Cells face a complex problem: how to transfer lipids and proteins between membrane compartments in an organized, timely fashion. Indeed, many thousands of membrane and secretory proteins must traffic out of the ER to different organelles to function, while others are retrieved from the plasma membrane having fulfilled their roles [Nat. Rev. Mol. Cell Biol. (2013) 14, 382–392]. This process is highly dynamic and failure to target cargo accurately leads to catastrophic consequences for the cell, as is clear from the numerous human diseases associated with defects in membrane trafficking [Int. J. Mol. Sci. (2013) 14, 18670–18681; Traffic (2000) 1, 836–851]. How then does the cell organize this enormous transfer of material in its crowded internal environment? And how specifically do vesicles carrying proteins and lipids recognize and fuse with the correct compartment?
Transport begins with the selection of cargo and budding of coated vesicles or carriers from donor membranes. These are then chaperoned through the cytoplasm by cytoskeletal machinery until they encounter acceptor membranes. At some point, uncoating occurs and vesicles are loosely docked or ‘tethered’, before SNARE proteins on opposing membranes interact. This promotes fusion leading to release of vesicular contents and integration of membrane proteins and lipids into the recipient organelle [1,2]. With the exception of the SNARE proteins, each of these steps requires the recruitment of peripheral membrane proteins that transiently generate ‘identity’ either on the donor or acceptor membrane before being turned over or recycled for another round of transport [3,4]. Work from numerous laboratories has led to the dissection of the membrane-trafficking steps described above and many of the important players have been identified; however, until recently, the process of ‘tethering’ was poorly understood. This mini-review will describe what we now know about this key step and will focus, in particular, on the coiled-coil family of vesicle tethers at the Golgi.
Located at the cytosolic surface of specific organelles, tethering proteins ensure that only the correct vesicles dock and deliver cargo to designated sites. Tethers can be broadly divided into two categories: multi-subunit tethering complexes and coiled-coil proteins [5,6]. Coiled-coil tethers are long, dimeric proteins containing extended regions of coiled-coil interspersed with unstructured, breaks or ‘hinges’. They are found mainly on the Golgi, with the exception of EEA1 on early endosomes [7,8] and TANGO1 on the ER-to-Golgi Intermediate Compartment . Multi-subunit tethering proteins, although exhibiting similarities with each other, are more diverse. They contain differing numbers of subunits and are located on organelles throughout the secretory and endocytic pathways [10,11]. Multi-subunit tethers are proposed to act more proximal to the membrane than coiled-coil proteins that may, based on their sequences, extend several hundred nanometers into the cytosol.
The majority of vesicle-tethering proteins are peripheral membrane proteins that are targeted to membranes by association with small GTPases of the Rab and Arf/Arl families . This means that they can cycle on and off organelles at specific sites, allowing them to contribute to the establishment of membrane micro-domains that may be regulated in response to changes in cellular demand. Both groups of tether interact with additional components of the membrane-trafficking machinery and with signaling molecules, including SNAREs, kinases, coat proteins, motor proteins and small GTPases, and are therefore implicated in processes besides tethering [13,14].
Central to both the secretory and endocytic pathways lies the Golgi apparatus, a cisternal ribbon-like organelle in mammals that receives vesicles from both the ER and the endocytic system, while co-ordinating the intra-Golgi vesicular trafficking of resident proteins between stacks [15,16]. With this ability to receive vesicles from so many different destinations, it is perhaps not surprising that the Golgi is decorated with an unusually large number of coiled-coil tethering proteins, collectively termed golgins . Golgins are ubiquitously expressed, hydrophilic proteins, with at least five existing in the last eukaryotic common ancestor, and are required both for vesicle tethering and for the maintenance of Golgi structure [6,18]. They decorate specific cisternae, where they are proposed to play distinct yet overlapping roles. The observation that besides the cis-Golgi tether, p115, ablation of individual golgins is not cell lethal, has led to the proposal that golgins have overlapping functions in vesicle tethering . Complex organisms have a larger complement of golgins, perhaps reflecting their need to regulate additional tissue or cell-specific trafficking pathways.
In mammals, the golgins found at the cis-Golgi comprise GM130 and GMAP-210, those at the Golgi rims are giantin, CASP, golgin-84 and TMF and those at the trans-Golgi are GCC88, GCC185, golgin-97 and golgin-245 (Figure 1). GM130 is localized by interacting with the scaffolding protein GRASP65 and also binds activated Rab1 [19–21], while GMAP-210 binds to Arf1-GTP via its C-terminal ‘GRAB’ domain . CASP, golgin-84 and giantin all have transmembrane domains that restrict them to the Golgi rims although exactly how this segregation within the membrane arises remains to be determined [23–25]. TMF binds to active Rab6 , while golgin-97, golgin-245, GCC88 and GCC185 localize via their ‘GRIP’ domains that bind Arl1-GTP at the trans-Golgi [27–29], with active Rab6 implicated in stabilizing a rather weak Arl1–GCC185 interaction, although the Rab6 requirement is debated [30,31]. All golgins therefore adopt a similar topology with a membrane-binding domain coupled to extensive coiled-coils that are thought to extend the N-terminus into the cytosol to capture in-coming vesicles. Despite their structural similarity, individual golgins lack sequence homology. Indeed, regions of high sequence conservation tend to be confined to their N- and C-termini. This means that golgins from different species vary in length, and has led to the suggestion that the bulk of their coiled-coil is present to act as a spacing unit .
Coiled-coil proteins at the Golgi.
Whether other Golgi localized coiled-coil proteins should be included in the golgin family remains unclear. Some, such as golgin-45, are rather small, while others have an atypical topology like golgin-160 that is localized via an N- rather than C-terminal Golgi-targeting domain [33,34]. Moreover, p115, a well-characterized tethering protein, has a large globular N-terminus consisting of armadillo repeats that is not found in other golgin members [35,36]. In contrast, golgin-104 (CCCP-1 in Caenorhabditis elegans), a newly described trans-Golgi coiled-coil protein implicated in regulating dense core vesicle biogenesis, may be included in the golgin family, having a C-terminal membrane-targeting domain and Rab2-binding site located within the same region [37–39]. The human genome also encodes additional uncharacterized proteins that are predicted to contain extensive regions of coiled-coil, and so, the list of golgins is likely to expand .
Coiled-coil proteins capture vesicles
The first evidence that a Golgi localized coiled-coil protein could act as a vesicle tether came in the late 1990s when Uso1, the yeast homolog of p115, was shown to promote docking of purified COPII vesicles onto isolated Golgi membranes [41,42]. This was followed by experiments, demonstrating that different golgin combinations could selectively capture subpopulations of COPI vesicles, with dimers of GM130-p115 and golgin-84-CASP interacting with vesicles containing different cargoes . Later, GMAP-210 was shown to mediate asymmetric tethering of highly curved liposomes to flatter membranes due to GMAP-210's intrinsic ability to sense differences in membrane curvature . These pioneering studies utilized elegant in vitro assays with purified components and liposomes and were focused on events occurring at the early Golgi.
In vivo analysis of the majority of golgins, however, lagged behind. One reason is that depletion of individual golgins by RNA interference leads to modest membrane-trafficking defects or to ambiguous outcomes such as Golgi fragmentation [26,45–47]. In addition, studies of human or animal models show that disease mutations in golgins are associated with defects in specific tissues rather than lethality [12,48]. These observations are thought to be due to the ability of golgins to compensate for each other, such that the loss of one protein can be alleviated by the collective action of others. This has made dissecting the role of individual proteins recalcitrant to analysis and, for quite some time, the premise that all golgins were vesicle tethers was inferred from the in vitro analysis of a limited number of early Golgi proteins and on their attractive topology.
Recently, the function of mammalian golgins has been revisited using an assay that couples their ectopic relocation with microscopic analysis of cargo proteins . This in vivo approach tests the sufficiency of golgins to capture vesicles over their necessity and is based on appending a mitochondrial-targeting sequence to their C-terminus that directs the golgins away from the Golgi and exploits the fact that SNARE proteins, which are required for Golgi–vesicle fusion, are not present . Thus, the normally transient vesicle-capturing step persists and can be analyzed. The resulting data show that, of the golgins tested, GM130, GMAP-210, golgin-84, TMF, GCC88, golgin-97 and golgin-245 all support vesicle tethering at the mitochondria when expressed individually. Moreover, the interactions are specific and different tethering proteins can capture transport carriers containing specific cargoes. Indeed, the location of the golgin within the stack dictates the type of vesicle that is captured at the ectopic site. Figure 1 shows the endogenous location of each of the golgins and how this relates to the specific classes of vesicles that they capture.
Three of the golgins tested failed to capture vesicles when located to the mitochondria. These are CASP, giantin and GCC185. For each, there is evidence that they play a role in vesicle transport at the Golgi. For example, cells lacking Coy1p, the yeast homolog of CASP, show a reduction in intra-Golgi transport, while depletion of GCC185 results in defects in endosome-to-Golgi trafficking [50–52]. Several theories can explain why these golgins fail to capture vesicles at the relocation site. First, they do not act as tethers per se, but play other important functions in vesicle transport such as the maintenance of Golgi structure or organization of microtubules. These would also lead to failure to accurately transport cargo. GCC185, for example, binds to the microtubule-associated CLASP proteins, and these were re-located to the mitochondria [49,53]. Second, the assay is performed in the presence of endogenous golgins, which may compete for vesicle binding with the exogenous proteins. Third, additional integral membrane proteins that are not re-routed to the mitochondria may be required to stabilize the golgin–vesicle interaction, or the binding to vesicles is regulated and is therefore not sustained long enough for analysis. Finally, since two of the proteins that fail to capture are integral membrane proteins, perhaps, the specific lipid environment of the Golgi is essential for vesicle capture by these proteins.
Vesicles are hooked by the N-terminus
As detailed above, tethering models predict that while the golgin C-terminus is responsible for membrane targeting, the N-terminus is ideally placed to capture vesicles. Evidence for this comes from work with GMAP-210, showing that the protein contains a specialized N-terminal amphipathic lipid-packing sensor or ‘ALPS’ motif within the first 38 amino acids. This motif selectively interacts with the highly curved membranes of vesicles and is proposed to act as a vesicle filter at the entrance to the Golgi [44,54]. Although this is an attractive model, other tethering mechanisms for GMAP-210 must exist in species outside of vertebrates since the ALPS motif is either missing or much shorter . Despite its failure to capture vesicles when re-located to mitochondria, the N-terminus of GCC185 can bind membranes in vitro . Atomic force microscopy shows that this region of the protein adopts a splayed conformation . This is also a feature of the GM130 dimer with electron micrographs showing that the N-terminus can switch between an open Y- and a closed I-shaped conformation . The topology of the N-termini suggests a model where golgins in the ‘open’ conformation can embrace in-coming vesicles.
Recently, mitochondrial relocation of golgins, combined with serial truncation and domain swapping experiments, has shown that the majority of golgins utilize a short N-terminal peptide (ranging from 21 to 49 amino acids) for vesicle capture . Several of the motifs share conserved residues, suggesting that they act through related mechanisms. For example, golgin-97 and golgin-245 located at the trans-Golgi share a highly conserved 21 amino acid vesicle-capturing motif that is distinct from GCC88, although all three golgins can capture endosome-derived carriers with overlapping cargoes. This implies that either there are two different vesicle populations that traffic between endosomes and the TGN that contain related cargo or that the same vesicles can be captured by two different mechanisms. Likewise, the N-termini of GMAP-210, golgin-84 and TMF, all of which capture intra-Golgi vesicles, share common features. For GMAP-210, this is in addition to the ALPS motif that is also present within this region. GMAP-210 may therefore capture two distinct types of vesicles by different mechanisms  (Figure 1).
How do the golgin N-terminal motifs select the correct vesicles? Are there receptors that specify vesicle populations? This has been addressed for two of the GRIP domain golgins, golgin-97 and golgin-245, by combining ectopic relocation to mitochondria with proximity labeling . These experiments identified TBC1D23, a catalytically inactive Rab GAP, as a protein that is unique to vesicles captured by these two golgins. Indeed, TBC1D23 binds directly to the tips of golgin-97 and golgin-245 via its N-terminus and to FAM21, a subunit of the WASH complex, via its C-terminus. It is proposed to act as a bridging factor between golgins at the trans-Golgi network (TGN) and the WASH complex on endosomally derived vesicles  (Figure 1, gray box). This, of course, opens up many additional questions. Since the WASH complex is known to play roles in vesicle trafficking both to the TGN and to the plasma membrane, how does binding this complex ensure that the correct subpopulation of vesicles is selected? Does the FAM21 subunit, a scaffolding platform for a variety of trafficking molecules including the retromer complex and sorting nexins, also interact with other unique TGN markers? Does TBC1D23 arrive with vesicles or is it already recruited to the Golgi by the golgins and how is this controlled? Are there adaptors for GRIP domain proteins besides TBC1D23? Indeed, there is potential for alternative adaptors that could control cell-, tissue- or even species-specific outcomes. One may ask why there is the need for such complex tethering machinery? Large protein assemblies seem to be a general theme in membrane trafficking. For example, the active zone at the pre-synaptic plasma membrane requires a large cohort of proteins to control synaptic vesicle docking including coiled-coil proteins such as ELKS (also known as CAST/Rab6IP2) [60,61]. This level of complexity may allow tethering to be finely controlled or allow tethers to span large volumes in a crowded environment.
After the initial capture
Once vesicles are captured at the tips of golgins, how are they delivered to the membrane for SNAREs to engage? One proposal is that golgins act as a ‘tentacular meshwork’ that captures vesicles and allows them to percolate through by a series of weak interactions until they reach the membrane . This does not imply that golgins are entirely rigid molecules and numerous reports have suggested that breaks in the coil could allow golgins to flex or bend [12,57,63,64]. Indeed, a completely static tether would be futile since the vesicle would be trapped distal to the membrane. A second feature of a tentacular model is that it prevents the ‘wrong’ vesicles from reaching the membrane as they must navigate through a forest of golgins that are in the way. This is an important point as vesicles destined for the Golgi arrive via cytoskeletal tracks, and so, it may be necessary to move them from the initial point of contact to the exact spot where fusion occurs.
In those cases where biophysical measurements have been made, golgins tend to be shorter than predicted by computational algorithms. For example, the length of giantin determined by sedimentation analysis is 250 nm while when fully extended it has the potential to span >500 nm . Likewise, GM130 adopts two conformations of differing length, with the length of the I-shaped molecule shorter than that predicted theoretically . On the other hand, electron micrographs show a ribosome-free exclusion zone around the Golgi and there is at least one report of vesicles in the vicinity contacted by long elongated structures [65–67]. Many experiments have recently been performed to test whether the coiled-coil regions of golgins are important for function and for flexibility. First, a truncated version of GMAP-210 known as ‘mini-GMAP’ encoding only the N- and C-terminal membrane-targeting domains and ∼30% of the coiled-coil was shown to support liposome tethering in vitro . However, the same protein could not rescue the in vivo functions of GMAP-210, indicating that additional features of the coiled-coil are important . Likewise, a truncation of GCC185 that removes the hinge regions results in a protein that fails to efficiently traffic mannose-6-phosphate receptors back to the Golgi. Analyzing the structure of full-length GCC185 reveals a centrally located ‘bubble’ around the hinge that is missing in the truncated version. This region allows the protein to ‘collapse’ bringing the N-terminus closer to the membrane . Since all golgins, including those with transmembrane domains, are Rab effectors, binding of GTPases may facilitate these transitions in vivo . Indeed, the Rab2-binding site in GMAP-210 is known to be important for some aspects of its function in cells . EEA1 is a coiled-coil protein that can tether endosomes through the interaction of its N-terminal Rab5-binding domain with one membrane and its C-terminal lipid-binding domain with another. Rotary shadowing EM of EEA1 reveals a filamentous molecule of ∼200 nm that upon the addition of Rab5 and a non-hydrolysable analog of GTP reduces in mean length to ∼120 nm. This is proposed to be due to Rab5-induced entropic collapse . Exactly how binding of Rab5 causes the coiled-coil to relax is intriguing; however, GTPases are ideal molecules to mediate such transient changes as they themselves are readily turned over by Rab GAPs. Thus, hydrolysis of bound GTP could be coupled with dissociation from the tether, release of the vesicle and return of the tether to its original conformation for another round of capture. Changes in golgin conformation may allow vesicles to move from longer to shorter tethers or to bind to SNARE proteins. In keeping with this, golgins are often found associated with other membrane-trafficking components. For example, both golgin-84 and TMF interact with subunits of the COG complex, a multi-subunit tethering protein; while p115 not only binds SNARE proteins but is also important for promoting SNAREpin assembly [70–73].
The evidence that specific coiled-coil proteins are vesicle tethers is now conclusive. Moreover, they deliver a degree of specificity to the membrane-trafficking process. The next steps will be to characterize more of the receptors and/or lipids on vesicles that are recognized by individual golgins and to better understand the transition from tethered state to SNARE engagement and fusion.
amphipathic lipid-packing sensor
conserved oligomeric Golgi
coat protein complex II
early endosome antigen 1
GTPase activating protein
soluble NSF attachment protein receptor
transport and Golgi organisation protein 1
TATA element modulatory factor
Wiskott-Aldrich syndrome protein and Scar homologue
The Author declares that there are no competing interests associated with this manuscript.