A TAle of Two Pathways: Tail-Anchored Protein Insertion at the Endoplasmic Reticulum

Abstract

Tail-anchored (TA) proteins are an essential class of integral membrane proteins required for many aspects of cellular physiology. TA proteins contain a single carboxy-terminal transmembrane domain that must be post-translationally recognized, guided to, and ultimately inserted into the correct cellular compartment. The majority of TA proteins begin their biogenesis in the endoplasmic reticulum (ER) and utilize two parallel strategies for targeting and insertion: the guided-entry of tail-anchored proteins (GET) and ER-membrane protein complex (EMC) pathways. Here we focus on how these two sets of machinery target, transfer, and insert TAs into the lipid bilayer in close collaboration with quality control machinery. Additionally, we highlight the unifying features of the insertion process as revealed by recent structures of the GET and EMC membrane protein complexes.

Integral membrane proteins are a structurally diverse class of proteins that comprise 25%–30% of all encoded eukaryotic genes (Krogh et al. 2001). They play critical roles in processes such as cellular communication, organelle biogenesis, trafficking, and cell adhesion. Though resident in all cellular membranes, the majority of integral membrane proteins are initially targeted to the endoplasmic reticulum (ER) for their biogenesis (Shao and Hegde 2011a).

The unifying structural feature of all integral membrane proteins is the presence of one or more transmembrane domains (TMs), a stretch of hydrophobic amino acids that typically spans the lipid bilayer as an α-helix. These TMs are not only functionally important, but also serve as a recognition and targeting signal for directing proteins to the correct cellular compartment (Park and Rapoport 2012; Guna and Hegde 2018). Integral membrane proteins begin their synthesis on the ribosome in the cytosol, an aqueous environment where TMs are prone to aggregation and nonspecific or off-target interactions. As a result, a highly coordinated process is required to ensure these TMs are recognized, shielded from the cytosol, targeted to the correct membrane, and ultimately integrated and assembled into the lipid bilayer where they are most energetically stable (White and von Heijne 2008; Shao and Hegde 2011a).

The biophysical and topological diversity of integral membrane proteins have required the evolution of multiple parallel biogenesis pathways (Wickner and Schekman 2005; Rapoport 2007). The first identified and most extensively studied is the cotranslational pathway, where proteins are targeted to the ER by the signal recognition particle (SRP) (Walter and Blobel 1981) and inserted into the membrane by the SEC61 translocation channel (Deshaies and Schekman 1987; Görlich and Rapoport 1993). Soon after its discovery, it was apparent that not all proteins can access this cotranslational pathway, and thus require alternative strategies for reaching the ER (Kutay et al. 1995). One important class of such proteins are tail-anchored (TA) proteins, which have a single TM immediately preceding their stop codon and must therefore be targeted and inserted post-translationally.

TA proteins are inserted into the ER primarily by two pathways: the guided-entry of tail-anchored proteins (GET), which relies on the hetero-oligomeric Get1/2 complex for insertion, and the ER-membrane protein complex (EMC), which is a multifunctional insertase that caters to a subset of TA proteins. However, it remains possible that alternative pathways contribute to TA biogenesis under particular cellular conditions (Aviram et al. 2016). Decades of elegant biochemical and genetic experiments have defined the molecular players required for these parallel pathways, but until recently the molecular details of the insertion step have remained elusive. Exploiting the recent technological advances in cryo-electron microscopy (cryo-EM), a series of contemporaneous structures of the EMC (Bai et al. 2020; Miller-Vedam et al. 2020; O'Donnell et al. 2020; Pleiner et al. 2020) and Get1/2 (McDowell et al. 2020) have revealed unparalleled mechanistic insight into how TA proteins are inserted into the bilayer. In this review, we summarize our understanding of TA targeting, handover, and insertion into the ER by the GET and EMC pathways, emphasizing the molecular details revealed by recent structural studies.

CHALLENGES OF TA PROTEIN BIOGENESIS

Despite only accounting for roughly 3%–5% of the eukaryotic proteome (Kalbfleisch et al. 2007; Kriechbaumer et al. 2009), TA proteins are essential for a variety of cellular processes including protein biogenesis and quality control, apoptosis, signal transduction, lipid biosynthesis, and vesicular trafficking (Chacinska et al. 2009; Südhof and Rothman 2009; Shamas-Din et al. 2013). TA proteins fulfill their functions in various cellular membranes, including mitochondria, peroxisomes, and chloroplasts (Kutay et al. 1993; Yabal et al. 2003). Typically, these functions are mediated by a variable amino-terminal domain that can engage in critical protein–protein interactions (Fig. 1A).

Figure 1.

TA targeting and insertion via the GET and EMC pathways. (A) Because of the positioning of their characteristic carboxy-terminal transmembrane domain (TM), TA proteins must be recognized and targeted to the correct cellular compartment post-translationally. Specific targeting to the ER is carried out by a cascade of cytosolic chaperones, which culminates in handoff to either the ER-resident EMC or the Get1/2 complex for insertion into the membrane. The TA biogenesis pathways rely on tight coupling with cellular quality control machinery to ensure robust and efficient clearance of substrates that fail to target or are mislocalized. (B) Specificity in TA protein targeting is determined by the biophysical properties of the substrate TM and its flanking residues. Discrimination of substrates to either the GET or EMC pathways is dictated primarily by TM tendency (as defined in Zhao and London 2006) and hydrophobicity (ΔG as defined in White and Wimley 1999), with the EMC responsible for TA proteins containing low-hydrophobicity TMs (Guna et al. 2018). Mitochondrial TAs typically contain TMs whose hydrophobicity overlaps substantially with ER-destined substrates of the EMC; how discrimination within this group of low-hydrophobicity TA proteins is achieved to ensure targeting to the correct cellular compartment remains unclear.

The unifying structural feature of TAs is a single TM within ∼50 amino acids of the carboxyl terminus (Jan et al. 2014; Chartron et al. 2016). This TM and its flanking sequences are necessary and sufficient for targeting and insertion of TAs into their diverse organelles (Whitley et al. 1996; Egan et al. 1999; Mullen and Trelease 2000; Beilharz et al. 2003; Masaki et al. 2003). During translation, this TM remains occluded within the ribosomal exit tunnel, which houses ∼40 amino acids (Voorhees et al. 2014), until termination and release from the ribosome (Kutay et al. 1993). TA proteins therefore cannot utilize the canonical cotranslational pathway for targeting, unlike the majority of integral membrane proteins, which contain a hydrophobic targeting signal close to their amino terminus (Halic and Beckmann 2005; Shan and Walter 2005).

During cotranslational targeting, TMs remain shielded by the ribosome or are immediately captured by SRP (Egea et al. 2005), which sits at the ribosomal exit tunnel where its M domain can directly shield the nascent TM (Zopf et al. 1990; Halic et al. 2006; Janda et al. 2010; Voorhees and Hegde 2015). Release of the TM from SRP can only occur upon binding to the SRP receptor, which triggers guanosine triphosphate (GTP) hydrolysis, thereby coupling TM release with arrival at the membrane (Miller et al. 1993; Egea et al. 2004; Focia et al. 2004). This tight coupling minimizes TM exposure with the bulk cytosol throughout the insertion process. Because TA proteins are released from the ribosome, their targeting requires higher levels of coordination to minimize exposure of their TM to the cytosol.

Post-translational delivery of TAs to the ER comes with a unique set of challenges. First, the TAs contain biophysically extremely diverse TMs that encompass a wide range of lengths, hydrophobicity, helical propensity, and flanking charge (Fig. 1B; Marty et al. 2014; Rao et al. 2016). How TAs are correctly targeted to a specific membrane within the cell using only information contained within the TM and its surrounding residues remains incompletely understood (Chio et al. 2017). Second, TA proteins may be transiently exposed to the cytosol upon release from the ribosome or during handover between chaperones, resulting in an increased propensity for aggregation. This possibility creates a need for tight coupling with quality control machinery throughout the biogenesis process (Shao et al. 2017). Both the GET and EMC pathways must address these challenges to ensure specific and robust targeting and insertion of their TA clients into the ER membrane.

DISCOVERY OF THE GET AND EMC PATHWAYS

The discovery of the GET and EMC targeting pathways occurred through markedly different routes. Discovery of the GET system was spurred by early experiments establishing that insertion of synaptobrevin, a TA protein that functions in vesicular trafficking, was SRP- and Sec61-independent (Kutay et al. 1995). A putative new pathway was proposed and shown to be protein-mediated and adenosine triphosphate (ATP)-dependent (Kutay et al. 1995; Steel et al. 2002; Yabal et al. 2003). Other studies suggested a range of potential insertion pathways, including unassisted insertion (Brambillasca et al. 2005, 2006; Colombo et al. 2009), an HSP70-mediated pathway (Abell et al. 2007), or SRP/Sec61 acting post-translationally (Abell et al. 2004). Eventually, a cytosolic factor involved in TA targeting was identified biochemically using a cell-free translational system (Stefanovic and Hegde 2007). Asna1, later renamed TRC40 (for TM-recognition complex protein of 40 kDa) was established as the first chaperone specifically catering to TA proteins.

The mammalian TRC40 is an ATPase and was found to associate with the ER membrane in an ATP-dependent manner, reconciling the early observation that TA protein targeting was energy-dependent (Stefanovic and Hegde 2007; Favaloro et al. 2008). Separately, a close ortholog of TRC40, Get3, was identified in budding yeast and found to genetically and physically interact with two genes, Get1 and Get2 (Ho et al. 2002; Schuldiner et al. 2005; Auld et al. 2006). Despite not being lethal in yeast, depletion of Get1, Get2, and Get3 resulted in pleiotropic phenotypes we now know reflect the loss of TA insertion (Schuldiner et al. 2008). It was soon established using a reconstituted system that Get1/2 was not only the Get3 receptor, but also the minimal insertion machinery for TA proteins (Fig. 2A; Mariappan et al. 2011; Zalisko et al. 2017). Subsequent experiments in budding yeast identified additional cofactors required for TA targeting: Get4, Get5, and Sgt2 (Jonikas et al. 2009; Costanzo et al. 2010). Eventually, mammalian homologs of Get1 (WRB), Get2 (CAML), Get4 (TRC35), Get5 (UBL4A), and Sgt2 (SGTA) were discovered (Vilardi et al. 2011, 2014; Xu et al. 2012; Yamamoto and Sakisaka 2012). It is now clear that the yeast and mammalian GET pathways rely on a similar set of core components, but for clarity we will primarily use yeast nomenclature.

Figure 2.

Architecture of the TA insertases for the GET (Get1/2) and EMC pathways. (A) Schematic of the hetero-oligomeric Get1/2 insertase complex, along with the cytosolic targeting chaperone, Get3. (B) The structure of the mammalian TRC40 (Get3) anchored at the membrane by the tetrameric WRB/CAML (Get1/2) complex as determined by cryo-electron microscopy (cryo-EM) (McDowell et al. 2020) (6SO5). Recent structures suggest that following substrate delivery by the homodimeric TRC40, WRB/CAML may function as a heterotetramer for transmembrane domain (TM) insertion into the ER via a conserved hydrophilic groove. (C) Schematic of the nine-subunit mammalian EMC. The composition of the EMC across all eukaryotes is highly conserved, except for the soluble subunit, EMC8/9. Yeast do not contain a homolog for EMC8, whereas EMC9 (a close paralog of EMC8, whose assembly is mutually exclusive with EMC8) is found only in higher eukaryotes. (D) Structure of the human EMC, as determined by cryo-EM, displayed from the side of the hydrophilic intramembrane vestibule (Pleiner et al. 2020) (6WW7). The EMC contains a conserved tripartite organization that includes (1) a large lumenal domain of largely unknown function; (2) 14 membrane-spanning TMs (nine of which are well-ordered and depicted here) that form a hydrophilic vestibule, accessible from the cytosol and thought to be important for TM insertion, and a hydrophobic crevice, accessible only from within the bilayer; and (3) a cytosolic domain centered around EMC2.

For many years, the majority of research in this area focused on addressing detailed mechanistic questions about TA targeting, handover, and insertion via the GET machinery. However, there were several clues that the GET pathway could not account for the targeting of all TA proteins. First, it was clear that TRC40 (Get3) could not engage all TAs, such as cytochrome b5 (Cyb5), which contains a poorly hydrophobic TM (Stefanovic and Hegde 2007). This observation was rationalized in part by earlier studies that found Cyb5 could autonomously insert into liposomes in vitro (Brambillasca et al. 2005, 2006). Despite rigorous experiments in reconstituted systems, lingering questions remained about whether this type of spontaneous pathway could efficiently operate in the crowded cytosol of the cell. Second, genetic studies showed only partial defects in yeast lacking GET components (Schuldiner et al. 2008; Jonikas et al. 2009), as well as persistent insertion of some TAs in Get3-deficient lysates, cells, and tissues (Rivera-Monroy et al. 2016; Casson et al. 2017). Third, bioinformatic analysis had identified Get1 as a member of the Oxa1 superfamily of insertases, which includes the bacterial YidC, the archaeal Ylp1, the plant Alb3, and the eukaryotic EMC3 and TMCO1 (Anghel et al. 2017). The presence of related but orphan insertases in the ER suggested that putative parallel pathways for membrane protein insertion remained to be discovered.

In hindsight, we can reconcile many of these observations with the discovery of the EMC pathway for TA insertion. The EMC was first identified through a genetic screen for protein folding factors in yeast where it was found to have a synthetic lethal relationship with the GET components (Jonikas et al. 2009) and was later shown to be universally conserved in all eukaryotes (Christianson et al. 2011; Wideman 2015). However, establishing a link between the EMC and TA insertion was challenging, and the molecular function of the EMC remained elusive for almost a decade after its identification. The key to unlocking its role as a membrane protein insertase was the observation that depletion of the EMC resulted in a commensurate decrease in the steady state levels of squalene synthase, a TA protein with a modestly hydrophobic TM (Volkmar et al. 2019). Biochemical reconstitution experiments definitively demonstrated that the EMC is both necessary and sufficient for insertion of squalene synthase and several other TA proteins with poorly hydrophobic TMs (Guna et al. 2018).

It is now clear that in addition to its role in TA insertion, the EMC is required for the cotranslational insertion of multipass membrane proteins that contain an N_exo TM (i.e., in which their amino terminus faces the ER lumen or extracellular environment), which includes many G-protein-coupled receptors (Chitwood et al. 2018; Shurtleff et al. 2018) and further functions as an intramembrane chaperone involved in stabilization of multipass membrane proteins containing polar or charged residues within their TMs (Shurtleff et al. 2018; Tian et al. 2019). The broad client base of the EMC at least in part rationalizes its highly pleotropic phenotypes and thus the difficulty in assigning its molecular function.

Based on several biochemical, genetic, and structural advances, we now know that although the smallest functional unit of Get1/2 is a heterodimer, the EMC contains eight subunits in yeast and nine in mammals (seven membrane-spanning and two peripherally associated; Fig. 2). Here we outline how this work has collectively led to our current understanding of TA biogenesis.

TARGETING A TA PROTEIN THROUGH THE AQUEOUS CYTOSOL

The first step in TA protein biogenesis is targeting to the ER membrane. Successful targeting requires both specificity, to ensure delivery to the correct organelle, and coordination, to prevent potentially cytotoxic exposure of the TM to the cytosolic milieu. Both substrate capture from the ribosome and targeting to the ER have been mechanistically delineated for the GET pathway and are largely mediated by the widely conserved ATPase Get3. Crystal structures show that Get3 acts as a symmetric homodimer that binds, shields, and releases a substrate TM in an ATP-dependent manner (Bozkurt et al. 2009; Hu et al. 2009; Mateja et al. 2009; Suloway et al. 2009; Yamagata et al. 2010). In archaea, Get3 has been observed to form a homotetramer (Suloway et al. 2012), although functional experiments suggest that the homodimer alone is sufficient for targeting in most eukaryotes. In the ADP (or apo) state, Get3's two α-helical substrate-binding domains are apart. ATP binding triggers a conformational rearrangement that brings the two α-helical domains together to create a composite, methionine-rich hydrophobic groove that can directly bind and shield substrate TMs (Fig. 3A; Mateja et al. 2015). Get3 contains binding sites for the cytosolic domains of ER-resident Get1 and Get2, thereby successful loading onto Get3 ensures directionality and specificity in targeting to the correct membrane (Mariappan et al. 2011; Stefer et al. 2011).

Figure 3.

Structural basis of TA protein capture in the cytosol. (A) Surface filling model of the crystal structures of the Saccharomyces cerevisiae Get3 homodimer in its nucleotide-free (Hu et al. 2009) 3H84 and ADP•AlFl⁻₄ bound (Mateja et al. 2009) (2WOJ) states. Hydrophobic residues are shown in yellow and more hydrophilic in green, revealing the methionine-rich substrate binding groove that forms between the two Get3 monomers upon ATP binding (represented by the ADP•AlFl⁻₄ structure). (B) Crystal structures of the cytosolic chaperone calmodulin in its open (Chattopadhyaya et al. 1992) (1CLL) versus closed, peptide-bound (Halling et al. 2009) (2VAY) states colored as in A. The peptide has been omitted to facilitate visualization of the substrate-binding groove. (C) Structure of the mammalian EMC as determined by cryo-electron microscopy (cryo-EM) depicting the lack of an obvious substrate binding groove within the cytosolic domain of the apo-complex (Pleiner et al. 2020) (6WW7). (D) Multiple sequence alignment of the cytosolic methionine-rich loop of EMC3. The conserved methionines (yellow), found in all Oxa1 superfamily insertases, were shown to be important for insertion by the EMC in human cells and may play a role in substrate capture in the cytosol.

Even in the presence of high concentrations of ATP, Get3 cannot independently capture substrates and relies on handover from an upstream targeting factor (Wang et al. 2010; Gristick et al. 2014; Mateja et al. 2015; Mock et al. 2015). The largely nonspecific chaperone Sgt2 (SGTA in mammals) is thought to sit immediately upstream of Get3. Sgt2 can associate with TAs destined to all cellular organelles (Wang et al. 2010; Mateja et al. 2015; Mock et al. 2015; Itakura et al. 2016; Lin et al. 2021), but if sufficiently hydrophobic, a TA is transferred to Get3, committing it to ER targeting (Shao et al. 2017). Whether Sgt2 directly captures substrates from the ribosome or additional upstream factors precede binding by Sgt2 remains unclear (Fig. 4; Cho and Shan 2018; Cho et al. 2021).

Figure 4.

TA protein targeting to the ER via the GET pathway. (A) In yeast, transfer of a substrate transmembrane domain (TM) from the nonspecific chaperone Sgt2 (yellow) to the ER-specific chaperone Get3 (green, in its ATP-bound state: “T”) is mediated by the Get4 and Get5 (orange/pink). Following substrate binding, ATP hydrolysis occurs, priming the substrate-engaged Get3 for delivery to the ER membrane. The ADP-bound (“D”) Get3 is initially captured by the flexible amino terminus of Get2 (not shown), and the TA substrate is subsequently released by binding of the coiled-coil domain of Get1 between the Get3 monomers. Finally, ATP binding to Get3 promotes recycling from the membrane, preparing it for another round of substrate binding in the cytosol. (B) In mammals, substrate transfer from SGTA (Sgt2 in yeast) to TRC40 (Get3) requires an additional factor, Bag6. The amino terminus of Bag6 is sufficient to mediate TA transfer, whereas the carboxyl-terminus ubiquitin-like (UBL) domain (labeled “L”) is responsible for recruiting the E3 ubiquitin ligase RNF126. If the handover process fails, the TA is rapidly ubiquitinated for degradation via the ubiquitin proteasome pathway.

Transfer of TAs from Sgt2 to Get3 requires additional factors known as Get4 and Get5 in yeast, and their homologs TRC35 and UBL4A in mammals (Fig. 3A). Metazoans further require an additional factor known as Bag6 (Fig. 3B; Mariappan et al. 2010; Mock et al. 2015; Shao et al. 2017). The carboxyl terminus of Bag6 is sufficient to recruit UBL4A, SGTA, and TRC35, thereby coordinating handover to TRC40. A recent structure of this pretargeting complex found that interaction of TRC40 with TRC35 is critical to prime TRC40 for substrate binding, providing insight into the molecular details of substrate transfer to TRC40 (Keszei et al. 2021).

The amino terminus of Bag6 contains a ubiquitin-like (UBL) domain that recruits the E3 ubiquitin ligase RNF126 to ensure that TAs that fail to correctly target are directly triaged for degradation (Rodrigo-Brenni et al. 2014). In yeast, Get4 and Get5 interact directly with the Get3 loading complex, suggesting that the role of Bag6 reflects the higher stringency for preventing aggregation required in higher eukaryotes (Wang et al. 2010, 2011; Rome et al. 2013; Mock et al. 2015). It was recently suggested that ubiquitination of GET substrates can also occur through a Bag6-independent mechanism, and that ubiquitination alone does not commit a TA protein to degradation. Instead, ubiquitinated TAs can still be transferred to TRC40, where they are deubiquitinated by UPS20 and UPS33 at the ER membrane immediately prior to insertion (Culver and Mariappan 2021). The fate of ubiquitinated TA proteins and the role of triage factors beyond Bag6 in TA targeting and quality control remain an area for future research.

In contrast to the GET pathway, very little is known about how targeting and delivery to the EMC is coordinated or occurs with high specificity. The physical properties of the Get3 binding groove suggest that there is a strong preference for TMs of particularly high hydrophobicity. Indeed, biochemical studies showed that Get3 does not efficiently bind TAs below a certain hydrophobicity threshold (Wang et al. 2010; Rao et al. 2016; Guna et al. 2018). These Get3-independent substrates account for roughly half of all ER-targeted TA proteins, which we have now established rely on the EMC for insertion.

In mammals, TAs that fail to bind TRC40 (Get3) were shown to primarily bind the cytosolic chaperone calmodulin (CaM) (Guna et al. 2018). CaM is a very abundant, ubiquitously expressed protein known for its central role in calcium-mediated signal transduction pathways (Berchtold and Villalobo 2014). However, CaM also contains a flexible methionine-rich helical scaffold (O'Neil and DeGrado 1990), which can bind a broad range of peptide sequences, particularly those of modest hydrophobicity (Fig. 3B). At physiologic Ca²⁺ levels, CaM can dynamically engage TMs and signal sequences (Meador et al. 1992; Martoglio et al. 1997; Shao and Hegde 2011b), including TMs of TAs which have not been captured by Get3 (Guna et al. 2018). Despite CaM's ability to capture TA substrates, its ability to act as a targeting factor is less clear because, unlike Get3, it has no known ER-resident receptor. However, in a minimal system it was shown that substrate-loaded recombinant CaM was sufficient for TA insertion into the ER membrane in an EMC-dependent manner. This observation suggests that dynamic release from CaM at physiologic Ca²⁺ levels permits engagement with the EMC for insertion before CaM reassociation (Guna et al. 2018).

The implications of CaM being the singular targeting factor for the EMC pathway raise several important questions. First, it is not clear how CaM alone can achieve targeting specificity. If the only requirements for CaM binding are a low-hydrophobicity TM that cannot engage TRC40, CaM cannot differentiate between mitochondrial- versus ER-destined TA proteins (Fig. 1B). Binding to CaM would thus be sufficient for delivery, and presumably insertion, of many mitochondrial TA proteins into the ER by the EMC. It is possible the cell simply tolerates some level of constitutive mistargeting, relying on quality control machinery to recognize and degrade incorrectly inserted TAs (Chen et al. 2014b; Okreglak and Walter 2014; McKenna et al. 2020). Alternatively, there may be additional cofactors that facilitate the specificity of EMC targeting that remain to be identified. Second, there is no evidence that CaM can associate with either the ribosome, EMC, or any other ER-resident receptor, leaving no obvious explanation on how directionality is achieved. Finally, unlike Get3, targeting to the EMC appears to be energy independent. Therefore, how substrate binding and release is coordinated with arrival at the ER membrane remains unclear. Addressing these questions will require a systematic understanding of the factors required for EMC targeting and insertion.

TA HANDOVER AND CAPTURE BY ER INSERTASES

Following targeting, a TA substrate must be released and transferred from a cytosolic chaperone to the appropriate ER resident machinery for insertion into the lipid bilayer. For the GET pathway, this is an intricately coordinated process whose molecular details are well-understood (Hegde and Keenan 2011). The Get3-TA complex is initially recruited to the ER membrane by a short two-helix motif at the amino terminus of Get2's flexible cytosolic tail (Mariappan et al. 2011; Stefer et al. 2011; Wang et al. 2011). This interaction tethers the Get3 dimer at the membrane, where the cytosolic coiled-coil domain of Get1 can engage and drive apart Get3's two α-helical domains, resulting in substrate release. Somewhere prior to Get3's interaction with Get1, ATP must be hydrolyzed or dissociate, as Get1's coiled-coil protrudes into the ATP-binding site of Get3. In support of this model, Get3 hydrolysis mutants cannot effectively release substrates (Stefanovic and Hegde 2007). Subsequent binding of ATP to the substrate-free Get3 releases it from Get1 for recycling back to the cytosol and another round of substrate targeting.

Given the homology of EMC3 to Get1, it was tempting to speculate that the EMC would capture substrates in an analogous way to the Get1/2 complex. Because Get3 is anchored at the membrane by interaction with the coiled-coil domain of Get1 (Mariappan et al. 2011; Stefer et al. 2011), it was postulated that EMC2, anchored at the membrane by the coiled-coil domain of EMC3, would perform a similar role in directly binding substrate TMs.

However, the recently determined structures of the EMC challenged this model. Analysis of the human and yeast EMC structures failed to reveal an obvious hydrophobic groove capable of TM binding (Fig. 3C; Bai et al. 2020; Miller-Vedam et al. 2020; Pleiner et al. 2020). The crystal structure of the EMC2-9 heterodimer identified a hydrophobic cleft that was proposed to accommodate an unfolded TM (O'Donnell et al. 2020), and cross-linking experiments with EMC2-8 suggested that this subcomplex could transiently interact with substrate. However, in the context of a more complete reconstituted cytosolic domain (including EMC3, EMC5, and EMC8), EMC2 does not appreciably cross-link to TMs (Pleiner et al. 2021). Thus, although it remains possible that large-scale rearrangements could expose a TM binding site within EMC2, the simplest model is that EMC2 functions primarily as an architectural scaffold rather than a substrate interaction site (Miller-Vedam et al. 2020; Pleiner et al. 2020). Further, unlike Get3, there is no evidence that EMC2 shuttles between the cytosol and ER, and when unassembled, EMC2 is unstable and rapidly degraded (Volkmar et al. 2019; Pleiner et al. 2021).

One characteristic of the Oxa1 superfamily of insertases is the presence of methionine-rich cytosolic loops (Borowska et al. 2015). Although the cytosolic loops of EMC3 and EMC7 were disordered and thus likely dynamic in all of the structures of the EMC, the loops are all enriched for methionines, many of which are well conserved across eukaryotes (Fig. 3D). Mutation of methionines in the loops of EMC3 to the polar amino acid, serine, were shown to induce a specific biogenesis defect to EMC-dependent, but not -independent substrates (Pleiner et al. 2020). We favor a model in which these cytosolic loops form a flexible methionine-rich landing site to orient and shuttle TMs into the bilayer. Such a methionine-rich network may be a general feature of Oxa1 family insertases. Using this type of strategy for TM capture may be particularly well-suited for the biophysically diverse, but generally moderately hydrophobic, substrates of the EMC and will require future work to define its precise functional and molecular role in substrate capture.

INSERTION OF TAs INTO THE LIPID BILAYER

After release from a cytosolic chaperone, a TA protein must be inserted into the ER membrane in the correct topology: its amino terminus in the cytosol, its carboxyl terminus in the ER lumen, and its TM spanning the hydrophobic core of the lipid bilayer. The primary energetic barriers to this process result from (1) transfer of the hydrophobic TM across the polar headgroups of the phospholipid bilayer such that it can partition into the membrane; and (2) translocation of the short soluble carboxy-terminal domain across the hydrophobic core of the membrane. Structures of Get1/2 and the EMC reveal striking similarities in how the entire Oxa1 family of insertases catalyze TM insertion (Fig. 5; Bai et al. 2020; McDowell et al. 2020; Miller-Vedam et al. 2020; O'Donnell et al. 2020; Pleiner et al. 2020).

Figure 5.

Comparison of the Oxa1 superfamily insertases. (A) The conserved three-helix fold characteristic of all Oxa1 superfamily insertases is displayed for EMC3 (Pleiner et al. 2020) (6WW7), WRB (Get1) (McDowell et al. 2020) (6SO5), and YidC (Kumazaki et al. 2014) (3WO6) in the context of their larger scaffold (gray). Polar and charged residues that contribute to formation of a hydrophilic groove within the membrane are highlighted in blue. (B) Superposition of the Oxa1 superfamily insertase fold.

Most strikingly, both Get1/2 and the EMC rely on a hydrophilic vestibule or groove that connects the cytosol with the hydrophobic core of the bilayer for insertion. At the center of these vestibules are the Oxa1 superfamily members, Get1 and EMC3, which interact with either Get2 or EMC6 to form a conserved six-TM insertase motif (Fig. 5). Interestingly, both insertase systems require a TM with low insertion propensity that must be inserted concomitant with assembly (EMC6 and CAML) (Carvalho et al. 2019; Inglis et al. 2020; Pleiner et al. 2020). Much like YidC (Chen et al. 2014a; Kumazaki et al. 2014), EMC3 and Get1 (Wang et al. 2014) position conserved polar and positively charged residues within the plane of the membrane, which are critical for insertase function. Although their precise mechanistic role remains to be determined, a polar or charged patch within the bilayer may be important for providing a hydrogen bonding surface for the soluble carboxyl termini of TA proteins, which would decrease the energetic barrier for translocating their extracellular domains across the hydrophobic core of the membrane and could further partially suppress insertion of mitochondrial TA proteins, which tend to contain positively charged carboxyl termini. Ultimately, EMC and Get1/2 represent additional examples of the many membrane protein translocases that utilize a hydrophilic vestibule to provide a protected environment for transferring a substrate TM past the polar headgroups for partitioning into or out of the lipid bilayer (Fig. 6; Kumazaki et al. 2014; Voorhees and Hegde 2016; Schoebel et al. 2017; Wu et al. 2020).

Figure 6.

Model for transmembrane domain (TM) insertion by the EMC and WRB/CAML. (A) TM insertion by the EMC is catalyzed by a combination of (1) a flexible methionine-rich cytosolic loop for substrate capture (yellow); (2) a hydrophilic vestibule composed of EMC3, EMC4, and EMC6 that contains a conserved positively charged patch (blue); and (3) a pronounced local thinning of the membrane. The hydrophilic groove is dynamically gated by the flexible TMs of EMC4, EMC7, and EMC10. (B) Insertion by WRB/CAML occurs through a qualitatively similar mechanism and also relies on a hydrophilic groove for inserting its TA substrates into the ER membrane. The presence of methionine-rich cytosolic loops and a hydrophilic vestibule that connects the cytosol and the membrane and local thinning of the bilayer appear to be conserved features of the Oxa1 superfamily class of insertases.

For both Get1/2 and the EMC, this hydrophilic vestibule is partially protected from the lipid environment. In the case of Get1/2, the vestibule is enclosed axially by the lumenal helix of Get1, whereas lateral access may be dynamically controlled by recruitment of a second Get1/2 heterodimer (Fig. 6A; McDowell et al. 2020). This observation raises additional questions about the precise role of the Get1/2 heterotetramer in TA insertion, including defining the path of a TA substrate into the bilayer given that a single Get1/2 heterodimer alone is sufficient for insertion (Zalisko et al. 2017).

In the case of the EMC, its hydrophilic vestibule is capped by the amphipathic lumenal helix of EMC3 and by a flexible “gate” formed by the three TMs of EMC4 and the single TMs of EMC7 and EMC10. The positioning of this gate relative to the hydrophilic vestibule formed by EMC3/6 appears to be highly variable, suggesting it may undergo dynamic rearrangement upon substrate engagement (Miller-Vedam et al. 2020). Although neither EMC4 nor EMC7 are core architectural subunits of the EMC (Volkmar et al. 2019), depletion experiments suggest they are functionally essential for TA insertion (Luo et al. 2002; Shurtleff et al. 2018; Lakshminarayan et al. 2020). Thus, their role in dynamically gating the EMC, potentially regulating access of substrates to the hydrophilic vestibule, and/or undergoing conformational changes upon substrate binding appears to be required for EMC function. Future work will be necessary to definitively establish the molecular role of the TMs of EMC4 and EMC7 in TM insertion.

Despite the marked advance the recent structures of the EMC represent, the precise path of a substrate TM into the bilayer via the EMC, and indeed Get1/2, is incompletely defined. Earlier cross-linking experiments established that TA substrates physically interact with the hydrophilic vestibule of YidC (Kumazaki et al. 2014) and Get1/2 (Wang et al. 2014). Further, EMC3 was shown to be the major interaction partner of TA substrates with the EMC (Pleiner et al. 2020). Thus, based on homology to YidC, and the striking evolutionary conservation of the polar and charged residues, it has been proposed that the hydrophilic vestibule of EMC represents the major path into the bilayer (Bai et al. 2020; Miller-Vedam et al. 2020; Pleiner et al. 2020). Indeed, the localization of the polar and charged residues of EMC3 and Get1 within the cytosolic leaflet of the bilayer create a hydrophobicity gradient that could drive insertion (Miller-Vedam et al. 2020), through a qualitatively similar mechanism to all Oxa1 superfamily insertases. In the case of the EMC, a putative role for the opposite intramembrane surface of the EMC in insertion has not been excluded (O'Donnell et al. 2020). However, the primary function of this hydrophobic surface, only accessible from within the bilayer, may instead more directly contribute to the EMC's function as an intramembrane chaperone (Miller-Vedam et al. 2020; Pleiner et al. 2020). Further, there has been speculation that TA substrates would bind the EMC between the TMs of EMC3 and EMC4 on their path into the bilayer (Bai et al. 2020). However, the TMs of EMC7 and EMC10 are likely also localized to this region (O'Donnell et al. 2020), and because of their dynamic nature, there is no consensus in their exact positioning. Systematic cross-linking and direct structural characterization of substrate binding will thus be critical to definitively determine the path of a TM into the membrane via the EMC.

Finally, structures of the EMC revealed a pronounced local thinning of the membrane, likely driven by the TMs of EMC3/6 with contributions from the amphipathic lumenal helices of EMC1, EMC3, and EMC7, which are embedded within the lumenal leaflet of the bilayer (Miller-Vedam et al. 2020; Pleiner et al. 2020). Similarly, the TMs surrounding the hydrophilic groove of Get1/2 are also markedly shorter and adopt a concave arrangement that could induce local membrane curvature (McDowell et al. 2020). It is thus likely that both the EMC and Get1/2 may induce local membrane thinning in a manner similar to many evolutionarily diverse protein translocases (Mitra et al. 2004) including Hrd1 (Wu et al. 2020), TatC (Rodriguez et al. 2013), and YidC (Chen et al. 2017). Decreasing the width of the hydrophobic core of the bilayer would facilitate translocation of a TA proteins’ soluble carboxyl terminus across the membrane into the ER lumen, decreasing the energetic barrier to insertion (Wu and Rapoport 2021). Once within the membrane, the thinned bilayer cannot accommodate a membrane-spanning TM, favoring its dissociation from the complex into the bilayer.

CONCLUDING REMARKS

Several outstanding questions remain in our understanding of TA biogenesis at the ER. First, it is still unclear how robust specificity to a particular targeting pathway, and therefore organelle, is established. The biophysical similarities of ER and mitochondrial and peroxisomal TAs makes it difficult to discriminate between substrates with high fidelity. This raises the possibility that the cell tolerates a certain degree of mistargeting, which is addressed by quality control machinery that recognizes and degrades failed or mislocalized TA proteins. Although coupling between the ubiquitin proteasome system and the cytosolic GET components is more clearly defined, related processes within the membrane for Get1/2 and the entire EMC pathway remain poorly understood. Further, it is not clear if either insertase possesses proofreading activity, akin to that of Sec61 (Voorhees and Hegde 2016), or if the primary driver of TA specificity remains chaperone binding alone. Addressing these questions will require detailed biochemical dissection of the Get1/2 and EMC insertion mechanism.

Second, we likely have an incomplete picture of the factors involved in the EMC targeting pathway. Although the lower hydrophobicity of EMC substrates may not necessitate the stringent coordination of the GET pathway, it is still difficult to envision how the EMC can efficiently insert substrates released stochastically in any orientation by CaM or SGTA. Two non–mutually exclusive possibilities are that additional cofactors are involved or that the cytosolic domains of the EMC play a greater role in capturing, orienting, and releasing substrates at the membrane. Systematic analysis of the EMC pathway factors will be needed to close these crucial gaps in our understanding of TA protein biogenesis.

Third, direct visualization of a substrate-engaged complex of any post-translational insertase has remained elusive, in part because of the technical challenges of trapping this normally transient intermediate. These structures would be particularly informative in understanding the molecular mechanism of how the EMC and Get1/2 capture and insert TMs. Such a structure may further clarify the role of a Get1/2 heterotetramer in insertion and definitively establish the path of a TA substrate through the bilayer.

Finally, beyond its role in TA insertion, several aspects of EMC function remain particularly enigmatic. For example, unlike other Oxa1 superfamily insertases, the EMC contains a large, conserved lumenal domain that has no clear function. The role of this domain, potentially as a platform for recruitment of auxiliary factors, remains an important area for further research. Further, the EMC has been recently shown to cotranslationally insert multipass membrane proteins that contain an N_exo TM (whose amino termini are localized to the extracellular environment; Chitwood et al. 2018). This raises the questions of how nascent multipass substrates are targeted and delivered to the EMC, and how the EMC cooperates with Sec61 for their insertion. Future work will be critical to define the most salient features of EMC's role in protein biosynthesis at the ER membrane.