Regulated exocytosis in chromaffin cells. A potential role for a secretory granule-associated ARF6 protein.

The ADP-ribosylation factor (ARF) GTP-binding proteins are believed to function as regulators of vesicular budding and fusion along the secretory pathway. To investigate the role of ARF in regulated exocytosis, we have examined its intracellular distribution in cultured chromaffin cells by subcellular fractionation and immunoreplica analysis. We found that ARF6 is specifically associated with the membrane of purified secretory chromaffin granules. Chemical cross-linking and immunoprecipitation experiments suggested that ARF6 may be part of a complex with βγ subunits of trimeric G proteins. Stimulation of intact chromaffin cells or direct elevation of cytosolic calcium in permeabilized cells triggered the rapid dissociation of ARF6 from secretory granules. This effect could be inhibited by AlF4− which selectively activates trimeric G proteins. Furthermore, a synthetic myristoylated peptide corresponding to the N-terminal domain of ARF6 strongly inhibited calcium-evoked secretion in streptolysin-O-permeabilized chromaffin cells. The possibility that ARF6 plays a role in the effector pathway by which trimeric G proteins control exocytosis in chromaffin cells is discussed.

The ADP-ribosylation factor (ARF) GTP-binding proteins are believed to function as regulators of vesicular budding and fusion along the secretory pathway. To investigate the role of ARF in regulated exocytosis, we have examined its intracellular distribution in cultured chromaffin cells by subcellular fractionation and immunoreplica analysis. We found that ARF6 is specifically associated with the membrane of purified secretory chromaffin granules. Chemical cross-linking and immunoprecipitation experiments suggested that ARF6 may be part of a complex with ␤␥ subunits of trimeric G proteins. Stimulation of intact chromaffin cells or direct elevation of cytosolic calcium in permeabilized cells triggered the rapid dissociation of ARF6 from secretory granules. This effect could be inhibited by AlF 4 ؊ which selectively activates trimeric G proteins. Furthermore, a synthetic myristoylated peptide corresponding to the N-terminal domain of ARF6 strongly inhibited calciumevoked secretion in streptolysin-O-permeabilized chromaffin cells. The possibility that ARF6 plays a role in the effector pathway by which trimeric G proteins control exocytosis in chromaffin cells is discussed.
Studies on diverse secretory cell types have highlighted the importance of GTP-binding proteins in Ca 2ϩ -regulated exocytosis. Proteins able to bind and hydrolyze GTP can be subdivided into different families including (i) trimeric G proteins, (ii) ras-related low molecular mass G proteins, and (iii) ADPribosylation factor (ARF) 1 proteins. Trimeric G proteins have been found associated with the membrane of secretory granules in various secretory cells (1)(2)(3), suggesting a role in exocytosis. Indeed, the participation of a plasma membrane-bound G i3 protein in the late stages of exocytosis in mast cells has been demonstrated (4). Direct control of exocytosis by G i and G o proteins has also been described in insulin-secreting cells (5) and in chromaffin cells (3,6). Thus regulated exocytosis appears as a possible effector pathway for trimeric G proteins although the mechanisms by which this class of G proteins relates to the exocytotic machinery remain to be elucidated.
Besides heterotrimeric G proteins, recent studies suggest that a subset of small GTPases of the ras superfamily may also participate in the control of calcium-dependent exocytosis. For example, rab3 isoforms which are specifically localized on the membrane of secretory vesicles (7,8), have been proposed to serve as a regulator of exocytosis in chromaffin cells (9) and in anterior pituitary cells (10). Other investigations have postulated the participation of ARF in exocytosis (11,12).
ARF is a monomeric GTP-binding protein that appears to control a wide range of vesicle transport and fusion steps along the secretory pathway. This may include budding, transport, and fusion steps in the Golgi complex (13,14), in the endoplasmic reticulum (15), and in the endocytotic pathway (16). Initial evidence for the involvement of ARF in the secretory pathway came from genetic studies in yeast Saccharomyces cerevisiae where deletion of the ARF gene resulted in a secretory defect (17). In mammalian cells, ARF proteins are classified into three families according to their size and sequence homology (18). ARF1, ARF2, and ARF3 form class I, ARF4 and ARF5 form class II, and ARF6 forms class III. The best characterized ARF protein is ARF1. It is localized on the Golgi apparatus and is involved in the recruitment of cytosolic coat proteins to membranes during the formation of transport vesicles (19). ARF6 has been recently localized in the endocytotic pathway in transfected mammalian cells, where it appears to control the fusion of endosomes (16).
The aim of the present work is to assess the possible involvement of ARF in regulated exocytosis. We report that ARF6 is specifically associated to the membrane of secretory chromaffin granules, presumably through an interaction with G␤␥ subunits. Stimulation of chromaffin cells triggered the dissociation of ARF6 from secretory granules. This dissociation can be blocked by aluminium fluoride, an activator of trimeric G proteins. We propose that ARF may be part of the effector pathway by which trimeric G proteins control the exocytotic pathway in chromaffin cells.

Culture of Chromaffin Cells and [ 3 H]Noradrenaline
Release Experiments-Bovine adrenal chromaffin cells were isolated and maintained in primary culture essentially as described previously (6,20). Cells were loaded with [ 3 H]noradrenaline (14.68 Ci/mmol, Dupont NEN) and washed with Locke's solution (6). To trigger release from intact cells, chromaffin cells were stimulated for an indicated period of time with Locke's solution containing 10 M nicotine (6). Permeabilization of chromaffin cells was performed with streptolysin-O (SLO; Institut Pasteur, Paris, France) for 2 min in Ca 2ϩ -free KG medium (3). Cells were then stimulated for 10 min, in the presence of the compound to be tested, in KG medium containing 20 M free Ca 2ϩ and 1 mM free Mg 2ϩ (3).
Subcellular Fractionation of Cultured Chromaffin Cells-Chromaffin granules were purified from cultured chromaffin cells on 1.6 M sucrose gradients as described previously (21). Granule membranes were obtained after three freezing/thawing steps and centrifugation at 50,000 ϫ g. Purified cytosol was obtained from the 20,000 ϫ g supernatant, which was further centrifuged 30 min at 100,000 ϫ g. Golgi membranes were prepared from cultured chromaffin cells essentially as described previously (22).
Alternatively, the crude membrane fraction (20,000 ϫ g pellet) was layered on a continuous sucrose density gradient (sucrose 1-2.2 M, 10 mM Tris, pH 7.4) and centrifuged for 1 h at 100,000 ϫ g. Twelve 1-ml fractions were collected from the top to bottom and analyzed for protein content. The distribution of the chromaffin granule marker chromogranin A in the fractions of the gradient was estimated by the ELISA technique (20). The plasma membrane marker Na ϩ /K ϩ -ATPase was assayed as described previously (3).
Immunoprecipitation Assays-Chromaffin granule membranes (2.5 mg) were immunoprecipitated with 50 l of anti-dopamine ␤-hydroxylase antiserum and 250 l of ImmunoPure immobilized protein A (Pierce) (7). Proteins linked to the beads were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and immunoblotting.
Cross-linking of Chromaffin Granule Membranes-Chromaffin granule membranes were cross-linked with 5 mM disuccinimyl suberate exactly as described previously (29). Membranes were collected by centrifugation at 100,000 ϫ g for 15 min, and proteins were analyzed by SDS-PAGE and immunoblotting.
Electrophoresis and Immunoblotting-One-dimensional SDS-PAGE was performed on 15% acrylamide gels in Tris-glycine buffer (30). Proteins were transferred to nitrocellulose sheets (30) at a constant current of 120 mA for 1 h. Blots were developed with secondary antibodies coupled to horseradish peroxidase (Amersham, Les Ulis, France), and immunoreactive bands were detected with the ECL system (Amersham). The conditions for separation of proteins by electrophoresis on two-dimensional gels were as described previously (30).

Association of ARF6 with Chromaffin Granule Membranes-
The intracellular distribution of ARF proteins in chromaffin cells was estimated by immunoreplica analysis using a monoclonal anti-ARF antibody (1D9). As illustrated in Fig. 1A, ARF was detected in the cytosol and in the Golgi apparatus, display-ing an apparent molecular mass of 20 kDa. ARF was also found associated to chromaffin granule membranes but with an apparent molecular mass of 18 kDa. The subtype of ARF proteins was further defined with several antibodies able to distinguish between the different classes of ARF. We found that the 20-kDa ARF protein detected in the cytosolic and Golgi fractions was specifically labeled with the affinity-purified anti-ARF1 antibody (Fig. 1A) and with the anti-sARFII antiserum (data not shown) but it was not recognized by the anti-ARF6 antibody (Fig. 1A) or by antisera against ARF5 or ARF6 (data not shown). In contrast, the 18-kDa ARF protein found associated to chromaffin granule membranes was selectively labeled with the anti-ARF6 antibody (Fig. 1A) and with anti-ARF6 antiserum (data not shown). Accordingly, ARF6 migrates slightly faster than ARF1 on SDS-polyacrylamide gel electrophoresis (31).
The specific association of ARF6 with chromaffin granule membranes was further substantiated by immunoblotting analysis of fractions collected from a sucrose density gradient layered with a crude membrane preparation (Fig. 1B). The highest immunosignal for ARF1 was detected in fractions 2 and 3 containing the plasma membranes as estimated by the Na ϩ / K ϩ -ATPase activity. Fractions 11 and 12, enriched in chromaffin granules revealed by the peak of chromogranin A, were strongly labeled with the anti-ARF6 antibody but showed no immunoreaction with anti-ARF1 antibody (Fig. 1B).
We also used an immunoadsorption procedure to exclude the association of ARF6 with a contaminating organelle that may FIG. 1. Immunochemical detection of ARF in subcellular fractions from cultured chromaffin cells. A, proteins (30 g) from chromaffin granule membranes (CGM), Golgi membranes (GM), and cytosol (C) were subjected to SDS-PAGE and immunodetection on nitrocellulose sheets using anti-ARF antibody 1D9, affinity-purified anti-ARF1, and anti-ARF6 (total IgG fraction) antibodies. B, 12 fractions collected from a continuous sucrose density gradient layered with the crude chromaffin membrane pellet were assayed for Na ϩ /K ϩ -ATPase and chromogranin A. Proteins (15 g) from each fraction were separated by SDS-PAGE and analyzed by immunoblotting with affinitypurified anti-ARF1 and anti-ARF6 (total IgG fraction) antibodies. Plasma membranes (Na ϩ /K ϩ -ATPase) were recovered in fractions 2 and 3 and chromaffin granules (chromogranin A) were present in fractions 11 and 12. C, chromaffin granule membranes obtained from fraction 12 were immunoprecipitated with anti-dopamine ␤-hydroxylase antibodies and protein A-Sepharose (IP). Control experiments were performed with protein A-Sepharose in the absence of anti-dopamine ␤-hydroxylase antibodies (C). Precipitated proteins were separated by SDS-PAGE, transferred and analyzed with anti-ARF 1D9 and anti-dopamine ␤-hydroxylase antibodies. D, chromaffin granule membranes (1 mg) were separated by two-dimensional SDS-PAGE (15% acrylamide), transferred on nitrocellulose membranes, and probed with anti-ARF6 antibodies (total IgG fraction). comigrate with chromaffin granules in the sucrose density gradients. Dopamine ␤-hydroxylase is a specific marker for chromaffin granules. Fig. 1C shows that preincubation of purified granule membranes with an anti-dopamine ␤-hydroxylase antibody followed by protein A-Sepharose addition resulted in the coadsorption of ARF6 and dopamine ␤-hydroxylase. Thus ARF6 is an associated component of the chromaffin granule membrane.
Two-dimensional gel electrophoresis of purified chromaffin granule membranes revealed the presence of two spots labeled with the anti-ARF6 antibody. These two spots displayed only a different isoelectric point value (respectively, 7.7 and 7.9), but no difference in the apparent molecular mass (Fig. 1D). Occasionally, the ARF-labeled band present in chromaffin granule membranes could also be resolved into two components on monodimensional gels (Fig. 1A). This observation may indicate some post-translational modification, such as phosphorylation or sulfation of the granule-associated ARF protein, although such modifications have not yet been reported for ARF proteins.
We performed protein cross-linking experiments to confirm the possible ARF6-G␤ interaction. Fig. 2B shows that crosslinking of chromaffin granule-associated proteins with disuccinimyl suberate resulted in the appearance of a 50-kDa adduct that reacted with antibodies against both ARF and G␤. With the exception of bands corresponding approximately to the mass of ARF dimers, no additional adducts were observed (Fig.  2B). Therefore, it is likely that this additional 50-kDa band is due to the cross-linking of ARF6 (18 kDa) to G␤ (34 kDa), which is in agreement with the idea that ARF6 and G␤ may be bound to each other in the granule membrane.

ARF6 may represent an important component of the regulated exocytotic machinery in chromaffin cells.
ARF6 Dissociates from Chromaffin Granule Membranes during Exocytosis-In order to compare the intracellular distribution of ARF6 in resting and stimulated chromaffin cells, secretory granules were purified by sucrose density gradients from a crude membrane fraction obtained either from intact cells stimulated with 10 M nicotine or from SLO-permeabilized cells challenged with 20 M free Ca 2ϩ . The amount of ARF6 associated to granule membranes was analyzed by immunodetection with the 1D9 anti-ARF antibody (Fig. 4). Stimulation of chromaffin cells strongly reduced the quantity of ARF6 associated with chromaffin granule membranes. In intact cells, nicotine reduced by 70% the amount of ARF6 bound to granules (Fig. 4). In permeabilized cells, calcium decreased the quantity of ARF6 recovered in the chromaffin granule membrane fraction by approximately 60% (Fig. 4). Parallel analysis with the anti-ARF6 antibody revealed a similar decrease in the amount of granule-associated ARF6 in stimulated cells (data not shown). Furthermore, the quantity of ARF6 detected with the 1D9 antibody in the crude membrane fraction remained unchanged in resting and nicotine-stimulated cells (data not shown), indicating that the reduction of ARF6 associated to the granule membrane fraction was not due to a possible degradation or a covalent modification of the protein in stimulated cells. These results suggest that the exocytotic process may be accompanied by the translocation of ARF6 from the secretory granule to some unknown subcellular compartment. Since ARF6 was not detected among the cytosolic proteins that escape through the pores created in the plasma membrane of permeabilized cells (data not shown), we assume that ARF6 translocates from the granule membrane to another membrane-bound compartment in stimulated cells.
We examined the time course of ARF6 translocation in nicotine-stimulated cells. As illustrated in Fig. 5, the dissociation of ARF6 from the chromaffin granules was rapid and almost complete within 1 min of stimulation. By comparison, maximal [ 3 H]noradrenaline release required 2-3 min of stimulation (Fig. 5), suggesting that the dissociation of ARF6 may represent an early event in the pathway of regulated exocytosis. Note that G␤ remained attached to the granule membrane during the entire stimulation period (Fig. 5).
The presence of G␣ o on the chromaffin granule membrane (3), which displays a high affinity for G␤␥ and may thereby compete with ARF for the binding to G␤ (32,33), led us to investigate whether AlF 4 Ϫ may interfere with the translocation of ARF6 in stimulated cells. AlF 4 Ϫ is an activator of trimeric G proteins which prevents the association of the ␣ subunit with ␤␥ subunits (34). In the presence of AlF 4 Ϫ , the amount of ARF6 detected on chromaffin granule membranes was comparable in resting and calcium-stimulated cells (Fig. 6), indicating that Ϫ (20 mM NaF ϩ 50 M AlCl 3 ). Chromaffin granule membranes (30 g) were subsequently purified and the content of ARF6 was estimated by immunodetection on nitrocellulose blots using anti-ARF 1D9 antibodies. B, the intensity of the ARF immunoreactive bands was quantified with the Fuji ␤ phosphoimager. Results are expressed relative to the amount of ARF detected in granule membranes in resting cells (incubated in the absence of Ca 2ϩ ). GTP␥S or AlF 4 Ϫ did not modify the amount of ARF recovered in chromaffin granule membranes in resting cells (data not shown).

AlF 4
Ϫ was able to inhibit the translocation of ARF6 induced by a rise of cytosolic calcium.
Studies with purified proteins indicate that ARF binds to G␤␥ in its GDP-liganded form (32,33). We examined the effect of the nonhydrolyzable GTP analogue, GTP␥S, on the attachment of ARF6 to chromaffin granule membranes. As shown in Fig. 6, the presence of GTP␥S in the incubation medium did not significantly affect the calcium-induced dissociation of ARF6 from secretory granules in permeabilized chromaffin cells. DISCUSSION The current general model of vesicular fusion suggests that the so-called SNARE complex functions in many fusion events throughout the secretory pathway, including transport at the endoplasmic reticulum and the Golgi apparatus, and exocytotic and endocytotic fusions (35). Cumulative evidence also implicates ARF as a molecular switch in the regulation of vesicular transport and fusion through early compartments of the secretory pathway (reviewed in Refs. 19, 36, and 37). These apparent similarities of the mechanisms involved in intracellular vesicular trafficking led us to investigate whether ARF may also be a component of the regulated exocytotic machinery. We report here the presence of an immunologically related ARF6 protein on the membrane of purified secretory chromaffin granules. Stimulation of chromaffin cells by a cholinergic agonist or a direct elevation of cytosolic calcium triggers a rapid decrease of the amount of ARF6 associated to granule membranes, suggesting that exocytotic fusion is accompanied by the release of ARF6 from secretory granules in chromaffin cells. It is noteworthy that a similar sequence of events has been described in the Golgi where the release of ARF from vesicles seems to be a prerequisite for fusion with the acceptor membrane (35).
ARF was originally discovered as a cofactor required for the ADP-ribosylation by cholera toxin of the heterotrimeric G protein G s (39). Further studies confirmed the close relationships between trimeric G proteins and ARF. For example trimeric G proteins seem to regulate the association of ARF and coat components onto the Golgi membrane (23,40,41). A collaboration among trimeric G proteins and ARF has also been suggested in the control of endosomal fusion (32). Finally direct interactions between ARF and the G␤␥ subunits have been described (32,33). Indeed, ARF proteins have more in common with G protein ␣ subunits than with other small GTPases when comparing GTP-binding concensus sequences and cotranslational modifications such as myristoylation of the N terminus (37). By cross-linking and coimmunoprecipitation experiments, we observed here that ARF6 interacts with G␤ in chromaffin granule membranes. Since G␤ exists only in close association with G␥ (38), we propose that the G␤␥ complex may represent the membrane receptor that could stabilize the interaction of ARF6 with secretory granules.
In chromaffin cells, trimeric G proteins control the ATP-dependent priming step and the late calcium-regulated fusion event of exocytosis (3,6). However, their downstream effectors in the exocytotic machinery remain unknown. We found that AlF 4 Ϫ can prevent the calcium-induced uncoupling of ARF6-G␤␥ on the secretory granule membrane in stimulated cells. Chromaffin granule membranes contain both G␣ o and G␤␥ subunits (3). Thus, upon the activation by AlF 4 Ϫ , G␣ o is likely to dissociate from G␤␥ which may thereby provide a site of interaction for ARF6 and prevent its translocation in response to calcium stimulation. Conversely, calcium may trigger the dissociation of ARF6 from the granule membrane by allowing the reassociation of G␣ o with G␤␥ which becomes then unable to retain ARF6. AlF 4 Ϫ , as well as mastoparan, inhibit calciumevoked catecholamine secretion in permeabilized chromaffin cells by a mechanism which can be selectively reversed by anti-G␣ o antibodies (6). This suggests that G␣ o blocks the exocytotic machinery when activated and dissociated from G␤␥. Thus, our data support a model in which exocytosis requires the inactivation of the granule-associated G o . In stimulated cells, the activation of G␣ o may be impaired by a calcium-dependent process and G␣ o reassociates with G␤␥. This triggers the release of ARF6 from G␤␥, an event which may be necessary for the subsequent exocytotic steps to occur.
We investigated the possible function of ARF6 in the exocytotic machinery by using synthetic peptides corresponding in sequence to the N-terminal residues of ARF1 and ARF6. The N-terminal region of ARF was chosen because it has been implicated as the putative effector domain. Indeed, the N terminus of ARF seems to be critical for its function since deletion of this region results in a global reduction of ARF activities (42). Furthermore, N-terminal ARF peptides have proved to be potent inhibitors of ARF functions, including endoplasmic reticulum to Golgi transport, intra-Golgi transport and endocytotic vesicle fusion (15,40,42,43). The strong inhibition of calcium-evoked catecholamine secretion by the myristoylated ARF6 (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13) peptide implicates ARF6 as a key component of the exocytotic pathway in chromaffin cells. Recent studies in HL60 cells have shown that ARF contributes to the restoration of secretion in permeabilized cytosol-depleted cells (44). Hence, the data presented here are convergent with the idea that ARF proteins may play a general role in the exocytotic pathway in endocrine and neuroendocrine cells.
In conclusion, based on subcellular fractionation techniques and immunological detection, the present results strongly suggest that ARF6 dissociates from secretory granules in stimulated chromaffin cells. Thus, an important future undertaking will be to confirm this observation by immunofluorescence and confocal microscopy since we cannot completely rule out some artefactual redistribution due to cell lysis. Nevertheless, the finding that ARF6 interacts with G␤ in the secretory granule membrane offers an attractive speculation: the dissociation of ARF6 from secretory granules may allow the interaction of ARF6 with its nucleotide exchange factor resulting in activation in the target membrane. In view of the current proposed functions for ARF proteins in vesicular transport, ARF6 is a good candidate to be involved in the assembly of a multisubunit complex required for fusion or to contribute to the generation of fusogenic lipids at the plasma membrane.