Comparative activity of ADP-ribosylation factor family members in the early steps of coated vesicle formation on rat liver Golgi membranes.

We have compared the abilities of mammalian ADP-ribosylation factors (ARFs) 1, 5, and 6 and Saccharomyces cerevisiae ARF2 to serve as substrates for the rat liver Golgi membrane guanine nucleotide exchange factor and to initiate the formation of clathrin- and coatomer protein (COP) I-coated vesicles on these membranes. While Golgi membranes stimulated the exchange of GTPγS for GDP on all of the ARFs tested, mammalian ARF1 was the best substrate, with an apparent Km of 5 μM. In all cases myristoylation of ARF was required for stimulation. Agents that inhibit the Golgi membrane guanine nucleotide exchange factor (the fungal metabolite brefeldin A and trypsin treatment) selectively inhibited the guanine nucleotide exchange on mammalian ARF1. Taken together, these data indicate that of the ARFs tested, only mammalian ARF1 is activated efficiently by the Golgi guanine nucleotide exchange factor. The other ARFs are activated mainly by another mechanism, possibly phospholipid-mediated. Once activated, all of the membrane-associated, myristoylated ARFs promoted the recruitment of coatomer to about the same extent. Mammalian ARFs 1 and 5 were the most effective in promoting the recruitment of the AP-1 adaptor complex, whereas yeast ARF2 was the least active. These data indicate that the specificity for ARF action on the Golgi membranes is primarily determined by the Golgi guanine nucleotide exchange factor, which has a strong preference for myristoylated mammalian ARF1.

guanine nucleotide exchange factor (GEF) that catalyzes the exchange of GTP for GDP on ARF (6 -8). Once it has bound GTP or its slowly hydrolyzable analog, GTP␥S, ARF facilitates the recruitment of the heterotetrameric AP-1 adaptor complex and clathrin onto the Golgi membranes (2,3). COPI-coated vesicle formation also proceeds following the initial recruitment of ARF onto the Golgi (4,5).
The family of mammalian ARFs (mARFs) has been grouped into three classes based on amino acid sequence (9). Class I is composed of mARFs 1-3, class II of mARFs 4 and 5, and class III has a single member, mARF6. All of the ARFs contain a glycine at position 2 that is a site for N-terminal myristoylation (10).
The majority of in vitro studies of the role of ARF in coat formation on the Golgi have utilized recombinant mARF1 or a mixture of cytosolic ARFs purified from various tissues. Consequently, little is known about the ability of the other ARFs to interact with the Golgi GEF and to facilitate the recruitment of AP-1 and coatomer, the protomer of the COPI coat, onto Golgi membranes. We have tested the ability of one mammalian ARF from each class, mARFs 1, 5, and 6, and Saccharomyces cerevisiae ARF2 (yARF2) to perform a number of the known functions required for coat formation on the Golgi membranes using an in vitro assay system. Our results indicate that myristoylated mARF1 has the highest affinity for the Golgi GEF. The other ARFs are activated by the Golgi membranes to some extent, mainly by a GEF-independent mechanism. Once activated, all of the ARFs tested promote the binding of coatomer and AP-1 but with variable efficiencies. nitrocellulose was from Schleicher & Schuell; enhanced chemiluminescence reagents for chemiluminescence were purchase from Amersham Corp.; isopropyl-1-thio-␤-D-galactopyranoside was from Amresco (Solon, OH); myristic acid was from Nu Chek Prep (Elysian, MN). All other reagents were the highest grade available.
Antibodies-The polyclonal antibody AE/1 to ␥-adaptin was provided by Linton Traub of our laboratory (16) and was used at a concentration of 1:5000 for immunoblotting. The monoclonal antibody M3A5 against ␤-COP was a gift from Thomas Kreis (University of Geneva, Geneva, Switzerland) (17) and was used at a dilution of 1 g/ml. Horseradish peroxidase-conjugated antibodies against mouse and rabbit immunoglobins were purchased from Amersham Corp.
Production and Purification of Recombinant ARFs-For the production of the mammalian ARFs, competent BL21(DE3) Escherichia coli were transformed with an ARF expression vector. When myristoylated ARF was made, the BL21 cells were co-transformed with either the S. cerevisiae or human NMT expression vector. In a typical preparation, 1-3 liters of bacterial culture were grown at 37°C until A 600 ϭ 0.8 -1.2 and then isopropyl-1-thio-␤-D-galactopyranoside (1 mM) was added. When myristoylated ARFs were prepared, myristic acid (500 M) and Brij 58 (0.5%) were also added. The culture was then grown overnight at 27-30°C. The cells were collected by centrifugation at 5000 ϫ g for 15 min at 4°C. Myristoylated or unmyristoylated yARF2 was produced as described previously (18). The bacterial pellets were either stored at Ϫ80°C or used immediately.
The bacteria were lysed by repeated cycles of freezing and thawing (19) and resuspended in column buffer (50 mM Tris-HCl, 1 mM DTT, 1 mM magnesium acetate, 10 M GDP, 0.02% sodium azide) plus protease inhibitors (1 g/ml leupeptin, 2 g/ml antipain, 10 g/ml benzamidine, 1 g/ml Trasylol, 1 g/ml chymostatin, 1 g/ml pepstatin) at pH 8.0 for mARFs 1 and 5 and yARF2 and at pH 7.0 for mARF6. The lysate was cleared by centrifugation at 10,000 ϫ g for 15 min at 4°C. The supernatant was collected and loaded onto a 20-ml DEAE-Sepharose column (mARF1 and 5 and yARF2) or a 20-ml SP-Sepharose column (mARF6) equilibrated with column buffer. The column was eluted at 90 ml/h with a 30-ml gradient from 0 to 200 mM NaCl for mARFs 1 and 5 and yARF2 and from 0 to 500 mM NaCl for mARF6. Aliquots of the column fractions were subjected to 15% SDS-PAGE followed by Coomassie Blue staining. The ARFs were detected as prominent 20-kDa bands, as confirmed by ligand blotting with [␣-32 P]GTP (data not shown). All of the ARFs except mARF6 were recovered primarily in the flow-through fractions; mARF6 eluted at approximately 400 mM NaCl. ARF-containing fractions were pooled, concentrated to about 3 ml using a Centriprep-10 (Amicon, Beverly, MA), and loaded onto a Superdex-75 column (1.5 ϫ 55 cm) equilibrated in column buffer, pH 7.5, adjusted to 10% sucrose. Column fractions were monitored for ARF content as described previously. ARF-containing fractions were pooled and concentrated to Ͼ2 mg of protein/ml. The sample was separated into small aliquots, frozen on dry ice/methanol, and stored at Ϫ80°C. Protein concentration was determined with the Bio-Rad (Bradford) Protein Assay, using standard I (Bio-Rad). The percentage of ARF in the final product was determined by densitometry scanning of a Coomassie Blue-stained gel using a Personal Densitometer (Molecular Dynamics Inc., Sunnyvale, CA) with Image-Quant software.
[ 3 H]Myristate Labeling of Bacteria-A 2-ml culture of bacteria expressing both ARF and NMT was grown as described above except that 25 l/ml [ 3 H]myristic acid (14 mCi/ml) was added instead of the unlabeled myristic acid. The bacteria were lysed as described above, resuspended in 20 -40-l column buffer at pH 7.5, and the lysate was cleared by centrifugation at 13,700 ϫ g for 5 min at 4°C in a microcentrifuge. The supernatant fraction was used without further purification.
Determination of ARF Myristoylation-2.5-5-g aliquots of the various ARFs were loaded alongside 5 g of [ 3 H]myristate-labeled lysate on a 13% (mARFs 1, 5, and yARF2) or a 15% (mARF6) SDS-polyacrylamide gel, measuring 25 cm long. The gels were electrophoresed until the 17-kDa See Blue prestained marker (Novex, San Diego, CA) was near the bottom of the gel. The gel was stained with Coomassie Blue, incubated in Amplify (Amersham Corp.), dried, and analyzed by autoradiography. The bands corresponding to myristoylated protein were identified by their co-migration with [ 3 H]myristate-labeled ARF and by their absence in unmyristoylated ARF preparations. The ratio of myristoylated ARF to total ARF in each preparation was determined by densitometry scanning of the Coomassie Blue-stained gel.
Preparation of Golgi-enriched Membranes and Cytosolic Fractions-Rat liver Golgi membranes and an AP-1/COPI-enriched fraction of rat liver cytosol were prepared as described previously (3), except that the latter was concentrated approximately 10-fold using a centriprep-10, frozen on dry ice in small aliquots, and stored at Ϫ80°C. The protein concentration of these fractions was determined with the Bio-Rad protein assay, using standard II. The final coat protein-enriched fraction had a concentration of AP-1 that was about 3 times that in cytosol, as determined by quantitative immunoblotting. There was no ARF detectable in this fraction as analyzed by ligand blotting with [␣-32 P]GTP.
Guanine Nucleotide Exchange Assay-A guanine nucleotide exchange assay was developed based on the assays of Northrup et al. (20) and Kahn and Gilman (21). Except where indicated, recombinant ARF (2.5 M) was added with or without Golgi membranes (50 g/ml) to a 100-l reaction mixture containing 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM DTT, 1 mM ATP, 1 mg/ml BSA, and 1 M [ 35 S]GTP␥S (3 ϫ 10 5 -1 ϫ 10 6 cpm). The reaction mixture was incubated for 10 min at 37°C, and then 1 ml of 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM DTT, at 4°C, was added to stop the reaction. The sample was then passed over a nitrocellulose filter in a sampling manifold (Model 3025 from Millipore, Bedford, MA), and the filter was washed 10 ϫ 1 ml with the buffer used to stop the reaction. Each experimental point was assayed in triplicate. The radioactivity bound to the filter was measured, and the amount of GTP␥S bound to protein was calculated. Background binding of [ 35 S]GTP␥S to the nitrocellulose, determined either by adding 5 mM GTP to a standard reaction mixture containing both ARF and Golgi membranes or by filtering a sample containing only [ 35 S]GTP␥S and buffer, was less than 0.1% of the radioactivity added.
BFA, when included, was added to 200 g/ml from a 10 mg/ml stock solution made fresh in ethanol. In these experiments, an equal volume of ethanol was added to the control reaction mixtures.
Protease-treated Golgi membranes were prepared as follows. Aliquots of Golgi membranes (2 mg/ml) in 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM DTT, 1 mM ATP, 1 mg/ml BSA were treated in one of three ways. One aliquot was incubated with trypsin (50 g/ml), another with trypsin inhibitor (2.5 mg/ml), and the third with both together for 15 min at room temperature. Trypsin inhibitor (2.5 mg/ml) was then added to the reaction containing trypsin alone, and the Golgi membranes were incubated for an additional 10 min at room temperature. The Golgi membranes were then used directly in the assay.
For the assays to determine the concentration dependence of guanine nucleotide exchange on mARFs 1 and 6 and yARF2, the Multiscreen Filtration System Vacuum Manifold (Millipore) was used. In these assays, the BSA concentration was reduced to 100 g/ml, and the [ 35 S]GTP␥S concentration was increased to 10 M. The reactions were stopped by immediately filtering 50 l of a 100-l reaction mixture onto a HA-High Protein and Nucleic Acid Binding Plate (Millipore). The filters were washed 7 ϫ 200 l with 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM DTT, at 4°C.
ARF and Coat Protein Recruitment Assay-The recruitment assays were performed in a total volume of 400 l in 1.5-ml presiliconized tubes to reduce background. The AP-1/COPI-enriched cytosolic fraction was precleared by centrifugation at 220,000 ϫ g for 20 min at 4°C before use in the assay. Golgi membranes (50 g/ml), recombinant myristoylated ARF to the concentration indicated in the figure legends, the AP-1/ COPI-enriched cytosolic fraction (5.6 mg/ml), and either GTP␥S (100 M) or GDP (1 mM) were mixed together on ice in assay buffer (25 mM Hepes, pH 7.0, 250 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM DTT, 1 mg/ml BSA). The reaction mixtures were incubated for 15 min at 37°C, and the assay was stopped by returning the tubes to ice. The reaction mixtures were then transferred to fresh 1.5-ml presiliconized tubes to further reduce the background. The Golgi membranes were reisolated as described previously (3) except that the final Golgi pellet was washed two times with assay buffer. The membranes were then solubilized in SDS-PAGE sample buffer (2.3% SDS, 62.5 mM Tris-HCl, pH 6.8, 5% ␤-mercaptoethanol, 10% sucrose), and analyzed by 13% SDS-PAGE followed by immunoblotting or Coomassie Blue staining. ARF binding was detected by Coomassie Blue staining; AP-1 binding was detected by immunoblotting using AE/1, a polyclonal antibody to ␥-adaptin, and coatomer binding was detected by immunoblotting using M3A5, a monoclonal antibody to ␤-COP. The immunoblots or Coomassie Blue-stained gels were quantified by densitometry scanning.
Electrophoresis and Immunoblotting-Electrophoresis and immunoblotting were performed as described previously (3). Gels to be analyzed by fluorography were incubated in gel dry (1% glycerol, 25% methanol) for at least 1 h, and then incubated in Amplify for 30 min. The gels were dried onto paper and exposed to X-Omat AR film (Eastman Kodak) using one or two intensifying screens.

Preparation of Recombinant Myristoylated
ARFs-Since it is difficult to fractionate tissue-purified ARF into individual fam-ily members, we used a bacterial expression system to produce homogeneous populations of individual, recombinant ARFs (11,22). Myristoylated ARFs were prepared by co-expressing the various ARFs with NMT, the enzyme that catalyzes the addition of myristate onto proteins (23). However, since the efficiency of myristoylation of the different ARFs varied considerably and one goal of this study was to determine the role of myristoylation in several functions mediated by the ARFs, it was important to determine the extent of myristoylation of each recombinant ARF preparation. ARF preparations can contain three types of molecules as follows: ARF that still contains its N-terminal methionine, which blocks myristoylation; unmyristoylated ARF that has its N-terminal methionine cleaved; and ARF that is myristoylated on its N-terminal glycine. For mARF1, these three forms can be clearly resolved by SDS-PAGE, as shown in Fig. 1, lanes 2 and 3 (24). In all cases, ARF made in bacteria co-expressing NMT had a faster migrating ARF band that was absent in unmyristoylated ARF preparations (Fig. 1, compare lanes 1 and 2, 5 and 6, 9 and 10, and 13 and 14) (18,24,25). This faster migrating band co-migrated with [ 3 H]myristate-labeled ARF (Fig. 1, lanes 2-4, 6 -8, 10 -12, and 14 -16). The ratio of myristoylated ARF to total ARF in each preparation could be determined by quantitating the Coomassie Blue-stained bands using densitometry scanning. The various mARF1 preparations were myristoylated between 10 and 50%, mARF6 preparations between 30 and 50%, and mARF5 and yARF2 preparations were completely myristoylated.
Effect of Myristoylation of ARF on Golgi Membrane-stimulated Guanine Nucleotide Exchange-We first tested the ability of rat liver Golgi membranes to stimulate guanine nucleotide exchange on myristoylated and unmyristoylated mARFs 1, 5, and 6, and yARF2. As summarized in Table I, each ARF species exhibited some spontaneous guanine nucleotide exchange. The Golgi membranes also bound a small amount of GTP␥S in the absence of added ARF. When these background values were taken into account, it was apparent that incubation of the unmyristoylated ARFs with Golgi membranes did not result in an increase in GTP␥S binding. By contrast, each myristoylated ARF species gave rise to a stimulation in GTP␥S binding over the background values. This increase has been shown to be due to increased GTP␥S binding to ARF, not to an ARF-induced increase in GTP␥S binding to another Golgi protein (6,8). However, the magnitude of this stimulation varied considerably among the various ARFs, with the exchange on mARF1 being by far the greatest. The effect of ARF concentration on the extent of stimulation was determined for the myristoylated forms of mARF1, mARF6, and yARF2 (Fig. 2). The apparent K m values for activation by the Golgi membranes were 5 M for mARF1 and 12.5 M for mARF6. An accurate K m value for the activation of yARF2 could not be determined, but it was greater than 20 M. The apparent V max values for myristoylated mARF1 and mARF6 were 400 and 250 pmol of GTP␥S bound/ min/mg Golgi membranes, respectively. Thus the catalytic efficiency (V max /K m ) was 4 times greater for myristoylated mARF1 than for myristoylated mARF6. The apparent V max for yARF2 could not be calculated.
Since the preparation of myristoylated mARF1 used in most of the experiments contained a significant fraction of the unmyristoylated species (80%), we tested whether the unmyristoylated mARF1 was an inhibitor of the Golgi membrane-stimulated guanine nucleotide exchange on the myristoylated protein. To do this, a large excess of unmyristoylated mARF1 was added to an assay containing myristoylated mARF1 and Golgi membranes. As shown in Table II, unmyristoylated mARF1 did not inhibit the Golgi membrane-stimulated guanine nucleotide exchange on myristoylated mARF1.

Effect of Brefeldin A and Trypsin on Golgi Membrane-stimulated Guanine Nucleotide Exchange on Myristoylated ARFs-
The effect of BFA on the Golgi membrane-stimulated guanine nucleotide exchange on the various myristoylated ARFs was tested next. This fungal metabolite inhibits the activity of the Golgi GEF and thereby provides a means to determine the role of the Golgi GEF in these exchange reactions (6 -8). The results of a typical experiment are shown in Table III. BFA inhibited the Golgi-stimulated guanine nucleotide exchange on myristoylated mARF1 by 72% while inhibiting the exchange on myristoylated mARFs 5 and 6 by only 20%. The exchange on myristoylated yARF2 was not inhibited. Similar results were obtained when Golgi membranes were treated with trypsin (Table IV). This protease has been shown to inactivate the Golgi GEF (7,8). In this representative experiment, trypsin treatment of the Golgi membranes inhibited guanine nucleotide exchange on myristoylated mARF1 by 92%, whereas the exchange on myristoylated mARFs 5 and 6 was only inhibited 19 and 30%, respectively. The exchange on myristoylated   Tables I, III, and IV are summarized graphically in Fig. 3, which presents the extent of Golgi-stimulated guanine nucleotide exchange activity for each ARF, expressed as a percent of the activity of myristoylated mARF1. The figure shows that myristoylated mARF5, mARF6, and yARF2 differ both quantitatively and qualitatively from myristoylated mARF1 in undergoing much less guanine nucleotide exchange which is almost completely insensitive to BFA or trypsin treatment of the Golgi membranes. The sensitivity to BFA and protease treatment displayed by myristoylated mARF1 is a hallmark of Golgi GEF activity.
Binding of ARFs to Golgi Membranes-Each of the myristoylated ARFs was tested for its ability to bind to Golgi membranes and to promote coat protein recruitment using an in vitro recruitment assay (3). In this assay, the recombinant myristoylated ARF is incubated with Golgi membranes, a rat liver cytosolic fraction enriched in AP-1 and COPI but depleted of ARF, and either GTP␥S or GDP. After incubation at 37°C for 15 min, the Golgi membranes are reisolated and analyzed for ARF binding and coat protein recruitment.
The binding of the various ARFs to the Golgi membranes as a function of myristoylated ARF concentration is shown in Fig.  4. A shows the Coomassie Blue staining pattern of Golgi membranes incubated with increasing concentrations of myristoylated ARF in the presence of GTP␥S. In all cases there was increased staining of a band at 20 kDa as the concentration of myristoylated ARF in the reaction increased. The identity of this band as ARF was confirmed by demonstrating that it bound [␣-32 P]GTP in a ligand blotting assay (data not shown). Very little ARF was found on the Golgi membranes when GDP was present instead of GTP␥S. When recombinant ARF was omitted, the Coomassie Blue-stained 20-kDa bands were not observed.
When the Coomassie Blue-stained bands migrating at 20 kDa were quantitated by densitometry scanning, it was possible to plot the amount of ARF bound as a function of myristoy-  (Fig. 4B). It is apparent that myristoylated mARF1 bound with the highest efficiency, myristoylated mARF6 and yARF2 bound significantly less well, and myristoylated mARF5 bound poorly.

Effect of ARFs on Coat Protein Recruitment onto Golgi
Membranes-After mARF1 binds to Golgi membranes, it promotes the recruitment of AP-1 and coatomer onto these membranes (2)(3)(4)(5). To evaluate the ability of the other ARFs to mediate these processes, the Golgi membranes used in the ARF binding experiments were assayed for their content of AP-1 and coatomer by immunoblotting. AP-1 recruitment was detected using an antibody to the ␥-adaptin subunit, whereas coatomer recruitment was detected with an antibody to the ␤-COP subunit. The resulting bands were quantitated by densitometry scanning. These values were compared with those obtained with myristoylated mARF1 assayed in the same experiment. The mARF1 curves were also used to normalize the values obtained in several independent experiments. When the data were plotted as the amount of ␥-adaptin or ␤-COP recruited versus the concentration of myristoylated ARF present in the reaction, it was apparent that mARF1 was much more effective than the other ARFs in promoting the binding of the coat proteins to the Golgi membranes (Figs. 5 and 6, A-C). These data reflect both the efficiency of the Golgi GEF in activating the various ARFs as well as the ability of the bound ARFs to facilitate coat protein recruitment. By plotting the data as coat protein recruited as a function of the amount of myristoylated ARF actually bound to the Golgi membranes in the same assay, it was possible to analyze the steps that occur after the ARFs bind to the Golgi membranes. When considered in this way, it was apparent that mARF5, once bound to the membranes, was somewhat more active than mARF1 in promoting ␥-adaptin recruitment, whereas mARF6 was slightly less active and yARF2 was much less active (Fig. 5, D-F). In contrast, mARF5 and yARF2 promoted ␤-COP recruitment to about the same extent as mARF1, and mARF6 appeared to be slightly less active (Fig. 6, D-F). It is also evident that ␤-COP binding to the Golgi membranes reaches saturation at a lower level of bound mARF1 than that required for maximal ␥-adaptin binding. This has been observed previously by Stamnes and Rothman (2). DISCUSSION During the formation of coated vesicles on Golgi membranes, ARF is activated by an as yet unidentified, membrane-associ-ated GEF which catalyzes the exchange of GTP for GDP. The activated, membrane-bound ARF in turn facilitates the recruitment of the AP-1 adaptor complex and the COPI coat complex onto the Golgi membranes. By comparing the ability of the Golgi membrane GEF to activate mARFs 1, 5, and 6, we have determined that it has specificity for mARF1. Several lines of evidence lead to this conclusion. First, although the Golgi membranes were able to stimulate the activation of all of the ARFs tested, the stimulation of mARF1 was much greater. Consistent with this, the apparent K m for activation of mARF1 by the Golgi membranes was much lower than for mARF6 or yARF2. In addition, we found that the Golgi membrane-dependent activation of mARF1, but not that of the other ARFs tested, was greatly inhibited by the addition of BFA, a compound known to inhibit the Golgi guanine nucleotide exchange activity in vitro  Tables I, III, and IV are graphically summarized for each myristoylated (myr) ARF as the percent of the value for myristoylated mARF1.
FIG. 4. ARF binding to Golgi membranes. A, Golgi membranes (50 g/ml), myristoylated (myr) ARF to the indicated concentration (after adjustment for the percentage of myristoylated protein in the preparation), and GTP␥S (100 M) or GDP (1 mM) were incubated together at 37°C for 15 min. In each experiment, the mARF5, mARF6, or yARF2 concentration curve was done in parallel to a mARF1 concentration curve. The Golgi membranes were reisolated, and one-half of the membranes were analyzed by SDS-PAGE (13%) and Coomassie Blue staining. B, the Coomassie Blue-stained bands from the experiments shown in A and one additional experiment for mARFs 5 and 6 and yARF2 were quantitated by densitometry scanning. The values obtained in the various experiments were normalized to each other using the ARF1 curve done in parallel. The amount of ARF bound was plotted versus the amount of myristoylated ARF added to the assay. (6 -8), and the binding of ARF, ␤-COP, and ␥-adaptin to the Golgi membranes in vivo (26 -29). Furthermore, trypsin treatment of the Golgi membranes, which is also known to destroy the activity of the Golgi GEF (7,8), markedly inhibited the activation of mARF1 but had only a small effect on the activation of mARFs 5 and 6 and no effect on yARF2 activation. Thus, for the following reasons we believe mARF1 (or class I ARFs) to be the most likely physiological mediator of coated vesicle formation on the Golgi membranes: 1) the Golgi membrane GEF has specificity for mARF1, or potentially for class I ARFs; 2) the cytosolic concentration of mARFs 1 and 3 is high relative to the other mammalian ARFs (30); and 3) mARF1 has been localized to Golgi membranes in vivo (15).
Because trypsin treatment almost completely destroyed the activity of the BFA-sensitive Golgi GEF, we think it unlikely that a protein on the Golgi membranes is responsible for the activation of the BFA-insensitive ARFs. Instead, we hypothesize that the Golgi membrane phospholipids are responsible. Phospholipids have been shown to affect the nucleotide state of ARF1 in vitro, stabilizing the GTP␥S-bound form of the myristoylated protein and slightly destabilizing the GDP-bound form (24,31). Consistent with this hypothesis, activation of yARF2 (which was completely insensitive to inhibition by BFA and trypsin) was unsaturable over the range of ARF concentrations tested, as might be expected for a nonenzymatic, phospholipid-mediated process.
Comparison of the ability of the Golgi GEF to activate myristoylated and unmyristoylated mARF1 showed that myristoylation is absolutely required for mARF1 activation. Addition of an excess of unmyristoylated mARF1 did not inhibit activation of myristoylated mARF1, suggesting that myristoylation is required for its recognition by the GEF. There are several possi-ble roles for the myristoyl group in this process. The myristoyl group may interact directly with the Golgi GEF, facilitating the protein-protein interaction. Alternatively, the myristoyl group may be required to maintain a conformation of ARF competent to interact with the Golgi GEF. Finally, recent work has suggested that ARF may need to associate with a membrane before it interacts with a GEF (32) and that the myristoyl group gives the GDP-bound form of the protein the required affinity for phospholipids (24).
These findings are consistent with the results obtained in overexpression studies in vivo (33). When ARF1(T31N), a mutant of mARF1 deficient in GTP binding, was overexpressed, it had a BFA-like effect on the cells. It has been hypothesized that this protein acts by binding irreversibly to the Golgi GEF, preventing activation of the wild type protein. However, overexpression of ARF1 (G2A,T31N), a double mutant also lacking the site of myristoylation, had no effect on the cells, suggesting that it was unable to block the active site of the Golgi GEF. It would be interesting to test whether these mARF1 mutants can act as inhibitors of the Golgi GEF in vitro.
In contrast to our results, others have found that unmyristoylated mARF1 can be activated by Golgi membranes in vitro, but its activation was much less efficient than that of the myristoylated protein (34). In addition, it has been reported that unmyristoylated mARF3 can be activated by a trypsinsensitive factor on the Golgi membranes (8). The reason for these discrepancies is unclear but may be due to differences in assay conditions. While our results demonstrate that the Golgi GEF is specific for mARF1, we also found that all of the ARFs tested could bind to the Golgi membranes in vitro, as had been shown earlier for mARFs 1, 3, and 5 (35,36). Therefore, the Golgi membranes themselves have no specificity for a single ARF. In addition, all FIG. 5. AP-1 recruitment onto Golgi membranes containing increasing amounts of bound myristoylated (myr) ARFs. Onehalf of the Golgi membranes from the experiments described in Fig. 4 were analyzed by SDS-PAGE and immunoblotting with AE/1, a polyclonal antibody to ␥-adaptin. The immunoblots were quantitated by densitometry scanning. The values for ␥-adaptin binding from two experiments containing a mARF5 (E), mARF6 (Ⅺ), or yARF2 (å) concentration curve were normalized using the mARF1 (q) concentration curves done in parallel. The normalized signals were then plotted versus the amount of ARF bound (determined as described in Fig. 4).

FIG. 6. Coatomer recruitment onto Golgi membranes containing increasing amounts of bound myristoylated (myr) ARFs.
The immunoblots from Fig. 5 were stripped and immunoblotted with M3A5, a monoclonal antibody to ␤-COP. The immunoblots were quantitated by densitometry scanning. The values for ␤-COP binding from two experiments containing a mARF5 (E), mARF6 (Ⅺ), or yARF2 (å) concentration curve were normalized using the mARF1 (q) concentration curves done in parallel. The normalized signals were then plotted versus the amount of ARF bound (determined as described in Fig. 4). of the mARFs tested could promote binding of AP-1 and coatomer, although to different extents. In particular, mARF6 promoted recruitment of the coat proteins less well than mARF1 and mARF5. Yeast ARF2 promoted the recruitment of AP-1 very poorly although it mediated the recruitment of coatomer reasonably well. These results suggest that there may be some specificity for activating the downstream events that result in coat recruitment.
Several models have been proposed for how ARF promotes coated vesicle formation on the Golgi membranes. In the simplest, ARF acts directly as a receptor for the vesicle coat protein (5). However, this model does not explain how coatomer is specifically recruited onto the membranes of the Golgi stack whereas AP-1 recruitment is confined to the trans-Golgi network (TGN). To address the localization of AP-1 recruitment, it has been proposed that ARF activates a putative docking protein present in the TGN (3). Upon activation, the docking protein would undergo a conformational change resulting in high affinity binding of AP-1. A variant of this model is that ARF activates a receptor present on the TGN, such as the mannose-6-P/IGF II receptor, to allow it to interact with AP-1 with greater affinity (37). It should be noted that in the latter two models, the effects of ARF on the target proteins could be direct or indirect.
In this regard, ARF has been identified as an activator of membrane-associated phospholipase D (PLD), including phospholipase D localized to the Golgi membranes (38 -44). Furthermore, the Golgi membranes of PtK1 and Madin-Darby canine kidney cells, which have high levels of endogenous PLD activity, form COPI-coated vesicles in the absence of added ARF and GTP (45) and bind ␤-COP in a BFA-insensitive manner (27,46,47). To explain this, Ktistakis and co-workers (45) have proposed that PLD acts as a downstream effector of ARF in COPI-coated vesicle formation and that the activated PLD in the Golgi membranes of the PtK1 and Madin-Darby canine kidney cells allows the requirement for exogenous ARF activation to be bypassed. They further suggest that the activated PLD degrades phosphatidylcholine in the Golgi membranes to phosphatidic acid which in turn brings about the recruitment of coatomer either directly or indirectly. Interestingly, ␥-adaptin binding to the Golgi membranes of these cells remains sensitive to BFA, consistent with ARF acting via two pathways to recruit these two types of coat proteins (27,47). The finding that maximal recruitment of coatomer and AP-1 in our assays is achieved at different levels of mARF1 binding (Figs. 5 and 6) is consistent with there being two activation pathways. Clearly, additional studies will be necessary to sort out these fundamental issues concerning the role of ARF in coat protein recruitment.