INTRODUCTION

The small GTP-binding protein ADP-ribosylation factor (ARF)¹ plays an essential role in the formation of clathrin-coated vesicles on the trans-Golgi network (TGN) and coatomer (COPI)-coated vesicles on the Golgi stacks (1). ARF is activated by a brefeldin A (BFA)-sensitive guanine nucleotide exchange factor (GEF) on the Golgi membranes that catalyzes the exchange of GDP for GTP on ARF (2-6). Once activated, ARF binds to the Golgi membranes and promotes binding of adaptor protein 1 (AP-1), a component of the clathrin coat, and coatomer, the protomer of the COPI coat (7-10). Furthermore, ARF activates phospholipase D (PLD) (11, 12), and recent evidence suggests that PLD may be an effector of ARF in COPI-coated vesicle formation (13) and promote vesicle transport from the ER to the Golgi (14) and budding of secretory vesicles from the TGN (15). The PLD isoform responsible for this is presumably PLD1, which localizes to the Golgi, ER, and early endosomes as opposed to PLD2, which localizes to the plasma membrane (16).

The amino acid sequences of ARFs are highly conserved between species. For example, Saccharomyces cerevisiae ARFs (yARFs) 1 and 2 are more than 70% identical to mammalian ARF1 (mARF1) (17, 18). Mammalian and yeast ARFs also exhibit some degree of functional conservation. Thus while the simultaneous disruption of the yARF1 and yARF2 genes is lethal (18), the double mutant can be partially rescued by expression of mammalian ARFs (19, 20). However, yARF2, in contrast to mARF1, is not activated by the BFA-sensitive rat Golgi GEF, although it is activated to a lesser extent by a BFA-resistant, GEF-independent mechanism probably mediated by Golgi membrane phospholipids. Once activated and bound to the Golgi membrane, yARF2 facilitates AP-1 binding much less efficiently than activated mARF1, although it is equivalent to mARF1 in initiating coatomer recruitment onto the membrane (21).

To begin to understand how mARF1 interacts with such a wide variety of target proteins, we constructed chimeras between mARF1 and yARF2 and tested their ability to interact with the Golgi GEF, to promote AP-1 recruitment onto Golgi membranes, and to activate purified human PLD1. Since yARF2 is deficient in each of these activities, we reasoned that the use of these chimeras should allow us to localize domains of mARF1 needed for these functions. Our results indicate that different domains of mARF1 are required for these various activities.

EXPERIMENTAL PROCEDURES

Brefeldin A (BFA), protease inhibitors, trypsin, dithiothreitol, antibiotics, 20 cetyl ether (Brij 58), ATP, and GDP were obtained from Sigma; GTPS from Boehringer Mannheim; [³⁵S]GTPS from NEN Life Science Products; [³H]myristic acid from ICN Biomedicals Inc. (Irvine, CA); Superdex-75 prep grade, DEAE-Sepharose Fast Flow, and SDS-PAGE low molecular weight markers from Pharmacia Biotech Inc.; nitrocellulose from Schleicher and Schuell; ECL reagents for chemiluminescence from Amersham Corp.; isopropyl-1-thio--D-galactopyranoside from Amresco (Solon, OH); myristic acid from Nu Chek Prep, Inc., (Elysian, MN); restriction enzymes from New England Biolabs (Beverly, MA); phospholipids from Avanti polar lipids; and L--dipalmitoylphosphatidylcholine [choline-methyl-³H]phosphatidylcholine from American Radiolabeled Chemicals (St. Louis MO). Phosphatidylinositol 4,5-bisphosphate was isolated as described (22). All other reagents were obtained from standard sources.

The S. cerevisiae myristoyl-CoA:protein N-myristoyltransferase expression vector pBB131 (23) and the S. cerevisiae ARF2 expression vector pRecA-ARF2 (24) were provided by Jeffrey Gordon (Washington University). The bovine ARF1 expression vector pOW12 (25) was a gift from Richard Kahn (Emory University, Atlanta), and the human ARF6 expression vector was supplied by Richard Klausner (National Institutes of Health).

pET-ARF2 was created by ligating the 538-base pair NdeI and BamHI fragment of pRecA-ARF2 containing the yeast ARF2 coding sequence into pET3c (Stratagene).

Mutations were created in the mARF1 and yARF2 coding sequences using the sequential PCR procedure (26). The mutated bases are in bold face. For all PCR reactions, Pfu polymerase (Stratagene) was used according to the manufacturer's directions. PCR reaction mixtures were subjected to 30-35 cycles of 1 min at 94 °C, 2 min at 55 °C, and 1 min at 72 °C and then incubated at 72 °C for 10 min to extend incomplete products. All sequences changed by PCR were determined by sequence analysis using the DNA Sequencing Kit, Dye Terminator Cycle Sequencing Ready Reaction (Perkin-Elmer). PCR products and restriction fragments were purified by excising the appropriate bands from an agarose gel, and recovering them with the QIAquick Gel Extraction Kit (Qiagen, Santa Clarita, CA). All ligations were performed using the DNA Ligation Kit (Takara Shuzo Co., Ltd., Otsu, Japan).

pET-mARF1* was created by site-directed mutagenesis of pOW12 using the mutagenic oligonucleotides 5-CTTTACGCTAGCAAGCTCTTCAAGGGCCTTTTTGG-3 and 5-CTTGCTAGCGTAAAGCCCCATATGTATATCTCCTT-3. The final PCR product was digested with BglII and BsrG1, and the 194-base pair fragment containing the mutated bases was used to replace the corresponding region of pOW12.

pET-mARF1*(K10S,G11N), pET-mARF1*(K15N), and pETmARF1*(K15A) were created by site-directed mutagenesis of pET-mARF1*. For pET-mARF1*(K10S,G11N), the mutagenic oligonucleotides were 5-CTCTTCTCGAATCTTTTTGGC-3 and 5-GCCAAAAAGATTCGAGAAGAG-3; for pET-mARF1*(K15N), they were 5-GCCTTTTTGGGGACAAAGAAATG-3 and 5-CATTTCTTTGTCCCCAAAAAGGC-3; for pET-mARF1*(K15A), they were 5-CCTTTTCGGCGCCAAGGAAATG-3 and 5CATTTCCTTGGCGCCGAAAAGG-3. The BglII and BsrGI fragments of the final PCR products were used to replace the corresponding region of pET-ARF1*.

pET-yARF2(N15K) was made by site-directed mutagenesis of pET-yARF2 using the oligonucleotides 5-TTTGGGAAGAAAGAAATGCG-3 and 5-TCTTTCTTCCCAAAAAGATTGCTG-3. The SphI and BsrGI fragment of the final PCR fragment was used to replace the corresponding region of pET-yARF2.

A BstB1 restriction site was engineered into pET-mARF1* and pET-yARF2 at base pair 280 (codon 94) of the respective coding sequences. For pET-mARF1*, the mutagenic oligonucleotides were 5-GGTTGATTCGAATGACAGAGAG-3 and 5-CTGTCATTCGAATCAACCACAAAG-3, and the BsrGI and HindIII fragment of the final PCR product was used to replace the corresponding region of pET-mARF1*. pET-yARF2 was mutagenized using 5-CGATTCGAATGATAGATCGCGT-3 and 5-CGATCTATCATTCGAATCGATGAC-3. The final PCR product was digested with BsrGI and BlpI, and the fragment was used to replace the corresponding region of pET-yARF2. All of the following constructs were made using these modified plasmids. The designation M is used when a region is derived from the mammalian ARF1 and Y is used when a region is derived from yeast ARF.

pET-M*YY was constructed by transferring the SphI and BsrG1 fragment of pET-mARF1* to pET-yARF2. pET-YMM was constructed by transferring the SphI and BsrGI fragment of pET-yARF2 to pET-mARF1*.

pET-M*YM and pET-YYM were constructed by transferring the BstB1 and HindIII fragment of pET-mARF1* to the corresponding regions of pET-M*YY and pET-yARF2, respectively. pET-YMY and pET-M*MY were constructed by ligating the BstBI and HindIII fragment of pET-yARF2 into the corresponding region of pET-YMM and pET-mARF1*, respectively.

The polyclonal antibody RY/1 to the µ1 subunit of AP-1 (27) was provided by Linton Traub of our laboratory and was used at a concentration of 1:5000 for immunoblotting. The monoclonal antibody M3A5, against -COR subunit of coatomer, was a gift from Thomas Kreis (University of Geneva, Geneva, Switzerland) (28) and was used at a dilution of 1 µg/ml. Horseradish peroxidase-conjugated antibodies against mouse and rabbit immunoglobins were purchased from Amersham Corp.

Recombinant myristoylated and unmyristoylated ARFs were prepared from Escherichia coli as described previously (21), except that yARF2 was expressed from pET-yARF2, and therefore its expression was induced like the mammalian ARFs. All of the mARF1/yARF2 chimeras were produced and prepared as described for mammalian ARF1. 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.

The extent of myristoylation of each ARF preparation was determined by a gel shift assay as described previously (21) and used to calculate the myristoylated ARF concentration in each experiment. All chimeras having yARF2 residues at positions 3-7 were well myristoylated (70-100%) when produced in bacteria co-expressing myristoyl-CoA:protein N-myristoyltransferase, consistent with the results of Randazzo and Kahn (29).

Rat liver Golgi membranes and an AP-1- and coatomer-enriched fraction of rat liver cytosol were prepared as described (10, 21). The final coat protein-enriched fraction had an AP-1 concentration about three times that in cytosol, as determined by quantitative immunoblotting. No ARF was detectable in this fraction as analyzed by ligand blotting with [-³²P]GTP.

A guanine nucleotide exchange assay was developed based on the assays of Northup et al. (30) and Kahn and Gilman (31). Recombinant ARF was added with or without 5 µg of Golgi membranes to a 100-µl reaction mixture containing 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol, 1 mM ATP, 100 µg/ml bovine serum albumin, and 10 µM [³⁵S]GTPS (3-10 × 10⁵ cpm). The reaction mixture was incubated for 10 min at 37 °C, and the reaction was stopped by passing 50 µl over a HA-High Protein and Nucleic Acid Binding Plate in the Multiscreen Filtration System Vacuum Manifold (Millipore). Filters were washed 7 times with 200 µl of 25 mM Hepes, pH 7.0, 50 mM potassium acetate, 2.5 mM magnesium acetate. Radioactivity bound to the filter was measured and the amount of GTPS bound to protein calculated. All reactions were done in duplicate or triplicate. Background binding of [³⁵S]GTPS to the nitrocellulose, determined by filtering a sample containing only [³⁵S]GTPS and buffer, was less than 0.1% of the radioactivity added. BFA, when included, was added to 50 µg/ml from a 5 mg/ml stock solution in ethanol, and an equal volume of ethanol was added to control reaction mixtures.

The recruitment assays were performed in a total volume of 200 µl in 1.5-ml siliconized tubes to reduce background. The AP-1/coatomer-enriched cytosolic fraction was cleared by centrifugation at 220,000 × g for 20 min at 4 °C before use in the assay. Golgi membranes (50 µg/ml), recombinant myristoylated ARF to the indicated concentration, the AP-1/coatomer-enriched cytosolic fraction (5.6 mg/ml), and either 50 µM GTPS or 1 mM GDP were mixed together in assay buffer (25 mM Hepes, pH 7.0, 250 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol, 1 mg/ml bovine serum albumin) on ice. After incubation for 15 min at 37 °C, the tubes were returned to ice, and the reaction mixtures were transferred to fresh 1.5-ml siliconized tubes to further reduce background. The Golgi membranes were sedimented by centrifugation (14,700 × g, 5 min, 4 °C), and the Golgi pellet was washed two times with 100 µl of 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 subjected to SDS-PAGE in 13% gels. Bound ARF was detected by Coomassie Blue staining, bound AP-1 by immunoblotting using RY/1, a polyclonal antibody that recognizes µ1, and bound coatomer by immunoblotting using M3A5, a monoclonal antibody that recognizes -COP. They were quantified by densitometry.

Electrophoresis and immunoblotting were performed as described previously (10). Gels to be analyzed by fluorography were incubated in 1% glycerol, 25% methanol for at least 1 h, and then in Amplify (Amersham Corp.) for 30 min. Gels were dried onto paper and exposed to X-Omat AR film (Eastman Kodak). GTP-ligand blotting was carried out essentially as described (32).

Human PLD1a was expressed in Sf9 insect cells using a baculovirus vector and then immunoaffinity purified using PLD1-specific antibodies (33).

Recombinant ARF was activated using GTPS as described previously (33). PLD activity was measured using the in vitro head group release assay as described previously (11, 22, 33). Assays were carried out in duplicate at least three times for each mutant ARF. Intra-assay variance was approximately 3%.

RESULTS

We previously reported that mARF1 is activated well by the Golgi GEF, whereas yARF2 is not (21). These two ARFs are 77% identical over their entire amino acid sequence but exhibit considerably less conservation of the 18 amino-terminal residues (Fig. 1). In addition, a peptide corresponding to amino acids 2-18 of mARF1 inhibits the activation of mARF1 by Golgi membranes in vitro (4) and slightly shorter peptides stimulate the formation of secretory granules on the TGN (34-36). Therefore, to begin the mapping of amino acids important for mARF1's activation by the Golgi GEF, we initially focused on residues in the amino-terminal regions of mARF1 and yARF2.

[View Larger Version of this Image (19K GIF file)]

In the first chimera, denoted mARF1*, amino acids 3-7 of mARF1 were changed to the corresponding residues of yARF2. Golgi membranes stimulated guanine nucleotide exchange on myristoylated mARF1 and mARF1* equally over a range of ARF concentrations (Fig. 2). The activation of myristoylated mARF1* by the Golgi membranes was inhibited 85% by BFA, an inhibitor of the Golgi GEF (Table I), indicating that this mutant ARF retains the ability to interact with Golgi exchange factor. In contrast, activation of myristoylated yARF2 by rat Golgi membranes was poorly inhibited by BFA (Table I).

[View Larger Version of this Image (16K GIF file)]

Table I. BFA sensitivity of Golgi membrane-stimulated guanine nucleotide exchange on mutant and wild type ARFs Myristoylated ARF was added to 5 µM. Values are calculated for the entire 100-µl reaction mixture and are the average of three determinations ±S.D. Background binding of GTPS to Golgi membranes and ARF incubated alone has been subtracted. Golgi-stimulated exchange Inhibition  BFA +BFA pmol GTPS % mARF1* 16.1  ± 0.9 2.4  ± 0.1 85 mARF1*(K10S, G11N) 16.3  ± 2.2 2.5  ± 0.7 84 mARF1*(K15N) 7.1  ± 0.8 2.0  ± 0.4 72 mARF1*(K15A) 9.7  ± 2.5 2.5  ± 0.8 74 yARF2(N15K) 7.2  ± 0.6 5.2  ± 0.2 28 yARF2 10.0  ± 3.6 8.2  ± 2.2 17

Since mARF1* was activated as well as mARF1, but was better myristoylated when produced in bacteria co-expressing myristoyl-CoA:protein N-myristoyltransferase (29), all additional mutants were made with yARF2 residues at positions 3-7. This is indicated by the * in their name. In the chimeras mARF1*(K10S,G11N) and mARF1*(K15N), amino acids 10, 11, and 15 of mARF1 were changed to the corresponding yARF2 residues. Mammalian ARF1*(K10S,G11N) was activated by the Golgi membranes to the same extent as mARF1*. By contrast, activation of mARF1* (K15N) was decreased to the yARF2 level (Fig. 3), but its activation was highly sensitive to BFA (Table I). This indicated that mARF1* (K15N) was being activated primarily through the Golgi GEF, although the extent of activation was less than mARF1. In contrast, activation of the reciprocal mutant, yARF2 (N15K) was not sensitive to BFA (Table I). Furthermore, when amino acid 15 of mARF1 was changed to an alanine instead of an asparagine, activation by the Golgi GEF approached that obtained with mARF1*, and this activation was very BFA-sensitive (Fig. 3 and Table I). These data indicate that lysine 15 is neither essential nor sufficient for activation by the Golgi GEF.

[View Larger Version of this Image (23K GIF file)]

In summary, the mutants mARF1*, mARF1*(K10S,G11N), and mARF1*(K15N) change all of the amino acids that differ between mARF1 and yARF2 in the region spanned by the inhibitory peptide (residues 2-18), and they are all activated by the Golgi GEF. Therefore, amino acids outside of this domain must be important for mARF1's interaction with the Golgi GEF.

To evaluate the role of domains outside of amino acids 2-18, the coding sequences of mARF1* and yARF2 were separated into three parallel regions using restriction sites at the codons for amino acids 34 and 94. As a result, region I contained amino acids 2-34, and regions II and III included residues 35-94 and 95-181, respectively. Chimeras encoding all possible combinations of yARF2 and mARF1* sequences in these regions were constructed, and they were named as diagrammed in Fig. 4. For example, the chimera M*YM had mARF1* (M*) sequences in region I, yARF2 sequences (Y) in region II, and mARF1 sequences (M) in region III.

[View Larger Version of this Image (25K GIF file)]

Activation of YMM by the Golgi membranes was slightly reduced compared with mARF1*, particularly at lower ARF concentrations (Fig. 5). This may be due to the presence of an asparagine at position 15, which impairs the activation of mARF1*(K15N), as shown in Fig. 3. Importantly, YMM activation was very sensitive to BFA (Table II). M*YY behaved similarly to yARF2 in terms of activation by the Golgi GEF (Fig. 5) and insensitivity to BFA (Table II). Taken together, the results with YMM and M*YY indicate that residues of mARF1 outside of region I are essential for activation by the Golgi GEF.

[View Larger Version of this Image (31K GIF file)]

Table II. BFA sensitivity of Golgi membrane-stimulated guanine nucleotide exchange on mARF1/yARF2 chimeric ARFs Myristoylated ARF was added to 5 µM. Values are calculated for the entire 100-µl reaction and are the average of three determinations ±S.D. Background binding of GTPS to Golgi membranes and ARF incubated alone has been subtracted. The experiments were repeated at least three times and a representative example is shown. Chimera Golgi-stimulated exchange Inhibition  BFA +BFA pmol GTPS % YMM 10.7  ± 1.1 2.0  ± 1.0 82 YYM 10.6  ± 1.4 5.4  ± 0.8 50 M*YM 8.4  ± 1.3 4.6  ± 0.5 45 M*MY 9.4  ± 2.5 7.5  ± 1.4 20 YMY 6.2  ± 0.8 5.5  ± 0.7 12 M*YY 5.6  ± 0.7 4.6  ± 0.9 17

We next tested whether either region II or III of mARF1 was capable of mediating activation by the Golgi GEF. YMY, which contains only region II of mARF1, was activated to approximately the same extent as yARF2 (Fig. 5), and its activation was not significantly inhibited by BFA (Table II). In contrast, YYM, which encodes only region III of mARF1, was activated by the Golgi membranes considerably better than yARF2, although not quite as well as mARF1* (Fig. 5). The activation of this chimeric ARF was inhibited by BFA to an intermediate extent (Table II). Therefore, region III of mARF1 contains amino acids critical for activation by the Golgi GEF.

Table III summarizes the effect of BFA on the activation of the various chimeric ARFs observed over multiple experiments. Activation of chimeras with regions II and III of mARF1 was inhibited 75-84%, whereas activation of chimeras having only region III was inhibited 45-46%. Chimeras with region III of yARF2 were inhibited less than 25%. These differences between the three groups of chimeric ARFs are highly significant. The results confirm that region III of mARF1 is essential for activation by the Golgi GEF and support a minor role for region II.

Table III. Summary of BFA sensitivity of mARF1/yARF2 chimera activation The average BFA sensitivity of selected chimeric ARFs was calculated from several independent determinations. The BFA sensitivity of YYM is significantly different than that of mARF1*, YMM, YMY, and yARF2 (p < 0.001); M*MY (p < 0.01); and M*YY (p < 0.05). The BFA sensitivity of M*YM is also significantly different than that of mARF1* and yARF2 (p < 0.001); YMM and YMY (p < 0.01); and M*MY (p < 0.02). Chimera Inhibition by BFA % mARF1*(n = 49) 84  ± 6 YMM (n = 6) 75  ± 9 YYM (n = 6) 46  ± 9 M*YM (n = 6) 46  ± 14 M*MY (n = 6) 21  ± 16 M*YY (n = 6) 24  ± 21 YMY (n = 5) 15  ± 5 yARF2 (n = 40) 10  ± 19

After myristoylated ARF has been activated, it binds to Golgi membranes and promotes the recruitment of the coat protein complexes AP-1 and coatomer (7-10, 21, 37). Since mARF1 facilitates AP-1 recruitment more efficiently than yARF2 (21), the mARF1/yARF2 chimeras were used to map the domains of mARF1 required for optimal AP-1 recruitment.

The ability of the chimeric ARFs to promote AP-1 binding was determined using an in vitro recruitment assay (10, 21). In this assay, purified rat liver Golgi membranes were incubated with an AP-1/coatomer-enriched fraction of cytosol, recombinant myristoylated ARF, and GTPS. The membranes were reisolated by centrifugation and analyzed by SDS-PAGE followed by immunoblotting or Coomassie Blue staining. AP-1 was detected with an antibody to its µ1 subunit, coatomer with an antibody to its -COP subunit, and ARF by Coomassie Blue staining. The immunoblots and stained gels were quantitated by densitometry. The amount of coat protein recruited was plotted versus the amount of ARF bound to the Golgi membranes since only the bound ARF is capable of facilitating coat protein recruitment. Coatomer binding served as a positive control, since mARF1 and yARF2 promote coatomer recruitment about equally (21).

Mammalian ARF1* promoted AP-1 binding almost as well as mARF1 and much better than yARF2 indicating that amino acids 3-7 of mARF1 are not required for efficient AP-1 recruitment (Fig. 6). Comparing M*MY and YYM with mARF1* revealed that M*MY promoted AP-1 recruitment similarly to mARF1* (Fig. 7A), whereas YYM was much less active (Fig. 7B). Therefore the amino-terminal half of mARF1 (residues 2-94) is important for optimal AP-1 recruitment.

[View Larger Version of this Image (23K GIF file)]

[View Larger Version of this Image (22K GIF file)]

To dissect further the amino-terminal half of mARF1, we tested the ability of M*YY and YMY to facilitate AP-1 recruitment (Fig. 8). It is apparent that M*YY behaved like yARF2 in this recruitment assay (Fig. 8A), whereas YMY promoted AP-1 binding to the same extent as mARF1* and better than yARF2 (Fig. 8, B and C). These results implicate amino acids 35-94 of mARF1 as being essential for efficient AP-1 recruitment. The YYM and M*YY chimeras promoted coatomer recruitment to the same extent as mARF1*, showing that the inefficient recruitment of AP-1 by these chimeras reflects a specific defect rather than a general loss of activity.

[View Larger Version of this Image (18K GIF file)]

Since PLD is a potential effector of ARF, we next examined the ability of various ARFs plus mARF1* to activate purified PLD1 in vitro. For these assays, the ARFs were loaded with GTPS artificially using a previously described technique (33), with an approximate efficiency of 20-30% (data not shown). Mammalian ARF1 and mARF1* activated PLD1 to the same maximal level (Fig. 9A). In contrast, mARF6 and yARF2 activated PLD1 to a lesser extent than mARF1 over the range of ARF concentrations tested.

[View Larger Version of this Image (17K GIF file)]

The results described above indicated that the mARF1/yARF2 hybrids could be used to map the domains of mARF1 necessary for effective PLD1 activation. Therefore, the chimeras described in Fig. 4 were evaluated for their ability to activate human PLD1 in vitro. As shown in Fig. 9B, the YMM chimera activated the PLD1 in the same manner as mARF1, whereas the M*YY chimera showed poor activity in the assay, similar to that of yARF2. Since the experiments shown in Fig. 9 revealed that the added ARFs were similarly potent but activated PLD1 to different maximal extents, the other mutant ARFs were tested at a concentration of 1 µM to assess maximal activation (Table IV). These experiments showed that the ability of the chimeric ARFs to activate PLD1 correlated with region II. Thus substitution of yeast region II for mammalian region II (M*YM) decreased the effectiveness of PLD1 activation to the level observed with yARF2, whereas the complementary chimera (YMY) activated PLD1 78 ± 10% as well as mARF1*. Consistent with these results, the YMM and M*MY chimeras activated PLD1 as effectively as mARF1*, and the reciprocal chimeras, M*YY and YYM, activated PLD1 to the same extent as yARF2.

Table IV. Activation of PLD1 by mARF1/yARF2 chimeras Purified recombinant PLD1 was combined with GTPS-activated myristoylated ARFs (1 µM), and PLD activation was measured using the in vitro head group release assay as described under "Experimental Procedures." Values are averages of at least three experiments, and are presented as the percent of the value obtained for mARF1*. Chimera PLD activation % mARF1* 100 yARF2 48  ± 13 YMM 98  ± 13 M*YY 56  ± 15 M*MY 101  ± 9 YYM 48  ± 17 M*YM 41  ± 22 YMY 78  ± 10

DISCUSSION

We have used mARF1/yARF2 chimeras along with point mutants of mARF1 to map the domains of ARF that are important for interaction with its targets. Our data indicate that residues 35-94 of mARF1 are required for optimal activation of PLD1 and efficient AP-1 recruitment, whereas the carboxyl half of mARF1 (amino acids 95-181) contains residues essential for interaction with the rat Golgi GEF. In addition, region II (amino acids 35-94) has a minor role in the interaction with the Golgi GEF.

Since a peptide corresponding to amino acids 2-18 of mARF1 inhibits activation of mARF1 by Golgi membranes in vitro (4), we were surprised to find that the region of mARF1 delineated by residues 2-34 did not contain unique amino acids critical for interaction with the Golgi GEF. However, a number of the amino acids located in positions 2-18 are conserved between mARF1 and yARF2 (Fig. 1), so it may be that the inhibitory peptide disrupts an important interaction mediated by one or more of the conserved residues. Alternatively, this peptide may have a nonspecific disruptive effect on Golgi membranes (38). It has also been reported that an amino-terminal mARF1 peptide facilitates AP-1 recruitment onto immature secretory granules in the TGN (35) and stimulates the formation of secretory vesicles (34, 36). In these studies high levels of myristoylated peptide were required for optimal activity (100 µM) (34, 35), raising the possibility that the effect was due to some residual activity in the amino terminus of ARF1.

Recently it has been shown that the physical association of mARF1 with phosphatidylinositol 4,5-bisphosphate, which regulates ARF1's interactions with ARF-GAP, is partially disrupted by mutating lysines 15 and 16 to leucines (39). The lysine at position 16 is shared by mARF1 and yARF2, but the latter ARF has an asparagine at position 15. Substitution of Lys-15 of mARF1 with an alanine had no significant effect on the interaction with the Golgi GEF, although the KI5N mutation caused a modest inhibition in this assay (Fig. 3). Since the effect of the single K15L mutation on phosphatidylinositol 4,5-bisphosphate association was not examined (39), we are unable to determine whether this lipid interaction is required for activation of mARF1 by the Golgi GEF.

The switch domains I and II of small GTP-binding proteins have different conformations depending upon whether GTP or GDP is bound (40, 41). These domains are often important for GEF interactions. For example, chimeric analysis demonstrated that amino acids 1-102 of Rab 3a, which encompass both putative switch domains, are sufficient to mediate maximal binding to its exchange factor Mss4 (42). In addition, Ras switch domain II and neighboring residues interact with several exchange factors (43-48). Regions outside of the switch domains also interact with GEFs, but the important domains are not conserved between different families of small GTP-binding proteins. The amino terminus of Rab3a is absolutely required for binding to Mss4 (42), and the helix  3/loop L7 region of Ras interacts with GEFs (47).

The GTP bound form of ARF has not been crystallized, so the switch domains of ARF can only be predicted by analogy with the Ras structure. Switch domain I of Ras corresponds to amino acids 44-51 of mARF1, and switch domain II corresponds to amino acids 70-86 (49, 50). Therefore neither of the putative switch domains of mARF1 falls into the region critical for activation by the Golgi GEF (amino acids 95-181), but both fall into a second region that had a slight contribution to the interaction (amino acids 35-94). It should be noted that the crystal structure of ARF has several significant differences from the crystal structure of Ras. Most strikingly, the switch I domain of Ras is a surface loop, whereas the corresponding region of ARF contributes mainly to a -strand unique to ARF (49, 50). Consequently, the accuracy of this comparison must await the structural determination of ARF in the GTP bound form.

The region of mARF1 containing the putative switch I and II domains (residues 35-94) was sufficient for both maximal activation of PLD1 and efficient AP-1 recruitment. Zhang et al. (51) measured the activation of partially purified rat brain PLD by mARF1 and human ARF-like 1 (ARL1) chimeras. Consistent with our results, they demonstrated that the first 73 amino acids of mARF1 were necessary and sufficient for PLD activation.

Since the same region of mARF1 (residues 35-94) was important for both maximal PLD1 activation and efficient AP-1 recruitment, we hypothesize that amino acids in this region interact with effector molecules. Several observations support this idea. First, purified recombinant ARF can activate purified recombinant PLD1, ruling out a requirement for other proteins (33). In the simplest model, ARF activates PLD directly. If this is the case, we have probably been mapping the ARF domains necessary to bind PLD. However, it is also possible that ARF activates PLD1 indirectly through effects on phospholipids present in the reaction mixture. Second, as stated above, this region of ARF encompasses the putative switch domains, and residues in the Ras switch domains directly interact with effectors (reviewed in Ref. 52). Finally, mARF1-GDP crystallizes as a dimer, and the point of contact between the two ARF molecules is in the unique -strand that corresponds to the Ras switch I domain (49, 50). Although the functional significance of this dimerization is unknown, it may reflect the importance of this region for protein-protein interactions.

PLD may be a downstream effector of ARF in COPI-coated vesicle formation (13, 15). We found that the ability of mARFs1, -6, and yARF2 to activate PLD1 did not correlate with their ability to promote coatomer binding to Golgi membranes. Yeast ARF2 and mARF6 activated PLD1 less effectively than mARF1 but promote coatomer recruitment as efficiently as mARF1 (21). One possibility is that PLD1 is the wrong PLD isoform. This seems unlikely, as PLD1 is found on Golgi membranes (16, 53). Regardless, these results do not rule out PLD involvement in coatomer recruitment. Each of the ARFs tested activated PLD1 to some extent, and perhaps only a low level of PLD activation is required for maximal coatomer recruitment.

Finally, analysis of the mARF1/yARF2 chimeras leads to an insight into the mechanism of coated vesicle formation. We previously found that purified rat liver Golgi membranes can activate ARF by two mechanisms. The first, mediated by the Golgi GEF, is inhibited by BFA and by trypsin treatment of the Golgi membranes (2-4, 21). The second mechanism is resistant to both BFA and trypsin, suggesting that the Golgi membrane phospholipids are activating ARF (21). Yeast ARF2 is only activated by the trypsin-resistant, GEF-independent mechanism, and yARF2 promotes AP-1 binding to the Golgi very inefficiently compared with mARF1, even if coat protein binding is adjusted for the amount of ARF on the Golgi membranes (21). One possible interpretation of these results is that ARF has to be activated by the Golgi GEF to function effectively in AP-1 recruitment. We can now exclude this possibility. The mARF1/yARF2 chimera YMY was not activated by the Golgi GEF, but after activation by the GEF-independent mechanism, it promoted AP-1 binding as well as mARF1*. This indicates that the mechanism of ARF activation has no influence on ARF's ability to function in downstream steps of coat formation.

While this paper was under review, Paris and colleagues (54) found that truncated mARF1 lacking the first 17 amino acids is activated by the BFA-insensitive exchange factor ARNO, demonstrating that the amino-terminal region of mARF1 is not the site of interaction with this exchange factor. This is consistent with our results.