Cargo sorting at the trans-Golgi network at a glance

The Golgi is an organelle of the eukaryotic endomembrane system that functions principally in the biogenesis and intracellular sorting of newly synthesized glycoproteins and lipids, herein termed ‘biosynthetic cargo’. It is composed of three or more flattened membrane sacs termed the cis, medial and trans cisternae, which adhere to each other in a polarized orientation to form the Golgi stack (Klumperman, 2011). The terminal Golgi compartment is a network of convoluted membranes termed the trans-Golgi network (TGN). Newly synthesized proteins and lipids produced in the endoplasmic reticulum (ER) are directed to the cis Golgi, where they enter the Golgi stack and are trafficked in the anterograde direction to the medial and trans cisternae and the TGN (Li et al., 2019). Within the Golgi stack, cargo-processing enzymes, primarily glycosyltransferases that attach carbohydrate moieties to biosynthetic cargo, reside in one or two cisternae. This arrangement establishes a processing pathway whereby sequentially acting Golgi-resident enzymes appropriately modify cargo proteins, via reactions that include elaboration of glycan chains on proteins and lipids, sulfation of proteins and proteolysis (Stanley, 2011; Stone et al., 2009; Tan and Gleeson, 2019; Thomas, 2002) (see Box 1).

Once biosynthetic cargo processing is complete, cargo is exported from the Golgi by coated vesicles and pleiomorphic uncoated vesicles, and subsequently transported to the plasma membrane, endosome or other Golgi compartments (De Matteis and Luini, 2008). Cargo sorting and processing in the Golgi is essential for the generation of bioactive proteins (such as hormones and cell surface receptors), organelle biogenesis, cell polarization and lipid homeostasis. Organismal physiology depends critically on these Golgi-mediated cargo-sorting processes as their disruption can contribute to a wide range of disorders, including neurodegenerative diseases, diabetes and cancer (Liu et al., 2021). This Cell Science at a Glance article and the accompanying poster summarize the conceptual advances made toward understanding mechanisms of cargo sorting and export from the trans Golgi and TGN (referred to hereafter as the trans/TGN) during secretion. While most of the cargo-sorting mechanisms presented here apply generally to all eukaryotic cells, the poster focuses on research of mammalian cell types.

Clathrin-coated vesicles (CCVs) traffic integral membrane proteins, predominantly precursors of lysosomal resident enzymes and proteins required for lysosome biogenesis, from the trans/TGN to organelles of the endo-lysosome system (see poster). The CCV coat is composed of two protein shells that are associated with the cytoplasmic leaflet of the vesicle membrane; the inner layer comprises a network of integral membrane proteins and peripheral cargo adaptor proteins that provide a platform for the assembly of clathrin triskelia into the second, outer layer (Wood and Smith, 2021). The best-characterized CCV cargo adaptors are those of the heterotetrameric clathrin adaptor protein family: AP-1, AP-2 and AP-3. These adaptors are comprised of four subunits: two large (one of α, γ, δ or ε and one β) subunits, a medium (μ) subunit and a small (σ) subunit (Sanger et al., 2019). For example, the AP-1 adaptor complex (which consists of β1 γ, μ1 and σ1 subunits) responsible for sorting at the trans/TGN has two γ (γ1 and γ2), two μ1 (μ1A and μ1B, also known in mammals as AP1M1 and AP1M2, respectively) and three σ1 (σ1A, σ1B and σ1C) isoforms, which are implicated in cell type-specific sorting functions of AP-1 (discussed below; Mattera et al., 2011). AP-1 binds two core trans/TGN membrane components, activated Arf1 GTPase (i.e. the GTP-bound form) and the signaling lipid phosphatidylinositol 4-phosphate (PI4P) (Kirchhausen et al., 2014). Additional CCV-associated proteins complement AP-1-mediated sorting by recognizing overlapping but distinct cargo clients. Of these, the Golgi-localized γ-ear-containing ADP-ribosylation factor-binding (GGA) proteins GGA1, GGA2 and GGA3 have prominent roles in lysosome biogenesis through the trafficking of mannose 6-phosphate (mannose 6-P) receptors (discussed below), with a principal role for GGA2 (Uemura and Waguri, 2020). Like AP-1, GGA proteins recognize sorting signals in the cytoplasmic segments of cargo proteins that conform to the sequence motif YXXΦ or [DE]XXXL[LI] (where X is any amino acid and Φ is a bulky hydrophobic amino acid) (Doray et al., 2007; Ohno et al., 1998, 1995; Puertollano et al., 2001; Zhu et al., 2001). Whereas AP-1 is present in two types of Golgi-derived CCVs, those containing only AP-1 and those containing both AP-1 and GGA proteins, GGA2 is specific to only AP-1-containing CCVs (Hirst et al., 2012).

The vesicle coat protein complex coatomer I (COPI) is the core component of Golgi retrograde pathways. COPI localizes to the rims of Golgi cisternae, whose edges are surrounded by vesicles, suggesting that these represent dynamic sites of protein sorting and export (Orci et al., 1997). Retrograde sorting of some trans/TGN-resident Golgi proteins is initiated by recruitment of COPI from the cytosol to Golgi membranes by Arf1-GTP and GOLPH3, to which COPI binds (Donaldson et al., 1992; Orci et al., 1993; Palmer et al., 1993; Serafini et al., 1991; Tu et al., 2012) (see poster). On Golgi membranes, COPI assembles into a polymeric membrane coat that contains binding sites for retrograde sorting signals present in the cytoplasmic segments of integral membrane client cargo proteins (Cosson and Letourneur, 1994; Letourneur et al., 1994). Cargo is captured by this nascent vesicle coat during the budding process, and once budded from the cisterna, the COPI coat dissociates and the vesicle fuses with an earlier, maturing cisterna, thus maintaining Golgi cisterna composition (Popoff et al., 2011).

In addition to direct capture of cargo by the COPI coat, the peripheral membrane protein GOLPH3 associates with COPI on Golgi membranes, and these interactions are necessary for retention of many residents of the early Golgi cisternae (Eckert et al., 2014; Tu et al., 2012, 2008). Loss of GOLPH3 increases the rate at which these Golgi residents are degraded in the lysosome, resulting in decreased levels of these proteins in the Golgi, perturbations to protein glycosylation and complex sphingolipid homeostasis (Chang et al., 2013; Rizzo et al., 2021; Wood et al., 2012), and ultimately secretion (Dippold et al., 2009; Rahajeng et al., 2019; Xing et al., 2016). GOLPH3 contains an evolutionarily conserved pattern of arginine residues (two or three within the first 12 amino acids of the N termini) that confers binding to coatomer (COPI) and is required for retention of Golgi residents (Eckert et al., 2014; Tu et al., 2012). Based on these observations, GOLPH3 has been proposed to be a cargo-selective adaptor for the COPI vesicle coat via coincident recognition of PI4P, COPI and sequence motifs – LXX[RK] and [FL][LIV]XX[RK] – present in many residents of early Golgi compartments (such as the cis or medial cisternae) (Chang et al., 2013; Dippold et al., 2009; Eckert et al., 2014; Rizzo et al., 2021; Tu et al., 2012; Welch et al., 2021; Wood et al., 2009). Alternatively, GOLPH3 and COPI might prevent packaging of Golgi residents into anterograde intra-Golgi transport vesicles to maintain Golgi residence. GOLPH3 expression is increased in cells derived from a wide variety of human tumors, reportedly contributing to cellular transformation through enhanced mitogenic signaling caused by altered sphingolipid metabolism (Farber-Katz et al., 2014; Rizzo et al., 2021; Scott et al., 2009).

The mechanisms for sorting of constitutively secreted soluble proteins and those destined for regulated secretion remain poorly understood, mostly because of a lack of an identified sorting receptor. Nevertheless, Ca²⁺ has emerged as a critical regulator in these processes (see poster).

In the constitutive secretion pathway, Ca²⁺ is pumped into the trans/TGN lumen, resulting in Ca²⁺-dependent protein aggregation that facilitates cargo sorting via SPCA1 (also known as ATP2C1), a trans/TGN-localized Ca²⁺ ATPase (von Blume et al., 2011). Conversely, knockdown of SPCA1 reduces Ca²⁺ levels in the TGN lumen and results in decreased secretion of soluble cargoes (Deng et al., 2018; von Blume et al., 2011). The Golgi-resident protein Cab45 (also known as SDF4) couples Ca²⁺ concentration to secretory activity as it oligomerizes in the presence of Ca²⁺ (Crevenna et al., 2016; von Blume et al., 2012) and binds to soluble secretory proteins (including lysozyme C and cartilage oligomeric matrix protein) (von Blume et al., 2012). Thus, sorting of constitutively secreted soluble cargo is regulated by Ca²⁺, although further studies are necessary to understand how Cab45 oligomers drive sorting.

In the regulated secretory pathway, secretory cargo is retained within cytoplasmic vesicles, including dense-core vesicles (DCVs), which fuse with the plasma membrane upon stimulation (see poster). Regulated secretion is crucial for the physiology of specialized secretory cells, such as pancreatic β-cells, cells of the neuroendocrine system and the immune system. Members of the granin family of proteins, including chromogranin A (CHGA or CgA), chromogranin B (CHGB or CgB) and secretogranin II (SCG2 or SgII), are thought to play a role in DCV biogenesis via Ca²⁺-dependent aggregation (Beuret et al., 2004; Colomer et al., 1996; Gerdes et al., 1989; Huh et al., 2003; Kim et al., 2001; Yoo, 1995). Biochemical studies of crude extracts derived from pituitary or adrenal glands and purified proteins have demonstrated that chromogranins can aggregate at acidic pH and high (millimolar) Ca²⁺ concentrations (Colomer et al., 1996; Gerdes et al., 1989; Yoo, 1995). This has led to the ‘sorting for entry’ (also termed ‘sorting by aggregation’) hypothesis, whereby chromogranins aggregate within the TGN milieu and, in the process, co-aggregate other secretory granule-destined cargoes, resulting in active sorting (Tooze, 1998).

The plasma membrane of polarized cells is itself polarized. In the case of epithelial cells, the apical plasma membrane faces the external environment, and the basolateral domain faces neighboring cells and the growth substratum, with intercellular tight junctions preventing diffusion of proteins and lipids between the two domains (see poster) (Cao et al., 2012; Stoops and Caplan, 2014). A longstanding goal regarding trafficking in polarized cells is to elucidate the mechanisms by which newly synthesized lipids and proteins are targeted to the correct plasma membrane domains and to understand the physiology of polarized cells in these terms.

At least four distinct Golgi-to-plasma membrane trafficking pathways have been described for polarized epithelial cells based on distinct sorting requirements for different cargo proteins (see poster) (Weisz and Fölsch, 2020). For basolateral trafficking, the most straightforward sorting mechanisms involve the recognition of tyrosine- or dileucine-based sorting motifs by AP-1, which mediates polarized sorting at the trans/TGN into CCVs (see above) (Caceres et al., 2019; Deborde et al., 2008; Gravotta et al., 2012). Some types of epithelial cells, including Madin–Darby canine kidney (MDCK)-derived cell lines that are commonly used for polarized cell sorting studies, express two forms of AP-1, AP-1A and AP-1B (Fölsch et al., 1999; Ohno et al., 1999), which differ solely in their distinct medium subunits (μ1A and μ1B). Curiously, AP-1A and AP-1B recognize tyrosine- or dileucine-based motifs (see above) with partly overlapping cargo-recognition preferences and so cooperate and compensate to accomplish basolateral sorting (Gravotta et al., 2012; Guo et al., 2013). Despite functional overlap, AP-1A localizes prominently to the trans/TGN and AP-1B to the recycling endosome (Fölsch, 2015). The role of AP-1 exclusively in basolateral trafficking has been challenged by recent reports showing that loss of AP-1 broadly effects both basolateral and apical protein targeting, highlighting the requirement for enhanced scrutiny when limited cargoes or cell types are employed (Caceres et al., 2019; Gravotta et al., 2019). One further heterotetrameric adaptor complex, AP-4, functions as a cargo adaptor for non-clathrin-coated vesicles at the trans/TGN, with a role in basolateral sorting that is apparently redundant to AP-1 (Simmen et al., 2002). Although little is known about the role of AP-4 in basolateral sorting at the trans/TGN, only AP-4 mediates sorting of the autophagy factor ATG9A from the trans/TGN to the nascent autophagosome (Davies et al., 2018; De Pace et al., 2018; Mattera et al., 2017). Additionally, several integral membrane proteins, notably the amyloid precursor (APP) and related proteins (likely APLP1 and APLP2), along with ATG9A, accumulate in the trans/TGN in AP-4-deficient cell lines (Burgos et al., 2010; Mattera et al., 2020), suggesting that they are AP-4 clients. Mutations in the genes encoding each of the AP-4 subunits cause a hereditary spastic paraplegia referred to as AP-4-deficiency syndrome, likely due to perturbed autophagy in neurons (Moreno-De-Luca et al., 2011; Tüysüz et al., 2014; Verkerk et al., 2009).

The apical plasma membrane of many types of epithelial cells is enriched in cholesterol and glycosphingolipids, such as galactosylceramide sulfate (also known as sulfatide) and Forssman glycolipid (Hansson et al., 1986; Koichi et al., 1974; Nichols et al., 1987), that are synthesized in the Golgi. Pioneering studies of lipid trafficking in MDCK cells employed synthetic fluorescent sphingolipid analogs that could be tracked and visualized in cells (van 't Hof and van Meer, 1990; van der Bijl et al., 1996; van Genderen and van Meer, 1995; van IJzendoorn and Hoekstra, 1998; van IJzendoorn et al., 1997), revealing that vesicles containing these fluorescent lipids are trafficked directly to the apical plasma membrane (van IJzendoorn et al., 2020).

In all cell types, sphingolipids reside exclusively in the exofacial and lumenal membrane leaflets of cellular membranes (Lorent et al., 2020; Verkleij et al., 1973), so it is not obvious how their identities and organization in the lumenal leaflet could be coupled to budding of secretory vesicles, which typically requires cytoplasmic factors. The ‘raft’ hypothesis postulates that phase condensation of cholesterol and sphingolipids within membrane bilayers forms a patch and/or domain of locally ordered lipids [the liquid-ordered (L_o) domain], termed a ‘lipid raft’ (Simons and Van Meer, 1988). Proteins bearing a glycosylphosphatidylinositol anchor (GPI-APs) and some palmitoylated integral membrane proteins co-segregate with ‘raft’ lipids in model membranes (Friedrichson and Kurzchalia, 1998; Levental et al., 2010; Varma and Mayor, 1998), and structural features of the membrane-spanning segments of model and native proteins have been identified whose segregation correlates with lipid rafts, with efficient targeting to the plasma membrane (Lorent et al., 2017; Sharpe et al., 2010; Yurtsever and Lorent, 2020). In addition, glycosylation of proteins has been shown to contribute to their clustering within these domains and export from the Golgi within apically targeted transport vesicles (see below) (Lebreton et al., 2019). Though it is obvious that the physicochemical properties of Golgi membranes are harnessed for the concentration and export of secretory products, it is now appreciated that vesicle-sized sphingolipid–cholesterol condensates are not present in biological membranes due to their molecular complexity (Anderson and Jacobson, 2002; Veatch and Keller, 2005). Nevertheless, lipid-based protein sorting mechanisms represent an active area of debate (Levental et al., 2020).

Glycosylation of cargo proteins plays a role in sorting at the trans/TGN. Two distinct glycosylation patterns exist on proteins: N-linked and O-linked. N-linked glycosylation is initiated in the ER and involves the covalent attachment of glycosidic linkages to the amide side chain of asparagine residues (Vagin et al., 2009). In contrast, O-linked glycosylation takes place in the Golgi by linkage of glycosyl chains to serine or threonine residues (Potter et al., 2006).

One of the best-characterized TGN sorting mechanisms, the sorting of lysosomal hydrolases by mannose 6-P receptors, relies on glycosylation (see poster). The N-linked glycosyl chains on lysosomal hydrolases are sequentially modified by two Golgi-resident enzymes, N-acetylglucosamine 1-phosphotransferase and N-acetylglucosamine-1-phosphodiester α-N-acetylglucosaminidase, to generate the mannose 6-P moiety (Kornfeld and Mellman, 1989). Mannose 6-P-tagged lysosomal hydrolases are then recognized by the mannose 6-P receptor and sorted into CCVs (Brown and Farquhar, 1984; Friend and Farquhar, 1967; Geuze et al., 1985; Hoflack and Kornfeld, 1985a,b; Lemansky et al., 1987). These vesicles deliver cargo–mannose 6-P complexes to the late endosome and/or lysosome, where the acidic lumenal pH drives dissociation of the enzymes from mannose 6-P receptors, allowing for their recycling back to the trans/TGN (Griffiths et al., 1988). Deficiency of N-acetylglucosamine 1-phosphotransferase leads to the lysosomal storage disorder mucolipidosis type 2 (Kornfeld and Mellman, 1989), causes secretion of lysosomal hydrolases (Gelfman et al., 2007; Vogel et al., 2009) and results in the formation of lysosomes filled with undigested substrates (Otomo et al., 2011).

N- and O-linked glycosylation have also been shown to contribute to apical sorting of both membrane-anchored and soluble proteins in polarized cells (see poster). Treatment of cells with inhibitors of early steps in glycan addition, expression of glycosylation-deficient mutant proteins and studies of glycosylation gain-of-function mutations have revealed a strong correlation between N-glycosylation and apical sorting of secreted proteins (Hendriks et al., 2004; Scheiffele et al., 1995; Urban et al., 1987; Vagin et al., 2009). There is limited data to delineate the role of O-linked glycosylation in protein sorting (Lebreton et al., 2019; Potter et al., 2006), although it has been shown to promote efficient exit of cargoes from the Golgi of HeLa cells (Sun et al., 2020). How these glycosylation signals promote sorting and exit at the trans/TGN remains unclear. Lectins, including VIP36 (also known as LMAN2) and galectin-3, have been implicated as putative glycan-sorting receptors (Delacour et al., 2006; Hara-Kuge et al., 2002); however, it is not clear to what extent endogenous levels of these proteins can contribute to sorting (Fullekrug et al., 1999).

Intracellular signaling pathways control numerous aspects of trans/TGN functions, including protein sorting, transport carrier formation and lipid homeostasis (Farhan and Rabouille, 2011) (see poster).

G proteins such as small GTPases act as molecular switches that facilitate the formation and transport of trans/TGN-derived carriers. Effectors generally interact with the GTP-bound form of the respective GTPase to induce a specific downstream signaling cascade that modulates cargo recognition, vesicular budding, membrane curvature, fission and transport of the vesicle to the cell surface on cytoskeletal tracks (Hutagalung and Novick, 2011).

The best-described GTPases regulating membrane trafficking are the Arf, Rab and Rho subfamily members of the Ras superfamily, which serve as membrane-associated platforms that recruit specific effectors to the trans/TGN, facilitating vesicle formation (Anitei and Hoflack, 2012; Bankaitis et al., 2012). These effectors include clathrin adaptor-associated proteins and F-actin, as well as microtubule-associated proteins that modulate membrane bending for budding and fission, and motors that transport the vesicles to their final destination (Anitei et al., 2010). Specifically, members of the Rab GTPase family are involved in TGN carrier budding, fission and movements of vesicles via cytoskeletal networks (both F-actin and microtubules). Rab6 connects myosin and kinesin motors to promote vesicular movement from the TGN to the cell periphery or focal adhesions (Eisler et al., 2018; Fourriere et al., 2019; Grigoriev et al., 2007, 2011; Hutagalung and Novick, 2011; Jasmin et al., 1992).

A role for heterotrimeric G proteins in trans/TGN carrier formation has also been described (Jamora et al., 1999; Pimplikar and Simons, 1994) but remains poorly understood. Heterotrimeric G proteins are associated with G-protein-coupled receptors (GPCRs), and it has been speculated that cargo exit from the Golgi is regulated by yet-to-be-identified GPCRs (Di Martino et al., 2019).

Protein kinases and phosphatases have been associated with secretory carrier formation at the trans/TGN (Mayinger, 2011). AP-1 recruitment to the membrane is coupled with protein phosphatase 2A-mediated dephosphorylation of the µ1 subunit of AP-1, promoting clathrin assembly. Once on the membrane, the AP-1 µ1 subunit is phosphorylated, which induces a conformational change and increases the binding to sorting signals in the cytosolic tails of cargo proteins (see poster). This process is a well-established example of how cyclical phosphorylation and dephosphorylation of AP-1 regulate its function from membrane recruitment until release into the cytosol (Ghosh and Kornfeld, 2003).

Protein kinase D (PKD, of which there are three mammalian isoforms with a similar modular structure, known as PRKD1, PRKD2 and PRKD3) is a key regulator of TGN lipid homeostasis (von Blume and Hausser, 2019) and TGN carrier fission (Bard and Malhotra, 2006). The C1a domain of PKD binds to diacylglycerol (DAG) (Baron and Malhotra, 2002), whereas the C1b domain binds to Arf1 (Pusapati et al., 2010) to tightly associate PKD with the TGN membrane (see poster). Fission of some secretory carriers from the TGN requires catalytic activation of PKD by protein kinase C (PKC)-mediated phosphorylation (Añel and Malhotra, 2005). Although several models describing how PKD could promote carrier fission have been proposed, the mechanism is still unknown (Campelo and Malhotra, 2012). Moreover, relevant substrates that influence fission have not yet been identified (Wakana and Campelo, 2021). Involvement of PKA–cyclic AMP signaling in export from the TGN has also been described, but the relevant substrates are also unknown (Pimplikar and Simons, 1994).

Rapid changes in particular lipids at the trans/TGN membranes play a significant role in cargo sorting, budding and fission (von Blume and Hausser, 2019). Lipid effectors are proteins that recognize these lipids and are recruited from the cytosol to the TGN membrane. Such lipids include phosphatidylserine (PS), phosphatidic acid (PA) and, in particular, PI4P. PI4P generated by phosphatidylinositol 4-kinase-IIIβ (encoded by PI4KB) recruits sorting factors such as AP-1, GGAs and GOLPH3 to TGN membranes (Bankaitis et al., 2012). It is also critical for determining TGN membrane composition by regulating non-vesicular lipid transfer of sphingolipids and cholesterol via ER–TGN contact sites by CERT (also known as CERT1) and OSBP, respectively (Hanada et al., 2003; Mesmin et al., 2013). CERT and OSBP activities are regulated by PKD phosphorylation, facilitating the crosstalk between lipid metabolism and protein signaling at the trans/TGN (Wang et al., 2003) (see poster).

A great deal is known about the features of cargo proteins that confer their sorting and trafficking from the Golgi to other organelles, yet much remains to be learned about the mechanisms that produce Golgi-derived specific transport carriers (see Box 2 for outstanding key questions). A large body of work suggests that in contrast to canonical, coat protein-driven budding mechanisms that are induced by cytoplasmic coat proteins, secretory carrier formation may be directed partially by the cargo itself. Aggregation of lumenal cargo proteins is controlled in part by the surrounding milieu, including pH and Ca²⁺ concentration, and can account for cargo sorting. Such aggregates could themselves provide a substrate around which membrane is wrapped to form a secretory carrier. In this scenario, association of cargo aggregates with the lumenal membrane leaflet could involve the lipid-binding activity of cargo proteins or, alternatively, integral membrane receptors that would capture cargo aggregates. Similarly, condensation of sphingolipids and cholesterol in the membrane, along with proteins with affinity for these condensates, likely explains sorting of membrane components such as ‘raft’ lipids and proteins, yet here too the mechanism(s) underlying carrier formation is unknown. Future insights into carrier formation may come from efforts to understand how the actin- and microtubule-based cytoskeletons influence Golgi membrane dynamics, which may initiate carrier budding and/or contribute to carrier fission.

We are grateful to colleagues for discussions and critical reading of the manuscript. The authors would like to thank Edward Felder (University of Ulm, Germany) for the collaboration to produce the electron micrograph shown in the poster.

Individual poster panels are available for downloading at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259110#supplementary-data