophilic domains of exported and membrane proteins were provided by the recent structure of the SecDF subcomplex of Thermus thermophilus solved at 3.3Å (197). The structure showed a large globular α-helical bundle with 12 transmembrane segments and a large periplasmic P1 domain (Fig. 10). Importantly, a possible proton-conducting path involving several aspartate residues was localized between TM4 of SecD and TM10 of SecF, and these residues are likely required for the pmf-driven translocation of a substrate protein chain, corroborating the role of SecDF in pmf-dependent translocation. It is conceivable that the motion of the P1 domain of SecD is controlled by the pmf, supporting the translocation of the hydrophilic domain of a membrane protein through SecYEG to the trans-side of the membrane where P1 binds the unfolded protein chain (197). Therefore, SecDF might function as a pmf-driven pulling translocation motor.

Figure 10.

Structure of SecDF. The crystal structure of SecDF of Thermus thermophilus with the 6-spanning SecD (green) and 6-spanning SecF (gold) (pdb: 3AQP). The arrow highlights the predicted proton conducting channel.

For this mechanism to work, SecDFYajC has to interact with SecYEG. Indeed, SecG was found to cross-link to SecDFYajC (193), supporting an earlier observation that depletion of SecDFYajC affects the SecG stability (198). SecDFYajC also forms a complex with YidC and has been proposed to function as a bridge to bring YidC to the SecYEG complex (199). SecDF binds to the large periplasmic domain (P1) of the E. coli YidC, and residues 215 to 254 of the P1 domain are sufficient to bind SecF protein as determined by coprecipitation (176).

For SecYEG/YidC substrates, the cooperation of SecYEG and YidC (facilitated or stabilized by SecDF) is coupled to YidC function in promoting the removal of TM segments of the inserting membrane protein from the Sec channel into the lipid bilayer, and for the folding of proteins (16, 167, 200). YidC contacts SecY at the lateral gate region as cross-links are formed between residues in TM2b, TM3, TM7, and TM8 of the SecY lateral gate region with YidC (19). Specifically, the SecY lateral gate residues were shown to contact YidC using site-directed cross-linking using the UV-activated p-benzoyl-L-phenylalanine. However, it is still not clear which residues of YidC are in contact with the SecYEG channel, although some YidC residues have been proposed based on a genetic screen (201), but not yet biochemically verified. Interestingly, YidC interacts with the SecY lateral gate in the absence of substrate and this interaction is disrupted in the presence of FtsQ nascent chains (19). The location of YidC at the lateral gate shows that YidC is optimally positioned to receive the TM segments as they exit the SecYEG channel.

When membrane proteins possess more than two membrane-spanning regions, multiple insertion events have to be initiated. As for single- and double-spanning membrane proteins, large periplasmic regions within multispanning membrane proteins render the protein strictly SecYEGA dependent for insertion. YidC is then thought to facilitate the subsequent release of TMs into the lipid phase and their folding. Closely spaced multispanning membrane proteins can also be inserted by just YidC (124), but it remains to be determined to what extent this happens in vivo.

CyoA is a protein that is SecYEGA/YidC dependent. Mature CyoA spans the membrane twice (Fig. 8, bottom panel), but is synthesized in a precursor form called pre-CyoA with an amino-terminal cleavable signal peptide. The amino-terminal leader peptide, the short periplasmic loop 1, and the first TM segment of preCyoA are inserted by the action of YidC in a pmf-independent manner, while the second TM region and the large periplasmic region require SecYEG and SecA similar to the insertion of LPase (Fig. 11, top panel; Table 1) (202, 203). During this process, the signal peptide is proteolytically cleaved by lipoprotein signal peptidase (signal peptidase 2).

Figure 11.

Topogenesis of multispanning membrane proteins. (Top) The N-terminal hairpin of CyoA is inserted by YidC alone, whereas the translocation of the C-terminal periplasmic domain requires SecYEG and SecA. After membrane insertion, the N-terminal signal peptide is cleaved off by signal peptidase 2. (Middle) F_oa requires SecYEGDF and YidC for insertion of both the N-terminal and the C-terminal TM segments. (Bottom) The membrane insertion of lactose permease (LacY) requires SecYEGDF. YidC is required for the proper folding of the permease.

The five-spanning protein F_oa (subunit a of F₁F_o ATP synthase) is inserted by the SecYEG/YidC complex (Fig. 8, bottom panel; Fig. 11, middle panel; Table 1). The translocation of the N-tail and the first periplasmic loop of F_oa is strongly inhibited in YidC and SecDF-depletion strains, respectively (120, 190); its insertion is also strongly inhibited by SecE depletion (125). Moreover, the pmf promotes the translocation of the negatively charged N tail but not the neutral P1 loop (120). Targeting of F_oa to the membrane occurs by the SRP pathway.

It is important to emphasize that, even though some membrane proteins have been found to contact YidC while being inserted via SecYEG (16, 18), their in vitro insertion may not be significantly reduced in the absence of YidC (90, 124, 135). In these cases, YidC probably has a kinetic function during SecYEG-dependent insertion, by enhancing the release of TMs from SecY and/or by facilitating their folding.

Insertion of MalF (an eight-spanning membrane protein) requires the SecYEGA complex but not YidC for membrane insertion (Fig. 8, bottom panel; Table 1). The periplasmic loops of MalF are translocated to the same extent with or without YidC (204). Translocation of the periplasmic loops requires SecYEG and SecA. While YidC does not play an insertion role for MalF, it has a chaperone function, because the folding of MalF was impaired when YidC was depleted as the MalF was unstable, and this led to a deficiency in the formation of the maltose transporter complex MalFGMalK2.

Lac permease (LacY), a 12-spanning membrane protein, is targeted by the SRP pathway and inserted at the SecYEG and YidC interface (Fig. 11, bottom panel; Table 1) (205, 206). Insertion absolutely requires SecYEG. Because all periplasmic loops are short, SecA is not needed for their membrane translocation. Lac permease, like MalF, requires YidC not for translocation, but rather is involved in postinsertional folding steps (Fig. 11, bottom panel) (207). Kaback and colleagues showed that the folding of the periplasmic P4 loop and a cytoplasmic region of the protein is perturbed under YidC depletion condition. Monoclonal antibodies that recognize folded epitopes in this region were impaired for binding to LacY. A more detailed study revealed that all the periplasmic loops were translocated when YidC was depleted; however, disulfide cross-linking studies confirmed that the folding of Lac permease was disrupted (208). As expected with YidC functioning as a chaperone, YidC contacts Lac permease during membrane insertion. Disulfide cross-linking experiments revealed that TM5 of YidC can interact with TM3, TM6, and TM9 segments of Lac permease during insertion (208). Recently, a study using single-molecule force spectroscopy showed that YidC prevents misfolding of LacY by stabilizing the unfolded state. From the unfolded state, LacY can be inserted in a stepwise manner into an immobilized liposome until the polypeptide chain is completely folded when YidC is present (209).

Although SecYEG, SecYEGA, YidC, and SecYEG/SecDFYajC/YidC translocases mediate insertion of most proteins, SecYEG associates with additional proteins, like the chaperone PpiD (194, 210, 211), but the exact impact of this interaction on membrane protein insertion is still unclear.

Finally, some membrane proteins with very short periplasmic regions like the potassium sensor protein KdpD can insert under SecYEG- and YidC-depleted conditions (212). This four-spanning membrane protein has periplasmic loops of 10 and 4 residues. Also, the potassium transporter protein KcsA was shown to insert into the membrane on its own (213, 214). Adding purified KcsA to liposomes results in efficient transmembrane insertion, which is different for purified Pf3 that inserts very poorly when added to liposomes lacking YidC (215). However, when the hydrophobic region of Pf3 coat was extended by 3 leucyl residues, membrane insertion was also independent of YidC and the pmf (215). Similarly, M13 procoat was routed into a Sec and YidC-independent insertion mechanism when the charged and hydrophilic residues in the periplasmic loop were removed and the hydrophobicity of the transmembrane segments was increased (126).

It is likely that proteins with very short periplasmic loops can insert to some extent autonomously into the membrane when YidC and/or SecYEG are absent. However, these proteins still might use the insertase or translocase in the presence of SecYEG and/or YidC. Considering that Pf3 can engage the YidC insertion pathway and that KdpD is recognized by SRP (35), it appears likely that spontaneous insertion serves only as a backup mechanism for some proteins if no translocon is available.

MEMBRANE TOPOLOGY AND POSITIVE INSIDE RULE

What determines that a protein spans the membrane with the correct orientation? Early studies have shown that the charged residues that flank the transmembrane segments control orientation (216, 217). Single-spanning membrane proteins that have their TM segment flanked by positively charged residues in the amino-terminal region span the membrane with their carboxyl terminus in the periplasm (218, 219). When the charged residues in Pf3 coat protein were changed into the opposite charge, the orientation of the protein was completely inverted (217). In conclusion, if the positively charged residues are on the amino-terminal side of the transmembrane segment, then the protein spans the membrane with their carboxy terminus facing the periplasmic space.

For proteins that span the membrane multiple times, their membrane topology is still determined mainly by positively charged residues that are adjacent to the TM segments, as has been experimentally demonstrated (216, 220–224). One popular model (225) suggests that the orientation of the first TM segment is decisive for dictating the orientation of a multispanning membrane protein. The first TM segment inserts with a specific orientation, and then the remaining TM segments are inserted in a sequential order in the opposite direction of the preceding TM segment. Membrane insertion of TM segments into the Sec translocon is a dynamic process (226–228). In the case of insertion into the ER, TM segments (with a final N_cytoC_lumen orientation) bind to the Sec translocon to translocate the amino terminus but then can undergo reorientation depending on the features of the TM segment and translocate the flanking C-terminal hydrophilic region (229, 230). Whether this reorientation occurs for type II TM segments during insertion into the bacterial SecYEG channel is not known.

While many proteins insert by a strictly cotranslational mechanism from the N terminus to the C terminus, the mechanism can be more complicated (226). For example, it was shown that a single positively charged residue introduced at the carboxyl terminus of EmrE can determine the orientation of the four-spanning protein (231). Apparently, in this case, the topology of the protein remains undefined until the carboxyl-terminal residues have been synthesized.

Phospholipids also contribute to the topology and folding of membrane proteins (232). The amino-terminal half of LacY was shown to mis-orient in the absence of phosphatidylethanolamine (PE) and to adopt an inverted conformation (233). Strikingly, the reintroduction of PE into the membrane leads to the adoption of a native-like protein conformation (234). The topological state of LacY (native or inverted) is dependent simply on the PE levels since the topological switch can occur in liposomes containing only LacY (i.e., does not require the SecYEG channel). The switch from the inverted to native conformation is reversible and occurs in both directions dependent on the lipid composition (235).

A very interesting group of membrane proteins are the ones that have a dual topology (236). In E. coli, there are six dual-topology membrane proteins known (EmrE, SugE, CrcB, YdgC, YnfA, and YbfB) (1). They exist in the inner membrane as dimers with one protomer in the inverted orientation (237). These proteins all have very defined TM segments, have few arginine and lysine residues flanking the TM segments, and are generally weakly determined by the positive inside rule.

ENERGETICS

The insertion of hydrophobic transmembrane segments into the lipid bilayer is a thermodynamically favored reaction (238). However, most membrane proteins also contain hydrophilic regions that have to traverse the bilayer to become localized in the periplasmic space. SecYEG and YidC support this reaction by reducing the required activation energy. Whereas long hydrophilic regions use ATP as an energy source to drive translocation by SecA, shorter hydrophilic regions can be translocated by coupling this unfavorable process to the favorable energetic process of inserting the hydrophobic domain into the lipid bilayer (126).

When negatively charged residues are present in the periplasmic regions of membrane proteins, the electrochemical membrane potential, generated by the pmf, is often required for their translocation (239, 240). The presence of the potential (periplasmic side positive) lowers the activation energy required to translocate a negative charge and, therefore, contributes to the translocation of negatively charged residues, thereby controlling the orientation of the membrane protein. The stimulatory role of the pmf to translocate negatively charged regions across the membrane has been shown for YidC-dependent substrates (241). In addition, the pmf has been shown to promote the translocation of Sec-dependent proteins (242, 243). For the latter, this may be due to SecDF promoting translocation in a pmf-dependent manner (see “Membrane topology and positive inside rule” above). SecDF has been shown in electrophysiology experiments to conduct protons and to display pmf-dependent conformational changes of SecD’s large periplasmic domain P1 (197).

INSERTION IN GRAM-POSITIVE BACTERIA AND ARCHAEA

Like Gram-negative bacteria, Gram-positive bacteria and archaea contain the SRP targeting components and the Sec translocon (151, 244, 245). Whereas the YidC homologues in Gram-positive bacteria are well known (173, 246, 247), YidC in archaea have been clearly identified only recently (248). Although the role of the SecYEG, SecDF, YidC, and SRP components in membrane protein insertion has not been intensively investigated in Gram-positive bacteria, they have been tested to some extent for protein export in these organisms (249–252).

Some Gram-positive species contain multiple gene copies for SecDF (254) and YidC (255). The SecYEG has been shown to be essential for life in the Gram-positive Bacillus and Streptococcus, whereas the SRP pathway components Ffh and FtsY are not essential in streptococci (23). Ffh and the FtsY can be deleted in Streptococcus mutans resulting in a diminished stress tolerance (23). One explanation for why SRP is not essential in this organism is that it has two YidC paralogs (YidC1 and YidC2), of which YidC2 functions in S. mutans in the targeting of proteins to the membrane by binding directly to ribosomes. Although YidC1 is dispensable for bacterial growth, YidC2 is absolutely required for growth under stress conditions (23). YidC2 might play an important role in the assembly of the F₁F_o ATP synthase, which would explain why YidC2 is important for growth under acid stress conditions. Under acid stress conditions, F₁F_o ATP synthase is believed to hydrolyze ATP and to help maintain cellular pH homeostasis by pumping protons out of the cytoplasm into the extracellular medium. Another distinctive feature of YidC2 in S. mutans is that it possesses a large C-terminal domain that is believed to function in ribosome binding (246). In this respect, it is similar to the mitochondrial Oxa1 and, in line with this, YidC2 can functionally substitute for Oxa1 in mitochondria when a mitochondrial matrix targeting signal is added (246).

There are also two gene copies of YidC in Bacillus subtilis called YidC1 (also known as SpoIIIJ) and YidC2; YidC2 is a housekeeping gene, while YidC1 has a more specialized function required for sporulation. In B. subtilis, only one YidC paralog is sufficient for vegetative growth (251, 256). Interestingly, the YidC2 gene is preceded by the gene encoding MifM that is inserted by SpoIIIJ (257). MifM is a single transmembrane protein that has a translational arrest domain in the C-terminal part of the protein. The failure of MifM to insert causes translational arrest that leads to changes in the conformation of the mRNA leading to upregulation of YidC2.

In archaea, three SRP components were found, namely Ffh, SRP19, and a 7S RNA (253, 258). The SRP receptor is composed of one polypeptide FtsY, as is found in bacteria. The SecYEβ complex and the ancillary SecDF component are present, while the SecA motor ATPase is missing.

DNA sequence alignments of yidC with archaeal genomes show a few weak candidates among the Euryarchaeota and it was unclear if there is an archaeal insertase at all (259). Recently, the structure of one of these putative homologues, DUF106, of Methanocaldococcus jannaschii was determined and it showed that DUF106 is structurally similar to membrane insertases (248). In addition, it was found to bind to nascent chains of subunit c of the ATP synthase by photo-cross-linking. The DUF106 protein structure shows only 3 TM regions and might resemble a minimal version of YidC (260). The 3 TMs show a structural homology to TM2, TM3, and TM6 of E. coli YidC. The protein also has a cytoplasmic coiled-coil region in between the first two conserved TM segments like the bacterial YidC proteins. Additional members in all archaeal phyla were found when using the coding sequence of DUF106 for sequence alignment studies suggesting that YidC is present in all the three domains of life (261).

Some cofactor-containing membrane proteins use the twin-arginine-translocation pathway for their insertion after folding (25, 262). The most prominent member of these proteins is the Rieske protein of Paracoccus and Synechocystis (263, 264). This protein has an iron-sulfur cluster bound and contains an N-terminal signal anchor that keeps the protein at the membrane surface. So far, no multispanning Tat-dependent protein has been discovered, and it remains unknown how a folded protein is actually inserted into the bilayer.

CONCLUSION AND OUTLOOK

In recent years, much progress has been made in revealing the molecular details of protein targeting and insertion into the bacterial membrane. Most or all of the key components have been identified and their structures have been determined at moderate or high resolution. Snap shots of the membrane insertion process are even available with a structure of a ribosome-bound SecYEG channel and an inserted nascent membrane protein chain. In addition, the structure of a substrate-engaged SecYEG/SecA complex is available showing the signal peptide in the lateral gate and the substrate hydrophilic region (to be translocated) passing through the pore ring. These and other breakthroughs have provided information on the mechanism of insertion of some simple proteins.

Still much needs to be accomplished. For example, a high-resolution structure of the holotranslocon needs to be determined to deepen our understanding of how SecYEG and YidC cooperate together to move transmembrane segments of membrane proteins through the machinery and eventually integrate into the lipid bilayer. It is important to solve the structure of the holocomplex with or without a substrate bound to the complex. Similarly, we need to have high-resolution structures of YidC in other states so we can gain insight into precisely how the substrate interacts with the hydrophilic groove and how the transmembrane segment of the substrate is bound to YidC in a transmembrane conformation. This information coupled to other techniques such as real-time experiments and molecular dynamic simulations will allow us to visualize how a membrane protein inserts into the membrane.

The mechanism of insertion of bacterial membrane proteins is still in its infancy. In general, it is not clear how the Sec holocomplex handles TM segments, especially the internal TM segments of multispanning membrane proteins and those TM segments that are quite hydrophilic. Although YidC contacts TM segments of membrane proteins after exiting the lateral gate of the SecYEG, how it functions as a chaperone to promote folding of the membrane proteins is not known. Intriguingly, a residue at the C terminus of a membrane protein can be a determinant of the topology and can reverse the topology of TM segments that precede it. How the previously inserted TM segments are reoriented after their Sec-mediated insertion is still a mystery and points to a dynamic and quite adaptable membrane insertion process.

It is clear that, to address these complex and dynamic events, continued X-ray crystallography, sophisticated cryo-EM methods, and a validation by biochemical studies will be required. Single-molecule FRET studies need to be exploited to determine the conformational changes in the substrate and the translocases during membrane insertion. Real-time imaging techniques will be needed to determine the location and dynamics of the insertases in living cells. Ultimately, these studies will provide significant new advances providing insights into the protein transport mechanism. Understanding the insertion mechanism will be of great interest also to the biotechnology industries and drug discovery since many key drug targets are multispanning membrane proteins.