**4.2 PSP binds specifically to phosphatidylinositol 3,4-bisphosphate**

The binding of a cargo protein to the vesicle membrane is of great interest in defining the mechanisms of sorting. However, the ability of PSP to bind granule membranes in the absence of any sorting receptor protein ruled out the most common model for sorting. Several secreted proteins have been shown to interact with lipid microdomains (e.g., CPE, SgIII, PC1/3), presumably by relatively non-specific hydrophobic interactions (Park and Loh 2008; Dikeakos et al. 2007), whereas other classes of protein bind to a highly specific lipid headgroup (Di Paolo and De Camilli 2006). Therefore, we tested the ability of PSP to bind specific lipids, and to bind liposomes. Parotid secretory granules were isolated, and the lysate supernatant containing soluble cargo proteins was used in lipid-overlay assays to determine whether PSP or other salivary proteins (amylase or PRP) bind specific lipids (Dowler et al. 2000). Importantly, none of the cargo proteins bound directly to the most abundant membrane lipids (phosphatidylcholine, phosphatidylethanolamine, cholesterol, or sphingomyelin). Similarly, acidic PRP never bound any lipid spots, and amylase showed little or no binding. In contrast, PSP bound with remarkable selectivity to phosphatidylinositol phosphates (PtdInsPs), but did not bind to unphosphorylated PtdIns (Fig. 4). We observed decreased PSP binding to PtdInsPs at more acidic conditions, but clear binding was still present at pH 6.0. Hence, this interaction could contribute to sorting of PSP in the TGN where the pH is approximately 6.2 (Arvan and Castle 1986), and may also contribute to retention as the pH increases during granule maturation.

Fig. 4. PSP binds to phosphatidylinositol phosphates. Lipid strips (Echelon Biosciences) were incubated with parotid granule soluble lysate at 2 μg/ml (Venkatesh et al. 2011). Bound protein was detected with antibodies to PSP or acidic PRP. A schematic of the lipid strips is shown on the left (filled circles represent PSP binding).

The inability of PSP to bind unphosphorylated PtdIns suggested that specific interactions with the headgroup were required. Therefore, we compared the binding of PSP to a dilution

into a specific membrane-dependent higher molecular weight band. Taking the opposite approach, we extensively digested parotid granule membranes with either trypsin or pronase to destroy all membrane-associated proteins, and subsequently found that exogenous PSP still binds quite effectively. These results indicated that PSP does not require a protein receptor for binding to the membrane. In contrast, exogenous amylase did not bind the trypsinized membranes, and was present in the unbound fraction only,

The binding of a cargo protein to the vesicle membrane is of great interest in defining the mechanisms of sorting. However, the ability of PSP to bind granule membranes in the absence of any sorting receptor protein ruled out the most common model for sorting. Several secreted proteins have been shown to interact with lipid microdomains (e.g., CPE, SgIII, PC1/3), presumably by relatively non-specific hydrophobic interactions (Park and Loh 2008; Dikeakos et al. 2007), whereas other classes of protein bind to a highly specific lipid headgroup (Di Paolo and De Camilli 2006). Therefore, we tested the ability of PSP to bind specific lipids, and to bind liposomes. Parotid secretory granules were isolated, and the lysate supernatant containing soluble cargo proteins was used in lipid-overlay assays to determine whether PSP or other salivary proteins (amylase or PRP) bind specific lipids (Dowler et al. 2000). Importantly, none of the cargo proteins bound directly to the most abundant membrane lipids (phosphatidylcholine, phosphatidylethanolamine, cholesterol, or sphingomyelin). Similarly, acidic PRP never bound any lipid spots, and amylase showed little or no binding. In contrast, PSP bound with remarkable selectivity to phosphatidylinositol phosphates (PtdInsPs), but did not bind to unphosphorylated PtdIns (Fig. 4). We observed decreased PSP binding to PtdInsPs at more acidic conditions, but clear binding was still present at pH 6.0. Hence, this interaction could contribute to sorting of PSP in the TGN where the pH is approximately 6.2 (Arvan and Castle 1986), and may also

emphasizing the specificity of the binding of PSP.

**4.2 PSP binds specifically to phosphatidylinositol 3,4-bisphosphate** 

contribute to retention as the pH increases during granule maturation.

strips is shown on the left (filled circles represent PSP binding).

Fig. 4. PSP binds to phosphatidylinositol phosphates. Lipid strips (Echelon Biosciences) were incubated with parotid granule soluble lysate at 2 μg/ml (Venkatesh et al. 2011). Bound protein was detected with antibodies to PSP or acidic PRP. A schematic of the lipid

The inability of PSP to bind unphosphorylated PtdIns suggested that specific interactions with the headgroup were required. Therefore, we compared the binding of PSP to a dilution series of each of the seven different phosphorylated forms of PtdInsPs. We found that native PSP binds 3- to 5-fold more to PtdIns(3,4)P2 compared to PtdIns(4,5)P2 or PtdIns(3,4,5)P3, and 10-fold greater than PtdIns(3,5)P2 or PtdIns(4)P (Fig. 5). PSP does not bind PtdIns(3)P, PtdIns(5)P, or PtdIns. Half-maximal binding of PSP was with approximately 35 pmoles PtdIns(3,4)P2. Parallel blots failed to detect any bound amylase or acidic PRP, both of which are abundant in the granule lysates. This high degree of specificity indicates that PSP binds the head group of PtdInsPs, analogous to known PtdInsP-binding proteins (Di Paolo and De Camilli 2006). For example, PSP is more selective than the PH-domain protein DAPP1 which binds PtdIns(3,4)P2 or PtdIns(3,4,5)P3 with similar avidity (Dowler et al. 2000).

Fig. 5. Binding of PSP to phosphatidylinositol phosphates. PtdInsP array membranes spotted with serially diluted lipids were used to define the binding of native PSP. Membranes were blocked and probed with anti-PSP as described (Venkatesh et al. 2011).

The experiments described above use a secretory granule lysate as the source of PSP. This leaves open the possibility that PSP binds indirectly to PtdInsP2. Therefore, PSP was expressed *in vitro* in rabbit reticulocyte lysates, and also was expressed in bacteria. Chloramphenicol acetyltransferase (CAT) was used as a negative control since CAT and rPSP are similar in size and also have similar acidic pI values. Rat PSP, and human PSP (human Splunc2, BPIFA2), each with a V5 tag, were translated *in vitro* and bound selectively to PtdIns(3,4)P2 and PtdIns(4,5)P2 demonstrating that this activity is conserved between species. CAT-V5 was unable to bind any of the lipids. Similarly, both human and rat PSP proteins were expressed in bacteria as glutathione-S-transferase (GST) fusion proteins having V5 tags. Again, GST-rPSP-V5 and GST-hPSP-V5 each bound strongly to PtdIns(3,4)P2 and did not bind PtdIns, whereas GST-V5 had no binding activity (Venkatesh et al. 2011). Bacterially expressed and GST-affinity purified rPSP-V5 also preferentially binds PtdIns(3,4)P2 (Fig. 6). The binding of *in vitro* synthesized, and bacterially expressed, rPSP and human PSP to PtdIns(3,4)P2 demonstrates that the interaction is independent of other parotid granule proteins.

Pleckstrin homology domain proteins bind phosphoinositides with a moderate to high affinity (Vicinanza et al. 2008). We used bacterially expressed rPSP-V5 to determine the affinity of PSP for PtdIns(3,4)P2. Bacterially expressed affinity-purified rPSP-V5 was incubated at 0.1 to 3.5 μg/ml with membranes spotted with 50 pmoles of PtdIns(3,4)P2. Bound protein was detected with anti-V5 antibody, and the intensity of the signal used to calculate the amount of free and bound protein from a PSP-V5 standard curve, as described

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 51

not detect any interaction of PSP with liposomes comprising PC, PE, and unphosphorylated PtdIns (77:20:3), nor did it bind to liposomes with PtdIns(3,5)P2 (Fig. 8). This suggests that PSP does not interact with membranes through the relatively non-specific hydrophobic interactions observed for amphipathic alpha helices, and shown to be important for sorting of certain cargo proteins (Kitamura et al. 1999; Dikeakos et al. 2007). Conversely, native rat PSP, in granule lysates, bound repeatably to lysosomes spiked with 3% PtdIns(3,4)P2 or PtdIns(4,5)P2 (Fig. 8), consistent with the previous results. Amylase present in the same granule lysates did not bind to any of these liposomes. PSP also bound to stabilized PIPosomes (from Echelon Biosciences Inc.) containing 5% PtdIns(3,4)P2. These results show that native PSP binds PtdIns(3,4)P2 in an intact lipid membrane. This binding is not detectably due to hydrophobic interactions of an alpha helical domain, does not require a different protein to act as a membrane tether or a transmembrane sorting receptor, nor does

Fig. 8. PSP binds to PtdIns(3,4)P2 and PtdIns(4,5)P2 in lipid bilayers. Parotid granule extract was incubated with liposomes spiked with 3% either PtdIns, PtdIns(3,4)P2, PtdIns(4,5)P2 or PtdIns(3,5)P2. After binding, liposomes were extensively washed and analyzed by SDS-PAGE and western blotting with anti-PSP. Each lane is a separate incubation. PSP bound to

The results described above suggest that PSP binds directly to the headgroup of PtdIns(3,4)P2 in a membrane. It is of interest to compare this activity to related proteins. Following the newly developed nomenclature (Bingle et al. 2011), rat PSP/BPIFA2E is a member of the BPI-fold superfamily. This diverse superfamily includes bactericidal/permeability-increasing protein (BPI), LPS-binding protein (LBP), cholesteryl ester transfer protein (CETP), and phospholipid transfer protein (PLTP). In addition, the superfamily includes the BPIFA subfamily (previously referred to as the SPLUNC family, containing PSP), and the BPIFB subfamily (previously termed LPLUNCs). The history of cloning the PSP/Splunc/BPIFA family, and changes in the nomenclature have recently been described in depth (Bingle et al. 2011). An important observation is that proteins in this superfamily have quite divergent amino acid sequences. For example, the optimal possible alignment of rat PSP to rat BPI gives a sequence identity of only 19%. However, there is strong conservation of secondary structure and predicted tertiary structures across the superfamily. The crystal structure of BPI shows a hollow boomerang-shaped structure (Beamer et al. 1997). The two halves (domains) of the boomerang show clear similarity to each other at both the primary sequence level, and at the structural level. The Long PLUNC (BPIFB) subfamily maps across both the two domains, whereas PSP and the rest of the Short PLUNC (BPIFA) subfamily consists of only one domain. Many of these proteins have important roles that involve binding to lipids. For example, both BPI and LBP bind to

PtdIns(3,4)P2 and PtdIns(4,5)P2 only. M: molecular size markers.

it require cholesterol-rich domains.

**4.3 The PSP family of BPI-fold proteins** 

Fig. 6. Bacterially expressed rPSP-V5 binds to PtdIns(3,4)P2. GST-PSP-V5 was affinity purified, and the rPSP-V5 was isolated separate from GST using PreScission protease. Protein overlay assays were performed using the purified rPSP-V5 (1 µg/ml) on nitrocelluose membranes spotted with lipids (200-6.25 pmoles/spot). Bound protein was detected with anti-V5 antibody. Bacterially expressed rPSP shows a similar pattern of specificity as native PSP.

(Venkatesh et al. 2011). The affinity was derived from the binding curve (Fig. 7). In three independent experiments, the binding affinity of PSP ranged from Kd= 1.85 x 10-10 to 3.9 x 10-11 M demonstrating a high affinity interaction. This is similar to the affinity of 5 x 10-9 M for TAPP1 binding PtdIns(3,4)P2 measured by a similar method (Dowler et al. 2000). The affinity of p47phox for PtdIns(3,4)P2 is reported as 3.8 x 10-8 (Karathanassis et al. 2002). In a direct comparison of bacterially expressed p47phox-V5 and rPSP-V5 we confirmed that PSP binds PtdIns(3,4)P2 more strongly than p47phox.

Fig. 7. Binding curve of PSP to PtdIns(3,4)P2. Bacterially expressed affinity-purified rPSP-V5 was incubated at different concentrations to define binding to PtdIns(3,4)P2 as described (Venkatesh et al. 2011).

We studied the interaction of PSP with lipid bilayers in liposomes for three reasons. Primarily, we wanted to determine if PSP binds specifically to PtdInsPs in intact membranes. In addition, it was important to test whether PSP can bind non-specifically to membranes due to hydrophobic interactions similar to helical peptides (Kitamura et al. 1999; Dikeakos et al. 2007) and to compare these two types of interaction. Synthetic liposomes were made by standard methods to contain phosphatidylcholine (PC), phosphatidylethanolamine (PE), and PtdIns (or PtdInsP) at a molar ratio of 77:20:3. We did

Fig. 6. Bacterially expressed rPSP-V5 binds to PtdIns(3,4)P2. GST-PSP-V5 was affinity purified, and the rPSP-V5 was isolated separate from GST using PreScission protease. Protein overlay assays were performed using the purified rPSP-V5 (1 µg/ml) on

specificity as native PSP.

(Venkatesh et al. 2011).

binds PtdIns(3,4)P2 more strongly than p47phox.

nitrocelluose membranes spotted with lipids (200-6.25 pmoles/spot). Bound protein was detected with anti-V5 antibody. Bacterially expressed rPSP shows a similar pattern of

(Venkatesh et al. 2011). The affinity was derived from the binding curve (Fig. 7). In three independent experiments, the binding affinity of PSP ranged from Kd= 1.85 x 10-10 to 3.9 x 10-11 M demonstrating a high affinity interaction. This is similar to the affinity of 5 x 10-9 M for TAPP1 binding PtdIns(3,4)P2 measured by a similar method (Dowler et al. 2000). The affinity of p47phox for PtdIns(3,4)P2 is reported as 3.8 x 10-8 (Karathanassis et al. 2002). In a direct comparison of bacterially expressed p47phox-V5 and rPSP-V5 we confirmed that PSP

Fig. 7. Binding curve of PSP to PtdIns(3,4)P2. Bacterially expressed affinity-purified rPSP-V5 was incubated at different concentrations to define binding to PtdIns(3,4)P2 as described

We studied the interaction of PSP with lipid bilayers in liposomes for three reasons. Primarily, we wanted to determine if PSP binds specifically to PtdInsPs in intact membranes. In addition, it was important to test whether PSP can bind non-specifically to membranes due to hydrophobic interactions similar to helical peptides (Kitamura et al. 1999; Dikeakos et al. 2007) and to compare these two types of interaction. Synthetic liposomes were made by standard methods to contain phosphatidylcholine (PC), phosphatidylethanolamine (PE), and PtdIns (or PtdInsP) at a molar ratio of 77:20:3. We did not detect any interaction of PSP with liposomes comprising PC, PE, and unphosphorylated PtdIns (77:20:3), nor did it bind to liposomes with PtdIns(3,5)P2 (Fig. 8). This suggests that PSP does not interact with membranes through the relatively non-specific hydrophobic interactions observed for amphipathic alpha helices, and shown to be important for sorting of certain cargo proteins (Kitamura et al. 1999; Dikeakos et al. 2007). Conversely, native rat PSP, in granule lysates, bound repeatably to lysosomes spiked with 3% PtdIns(3,4)P2 or PtdIns(4,5)P2 (Fig. 8), consistent with the previous results. Amylase present in the same granule lysates did not bind to any of these liposomes. PSP also bound to stabilized PIPosomes (from Echelon Biosciences Inc.) containing 5% PtdIns(3,4)P2. These results show that native PSP binds PtdIns(3,4)P2 in an intact lipid membrane. This binding is not detectably due to hydrophobic interactions of an alpha helical domain, does not require a different protein to act as a membrane tether or a transmembrane sorting receptor, nor does it require cholesterol-rich domains.

Fig. 8. PSP binds to PtdIns(3,4)P2 and PtdIns(4,5)P2 in lipid bilayers. Parotid granule extract was incubated with liposomes spiked with 3% either PtdIns, PtdIns(3,4)P2, PtdIns(4,5)P2 or PtdIns(3,5)P2. After binding, liposomes were extensively washed and analyzed by SDS-PAGE and western blotting with anti-PSP. Each lane is a separate incubation. PSP bound to PtdIns(3,4)P2 and PtdIns(4,5)P2 only. M: molecular size markers.

#### **4.3 The PSP family of BPI-fold proteins**

The results described above suggest that PSP binds directly to the headgroup of PtdIns(3,4)P2 in a membrane. It is of interest to compare this activity to related proteins. Following the newly developed nomenclature (Bingle et al. 2011), rat PSP/BPIFA2E is a member of the BPI-fold superfamily. This diverse superfamily includes bactericidal/permeability-increasing protein (BPI), LPS-binding protein (LBP), cholesteryl ester transfer protein (CETP), and phospholipid transfer protein (PLTP). In addition, the superfamily includes the BPIFA subfamily (previously referred to as the SPLUNC family, containing PSP), and the BPIFB subfamily (previously termed LPLUNCs). The history of cloning the PSP/Splunc/BPIFA family, and changes in the nomenclature have recently been described in depth (Bingle et al. 2011). An important observation is that proteins in this superfamily have quite divergent amino acid sequences. For example, the optimal possible alignment of rat PSP to rat BPI gives a sequence identity of only 19%. However, there is strong conservation of secondary structure and predicted tertiary structures across the superfamily. The crystal structure of BPI shows a hollow boomerang-shaped structure (Beamer et al. 1997). The two halves (domains) of the boomerang show clear similarity to each other at both the primary sequence level, and at the structural level. The Long PLUNC (BPIFB) subfamily maps across both the two domains, whereas PSP and the rest of the Short PLUNC (BPIFA) subfamily consists of only one domain. Many of these proteins have important roles that involve binding to lipids. For example, both BPI and LBP bind to

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 53

Fig. 9. PtdIns(3,4)P2 co-localizes to secretory granules of parotid acinar cells. Rat parotid tissue sections were probed with anti-rPSP (red) and monoclonal anti-PtdIns(3,4)P2 (green). A merge of the two images shows co-localization (C). The dotted line marks the boundary of the acinus. The boxed area of the Nomarski image (D) is enlarged in the inset. The scale bar

The observation that PtdIns(3,4)P2 is present in parotid granule membranes supports our model that PSP binds the membrane by interacting with this lipid. However, the PI-kinases that produce PtdInsPs are located on the cytosolic leaflet of the TGN and granule

The coat protein complex on budding vesicles of the TGN includes translocases which flip phospholipids to maintain lipid asymmetry, and can contribute to bending of membranes (van Meer and Sprong 2004; Natarajan et al. 2009; Contreras et al. 2010). Translocases (flippases) are reported on the TGN and post-Golgi vesicles, and are linked to vesicle budding (Muthusamy et al. 2009). Translocases on yeast TGN are involved in a clathrindependent pathway, vesicle bud formation, and membrane trafficking (Graham 2004; Daleke 2007; Natarajan et al. 2009). Translocase activity is also present on pig gastric parietal cell secretory vesicles (Suzuki et al. 1997), and adrenal chromaffin granules (Zachowski et al. 1989). Three types of lipid translocases have been described at the TGN or on post-Golgi vesicles. P-type ATPase translocases are present on the TGN and secretory granules (Suzuki et al. 1997), and are important for secretion (Muthusamy et al. 2009). Alternatively, the ATPbinding cassette (ABC) superfamily of transporters includes lipid translocases which can be found on the TGN, lysosomes, and secretory vesicles of lung type II cells (Stahlman et al. 2007). Similarly, the phospholipid scramblase family mediates bidirectional flipping of

**4.5 Lipid translocases in parotid granule membranes** 

membranes, yet PSP is present only inside the secretory granule.

is 5 µm.

lipopolysaccaride (LPS) by the lipid A region containing multiple acyl chains. The crystal structure of BPI shows the presence of two bound phosphatidylcholine molecules, each with the acyl chains deeply embedded in the hollow protein tube (Beamer et al. 1997). A similar model is found for CETP, and PLTP (Huuskonen et al. 1999; Qiu et al. 2007). This is distinctly different from PSP binding to lipids observed in our results. Using two different assays, PSP does not bind to phosphatidylcholine. Further, PSP requires critical interactions with the headgroup of the lipid, which are not apparent in BPI binding to phosphatidylcholine. Based on these differences, it does not appear that PSP binding to PtdIns(3,4)P2 uses the same mechanism as the well-characterized binding of BPI to lipids. As an initial hypothesis, we predict that separate binding sites on PSP will be identified for binding PtdIns(3,4)P2 and lipopolysaccaride. PSP could bind PtdIns(3,4)P2 in membranes during trafficking, and subsequently use a different interaction to bind LPS in the saliva.

#### **4.4 Phosphoinositides in parotid granule membranes**

The observed binding of PSP to PtdIns(3,4)P2 could support sorting of PSP; and in addition PSP could act as a membrane tether (or chaperone) to aid in sorting of other cargo. However, PtdIns(3,4)P2 binding can mediate selective trafficking into granules only if the presence of PtdInsP2 is somehow linked to the identity of the forming granule. The idea that a specific type of rare lipid may direct the sorting of cargo proteins is an entirely novel suggestion. However, it is just an extrapolation of the role of PtdInsPs on the other side of the membrane. In the following sections we address two key questions. Is PtdIns(3,4)P2 present on parotid secretory granule membranes? Can PtdIns(3,4)P2 cross to the luminal side of the granule membrane?

PtdInsPs have highly specific intracellular distributions, anchoring critical proteins to specific membranes (Graham and Burd 2011). PtdIns(3,4)P2 could reasonably be present in the TGN since both PI3-kinase and PI4-kinase are bound at the TGN, and both PI-kinase activities are required for regulated secretion (Meunier et al. 2005; Low et al. 2010). However, PtdIns(3,4)P2 has tended to be neglected in studies of PtdInsP distribution, so little information is available. Nonetheless, immunofluorescence of a transfected PtdIns(3,4)P2-binding protein (TAPP1) showed clear localization to a Golgi-like structure adjacent to the nucleus (Hogan et al. 2004). We isolated parotid gland secretory granules, and methanolic extracts of purified granule membranes were spotted on PVDF and probed with antibodies to specific phosphoinositides. As expected, we detected PtdIns(4)P which is reported to be on vesicles and TGN of several cell types. In addition, strong immunoreactivity was observed for PtdIns(3,4)P2 (Venkatesh et al. 2011). Standard curves of PtdIns(4)P and PtdIns(3,4)P2 were used to calculate the amounts of each lipid. PtdIns(4)P is abundant in the TGN, however, we find that PtdIns(3,4)P2 is present at a slightly higher amount than PtdIns(4)P in parotid granule membranes.

As a separate approach, we used immunofluorescence to localize PtdIns(3,4)P2 within the parotid acinar cell. Anti-PSP labels the secretory granules, which collect near the center of the acinus, in the apical end of each cell, but did not label the basal ends. Anti-PtdIns(3,4)P2 also labeled the apical end of parotid acinar cells, giving a similar pattern. Superimposing the images shows that PSP and PtdIns(3,4)P2 co-localize to the secretory granules (Fig. 9; yellow and orange). Therefore, using either biochemical or histological methods, we consistently find that PtdIns(3,4)P2 is present in parotid granule membranes.

lipopolysaccaride (LPS) by the lipid A region containing multiple acyl chains. The crystal structure of BPI shows the presence of two bound phosphatidylcholine molecules, each with the acyl chains deeply embedded in the hollow protein tube (Beamer et al. 1997). A similar model is found for CETP, and PLTP (Huuskonen et al. 1999; Qiu et al. 2007). This is distinctly different from PSP binding to lipids observed in our results. Using two different assays, PSP does not bind to phosphatidylcholine. Further, PSP requires critical interactions with the headgroup of the lipid, which are not apparent in BPI binding to phosphatidylcholine. Based on these differences, it does not appear that PSP binding to PtdIns(3,4)P2 uses the same mechanism as the well-characterized binding of BPI to lipids. As an initial hypothesis, we predict that separate binding sites on PSP will be identified for binding PtdIns(3,4)P2 and lipopolysaccaride. PSP could bind PtdIns(3,4)P2 in membranes during trafficking, and subsequently use a different interaction to bind LPS in the saliva.

The observed binding of PSP to PtdIns(3,4)P2 could support sorting of PSP; and in addition PSP could act as a membrane tether (or chaperone) to aid in sorting of other cargo. However, PtdIns(3,4)P2 binding can mediate selective trafficking into granules only if the presence of PtdInsP2 is somehow linked to the identity of the forming granule. The idea that a specific type of rare lipid may direct the sorting of cargo proteins is an entirely novel suggestion. However, it is just an extrapolation of the role of PtdInsPs on the other side of the membrane. In the following sections we address two key questions. Is PtdIns(3,4)P2 present on parotid secretory granule membranes? Can PtdIns(3,4)P2 cross to the luminal

PtdInsPs have highly specific intracellular distributions, anchoring critical proteins to specific membranes (Graham and Burd 2011). PtdIns(3,4)P2 could reasonably be present in the TGN since both PI3-kinase and PI4-kinase are bound at the TGN, and both PI-kinase activities are required for regulated secretion (Meunier et al. 2005; Low et al. 2010). However, PtdIns(3,4)P2 has tended to be neglected in studies of PtdInsP distribution, so little information is available. Nonetheless, immunofluorescence of a transfected PtdIns(3,4)P2-binding protein (TAPP1) showed clear localization to a Golgi-like structure adjacent to the nucleus (Hogan et al. 2004). We isolated parotid gland secretory granules, and methanolic extracts of purified granule membranes were spotted on PVDF and probed with antibodies to specific phosphoinositides. As expected, we detected PtdIns(4)P which is reported to be on vesicles and TGN of several cell types. In addition, strong immunoreactivity was observed for PtdIns(3,4)P2 (Venkatesh et al. 2011). Standard curves of PtdIns(4)P and PtdIns(3,4)P2 were used to calculate the amounts of each lipid. PtdIns(4)P is abundant in the TGN, however, we find that PtdIns(3,4)P2 is present at a slightly higher

As a separate approach, we used immunofluorescence to localize PtdIns(3,4)P2 within the parotid acinar cell. Anti-PSP labels the secretory granules, which collect near the center of the acinus, in the apical end of each cell, but did not label the basal ends. Anti-PtdIns(3,4)P2 also labeled the apical end of parotid acinar cells, giving a similar pattern. Superimposing the images shows that PSP and PtdIns(3,4)P2 co-localize to the secretory granules (Fig. 9; yellow and orange). Therefore, using either biochemical or histological methods, we

consistently find that PtdIns(3,4)P2 is present in parotid granule membranes.

**4.4 Phosphoinositides in parotid granule membranes** 

amount than PtdIns(4)P in parotid granule membranes.

side of the granule membrane?

Fig. 9. PtdIns(3,4)P2 co-localizes to secretory granules of parotid acinar cells. Rat parotid tissue sections were probed with anti-rPSP (red) and monoclonal anti-PtdIns(3,4)P2 (green). A merge of the two images shows co-localization (C). The dotted line marks the boundary of the acinus. The boxed area of the Nomarski image (D) is enlarged in the inset. The scale bar is 5 µm.

#### **4.5 Lipid translocases in parotid granule membranes**

The observation that PtdIns(3,4)P2 is present in parotid granule membranes supports our model that PSP binds the membrane by interacting with this lipid. However, the PI-kinases that produce PtdInsPs are located on the cytosolic leaflet of the TGN and granule membranes, yet PSP is present only inside the secretory granule.

The coat protein complex on budding vesicles of the TGN includes translocases which flip phospholipids to maintain lipid asymmetry, and can contribute to bending of membranes (van Meer and Sprong 2004; Natarajan et al. 2009; Contreras et al. 2010). Translocases (flippases) are reported on the TGN and post-Golgi vesicles, and are linked to vesicle budding (Muthusamy et al. 2009). Translocases on yeast TGN are involved in a clathrindependent pathway, vesicle bud formation, and membrane trafficking (Graham 2004; Daleke 2007; Natarajan et al. 2009). Translocase activity is also present on pig gastric parietal cell secretory vesicles (Suzuki et al. 1997), and adrenal chromaffin granules (Zachowski et al. 1989). Three types of lipid translocases have been described at the TGN or on post-Golgi vesicles. P-type ATPase translocases are present on the TGN and secretory granules (Suzuki et al. 1997), and are important for secretion (Muthusamy et al. 2009). Alternatively, the ATPbinding cassette (ABC) superfamily of transporters includes lipid translocases which can be found on the TGN, lysosomes, and secretory vesicles of lung type II cells (Stahlman et al. 2007). Similarly, the phospholipid scramblase family mediates bidirectional flipping of

Phosphatidylinositol Bisphosphate Mediated Sorting of Secretory Granule Cargo 55

As discussed above, a central aspect of selective trafficking is ensuring that the vesicle targeting information encoded by the granule coat matrix directs the choice of which cargo proteins fill the lumen of the granule. Current models for selective sorting all rely on interactions with a transmembrane protein to convey that information, directly or indirectly. Our results with parotid secretory granules suggest the possibility of a variant of this model. Rather than a transmembrane sorting receptor, translocation (flipping) of a rare lipid,

Both PI3-kinase and PI4-kinase are present on the TGN or vesicles in several cell types, and create specific PtdInsPs in the outer leaflet of the membrane. In addition, lipid translocases, or flippase activity, has been reported on mammalian secretory granules. Our results demonstrate that PtdIns(3,4)P2 is present in parotid secretory granule membranes. We observe that PtdIns(3,4)P2 can translocate to the luminal bilayer of the granule membrane. Further, we find that PSP binds strongly to PtdIns(3,4)P2 in the membrane. Since the translocase likely is recruited and localized by the coat proteins, we suggest that the translocase may create a local region of higher concentration of PtdIns(3,4)P2 in the luminal leaflet within the forming vesicle bud, compared to other areas of the TGN. This may serve to localize PSP within the budding vesicle, thereby sorting it for secretion. The membranebound PSP may in turn act as a sorting chaperone for other cargo proteins. This hypothetical model has the advantage that it suggests specific interactions which can be tested for a role

Fig. 11. A hypothetical model for how PtdIns(3,4)P2 could mediate sorting of PSP.

activity. J. Cell Biol. 103(4): 1257-1267.

Work described in this chapter was supported by NIH NIDCR grants DE012205 and

Arvan, P. and Castle, J. (1986) Isolated secretion granules from parotid glands of chronically

stimulated rats possess an alkaline internal pH and inward-directed H+ pump

**5. Conclusions: Lipids as sorting receptors** 

in sorting for secretion.

**6. Acknowledgments** 

DE019243.

**7. References** 

PtdIns(3,4)P2, may convey the character of the coat matrix (Fig. 11).

phospholipid across the membrane, and PLSCR1 is present on neutrophil secretory vesicle membranes (Frasch et al. 2004). Un-phosphorylated PtdIns can be translocated by a flippase, however, none of the translocase families have been tested for the ability to flip PtdInsPs.

Our data suggest that PtdIns(3,4)P2 may be present on the luminal face of the parotid granule membrane as a binding site for PSP, however, it is unclear how it would get there. Therefore, intact rat parotid secretory granules were incubated with fluorescent NBDtagged PtdInsP to measure flipping, according to (Natarajan and Graham 2006; Natarajan et al. 2009). In this method, added NBD-lipids integrate rapidly into the outer leaflet of the granule membrane on ice. After incubation at 37 oC to allow flipping, label remaining in the outer leaflet is quantitatively destroyed by addition of BSA and dithionite, however, NBDlipids which had translocated to the luminal leaflet are protected by the membrane. We found that incubation for 1 hour at 37 oC allowed 15% of integrated NBD-PtdIns(3,4)P2 to translocate to the protected inner leaflet (Fig. 10). Less than 2% of the PSP leaked from the granules after incubation at 37 oC for 1 h, indicating that the granules remained sealed during the assay.

Unphosphorylated PtdIns is translocated by a flippase at a rate similar to phosphatidylcholine (Vishwakarma et al. 2005). This provides us with a benchmark for comparison with the extent of flipping of PtdInsPs. In our experiments, approximately 10% of PtdIns was flipped to the inner leaflet of parotid granules, whereas 15-18% of PtdIns(4)P, PtdIns(3,4)P2, or PtdIns(3,5)P2 was translocated (Fig. 10). Hence, the flippase activity is relatively non-selective, and the phosphate groups do not inhibit translocation. Parotid granule membranes support flipping of phosphorylated forms to a greater extent than PtdIns. Taken together, our results demonstrate that PtdIns(3,4)P2 is present in the granule membrane, and can flip to the inner face of the membrane. Further, the presence of PtdInsPs on the luminal leaflet of membranes raises the possibility that other intra-organelle proteins may be localized by PtdInsP anchors.

Fig. 10. A. Parotid secretory granules translocate PtdIns(3,4)P2. Purified intact parotid secretory granules were incubated with fluorescent NBD-labeled PtdIns(3,4)P2 to measure translocation, as described (Natarajan and Graham 2006; Venkatesh et al. 2011). Data show the amount of lipid flipped to the inner leaflet. Data are Mean±SE of 3 experiments in triplicate. \*p<0.01 compared with 0h. B. Several PtdInsPs translocate parotid granule membranes. Data are the amount of flipped lipid at 0 or 1 hour. Data are Mean ± SE of 3-6 experiments in triplicate. \*p<0.05, \*\*\*p<0.01 compared to PtdIns.
