**2. Biogenesis during trans-Golgi network vesicle trafficking**

Secreted proteins are translated at the rough endoplasmic reticulum and transit from the ER, through the ER-Golgi Intermediate Compartment (ERGIC) to the Golgi and subsequently the trans-Golgi network (TGN) (Shilo and Schejter 2011). Post-translational modifications such as glycosylation occur in the ER and Golgi. Membranes on the trans side of the Golgi apparatus form dynamic tubular reticular structures having a large surface area (De Matteis and Luini 2008). This network of saccules and tubes is continuously remodeled such that both the structure and size of the TGN varies depending on the secretory activity and the cell type (Trucco et al. 2004). Selective trafficking at the TGN will sort cargo into vesicles or carrier tubules (De Matteis and Luini 2008), and this sorting requires the genesis of carrier vesicles targeted to specific membranes within the cell. As noted above, these vesicles are distinguished by the combination of coat proteins on the cytosolic face of the membrane which determine the target membrane for that vesicle. In some pathways additional sorting occurs at the recycling endosome (reviewed in (De Matteis and Luini 2008; Santiago-Tirado and Bretscher 2011)).

The initiating event in vesicle biogenesis may be driven by the local membrane lipid composition where asymmetry in the types of lipid in the two faces of the membrane bilayer can induce bending (van Meer and Sprong 2004). Initiation may also involve membrane rafts, which are reported to be present in the TGN of all cells (Park and Loh 2008), and on membranes of secretory granules (Hosaka et al. 2004; Lang 2007; Guerriero et al. 2008). While membrane rafts are well characterized to play important roles in endocytosis at the plasmalemma, their role in vesicle formation at the TGN is not as well understood. Many vesicle coat proteins have been localized to lipid rafts on vesicles (Puri and Roche 2006). This has been suggested to be important for the formation of the coat protein complexes on the TGN for the initial creation of vesicle buds and selective trafficking (Simons and Sampaio 2011). For example, SNARE proteins are enriched in cholesterol-dependent rafts in beta-cells and PC-12 cells (Lang 2007). However, technical issues have called into question the validity of some methods for isolation of 'lipid rafts', leading to a more stringent definition and the term 'membrane rafts' (Lang 2007). Nonetheless, even with the more stringent approach, SNAREs such as VAMP2 and VAMP3 are enriched in membrane rafts of vesicles. Membrane rafts can contain different complements of proteins due to specific protein interactions. One model for the formation of vesicle buds is that membrane rafts on the TGN (with associated transmembrane cargo proteins) coalesce creating a lipid domain that is favorable to bending the membrane, and containing transmembrane proteins which can facilitate the decoration of the cytosolic face of the membrane with adaptor and other coat proteins (De Matteis and Luini 2008; Simons and Sampaio 2011). Testing the relevance of this model to selective sorting for secretion in living cells is important, and requires determining whether trafficking vesicles in different pathways (e.g., regulated secretion versus constitutive secretion pathways) contain different types of membrane rafts, or that

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 43

has not been well characterized in any cell type. Recently, PtdIns(3,4)P2 was found to be transiently synthesized at the plasmalemma as a second messenger of platelet-derived growth factor (PDGF) (Hogan et al. 2004). Importantly, our understanding of the roles of PtdInsPs derives primarily from studies of yeast and a few mammalian cell lines. It is unclear whether these generalizations carry over to well differentiated cells in tissues. For example, we find high levels of expression of PtdIns(3,4)P2 in parotid secretory granule

Discrete and dynamic localization of PtdIns-kinases (PI-kinases) and PI-phosphatases regulate the production of PtdInsPs. Of these enzyme families, PI4-kinases have the dominant roles in Golgi function and secretion, although PI3-kinases and PI5-kinases also have roles in secretion (Wang et al. 2003; Roth 2004; Balla and Balla 2006). Several PI4 kinases are localized to the Golgi, TGN, or endosomes, and form complexes with several coat proteins including GGA and FAPP1/2 (D'Angelo et al. 2008). Careful immunofluorescence co-localization studies with MDCK cells have found PI4KIIα in the TGN and PI4KIIIβ in the cis/medial Golgi (Weixel et al. 2005). This is consistent with PtdIns4P being the most abundant PtdInsP in the Golgi. PI4KIIIβ is recruited by the coat protein Arf1 to the Golgi, and in turn PI4KIIIβ recruits Rab11 and the PtdIns4P-binding protein FAPP1 (Godi et al. 2004). Knock-down of PI4KIIα by RNAi has little effect on intra-Golgi transit, but inhibits TGN export of vesicles (Wang et al. 2003). These studies are interpreted as showing that PI4-kinases produce PtdIns4P on the cytosolic leaflet of the TGN and vesicles, which anchors essential adaptor proteins (Santiago-Tirado and Bretscher 2011). Importantly, class I PI3-kinases are also present on Golgi membranes, and are essential for tumor necrosis factor (TNF) secretion by macrophages (Low et al. 2010). Similarly, PI3 kinase was localized to secretory granules of PC12 cells by both immunofluorescence and cell fractionation (Meunier et al. 2005). Transfection of a PtdIns3P-binding domain (FYVE), or a catalytically-inactive PI3-kinase, blocked regulated secretion, possibly by interfering with fusion of the granule at the plasmalemma (Meunier et al. 2005). In summary, certain PIkinases decorate the Golgi and TGN, interact with coat proteins, and produce PtdInsPs. These PtdInsPs are important for vesicle trafficking, however, the identities and roles of PIkinases have not been investigated in cells that are highly specialized for bulk exocrine

The presence of multiple different trafficking pathways for secretion has been described in many cell types (Dikeakos et al. 2007; De Matteis and Luini 2008; Park et al. 2008; Folsch et al. 2009; Perez et al. 2010; Stanley and Lacy 2010; Lacy and Stow 2011; Santiago-Tirado and Bretscher 2011). The best characterized of these pathways have a specific cargo protein that is a unique marker of that route, which is essential for molecular characterization (Lara-Lemus et al. 2006). For example, the cytolytic protein perforin of natural killer cells undergoes polarized secretion into the immune synapse, whereas the same cells secrete TNF in a non-polarized pattern (Reefman et al. 2010) thereby marking a different pathway. Also, some pituitary gonadotropes segregate luteinizing hormone into separate granules than follicle-stimulating hormone for different regulated secretion dynamics (Nicol et al. 2004). Alternatively, pathways can be marked by the use of different coat proteins. For example, the coat protein FAPP1 is essential for secretion by the basolateral pathway, whereas FAPP2

membranes.

secretion, such as the parotid.

**3. Selective trafficking for secretion** 

some pathways lack membrane rafts. Importantly, Guerriero et al. recently found that raftindependent and raft-associated proteins collect in distinct sites at the Golgi, and likely enter different vesicles (Guerriero et al. 2008).

#### **2.1 Phosphatidylinositol phosphates in biogenesis of secretory vesicles**

In all cell types, the earliest events that are strongly linked to the biogenesis of secretory vesicles is the binding of coat proteins to phosphatidylinositol phosphate lipids (PtdInsP) and PI-kinases in the TGN membrane. The phosphoinositides and small GTPases of the Arf and Rab families define the identity of the membrane and recruit additional coat proteins (Di Paolo and De Camilli 2006; De Matteis and Luini 2008; Santiago-Tirado and Bretscher 2011). PtdInsPs are recognized as being critical for selective trafficking of vesicles within cells (Di Paolo and De Camilli 2006; D'Angelo et al. 2008; Vicinanza et al. 2008; Graham and Burd 2011). Phosphatidylinositol comprises less than 10% of membrane phospholipids, and the individual phosphorylated forms total less than 1.5% of lipids (Di Paolo and De Camilli 2006; Roth 2004). Phosphatidylinositol can be phosphorylated on any combination of carbons 3, 4, or 5 of the inositol ring (Fig. 1). The most abundant, PtdIns4P, occurs at approximately 0.05% of membrane lipids, whereas the low abundance forms such as PtdIns(3,4)P2 or PtdIns(3,4,5)P3 are approximately 0.0001% each (Cullen 2011). Subcellular pools of the 7 different PtdInsPs have diverse regulatory roles in cytoskeleton remodeling, second messenger signaling, endosomal trafficking, membrane trafficking, osmotic stress, nuclear signaling, and other aspects of cell physiology (Godi et al. 2004; Balla and Balla 2006; Di Paolo and De Camilli 2006). Cellular effects can be mediated by signaling through production of second messengers (diacylglycerol, and inositol trisphosphate). However, many cellular effects are mediated by the localized anchoring of cytosolic proteins having specific PtdInsP-binding domains (e.g., PH, FYVE, PX, ENTH-domains) (De Matteis et al. 2005; Balla and Balla 2006). For example, Arf1 directly binds and recruits PI4-Kinases to the TGN. This produces PtdIns4P which aids in recruitment of coat proteins such as AP1, GGAs, VAMPs, and FAPP1/2 most of which interact with PtdIns4P as well as other proteins in the coat matrix (Balla and Balla 2006). This interaction provides identity to the membrane (reviewed in (Santiago-Tirado and Bretscher 2011)). PI3K-C2α is also present on the TGN. Each type of PtdInsP is localized to specific membranes within the cell. The most abundant phosphoinositides, PtdIns4P, and PtdIns(4,5)P2, are predominantly localized to the Golgi and plasma membrane, respectively, whereas PtdIns3P and PtdIns(3,5)P2 are predominantly localized to early and late endosomes (Di Paolo and De Camilli 2006; Santiago-Tirado and Bretscher 2011). PtdIns(3,4)P2 is rare in resting cells, but is present in the plasmalemma and multivesicular and early endosomes (Roth 2004; Di Paolo and De Camilli 2006). PtdIns(3,4)P2 is not noted as being present in the Golgi or TGN (De Matteis et al. 2005), however, this

Fig. 1. Structure of Phosphatidylinositol (3,4)bisphosphate. PtdIns(3,4)P2 is an exceedingly rare membrane lipid that is highly localized to specific subcellular membranes within the cell.

some pathways lack membrane rafts. Importantly, Guerriero et al. recently found that raftindependent and raft-associated proteins collect in distinct sites at the Golgi, and likely enter

In all cell types, the earliest events that are strongly linked to the biogenesis of secretory vesicles is the binding of coat proteins to phosphatidylinositol phosphate lipids (PtdInsP) and PI-kinases in the TGN membrane. The phosphoinositides and small GTPases of the Arf and Rab families define the identity of the membrane and recruit additional coat proteins (Di Paolo and De Camilli 2006; De Matteis and Luini 2008; Santiago-Tirado and Bretscher 2011). PtdInsPs are recognized as being critical for selective trafficking of vesicles within cells (Di Paolo and De Camilli 2006; D'Angelo et al. 2008; Vicinanza et al. 2008; Graham and Burd 2011). Phosphatidylinositol comprises less than 10% of membrane phospholipids, and the individual phosphorylated forms total less than 1.5% of lipids (Di Paolo and De Camilli 2006; Roth 2004). Phosphatidylinositol can be phosphorylated on any combination of carbons 3, 4, or 5 of the inositol ring (Fig. 1). The most abundant, PtdIns4P, occurs at approximately 0.05% of membrane lipids, whereas the low abundance forms such as PtdIns(3,4)P2 or PtdIns(3,4,5)P3 are approximately 0.0001% each (Cullen 2011). Subcellular pools of the 7 different PtdInsPs have diverse regulatory roles in cytoskeleton remodeling, second messenger signaling, endosomal trafficking, membrane trafficking, osmotic stress, nuclear signaling, and other aspects of cell physiology (Godi et al. 2004; Balla and Balla 2006; Di Paolo and De Camilli 2006). Cellular effects can be mediated by signaling through production of second messengers (diacylglycerol, and inositol trisphosphate). However, many cellular effects are mediated by the localized anchoring of cytosolic proteins having specific PtdInsP-binding domains (e.g., PH, FYVE, PX, ENTH-domains) (De Matteis et al. 2005; Balla and Balla 2006). For example, Arf1 directly binds and recruits PI4-Kinases to the TGN. This produces PtdIns4P which aids in recruitment of coat proteins such as AP1, GGAs, VAMPs, and FAPP1/2 most of which interact with PtdIns4P as well as other proteins in the coat matrix (Balla and Balla 2006). This interaction provides identity to the membrane (reviewed in (Santiago-Tirado and Bretscher 2011)). PI3K-C2α is also present on the TGN. Each type of PtdInsP is localized to specific membranes within the cell. The most abundant phosphoinositides, PtdIns4P, and PtdIns(4,5)P2, are predominantly localized to the Golgi and plasma membrane, respectively, whereas PtdIns3P and PtdIns(3,5)P2 are predominantly localized to early and late endosomes (Di Paolo and De Camilli 2006; Santiago-Tirado and Bretscher 2011). PtdIns(3,4)P2 is rare in resting cells, but is present in the plasmalemma and multivesicular and early endosomes (Roth 2004; Di Paolo and De Camilli 2006). PtdIns(3,4)P2 is not noted as being present in the Golgi or TGN (De Matteis et al. 2005), however, this

Fig. 1. Structure of Phosphatidylinositol (3,4)bisphosphate. PtdIns(3,4)P2 is an exceedingly rare membrane lipid that is highly localized to specific subcellular membranes within the

**2.1 Phosphatidylinositol phosphates in biogenesis of secretory vesicles** 

different vesicles (Guerriero et al. 2008).

cell.

has not been well characterized in any cell type. Recently, PtdIns(3,4)P2 was found to be transiently synthesized at the plasmalemma as a second messenger of platelet-derived growth factor (PDGF) (Hogan et al. 2004). Importantly, our understanding of the roles of PtdInsPs derives primarily from studies of yeast and a few mammalian cell lines. It is unclear whether these generalizations carry over to well differentiated cells in tissues. For example, we find high levels of expression of PtdIns(3,4)P2 in parotid secretory granule membranes.

Discrete and dynamic localization of PtdIns-kinases (PI-kinases) and PI-phosphatases regulate the production of PtdInsPs. Of these enzyme families, PI4-kinases have the dominant roles in Golgi function and secretion, although PI3-kinases and PI5-kinases also have roles in secretion (Wang et al. 2003; Roth 2004; Balla and Balla 2006). Several PI4 kinases are localized to the Golgi, TGN, or endosomes, and form complexes with several coat proteins including GGA and FAPP1/2 (D'Angelo et al. 2008). Careful immunofluorescence co-localization studies with MDCK cells have found PI4KIIα in the TGN and PI4KIIIβ in the cis/medial Golgi (Weixel et al. 2005). This is consistent with PtdIns4P being the most abundant PtdInsP in the Golgi. PI4KIIIβ is recruited by the coat protein Arf1 to the Golgi, and in turn PI4KIIIβ recruits Rab11 and the PtdIns4P-binding protein FAPP1 (Godi et al. 2004). Knock-down of PI4KIIα by RNAi has little effect on intra-Golgi transit, but inhibits TGN export of vesicles (Wang et al. 2003). These studies are interpreted as showing that PI4-kinases produce PtdIns4P on the cytosolic leaflet of the TGN and vesicles, which anchors essential adaptor proteins (Santiago-Tirado and Bretscher 2011). Importantly, class I PI3-kinases are also present on Golgi membranes, and are essential for tumor necrosis factor (TNF) secretion by macrophages (Low et al. 2010). Similarly, PI3 kinase was localized to secretory granules of PC12 cells by both immunofluorescence and cell fractionation (Meunier et al. 2005). Transfection of a PtdIns3P-binding domain (FYVE), or a catalytically-inactive PI3-kinase, blocked regulated secretion, possibly by interfering with fusion of the granule at the plasmalemma (Meunier et al. 2005). In summary, certain PIkinases decorate the Golgi and TGN, interact with coat proteins, and produce PtdInsPs. These PtdInsPs are important for vesicle trafficking, however, the identities and roles of PIkinases have not been investigated in cells that are highly specialized for bulk exocrine secretion, such as the parotid.

#### **3. Selective trafficking for secretion**

The presence of multiple different trafficking pathways for secretion has been described in many cell types (Dikeakos et al. 2007; De Matteis and Luini 2008; Park et al. 2008; Folsch et al. 2009; Perez et al. 2010; Stanley and Lacy 2010; Lacy and Stow 2011; Santiago-Tirado and Bretscher 2011). The best characterized of these pathways have a specific cargo protein that is a unique marker of that route, which is essential for molecular characterization (Lara-Lemus et al. 2006). For example, the cytolytic protein perforin of natural killer cells undergoes polarized secretion into the immune synapse, whereas the same cells secrete TNF in a non-polarized pattern (Reefman et al. 2010) thereby marking a different pathway. Also, some pituitary gonadotropes segregate luteinizing hormone into separate granules than follicle-stimulating hormone for different regulated secretion dynamics (Nicol et al. 2004). Alternatively, pathways can be marked by the use of different coat proteins. For example, the coat protein FAPP1 is essential for secretion by the basolateral pathway, whereas FAPP2

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 45

(Saito et al. 2011). Within the lumen, phogrin can bind to carboxypeptidase E (CPE), and contributes to sorting of the complex (Saito et al. 2011). Hence, information about the identity of the vesicle encoded by the coat proteins, is related through phogrin to determine the luminal cargo. Phogrin also has PI-phosphatase activity which may be important for

Other sorting receptors have been reported in neural and endocrine cells. CPE itself binds the granule membrane. The C-terminus of CPE can span the membrane, although only 5 amino acid residues are cytosolic (Dhanvantari et al. 2002). Recycling of CPE from the plasmalemma requires ARF6 apparently due to direct binding to the coat protein. CPE acts as a sorting receptor for proopiomelanocortin (POMC) and proBDNF trafficking into the regulated secretion pathway (reviewed in (Park and Loh 2008)). Secretogranin III (SgIII) can also serve as a sorting receptor. Despite the absence of a transmembrane domain, SgIII binds to granule membranes or to cholesterol-rich liposomes, and anchors chromograninA (CgA) to the membrane (Hosaka et al. 2004). SgIII also interacts with CPE, POMC, and adrenomedullin (Hosaka et al. 2005; Han et al. 2008). It is unclear how SgIII is targeted to specific granules, but this may occur due to selective interactions with membrane rafts, or

Dikeakos et al. (Dikeakos et al. 2007; Dikeakos and Reudelhuber 2007) have shown that a hydrophobic patch in short amphipathic alpha helices is sufficient to sort cargo proteins. Helical domains are implicated in sorting of somatostatin, CPE, prohormone convertase enzymes (PC1/3, PC2), and chromogranin A (CgA). The proposed mechanism for this sorting is that the hydrophobic patch of the helix embeds into the membrane of the forming granule (Dikeakos and Reudelhuber 2007). One of the sorting-competent helical domains (Hels13-5) bound liposomes composed of phosphatidylcholine (PC) and phosphatidylglycerol (PG) with a Kd=9.7 μM (Kitamura et al. 1999). This Hels13-5 peptide integrated into the non-polar layer to a variable extent depending on the pH of the liposome. No cholesterol was necessary for this interaction. In addition, the helical domain of PC1/3 was shown to directly interact with CHAPS detergent micelles (Dikeakos et al. 2009). Characterization of both natural and artificial helices which confer sorting of a cargo protein at the TGN is an important step forward, however, at this point it is unclear how the identity of the vesicle (determined by the cytosolic coat proteins) directs selective sorting of

For many years, aggregation of secreted proteins has been seen as critical to sorting into the regulated secretory pathway in neuroendocrine cells (Gorr et al. 2005). Dense-core secretory granules of the regulated pathway contain large aggregates comprising chromogranins secretogranins and other secreted proteins, and which are not present in the constitutive secretion pathway. Aggregation of granins and many hormones can be demonstrated *in vitro* in a Ca++ and pH-dependent fashion. The pH of the TGN is approximately 6.2. In AtT20 cells the pH decreases further as the granule matures, reaching pH 5.5 in mature secretory granules (Wu et al. 2001). The acidic pH and high Ca++ present in the regulated secretory pathway is essential for aggregation. This relatively non-specific interaction may allow trafficking of large aggregates of proteins even where only a few specific interactions with transmembrane sorting receptors are present. Nonetheless, it must be recognized that aggregation is a fairly ill-defined concept, and it will be necessary to characterize these interactions in order to determine how some cargo is excluded from the aggregate to be

regulating glucose-stimulated insulin secretion (Caromile et al. 2010).

due to interaction with CPE and, indirectly, phogrin.

such cargo into the correct pathway.

sorted into the constitutive pathway.

directs vesicles to the apical pathway (Godi et al. 2004; Vieira et al. 2005). Importantly, mutation of proteins involved with sorting between trafficking pathways can cause disease. This is also seen with non-genetic diseases, such as pancreatitis during which inappropriate basolateral (endocrine) secretion of cargo proteins occurs. Hence, pancreatic amylase or lipase in the serum are standard clinical markers of this disease. Given the presence of multiple different pathways for secretion, the key issue is to understand the molecular interactions of proteins destined for each pathway, which cause sorting to the correct immature granule or tubule as it forms, or which cause retention of the protein in the granule as it matures.

The best characterized model for specific sorting of soluble cargo proteins involves a transmembrane sorting receptor protein (Fig. 2). The receptor protein is present in the TGN membrane and is able to interact with coat proteins on the cytosolic side of the vesicle bud, and can also bind luminal cargo proteins. The transmembrane receptor is localized to the vesicle bud by the appropriate coat proteins, and in turn, selects the correct cargo. Hence, a single protein serves to coordinate the identity of the vesicle with the luminal contents. This model is exemplified by mannose-6-phosphate sorting receptors (MPRs), which are type I transmembrane receptors present in the TGN (Ghosh et al. 2003). Both the cation-dependent (MPR300) and the cation-independent (MPR46) MPRs deliver lysosomal enzymes from the TGN to endosomes for subsequent transfer to lysosomes. The cytosolic tails of MPRs have specific binding sites for multiple adaptor proteins, including AP1, AP2, GGAs and PACs1 (reviewed in (Ghosh et al. 2003)). In addition, the portion of MPR in the lumen of the vesicle binds to mannose 6-phosphate tags on cargo proteins. The mannose 6-phosphate is a specific posttranslational modification on the N-glycans of over 60 of acid hydrolases which need to be transported to lysosomes. Failure of this sorting causes lysosomal sorting disease, Mucolipidosis type II alpha/beta (I-cell disease) (Ghosh et al. 2003). A similar sorting mechanism has been described for stabilin-1 which binds GGAs and serves as a sorting receptor for a chitinase-like enzyme.

Fig. 2. Transmembrane sorting receptor model. The coat proteins localize a transmembrane sorting receptor, which collects the appropriate luminal cargo.

Various mechanisms have been suggested for sorting of secreted proteins into the regulated secretory pathway, as opposed to the constitutive secretory pathway or trafficking to intracellular targets (Dikeakos and Reudelhuber 2007; Park et al. 2008). Some proteins are secreted by the model discussed above. Phogrin (Ptprn2) is a type I transmembrane receptor present in the TGN and secretory granules of endocrine and neuroendocrine cells. Phogrin contains tyrosine and leucine motifs in the C-terminal (cytosolic) tail, which are important for localization to secretory granules, likely by interaction with coat proteins such as AP1

directs vesicles to the apical pathway (Godi et al. 2004; Vieira et al. 2005). Importantly, mutation of proteins involved with sorting between trafficking pathways can cause disease. This is also seen with non-genetic diseases, such as pancreatitis during which inappropriate basolateral (endocrine) secretion of cargo proteins occurs. Hence, pancreatic amylase or lipase in the serum are standard clinical markers of this disease. Given the presence of multiple different pathways for secretion, the key issue is to understand the molecular interactions of proteins destined for each pathway, which cause sorting to the correct immature granule or tubule as it forms, or which cause retention of the protein in the

The best characterized model for specific sorting of soluble cargo proteins involves a transmembrane sorting receptor protein (Fig. 2). The receptor protein is present in the TGN membrane and is able to interact with coat proteins on the cytosolic side of the vesicle bud, and can also bind luminal cargo proteins. The transmembrane receptor is localized to the vesicle bud by the appropriate coat proteins, and in turn, selects the correct cargo. Hence, a single protein serves to coordinate the identity of the vesicle with the luminal contents. This model is exemplified by mannose-6-phosphate sorting receptors (MPRs), which are type I transmembrane receptors present in the TGN (Ghosh et al. 2003). Both the cation-dependent (MPR300) and the cation-independent (MPR46) MPRs deliver lysosomal enzymes from the TGN to endosomes for subsequent transfer to lysosomes. The cytosolic tails of MPRs have specific binding sites for multiple adaptor proteins, including AP1, AP2, GGAs and PACs1 (reviewed in (Ghosh et al. 2003)). In addition, the portion of MPR in the lumen of the vesicle binds to mannose 6-phosphate tags on cargo proteins. The mannose 6-phosphate is a specific posttranslational modification on the N-glycans of over 60 of acid hydrolases which need to be transported to lysosomes. Failure of this sorting causes lysosomal sorting disease, Mucolipidosis type II alpha/beta (I-cell disease) (Ghosh et al. 2003). A similar sorting mechanism has been described for stabilin-1 which binds GGAs and serves as a sorting

Fig. 2. Transmembrane sorting receptor model. The coat proteins localize a transmembrane

Various mechanisms have been suggested for sorting of secreted proteins into the regulated secretory pathway, as opposed to the constitutive secretory pathway or trafficking to intracellular targets (Dikeakos and Reudelhuber 2007; Park et al. 2008). Some proteins are secreted by the model discussed above. Phogrin (Ptprn2) is a type I transmembrane receptor present in the TGN and secretory granules of endocrine and neuroendocrine cells. Phogrin contains tyrosine and leucine motifs in the C-terminal (cytosolic) tail, which are important for localization to secretory granules, likely by interaction with coat proteins such as AP1

sorting receptor, which collects the appropriate luminal cargo.

granule as it matures.

receptor for a chitinase-like enzyme.

(Saito et al. 2011). Within the lumen, phogrin can bind to carboxypeptidase E (CPE), and contributes to sorting of the complex (Saito et al. 2011). Hence, information about the identity of the vesicle encoded by the coat proteins, is related through phogrin to determine the luminal cargo. Phogrin also has PI-phosphatase activity which may be important for regulating glucose-stimulated insulin secretion (Caromile et al. 2010).

Other sorting receptors have been reported in neural and endocrine cells. CPE itself binds the granule membrane. The C-terminus of CPE can span the membrane, although only 5 amino acid residues are cytosolic (Dhanvantari et al. 2002). Recycling of CPE from the plasmalemma requires ARF6 apparently due to direct binding to the coat protein. CPE acts as a sorting receptor for proopiomelanocortin (POMC) and proBDNF trafficking into the regulated secretion pathway (reviewed in (Park and Loh 2008)). Secretogranin III (SgIII) can also serve as a sorting receptor. Despite the absence of a transmembrane domain, SgIII binds to granule membranes or to cholesterol-rich liposomes, and anchors chromograninA (CgA) to the membrane (Hosaka et al. 2004). SgIII also interacts with CPE, POMC, and adrenomedullin (Hosaka et al. 2005; Han et al. 2008). It is unclear how SgIII is targeted to specific granules, but this may occur due to selective interactions with membrane rafts, or due to interaction with CPE and, indirectly, phogrin.

Dikeakos et al. (Dikeakos et al. 2007; Dikeakos and Reudelhuber 2007) have shown that a hydrophobic patch in short amphipathic alpha helices is sufficient to sort cargo proteins. Helical domains are implicated in sorting of somatostatin, CPE, prohormone convertase enzymes (PC1/3, PC2), and chromogranin A (CgA). The proposed mechanism for this sorting is that the hydrophobic patch of the helix embeds into the membrane of the forming granule (Dikeakos and Reudelhuber 2007). One of the sorting-competent helical domains (Hels13-5) bound liposomes composed of phosphatidylcholine (PC) and phosphatidylglycerol (PG) with a Kd=9.7 μM (Kitamura et al. 1999). This Hels13-5 peptide integrated into the non-polar layer to a variable extent depending on the pH of the liposome. No cholesterol was necessary for this interaction. In addition, the helical domain of PC1/3 was shown to directly interact with CHAPS detergent micelles (Dikeakos et al. 2009). Characterization of both natural and artificial helices which confer sorting of a cargo protein at the TGN is an important step forward, however, at this point it is unclear how the identity of the vesicle (determined by the cytosolic coat proteins) directs selective sorting of such cargo into the correct pathway.

For many years, aggregation of secreted proteins has been seen as critical to sorting into the regulated secretory pathway in neuroendocrine cells (Gorr et al. 2005). Dense-core secretory granules of the regulated pathway contain large aggregates comprising chromogranins secretogranins and other secreted proteins, and which are not present in the constitutive secretion pathway. Aggregation of granins and many hormones can be demonstrated *in vitro* in a Ca++ and pH-dependent fashion. The pH of the TGN is approximately 6.2. In AtT20 cells the pH decreases further as the granule matures, reaching pH 5.5 in mature secretory granules (Wu et al. 2001). The acidic pH and high Ca++ present in the regulated secretory pathway is essential for aggregation. This relatively non-specific interaction may allow trafficking of large aggregates of proteins even where only a few specific interactions with transmembrane sorting receptors are present. Nonetheless, it must be recognized that aggregation is a fairly ill-defined concept, and it will be necessary to characterize these interactions in order to determine how some cargo is excluded from the aggregate to be sorted into the constitutive pathway.

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 47

the molecular interactions of PSP in the secretory granule in an attempt to understand how

We analyzed rat parotid granule membranes by mass spectrometry with the goal of identifying integral membrane proteins that may be candidate sorting receptors in the parotid gland. Parotid gland homogenates were fractionated on sucrose gradients to isolate the secretory granules, which were osmotically lysed. Membranes were washed and further enriched by an additional sucrose gradient. Sucrose gradient-purified granule membranes were electrophoresed on polyacrylamide gels, and trypsinized peptides identified by MS/MS as described (Uriarte et al. 2008). Numerous integral or membranebound proteins were identified, including several involved in vesicle trafficking and cytoskeletal proteins, as expected. However, potential sorting receptors such as SgIII (Han et al. 2008) or carboxypeptidase E (Dhanvantari et al. 2002) were not identified in parotid membranes by this method. Nonetheless, one salivary cargo protein, PSP, was identified. Other abundant soluble cargo proteins such as amylase and acidic Proline-Rich Protein (PRP) were not detected. To confirm the binding of PSP to granule membranes, western blot analysis was done with independent preparations of purified and extensively washed granule membranes. This confirmed that PSP is selectively bound, whereas amylase and PRP are absent from purified membranes (Fig. 3) (Venkatesh et al. 2011). While this approach failed to identify candidate sorting receptor proteins, it did demonstrate that PSP is a good marker for interactions with the membrane. In other cell types, putative sorting proteins such as secretogranin III (Hosaka et al. 2004), carboxypeptidase E (Dhanvantari et al. 2002), PC1/3, and PC2 (Jutras et al. 2000) are also associated with

Fig. 3. Western blots of purified secretory granule membranes, probed with either antiamylase or anti-PSP. Lanes contain either intact granules (G), purified granule membranes (M), or soluble cargo protein lysate (L). Equal proportions (0.5%) of each fraction was analyzed. The star indicates PSP in the purified membrane fraction. Mw: molecular size

Given the existing models for sorting receptors, we tested whether PSP was bound to a sorting receptor protein in the secretory granule membrane. Numerous experiments were done attempting to crosslink PSP to a membrane protein; however, PSP never crosslinked

**4.1 Parotid secretory protein (PSP) binds to secretory granule membranes** 

sorting may be controlled (Venkatesh et al. 2011).

isolated secretory granule membranes.

standards.
