COPI Budding within the Golgi Stack

Role in Mitosis and Golgi Positioning

In addition to its role in interphase, coatomer has an active role during mitosis. COPI activity, concomitantly with repression of COPII vesicle formation (Farmaki et al. 1999), strongly contributes to the fragmentation of the Golgi apparatus into vesicles (Misteli and Warren 1994; Tang et al. 2008). In addition, recruitment of coatomer by a nucleoporin may induce the breakdown of the nuclear envelope (Liu et al. 2003).

An interaction of coatomer with cdc42 (Wu et al. 2000) and dynein was attributed to positioning of the Golgi (Chen et al. 2005; Hehnly et al. 2010).

Role in the Endocytic Pathway?

The early secretory and the endocytic pathways are similar and mirror each other as newly synthesized protein from the ER and proteins of the plasma membrane follow a similar trafficking scheme. Both protein populations are first transferred to a sorting station, the Golgi apparatus or the early endosomes, before being either recycled back to their starting compartment (ER or plasma membrane), or further transported along the late secretory pathway or to the late endosomal or lysosomal compartments. Pools of coatomer have also been identified at endosomal membranes (Whitney et al. 1995; Aniento et al. 1996; Gu and Gruenberg 2000). In this context it is of note that coatomer has been implicated to take part in the maturation of early endosomes (Whitney et al. 1995; Aniento et al. 1996; Daro et al. 1997; Gu et al. 1997; Gu and Gruenberg 2000; Gabriely et al. 2007), and/or recycling toward the plasma membrane (Daro et al. 1997; Razi et al. 2009). Moreover, coatomer may participate in the maturation of specialized endosomes such as phagosomes (Botelho et al. 2000; Beron et al. 2001; Hackam et al. 2001), autophagosomes (Razi et al. 2009), and peroxisomes (Lay et al. 2006). It is, however, not known if in these endosomal functions coatomer directly serves as a coat, or if these processes directly depend on the service of COPI-coated vesicles. Furthermore, recruitment of coatomer to endosomes, in contrast to Golgi binding, was reported to be pH-sensitive (Aniento et al. 1996), and to require only a subset of coatomer subunits (Whitney et al. 1995; Aniento et al. 1996), suggesting differences in mechanisms underlying the function of the complex in the endocytic and in the early secretory pathways.

Coat Recruitment

The small Ras-like GTPase ADP-ribosylation factor 1 (Arf1) plays a central role in the formation of COPI-coated vesicles at the Golgi apparatus. However, action of Arf1 is not limited to this organelle and can function at various intracellular compartments, involving different sets of effectors (Nie et al. 2003; Donaldson et al. 2005; Volpicelli-Daley et al. 2005). It is thus very likely that additional factors are required to provide a spatio-temporal regulation of coatomer recruitment at specific sites suitable for COPI vesicle biogenesis. Like most of the small GTPases, Arf1 cycles between a GDP-loaded inactive cytosolic state and an active, membrane anchored, GTP-loaded state (Randazzo et al. 1993). This molecular switch is positively regulated by guanine nucleotide exchange factors (GEFs), which catalyze exchange of GDP with GTP. On activation of an intrinsic low GTPase activity by Arf GTPase activating proteins (ArfGAPs), Arf1 is converted into its inactive GDP form and released from the membrane.

Specific recruitment of the small GTPase to the membrane is a prerequisite for its function. In agreement with its role in COPI vesicle formation, Arf1 contains Golgi localization signals encoded in its sequence. Arf1-GDP binds to a dimeric complex of members of the p24 family, a group of type I transmembrane proteins. In vitro cross-linking experiments reveal a specific interaction between a carboxy-terminal site of Arf1 with a dimer of p23 or p24 (Gommel et al. 2001). Foerster resonance energy transfer (FRET) experiments confirmed the existence of this interaction in the living cell (Majoul et al. 2001). In addition, at the cis-Golgi, Arf1 is recruited to the membrane via membrin, a SNARE protein. This interaction is mediated by a MXXE motif within the GTPase. Mutation of this motif impaired recruitment of Arf1 to the cis-Golgi without affecting its binding to the trans-Golgi (Honda et al. 2005). Thus, other signals within Arf1 must mediate targeting of the GTPase to the trans-Golgi.

Arf1 is activated on membranes by large ArfGEFs that are members of the Sec7 super family. Based on sequence homologies, mammalian ArfGEFs can be classified into the following groups: (1) Golgi Brefeldin A (BFA)-resistance factor 1/BFA-inhibited GEF (GBF/BIG), (2) Arf nucleotide binding site opener (ARNO)/cytohesins, (3) exchange factor for Arf6 (EFA6), (4) BFA-resistant Arf GEF (BRAG), and (5) F-box only protein 8 (FBX8) (Casanova 2007).

GBF1 and the BIGs, as well as their yeast homologs (Gea1/2 and Sec7p, respectively) activate Arf1 at the Golgi membrane (Claude et al. 1999; Spang et al. 2001; Kawamoto et al. 2002; Zhao et al. 2002), whereas the other ArfGEFs act at post-Golgi compartments. The mammalian Golgi GEFs are not evenly distributed across the stack of cisternae. BIG1 and BIG2 are present at the trans-Golgi, the TGN, and the recycling endosomes (Mansour et al. 1999; Shinotsuka et al. 2002; Shin et al. 2004). Their mode of recruitment to the membrane is not well understood. GBF1 localizes to IC and Golgi, acting as a direct mediator of retrograde transport between these compartments and the ER (Kawamoto et al. 2002; Garcia-Mata et al. 2003; Zhao et al. 2006). This localization is directly dependent on an interaction between the GEF and Rab1b (Monetta et al. 2007). Recruitment of GBF1 also requires PI4P, and it was proposed that Rab1 contributes to GBF1 recruitment by locally activating phosphatidylinositol 4-kinase (PI4KIIIalpha) (Dumaresq-Doiron et al. 2010).

Once both Arf1 and its exchange factor are recruited to the membrane, a Sec7 domain within the ArfGEF triggers the activation of the small GTPase. A critical feature of this catalytic domain is its “glutamic finger,” a conserved glutamate residue exposed at the tip of a hydrophilic loop between helices 6 and 7 (Beraud-Dufour et al. 1998). Nucleotide exchange is mediated by electrostatic competition of this glutamate side chain with the nucleotide’s β–phosphate. On exchange of GDP to GTP, a conformational change within Arf1 leads to exposure of its amino-terminal amphipathic and myristoylated helix, which in turn causes insertion of this helix into the lipid bilayer and thereby secures membrane anchorage of Arf1 (Franco et al. 1996; Antonny et al. 1997). This model was challenged by recent structural data. Nuclear magnetic resonance spectrometry of full-length myristoylated Arf1 embedded into bicelles showed an unexpected localization of the myristic acid residue. Instead of being inserted within the bilayer parallel to the phospholipid-acyl chains, the myristic fatty acyl chain appears on top of the bicelles, perpendicular to the phospholipids (Liu et al. 2010). Whether this orientation is because of the physicochemical nature of the bicelles, or reflects Arf1’s positioning on physiological membranes, is presently not known.

Recruitement of coatomer to Golgi membranes is tightly correlated to Arf1 activation (Donaldson et al. 1991; Serafini et al. 1991a; Palmer et al. 1993). The complex is recruited to the Golgi membrane en bloc (Hara-Kuge et al. 1994), in contrast to clathrin/AP and COPII that are composed of two successively recruited layers. Coatomer forms multiple interfaces with Arf1-GTP: site-directed photolabeling studies highlighted specific contacts between the GTPase and the subunits β′-COP, β-COP, δ- COP, as well as the trunk domain of γ-COP (Zhao et al. 1997, 1999; Sun et al. 2007). Yeast two hybrid analysis further points to an interaction of Arf1 with ε-COP (Eugster et al. 2000). Three to four Arf1 molecules bind to one coatomer complex (Serafini et al. 1991a; Beck et al. 2009a). Thus, interaction of coatomer with Arf provides multiple protein–protein interfaces. Binding of γ-COP, via its trunk and appendage domains, to dimers of p24 transmembrane protein family members (COPI transmembrane machinery proteins, see later) further stabilizes coatomer on the Golgi membrane (Harter and Wieland 1998; Bethune et al. 2006). Members of the p24 family carry variants of a dilysine motif in combination with a double F motif, FFXX(K/R)(K/R)X_n (n ≥ 2). Binding of coatomer to these signals depends more strongly on the phenylalanine than on the dilysine residues (Fiedler et al. 1996b; Sohn et al. 1996). These machinery signatures are recognized exclusively by γ-COP (Bethune et al. 2006). p24 proteins are present in COPI vesicles in amounts stoichiometric to coatomer (Sohn et al. 1996). In a yeast strain defective for vesicle fusion, a p24 knockout reduced the number of COPI-coated vesicles (Stamnes et al. 1995). These data suggested that type I transmembrane proteins represent membrane receptors for coatomer and are actively required for the formation of COPI vesicles (Stamnes et al. 1995; Sohn et al. 1996). Surprisingly, however, in yeast the deletion of all p24 proteins showed only a reduction in the rate of transport of some cargo proteins (Springer et al. 2000). More recently, additional biochemical evidence was reported that in yeast the p24 complex participates in retrograde transport from Golgi to ER, by promoting the formation of COPI vesicles (Aguilera-Romero et al. 2008). p24 members are indispensible in mammals: knockout of p23 is lethal at the earliest possible time point in the development of a mouse embryo (Denzel et al. 2000).

Specific binding of the γ-COP trunk domain to dimers of p23 and p24 (not p25, p26, or p27), results in a conformational change in γ-COP (Reinhard et al. 1999; Bethune et al. 2006) that is transmitted to the α-COP subunit (Langer et al. 2008). This spatial rearrangement causes aggregation of the complex and is likely to initiate coatomer polymerization (Reinhard et al. 1999), providing the energy to bend the membrane and sculpting a COPI-coated bud (R Beck et al., unpubl.).

A machinery component of COPI vesicles with another dilysine motif is the seven-helix transmembrane protein KDEL receptor that functions as a transmembrane adaptor. Its cytosolic tail interacts with coatomer via a KKXSXXX signal, active only when its serine residue is phosphorylated (Cabrera et al. 2003), whereas its luminal part interacts with soluble proteins that harbor a C-terminal KDEL-sequence (Lewis and Pelham 1992). As a result, KDEL-proteins are included into COPI vesicles and retrieved to the ER (Pelham 1991; Majoul et al. 1998), where they dissociate from the KDEL receptor, probably because of a difference of pH between Golgi and ER (Wilson et al. 1993). The free KDEL-receptor is then cycled back to the Golgi.

Identification of a subpopulation of COPI vesicles, which lacks p24 proteins (Malsam et al. 2005), raises the question of whether these carriers contain transmembrane machinery components, other than p24 family proteins, that are crucial for their biogenesis.

Cargo Sorting/Sorting Motifs

Proteins are directed into COPI vesicles by various mechanisms based on direct or indirect binding to the coat.

Membrane proteins to be included into COPI vesicles can also be recognized directly by coatomer through sorting motifs present in their sequence. The first signals characterized were dilysine motifs present at the extreme carboxyl terminus of membrane proteins (Nilsson et al. 1989). They bind directly to coatomer at sites different to the above machinery proteins and induce the retrieval from the Golgi to the ER of the host protein (Cosson and Letourneur 1994; Letourneur et al. 1994). Various subsets of dilysine motifs exist that are recognized differentially by the coatomer complex (Schroder-Kohne et al. 1998). In contrast to the p24 familiy carboxy-terminal signatures, KKXX motifs interact with the WD40 domain of α-COP, and KXKXX binds to a similar domain within β′-COP (Eugster et al. 2004). Thus structural differences constitute the molecular basis for coatomer to discriminate machinery components that are recycled from cargo proteins that are transported and delivered unidirectionally. The affinity between coatomer and such cargo sorting signals depends on the nature of the X amino acids following the lysine residues (Zerangue et al. 2001), thus giving rise to differences in efficiency of retrieval to the ER.

Together with a modulation of the efficiency of ER exit, these tuned affinities may provide a mechanism to control the steady-state localization of membrane proteins at the ER-Golgi interface (Zerangue et al. 2001).

Furthermore, Arginine-based motifs that conform to the consensus sequence (Φ/Ψ/R)RXR (where Φ/Ψ is an aromatic or bulky hydrophobic residue) (Zerangue et al. 1999) are recognized by coatomer subunits β- and δ-COP (Michelsen et al. 2007). In contrast to dilysine motifs, they are not restricted to carboxyl termini and can have a more flexible positioning within the cytosolic tail of the host protein (Shikano and Li 2003). These signals can control the maturation of membrane protein complexes (Michelsen et al. 2005). As long as a complex is not fully assembled, such arginine motifs present in its subunits are exposed, inducing retrograde transport to the Golgi. On complete assembly of a complex, the signals become masked, allowing export of the complex to the cell surface (Zerangue et al. 1999). Several mechanisms for masking of the signals have been proposed, including steric masking by a partner subunit (Michelsen et al. 2005), inactivation of the signal by phosphorylation of nearby residues (Scott et al. 2001), or competition for motif binding of coatomer with 14-3-3 proteins (O’Kelly et al. 2002; Yuan et al. 2003), or PDZ-domain proteins (Standley et al. 2000).

Additional sorting motifs are based on aromatic residues. A “δL” motif confers binding to δ-COP and retrieval to the ER (Cosson et al. 1998), and a FXXXFXXXFXXLL motif in the Dopamin1 receptor mediates interaction with γ-COP, which is necessary for physiological trafficking of the receptor (Bermak et al. 2002). However, not all the membrane proteins transported within COPI vesicles carry coatomer-interacting motifs. Notably, this is the case of glycosylation enzymes with their tails lacking known sorting signal. Recent studies in yeast implicated Vps74 as an essential factor for the packaging of glycosylation enzymes into transport carriers (Schmitz et al. 2008; Tu et al. 2008). By binding simultaneously to coatomer and the cytosolic tails of glycosyltransferases, Vps74 works as a coat-cargo adaptor (Tu et al. 2008). Although further studies will be needed to unravel similar mechanisms in mammalian cells, Vps74 suggests the existence of a set of similar cargo adaptors in higher eukaryotes.

Inclusion of Cargo into COPI Vesicles

Two modes of incorporation of membrane proteins into COPI vesicles are observed. Proteins directly involved in the budding process are incorporated into vesicles in a GTP-independent manner and are further referred to as machinery components (Nickel et al. 1998; Malsam et al. 1999). Cargo proteins, on the other hand, require GTP hydrolysis for their uptake into the transport carrier. Indeed, in the presence of GTPγS, a poorly hydrolysable form of GTP, or Arf1-Q71L (an Arf1 variant locked in its GTP-loaded state), COPI vesicles can still form, but appear devoid of anterograde and retrograde cargo (Nickel et al. 1998; Lanoix et al. 1999; Malsam et al. 1999; Pepperkok et al. 2000). In contrast, uptake of p23, p24, or KDEL receptor is still effective under these conditions (Nickel et al. 1998). As previously described, activation of Arf1 is a prerequisite for the recruitment of coatomer to the membrane. As a corollary, hydrolysis of Arf1-GTP into Arf1-GDP leads to membrane uncoating (see later). This seems in contradiction with a role of GTPase activity in cargo sorting. In in vitro experiments with the COPII coat, priming complexes weakly bound to unassembled cargo are instantaneously released, whereas those interacting more strongly to oligomerized cargo remain transiently associated to the membrane, increasing their probability to be incorporated into a final COPII vesicle (Sato and Nakano 2005). This characteristic may provide the molecular basis of a GTPase-driven kinetic proofreading mechanism (Sato and Nakano 2007; Tabata et al. 2009). Kinetic analysis of the COPI system performed in living cells revealed that Arf1 dissociates faster from membranes than coatomer (Presley et al. 2002), suggesting that coatomer is not immediately released after Arf1 inactivation but stays metastably associated to the membrane for an additional period of time. This difference is likely because of the additional binding of coatomer to membrane proteins (e.g., p24 family proteins), and to lateral interactions of the polymerized complexes within the coat network. It is open at present whether the different dissociation kinetics of Arf and coatomer would reflect a kinetic proofreading mechanism, a model that was suggested by Weiss and Nilsson (2003).

A different, but not mutually exclusive, mechanism has been proposed for specific enrichment of cargo in a nascent COPI vesicle based on the curvature sensitivity of ArfGAP1 (Liu et al. 2005). Preferential binding of the protein to positively curved membranes (Bigay et al. 2003) increases local ArfGAP1 activity. As a result, GTP hydrolysis, and thus coat release, from the positive curvature of a forming bud would be increased. This GTPase activity would lead to a flux of coatomer from the rim, where Arf and coatomer are recruited to the center of a growing bud, thereby mediating cargo concentration (Liu et al. 2005). Several machinery components can locally modulate GTP hydrolysis. Coatomer itself can stimulate ArfGAP1 (Goldberg 1999; Szafer et al. 2001), whereas the cytosolic tails of p23 and p24 slow down the hydrolysis of Arf1 (Goldberg 2000; Lanoix et al. 2001). Interaction between a KDEL receptor and ArfGAP1 may recruit an additional activating protein to the membrane, and thus increase local GTPase activity (Aoe et al. 1997, 1998). This would give rise to another, less characterized, level of regulation.

Budding and Scission

More than 20 years after the discovery of COPI-coated vesicles (Orci et al. 1986; Malhotra et al. 1989), there is still an ongoing discussion about the structural components of this class of vesicular carrier (Beck et al. 2009b; Hsu and Yang 2009; East and Kahn 2010). The two cytosolic components essential for COPI vesicle formation in vitro are Arf1 and the heptameric coat complex coatomer (Serafini et al. 1991a; Orci et al. 1993).

In a chemically defined in vitro system, using synthetic liposomes with a nonphysiological lipid composition, Arf1-GTP and coatomer alone were sufficient to induce vesicle formation (Spang et al. 1998). Nevertheless, in the presence of a cytoplasmic domain of a p24 family protein, vesicle formation is stimulated, and independent of a wide range of lipid compositions (Bremser et al. 1999). As described above, the p24 proteins were initially found to be a major component of COPI-coated vesicles (Stamnes et al. 1995; Sohn et al. 1996) and serve as membrane machinery for these carriers.

More recently a membrane surface activity was observed of the activated, GTP-loaded form of Arf1 on synthetic membranes in the absence of coatomer (Beck et al. 2008; Krauss et al. 2008; Lundmark et al. 2008). This activity results in tubulation of giant unilamelar vesicles (GUVs), or of membrane sheets tethered to glass surfaces, and strictly depends on a dimerization of the small GTPase. This dimerization is essential in the living cell, as a yeast strain devoid of Arfs (Takeuchi et al. 2002) cannot be rescued with an Arf point mutant unable to dimerize (Beck et al. 2008). The surface activity of Arf1 is a prerequisite for COPI vesicle formation (Beck et al. 2008). Thus, during formation of a COPI bud, two different activities have the potential to deform a flat donor membrane into a curved bud. These are the polymerization of coatomer on its recruitment and/or the dimerization-dependent membrane surface activity of activated Arf.

Cryo electron microscopy and tomography of liposomes incubated with Arf wt, GTP and coatomer revealed production of regular free COPI vesicles, as expected from earlier work (Reinhard et al. 2003). With the nondimerizing mutant of Arf, however, hardly any free vesicles are formed. Instead, the liposomes are transformed into flower-like structures that represent multiple buds linked by a continuous bilayer, indicating a block in scission. It is of note that neither mechanical shearing nor high salt conditions are needed to create a free COPI-coated vesicle (R Beck et al., unpubl.). In the mechanistically similar COPII system, a membrane surface activity of Sar1 was revealed earlier (Bielli et al. 2005; Lee et al. 2005). Here, the fission step in the formation of COPII-coated vesicles from liposomes was dependent on the N-terminal amphipathic helix of the small GTPase (Lee et al. 2005). Most recently, a regulated scaffold assembly by Sar1 was observed (Long et al. 2010), opening a possibility that Sar1 assembly on membranes controls constriction of the bud neck similar to the fission mechanism proposed for dynamin in the clathrin system (Bashkirov et al. 2008; Pucadyil and Schmid 2008). However, the dynamin mechanism requires hydrolysis of GTP for the scission process, and in various reports the formation of COPII vesicles was described in the presence of nonhydolyzable analogs of GTP (Barlowe et al. 1994; Oka and Nakano 1994). Likewise, Arf1-dependent formation of COPI vesicles is independent of GTP hydrolysis (Malhotra et al. 1989). Taken together, this indicates that the small GTPases Arf1 and Sar1 not only mediate coat assembly but also promote the fission step to release a nascent vesicle in a mode other than that proposed for dynamin. What mechanism can explain these two functions of a small GTPase?

A Model for Membrane Separation Based on Cooperation of Coatomer with Arf1

Within the growing coat, Arf interacts with the membrane via its myristoylated amphipathic helix, and with its covering layer of coatomer via several defined interfaces (Zhao et al. 1997, 1999; Sun et al. 2007). Two Arf1 dimers are bound to one coatomer complex (Serafini et al. 1991b; Beck et al. 2008) (R Beck et al., unpubl.). In the positive curvature of the growing bud, Arf’s interaction with the membrane is energetically favorable. As the bud becomes more complete, a neck forms at the site of Arf and coatomer recruitment, with increasingly negative curvature, which is highly unfavorable for Arf binding. This energy strain could be relaxed by diffusion of the small GTPase from the negatively curved zone. If the stability of the complex of the dimerized Arfs with the covering network of coatomer is strong enough, however, to prevent Arf from escaping, the local high-energy state can only be relaxed by fusion of the adjacent membranes in the neck, causing membrane separation. The need for Arf to dimerize is simply explained by comparing the stability of a complex of coatomer formed with either two dimers of Arf (Arf wt) or with four monomers (nondimerizing mutant). With Arf wt the complex’s stability is higher because of the energy released by the two Arf dimerization interfaces (R Beck et al., unpubl.).

Additional Proteins Implicated in COPI Fission

A variety of proteins have been shown to be involved in COPI vesicle formation. Among those, Brefeldin-A ADP-ribosylated Substrate (BARS) (Yang et al. 2005), endophilin (Yang et al. 2006), and ArfGAP1 (Yang et al. 2002) are thought to play a role in membrane scission and/or act as a coat component. Some of these functions have been challenged by more recent investigations (Gallop et al. 2005; Beck et al. 2009b).

Whereas the basic mechanisms of the core machinery as deduced from experiments in reconstituted systems seem quite clear, the mechanisms underlying the contributions of such additional components that might play important roles in the living cell represent challenges for future research.

We apologize to those colleagues whose work we could not cite because of the limited space. We thank Thomas Soellner and Patricia McCabe for critical reading the manuscript, and the German Research Council for supporting our work (SFB 638, projects A10 and A16, and GRK 1188). V.P. is supported by a FEBS long-term fellowship.