Golgi Compartmentation and Identity

Abstract

Recent work supports the idea that cisternae of the Golgi apparatus can be assigned to three classes, which correspond to discrete stages of cisternal maturation. Each stage has a unique pattern of membrane traffic. At the first stage, cisternae form in association with the ER at multifunctional membrane assembly stations. At the second stage, cisternae synthesize carbohydrates while exchanging material via COPI vesicles. At the third stage, cisternae of the trans-Golgi network segregate into domains and produce transport carriers with the aid of specific lipids and the actin cytoskeleton. These processes are coordinated by cascades of Rab and Arf/Arl GTPases.

Introduction

The Golgi apparatus is fascinating but enigmatic. A working assumption is that operational principles of the Golgi are conserved between species, and that model organisms can serve to illuminate these principles [1, 2]. From this perspective, the goal is to synthesize diverse lines of evidence into a “generic” picture of how the Golgi works.

Such a multifaceted approach has been applied to understanding the flow of membrane and cargo through the Golgi [3, 4]. Perhaps the most widely embraced model for Golgi traffic is cisternal maturation, which proposes that cis-Golgi cisternae arise de novo, then progressively mature into trans-Golgi cisternae while carrying secretory cargoes forward, then ultimately disintegrate into transport carriers at the trans-Golgi network (TGN) stage. As the cisternae mature, COPI vesicles are thought to recycle resident Golgi membrane proteins. Maturation of Golgi cisternae has been directly observed in Saccharomyces cerevisiae, which has non-stacked Golgi cisternae that can be optically resolved [5, 6]. In plant and animal cells, the ability of large cargoes to transit through the Golgi provides some of the clearest evidence that Golgi cisternae serve as anterograde carriers [7, 8]. However, experimental and theoretical arguments suggest that other data are inconsistent with a simple maturation mechanism [9 •, 10], and alternatives such as the cisternal progenitor model and the kiss-and-run model have been proposed [11, 12], so this issue remains unresolved.

A pair of recent studies tested the cisternal maturation model in mammalian cells [13 ••, 14 ••]. Both groups used transmembrane proteins fused to a multiple copies of a dimerizing mutant of FK506-binding protein (FKBP). In one case, an FKBP-tagged secretory cargo self-associated to form “staples” that linked the two faces of a cisterna together [13 ••]. Those staples did not progress through the Golgi, consistent with the idea that Golgi cisternae are long-lived compartments. In the other case, an FKBP-tagged resident Golgi protein aggregated in the plane of the membrane [14 ••]. Those aggregates moved forward through the Golgi, and when they were subsequently dissolved, the tagged Golgi proteins recycled to earlier cisternae, consistent with the idea that Golgi cisternae carry large cargoes forward while recycling resident proteins. The discrepancies between the two studies presumably center on technical details. Resolution of this debate will be an important step toward a unified picture of Golgi traffic.

Our review focuses on Golgi compartmentation and identity. These issues are closely linked to Golgi traffic, and the cisternal maturation model will serve as the basis for the discussion, but we will highlight open questions.

The Golgi as a tripartite organelle

In both stacked and nonstacked Golgi organelles, the cisternae vary in composition [15, 16]. Mammalian cells contain multiple Golgi cisternae plus the ER-Golgi intermediate compartment (ERGIC) [17]. These findings raise the question of how many fundamentally different types of compartments exist in the Golgi. There are compelling reasons to distinguish between the TGN and earlier cisternae [18, 19], and an early review suggested that the Golgi is actually a tandem pair of organelles [20], which would now be designated the Golgi stack and the TGN. However, differences can also be seen among cisternae that precede the TGN [21, 22], leading to the suggestion that the Golgi consists of four or more different kinds of compartments [15, 23]. This uncertainty highlights a major gap in our understanding.

We recently updated an old concept [19] by postulating that Golgi cisternae can be divided into three classes, which represent discrete stages of maturation [24]. Each stage is defined by the membrane traffic pathways to and from the cisternae (Fig. 1). At the “cisternal assembly” stage, which includes the ERGIC and the cis-Golgi cisternae, ER-derived COPII vesicles deliver secretory proteins to the nascent cisterna, while vesicles of the COPIa type bud from the nascent cisterna to recycle trafficking components and resident ER proteins back to the ER. At the “carbohydrate synthesis” stage, which includes the medial and trans-cisternae, vesicles of the COPIb type transfer resident Golgi proteins between cisternae. This transfer is presumably nondirectional for each individual vesicle, but as young cisternae acquire the ability to receive COPIb vesicles while old cisternae lose this ability, COPIb vesicles mediate a net recycling of resident Golgi proteins from old to young cisternae. Finally, at the “carrier formation” stage, which corresponds to the TGN as traditionally defined, the cisternae produce clathrin-coated vesicles and secretory carriers while receiving incoming traffic from endosomes. Within a given stage, cisternae can evolve in terms of their resident protein composition, but they share a common machinery for membrane traffic. Current knowledge of the Golgi traffic machinery is consistent with the three-stage model [24].

Figure 1.

We propose that the Golgi can be divided into three stages of maturation. During the Cisternal Assembly stage, COPII vesicles bud from the ER and fuse with one another to generate a new Golgi cisterna, while COPIa vesicles bud from this nascent cisterna to recycle trafficking proteins and resident ER proteins to the ER. This stage includes cis-Golgi cisternae, as well as ERGIC membranes in animal cells. During the Carbohydrate Synthesis stage, most of the oligosaccharide remodeling reactions take place, and resident Golgi proteins are transferred between cisternae by COPIb vesicles. This stage includes cisternae that are often designated medial and trans. During the Carrier Formation stage, the cisterna disintegrates to produce cargo carriers that include clathrin-coated and secretory vesicles. This stage includes cisternae that are traditionally designated TGN. Adapted from [24].

It remains to be seen whether this view of the Golgi as a tripartite organelle will endure. The classification of Golgi cisternae has been an elusive quest, and indeed, the names of the three stages shown in Fig. 1 are imperfect. For instance, even though carbohydrate synthesis occurs mainly at the second stage, Golgi glycosylation in yeast is initiated in the earliest Golgi compartments [23], and terminal glycosylation in mammalian cells can occur in the TGN [25]. Yet diverse eukaryotes may all have three types of Golgi cisternae as defined by their membrane traffic machinery. According to this hypothesis, an understanding of Golgi compartmentation and identity will rely on deciphering the mechanisms that generate the three stages.

To the Golgi: the cisternal assembly stage

Golgi cisternae are thought to assemble by the homotypic fusion of COPII vesicles, which are produced at long-lived domains known as transitional ER (tER) sites or ER exit sites [26]. Intermediates in this cisternal assembly pathway were visualized by electron tomography of plant and algal Golgi stacks [27 ••]. An assembling cisterna produces COPIa vesicles that recycle trafficking components to the ER [26, 28]. These retrograde carriers may be targeted to tER sites [29 •, 30]. Regardless of whether tER sites are unidirectional or bidirectional portals, they are clearly associated with early Golgi or ERGIC membranes in diverse eukaryotes [17, 31, 32].

A long-standing mystery is how COPII vesicles become clustered at tER sites. Normal tER organization requires Sec16, a peripherally membrane-associated COPII binding protein [33-35]. Sec16 has been proposed to act as a scaffold that clusters COPII components at tER sites [34-37]. However, recent data challenge this notion, suggesting instead that Sec16 is recruited by individual COPII vesicles as a regulator of COPII coat dynamics [37, 38 •, 39, 40]. When Sec16 was inactivated in the yeast Pichia pastoris, tER sites became smaller and more dynamic, but they persisted [38 •]. The implication is that a Sec16-independent mechanism establishes tER sites. This mechanism might involve the tethering of nascent and budded COPII vesicles to adjacent early Golgi or ERGIC membranes [41] (Fig. 2). If multiple COPII vesicles are all tethered to the same membrane structure, these vesicles will naturally cluster. Some of the tethering events may be controlled by cycles of phosphorylation and dephosphorylation of COPII coat subunits [42 •]. Studies of Golgi stack formation suggest that tethering proteins help to nucleate cisternal assembly [43, 44]. The tethering model postulates that tER sites form in conjunction with early Golgi or ERGIC membranes by an integrated self-organization pathway [41].

Figure 2.

The ER contains multifunctional membrane assembly stations. COPII vesicles bud from tER sites, which may be generated by the tethering of nascent COPII vesicles to adjacent cis-Golgi or ERGIC membranes. A subset of the tER sites are associated with preautophagosomal structures.

The tER plus the early Golgi/ERGIC seems to be part of a larger structure that includes nascent autophagosomes (Fig. 2). Starved yeast cells contain a compartment for unconventional protein secretion (CUPS) that associates with tER sites and contains autophagy-related proteins [45], and pre-autophagosomal structures are functionally and physically linked to tER sites in yeast and mammalian cells [46 ••, 47 •, 48 ••]. COPII vesicles evidently contribute to forming these structures [49]. Moreover, the yeast Ypt1 GTPase, which is implicated in ER-to-Golgi transport, has an additional autophagy-related function at the ER [50 •, 51 ••, 52 •]. Thus, Golgi cisternae may form together with other compartments at a multifunctional membrane assembly station (Fig. 2).

Through the Golgi: the carbohydrate synthesis stage

After cisternal assembly is complete, the main task is to synthesize and remodel carbohydrates. Conversion of a Golgi cisterna to the carbohydrate synthesis stage involves multiple changes: a loss of the ability to receive COPII vesicles, a transition from producing COPIa to COPIb vesicles, and a gain of the ability to receive COPIb vesicles (Fig. 1). At the carbohydrate synthesis stage, COPIb vesicles are thought to exchange resident Golgi glycosylation enzymes between cisternae, thereby generating gradients of enzyme composition across Golgi stacks [53]. However, fundamental questions remain about intra-Golgi traffic at this stage.

One question concerns the molecular distinction between COPIa and COPIb vesicles. This terminology comes from morphological studies of the plant secretory pathway [22], but studies of mammalian cells also provide evidence for two classes of COPI vesicles [54-57]. In mammalian cells, COPI subunit isoforms apparently interact with specific partners during different COPI-dependent pathways [58 •]. More generally, the properties of a COPI vesicle may be dictated by the composition of the parental compartment, with the same basic machinery generating either COPIa or COPIb vesicles at different stages of maturation.

A second question concerns the directionality of COPIb vesicles. If these vesicles move one step at a time from older to younger cisternae, the observed gradients of resident Golgi proteins across the stack can be conveniently explained [53]. However, no mechanism for such orderly movement has been described. A plausible alternative is that COPIb vesicles “percolate” stochastically between cisternae at the carbohydrate synthesis stage [59]. To explain how stochastic vesicle movement generates gradients of protein composition across the stack, one can postulate that time-dependent changes in cisternal pH or lipid composition kinetically trap resident Golgi proteins in cisternae of a particular age [24, 60].

Finally, a third question is whether COPI vesicles contain secretory cargoes. If so, vesicle-mediated transport could provide an anterograde intra-Golgi transport pathway that is complementary to cisternal maturation [61]. The literature on this topic has been contradictory [4, 62], but a study of transfer between Golgi stacks in fused cells suggested that COPI vesicles could carry both resident Golgi enzymes and small secretory cargoes [63 •]. This finding supports the idea that unlike COPII vesicles, COPI vesicles can act as long-range carriers.

Some abundant secretory proteins such as albumin and insulin are not glycosylated, so cells could conserve resources by enabling these proteins to move quickly through the carbohydrate synthesis stage. The Golgi stacks of actively secreting mammalian cells contain narrow tubular connections between heterologous cisternae [64-66], and such tubules could permit the rapid anterograde transport of small soluble cargoes [67, 68]. Intriguingly, the COPI machinery can cooperate with lipid modifying enzymes to generate either vesicles or tubules [69]. Heterologous tubular connections between cisternae have been convincingly demonstrated only in mammalian cells, so this mechanism might be a specialization rather than a conserved property of the Golgi.

In a Golgi stack, the number of cisternae at the carbohydrate synthesis stage varies from one or two to a dozen or more, depending on the cell type [27 ••]. The mechanisms that regulate the number of cisternae per stack are obscure. However, a study of S. cerevisiae revealed that altering the kinetics of cisternal maturation influenced the copy number of Golgi cisternae, suggesting that an analogous process operating in a stacked Golgi could modulate the number of cisternae per stack [70 •].

What holds the cisternae together in a stacked Golgi? The most prominent candidates have been peripheral membrane proteins of the GRASP family [71]. GRASP proteins promote homotypic interactions between apposing membranes [72-74]. However, plant cells have Golgi stacks but lack GRASP proteins, and deletion of the P. pastoris GRASP homolog Grh1 did not perturb Golgi stacking [31]. The results with mammalian cells have been conflicting, with some researchers finding that GRASP inactivation disrupted Golgi stacking while others saw effects only on the lateral linking of Golgi stacks [71, 75]. It seems likely that cisternal stacking depends on weak adhesive interactions that involve a variety of proteins in different cell types. For example, in mammalian cells, GRASPs seem to cooperate with coiled-coil “golgin” proteins to mediate cisternal stacking [76]. Golgi stacking has been repeatedly lost during eukaryotic evolution [77], implying that this structural feature is dispensable for the basic operation of the Golgi.

From the Golgi: the carrier formation stage

At the final stage, Golgi cisternae produce transport carriers that deliver secretory cargo proteins to plasma membrane domains, regulated secretory granules, and endosomes or lysosomes/vacuoles. These sorting events occur in cisternae that are traditionally designated the TGN [19, 78]. The definition of the TGN is sometimes expanded to include earlier cisternae [79], but we will treat the TGN as being synonymous with the carrier formation stage. TGN cisternae differ from younger cisternae in that they produce clathrin-coated vesicles but do not produce or receive COPI vesicles [18, 24]. Moreover, only TGN cisternae receive membrane traffic from endosomes [80]. In mammalian cells and P. pastoris, TGN cisternae peel off from the Golgi stack, and in plant cells, the TGN is often fully dissociated from the Golgi stack [22, 81-83], implying that stacking interactions are reduced or eliminated during the carrier formation stage.

The transition from the carbohydrate synthesis stage to the carrier formation stage seems to be abrupt. Electron tomography of the mammalian Golgi revealed a sharp morphological divide in which the trans-most cisterna produced only clathrin-coated vesicles while earlier cisternae produced only COPI vesicles [18]. Moreover, video microscopy of the S. cerevisiae Golgi showed a rapid conversion in which cisternae lost Vrg4, which participates in carbohydrate synthesis, while acquiring Sec7, which participates in clathrin-coated vesicle formation [5]. These results support the idea that Golgi maturation involves “quantized” transitions between stages.

TGN cisternae segregate both resident proteins and secretory cargoes into membrane domains, which can generate multiple types of transport carriers [78, 84]. These processes rely on the localized metabolism of specific lipids [85-87], and on the regulated flipping of phospholipids between the two leaflets of the membrane [88]. A lipid crucial for the carrier formation stage is phosphatidylinositol-4 phosphate (PI4P), which binds multiple effectors and helps to coordinate exocytic vesicle formation [85, 86, 89, 90]. TGN domain formation also involves luminal Ca²⁺ gradients generated by interactions of calcium pumps with the actin cytoskeleton [91 •, 92 ••]. Thus, a variety of processes contribute to the functional segregation of the TGN.

A puzzle concerns how proteins are localized to the TGN. As a cisterna matures from the carbohydrate synthesis stage to the carrier formation stage, it receives peripheral and transmembrane TGN proteins from maturing TGN cisternae [5, 93 ••]. Moreover, mutant forms of some plasma membrane proteins accumulate in the TGN [94, 95 •], consistent with the existence of active recycling from older to younger TGN cisternae. Such a recycling pathway could explain why secretory proteins exit the Golgi with exponential rather than linear kinetics [10]. Components that might play a role in recycling from older to younger TGN cisternae include clathrin-coated vesicles containing the AP-1 adapter [96, 97] and a subcomplex of the COG tether [98]. Much remains to be learned about membrane dynamics at the carrier formation stage.

Role of GTPases in Golgi compartmentation and identity

Golgi-associated GTPases of the Rab and Arf/Arl families are likely to coordinate multiple events: the recruitment of resident Golgi proteins at each stage of maturation, the transitions between stages, and the partitioning of Golgi cisternae into membrane domains. This topic was recently reviewed in depth [99], so we will highlight only a few key points.

Rab GTPases, many of which have the prefix “Ypt” in yeast, act throughout the secretory pathway. They regulate the tethering of vesicles to their target compartments and the recruitment of motor proteins to secretory vesicles. As a consequence of these and other activities, Rab GTPases help to define the identities of Golgi cisternae [100-102]. It has been proposed that Rab proteins also define membrane domains within individual cisternae [12]. An appealing hypothesis is that Rab proteins operate in cascades, with a given Rab recruiting the guanine nucleotide exchange factor (GEF) for the next Rab, which in turn recruits a GTPase activating protein (GAP) that inactivates the first Rab [99, 101, 102]. Such a Rab cascade could drive the conversion from one Golgi stage to the next. Evidence was presented for a cascade at a late stage of the yeast Golgi, with the late-acting Ypt32 recruiting a GAP for the early-acting Ypt1 [103]. Ypt32 also recruits a GAP for Ypt6, which acts at intermediate and late stages of the yeast Golgi [104 ••]. In mammalian cells, the Rab33B GTPase in the medial Golgi recruits a GEF for the trans-Golgi/TGN protein Rab6, which is the homolog of yeast Ypt6 [105 •]. Thus, Rab proteins seem to be central players in choreographing Golgi membrane dynamics.

The Arf/Arl family of GTPases is equally important for Golgi function. Class I and II Arf proteins recruit the COPI coat at ERGIC and Golgi compartments, with the aid of the GEF Gea1/2 in yeast or GBF1 in mammalian cells [106]. Arf proteins also recruit clathrin adaptors at the TGN with the aid of the GEF Sec7 in yeast or BIG1/2 in mammalian cells [93 ••, 106]. In addition, Arf-like Arl GTPases cooperate with Rab GTPases to recruit vesicle tethers at the TGN [107-109]. The Arf GEFs show interesting behavior in two respects. First, Gea2 and GBF1 are apparently present on all Golgi compartments, including the TGN [110 •, 111 •], presumably because they remain associated with cisternae throughout the maturation process. Second, Arf and Arl help to recruit the TGN-localized Arf GEFs [112 ••, 113 ••]. The implication is that like Rab GTPases, Arf and Arl GTPases operate in cascades during Golgi maturation [114].

Conclusions

A three-stage model for Golgi maturation and function is useful for guiding ongoing research. These stages are defined by distinct membrane trafficking pathways. At each stage, a set of GTPases recruits specific components while regulating the transition to the next stage. Challenges for the future include identifying the components that establish each stage, and determining how membrane and proteins are recycled within and between the three stages.