Sending out molecules from the TGN

Abstract

The sorting of secreted cargo proteins and their export from the trans-Golgi network (TGN) remains an enigma in the field of membrane trafficking, although the sorting mechanisms of many transmembrane proteins have been well described. The sorting of secreted proteins at the TGN is crucial for the release of signaling factors as well as extracellular matrix proteins. These proteins are required for cell-cell communication and integrity of an organism. Missecretion of these factors can cause diseases such as neurological disorders, autoimmune disease, or cancer. The major open question is how soluble proteins that are not associated with the membrane are packed into TGN derived transport carriers to facilitate their transport to the plasma membrane. Recent investigations have identified novel types protein and lipid machinery that facilitate the packing of these molecules into a TGN derived vesicle. In addition, novel research has uncovered an exciting link between cargo sorting and export in which TGN structure and dynamics as well as TGN/endoplasmic reticulum contact sites, play a significant role. Here, we have reviewed the progress made in our understanding of these processes.

Introduction

Secreted proteins are estimated to be encoded by approximately 15 % of all human genes [1]. These molecules play pertinent roles contributing to physiology and development or organisms. Signaling molecules such as cytokines and chemokines, growth factors, enzymes, coagulation factors, and hormones such as insulin are examples of a few crucial secreted proteins. Some of the secreted proteins are also the main constituents of the extracellular matrix and provide the essential physical scaffolding for cellular components and act as the initiators of biochemical and biomechanical homeostasis, morphogenesis, and differentiation [2].

Soluble proteins are synthesized, glycosylated, folded, and quality checked in the endoplasmic reticulum (ER) before being loaded into vesicular carriers for transport to the Golgi apparatus [3],[4]. Subsequently, these molecules travel across the Golgi stacks and are sorted in the trans-Golgi network (TGN) into specific transport carriers and routed to their designated cellular compartments or for secretion [5],[6]. Important intracellular destinations for such carriers include the cell surface, early/sorting endosomes, late endosomes, and recycling endosomes [7],[8]. Another layer of complexity is seen in polarized cells, where the soluble proteins are secreted multi-directionally from the cell or delivered to specific plasma membrane domains such as the apical and basolateral membrane domains in polarized epithelial cells or axons/dendrites in neurons.

Certain cell types are specialized in regulated secretion and have developed mechanisms to accommodate their cargoes within storage granules and release them in response to various stimuli (regulated pathways). Regulated secretion is a distinctive feature of professional secretory cells and facilitates the packing of proteins into secretory granules. These cell types include neuroendocrine cells such as insulin secreting cells or neurons containing chemical synapses.

Some proteins are secreted unconventionally that means independently of the classical secretory pathway (ER/Golgi). This process is, however, is beyond the scope of this review and it has been discussed intensively elsewhere [9,10].

Research in the last decade has shed light on the importance of the TGN structure and the role of ER/TGN contact sites in the biogenesis of secretory cargo carriers. To facilitate sorting, the soluble cargo molecules must be concentrated and separated from other cargoes and TGN resident proteins and become attached to the membrane to form a cargo containing vesicle. Simultaneously, the vesicle requires a precise “molecular make up” such as specific SNARE proteins and Rab proteins for targeting. How could this work? This simplest model includes a cargo receptor that recognizes a specific feature of the cargo molecule. This has been shown for soluble lysosomal hydrolases that contain a Mannose-6-Phosphate (M6P) that is recognized by the M6P receptor and subsequently packed into clathrin-coated vesicles towards endosomal compartments (Figure 1A). Since there is no consensus sorting motif for secreted proteins, we predict that most of the proteins are sorted in a manner that is independent of a specific sorting receptor. We further propose that these cargo molecules must either have a lipid-binding site or the capability to associate with a luminal adaptor protein that binds lipids. The concentration of cargo at a specific export domain might be attributed to the oligomerization of the cargo proteins that occurs only under specific conditions. An example is the low pH and high Ca²⁺ concentration in the milieu of the TGN as it has been proposed for Cab45 and chromogranins.

Figure 1. Schematic view of protein sorting at the trans-Golgi Network (TGN).

Diagram of the Golgi apparatus and TGN that form membrane contact sites (MCS) with the ER. (A) Schematic model of sorting lysosomal hydrolases into clathrin-coated vesicles the best understood sorting process of soluble proteins at the TGN. (B) Schematic representation of Ca²⁺ and pH dependent “sorting by aggregation” model in neuroendocrine cells. (C) Schematic model of sorting of the PAUF protein into sphingomyelin and cholesterol-rich CARTs. (D) Schematic model depicts the Cab45-dependent client sorting of lysozyme C into sphingomyelin reach vesicles. (E) Schematic model of sorting of Lipoprotein lipase (LPL) protein by Heparan Sulfate modification of Syndecan Iintosphingomyelin reach vesicles. (F) ER/TGN membrane contact site that serve as hotspots for inter-organelle lipid transfer and could play role in Ca²⁺-dependent sorting and signaling between compartments. The Figure was created using Biorender.

The lipid environment and the TGN milieu is at least partly controlled by the ER/TGN contact sites. Therefore, these structures are necessary for secretory cargo sorting and export. Another important question is the sensing of the cargo proteins in the TGN. In this regard, Di Martino et al. have proposed a novel GPCR (GPRC5A) in the TGN that binds to basolateral cargo molecules and induces a signaling cascade resulting in the activation of PLCb3. This process in turn induces DAG production and PKD recruitment and the subsequent formation of basolateral carriers [72].

The main issue that we would like to address in this review is the sorting of soluble molecules that do not directly contact the membrane into a transport carrier that buds from the TGN. Moreover, we wish to discuss the structural requirements of the endomembrane system that controls these processes.

Features and functional subdomains of the TGN

Structure of the TGN

Protein sorting is a spatial segregation process in which the cargo sorting machinery concentrates specific cargo proteins in particular membrane subdomains from which other proteins, such as Golgi resident proteins, are excluded. The trans Golgi consists of TGN cisternae and further develops into an anastomosing tubular network called the TGN (Figure 1). In the classical view, cargo proteins remain mixed throughout the Golgi complex until they reach the TGN and are packaged into their specific transport carrier [11]. An emerging modern view is that the TGN is organized in the form of a multi-domain structure. The model proposes that the budding of each type of carrier is spatially segregated and that the cargo export occurs from different trans cisternae and the TGN (Figure1). In this context, the extended definition of TGN in mammalian cells as a compartment consisting of the last three cisternae (cisternal TGN) in the Golgi ribbon and the exit tubules extending from them (tubular TGN) is gaining prominence[12].

Recent studies in plant and yeast cells have revealed the spatiotemporal pattern in the assembly dynamics of various Golgi/TGN-resident proteins with the aid of high-speed and high-resolution spinning-disk confocal microscopy [13,14]. These live-cell imaging experiments have provided additional support for the cisternal maturation model in these organisms. This model proposes that the cargo proteins remain enclosed in their respective Golgi cisternae while the cisternae undergo in line-up changes and mature from an earlier to a later one [15–17]. Accordingly, in yeast the TGN can be subdivided into sub-stages, namely the “early TGN” thatreceives retrograde traffic and the “late TGN” that generates the TGN-derived carriers [13]. Glick and collegues put forward the idea that different Golgi compartments can be defined as discrete kinetic stages in the maturation process [15]. When the cargo molecules reach the TGN, a small amount of cargo recycles to the earlier zone, which could represent a TGN quality control system [17]. In the late TGN stage adaptor proteins exhibit diverse assembly and disassembly kinetics [13]. Lei Lu et al. have developed a novel technology called Golgi protein localization by imaging centers of mass (GLIM) to monitor the subdomains in the Golgi apparatus. GLIM analyzes protein localization based on the center of the fluorescent masses in nocodazole-induced Golgi ministacks by measuring the axial position or localization quotient of a Golgi-localized protein [18]. This method yields a localization accuracy in the low nanometer scale (around 30 nm) by conventional light microscopy and allows the systematic and quantitative localization analysis of Golgi compartments and cargo proteins [18]. Based on these methods they proposed a model in which the cargo proteins destined to the plasma membrane leave at the cisternal TGN whereas others enter the tubular TGN in a signal-specific manner as clathrin and accessory proteins localize in this region [19][19]. However, this view is contradictory to the hypothesis according to which the tubular TGN is a Golgi exit site for proteins destined to reach various sites inside the cell, including the plasma membrane for secretion of soluble proteins [11]. Lei Lu and colleagues have suggested that the tubular TGN could serve as an exit site for endosome/lysosome or secretory granule-targeted proteins [19].

The TGN forms contact sites with the ER

Another prominent functional feature of the cisternal and tubular TGN is that it forms contact sites with the ER in which the solutes and lipids are transferred to mediate non-vesicular inter organelle communication (Figure 1). Currently, 10 proteins have been described to be present, to act, or to require ER/TGN contact sites for their functioning [20–22]. These include tethering proteins such as vesicle- associated membrane proteins A and B (VAP-A a4nd VAP-B), that tether ER and cisternal and tubular TGN membranes as well as lipid transfer proteins such as ceramide transfer protein (CERT,transfer of ceramide, Figure 1F), Oxysterol-binding protein 1 (OSBP1, counter exchange of cholesterol and PI4P, Figure 1F), OSBP-related protein 9 (ORP9, transfer of cholesterol), OSBP-related protein 10 (ORP10) transfer of phosphatidylserine), and Nir2 (phosphatidylinositol transfer). For instance, CERT possesses dual organelle targeting motifs including two phenylalanines in an acidic track (FFAT) domain, which binds the ER proteins VAP-A and VAP-B and a pleckstrin homology (PH) domain, which allows TGN recognition by Phosphatidylinisitol-4-phosphate (PI4P) or Arf1-GTP on TGN membrane [23,24]. Ceramide species are recognized by a START domain of CERT. Subsequently they are transferred to the TGN for the synthesis of sphingomyelin (SM) and glycosphingolipids by enzymes present in the TGN [25].

Another important function of the ER/TGN contact site is the control and maintenance of PI4P levels in the TGN. The TGN is enriched with PI4P, which acts as a localization signal for the protein machinery involved in sorting and budding events. OSBP1 counter transports PI4P from the TGN and cholesterol from the ER and the process is mediated by the ORD domain of OSBP1 that binds PI4P or cholesterol in a mutually exclusive manner [26]. The PI4P phosphatase Sac1 is localized in the ER and has two modes of action, in cis and trans. In the trans mode that usually oppose lipid traffic, its action requires the adaptor protein FAPP1 to interact with VAP proteins [27]. The ER/TGN contact site also has a major function in vesicular transport since the depletion of VAP tethering proteins impairs the trafficking of cargoes to the plasma membrane [28]. Depletion of the TGN PI4P pool by recruiting Sac1 to the TGN by a rapamycin inducible system leads to a defective secretion to the plasma membrane [29]. FAPP2, which tubulates the membrane and generates microdomains at the TGN and GOLPH3 which binds to PI4P at the TGN by MYO18A, are some of the P14 effector proteins. The interaction between GOLPH3 and MYO18A creates a mechanical force to detach the vesicular carriers from the TGN [30]. Another potential role of the ER/TGN contact sites is the transfer of Ca²⁺ from the ER. The TGN has a low steady-state Ca²⁺ level, and cargo sorting and export require a short transient Ca²⁺ influx in the millimolar range mediated by the TGN Ca²⁺ ATPase SPCA1 [31–34]. As the cytosol has very low Ca²⁺ concentrations, it is anticipated that the process could occur in a similar manner as it has been described for ER-mitochondria contact sites in which Ca²⁺ levels can rise very high [35]. The ER-mitochondria contact sites accommodate inositol 1,4,5 triphosphate receptors (IP3R) in the ER and voltage dependent anion channels in the outer mitochondrial membrane as well as the mitochondrial Ca²⁺ uniporter MICU1 at the inner mitochondria membrane [36]. However, it remains to be experimentally elucidated if the transfer occurs from the ER to the TGN.

Link between TGN structure and secretory cargo sorting

The sorting of soluble proteins in the trans Golgi/TGN is poorly understood. The major challenge is the lack of understanding of how these proteins are connected to a specific membrane domain to bud off a transport carrier. The historic view is that these molecules follow the bulk-flow of the secretory pathway. Any luminal protein lacking a sorting signal that directs it to the lysosomes would be transported out of the Golgi to the plasma membrane by default [37,38]. However, recent research has provided evidence that many of these molecules are actively sorted and that the TGN structures present in specific membrane contact sites play a major role in the process.

The unique structure of the TGN itself contributes tocargo sortingand export. It has been proposed that cargo export occurs in the highly curved cisternal rims of the TGN cisternae while the Golgi enzymes such as glycosylating proteins are localized in the flat interior regions of the cisternae. In agreement with this hypothesis, the trafficking machinery resides in the periphery of the stacks and in the highly curved cisternal rims where the export machinery and SNARE proteins are also localized [39,40]. Furthermore, this hypothesis has been strengthened by the fact that palmitoylated transmembrane proteins are concentrated in the highly curved rims of the cis Golgi, which promote their forward transport [41] and lower the energy needed to bend the local membrane to yield pinched-off carriers. Studies have also suggested that glycosylation and budding machinery are laterally segregated in the Golgi cisternae [42].

Post translational modifications, such as glycosylation and phosphorylation, alter the features of the cargo protein. For instance, O-glycosylation has been proposed as an export signal for Golgi exit. In this model, glycosylation of the luminal domain of the cargo molecule increases its hydrodynamic volume and probably excludes it from the “enzyme matrix” located in the interior of the Golgi cisternae. This process pushes the glycosylated protein to the rim of the cisternae where the trafficking components are present to pack the protein in a specific carrier [43]. The working of these processes or individual cargo molecules is yet to be deciphered.

Sorting into SM-rich vesicles

Cholesterol and Sphingolipids can form liquid-ordered membrane nanodomains that have been suggested to serve as platforms to sort specific proteins in the TGN. Recent experiments in which sphingomyelin (SM) metabolism was perturbed strongly suggest that such domains are required for functional cargo sorting and transport carrier biogenesis at the TGN [44–48]. TGN export of several soluble secretory proteins such as Cartilage Oligomeric Protein (COMP), Lysozyme C (LyzC), Matrix Gla Protein (MGP), and Thrombospondin 1 (TSP1) is Ca²⁺ dependent. Furthermore, a SM-rich membrane domain is required. This process involves the TGN localized Ca²⁺ ATPase SPCA1, which upon binding to cofilin and actin, locally pumps Ca²⁺ into the lumen of the TGN [49–51]. This transient local Ca²⁺ influx induces the oligomerization of the Ca²⁺ binding protein Cab45, which in turn leads to the concentration of the secretory molecules at a specific domain of the TGN in which they are packaged into sphingomyelin-rich vesicles (Figure 1D) [34,48,52]. The SPCA1 Ca²⁺ pumping activity is dependent upon the Golgi sphingomyelin synthesis pathway and SPCA1 is associated with sphingomyelin. Furthermore, cofilin and F-actin modulate the activity of SPCA1 by an unknown mechanism. Hence, the ER/TGN contact sites might be crucial. As previously stated, the steady-state Ca²⁺ concentration in the TGN is very low (100 μM). Cab45 requires around 1 mM of Ca²⁺ for its oligomerization to capture the cargo molecules for sorting. A significant question here is the source of the Ca²⁺ ions. We anticipate that Ca²⁺ is transferred via the ER/TGN contact sites in which an IP3 receptor in the ER membrane could release these ions into the ER/TGN contact site for being pumped into the TGN by SPCA1 (Figure 1F). In line with these observations, SM is generated by SMS1 converting ceramide and Phosphatidylcholine (PC) to SM and diacylglycerol (DAG) in the TGN. Importantly, SPCA1, Sphingomyelin Synthase 1 (SMS1), and SM populate identical regions of the TGN [48]. Therefore, local hotspots of SPCA1 activity linked to the local synthesis of SM, and the proximity to ER/TGN contact sites could determine the export of Ca²⁺/Cab45 dependent cargoes. This model is also supported by the fact that the depletion of VAP-A or VAP-B that tether ER/TGN exerts a profound impact on protein secretion [28] and the export of Cab45 clients (unpublished). Phosphorylation of Cab45 by the Golgi kinase Fam20C in the TGN plays a key role in modulating the budding of SM-rich vesicles by modulating the oligomerization capacity of Cab45 [53]. A recent work has reported that the soluble secretory protein lipoprotein lipase (LPL) is sorted and secreted in SM-rich vesicles, which further highlights the importance of SM in cargo export. Sorting of LPL into these vesicles requires their interaction of LPL with heparan sulfate proteoglycans. The membrane-spanning protein syndecan-1 (SDC-1) acts as a sorting receptor for LPL. Interestingly, the sorting of LPL by SDC-1 depends on the physical property of the transmembrane domain, which is sufficient to drive its incorporation into SM-rich secretory vesicles (Figure 1E) [54].

Sorting by aggregation

Another sorting process that has been intensely debated over the last three decades is the sorting of bioactive peptides such as insulin, which probably relies on ER/TGN contact sites [55]. Insulin secretory granule (SG) cargo is packaged in nascent granules that bud off the TGN. It has been proposed that members of the granin family such as chromogranin A, B and VGF are major regulators of this process [56,57]. These molecules aggregate at an acidic pH (pH 5.5) and under millimolar Ca²⁺ concentrations in vitro. Moreover, the ectopic expression of chromogranins can lead to the appearance of SG like structures in non-secretory cells [58]. This observation has led to a proposed hypothesis called “sorting by aggregation” or “sorting for entry”, according to which low pH and high Ca²⁺ can drive the aggregation of granins and associated proteins [59]. These aggregates would bind bioactive peptides and exclude other soluble proteins, thereby driving SG biosynthesis (Figure 1B). This process possibly occursat theER/TGN contact sites since it requires the transfer of high Ca²⁺ concentrations. Further evidence for its occurrence at the ER/TGN contact sites is provided by the fact that granin aggregates require high cholesterol levels for driving the maturation of SGs [60,61]. In this context it is interesting to note that OSBP1 seems to directly control insulin granule biosynthesis [61]. However, the interaction of these separated aggregates with the membrane to detach themselves from the TGN and form an immature secretory storage granule remains unclear.

Sorting and export of proteins by CARTS

Recent work has shown that the soluble pancreatic adenocarcinoma upregulated factor (PAUF) is exported from the TGN by “carriers of the TGN to the cell surface” (CARTS) [62]. The trafficking of these carriers occurs in a microtubule and kinesin-5-dependent manner. CARTS fission from the TGN requires protein kinase D (PKD) [62,63]. It is noteworthy that CARTS biogenesis also requires the ER/TGN contact sites that are maintained by VAP proteins [64]. A recent study has established that SCAP, which regulates cholesterol biosynthesis in the ER, is a central modulator of CARTS biogenesis. Researchers have provided evidence that SCAP interacts with VAP-OSBP at the ER-TGN contact sites via Sac1 and regulates the countertransport of cholesterol and PI4P in an ER cholesterol-dependent manner. Furthermore, it has been proposed that cholesterol and SM organize the lipid nanodomains at the TGN, which serve as platforms for the molecular machinery, that sort and process cargo. DAG which is synthesized with SM from ceramide and phosphatidylcholine recruits PKD for membrane fission. The data suggests that putative nascent CARTS bud in proximity to ER-TGN contact sites and determine the position for membrane fission [65]. To understand sorting of PAUF into these vesicles, the CARTS were identified by enrichment of TGN derived vesicles using a TGN46 antibody [62]. Since TGN46 traffics from the TGN to the cell surface and back to the TGN by clathrin-mediated endocytosis, it could be an attractive candidate for a cargo receptor (Figure 1C) [66]. However, whether TGN46 is capable of acting as a cargo receptor for PAUF is yet to be elucidated.

Conclusions

Soluble protein sorting at the TGN in concert with membrane dynamics needs to be further investigated. Rapid progress in recent years, specifically in the imaging technology, is likely to boost research in the field. High-end cryo-EM tomography is bound to provide additional insights into the ultrastructure of the TGN. Pan expansion microscopy is a novel technique that can aid in fathoming the distribution of specific proteins at the nanoscale in the ultrastructural context of the cell. Decrowding the intracellular space through physical expansion while retaining the protein bulk can resolve the local protein densities and reveal the cellular nanoarchitecture by standard light and immunofluorescence microscopy [67]. Genome-editing can allow the tagging of endogenous proteins and the analysis of their dynamics by high-speed and high-resolution microscopy [68].