A distinctive feature of eukaryotic cells is the presence of an endomembrane system enclosing various intracellular compartments. Transport between these compartments is mediated by coat protein complexes such as COPI, COPII, and clathrin (1). Coat protein complex II (COPII) is involved in transporting proteins from the endoplasmic reticulum (ER) to the Golgi apparatus. The basic functional units of the COPII coat are the proteins Sar1, Sec23/24, and Sec13/31 (2). X-ray crystal structures of each of these proteins have been solved, and cryogenic electron microscopy (EM) structures have been solved of assembled COPII coats and cages. Here we will describe structural features of the COPII coat protein complex and discuss how these characteristics give rise to their functional mechanisms.
In eukaryotes, coat protein complex II (COPII) proteins are involved in transporting cargo proteins from the endoplasmic reticulum (ER) to the Golgi apparatus. The COPII proteins, Sar1, Sec23/24, and Sec13/31 polymerize into a coat that gathers cargo proteins into a coated vesicle. Structures have been recently solved of individual COPII proteins, COPII proteins in complex with cargo, and higher-order COPII coat assemblies. In this review, we will summarize the latest developments in COPII structure and discuss how these structures shed light on the functional mechanisms of the COPII coat.
The Structure of Sar1
Sar1 is involved in regulating the formation of COPII vesicles at the ER. Sar1 is a GTPase in the Ras superfamily of GTPases and, as such, functions as molecular switch controlled by the exchange of GDP and GTP. Sar1 is cytosolic in its GDP form and is recruited to the ER by the exchange of GDP for GTP. The GTP exchange factor (GEF) for Sar1 is Sec12, which is a membrane protein that is localized to the ER. Thus, the localization and activity of Sec12 promotes the localization of Sar1 to the ER. Upon exchanging GDP to GTP, Sar1 exposes an amphipathic α-helix which becomes buried in the ER membrane. Sar1•GTP then recruits Sec23/24 which in turn recruits Sec13/31, and these together promote the formation of a COPII-coated vesicle.
The structure of Sar1 has been characterized by X-ray crystallography (3–5) (Figure 1, Table 1). This revealed Sar1 to have four distinct domains: a guanine nucleotide binding pocket, switch I, switch II, and N-terminal amphipathic α-helix each of which play a specific role in Sar1 function (Figure 1). The core of the protein is comprised of six β sheets (five parallel [β6,5,4,1,3] and one antiparallel [β2]) that form the guanine nucleotide-binding pocket (the G domain) typical of the Ras superfamily with a characteristic highly conserved nucleotide-binding motif GxxxxGKT39 (5). The G domain is essential for Sar1 function; mutation of Thr39 to Asn creates a GDP restricted dominant negative mutant by interfering with the association to its guanine nucleotide exchange factor, Sec12 (6). The switch I domain mediates the interactions between the guanine nucleotide and Mg2+. Switch I encompasses residues 44–57 which are located close to the guanine nucleotide (Figure 1, red). Mg2+ is coordinated by the β phosphate of GDP and the hydroxyl of a conserved Thr54 residue in switch I (3). Switch II (residues 76–89) plays an important role in the hydrolysis of GTP (Figure 1, blue). In the structure of Sar1 bound to GppNHp, Mg2+ is coordinated to the γ phosphate position and interacts with the backbone of Gly76 (4). If His79 of the switch II region is mutated, so that it loses its ablility to H-bond with Asp34, Asp34 swings away from His79 and Sar1 loses the ability to hydrolyze GTP thus inhibiting the disassembly of the COPII complex (5). The N-terminal amphipathic α-helix (α1′) is essential for anchoring Sar1 to the lipid bilayer (5,7,8). Sar1 in the GDP form keeps the α1′ helix retracted into a pocket formed by the β2–β3 hairpin linker region (Figure 1, orange). In the GDP bound conformation, a conserved Asp upstream from switch II mimics the charges of the γ phosphate in GTP. Exchange of GDP for GTP causes a two residue shift that pulls switches I and II up to the active conformation (Figure 1B). This causes a 7 Å shift of the linker (Figure 1) eliminating the binding pocket for α1′ and causing it to shift into a conformation that exposes the hydrophobic region which facilitates binding to the ER membrane (9).
|Protein complexes||PDB ID/EMDB ID||References|
|Sar1H79G • GDP||2FA9||1|
|Sar1 • GDP, low Mg2+||2FMX||1|
|Sec23 • Sar1 • GppNHp||1M2O||2|
|Sar1 • GDP||1F6B||3|
|Sec31(fragment) • Sec23 • Sar1 •|
|Sec23 • Sar1 • GppNHp||1M2O||2|
|Sec23a/Sec24d • syntaxin 5 IxM|
|exit code peptide||3EFO||5|
|Sec23a/Sec24d • membrin IxM|
|exit code peptide||3EG9||5|
|Sec24 • Bet1 LxxLE exit code|
|Sec24 • Sed5 YNNSNPF exit code|
|Sec24 • Sys1 DxE exit code peptide||1PD1||6|
|Sec23/24 • Sec22||2NUT||7|
|Sec31 (residues 370-763) • Sec13||2PM6||8|
|Sec31 (residues 1-411) • Sec13||2PM9||8|
|COPII coat and cage|
|Sec13/31 • Sec23/24 icosidodecahedron||1511||10|
In addition to tethering Sar1 to the ER membrane, α1′ may be involved in membrane deformation and vesicle fission (10). Upon membrane binding, the hydrophilic face of the Sar1 amphipathic helix lies close to the phospholipids and the hydrophobic face buries in the membrane interior disrupting the lipid interactions and causing a deformation of the membrane. Addition of excess Sar1 to semi-permeabilized cells induces the formation of tubules from the ER, and likewise, addition of Sar1 to liposomes in vitro without Sec23/24 or Sec13/31 causes the liposomes to be extruded into long tubules (10–12). Mutation of the hydrophobic residues of the α1′ were shown to reduce the tubulation activity (10). In addition, it was also observed that these mutations lead to a reduction in the formation of free vesicles in an in vitro budding assay, despite the fact that Sec23/24 and Sec13/31 were recruited to the membrane (10). Given these data, it has been proposed that Sar1 may have a role in both the membrane deformation and vesicle fission (10,13). In the early stages of vesicle formation Sar1 recruits Sec23/24 and together these may create a local region of curvature that is later organized into a more global curvature by Sec13/31 (10,13,14). After a vesicle is formed but has not yet budded from the ER, Sar1 may be involved in the fission of the vesicle from the donor membrane, and this role would be analogous to dynamin in the clathrin-coated vesicle system (15).
The Structure of Sec23/24
The Sec23/24 heterodimer is responsible for binding cargo and serving as the Sar1 GTPase activating protein (GAP) (16). As such, it plays a central role in the organization of the COPII coat. The importance of this heterodimer is evident in the fact that human diseases cranio-lenticulo-sutural dysplasia, and congenital dyserythropoietic anemia type II result from mutations in Sec23 (17,18). Sec23/24 is recruited to the ER membrane by Sar1. Crystal structures of the Sec23/24 heterodimer revealed it to have a slightly curved surface that presumably facilitates interaction with and curvature of the ER membrane. The heterodimer of Sec23/24 has a general bow tie shape as revealed first by Lederkremer et al. (19). The dimer is 150 Å long and 100 Å wide at the widest point narrowing down to 30 Å wide at the ‘knot’, where Sec23 and Sec24 come together. It is 40 Å tall (away from the membrane) except for a protrusion on each protein that extend up to 60 Å. The Sec23/24 dimer has a concave inner face that closely matches the arc of the 60 nm cages, likely favoring deformation of the membrane (4). Moreover, binding Sar1•GTP causes a subtle rotation around trunk domains increasing the curvature of Sec23/24 by 1.5°(4). The concave inner face (membrane facing) is generally positively charged enforcing the requirement for acidic phospholipid for Sec23/24-Sar1 binding (20), while the outside is generally more acidic. A single Sec23/24 complex would cover ∼8200 Å2 roughly 0.7% of the surface area of a 60 nm vesicle.
The individual structures of Sec23 and Sec24 are very similar even though they share only 14% sequence identity (Figure 2A). Both Sec23 and Sec24 fold into five distinct domains: β barrel, zinc finger, α-helical region, trunk domain, and carboxyl-terminal domain (4) (Figure 2B). The β barrel, 180 residues long, lies roughly parallel to the membrane. The zinc finger is 55 residues and is situated next to the β barrel. The Zn cluster is situated towards the membrane contributing to the basic inner surface of the protein. The trunk domain is located between β1 and β19 of the β barrel, and presumably lies along the lipid membrane. The trunk domain plays an important role in the binding of Sar1 and is involved in the interaction of Sec23 with Sec24. The two proteins dimerize through contacts between β14 and the β14–β15 loop of their respective trunk domains. Furthermore, Phe385 and Pro387 form van der Waals interactions with conserved residues (181–183) of Sec23. This presumably allows multiple isoforms of Sec24 to form dimers with Sec23.
The GTPase activity of Sar1 is activated by the carboxyl-terminal gelsolin-like domain of Sec23. Sec23 identifies Sar1•GTP through interactions between switches I and II and the linker region of Sar1 and the helical and trunk domains of Sec23 (Figure 2C,D). Sec23 activates the Sar1 GTPase by inserting an arginine side chain into the active site of Sar1 neutralizing negative charges of the phosphate; this mechanism is referred to as an arginine finger and is a feature of many Ras family proteins (21). In addition to the arginine finger, the active fragment of Sec23 inserts Trp922 and Asn923 with the indole ring of Trp922 oriented close to parallel with the imidazole ring of His77 of Sar1. His77 bonds to a nucleophilic water molecule and is solvent accessible on one side of the His77 indole ring (4).
The lifetime of GTP bound to Sar1 is around 30 seconds when Sar1 is bound to Sec23, and the hydrolysis is accelerated by an additional order of magnitude by the binding of Sec13/31 (22). A crystal structure of Sec23 and Sar1 together with a fragment of Sec31 (Figure 2C,D) show that this enhancement is because of conformational changes in Sec23 and Sar1 induced by the binding of Sec31 (23). In the Sec23 Sar1•GTP complex Gln720 of Sec23 is oriented away from the active site. Upon binding of Sec31, the Gln720 swings towards the active site and forms hydrogen bonds with Trp922 and Asn923. These new bonds improve the orientation with His77 of Sar1 and shield the solvent accessible side of the H77 imidazole ring. This indicates that Sec31 plays a structural role in improving the rate of hydrolysis and does not provide any catalytic activity itself (23).
Structures of Sec23/24 in Complex with Cargo
Sec24 functions as the primary cargo adaptor element of the COPII coat. Initial work on yeast Sec24 identified three separate binding sites (A, B, and C) that controlled which cargo was selected (24,25). The presence of multiple binding sites allows for diversity in the number of exit codes that Sec24 can bind, and multiple Sec24 isoforms expand the range and specificity of Sec24/cargo interactions. Here we will discuss the structural interactions that facilitate the binding of cargo by Sec24. In order to simplify the naming, for the purposes of this review, we will simply refer to the binding sites on Sec24 by the exit code sequences they recognize.
One of the first exit codes to be determined is the sequence DxE in the vesicular stomatitus virus G protein (26). The DxE exit code is also found in the SNARE protein Sys 1. Crystallographic analysis of a Sys 1 peptide containing the DxE sequence has localized the binding site on Sec24 to a groove on the membrane proximal surface of Sec23/24 (24) (Figure 2D, purple ribbon). In mammals, the DxE sequence is recognized by Sec24 isoforms a and b which are 75% identical in sequence, respectively. The same binding site that binds the DxE sequence also recognizes the sequence LxxL/ME (24) (Figure 2D, purple ribbon). This binding pocket is modified in Sec24 isoforms c and d, where Leu582 is substituted to aspartic acid which prevents binding of the DxE and LxxL/ME sequence.
An exit code that is found in the SNAREs membrin and syntaxin is IxM. This sequence is recognized by the Sec24 c and d isoforms, where a binding pocket is formed on the distal side of Sec24 in the helical domain (27) (Figure 2D, blue ribbon). This binding site is obstructed in Sec24a and Sec24b due to a short β strand on the N-terminus that runs across the pocket and inhibits the binding of the IxM sequence.
The binding site for the SNARE Sec22 is unusual because it does not seem to recognize a specific sequence but rather a conformational epitope (28) (Figure 3C,D, yellow). Sec22 has ‘open’ and ‘closed’ conformations and only binds its cognate partner SNARES in its open conformation and only binds to Sec23/24 when it is closed. This indicates that COPII will only transport Sec22 when it is not interacting with other SNAREs, and this could serve as a timing device for SNARE transport. Sec22 binds specifically to Sec24a and Sec24b isoforms but is excluded from Sec24c and Sec24d. Furthermore the binding site for Sec22 is shared between Sec23 and Sec24. To date this is the only COPII cargo identified to require binding to both Sec23 and Sec24.
A site recognizing the sequence YNNSNPF has been characterized in yeast but that site appears to be unique to yeast SNARE Sed5 (24) (Figure 2D, orange ribbon). The mammalian homolog to Sed5, syntaxin 5, does not contain the YNNSNPF sequence and instead appears to bind to the IxM site on Sec24c and Sec24d (27). In yeast Sed5, the YNNSNPF sequence is only exposed in the open conformation for this SNARE, and this conformation is promoted by the formation of the tSNARE complex (i.e. when it is bound to Bos1 and Sec22). It is unclear however, how this relates to the Sed5 homolog syntaxin 5 in mammals since the YNNSNPF sequence is only found in Sed5.
The Structure of Sec13/31
Two complementary structures were recently solved that elucidated the structure of the outermost layer of the COPII coat: what we refer to as the COPII cage. A crystallographic study determined the atomic structures of Sec13 and Sec31 and showed how they are organized into a heterotetramer, and a cryoEM study showed that Sec13/31 can self-assemble into a cage with the potential to form a variety of structures that can accommodate cargos of widely varying shapes and sizes (14,29,30). Together these studies revealed the structural rules that dictate the formation of the COPII cage.
Sec13 and Sec31 have long been known to associate with each other (31), and evidence supported the idea that they form a heterotetramer with two copies each of Sec13 and Sec31 (19). Fath and colleagues (29) determined two separate structures that when combined revealed the complete structure of the Sec13/31 heterotetramer. The first structure consisted of the N-terminal domain of Sec31 bound together with Sec13. The other structure consisted of Sec13 bound to the C-terminal sequence of Sec31. They used Sec13 as a common alignment region and put the two structures together to form the complete Sec13/31 heterotetramer (Figure 3A).
The Sec13/31 structure is composed primarily of WD40 domains and α-solenoid motifs. WD40 domains consist of a repeating β-propeller motif with the complete domain typically being formed of seven β-propellers (Figure 3B). Sec13 is composed almost entirely of a WD40 domain, and Sec31 has a WD40 domain at its N-terminus with the rest of the protein being comprised primarily of α-solenoid repeats and a small unresolved proline-rich region. The WD40 domain of Sec13 is unusual in that it contains only six blades instead of the usual seven. Remarkably, Sec31 interacts with Sec13 by donating a seventh β-propeller blade into the WD40 domain of Sec13 (Figure 3B). Another overlapping motif is observed in the Sec13/31 structure in the way that Sec31 forms a dimer with another Sec31. They form an interlocking and overlapping fold, where the C-termini of adjacent Sec31s in the heterotetramer loop back over each other, such that their adjacent α-solenoid motifs pack against each other and form a continuous α-solenoid bundle via interprotein contacts (Figure 3C). Altogether, the protein/protein interactions in the Sec13/31 heterotetramer are characterized by overlapping and interlocking folds. One is tempted to speculate that the degree of overlap may be important for stabilizing the heterotetramer in the cell so that the COPII coat can withstand the strains that may be encountered in the cell during vesiculation.
The Structure of the COPII Cage
Sec13/31 self-assembles into a cage-like structure in vitro, and these structures were solved by cryoEM and single particle reconstruction. This revealed the cage to be a 600 Å diameter cuboctahedron (30) (Figure 4A). The diameter of the in vitro assembled cages is quite similar to that of COPII-coated vesicles which range from 500 to 900 Å(20). The cage structure exhibited a unique design for cage complexes with four edges combining to form the vertices of the cages as opposed to three edges like clathrin. The heterotetramer architecture of Sec13/31 could be observed in the edges of the COPII cage structure; the edge structure has two roughly spherical domains at one end, and these are related to two other spherical domains by a 180° rotation around a continuous curving region of density connecting the two (30). In the COPII cage, the Sec13/31 heterotetramer is positioned off-center with respect to the cuboctahedral vertices; that is, one end of the heterotetramer is closer to the vertex than the other end giving rise to a polarity in the edges. The end that is closer to the vertex is referred to as the plus end and the other end is the minus end (Figure 4B). The cages appear to be held together by relatively small-area protein–protein interactions at the vertices, and these interactions contrast greatly with the extensive overlapping arms of clathrin triskelia in the clathrin cage (32). The vertices in the COPII cage are held together by interactions between adjacent Sec13 and Sec31 WD40 domains (29,30) (Figure 4B). In contrast, clathrin cages exhibit extensively overlapping triskelia where every individual triskelion leg interacts with three different cage edges and vertices (32,33).
The specific interactions mediating the formation of the COPII cage were revealed by combining the crystal structure of Sec13/31 with the EM structure of the COPII cage. The Sec13/31 edge element, as it was crystallized, did not fit into the EM density. The EM density has a 135° bend in the middle of the edge while the crystal structure is much closer to straight with a bend angle of 165°. A curve could be modeled into the crystal structure, and after doing this, it fits very well into the contours of the EM structure (14,29). Analysis of the crystal structure fit inside the EM density showed that the vertices of the COPII cage are formed by interactions between adjacent Sec13 and Sec31 WD40 domains. Four putative contact interfaces have been identified that occur between WD40 domains to facilitate the formation of the cage. Contact I (cI) occurs between the WD40 domains of two adjacent Sec31 plus ends. Contacts II and III (cII and cIII) occur between WD40 domains of adjacent Sec31s at one plus end and one minus end, and contact IV (cIV) occurs between WD40s of Sec13 at the plus end and the minus end of Sec31 (14,29) (Figure 4B). Thus it appears that interactions between WD40 domains facilitate the assembly of the COPII cage.
A larger COPII complex was observed when COPII cages were assembled together with Sec23/24 (14) (Figure 4C). These structures were solved by cryoEM and single particle reconstruction, and this revealed a 1000 Å diameter two-layered COPII assembly. Sec13/31 was observed to form the outer layer with Sec23/24 decorating the Sec13/31 cage beneath the vertices. The outer cage structure showed many similarities to the cuboctahedral COPII cage assembled from Sec13/31 alone; the vertices are formed from the intersection of four edges, and the Sec13/31 heterotetramers are off-center with respect to the vertices. Even though these features are the same, the cage is quite a bit larger in diameter (1000 Å versus 600 Å in the cuboctahedron structure) and forms with a icosidodecahedral geometry (Figure 4C). The cage accommodates this larger size by changing one of the two angles between edges at the vertices. The angles at the vertices are identified as α and β, where α is the angle between a plus end and a minus end in a clockwise manner and β is the angle between a minus end and a plus end in a clockwise manner (Figure 4B,D,E ). α appears to be fixed at 90°, so this means that cages of different sizes are generated by varying the β angle. A demonstration of this is seen in the transition from the cuboctahedral cage to the icosidodecahedral cage. The 600 Å diameter cuboctahedral cage has β angles of 90° and β expands to 108° in the 1000 Å diameter icosidodecahedral cage. It is not known what dictates the formation of particular β angles, but Sec23/24 is located right beneath the cage vertices (Figure 4C,E ). It is speculated that Sec23/24 may be involved in influencing the β angle in response to cargo of different shapes and sizes (14).
The positioning of Sec23/24 with respect to Sec13/31 also has relevance for the activation of the Sar1 GTPase. In the icosidodecahedral COPII coat structure, a cluster of four Sec23/24 dimers lies right beneath the Sec13/31 vertices. The resolution of the EM structure was too low to unambiguously determine the positioning of Sec23/24 with respect to Sec13/31. Nonetheless, the approximate locations of Sec23/24 indicate that it takes on two different orientations relative to Sec13/31. One of the Sec23/24s in the asymmetric unit of the cluster is much closer to the proline-rich region of Sec31 that increases the Sar1 GAP activity than the other Sec23/24. This positioning suggests that only one of the two Sec23/24s in the asymmetric unit interacts with Sec31 and the other interacts with its neighboring Sec23/24. If this is the case, only one of the Sec23/24 dimers will have its GAP activity stimulated by Sec31, and thus one Sar1 will have a short resident time and the other will be longer lived. This idea dovetails nicely with a study that showed that in the absence of Sec13/31, Sar1 GTP hydrolysis precedes complete Sec23/24 dissociation from membrane and cargo (34). The authors of this study suggested that multiple rounds of Sar1 GTP hydrolysis may serve to ‘proofread’ cargo before incorporation into a vesicle (34,35). In the model where Sar1 can have two residence times on a coated vesicle, the Sar1 molecules with a short resident time would presumable fill the proofreading role. Further experiments including higher resolution cryoEM reconstructions and reconstructions in the presence of Sar1 are required to test this model.
Transport of Large Cargoes
The variability of β angles has relevance to the transport of large or unusually shaped cargos such as procollagen and chylomicron particles. Procollagen is 3000 Å in length (36) while chylomicron particles are greater than 4000 Å in diameter both of which are too large to fit in the 1000 Å COPII cage (37). Indeed, early evidence suggested that chylomicron particles and procollagen were transported by some other means than COPII vesicles (37,38), but more recent genetic and biochemical evidence support a role for COPII in transporting these unusual cargos. Mutations in Sar1B have been shown to cause chylomicron retention disease (39,40), which is characterized by buildup of chylomicron particles in the ER of enterocytes. The failure of chylomicrons to exit the ER because of a Sar1 mutation strongly implicates COPII in their secretion. COPII also appears to be involved in the secretion of collagen. Depletion of Sec13/31 leads to the inhibition of collagen secretion in fibroblasts and causes skeletal abnormalities in zebrafish (41). These abnormalities are phenotypically similar to developmental defects observed in cranio-lenticulo-sutural dysplasia, a human disease that derives from a F382L substitution in Sec23A (17,42). Finally Saito et al. (43) have recently identified that the integral membrane protein TANGO1 serves as an ER lumenal receptor that is involved in loading procollagen into COPII-coated vesicles. These data support a role for the COPII coat in the secretion of large cargos, but the question still remains if COPII is taking on a structural role, or if it simply serves to organize ER membrane domains that produce some other carrier besides a typical COPII-coated vesicle (38). If the COPII cage is to accommodate these large procollagen or chylomicron particles, which are bigger than the 1000 Å diameter COPII coat structure, then the coat must be able to expand to accommodate them. In other words, the β angle must be able to take on angles larger than 108°. If β can accommodate angles up to 120°, then any number of structures could be imagined from planar polygonal sheets to helical tubes. The fact that β can take on different angles suggests that the COPII coat has the capacity to accommodate large cargo, but such structures have not yet been observed in vitro.
Conclusions and Future Perspectives
The last decade has seen much progress toward a structural understanding of the mechanisms employed by the COPII coat, but we have only scratched the surface for these complex and dynamic assemblies. Many questions still remain such as what are the molecular mechanisms by which the coat assembles and disassembles? How is vesicle fission catalyzed? What is the structure of a complete COPII-coated vesicle in complex with cargo? Answering these questions will require a combination of traditional biochemistry and crystallography with cutting edge biophysical and structural techniques.
This work was supported by AHA grant #0835300N.