Secretory Protein Biogenesis and Traffic in the Early Secretory Pathway
Charles K. Barlowe, Elizabeth A. Miller


The secretory pathway is responsible for the synthesis, folding, and delivery of a diverse array of cellular proteins. Secretory protein synthesis begins in the endoplasmic reticulum (ER), which is charged with the tasks of correctly integrating nascent proteins and ensuring correct post-translational modification and folding. Once ready for forward traffic, proteins are captured into ER-derived transport vesicles that form through the action of the COPII coat. COPII-coated vesicles are delivered to the early Golgi via distinct tethering and fusion machineries. Escaped ER residents and other cycling transport machinery components are returned to the ER via COPI-coated vesicles, which undergo similar tethering and fusion reactions. Ultimately, organelle structure, function, and cell homeostasis are maintained by modulating protein and lipid flux through the early secretory pathway. In the last decade, structural and mechanistic studies have added greatly to the strong foundation of yeast genetics on which this field was built. Here we discuss the key players that mediate secretory protein biogenesis and trafficking, highlighting recent advances that have deepened our understanding of the complexity of this conserved and essential process.

LIKE all eukaryotes, yeast cells segregate various physiological functions into distinct subcellular compartments. A key challenge is thus ensuring that appropriate proteins are delivered to the correct subcellular destination, a process that is driven by discrete sorting signals that reside in the proteins themselves. Perhaps the most prevalent type of sorting signal is that directing a protein to the secretory pathway, which handles the various proteins that are destined for the extracellular environment or retention in the internal endomembrane system. Approximately one-third of the yeast proteome enters the secretory pathway. Protein secretion is not only essential for cellular function but also provides the driving force for cell growth via delivery of newly synthesized lipid and protein that permits cell expansion. Secretory proteins enter this set of interconnected organelles at the endoplasmic reticulum (ER), which regulates protein translation, protein translocation across the membrane, protein folding and post-translational modification, protein quality control, and forward traffic of suitable cargo molecules (both lipid and protein). Once contained within the secretory pathway, proteins are ferried between compartments via transport vesicles that bud off from one donor compartment to fuse with a downstream acceptor compartment, thereby mediating directional traffic of both lipid and protein. The forward-moving, or anterograde pathway is balanced by a reverse, or retrograde pathway that returns escaped resident proteins and maintains the homeostasis of individual organelles. Early yeast screens pioneered the genetic dissection of the eukaryotic secretory pathway and were rapidly followed by biochemical approaches that permitted the molecular dissection of individual processes of protein biogenesis and traffic. Here we discuss the methodologies that have yielded great insight into the conserved processes that drive protein secretion in all eukaryotes and describe the fundamental processes that act to ensure efficient and accurate protein secretion. The reader is also referred to earlier comprehensive reviews on these topics (Kaiser et al. 1997; Lee et al. 2004) as we focus our coverage on more recent advances.

Expanding Methodologies: From a Parts List to Mechanisms and Back to More Parts

Classic screens lay the groundwork; in vitro reconstitution defines mechanism

There is no doubt that early seminal yeast genetics approaches laid the foundation upon which our understanding of protein secretion is built. From the original Novick and Schekman screens that identified a host of secretion-defective (sec) mutants (Novick and Schekman 1979; Novick et al. 1980), to additional more targeted approaches from the Schekman (more secs; Deshaies and Schekman 1987; Wuestehube et al. 1996), Gallwitz (ypt; Gallwitz et al. 1983), Ferro-Novick (bet; Newman and Ferro-Novick 1987), Jones (pep; Jones 1977), Stevens (vps; Rothman et al. 1989), and Emr (vps; Bankaitis et al. 1986) labs that expanded the repertoire of mutants with defects in secretory protein and membrane biosynthesis, the field has been blessed with an abundance of reagents that permitted the characterization of each branch of the secretory pathway (Schekman and Novick 2004). Many of these processes are essential, conserved, and have direct relevance to issues of human health, yet yeast genetics approaches remain at the forefront in deciphering molecular mechanisms, unraveling cellular redundancy and complexity, and appreciating the cross-talk between different branches of the pathway. The strength of yeast as a model system to probe this complexity lies in the combination of facile genetics and robust biochemistry that are afforded by this remarkable organism. Indeed, the field has a long history of capitalizing on yeast mutants to inform biochemical reconstitution approaches that in turn inform new genetic screening approaches.

The most pertinent example of the strength of this approach is the mechanistic description of the COPII coat proteins that drive vesicle formation from the endoplasmic reticulum. Classic epistasis analyses of the Novick and Schekman sec mutants (Novick et al. 1980), placed the early sec genes in order within the secretory pathway: sec12, sec13, sec16, and sec23 mutants blocked formation of transport vesicles and induced proliferation of the ER, whereas sec17, sec18, and sec22 mutants blocked vesicle fusion and caused accumulation of vesicles (Novick et al. 1981; Kaiser and Schekman 1990). The subsequent development of in vitro assays relied in part on the use of these mutants in biochemical complementation assays (Baker et al. 1988; Ruohola et al. 1988). Recapitulation of ER–Golgi traffic in permeabilized yeast cells was perturbed in sec23 mutants, but could be restored by incubation with cytosol prepared from wild-type cells, placing Sec23 as a soluble factor required for transport vesicle formation (Baker et al. 1988). Further refinement of these in vitro transport assays permitted the dissection of different transport stages (Rexach and Schekman 1991) and allowed the biochemical characterization of the COPII coat proteins (Barlowe et al. 1994) that generate transport intermediates, and the membrane-bound and cytosolic factors required for tethering and fusion steps that consume vesicles at the Golgi membrane (Barlowe 1997; Cao et al. 1998). Further mechanistic dissection came from even more refined reconstitution systems that permitted the identification of the minimal machinery required to generate COPII vesicles from synthetic liposomes (Matsuoka et al. 1998a,b) and defined the dynamics of individual events using real-time assays (Antonny et al. 2001).

Similar reconstitution of the COPI-mediated Golgi–ER retrograde pathway in yeast lagged somewhat behind, in part due to equivalent biochemical experiments that were under way in mammalian cells (Balch et al. 1984; Waters et al. 1991). Furthermore, due to rapid perturbation in forward (ER–Golgi) traffic when the retrograde pathway is blocked, for some time there was confusion over the directionality of COPI-mediated events (Gaynor and Emr 1997). Despite these difficulties, in vitro reconstitution of COPI-coated vesicle formation was ultimately achieved (Spang and Schekman 1998) and has been similarly dissected in minimal systems using synthetic liposomes (Spang et al. 1998).

In contrast to the genetics-informed biochemical approaches described above, minimal reconstitution of the membrane fusion events that drive vesicle consumption took a slightly different path. Armed with the knowledge that fusion is driven by proteins known as SNAREs (soluble N-ethylmaleimide-sensitive factor attachment protein receptors) and with the full description of yeast SNAREs in hand from computational analyses of the yeast genome, Rothman and colleagues established liposome-based assays that demonstrated compartment specificity of different SNARE pairs (McNew et al. 2000). That this biochemical approach largely recapitulated known pathways, previously defined by genetic means, serves to highlight the success of mutually informed genetic and biochemical approaches to fully dissect the molecular mechanisms of budding and fusion events.

Dynamics and organization revealed by live cell imaging

With budding and fusion machineries well described in minimal systems, it became apparent that there were still pieces of the puzzle missing, including the roles of some essential proteins (e.g., Sec16; Espenshade et al. 1995) that remained unexplained in terms of functionality. Furthermore, some of the more pressing mechanistic questions could not be answered by biochemical means. For example, the mode of protein and lipid traffic through the Golgi remained controversial: did COPI vesicles mediate forward traffic or did proteins proceed through the Golgi by a process of maturation of individual cisternae? These questions were addressed in part by the Glick and Nakano labs using high-resolution time-lapse imaging of living yeast cells (Losev et al. 2006; Matsuura-Tokita et al. 2006). Such experiments defined discrete sites of vesicle formation, known as transitional ER (tER) or ER exit sites (ERES), that are dynamic in nature, can form de novo but also fuse with each other, and have clear relationships with downstream Golgi elements (Bevis et al. 2002; Shindiapina and Barlowe 2010). Furthermore, imaging of distinct Golgi elements lent support for the cisternal maturation model of protein secretion, although direct imaging of cargo molecules remains to be fully demonstrated. Recent advances in superresolution imaging hold great promise in further understanding the nature of these subdomains and their relationships with distinct protein machineries and membrane compartments, although some limitations will still apply, especially with respect to the problem of detecting transient cargo molecules that are in flux through the system.

New technologies yield new players and define interplay between pathways

Since the yeast community entered the postgenomic world, a host of new tools has opened up many new approaches: the haploid deletion collection represents an accessible large-scale analysis platform for novel screens (Tong et al. 2001), the GFP- (Huh et al. 2003) and TAP-tagged (Ghaemmaghami et al. 2003) fusion databases documented the localization and abundance of many gene products, and microarray analyses of gene expression changes allow the dissection of cell-wide changes to a given perturbation (Travers et al. 2000). These new tools are being used with remarkable imagination, often capitalizing on the facile nature of yeast genetics, to define the interplay between related pathways in exciting ways. For example, microarray analysis of the changes in gene expression that occur upon induction of ER stress via the unfolded protein response (UPR) identified upregulation of machineries involved in ER-associated degradation (ERAD), ultimately leading to the appreciation that these discrete pathways are intimately coordinated to manage the burden of protein within the ER (Travers et al. 2000). A second example derives from the development of synthetic genetic array (SGA) technology, which allows the rapid generation of haploid double mutant strains (Tong et al. 2001). Although the piecemeal application of this technology was informative for individual genes, the broader application to an entire pathway was revolutionary in terms of being able to define novel functions based on shared genetic fingerprints. The first so-called epistatic miniarray profile (E-MAP) made pairwise double mutations among almost 500 early secretory pathway components, quantifying the phenotypic cost of combined mutations (Schuldiner et al. 2005). Analysis of the shared patterns of genetic interactions revealed (perhaps not surprisingly) that components in common pathways shared similar profiles, which allowed the assignation of novel functions to previously uncharacterized and enigmatic proteins. An elaboration on the E-MAP approach made elegant use of a fluorescent reporter system to first assess the UPR state of individual strains in the genomic deletion collection and then to probe how UPR activation changes in double mutant backgrounds, yielding a more subtle understanding of genetic interactions than gross life and death dichotomies, which usually form the basis of synthetic interactions (Jonikas et al. 2009). With further development of such reporters on cell status, this area of cross-talk between pathways will become more and more integrated, allowing a detailed picture of cellular physiology. However, as these new technologies yield new functional clues to previously uncharacterized genes, we need to continue to use and develop biochemical tools that allow true mechanistic insight. Again, the strength of the yeast system is the use of both genetic and biochemical tools to mutually inform new discoveries.

Secretory Protein Translocation and Biogenesis

Polypeptide targeting and translocation

The first step in biogenesis of most secretory proteins is signal sequence-directed translocation of the polypeptide into the ER. Both cotranslational and post-translational mechanisms operate in yeast to target diverse sets of soluble and integral membrane secretory proteins to the ER (Figure 1). The cotranslational translocation process is initiated when a hydrophobic signal sequence or transmembrane sequence is translated and recognized by the signal-recognition particle (SRP) for targeting to the SRP receptor at ER translocation sites (Figure 1a). In the case of post-translational translocation, cytosolic chaperones play a critical role in binding hydrophobic targeting signals to maintain the nascent secretory protein in an unfolded or loosely folded translocation competent state until delivery to the ER membrane (Figure 1b). Progress on identification and characterization of the translocation machinery will be described in turn below as the start of a continuum of events in biogenesis of secretory proteins.

Figure 1 

Membrane translocation of secretory proteins. Three well-characterized pathways operate to deliver secretory proteins to the ER for membrane translocation. (A) The signal recognition particle (SRP) recognizes a hydrophobic signal sequence or transmembrane segment during protein translation followed by targeting of the ribosome–nascent chain complex to the SRP receptor for cotranslational membrane insertion. (B) Post-translational insertion of secretory proteins depends on cytosolic Hsp70 ATPases such as Ssa1 to maintain the nascent secretory protein in an unfolded translocation competent state until delivery to the Sec63 complex formed by Sec62/Sec63/Sec71/Sec72. The Sec61 complex forms an aqueous channel for both post- and cotranslational polypeptide translocation. Kar2, a luminal Hsp70 ATPase, facilitates directed movement and folding of nascent polypeptides. (C) In GET-mediated insertion of C-terminal tail-anchored proteins, the Sgt2–Get4–Get5 complex targets nascent polypeptides to Get3 for Get1/Get2 dependent translocation. Tail-anchored proteins are integrated into the membrane in Sec61-independent pathway.

Genetic approaches in yeast uncovered key components in both the co- and post-translational translocation pathways. Appending a signal sequence to the cytosolic enzyme encoded by HIS4 targets this enzyme to the ER where it cannot function and produces histidine auxotrophy. A genetic selection for mutants that are partially defective in translocation of this signal peptide-bearing fusion protein, and therefore restore histidine prototrophy was used to identify conditional mutations in three essential genes; SEC61, SEC62, and SEC63 (Deshaies and Schekman 1987; Rothblatt et al. 1989). Sequencing indicated that all three genes encode integral membrane proteins, with the 53-kDa Sec61 protein a central component that contained 10 transmembrane segments and striking sequence identity with the Escherichia coli translocation protein SecY (Stirling et al. 1992; Jungnickel et al. 1994). Similar genetic selection approaches using the HIS4 gene product fused to integral membrane proteins identified SEC65, which encodes a component of the SRP (Stirling and Hewitt 1992; Stirling et al. 1992) as well as mutations in SEC71 and SEC72 (Green et al. 1992).

Concurrent with these genetic approaches, cell-free reconstitution assays that measured post-translational translocation of radiolabeled pre-pro-α-factor into yeast microsomes were used to dissect molecular mechanisms in this translocation pathway (Hansen et al. 1986; Rothblatt and Meyer 1986). Fractionation of cytosolic components required in the cell-free assay revealed that Hsp70 ATPases stimulated post-translational translocation (Chirico et al. 1988). Yeast express a partially redundant family of cytosolic Hsp70s encoded by the SSA1SSA4 genes that are collectively essential. An in vivo test for Hsp70 function in translocation was demonstrated when conditional expression of SSA1 in the background of the multiple ssaΔ strain resulted in accumulation of unprocessed secretory proteins as Ssa1 was depleted (Deshaies et al. 1988). ATPase activity of Hsp70 family members is often stimulated by a corresponding Hsp40/DnaJ partner and in the case of polypeptide translocation in yeast the YDJ1 gene encodes a farnsylated DnaJ homolog that functions in ER translocation (Caplan et al. 1992). Ydj1 has been shown to directly regulate Ssa1 activity in vitro (Cyr et al. 1992; Ziegelhoffer et al. 1995) and structural studies indicate that Ydj1 binds to three- to four-residue hydrophobic stretches in nonnative proteins that are then presented to Hsp70 proteins such as Ssa1 (Li et al. 2003; Fan et al. 2004). Finally, genetic experiments connect YDJ1 to translocation components in addition to multiple other cellular pathways presumably due to action on a subset of secretory proteins (Becker et al. 1996; Tong et al. 2004; Costanzo et al. 2010; Hoppins et al. 2011).

Several lines of experimental evidence indicate that the multispanning Sec61 forms an aqueous channel for polypeptide translocation into the ER. Initial approaches probing a stalled translocation intermediate in vitro revealed that direct cross-links formed only between transiting segments of translocation substrate and Sec61 (Musch et al. 1992; Sanders et al. 1992; Mothes et al. 1994). Purification of functional Sec61 complex revealed a heterotrimeric complex consisting of Sec61 associated with two ∼10-kDa proteins identified as Sss1 and Sbh1 (Panzner et al. 1995). For efficient post-translational translocation, the Sec61 complex assembles with another multimeric membrane complex termed the Sec63 complex, which consists of the genetically identified components Sec63, Sec62, Sec71, and Sec72 (Deshaies et al. 1991; Brodsky and Schekman 1993; Panzner et al. 1995). Purification of these complexes combined with proteoliposome reconstitution approaches have demonstrated that the seven polypeptides comprising the Sec61 and Sec63 complexes plus the lumenal Hsp70 protein, Kar2, are sufficient for the post-translational mode of translocation (Panzner et al. 1995). Further biochemical dissection of this minimally reconstituted system in addition to crystal structures of the homologous archaeal SecY complex (Van den Berg et al. 2004) have provided molecular insights into the translocation mechanism (Rapoport 2007). Current models for post-translational translocation suggest that the hydrophobic N-terminal signal sequence is recognized and bound initially by the Sec63 complex, which then transmits information through conformational changes to the Sec61 complex and to lumenally associated Kar2 (Figure 1b). In a second step that is probably coordinated with opening of the translocation pore, the signal sequence is detected at an interface between membrane lipids and specific transmembrane segments in Sec61 where it binds near the cytosolic face of the channel (Plath et al. 1998). Opening of the pore would then permit a portion of the hydrophilic polypeptide to span the channel where association with lumenal Kar2 would capture and drive directed movement in a ratcheting mechanism through cycles of ATP-dependent Kar2 binding (Neupert et al. 1990; Matlack et al. 1999). Well-documented genetic and biochemical interactions between Kar2 and the lumenal DnaJ domain in Sec63 are thought to coordinate directed movement into the ER lumen (Feldheim et al. 1992; Scidmore et al. 1993; Misselwitz et al. 1999). The N-terminal signal sequence is thought to remain bound at the cytosolic face of the Sec61 complex as the nascent polypeptide chain is threaded through the pore where, at some stage, the signal sequence is cleaved by a translocon-associated signal peptidase for release into the lumen (Antonin et al. 2000).

Of course, a major pathway for delivery of nascent secretory proteins to the ER employs the signal recognition particle in a co-translational translocation mechanism. Here, the ribosome–nascent chain–SRP complex is targeted to Sec61 translocons through an initial interaction between SRP and the ER-localized SRP receptor (SR) encoded by SRP101 and SRP102 (Ogg et al. 1998). In an intricate GTP-dependent mechanism, paused SRP complexes bound to SR transfer ribosome–nascent chains to Sec61 tranlocons as polypeptide translation continues in a cotranslational translocation mode (Wild et al. 2004). Genetic screens uncovered the Sec65 subunit of SRP, and purification of native SRP identified the other core subunits termed Srp14, Srp21, Srp54, Srp68, and Srp72 in addition to the RNA component encoded by SCR1 (Hann and Walter 1991; Brown et al. 1994). Somewhat surprisingly, deletion of the SRP components in yeast produced yeast cells that grow slowly but remain viable. These findings indicate that the SRP-dependent pathway is not essential, unlike the core translocation pore components, and indicates that other cytosolic machinery can manage delivery of all essential secretory proteins to the translocon. Although yeast cells can tolerate complete loss of the SRP pathway, it became clear that certain secretory proteins displayed a preference for the SRP-dependent route, whereas others were efficiently translocated into the ER in a post-translational mode (Hann et al. 1992; Stirling and Hewitt 1992). In general, integral membrane proteins and signal sequences of relatively high hydrophobicity preferentially engage the SRP-dependent pathway, whereas soluble and lower hydrophobicity signal sequences depend on a Sec63-mediated post-translational mode of translocation (Ng et al. 1996).

More recently a third post-translational translocation pathway to the ER has been characterized in yeast and other eukaryotes whereby short integral membrane proteins and C-terminal tail-anchored proteins are integrated into the membrane (Figure 1c) (Stefanovic and Hegde 2007; Schuldiner et al. 2008). For this class of proteins, transmembrane segments are occluded by the ribosome until translation is completed, thereby precluding SRP-dependent targeting. Bioinformatic analyses suggest that up to 5% of predicted integral membrane proteins in eukaryotic genomes may follow this SRP-independent route including the large class of SNARE proteins that drive intracellular membrane fusion events and are anchored by C-terminal membrane domains. Interestingly, this post-translational targeting pathway operates independently of the Sec61 and Sec63 translocon complexes (Steel et al. 2002; Yabal et al. 2003) and instead depends on recently defined soluble and membrane-bound factors. Large-scale genetic interaction analyses in yeast identified a clustered set of nonessential genes that produced Golgi-to-ER trafficking deficiencies that were named GET genes (Schuldiner et al. 2005). Get3 shares high sequence identity with the transmembrane domain recognition complex of 40 kDa (TRC40) that had been identified through biochemical strategies in mammalian cell-free assays as a major interaction partner for newly synthesized tail-anchored proteins (Stefanovic and Hegde 2007; Favaloro et al. 2008). Subsequent synthetic genetic array analyses and biochemical approaches in yeast (Jonikas et al. 2009; Battle et al. 2010; Chang et al. 2010; Chartron et al. 2010; Costanzo et al. 2010) have implicated five Get proteins (Get1–5) and Sgt2 in this process. Current models for the GET targeting pathway in yeast suggest that a Sgt2Get4Get5 subcomplex loads tail-anchored substrates onto the targeting factor Get3 (Figure 1c). The Get3-bound substrate then delivers these newly synthesized proteins to an integral membrane Get1/Get2 complex. In an ATP-dependent process, Get3 in association with Get1/Get2 then inserts the hydrophobic segment to span across the ER membrane bilayer (Shao and Hegde 2011). Although structural and biochemical studies are rapidly advancing our understanding of the GET-dependent targeting pathway, the mechanisms by which tail-anchored proteins are inserted into ER membrane bilayer remain to be defined.

Maturation of secretory proteins in the ER: signal sequence processing

For the many secretory proteins that contain an N-terminal signal sequence, the signal peptidase complex (SPC) removes this domain by endoproteolytic cleavage at a specific cleavage site during translocation through the Sec61 complex (Figure 2a). The SPC consists of four polypeptides termed Spc1, Spc2, Spc3, and Sec11 (Bohni et al. 1988; YaDeau et al. 1991). Spc3 and Sec11 are essential integral membrane proteins that are required for signal sequence cleavage activity, with the Sec11 subunit containing the protease active site (Fang et al. 1997; Meyer and Hartmann 1997). Based on structural comparisons with E. coli leader peptidase, the active site of SPC is thought to be located very near the lumenal surface of the ER membrane and presumably close to translocon exit sites. The Spc1 and Spc2 subunits are not required for viability; however, at elevated temperatures the corresponding deletion strains accumulate unprocessed precursors of secretory proteins in vivo (Fang et al. 1996) and are required for full enzymatic activity of the SPC in vitro (Antonin et al. 2000). Interestingly, Spc2 is detected in association with the Sbh1 subunit of the Sec61 complex and is thought to physically link the SPC and Sec61 complex (Antonin et al. 2000). Given that SEC11 was identified in the original SEC mutant screen as required for ER-to-Golgi transport of secretory proteins, signal sequence cleavage is regarded as an essential step for maturation of secretory proteins that contain N-terminal signal sequences.

Figure 2 

Folding and maturation of secretory proteins. A series of covalent modifications and folding events accompany secretory protein biogenesis in the ER. (A) Signal peptidase complex consisting of Spc1/Spc2/Spc3/Sec11 cleaves hydrophobic signal sequences during polypeptide translocation. (B) Coincident with polypeptide translocation and signal sequence cleavage, N-linked core-oligosaccharide is attached to consensus N-X-S/T sites within the transiting polypeptide by the multisubunit oligosaccharyl transferase complex. (C) In the oxidizing environment of the ER lumen, disulfide bond formation is reversibly catalyzed by protein disulfide isomerases (such as Pdi1) with Ero1 providing oxidizing equivalents. (D) Trimming of individual glucose and mannose residues from the attached core-oligosaccharide assists protein folding and quality control processes, which involve the calnexin family member Cne1. For terminally misfolded glycoproteins, sequential trimming of mannose residues by Mns1 and Htm1 generates a signal for ER-associated degradation.

Maturation of secretory proteins in the ER: protein glycosylation

In addition to signal sequence cleavage, attachment of asparagine-linked oligosaccharide to nascent glycoproteins occurs concomitantly with polypeptide translocation through the Sec61 pore (Figure 2b). The addition of core oligosaccharides to consensus Asn-X-Ser/Thr sites in transiting polypeptides is catalyzed by the oligosaccharyltransferase (OST) enzyme. OST is composed of eight integral membrane polypeptides (Ost1, Ost2, Ost3 or Ost6, Ost4, Ost5, Wbp1, Swp1, and Stt3) and is also detected in complex with the Sec61 translocon (Kelleher and Gilmore 2006). Indeed, for N-linked glycosylation sites that are near signal sequence cleavage sites, cleavage must occur before addition of N-linked oligosaccharide, demonstrating the sequential stages of polypeptide translocation, signal sequence cleavage, and N-linked glycosylation (Chen et al. 2001). The Stt3 subunit is critical for catalytic activity and in addition to Stt3, most of the OST subunits are required for cell viability, indicating a critical role for N-linked glycosylation in maturation of secretory proteins. OST transfers a 14-residue oligosaccharide core en bloc to most (but not all) Asn-X-Ser/Thr sites in transiting polypeptides. The 14-residue oligosaccharide core is assembled on the lipid-linked carrier molecule dolichylpyrophosphate in a complex multistep pathway (Burda and Aebi 1999).

The precise role(s) for N-linked glycosylation of secretory protein is not fully understood because in many instances mutation of single and multiple sites within a given protein produces only mild consequences. Hydrophilic N-linked glycans influence thermodynamic stability and solubility of proteins, and in the context of nascent secretory proteins in the ER, the N-linked structure is also thought to be an integral part of a system that assists in protein folding and quality control to manage misfolded glycoproteins (Schwarz and Aebi 2011). This quality control process will be explored further after covering other folding and post-translational modification events in secretory protein maturation.

In addition to N-linked glycosylation, some secretory proteins undergo O-linked glycosylation through attachment of mannose residues on Ser/Thr amino acids by protein O-mannosyltransferases (Pmts). Saccharomyces cerevisiae contains a family of seven integral membrane mannosyltranferases (Pmt1Pmt7) that covalently link mannose residues to Ser/Thr residues using dolichol phosphate mannose as the mannosyl donor (Orlean 1990; Willer et al. 2003). Both O-linked mannose residues and N-linked core oligosaccharides added in the ER are extended in the Golgi complex by the nine-membered KRE2/MNT1 family of mannosyltranferases that use GDP-mannose in these polymerization reactions (Lussier et al. 1997a,b). O-linked mannosyl modification of secretory proteins in the ER is essential in yeast (Gentzsch and Tanner 1996) and required for cell wall integrity as well as normal morphogenesis (Strahl-Bolsinger et al. 1999). The role of O-linked glycosylation in ER quality control processes remains unclear although investigators have reported influences of specific pmt mutations on turnover rates of misfolded glycoproteins (Harty et al. 2001; Vashist et al. 2001; Hirayama et al. 2008; Goder and Melero 2011) and the PMT genes are upregulated by activation of the UPR (Travers et al. 2000).

Maturation of secretory proteins in the ER: glycosylphosphatidylinositol anchor addition

Approximately 15% of proteins that enter the secretory pathway are post-translationally modified on their C terminus by addition of a lipid-anchored glycosylphosphatidylinositol (GPI) moiety. The synthesis and attachment of GPI anchors occur in the ER through a multistep pathway that depends on >20 gene products (Orlean and Menon 2007). GPI synthesis and attachment are essential processes in yeast and GPI anchored proteins on the cell surface are thought to play critical roles in cell wall structure and cell morphology (Leidich et al. 1994; Pittet and Conzelmann 2007). As with assembly of the N-linked core oligosaccharide, the GPI anchor is fully synthesized as a lipid anchored precursor and then transferred to target proteins en bloc by the GPI transamidase complex (Fraering et al. 2001). The GPI-anchoring machinery recognizes features and signals in the C terminus of target proteins that result in covalent linkage to what becomes the terminal amino acid (termed the ω residue) and removal of the ∼30-amino-acid C-terminal GPI signal sequence (Udenfriend and Kodukula 1995). Bioinformatic approaches are now reasonably effective in predicting GPI anchored proteins. These algorithms scan for open reading frames that contain an N-terminal signal sequence and a C terminus that consists of an ω residue bracketed by ∼10 residues of moderate polarity plus a hydrophobic stretch near the C terminus of sufficient length to span a membrane bilayer (Eisenhaber et al. 2004). GPI precursor proteins that do not receive GPI-anchor addition and removal of their C-terminal hydrophobic signal are not exported from the ER (Nuoffer et al. 1993; Doering and Schekman 1996) and are probably retained through an ER quality control mechanism.

Maturation of secretory proteins in the ER: disulfide bond formation

Most secretory proteins contain disulfide bonds that form when nascent polypeptides are translocated into the oxidizing environment of the ER lumen. A family of protein-disulfide isomerases that contain thioredoxin-like domains catalyze the formation, reduction, and isomerization of disulfide bonds to facilitate correct protein folding in the ER lumen (Figure 2c). In yeast, Pdi1 is an essential protein disulfide isomerase that is required for formation of correct disulfide bonds in secretory and cell surface proteins (Farquhar et al. 1991; Laboissiere et al. 1995). Pdi1 obtains oxidizing equivalents for disulfide formation from the essential flavoenzyme Ero1, which is bound to the luminal face of the ER membrane (Sevier et al. 2007). Ero1 and Pdi1 form the major pathway for protein disulfide bond formation by shuttling electrons between Ero1, Pdi1, and substrate proteins (Tu and Weissman 2002; Gross et al. 2006). In reconstituted cell-free reactions, FAD-linked Ero1 can use molecular oxygen as the electron acceptor to drive Pdi1 and substrate protein oxidation. The electron acceptor(s) used by Ero1 in vivo remain to be fully characterized (Hatahet and Ruddock 2009).

In addition to Pdi1, yeast express four other nonessential ER-localized protein disulfide isomerase homologs, Mpd1, Mpd2, Eug1, and Eps1. Overexpression of Mpd1 or mutant forms of Eug1 can partially compensate for loss of Pdi1 (Norgaard et al. 2001; Norgaard and Winther 2001). In addition to oxidoreductase activity, Pdi1 can act as a molecular chaperone in protein folding even for proteins that lack disulfide bonds (Wang and Tsou 1993; Cai et al. 1994). More recently, Pdi1 and other members of this family were reported to interact with components of the ER folding machinery including calnexin (Cne1) and Kar2 (Kimura et al. 2005) as well as the quality control mannosidase enzyme Htm1 (Gauss et al. 2011). Growing evidence indicates that this family of protein disulfide isomerases contains different domain architectures (Vitu et al. 2008) to dictate interactions with specific ER-chaperone proteins and thus shepherd a broad range of client proteins into folded forms or into ER-associated degradation pathways (Figure 2d).

Glucosidase, mannosidase trimming, and protein folding

The initial 14-residue N-linked core oligosaccharide that is attached en bloc to nascent polypeptides is subsequently processed by glycosylhydrolases in a sequential and protein conformation-dependent manner to assist protein folding and quality control in the ER lumen (Helenius and Aebi 2004). The Glc3Man9GlcNAc2 glycan, which comprises the N-linked core, is rapidly processed by glucosidase I (Gls1/Cwh41) and glucosidase II (Gls2/Rot2) enzymes to remove the three terminal glucose residues and generate Man9GlcNAc2. Molecular chaperones collaborate in protein folding during these glucose-trimming events and Rot1 alone has been shown to possess a general chaperone activity (Takeuchi et al. 2008). In many cell types, a calnexin-dependent folding cycle operates to iteratively fold and monitor polypeptide status through the coordinated activities of glucosidase I, glucosidase II, UDP-glucose;glycoprotein glucosyltransferase (UGGT), and calnexin (Cne1). After removal of terminal glucose residues by the glucosidase enzymes, UGGT can add back a terminal glucose to the glycan if the polypeptide is not fully folded to generate the Glc1Man9GlcNAc2 structure. This Glc1Man9GlcNAc2 form of an unfolded protein binds to calnexin, which keeps the nascent polypeptide in an iterative folding cycle. Once fully folded, UGGT does not act after glucosidase II and the nascent protein exits the cycle (Helenius and Aebi 2004). This calnexin cycle operates in many eukaryotes but it is currently unclear how or if the cycle works in yeast since deletion of Cne1, Gls1, Gls2, or Kre5 (potential UGGT-like protein) do not produce strong delays in biogenesis of secretory proteins but are known to produce defects in biosynthesis of cell wall β-1,6-glucan (Shahinian and Bussey 2000). Although a precise molecular understanding of the calnexin cycle components in yeast folding remains to be determined, there are clear genetic (Takeuchi et al. 2006; Costanzo et al. 2010) and biochemical (Xu et al. 2004; Kimura et al. 2005) interactions that indicate a coordinated role for these factors in protein folding.

In addition to the glucose trimming of core oligosaccharide, two additional ER-localized mannosidase enzymes termed Mns1 and Htm1 remove terminal mannose residues from the Man9GlcNAc2 glycan-linked structure (Figure 2d). Mns1 and Htm1 are related enzymes with distinct specificities. Mns1 removes the terminal mannosyl residue of the B branch of Man9GlcNAc2 and it is typically the Man8GlcNAc2 processed form of fully folded glycoproteins that is exported from the ER (Jakob et al. 1998). Htm1 is thought to act after Mns1 on terminally misfolded proteins (or misfolded proteins that have lingered in the ER folding cycle for too long) to remove the outermost mannosyl residue from the C branch of the glycan to generate Man7GlcNAc2 (Clerc et al. 2009). This form of the glycan is then recognized by the ER lectin Yos9 and targets misfolded proteins for ER-associated degradation (Carvalho et al. 2006; Denic et al. 2006). Although Mns1- and Htm1-deficient cells appear to transport folded secretory proteins at normal rates, both display significant delays in turnover of terminally misfolded glycoproteins (Jakob et al. 1998, 2001), which serves to highlight an important role for mannosidase activity in ER quality control.

Folding of nascent polypeptides throughout translocation and within the ER is also managed by Hsp70 ATPase systems, which handle partially folded intermediates. In general, Hsp70 proteins hydrolyze ATP when binding to exposed hydrophobic stretches in unfolded polypeptides to facilitate protein folding. The Hsp70 remains bound to unfolded substrates until ADP is released with this Hsp70 ATPase cycle governed by specific DnaJ-like proteins that stimulate ATP hydrolysis and nucleotide exchange factors that drive ADP release (Hartl 1996; Bukau and Horwich 1998). In yeast, the Hsp70 Kar2 plays a prominent role in ER folding in concert with the related Hsp70 protein Lhs1 (Rose et al. 1989; Baxter et al. 1996; Brodsky et al. 1999; Steel et al. 2004). For Kar2, the known DnaJ-like stimulating factors include Sec63, Scj1, and Jem1 (Schlenstedt et al. 1995; Nishikawa and Endo 1997), whereas the GrpE family member Sil1 and, surprisingly, the unrelated ATPase Lhs1 serve as nucleotide exchange factors (Hale et al. 2010). Complexity in regulating the Kar2 ATPase cycle probably reflects the range of unfolded substrates that Kar2 must handle in maintaining ER homeostasis, and there are likely to be additional factors that couple Kar2 activity to other specific ER processes. As mentioned above, Kar2 chaperone activity is tightly linked with the PDI, calnexin, and glycan trimming pathways (Figure 2d). Finally, Kar2 also plays a prominent role in ER-associated degradation (ERAD) pathways to dispose of terminally misfolded proteins (Nishikawa et al. 2001). Although our understanding of Kar2 biochemical activity is advanced, the coordinated control of Kar2-dependent folding and modification cycles in the context of an ER lumenal environment remains a challenging area.

ERAD of misfolded and unassembled proteins proceeds through a series of pathways that remove targeted proteins from the ER for ubiquitin- and proteasome-dependent degradation in the cytoplasm. ERAD is thought to play a key role in ER homeostasis and cellular physiology. Since these pathways divert misfolded secretory proteins from their routes of biogenesis, this important topic is beyond the scope of this current review and the reader is referred to excellent recent reviews (Vembar and Brodsky 2008; Smith et al. 2011).

Control of ER homeostasis by the Unfolded Protein Response

Much of the folding and biogenesis machinery in the ER is under a global transcriptional control program referred to as the UPR. The yeast UPR is activated by an increase in the level of unfolded proteins in the ER, which can be experimentally induced by treatment with inhibitors of ER protein folding (e.g., tunicamycin, dithiothreitol) or by overexpression of terminally misfolded proteins (Bernales et al. 2006). Regulation of the UPR was initially examined through identification of a 22-nucleotide segment in the KAR2 promoter region, termed the unfolded protein response element (UPRE), which was required for UPR activation of Kar2 expression. Fusion of this KAR2 promoter element to a lacZ reporter provided an elegant screen for gene mutations that blunted UPR reporter expression (Cox et al. 1993; Mori et al. 1993). Genetic screening led to the discovery that IRE1, HAC1, and RLG1 were required for a robust UPR under ER stress conditions (Cox and Walter 1996; Sidrauski et al. 1996). Further studies revealed that IRE1 encodes an ER transmembrane protein with cytosolic kinase/ribonuclease domains and a lumenal sensor domain that together are thought to serve as readout on unfolded protein levels. HAC1 encodes a basic leucine zipper transcription factor that binds to UPRE-containing segments of DNA and induces their expression (Cox and Walter 1996). Surprisingly, RLG1 encodes a tRNA ligase that is required for the nonconventional splicing of HAC1 pre-mRNA. Structural and mechanistic dissection of these core components is now advanced. Current models indicate that the Ire1 lumenal domain interacts with Kar2 and unfolded proteins to sense protein folding status (Bertolotti et al. 2000; Pincus et al. 2010; Gardner and Walter 2011). When unfolded proteins accumulate in the ER, Ire1 forms oligomers that activate the cytoplasmic kinase and ribonuclease domains. Activated Ire1 ribonuclease then acts on HAC1 pre-mRNA to remove a nonconventional intron and this splicing intermediate is then ligated by the Rlg1 ligase to produce mature HAC1 mRNA. Translation of HAC1 message produces Hac1 protein, which is a potent transcriptional activator of UPR target genes (Bernales et al. 2006).

In addition to Kar2, the UPR was known to induce other ER folding components including Pdi1 and Eug1 (Cox et al. 1993; Mori et al. 1993). To comprehensively assess the transcriptional profile of the yeast UPR, DNA microarray analysis was powerfully applied to monitor mRNA levels under ER stress conditions (Travers et al. 2000). Comparing transcription profiles in wild-type, ire1Δ, and hac1Δ strains after UPR induction revealed 381 genes that passed stringent criteria as UPR targets. Not surprisingly, 10 genes involved in ER protein folding were identified as UPR targets and included JEM1, LHS1, SCJ1, and ERO1. In addition, dozens of genes involved in ER polypeptide translocation, protein glycosylation, and ER-associated degradation were induced. Perhaps more surprisingly, 19 genes involved in lipid and inositol metabolism as well as 16 genes encoding proteins that function in vesicle trafficking between the ER and Golgi were upregulated by the UPR. These findings highlight a global role for the UPR in regulating ER homeostasis through balancing ER lipid and protein biosynthetic rates. In the context of cellular physiology, the UPR is now thought to serve a central role in sensing and integrating secretory pathway function to finely tune ER capacity in response to cellular demands (Walter and Ron 2011).

Transport From the ER: Sculpting and Populating a COPII Vesicle

Once secretory proteins have completed their synthesis and modification regimes, they become competent for forward traffic through the secretory pathway, a process mediated by a series of transport vesicles that bud off from one compartment, traverse the cytoplasm, and fuse with a downstream organelle (Figure 3). ER-derived vesicles are created by the COPII coat that, like other coat protein complexes, is charged with the dual tasks of creating a spherical transport vesicle from a planar donor membrane and populating the nascent vesicle with the appropriate cargoes. Biochemical characterization of this process, first from complex microsomal membranes using purified COPII coat proteins (Barlowe et al. 1994), then in more reduced form from synthetic liposomes (Matsuoka et al. 1998b), and subsequently at the structural level through cryo-EM (Stagg et al. 2006) and X-ray crystallography (Bi et al. 2002; Fath et al. 2007), has been remarkably fruitful in defining the molecular basis of these events. What has emerged is an elegant mechanism whereby the minimal COPII machinery, composed of five proteins (Sar1, Sec23, Sec24, Sec13, and Sec31), suffices to fulfill these multiple functions. However, recent insights into how this process is regulated suggest there is still much to learn about coat dynamics in the cell, and the precise physical basis for various steps, including membrane scission during vesicle release, vesicle uncoating, and the formation of large transport carriers capable of shuttling large cargoes.

Figure 3 

Coat assembly drives vesicle formation. Both the COPII (left) and COPI (right) coats are directed in their assembly by small GTPases of the Arf/Sar1 family. In the COPII coat, Sar1 is activated by its guanine nucleotide exchange factor (GEF), Sec12, which localizes to the ER membrane. Activated Sar1–GTP recruits the Sec23/Sec24 dimer, which corresponds to the “inner coat” layer and provides the cargo-binding function. A heterotetramer of Sec13/Sec31 is subsequently recruited, forming the “outer coat” and polymerizing into a lattice-like structure that drives membrane curvature. In the COPII cage formed by Sec13/Sec31, four molecules of Sec31 assemble head-to-head via β-propeller domains to form the “vertex” of the cage (inset). The COPI coat assembles upon activation of Arf1, which is driven by either of the redundant GEFs, Gea1 or Gea2. Arf1 in turn recruits the inner coat complex of Sec21/Sec26/Ret2/Ret3, which has homology to the clathrin AP-2 adaptor complex. The COPI outer coat is formed by Sec27/Ret1/Sec28, which assembles in a triskelion structure via interactions of three β-propeller domains of Sec27 (inset).

Structure and assembly of the COPII coat

COPII coat assembly (Figure 3) is initiated by the local recruitment and activation of the small G protein, Sar1 (Nakano and Muramatsu 1989; Barlowe et al. 1993) upon exchange of GDP for GTP, catalyzed by an ER membrane protein, the guanine nucleotide exchange factor (GEF) Sec12 (Nakano et al. 1988; d’Enfert et al. 1991). GTP loading on Sar1 exposes an amphipathic α-helix that likely induces initial membrane curvature by locally expanding the cytoplasmic leaflet relative to the lumenal leaflet (Lee et al. 2005). GTP-bound, membrane-associated Sar1 subsequently recruits the heterodimeric complex of Sec23 and Sec24 (Matsuoka et al. 1998b). Sec23 is the GTPase-activating protein (GAP) for Sar1 (Yoshihisa et al. 1993), contributing a catalytic arginine residue analogous to GAP stimulation in many Ras-related G proteins (Bi et al. 2002). Sec24 provides the cargo-binding function of the coat, containing multiple independent domains that interact directly with specific sorting signals on various cargo proteins (Miller et al. 2002, 2003; Mossessova et al. 2003). The Sar1/Sec23/Sec24 “prebudding” complex in turn recruits the heterotetrameric complex of Sec13 and Sec31 (Matsuoka et al. 1998b). Sec31 also contributes to the GTPase activity of the coat by stimulating the GAP activity of Sec23 (Antonny et al. 2001; Bi et al. 2007). Thus, the fully assembled coat is composed of two distinct layers: the “inner” membrane proximal layer of Sar1/Sec23/Sec24 that intimately associates with lipid headgroups (Matsuoka et al. 2001) and contributes cargo-binding function, and the “outer” membrane distal layer composed of Sec13/Sec31. Both layers contribute to the catalytic cycle of Sar1 and endowing maximal GTPase activity when the coat is fully assembled (Antonny et al. 2001).

Our mechanistic understanding of COPII coat action has been significantly enhanced by the structural characterization of the different coat components. A structure of the Sec23/Sec24 dimer showed a bow-tie shaped assembly with a concave face that is presumed to lie proximal to the membrane and is enriched in basic amino acids (Bi et al. 2002). These charged residues may facilitate association with the acidic phospholipid headgroups of the ER membrane. Subsequent structural, genetic, and biochemical analyses of Sec24 revealed multiple discrete sites of cargo interaction dispersed around the perimeter of the protein (Miller et al. 2003; Mossessova et al. 2003). Structural analysis of the outer coat was facilitated by the observation that under some conditions, the purified coat proteins can self-assemble into “cages” of the approximate size of a COPII vesicle (Antonny et al. 2003). Further experiments using mammalian Sec13/Sec31 recapitulated this self-assembly reaction and led to a cryoelectron microscopy structure of the COPII cage, which forms a lattice-like structure with geometry distinct from that of the clathrin coat (Stagg et al. 2006). Heterotetrameric Sec13/Sec31 complexes form straight rods, known as “edge” elements, four of which come together at “vertex” regions to drive cage assembly (Figure 3). Subsequent crystal structures of Sec13 and a portion of Sec31 revealed an unexpected domain arrangement within the edge element, whereby Sec31 forms both the dimerization interface along the edge element and the vertex assembly unit with Sec13 sandwiched between these structural elements (Fath et al. 2007). However, the fragment of Sec31 that fits well into the density of the cryo-EM structure represents only about half of the protein: an additional proline-rich domain contains the GAP-stimulatory activity of Sec31. Again, the crystal structure of this region bound to Sar1/Sec23 has yielded great insight into the mechanism of GAP activity, whereby the active fragment of Sec31 lies along the membrane-distal surface of Sec23/Sar1 and optimizes the orientation of the catalytic histidine of Sar1 (Bi et al. 2007).

The ability of Sec13/Sec31 to assemble into a spherical structure that matches closely the size of a COPII vesicle suggests that the primary membrane bending force may come from the scaffolding effect of this structure on the ER membrane. Indeed, when the curvature-inducing amphipathic helix of Sar1 is replaced with an N-terminal histidine tag to drive recruitment to Ni-containing liposomes, subsequent recruitment of Sec23/Sec24 and Sec13/Sec31 is sufficient to drive the generation of spherical buds that remain attached to the donor liposome (Lee et al. 2005). Thus an additional function of the Sar1 helix is to drive vesicle scission, a model supported by experiments that link GTPase activity to vesicle release in a manner analogous to that proposed for dynamin (Pucadyil and Schmid 2009; Kung et al. 2012). Although the concave face of Sec23/Sec24 may also contribute to membrane curvature, it has been suggested that the relatively paltry dimer interface between these two molecules is not robust enough to impart curvature despite an intimate interaction with the lipid bilayer (Zimmerberg and Kozlov 2006). Thus, although Sar1 and Sec23/Sec24 may participate in membrane curvature, the majority of membrane bending force likely comes from Sec13/Sec31. Indeed, recent genetic and biochemical experiments support this model: Sec31 likely forms all the contacts needed to make the COPII cage (Fath et al. 2007) with Sec13 providing structural rigidity to the cage edge element to overcome the membrane bending energy of a cargo-rich membrane (Copic et al. 2012).

Cargo capture: stochastic sampling vs. direct and indirect selection

The fundamental function of vesicles is to ensure directional traffic of protein cargoes, making cargo capture an integral part of coat action. To some extent, cargo can enter into vesicles in a nonspecific manner known as bulk flow, whereby stochastic sampling of the ER membrane and lumen occurs during vesicle formation, capturing local molecules by chance. Although this mode of transport could traffic some abundant cargoes, the random nature of this process cannot explain the efficiency with which some ER export occurs. In particular, some cargoes are dramatically enriched in vesicles above their prevailing concentration in the ER, suggesting a more efficient and selective packaging process. Although the concentrative mode of cargo selection has gained favor in the last decade, recent experiments reevaluating the potential for bulk flow to explain forward traffic of some proteins warrants a more detailed analysis of the potential prevalence of this nonspecific pathway, especially with respect to abundant, nonessential proteins where the efficiency of secretion may not be central to cellular viability (Thor et al. 2009).

Selective enrichment of cargo in transport vesicles via specific sorting signals is a common paradigm in intracellular protein trafficking, first characterized in endocytosis. Deciphering a similar mode of transport for the entire spectrum of cargoes handled by the COPII coat, however, has been hindered by the absence of a single common signal used by the entire secretome. Instead, multiple signals seem to drive selective capture, meaning the COPII coat must recognize various signals employed by structurally diverse cargoes. Such signals range from simple acidic peptides (Malkus et al. 2002) to folded epitopes (Mancias and Goldberg 2007) and can act either by interacting directly with the COPII coat or by binding to a cargo adaptor that links them to the coat indirectly (Figure 4) (Dancourt and Barlowe 2010).

Figure 4 

Cargo selection can be direct or indirect. Selective cargo capture during vesicle formation can occur via direct interaction of cargo molecules with the COPI and COPII coats. ER export signals (e.g., DxE, LxxLE, and YxxNPF) interact directly with Sec24 to facilitate capture into COPII vesicles. Similarly, dilysine and diaromatic signals mediate interaction with the COPI coat to direct retrograde traffic back to the ER. Soluble secretory proteins may be captured indirectly via specific cargo receptors that serve to recognize the transport-competent cargo and link it to the coat. Erv29 is the cargo receptor for many soluble secretory proteins. Soluble ER residents are returned back to the ER via a similar cargo receptor system, driven by Erd2, which recognizes HDEL signals. Membrane proteins may also require cargo adaptor proteins such as Erv14 and Rer1, although the basis for cargo recognition is not as well defined.

Genetic, biochemical, and structural data support Sec24 as the cargo binding adaptor for the COPII coat, forming a relatively static platform that has multiple binding sites for interaction with distinct sorting signals. The so-called A site binds the SNARE, Sed5, via a NPF motif (Mossessova et al. 2003; Miller et al. 2005); the B site is most diverse, recognizing acidic sorting signals such as those found on the SNARE, Bet1, the Golgi membrane protein, Sys1, and unknown signals on additional cargoes (Miller et al. 2003; Mossessova et al. 2003); the C site binds a folded epitope formed by the longin domain of the SNARE, Sec22 (Miller et al. 2003; Mancias and Goldberg 2007). The repertoire of binding sites is further expanded by the presence of additional Sec24 isoforms, the nonessential Iss1 and Lst1 proteins (Roberg et al. 1999; Kurihara et al. 2000; Peng et al. 2000). Sec24–cargo interactions are in general fairly low affinity (Mossessova et al. 2003), which is compatible with the transient nature of the association of cargo with coat: proteins must bind during vesicle formation but must also be released prior to vesicle fusion to allow coat recycling and exposure of fusogenic domains. The possibility remains that additional layers of regulation impact coat dissociation from cargo molecules after vesicle release: Sec23 is both ubiquitinated (Cohen et al. 2003) and phosphorylated (Lord et al. 2011) and similar activity on Sec24 may promote uncoupling of coat from cargo.

Some cargoes, by topology or preference, do not interact directly with Sec24 but instead use adaptor/receptor proteins to link them to the coat indirectly (Dancourt and Barlowe 2010). Some of these adaptors likely function as canonical receptors, binding to their ligands in one compartment and simultaneously interacting with Sec24 to couple cargo with coat, then releasing their ligand in another compartment, perhaps as the result of a change in ionic strength or pH of the acceptor organelle (Figure 3). Although their precise mechanisms of ligand binding and release remain to be fully explored, such receptors include Erv29, which mediates traffic of soluble secretory proteins like pro-α-factor and CPY (Belden and Barlowe 2001) and Emp46/Emp47, which are homologous to the mammalian ERGIC-53 family of proteins that mediate traffic of coagulation factors (Sato and Nakano 2002). Other receptors function to enrich vesicles with membrane protein cargoes. The p24 proteins, Emp24, Erv25, Erp1, and Erp2, are required for efficient ER export of GPI-anchored proteins, whose lumenal orientation precludes direct coupling to the COPII coat (Belden and Barlowe 1996; Muniz et al. 2000; Belden 2001). Others, like Erv26 (Bue et al. 2006; Bue and Barlowe 2009) and Erv14 (Powers and Barlowe 1998; Powers and Barlowe 2002; Herzig et al. 2012), mediate efficient export of transmembrane proteins that have cytoplasmically oriented regions but either do not contain ER export signals or require additional affinity or organization to achieve efficient capture. The requirement for receptors for such transmembrane cargoes remains unexplained, but may derive from the ancestral history of the cargoes whereby previously soluble proteins became membrane anchored as a result of gene fusion events (Dancourt and Barlowe 2010). Alternatively, the receptor proteins may provide additional functionality required for efficient ER egress, like a chaperoning function that would protect the long transmembrane domains of plasma membrane proteins from the relatively thinner lipid bilayer characteristic of the ER (Sharpe et al. 2010). Indeed, some cargo proteins have specific chaperoning needs, with ER resident proteins that are not themselves captured into COPII vesicles likely functioning to promote assembly and folding of polytopic membrane proteins. For example, the amino acid permeases all depend on an ER resident, Shr3, for correct folding and quaternary assembly, which is itself a prerequisite for COPII capture (Ljungdahl et al. 1992; Kuehn et al. 1996; Gilstring et al. 1999; Kota et al. 2007).

Regulation of COPII function: GTPase modulation, coat modification

The GTPase activity of the coat is the primary mode of regulation, known to govern initiation of coat assembly/disassembly through canonical GEF and GAP activities of Sec12 (d’Enfert et al. 1991) and Sec23 (Yoshihisa et al. 1993) respectively, but also contributing to additional functions, like discrimination of relevant cargo proteins (Sato and Nakano 2005) and vesicle scission (Bielli et al. 2005; Lee et al. 2005). Unlike other coat systems, the COPII coat uses a combinatorial GAP activity that is provided by components of the coat themselves, Sec23 (Yoshihisa et al. 1993) and Sec31 (Antonny et al. 2001). The effect of this autonomous GAP in minimal systems is that as soon as the coat fully assembles, GTP is hydrolyzed and the coat is rapidly released (Antonny et al. 2001), creating a paradox as to how coat assembly might be sustained for a sufficient length of time to generate vesicles. One solution to this conundrum is that constant Sec12 GEF activity feeds new coat elements into a nascent bud (Futai et al. 2004; Sato and Nakano 2005); coat release from the membrane might also be delayed by the increased affinity afforded by cargo proteins (Sato and Nakano 2005). However, recent findings suggest that a GAP inhibitory function, contributed by the peripheral ER protein, Sec16, also modulates the activity of the coat (Kung et al. 2012; Yorimitsu and Sato 2012). Sec16 is a large, essential protein that associates with the cytoplasmic face of the ER membrane at ERES (Espenshade et al. 1995; Connerly et al. 2005). It interacts with all of the COPII coat proteins (Gimeno et al. 1996; Shaywitz et al. 1997) and is thus thought to scaffold and/or organize coat assembly at these discrete domains (Supek et al. 2002; Shindiapina and Barlowe 2010). In addition to this recruitment function, a fragment of Sec16 dampens the GAP-stimulatory effect of Sec31, probably by preventing Sec31 recruitment to Sar1/Sec23/Sec24 (Kung et al. 2012). The GAP-inhibitory effect of Sec16 was diminished in the context of a point mutation in Sec24 (Kung et al. 2012), raising the tantalizing possibility that cargo engagement by Sec24 could trigger interaction with Sec16 to inhibit the full GTPase activity of the coat in such a manner that a vesicle is initiated around a cargo-bound complex of Sar1/Sec23/Sec24/Sec16 (Springer et al. 1999).

Another poorly explored aspect of COPII regulation is post-translational modification of the coat. Sec23 is a target for ubiquitination and is seemingly rescued from degradation by the action of the ubiqutin protease complex, Bre5/Ubp3 (Cohen et al. 2003). Whether this activity only controls expression levels of the protein or contributes more subtly to regulate protein–protein interactions remains to be tested. Furthermore, the potential ubiquitination of other COPII coat components also warrants investigation: recent experiments in mammalian cells identified Sec31 as a target for a specific monoubiquitination event that is important for ER export of collagen fibers (Jin et al. 2012). Whether yeast Sec31 is similarly modified by the equivalent E3 ubiquitin ligases, and how such a modification might influence coat action, perhaps by contributing to the structural integrity of the coat to drive membrane bending around rigid cargoes, remains to be tested. Like ubiquitination, the role of coat phosphorylation is only starting to be explored. It has long been known that Sec31 is a phosphoprotein and that dephosphorylation specifically impacted vesicle release (Salama et al. 1997). However, despite the many sites of Sec31 phosphorylation being revealed by high throughput phosphoproteomics, the precise function of these modifications remains unclear. In contrast, progress has recently been made in understanding phosphorylation of Sec23 and how this event probably influences the directionality of vesicle traffic by controlling sequential interactions with different Sec23 partners (Lord et al. 2011). It is tempting to speculate that similar phosphorylation of Sec24 might also regulate coat displacement from cargo molecules to further promote coat release and expose the fusogenic SNARE proteins, that would otherwise be occluded by their interaction with the coat. Indeed, at least partial uncoating of COPII vesicles is required for fusion to ensue since when GTP hydrolysis is prevented, vesicles fail to fuse (Barlowe et al. 1994). Whether additional protein–protein interactions or post-translational modifications contribute to coat shedding remains to be seen.

Higher-order organization of vesicle formation

Although the minimal COPII coat can drive vesicle formation from naked liposomes (Matsuoka et al. 1998b), this process in vivo is likely tightly regulated to enable both efficient vesicle production and adaptability to suit the secretory burden of the cell (Farhan et al. 2008). In part, this regulation occurs at the level of the subdivision of the ER into discrete ERES from which vesicles form. These small domains are marked by both the COPII coat proteins themselves and accessory proteins such as Sec16, and, in some cells, Sec12 (Rossanese et al. 1999; Connerly et al. 2005; Watson et al. 2006). ERES are located throughout the ER, with a seemingly random distribution that may in fact correspond to regions of high local curvature induced by the ER membrane proteins, Rtn1, Rtn2, and Yop1 (Okamoto et al. 2012). In related yeasts, these sites are dynamic, with the ability to form de novo, fuse, and divide (Bevis et al. 2002). Although the precise mechanisms that regulate the steady state distribution and size of these domains remain unclear, activity of both Sec12 and Sec16 seems to play a role (Connerly et al. 2005), as does the lipid composition of the ER (Shindiapina and Barlowe 2010). In mammalian cells, misfolded proteins that are incompetent for forward traffic are excluded from ERES (Mezzacasa and Helenius 2002) and this also seems to be true for some proteins in yeast, most notably GPI-anchored proteins with lipid anchors that have not been adequately remodeled, which are not concentrated at ERES but instead remain dispersed within the bulk ER (Castillon et al. 2009).

Vesicle Delivery to the Golgi

After release of COPII vesicles from ER membranes, tethering and fusion machineries guide ER-derived vesicles to Golgi acceptor membranes through the action of over a dozen gene products (Figure 5). Although ER–Golgi transport can be separated into biochemically distinct stages using cell-free assays, evidence suggests that these events may be organized in a manner that couples the budding and fusion stages. In general, budded vesicles become tethered to Golgi membranes through the action of the Ypt1 GTPase and tethering proteins Uso1 and the transport protein particle I (TRAPPI) complex. Membrane fusion between vesicle and Golgi acceptor membranes is then catalyzed through assembly of SNARE protein complexes from the apposed membrane compartments. How the budding, tethering, and fusion events are coordinated in cells remains an open question, although genetic, biochemical, and structural studies have advanced our understanding of underlying molecular mechanisms in vesicle tethering and membrane fusion described below.

Figure 5 

Vesicle tethering and fusion. Anterograde delivery of COPII-coated vesicles is mediated by a variety of tethering and fusion complexes. The TRAPP complex binds to Sec23 on the surface of a COPII vesicle and mediates local activation of the Rab family member, Ypt1. Ypt–GTP recruits downstream effectors such as the long coiled-coil tether, Uso1. A Golgi-localized kinase, Hrr25, phosphorylates Sec23 and displaces TRAPP, perhaps contributing to coat shedding. Removal of the coat exposes the fusogenic SNARE proteins, which assemble to drive membrane mixing. In the retrograde pathway, COPI-coated vesicles employ the DSL1 complex, composed of Dsl1/Sec39/Tip20, to recognize the incoming vesicle and coordinate coat release and SNARE pairing.

Vesicle tethering

Initial cell free transport assays, coupled with genetic approaches, placed ER–Golgi transport requirements into distinct vesicle budding and vesicle consumption/fusion stages (Kaiser and Schekman 1990; Rexach and Schekman 1991). Ypt1, identified as a founding member of the Rab family of GTPases, was implicated in the vesicle targeting stage in the ER–Golgi transport pathway (Schmitt et al. 1988; Segev et al. 1988; Baker et al. 1990). In reconstituted vesicle fusion reactions, Ypt1 was found to act in concert with the extended coil-coiled domain protein Uso1 to tether COPII vesicles to Golgi acceptor membranes (Nakajima et al. 1991; Barlowe 1997). In these assays, freely diffusible COPII vesicles could be tethered to and sedimented with washed Golgi acceptor membranes upon addition of purified Uso1. Interestingly, the Uso1- and Ypt1-dependent tethering stage does not appear to require the downstream SNARE protein fusion machinery (Sapperstein et al. 1996; Cao et al. 1998).

In addition to the extended structure of Uso1, which is predicted to span a distance of >180 nm (Yamakawa et al. 1996), the multisubunit TRAPPI complex is required for COPII-dependent transport to Golgi acceptor membranes (Rossi et al. 1995; Sacher et al. 1998). In vitro assays revealed that TRAPPI can also function to physically link COPII vesicles to Golgi membranes (Sacher et al. 2001). Structural analyses show that TRAPPI is a 170-kDa particle consisting of six subunits (Bet3, Bet5, Trs20, Trs23, Trs31, and Trs33) that assemble into a flat bilobed arrangement with dimensions of ∼18 nm × 6 nm × 5 nm (Kim et al. 2006). Bet3 can bind directly to Sec23, and with TRAPPI peripherally bound to membranes, this activity is thought to link partially coated COPII vesicles to Golgi acceptor membranes (Cai et al. 2007). In a recent study, the Golgi-associated Hrr25 kinase was reported to phosphorylate Sec23/Sec24 and regulate interactions between Sec23 and TRAPPI to control directionality of anterograde transport (Lord et al. 2011). Moreover, TRAPPI functions as a GEF for Ypt1 in a manner that is thought to generate activated Ypt1 on the surface of Golgi acceptor membranes and/or COPII vesicles (Jones et al. 2000; Wang et al. 2000; Lord et al. 2011). A subassembly of TRAPPI consisting of Bet3, Bet5, Trs23, and Trs31 binds Ypt1p and catalyzes nucleotide exchange by stabilizing an open form of this GTPase (Cai et al. 2008). TRAPPI does not appear to interact directly with Uso1, although Ypt1 activation could serve to coordinate the long-distance tethering mediated by Uso1 with a closer TRAPPI-dependent tethering event. The precise orientation of TRAPPI on Golgi and vesicle membranes is not known, but current models suggest that this multisubunit complex links COPII vesicles to the cis-Golgi surface and serves as a central hub in coordinating vesicle tethering with SNARE-mediated membrane fusion.

Genetic and biochemical evidence indicate that other coiled-coil domain proteins also act in COPII vesicle tethering and/or organization of the early Golgi compartment in yeast. The GRASP65 homolog Grh1 is anchored to cis-Golgi membranes through N-terminal acetylation and forms a complex with another coiled-coil domain protein termed Bug1 (Behnia et al. 2007). Grh1 and Bug1 are not essential, but deletion of either protein reduces COPII vesicle tethering and transport levels in cell-free assays and the grh1Δ and bug1Δ mutants display negative genetic interactions with thermosensitive ypt1 and uso1 mutants (Behnia et al. 2007). These findings suggest a redundant network of coiled-coil proteins that act in tethering vesicles and organizing the cis-Golgi compartment. Indeed, additional coiled-coil proteins including Rud3 and Coy1 localize to cis-Golgi membranes and are implicated in organization of the cis-Golgi and interface with COPII vesicles (VanRheenen et al. 1999; Gillingham et al. 2002, 2004). Although some double deletion analyses have been performed with these genes, multiple deletions may be required to severely impact this redundant network.

SNARE protein-dependent membrane fusion

Fusion of tethered COPII vesicles with cis-Golgi membranes depends on a set of membrane-bound SNARE proteins. Several lines of evidence indicate that the SNARE proteins Sed5, Bos1, Bet1, and Sec22 catalyze this membrane fusion event in yeast (Newman et al. 1990; Hardwick and Pelham 1992; Sogaard et al. 1994; Cao and Barlowe 2000). The SNARE protein family is defined by a conserved ∼70-amino-acid heptad repeat sequence, termed the SNARE motif, which is typically adjacent to a C-terminal tail-anchored membrane segment (Rothman 1994; Fasshauer et al. 1998). Cognate sets of SNARE proteins form stable complexes through assembly of their SNARE motifs into parallel four-helix coiled-coil structures (Hanson et al. 1997; Sutton et al. 1998). The close apposition of membranes that follows assembly of SNARE complexes in trans is thought to drive membrane bilayer fusion (Weber et al. 1998). Structural studies of the four-helix bundle reveal that the central or “zero layer” consists of ionic residues such that three of the SNARE proteins contribute a glutamine residue, and are thus termed Q-SNARES, whereas the fourth helix contains an arginine residue, and is known as the R-SNARE (Fasshauer et al. 1998; Sutton et al. 1998). Further refinement of the Q-SNARE proteins based on sequence conservation identifies each as a member of the Qa, Qb, or Qc subfamily (Kloepper et al. 2007). SNARE-dependent membrane fusion is though to proceed through a conserved mechanism in which three Q-SNARES (Qa, Qb, and Qc) and one R-SNARE zipper together from the N-terminal side of the SNARE motif toward the membrane (Sudhof and Rothman 2009). In the case of COPII vesicle fusion with Golgi membranes, Sed5 serves as the Qa-SNARE, Bos1 the Qb-SNARE, Bet1 the Qc-SNARE, and Sec22 the R-SNARE. Furthermore, this SNARE set is sufficient to catalyze membrane fusion when reconstituted into synthetic proteoliposomes (Parlati et al. 2000).

In addition to Sed5, Bos1, Bet1, and Sec22, other regulatory factors are required to control fusion specificity and govern SNARE complex assembly/disassembly. Members of the Sec1/Munc18-1 (SM) family of SNARE-binding proteins regulate distinct SNARE-dependent fusion events (Sudhof and Rothman 2009). The SM family member Sly1 is required for fusion of COPII vesicles with Golgi membrane in yeast (Ossig et al. 1991; Cao et al. 1998). SLY1 was initially identified as a suppressor of loss of YPT1 function when the gain-of-function SLY1-20 allele was isolated in a selection for mutations that permit growth in the absence of YPT1 (Dascher et al. 1991). Sly1 binds directly to Sed5 and increases the fidelity of SNARE complex assembly between Sed5, Bos1, Bet1, and Sec22 compared to noncognate SNARE complexes (Peng and Gallwitz 2002). Crystallographic studies of Sly1 reveal a three-domain arch-shaped architecture that binds a 45-amino-acid N-terminal domain of Sed5 as observed for other SM protein interactions with Qa-SNAREs (Bracher and Weissenhorn 2002). Working models for Sly1 and SM protein function in general are based on multiple binding modes wherein Sly1 initially bound to the N terminus of Sed5 would subsequently bind to other cognate SNARE proteins to regulate assembly and ultimately to act as a clamp in stabilizing a trans-SNARE complex (Furgason et al. 2009; Sudhof and Rothman 2009).

After SNARE-mediated membrane fusion is complete, stable four-helix bundles of cis-SNARE complexes are now present on the acceptor membrane compartment. To recycle assembled Sed5Bos1Bet1Sec22 complexes for use in additional rounds of membrane fusion, the general fusion factors Sec17 and Sec18 catalyze SNARE complex disassembly (Sogaard et al. 1994; Bonifacino and Glick 2004). Sec18 belongs to the AAA family of ATPase chaperones and uses the energy of ATP hydrolysis to separate stable cis-SNARE complexes. Sec17 is thought to recruit Sec18 to SNARE protein complexes and couples ATPase dependent disassembly of cis-SNARE complexes (Bonifacino and Glick 2004). How Sec17/Sec18-mediated disassembly is coordinated with coat-dependent capture of SNARE proteins into vesicles and Sly1-dependent assembly of trans-SNARE complexes during fusion remain open questions.

A concerted model for COPII vesicle tethering and fusion

Although distinct stages in vesicle tethering and fusion can be defined through biochemical and genetic analyses, these are likely concerted reactions in a continuum of events through the early secretory pathway (Figure 5). The multisubunit TRAPPI may serve as an organizational hub on cis-Golgi membranes or vesicles to coordinate vesicle tethering and fusion events. TRAPPI interactions with the COPII subunit Sec23, with the Ypt1 GTPase and potentially with SNARE proteins (Jang et al. 2002; Kim et al. 2006) could link tethering and fusion stages. TRAPPI-activated Ypt1 could recruit Uso1 to Golgi membranes and as COPII vesicles emerge from the ER, Uso1 could forge a long-distance link between newly formed vesicles and acceptor membranes. With tethered vesicles aligned to fusion sites, TRAPPI interactions with vesicle-associated Sec23 and Golgi SNARE machinery would then position vesicles in closer proximity to acceptor membranes. TRAPPI-bound vesicles could transmit signals to the SNARE machinery by direct contact or perhaps through generation of elevated levels of activated Ypt1. The result of such a signal may be to disassemble cis-SNARE complexes or to generate a Sly1Sed5 conformation that promotes assembly of fusogeneic SNARE complexes. Assembly of trans-SNARE complexes would then presumably lead to rapid hemifusion followed by bilayer fusion and compartment mixing.

Traffic Within the Golgi

Transport through the Golgi complex

Newly synthesized secretory proteins arrive at the cis-Golgi in COPII vesicles and after membrane fusion progress through the Golgi complex. Secretory cargo may receive outer-chain carbohydrate modifications and proteolytic processing in a sequential manner as cargo advances through distinct Golgi compartments. For glycoproteins, the N-linked core carbohydrate is extended by addition of α-1,6-mannose residues in the cis-Golgi and by addition of α-1,2- and α-1,3-mannose residues in the medial compartment. Kex2-dependent proteolytic processing of certain secretory cargo occurs in the trans-Golgi compartment. Each of these events can be resolved by blocking membrane fusion through inactivation of the thermosensitive sec18-1 allele (Graham and Emr 1991; Brigance et al. 2000). In support of this sequential organization, distinct Golgi compartments can be visualized through fluorescence microscopy or immuno-EM by monitoring components of the glycosylation and processing machinery (Franzusoff et al. 1991; Preuss et al. 1992; Wooding and Pelham 1998; Rossanese et al. 1999). However, genetic and morphological approaches have not uncovered a vesicle-mediated anterograde transport pathway through distinct compartments of the yeast Golgi complex. Instead, a model of cisternal maturation, in which Golgi cisternae are the anterograde carriers of secretory cargo, is most consistent with a range of experimental observations (Bonifacino and Glick 2004). In the cisternal maturation model, Golgi cisterna containing nascent secretory cargo are formed at the cis-face of the Golgi and mature into a medial and then trans-compartment as resident Golgi glycosylation and processing proteins are dynamically retrieved in retrograde vesicles to preceding cisternae. Indeed, the dispersed organization of Golgi compartments in S. cerevisiae are resolvable by fluorescence microscopy and provided a powerful test of the maturation model through live cell imaging of cis- and trans-Golgi proteins labeled with different fluorescent tags. In such a dual labeled strain, a cis-compartment should be observed to change color to a trans-compartment over the time period required for secretory cargo to transit the Golgi complex. Strikingly, two independent research groups using time resolved high resolution microscopy documented individual cisterna transitioning from early to late compartments in accord with the cisternal maturation model (Losev et al. 2006; Matsuura-Tokita et al. 2006).

In addition to retrograde transport from cis-Golgi to ER (discussed below), the COPI coat is thought to mediate retrograde transport within the Golgi complex to retrieve recycling Golgi machinery to earlier compartments as Golgi cisternae mature (Bonifacino and Glick 2004). In current working models, anterograde-directed COPI vesicles are targeted to preceding Golgi compartments by the conserved oligomeric Golgi (COG) complex, a large multisubunit tethering complex identified through a combination of genetic and biochemical approaches (Miller and Ungar 2012). COG consists of eight subunits and belongs to the larger CATCHR (complex associated with tethering containing helical rods) family of tethering factors that includes the exocyst and GARP complexes (Yu and Hughson 2010). In intra-Golgi retrograde transport, the COG complex appears to operate as a tethering and fusion hub with multiple interactions that link COG to the γ-COPI subunit, to Ypt1 and to Golgi SNARE proteins (Suvorova et al. 2002). More specifically, fusion of retrograde-directed COPI vesicles with cis-Golgi membranes is thought to depend on COG complex interactions with a distinct SNARE complex consisting of Sed5 (Qa), Gos1 (Qb), Sft1 (Qc), and Ykt6 or Sec22 as the R-SNARE (Shestakova et al. 2007). Mutations in COG complex subunits disrupt Golgi transport and glycosylation of secretory cargo, fully consistent with this model. However, at this stage there are no cell-free assays to measure COG-dependent fusion of COPI vesicles to fully dissect underlying molecular mechanisms (Miller and Ungar 2012).

Lipid requirements for Golgi transport

While the protein machinery underlying Golgi transport has received much attention, the role of specific lipid biosynthetic and transfer pathways in Golgi trafficking remain relatively understudied. One of the first connections for a lipid requirement in transport through the Golgi complex was the identification and characterization of Sec14 as an essential phosphatidylinositol/phosphatidylcholine (PI/PC) transfer protein in yeast (Novick et al. 1981; Bankaitis et al. 1989; Cleves et al. 1991). The trafficking blocks associated with Sec14 deficiencies lead to an accumulation of Golgi membranes and Golgi forms of secretory cargo. Sec14 probably does not play a major role in transporting bulk phospholipids but rather is thought to function in regulating phospholipid homeostasis through presentation of PIs to modifying activities such as the PI4 kinases (Schaaf et al. 2008). Interestingly, PI4P levels in the Golgi complex also play a critical role in Golgi structure and function as demonstrated by mutations in the essential PI4 kinase Pik1, which block transport through the Golgi (Walch-Solimena and Novick 1999; Audhya et al. 2000). More recently, a direct requirement for PI4P levels on Golgi organization has been documented through characterization of the Golgi-localized PI4P binding protein encoded by VPS74 (Schmitz et al. 2008; Tu et al. 2008). Loss of Vps74 function results in mislocalization of Golgi mannosyltransferases from early Golgi compartments to the vacuole. Vps74 appears to bind to cytoplasmic sorting signals contained on Golgi resident enzymes and to the COPI coat in addition to PI4P in sorting Golgi-localized proteins into retrograde-directed vesicles. In this manner, PI4P levels and Vps74 may function together in dynamic recycling of Golgi modification enzymes as cisterna containing nascent secretory cargo mature in accord with Golgi maturation models. Indeed, the polarized distribution of PI4P across the Golgi with increasing concentrations from cis- to trans-compartments appears to play several important roles in organization and transport through the Golgi complex (Graham and Burd 2011).

The Return Journey: Retrograde Traffic via COPI Vesicles

Although it remains to this day somewhat controversial as to the precise function (and thus direction) of COPI-mediated vesicular traffic within the Golgi (Emr et al. 2009), the role of these vesicles in retrograde Golgi–ER transport is well established. This is despite the original confusion in the field as to the directionality of COPI-mediated traffic: yeast COPI mutants generally have anterograde trafficking defects that probably stem from indirect effects of blocking retrograde transport rather than impacting forward traffic directly (Gaynor and Emr 1997). Although one COPI component, Sec21, was identified in the original sec mutant screen (Novick et al. 1980), advances in understanding this step of the secretory pathway largely lagged behind and was informed by the biochemical advances made in mammalian systems (Serafini et al. 1991). Once Sec21 was cloned and realized to be an ortholog of the mammalian coatomer complex (Hosobuchi et al. 1992), biochemical analyses allowed the identification of all equivalent yeast subunits, which were in turn also subsequently identified in a variety of genetic screens as additional sec/ret/cop mutants (Duden et al. 1994; Cosson et al. 1996). The major advances in dissecting the mechanisms of retrograde traffic have continued to be led by biochemical approaches (Spang et al. 1998; Spang and Schekman 1998), with many recent high resolution structures of the relevant coat (Lee and Goldberg 2010; Faini et al. 2012; Yu et al. 2012) and tether proteins (Ren et al. 2009; Tripathi et al. 2009). Given the strong homology between the mammalian and yeast proteins, it seems likely that the global structure of the yeast COPI coat is broadly similar to that of mammals (Yip and Walz 2011). Indeed, current approaches make good use of yeast genetics approaches to test functional relevance of the structural data, yielding insight into areas including cargo selection (Michelsen et al. 2007), directionality of vesicle delivery (Kamena and Spang 2004), and coat/tether influences on vesicle fusion (Zink et al. 2009).

Composition and structure of the COPI coat

Originally characterized from mammalian cells as a single coat protomer, or coatomer (Waters et al. 1991), the COPI coat is composed of seven subunits: α-, β-, β′-, γ-, δ-, ε-, and ζ-COP that correspond to the yeast proteins Cop1/Sec33/Ret1, Sec26, Sec27, Sec21, Ret2, Sec28, and Ret3, respectively. Although found as a large cytosolic complex, it is now appreciated that, like the COPII coat, COPI comprises two separable layers: an inner layer that functions in cargo binding composed of γ-, δ-, ζ-, and β-COP and an outer layer formed by α-, β′-, and ε-COP (Figure 3). Furthermore, significant sequence homology was apparent between the inner COPI coat and the adaptor subunits of the clathrin coat system. Indeed, a recent structural analysis of the γ/ζ subcomplex of the inner COPI coat shows clear homology with the α/σ subunits of the AP2 clathrin adaptor, with Arf1 bound at a site that corresponds spatially to the PI(4,5)P2 binding site on AP2 (Yu et al. 2012). Although the structure of the β/δ subcomplex remains to be determined, homology modeling suggests that it adopts a conformation very similar to the β2–AP2 subunit, and biochemical analyses suggest that a second Arf1 molecule can bind to the PI(4,5)P2 binding site on β2–AP2 (Yu et al. 2012). Unlike the inner coat, which is most similar to the clathrin coat adaptors, the outer COPI coat shows homology with both clathrin and COPII coats, with β-propeller and α-solenoid domains forming the building blocks of the putative cage. Structural analysis of stable fragments of the α-/β′-COPI subcomplex supports the concept that the global architecture of the COPI coat is intermediate between that of the COPII and clathrin coats: the individual β-barrel and α-solenoid structures most closely resemble the Sec13/Sec31 structure of the COPII cage but they assemble in a clathrin-like triskelion (Lee and Goldberg 2010). It remains unclear exactly how the inner and outer layers come together, either in solution prior to assembly on the membrane or during vesicle formation, although purified yeast coatomer examined by single particle electron microscopy suggests a somewhat flexible configuration that would need to stabilize during polymerization or oligomerization on the surface of the membrane (Yip and Walz 2011). This concept of structural flexibility for the COPI coat is supported by recent EM analysis of COPI vesicles budded from synthetic liposomes, which showed striking structural diversity of coat arrangement on the surface of the budded vesicles (Faini et al. 2012). Although all the crystallographic, and much of the biochemical analysis of the COPI coat has employed mammalian proteins, the yeast orthologs are highly likely to adopt similar conformations. Indeed, the known structures are consistent with the nonessential nature of Sec28; its ortholog, ε-COP, is a helical structure that interacts with α-COPI but likely does not form part of the cage (Hsia and Hoelz 2010; Lee and Goldberg 2010), probably rendering it dispensable in vivo despite some destabilization of Cop1 (α-COP) in the sec28 mutant (Duden et al. 1998).

Like the COPII coat, COPI assembly on the membrane is initiated by a small GTPase, Arf1, which in addition to the N-terminal amphipathic α-helix also contains a myristoyl group that facilitates membrane anchorage (Antonny et al. 1997a). GDP–GTP exchange on Arf1 and its paralogs makes use of a common structural motif, the Sec7 domain, named for the late Golgi GEF that is the target of the fungal metabolite, Brefeldin A (Sata et al. 1998, 1999). In Golgi–ER retrograde traffic, two redundant GEFs, Gea1 and Gea2, each with a Sec7 domain, likely initiate coat assembly by triggering local recruitment of Arf1 (Peyroche et al. 1996; Spang et al. 2001). Unlike the COPII system, the GAP activity for the COPI coat is not an integral part of the coat itself, but is instead contributed by a separate protein, known (not surprisingly) as ArfGAP1 in mammalian cells. In yeast Arf–GAP activity derives from two distinct proteins, Gcs1 and Glo3, with partially overlapping roles (Poon et al. 1996, 1999). Mammalian ArfGAP1 employs a lipid-packing sensor domain to regulate its activity according to membrane curvature, becoming active on highly curved membranes, likely after vesicle formation has completed or at least progressed enough as to permit Arf release without destabilizing the coat (Bigay et al. 2003, 2005). Yeast Gcs1 also showed a binding preference for conical lipids, suggesting a similar mechanism could regulate GTPase activity of the yeast COPI coat (Antonny et al. 1997b). However, curvature-responsive activity may not be the only mode of regulation of the COPI GTPase cycle. Coatomer itself also seems to influence ArfGAP activity (Goldberg 1999), although the mechanism remains to be fully defined (Luo and Randazzo 2008). Furthermore, the ability of some sorting signals on cargo proteins to inhibit the coatomer-stimulated GAP activity directly links coat recruitment to cargo selection (Springer et al. 1999; Goldberg 2000), an appealing model whereby the coat stably associates with the membrane only when bound to cargo proteins (Springer et al. 1999). Further complicating the problem, is evidence that implicate ArfGAP proteins as positive regulators of the COPI coat rather than negative regulators: overexpression of any of the four yeast ArfGAPs suppressed the lethality of an arf1 mutant (Zhang et al. 1998, 2003). Further yeast experiments also support an active role for Gcs1 and Glo3 in cargo selection, acting on SNARE proteins prior to incorporation into vesicles to promote Arf1 and coatomer interaction (Rein et al. 2002; Schindler and Spang 2007; Schindler et al. 2009). Clearly, the precise role of the GAP in the COPI system remains to be fully understood, complicated by conflicting results from different labs and/or systems and may in fact be multifaceted by serving both positive and negative roles at different stages during the vesicle formation process (Spang et al. 2010).

Cargo capture: sorting signals, cargo adaptors, and coat stimulators

Like other vesicle trafficking events, retrieval of ER resident proteins via COPI vesicles employs sorting signals, most notably the canonical retrieval motifs, HDEL for soluble lumenal cargoes and K(X)KXX for membrane proteins (Figure 4). Soluble proteins bind to a retrieval receptor, Erd2 (Semenza et al. 1990), which couples them to the COPI coat to facilitate retrograde traffic. The COPI coat can discriminate between similar but distinct motifs, including the canonical K(X)KXX, which must be located at the C terminus of the cargo and membrane-proximal to ensure efficient retrieval, R-based motifs that only function when spaced some distance from the membrane surface, and other basic motifs that remain to be fully dissected (Cosson et al. 1998; Shikano and Li 2003). Yeast two-hybrid experiments and subsequent mutagenesis analyses suggest that the R-based motif binds at the interface between the β- and δ-COP subunits (Sec26 and Ret2, respectively), in a manner that is distinct from KKXX binding to the coat (Michelsen et al. 2007). The site of KKXX recognition remains somewhat unclear. Multiple lines of evidence support a role for the α-/β′-/ε-COP complex in KKXX binding (Cosson and Letourneur 1994; Letourneur et al. 1994; Fiedler et al. 1996), whereas direct cross-linking studies implicate the γ-COP subunit in KKXX binding (Harter et al. 1996; Harter and Wieland 1998).

In addition to retrieval motifs based on basic residues, diaromatic retrieval signals have also been identified, perhaps best characterized for the p24 family of proteins, albeit largely using the mammalian family members (Strating and Martens 2009). This class of signal likely binds to the inner COPI coat via the γ-COP subunit, causing a conformational change that may open up the cargo adaptor platform to become receptive to additional cargo clients (Béthune et al. 2006; Strating and Martens 2009). Yet another mode of cargo binding is represented by the SNARE proteins that drive membrane fusion. Unlike SNARE interaction with the COPII coat, direct binding of SNARE sorting signals with COPI components has not been observed. Instead, SNARE incorporation into COPI vesicles depends on the activity of the Arf–GAP, Glo3, although the precise function of Glo3 in promoting a SNARE configuration that is favorable for vesicle capture remains to be fully dissected (Rein et al. 2002).

As with the COPII coat, capture of cargo proteins into retrograde COPI vesicles sometimes requires the action of cargo adaptors. The first of these described was the HDEL receptor, Erd2, described above, where the lumenal domain likely provides ligand-binding function (Scheel and Pelham 1998), with changing pH conditions likely driving binding and release in the appropriate compartments (Wilson et al. 1993). Another well-described cargo adaptor is the membrane protein Rer1 (Nishikawa and Nakano 1993; Sato et al. 1995), which is important for the efficient retrieval, and thus steady-state ER localization, of some ER resident proteins, including the COPII GEF, Sec12, and the translocon components, Sec63 and Sec71 (Sato et al. 1997). The reason these proteins would require an escort back to the ER rather than employing their own retrieval motifs is unclear, but Rer1 seems to bind these clients within their transmembrane domains, via polar residues embedded within the hydrophobic environment (Sato et al. 1996, 2001). Sec12 and Sec71 appear to use different sites on Rer1 to facilitate retrograde traffic, since mutation of the Sec12-binding site had no effect on Sec71 retrieval, suggesting that Rer1 forms a multivalent cargo receptor that has the capacity to bind multiple cargo clients simultaneously (Sato et al. 2003).

Yet another important player in COPI vesicle formation is the class of proteins that seem to serve as coat nucleators, increasing or stabilizing the recruitment of the COPI coat on the Golgi to stimulate retrograde traffic. Although the mechanistic details remain to be fully understood, two classes of protein seem to stimulate retrograde traffic by modulating the ability of the COPI coat to form vesicles. The first description of this function was for a membrane protein, Mst27, which suppresses the lethality of a sec21-1 mutant when overexpressed (Sandmann et al. 2003). Mst27 and its related binding partner, Mst28, both bind to yeast coatomer via KKXX motifs and this function is required for the sec21-1 suppression. Although the endogenous function of Mst27/Mst28 is unclear, the ability of these cargo proteins to stimulate vesicle production was one of the first concrete pieces of evidence that cargo abundance can directly influence vesicle formation. More recently, a similar role has been postulated for the abundant class of p24 proteins; genetic interactions between EMP24 and various COPI components, including SEC21 and the Arf–GAP, GLO3, are suggestive of a functional relationship, and membranes isolated from emp24Δ cells are diminished in their ability to form COPI vesicles in vitro (Aguilera-Romero et al. 2008). Since some of the mammalian p24 proteins showed a capacity to modulate the GTPase activity of the COPI coat (Goldberg 2000), it is tempting to link these observations: by slowing the GTPase activity of Arf1, the COPI coat might be stabilized on the membrane, prolonging the cargo-engagement step and perhaps stimulating coat oligomerization to enhance vesicle production.

Vesicle delivery: DSL-mediated tethering and SNARE-mediated fusion

Like other vesicle trafficking steps, the final stages of delivery of COPI vesicles employ a long-distance tether to bring the vesicle into proximity of the acceptor membrane and SNARE proteins to drive membrane fusion (Spang 2012). The ER-localized tethering complex, the Dsl1 complex, performs the tethering function, recognizing COPI vesicles via their intact coat, and also participates in the fusion event by proofreading the SNARE pairing that occurs prior to fusion (Figure 5). Originally identified as a mutant that was dependent on the presence of the dominant sly1-20 allele, dsl1 mutants showed accumulation of vesicles at restrictive temperature and were suppressed by overexpression of SEC21, although they also showed ER–Golgi transport defects, making a precise function difficult to discern (VanRheenen et al. 2001). Dsl1 forms a complex with Dsl3/Sec39 and Tip20 to form the Dsl1 complex, another member of the CATCHR family of tethering complexes noted for their extended helical rod structures (Lees et al. 2010). Further genetic and biochemical dissection of these proteins converged on a role in retrograde transport from the Golgi to the ER: tip20 and dsl1 mutants showed genetic interactions with a variety of ER–Golgi SNAREs (Sweet and Pelham 1993; Andag et al. 2001; Kraynack et al. 2005), tip20 mutants showed defects in fusion of COPI vesicles (Kamena and Spang 2004), the Dsl1 complex was localized to the ER (Kraynack et al. 2005), and Dsl1 interacts directly with multiple components of the COPI coat (Andag and Schmitt 2003).

Recent structural analyses have generated an appealing mechanistic model by which the extended Dsl1 complex performs three functions by virtue of its ability to interact with both the COPI coat and the fusogenic SNAREs (Ren et al. 2009; Tripathi et al. 2009; Zink et al. 2009). A composite crystal structure suggests that a long stalk, formed largely by Sec39, extends away from the ER membrane, with Dsl1 located at the membrane-distal end to “catch” incoming COPI vesicles via an unstructured loop that would interact directly with the coat via an α-helical structure formed by α- and ε-COPI (Ren et al. 2009; Hsia and Hoelz 2010). Sec39 itself binds to the N-terminal domain of the ER resident SNARE, Use1, via a region that likely lies proximal to the membrane (Tripathi et al. 2009), and Tip20 contains a second SNARE-binding site, interacting with the N-terminal domain of Sec20 (Ren et al. 2009). In addition to binding individual SNAREs, the Dsl1 complex also promotes SNARE assembly and thus may serve two roles in fusion: maintaining individual SNAREs in an unpaired, receptive state, and scaffolding assembly of the fusogenic SNARE complex to promote fusion (Kraynack et al. 2005; Ren et al. 2009). An additional role in vesicle uncoating is suggested by the tendency of vesicles to accumulate en masse under conditions of Dsl1 depletion (Zink et al. 2009): COPI shedding might be assisted by a Dsl1–COPI interaction that would prevent repolymerization of disassembled coat subunits, or could be driven by conformational changes in the Dsl1 complex that would capitalize on the ability of Dsl1 to interact with both the outer α-/ε-COPI domain and a second site on the inner δ-COP subunit to prize the coat from the membrane (Ren et al. 2009; Zink et al. 2009). Indeed, negative stain EM images of the Dsl1 complex suggest a variety of possible configurations, although the mechanistic impact of the different conformations with respect to coat and SNARE binding remain to be tested (Ren et al. 2009). Clearly, the Dsl1 complex is a multifunctional tether that may serve as a useful paradigm for other vesicle “tethering” systems that may contribute to multiple layers of vesicle uncoating, docking, and fusion in addition to their canonical long-distance vesicle trapping function.


Having moved from the “parts list” generated by numerous genetic screens to molecular mechanisms defined by in vitro assays, where is the field currently heading? Emerging questions currently center on how the varied processes that drive protein secretion are coordinated and regulated, both at the molecular level and at the higher-order organizational level. The biosynthesis of secretory proteins can be thought of as a series of simple events (translation/translocation, post-translational modification, chaperone binding, forward transport) but are these events more closely entwined than we currently appreciate? How are protein quality control decisions made: are they a simple outcome of a tug of war between the ER-associated degradation machinery and the forward transport machinery? Adding a dominant ER export signal to a misfolded protein could drive forward traffic (Kincaid and Cooper 2007), but the converse experiment of blocking ERAD of a different misfolded substrate did not lead to its secretion (Pagant et al. 2007). Understanding the interplay between the folding, degradation, and export machineries will be key in appreciating the intricate regulation of secretory protein production and how the different machineries might be coregulated to cope with the changing secretory burden of the cell under different environmental conditions.

Additional questions stem from our relatively poor understanding of how the early secretory pathway is organized and how this organization is maintained. Although it is clear that ER exit sites form discrete subdomains of the ER (Rossanese et al. 1999; Shindiapina and Barlowe 2010), what is the functional significance of this organization? Is the segregation of cargo molecules into different ER exit sites (Muniz et al. 2001) driven by active processes, or does it reflect the passive influence of specific lipid and protein requirements for subsets of cargo molecules? Similarly, do all secretory cargo proteins follow the same route through the Golgi, or are specific itineraries devised for distinct cargoes that might also be driven by specific lipid microenvironments and/or post-translational modification needs? Larger-scale questions also remain: How is the cis-Golgi founded, through homotypic fusion of COPII vesicles, by heterotypic fusion of COPII and COPI vesicles, or by templating from an existing cis-Golgi fragment that expands through delivery of COPII and COPI vesicles? Electron tomography of yeast cells show distinct transport vesicles and Golgi cisternae but no apparent intermediates (West et al. 2011). How are vesicles targeted to the correct destination: Is there a role for the cytoskeleton in vesicle delivery, and how do COPI vesicles that bud from the Golgi find the proper acceptor compartment? Indeed, are there multiple types of COPI vesicles that drive different transport events between different Golgi cisternae and do tubular elements play a role in lipid and protein traffic as they appear to do in mammalian cells? Finally, how are the protein and lipid needs of the cell sensed and maintained to ensure efficient protein secretion, which lies at the heart of cell growth to permit cell division, and how are the rates of anterograde and retrograde traffic balanced to maintain the correct morphology and distribution of the various secretory organelles? As in the past, the facile genetics and accessible biochemistry of the yeast system still hold promise in answering these questions, with the development of new tools serving to strengthen the field and provide new avenues for further exploration.


  • Communicating editor: T. Davis

  • Received June 14, 2012.
  • Accepted September 25, 2012.

Literature Cited