Heterotrimeric G proteins interact with the small GTPase ARF. Possibilities for the regulation of vesicular traffic.

Trimeric G proteins have emerged as important regulators of membrane trafficking. To explore a role for G beta gamma in endosome fusion, we have taken advantage of beta-adrenergic receptor kinase (beta ARK), an enzyme translocated to membranes by interaction with G beta gamma. The COOH terminus of beta ARK (beta ARKct) has a G beta gamma-binding domain which blocks some G beta gamma-mediated processes. We found that beta ARKct and peptide G, a peptide derived from beta ARKct, inhibit in vitro endosome fusion. Interestingly, peptide G and ARF share sequence similarity. Peptide G and beta ARKct reversed ARF-mediated inhibition of endosome fusion and blocked ARF binding to membranes. Using an ARF fusion protein, we show that both G beta gamma and G alpha s interact with the small GTPase ARF, an interaction that is regulated by nucleotide binding. We conclude that G proteins may participate in the regulation of vesicular trafficking by directly interacting with ARF, a cytosolic factor required for transport.

Trimeric G proteins have emerged as important regulators of membrane trafficking. To explore a role for G␤␥ in endosome fusion, we have taken advantage of ␤-adrenergic receptor kinase (␤ARK), an enzyme translocated to membranes by interaction with G␤␥. The COOH terminus of ␤ARK (␤ARKct) has a G␤␥-binding domain which blocks some G␤␥-mediated processes. We found that ␤ARKct and peptide G, a peptide derived from ␤ARKct, inhibit in vitro endosome fusion. Interestingly, peptide G and ARF share sequence similarity. Peptide G and ␤ARKct reversed ARF-mediated inhibition of endosome fusion and blocked ARF binding to membranes. Using an ARF fusion protein, we show that both G␤␥ and G␣s interact with the small GTPase ARF, an interaction that is regulated by nucleotide binding. We conclude that G proteins may participate in the regulation of vesicular trafficking by directly interacting with ARF, a cytosolic factor required for transport.
Vesicular membrane trafficking among intracellular compartments is now recognized to involve multiple small GTPbinding proteins including members of the Ras-like superfamily such as Rab, ARF, and Sar1 (reviewed by Goud and McCaffrey, 1991;Pryer et al., 1992;Nuoffer and Balch, 1994). The ARF family, which includes several distinct ARF proteins, seems to control the assembly of coat components on transport vesicles. ARF (ADP-ribosylation factor) was originally discovered as a cofactor required for the ADP-ribosylation by cholera toxin of the heterotrimeric G protein G s (Kahn and Gilman, 1984). The initial evidence for a role for ARF in vesicular transport came from genetic studies in yeast where deletion of the ARF1 gene resulted in a secretory defect (Stearns et al., 1990a(Stearns et al., , 1990b. Using several in vitro assays that reconstitute transport between different compartments, it has been shown that ARF is an essential component required for transport (Balch et al., 1992;Lenhard et al., 1992;Donaldson and Klausner, 1994). ARF is also required for the assembly of the coat complex on non-clathrin-coated vesicles (COP-coated vesicles) mediating transport between Golgi compartments (reviewed by Rothman and Orci, 1992;Kreis and Pepperkok, 1994;Donaldson and Klausner, 1994) and in the association of AP-1 adaptor complex to Golgi membranes, raising the possibility that ARF may also be required for the assembly of clathrin coats at the trans-Golgi network (Stamnes and Rothman, 1993;Traub et al., 1993).
A growing body of evidence indicates that heterotrimeric GTP-binding proteins (G proteins) play a crucial role in vesicular trafficking (reviewed by Bomsel and Mostov, 1992;Burgoyne, 1992;Nuoffer and Balch, 1994). Previous work from our laboratory indicates that fusion among endosomes and between phagosomes and endosomes is controlled by G proteins (Colombo et al., 1992a(Colombo et al., , 1994aBeron et al., 1995). Moreover, multiple G proteins seem to participate in different steps of transport (Stow et al., 1991;Leyte et al., 1992;Carter et al., 1993). We have reported that one of the G proteins involved in endosomal fusion is G␣ s (Colombo et al., 1994b). The role of G␣ s has also been implicated in trafficking in polarized cells (Pimplikar and Simons, 1993;Bomsel and Mostov, 1993;Barroso and Sztul, 1994;Hansen and Casanova, 1994) and in the secretory pathway . However, the actual mechanism by which these proteins regulate traffic remains poorly understood.
Classically, trimeric G proteins transduce extracellular signals to appropriate effector molecules inside the cell. G proteins are comprised of three subunits, G␣, G␤, and G␥. Binding of GTP causes the activation of the G protein and the subsequent dissociation of G␣ from G␤␥ (Gilman, 1987). It is now widely accepted that signals by both G␣ and G␤␥ are physiologically relevant. Several recent reports clearly demonstrate the prominent involvement of G␤␥ in several transmembrane signaling systems. An increasing number of G protein-coupled effectors which appear to be modulated by G␤␥ subunits have been identified (reviewed by Clapham and Neer, 1993;Sternweis, 1994). On the other hand, G␤␥ specifically mediates the translocation of cytosolic ␤-adrenergic receptor kinase (␤ARK), 1 one of the G protein-coupled receptor kinases, to the plasma membrane. This translocation allows the phosphorylation of activated receptors as part of the desensitization process . A fragment of ␤ARK corresponding to the last 222 C-terminal amino acids was found to contain the "G␤␥-binding domain" (Pitcher et al., 1992). A fusion protein corresponding to this G␤␥-binding domain blocks binding of ␤ARK to G␤␥  and prevents receptor phosphorylation. It has recently been shown that this reagent interferes with multiple G␤␥-mediated processes such as G␤␥-dependent activation of adenylyl cyclase type II, ␤ARK2 regulated olfactory signal transduction, and atrial K ϩ channel activation (Reuveny et al., 1994;Boekhoff et al., 1994;Inglese et al., 1994).
In an attempt to study the possible role of G␤␥ in the mechanism or regulation of endosome fusion we used ␤ARK Cterminal fusion protein and peptides derived from the G␤␥binding domain in a cell-free assay that reconstitutes fusion between endosomes. His6-␤ARK fusion protein completely blocked endosome fusion while His6-rhodopsin kinase had no effect. A single 28-amino acid peptide (Peptide G) derived from the targeting domain of ␤ARK was also found to inhibit fusion. Alignment of the cytosolic small GTP-binding protein ARF and peptide G reveals that they share sequence similarity. Our results suggest that ␤ARK COOH terminus and peptide G inhibit endosome fusion by blocking the interaction of G␤␥ with ARF, a cytosolic factor required for endosome fusion. In order to address this provocative hypothesis, we constructed GST-ARF fusion proteins and studied their direct interaction with purified G proteins. Our results indicate that both G␤␥ and G␣ s interact with the small GTPase ARF. Activation of G␣ s by either GTP␥S or aluminofluoride complexes completely blocked ARF-G␣ interaction, indicating that the heterotrimer is the most likely candidate for ARF-G protein interaction. Our results suggest that a direct collaboration among heterotrimeric G proteins and ARF may regulate vesicular transport.

EXPERIMENTAL PROCEDURES
Cells and Materials-J774, E-clone (mannose receptor positive), a macrophage cell line, was grown to confluence in minimum essential medium containing Earle's salts and supplemented with 10% fetal calf serum. HDP-1, a mouse IgG1 monoclonal antibody specific for dinitrophenol was isolated and mannosylated as described previously (Diaz et al., 1988;Colombo et al., 1992b). ␤-Glucuronidase was isolated from rat preputial glands and derivatized with dinitrophenol (DNP) using dinitrofluorobenzene (Diaz et al., 1988). Cytosol from J774 was the high speed supernatant of a cell homogenate obtained as described (Diaz et al., 1988) and stored at Ϫ80°C. Cytosol samples (200 l) were gel filtered through 1-ml Sephadex G-25 spin columns just before use in the fusion assay. Protein concentration after filtration was 3-5 mg/ml. The His6-fusion proteins, His6-RK carboxyl terminus and His6-␤ARK1 carboxyl terminus containing the terminal 91 amino acids of RK and the terminal 222 amino acids of ␤ARK1, were prepared and purified as described . Peptides G 1 , G 2 , and G 1 Ј, corresponding to specific ␤ARK1 and ␤ARK2 sequences were synthesized and purified as described previously . Recombinant myristoylated ARF1 and ARF4 were prepared and purified essentially as described . G␤␥ subunits were purified from bovine brain as described previously (Casey et al., 1989). Recombinant G␣ subunits were a generous gift from Dr. M. Linder (Washington University, St. Louis, MO) and Dr. J. Garrison (University of Virginia, Charlottesville, VA). All other chemicals were obtained from Sigma.
Preparation of Endocytic Vesicles-Early endosomes were loaded with mannosylated anti-DNP IgG or with DNP-␤-glucuronidase by a 5 min uptake at 37°C as described previously (Diaz et al., 1988;Colombo et al., 1992b). After ligand uptake, the macrophages (1 ϫ 10 8 cells) were washed sequentially with 150 mM NaCl, 5 mM EDTA, 10 mM phosphate buffer, pH 7.0, and with 250 mM sucrose, 0.5 mM EGTA, 20 mM HEPES-KOH, pH 7.0 (homogenization buffer), and homogenized in the latter buffer (2 ml) using a cell homogenizer (Colombo et al., 1992b). Homogenates were centrifuged at 800 ϫ g for 5 min to eliminate nuclei and intact cells, and then pelleted for 1 min at 37,000 ϫ g in a Beckman L 100 microcentrifuge. The supernatants were centrifuged for additional 5 min at 50,000 ϫ g. The pellets of this second centrifugation were enriched with 5-min endosomes. Endosomal fractions containing each probe were resuspended in homogenization buffer and then mixed in the presence of DNP-BSA as scavenger. The samples were quickly frozen in liquid nitrogen and stored at Ϫ80°C.
In Vitro Fusion Assay-Endosomal fractions were quickly thawed and mixed with fusion buffer (250 mM sucrose, 0.5 mM EGTA, 20 mM HEPES-KOH, pH 7.0, 1 mM dithiothreitol, 1.5 mM MgCl 2 , 50 mM KCl, 1 mM ATP, 8 mM creatine phosphate, 31 units/ml creatine phosphokinase, and 0.25 mg/ml DNP-BSA), supplemented with gel filtered cytosol. The samples were incubated at 37°C for 45 min and the reaction was stopped by cooling on ice. To measure the immune complexes formed, the vesicles were solubilized by adding 50 l of solubilization buffer (1% Triton X-100, 0.2% methylbenzethonium chloride, 1 mM EDTA, 0.1% BSA, 0.15 M NaCl, 10 mM Tris-HCl, pH 7.4) containing 50 g/ml DNP-BSA. For immunoprecipitation the samples were transferred to multiwell plates coated with rabbit anti-mouse IgG. After 30 -45 min of incubation at room temperature, the wells were washed three times with 300 l of solubilization buffer, and ␤-glucuronidase activity was measured using 4-methylumbelliferyl ␤-D-glucuronide as substrate in a Microplate fluorometer 7600, Cambridge Technology, Inc. (Colombo et al., 1992b). Fusion was expressed in arbitrary fluorescence units.
ARF Binding Assay-An enriched endosomal fraction was prepared by differential centrifugation as described previously (Colombo et al., 1992b). The endosomal fraction (10 -20 g of total protein) was resuspended in the fusion buffer described above, containing gel filtered cytosol (1-2 mg protein/ml). Incubations were carried out in 1.5-ml tubes (Beckman, polyallomer). Incubation volumes were 50 l. After 5 min of preincubation with the reagents to be tested, 20 M GTP␥S was added, and the samples were incubated for additional 20 min at 37°C. After incubation, the samples were washed with 1 ml of homogenization buffer containing 20 M GTP␥S and 1 mM MgCl 2 . The membranes were recovered by centrifugation for 5 min at 50,000 ϫ g. Proteins were subjected to SDS-polyacrylamide gel electrophoresis under reducing conditions and transfered onto nitrocellulose in 25 mM Tris, pH 8, 192 mM glycine, and 5% methanol at 150 mA for 1 h. ARF was detected using a polyclonal affinity purified antibody against ARF, kindly provided by J. Rothman (diluted 1:500) and horseradish peroxidase-conjugated goat anti-rabbit IgG (diluted 1:5,000). The visualization was performed using the ECL detection system (Amersham Corp.) according to the manufacturer's instructions.
Construction and Isolation of GST Fusion Proteins-cDNA corresponding to human ARF4 (a gift from Richard Kahn, NIH), and ARF4 with the first 17 amino acids deleted were amplified by the polymerase chain reaction using 5Ј primers containing BamHI sites. The GST gene fusion vector pGEX-3T (Pharmacia Biotech Inc.) was used to construct cDNAs in which the amplified cDNAs were ligated with the 3Ј-end of the coding region of GST. The clones used in these experiments were verified by sequencing using Sequenase (U. S. Biochemical Corp.). Fusion proteins constructs were introduced into the Escherichia coli strain JM101 and induced with isopropyl-1-thio-␤-D-galactopyranoside to produce GST fusion proteins.
Recombinant C-terminal half of ARF1 (ARF1ct) protein was expressed as follows: the C-terminal half of the ARF1 cDNA was amplified by polymerase chain reaction. The 5Ј-oligonucleotide primer contained a BamHI linker followed by 14 nucleotide residues downstream of nucleotide residue 307. The 3Ј-oligonucleotide primer contained an EcoRI linker followed by 16 nucleotide residues complementary to the carboxyl-terminal end of ARF1 cDNA. The amplified cDNA was digested with the restriction enzymes BamHI and EcoRI and then subcloned into the bacterial expression vector pGEX-3T. The recombinant protein was expressed as fusion protein in the E. coli strain JM109, with the NH 2 -terminal end fused to GST and induced with isopropyl-1-thio-␤-D-galactopyranoside to produce GST fusion proteins.
The fusion proteins were purified by glutathione-Sepharose either by standard techniques or using the Sarkosyl method (Frangioni and Neel, 1993). The samples were dialyzed against PBS and, if necessary, concentrated in a Centricon-10 (Amicon). GST-␤ARK COOH-terminal and GST-RK COOH-terminal fusion proteins were constructed and purified as described previously . GST-Rab5 fusion protein, constructed and purified as described (Barbieri et al., 1994), was kindly provided by Mary K. Cullen (Washington University, St. Louis, MO).
Detection of Binding of G␤␥ to Fusion Proteins-Binding of G␤␥ subunits to the purified GST fusion proteins was done essentially as described previously (Pitcher et al., 1992;Touhara et al., 1994). Briefly, purified bovine brain G␤␥ subunits (200 -300 nM) were incubated with the GST fusion proteins (600 -700 nM) for 30 min on ice in PBS containing 0.01% lubrol. When indicated, purified recombinant G␣ s were also added to the binding assay. Glutathione-Sepharose (20 l of a 50% slurry in PBS/lubrol, Sigma) was added, and incubation was continued on ice for 60 min. The Sepharose beads containing bound GST or GST fusion proteins were subsequently washed four times with PBS/lubrol (400 l), subjected to SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes as described above. Antibodies against G␤␥ subunits kindly provided by Gary Johnson (National Jewish Center, Denver) were used at a 1:250 dilution. G␣ s was detected using a polyclonal affinity purified antibody against the COOH-terminal end of G␣ s , kindly provided by Dr. Gary Johnson. Blots were developed with goat anti-rabbit IgG coupled to horseradish peroxidase and detected with the ECL detection system (Amersham) according to the manufacturer's instructions.

RESULTS
The G␤␥-binding Domain of ␤ARK Inhibits Endosome Fusion-Previous work (Colombo et al., 1992a(Colombo et al., , 1994a(Colombo et al., , 1994b from our laboratory indicates that heterotrimeric G protein(s) regulate fusion among endosomes. G␣ has long been associated with signal transduction pathways. More recently, G␤␥ has emerged as a major participant in signal transduction via its interaction with several effectors within the cell (reviewed by Clapham and Neer, 1993;Sternweis, 1994). ␤ARK is a cytosolic enzyme that is targeted by G␤␥ to the membrane . A fragment of ␤ARK corresponding to 222 amino acids of the COOH-terminal domain contains the targeting domain for binding to G␤␥ subunits (Pitcher et al., 1992). A fusion protein corresponding to this targeting domain blocks the binding of ␤ARK to G␤␥  and other G␤␥-mediated processes.
To assess the possible involvement of G␤␥ in the mechanism of endosome fusion, the COOH-terminal ␤ARK fusion protein was tested in the in vitro endosome fusion assay. Fig. 1A shows that a 6His-COOH-terminal ␤ARK1 fusion protein (␤ARK1ct) completely blocks fusion between endosomes (closed circles). The inhibitory potency of ␤ARK1ct in the in vitro fusion assay (EC 50 10 -15 M) was similar to the inhibitory activity against G␤␥ activation of ␤ARK . Interestingly, the 6His-␤ARK2ct corresponding to the same region of ␤ARK2, another member of the G protein-coupled kinase family, was a better inhibitor of endosome fusion (triangles). In contrast, no effect was observed with the COOH-terminal domain of rhodopsin kinase (open circles). This result is consistent with earlier observations showing that the COOH-terminal domain of RK (RKct) does not bind to G␤␥. RKct lacks the G␤␥-binding domain and consequently does not interact with G␤␥ subunits (Pitcher et al., 1992). The differential effect observed with ␤ARKct and RKct fusion proteins appears to rule out any nonspecific effect of these polypeptides.
In order to identify the critical regions involved in ␤ARK binding to G␤␥, Koch and collaborators (1993) synthesized several peptides corresponding to the targeting domain. A single 28-amino acid peptide (Peptide G 1 ) derived from the targeting domain of ␤ARK1 was found to inhibit G␤␥ activation of ␤ARK with an IC 50 of 76 M. In contrast, peptide GЈ 1 , containing only the first 15 amino acid residues of peptide G 1 , was inactive. Fig. 1B shows that peptide G 1 was also inhibitory of endosome fusion with a similar EC 50 (closed circles). No inhibitory effect was observed with peptide GЈ 1 (open circles). As observed with ␤ARK2, peptide G 2 corresponding to the same region of ␤ARK 2 was a more potent inhibitor of endosome fusion (triangles).
Since the COOH-terminal domain of ␤ARK selectively binds to G␤␥, the inhibitory effect observed with the fusion protein and with peptide G suggests that a G␤␥-mediated process is involved in in vitro endosome fusion. The results further suggest that ␤ARKct and peptide G are likely blocking the interaction of G␤␥ subunits with a factor(s) required for in vitro endosome fusion.
The Small GTP-binding Protein ARF and the G␤␥-binding Domain of ␤ARK Share Sequence Similarity-ARF is a Raslike small GTP-binding protein that was originally identified as the protein cofactor required for efficient ADP-ribosylation of G␣ s by cholera toxin (Kahn and Gilman, 1984). It is now clear that ARF has an important role in vesicular transport (reviewed by Nuoffer and Balch, 1994;Rothman and Orci, 1992). Work in our laboratory indicates that ARF is required for in vitro endosome fusion and that in the presence of GTP␥S, ARF inhibits fusion (Lenhard et al., 1992). Recently, we have shown that ARF plays a regulatory role in receptor-mediated endocytosis (D' Souza-Schorey et al., 1995). Given that ARF is a cytosolic protein involved in fusion between endosomes we compared the sequence of peptides G with members of the ARF family. When peptides G were aligned with ARF a surprising similarity was found among the sequences (Fig. 2). A segment of five amino acids (ELRDA) from peptide G 1 was identical to a fragment corresponding to amino acids 115-119 of ARF1 (see box in Fig. 2). Equivalent sequence similarity was observed with other members of the ARF family.
␤ARKct and Peptide G Interferes with ARF Binding to Membranes-Previous work has shown that addition of GTP␥S inhibits several assays that reconstitute vesicular transport including transport through the Golgi and fusion between endosomes, in a cytosol-dependent fashion Mayorga et al., 1989). The sensitivity to GTP␥S of several cell-free assays is conferred in part by ARF, a cytosolic protein (Taylor et al., 1992). Since peptide G and ARF have sequences in common, we speculated that peptide G would compete with ARF function. If that were the case, addition of peptide G might be expected to compete both the GTP␥Sand ARF-dependent inhibition of fusion. As predicted the inhibitory effect of GTP␥S was reversed by addition of increasing concentrations of peptide G 1 (Fig. 3A). Similarly, the inhibitory effect of ARF was reversed by addition of peptide G 1 (Fig. 3B). Moreover, ␤ARKct also reversed the inhibitory effect of both GTP␥S and ARF (data not shown). The observation that both peptide G and the fusion protein containing the COOH-terminal domain of ␤ARK produce a similar effect rules out the possibility that the effects observed in our assay are due to detergent-like effects sometimes attributed to certain peptides.
As another approach to directly show that peptide G was competing with ARF for interaction with membranes, we studied the binding of ARF to endosomal membranes by Western blot assay. Fig. 3C shows that incubation of enriched endosomal membranes with cytosol in the presence of 20 M GTP␥S resulted in binding of ARF (lane a). Preincubation of the membranes for 5 min at 37°C before the addition of GTP␥S with peptide G 1 or G 2 (lanes b and c) inhibited the binding of ARF to crude endosomal membranes. As expected no inhibition of ARF binding was observed with the control peptide G 1 Ј (lane d).
Taken together our results indicate that peptide G interferes with ARF function by blocking the interaction of this protein with the membranes.
Binding of G␤␥ Subunits to ARF Fusion Proteins-Based on the sequence similarity between peptide G and ARF and, since peptide G blocks binding of ␤ARK to G␤␥ subunits , the results suggest that G␤␥ is one of the membrane components that may interact with ARF. In order to address this question, we performed an in vitro binding assay using a GST-ARF fusion protein to study the direct interaction between the proteins. We constructed GST-ARF4 and an aminoterminal deletion mutant GST-ARF4 (⌬1-17) with the first 17 amino acids deleted. In order to better define the domain that is involved in the interaction of ARF with G␤␥, a third fusion protein GST-ARF1ct, which contained the carboxyl-terminal half of ARF1, was constructed based on the sequence alignments between peptide G and ARF. Fig. 4 shows a diagrammatic representation of the GST-ARF fusion proteins used in the in vitro binding assay. As shown in Fig. 5A both ARF mutants, ARF4 (⌬1-17) and ARF1ct (lanes 4 and 5, respectively), bound G␤␥ although to a lesser extent than GST-␤ARK1ct (lane 1, positive control). Also, GST-ARF4 bound G␤␥ to an extent similar to the mutated forms of ARF (data not shown). The corresponding region of RK (GST-RKct), which does not bind G␤␥, and GST alone were used as negative controls (lanes 2 and 3, respectively). In order to assess the specificity of the interaction between G␤␥ and ARF, Rab5, another small GTPbinding protein involved in fusion among endosomes, was tested in the binding assay. GST-Rab5 was negative for binding to G␤␥ subunits (lane 6). Although these results indicate that the binding of G␤␥ to immobilized ARF is specific, only a small amount of the available G␤␥ subunits bound to GST-ARF. However, the binding of G␤␥ to GST-ARF4 (⌬1-17) was markedly increased by addition of recombinant G␣ subunits such as G␣ s and G␣ i3 (Fig. 5A, lanes 7 and 8), suggesting that ARF interacts more efficiently with the heterotrimer than with G␤␥ alone. This was an unexpected observation given previous results showing that G␣ completely inhibited the binding of G␤␥ to ␤ARK . Nevertheless, there is a precedent for a possible G␣-ARF association given the fact that ARF is the co-factor necessary for the ADP-ribosylation of G s by cholera toxin (Kahn andGilman, 1984, 1986). Therefore, G␣ may associate directly with ARF increasing the binding of G␤␥.
G␤␥ binding to ARF-GST was specifically competed by purified recombinant ARF1 and ARF4 (Fig. 5B), but not by BSA indicating the specificity of the ARF-G␤␥ association.
G␣ s in the GDP-bound Form Interacts with ARF-As mentioned above addition of G␣ subunits enhanced the binding of G␤␥ to ARF. In order to study the possibility that G␣ was also part of the complex, we added recombinant G␣ s to the assay in the presence of G␤␥ subunits. As shown in Fig. 6A, G␣ was detected using a specific antibody generated against the COOH-terminal domain of G␣ s . We next asked if G␣ s was able to associate with ARF in the absence of G␤␥ and if this interaction were specific. Fig. 6B shows that G␣ s binds to GST-ARF in the absence of G␤␥, and that a marked increase in binding is observed when both subunits were added to the assay. Essentially, no binding was observed when GST alone was used, indicating the specificity of the protein association.
It is known that GTPases function as molecular switches changing their conformation when they are activated. Aluminum fluoride (AlF) is a classical activator of heterotrimeric G proteins but does not activate members of the small GTPase family such as ARF . Therefore, in order to independently activate the heterotrimeric G protein, the effect of AlF was tested in the binding assay. As shown in Fig. 6C, activation of G␣ s by AlF completely blocked ARF-G␣ s association both in the presence or the absence of G␤␥. As expected, AlF did not affect ARF-G␤␥ association.
Given that both ARF and heterotrimeric G proteins are regulated by nucleotide binding we next studied the effect of either GTP␥S or GDP␤S. Similar to the effect observed with AlF, FIG. 2. Homology of peptides G 1 and G 2 to ARFs. Peptide G 1 , a 28-amino acid peptide corresponding to ␤ARK1 residues Trp 643 to Ser 670 and peptide G 2 corresponding to the same region of ␤ARK2 were aligned with the COOH-terminal half of members of the ARF family using the J. Hein method with PAM250 residue weight table. A segment of five amino acids (ELRDA) from peptide G 1 , identical to a fragment corresponding to amino acids 115-119 of ARF1, is boxed. Equivalent sequence similarity was observed with other members of the ARF family. Sequence positions for the rightmost residue of each polypeptide are given in the right-hand column.
GTP␥S almost completely blocked G␣ s -ARF association (data not shown); essentially no effect was observed with GDP␤S. Our results clearly indicate that G␣ s in the GDP-bound form associates with ARF either in the presence or the absence of G␤␥ subunits. Activation of G␣ s by either GTP␥S or AlF completely blocked ARF-G␣ interaction, indicating that the heterotrimer is the most likely candidate for ARF-G protein interaction. GTP␥S inhibited ARF-G␤␥ association (Fig. 6D) suggesting that ARF interacts with G␤␥ in the GDP-bound state.

DISCUSSION
The ␤␥ subunits of heterotrimeric G proteins modulate the activity of several signal-transducing effector molecules such as phospholipase C, phospholipase A2, certain isoforms of adenylate cyclase and cardiac muscarinic potassium channels (reviewd by Clapham and Neer, 1993). G␤␥ also mediates the membrane translocation of the ␤-adrenergic receptor kinases (␤ARK1 and ␤ARK2) where they phosphorylate activated receptors . The COOH-terminal domain of ␤ARK (␤ARKct) contains the targeting domain for binding to G␤␥ (Pitcher et al., 1992), and a fusion protein corresponding to this targeting domain blocks the binding of ␤ARK to G␤␥ (Koch FIG. 3. Peptide G 1 reverses GTP␥S-and ARF-mediated inhibition of fusion by inhibiting ARF binding to the membranes. A, endosome fusion was tested in the presence of 0.8 mg/ml cytosol supplemented with 20 M GTP␥S to inhibit fusion. The inhibitory effect of GTP␥S was reversed by addition of increasing concentrations of peptide G 1 . Endosome fusion was measured as described under "Experimental Procedures." Fusion is expressed in relative units. B, endosomal vesicles were resuspended in cytosol (0.2 mg/ml) containing 20 M GTP␥S. Fusion was assessed in the presence (closed circles) or the absence (open circles) of 15 g/ml purified recombinant myristolated ARF1. The inhibitory effect of ARF was reversed by addition of increasing concentrations of peptide G 1 . The results are representative data of a experiment performed three times. C, enriched endosomal fraction (10 -20 g of total protein) was resuspended in fusion buffer, containing 1 mg/ml cytosolic proteins. Samples were incubated for 5 min at 37°C in the presence of: lane a, no additions; lane b, 50 M peptide G 1 ; lane c, 25 M peptide G 2 ; lane d, 50 M peptide G 1Ј (control peptide). After preincubation, 20 M GTP␥S was added, and the samples were incubated for additional 20 min at 37°C. After incubation, the samples were washed with 1 ml of homogenization buffer containing 20 M GTP␥S and 1 mM MgCl 2 , and the membranes were recovered by centrifugation for 5 min at 50,000 ϫ g. The membrane proteins were resolved by SDS-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and probed with anti-ARF antibodies. Data represent one of three similar experiments.
In this report we present evidence that the COOH-terminal portion of ␤ARK (␤ARKct) and peptides corresponding to the G␤␥-targeting domain of ␤ARK inhibit in vitro endosome fusion. The results suggest that a G␤␥-mediated signal is involved in either the mechanism or the regulation of endosome fusion. Indeed, our results suggest that ␤ARKct and peptides from the G␤␥-binding domain (peptides G) block the interaction of G␤␥ with a factor(s) required for endosome fusion. We believe that one of these factors is ARF for the following reasons: (i) peptide G and ARF share sequence homology, (ii) peptide G reverses GTP␥Sand ARF-mediated inhibition of endosome fusion, (iii) peptide G inhibits ARF binding to membranes. Supporting evidence for a direct interaction between ARF and G␤␥ was provided by an in vitro binding assay using ARF-GST fusion proteins. Our study establishes that G␤␥ binds to ARF and that this interaction is specifically competed by purified recombinant ARF and enhanced by G␣.
While the binding of G␤␥ to immobilized ARF is specific, only small amounts of the available G␤␥ subunits bound to GST-ARF. However, the binding was increased by the addition of G␣. A trivial explanation is that most of the G␤␥ has been simply denatured during its preparation. Another possibility is that ARF 4 binds only to a specific subset of the G␤␥ combinations comprising the heterogeneous preparation isolated from bovine brain. An interesting possibility is that G␤␥ may require interaction with another protein to be in the right conformation for binding. The G␤␥-binding domain of ␤ARK shares homology with the novel pleckstrin homology domain (PH domain). This domain is found in a variety of signaling molecules such Ras-GAP, Ras-GRF, SOS, and others (Shaw, 1993;Musacchio et al., 1993). Recently, it has been shown that proteins with PH domains bind to G␤␥ in vitro . Proteinprotein interactions between proteins containing a PH domain and G␤␥ may play a significant role in cellular signaling. Although the presence of a PH domain has not been described for ARF, it is tempting to speculate that putative ARF accessory proteins such as ARF-GAP or ARF-GRF may indeed contain such a domain and that they may regulate ARF activity in conjunction with G␤␥. Current models for the interaction between ARF and target membranes propose that activation of ARF by a protease-and brefeldin A-sensitive membrane-bound nucleotide-exchange factor (Helms and Rothman, 1992;Donaldson et al., 1992b;Randazzo et al., 1993) results in association of ARF-GTP with the lipid bilayer. Our results indicating that ARF in the GDP form interacts with G␤␥ suggest that these proteins may form a multimeric complex that allows the interaction of ARF with its nucleotide exchange factor resulting in ARF activation.
The results presented in this report are the first direct evidence indicating that both G␣ s and G␤␥ associates directly with ARF. There is a precedent for this connection in that ARF is the co-factor necessary for the ADP-ribosylation of G s by cholera toxin and a possible interaction with G␣ s has been previously suggested (Kahn andGilman, 1984, 1986). Interest- FIG. 6. A, binding of G␣ s to GST-ARF. Purified bovine brain G␤␥ (300 nM) with or without purified recombinant G␣ s (700 nM) was incubated with GST-ARF4(⌬1-17) as described in Fig. 5. B, G␣ s interacts with ARF in the presence or absence of G␤␥. GST-ARF4(⌬1-17) or GST alone was incubated with purified bovine brain G␤␥ (300 nM), purified recombinant G␣ s (500 nM) or both as described. C, activation of G␣ s by AlF blocks G␣ s -ARF association. GST-ARF4(⌬1-17) was incubated with 300 nM of purified bovine G␤␥ subunits and/or with 700 nM recombinant G␣ s for 30 min at 30°C in PBS containing 0.01% lubrol and 10 mM MgCl 2 in the presence or the absence of AlF (100 M AlNH 4 (SO 4 ) 2 ϩ 10 mM KF). D, activation of ARF by GTP␥S inhibits G␤␥-ARF association. GST-ARF4(⌬1-17) was preincubated for 90 min at 37°C in 50 mM HEPES-K, pH 7.5, containing 0.01% lubrol, 1 mM DTT, and 10 mM MgCl 2 in the presence of 50 M GTP␥S, 50 M GDP␤S or no additions. Nucleotide exchange on ARF was stopped by cooling at 4°C. Subsequently, 300 nM of purified bovine G␤␥ subunits were added, and the samples were incubated for additional 30 min at 4°C. The binding of the proteins to glutathione-Sepharose and the detection were performed as described under "Experimental Procedures." Western blot analysis showing G␤␥ and/or G␣ s binding to GST fusion proteins. Data represent one of three similar experiments. ingly, during the purification of ARF from bovine brain, ARF eluted in two peaks, one coincidental with G s . Addition of AlF was necessary to obtain a single peak of ARF activity (Kahn andGilman, 1984, 1986). The more likely target for AlF is the GDP-form of G␣. In agreement with the results of Kahn and Gilman, our data indicate that G␣ s in the GDP-bound conformation associates with ARF since activation of G␣ s by either GTP␥S or AlF completely blocked ARF-G␣ interaction. Recently, Finazzi and collaborators (1994) have shown that AlF plus GTP stabilizes the active state of ARF by preventing the rapid hydrolysis of the GTP loaded onto ARF. These authors have postulated that an AlF-sensitive target may lead to a persistent activation of ARF by inhibiting an ARF GAP or by making the ARF-GTP either insensitive or inaccessible to ARF GAP. While the exact role and mechanism of action of G␣ remains to be defined, our results of complete inhibition of G␣-ARF association by AlF suggest the intriguing possibility that G␣ in the GDP-bound form may regulate ARF GTPase activity.
An interesting outcome of our experiments relates to the role of the amino-terminal domain of ARF in mediating ARF function. It has been reported that the amino terminus of ARF is critical for function since deletion of this domain results in a global reduction of ARF activities . A synthetic peptide derived from the amino terminus of ARF inhibits ARF activity including cholera toxin activation, as well as intra-Golgi transport  and fusion between endosomes (Lenhard et al., 1992). Moreover, the amino-terminal 13 residues of ARF1 are required for cofactor activity in the ADP-ribosylation by cholera toxin when G s is the substrate (Randazzo et al., 1994). However, Vaughan and collaborators (Hong et al., 1994) have shown that the amino terminus of ARF is not necessary for in vitro activation of cholera toxin using as a substrate agmatine. Although the basis for this disparity is not clear, this latest result suggests that other domains, besides the amino terminus, are likely involved in the interaction of ARF with the toxin. Our results indicate that the ARF domain involved in ARF-G␤␥ interaction does not require the amino-terminal 17 amino acids since G␤␥ binds to GST-ARF4 and to the truncated ARF mutants (ARF4⌬1-17 and ARF1ct) to a comparable extent. However, we cannot rule out the possibility that the presence of GST at the amino terminus may interfere with the proper folding and binding capacity of this domain.
It has been demonstrated that ARF plays an essential role in regulating coatomer binding (Donaldson et al., 1992a;Palmer et al., 1993) and AP-1 recruitment onto Golgi membranes (Traub, et al., 1993;Stamnes et al., 1993). Moreover, a number of studies have provided evidence for the involvement of heterotrimeric G proteins in coat assembly (Donaldson et al., 1991;Ktistakis et al., 1992). Association of ARF and ␤-COP with Golgi membranes is sensitive to a number of reagents that modulate heterotrimeric G protein function (Donaldson et al., 1991;Ktistakis et al., 1992). In addition to GTP␥S, AlF, known to specifically activate trimeric G proteins (Kahn, 1991), enhances the binding of ␤-COP to Golgi membranes (Serafini et al., 1991). These findings and the observation that G␤␥ inhibits both ARF and ␤-COP binding (Donaldson et al., 1991) suggest that G proteins regulate coat protein binding. We have also recently shown that both heterotrimeric G proteins and ARF regulate priming of endosomal membranes for fusion (Lenhard et al., 1994). Addition of G␤␥ resulted in inhibition of GTP␥Smediated priming of endosomes. In contrast, addition of ARF to the assay enhanced priming in the presence of cytosol. These observations suggest that ARF enhances binding of cytosolic factors required for fusion onto the endosomal membrane. Al-though the linkage between ARF binding and coat assembly with heterotrimeric G proteins has been proposed based on the data summarized above, to date no direct evidence for the interaction between ARF and trimeric G proteins has been presented. Our data would support a model in which heterotrimeric G proteins regulate binding of essential proteins at least in part, by directly interacting with ARF.
Finally, several recent observations implicate a signal transduction mechanism in the regulation of vesicular traffic. The findings from Bomsel and Mostov (1993) indicating that binding of dIgA to the pIgR stimulates the formation of transcytotic vesicles suggest that ligand binding generates a signal that is transduced to the intracellular sorting machinery. Interestingly, in Chinese hamster ovary cells transfected with muscarinic receptors, endosomal trafficking was inhibited by carbachol (Haraguchi and Rodbell, 1991). More specifically, antigeninduced activation of the IgE receptor and activation of protein kinase C regulate the GTP-dependent binding of ARF and ␤-COP to Golgi membranes (De Matteis et al., 1993). Furthermore, the recent identification of phospholipase D as an effector of ARF (Brown et al., 1993;Kahn et al., 1993) raises the possibility that a novel signal transduction pathway may regulate intracellular membrane traffic. Our results of a direct interaction between ARF and trimeric G proteins suggest that ARF may be a nexus linking heterotrimeric G proteins and downstream effectors (i.e. PLD). Given the enormous potential for specificity with 24 possible combinations of G␤␥ and several ARFs, our present observations, together with those of others, provide a novel prospect by which trimeric G proteins and ARF provide fine control of vesicular traffic and its response to extracellular signals.