Molecular Characterization of the GTPase-activating Domain of ADP-ribosylation Factor Domain Protein 1 (ARD1)*

ADP-ribosylation factors (ARFs) are ; 20-kDa guanine nucleotide-binding proteins recognized as critical components in intracellular vesicular transport and phospholipase D activation. Both guanine nucleotide-ex-change proteins and GTPase-activating proteins (GAPs) for ARFs have been cloned recently. A zinc finger motif near the amino terminus of the ARF1 GAP was required for stimulation of GTP hydrolysis. ARD1 is an ARF family member that differs from other ARFs by the presence of a 46-kDa amino-terminal extension. We had reported that the ARF domain of ARD1 binds specifically GDP and GTP and that the amino-terminal extension acts as a GAP for the ARF domain of ARD1 but not for ARF proteins. The GAP domain of ARD1, synthesized in Escherichia coli , stimulated hydrolysis of GTP bound to the ARF domain of ARD1. Using ARD1 truncations, it appears that amino acids 101–190 are critical for GAP activity, whereas residues 190–333 are involved in physical interaction between the two domains of ARD1 and are required for GTP hydrolysis. The GAP function of the amino-terminal extension of ARD1 required two arginines, an intact zinc finger motif, and a group of residues which resembles a sequence present in Rho/ Rac GAPs. Interaction between the two domains of ARD1 required two negatively charged residues (Asp 427 and Glu 428 ) located in the effector region of the ARF domain and two basic amino acids (Arg 249 and Lys 250 ) found in the amino-terminal extension. The GAP domain of ARD1 thus is similar

2). ARF1 acts as a key regulator of the interactions of non-clathrin coat protein (coatomer) with Golgi stacks (3) and of clathrin adaptor particles with the trans-Golgi network (4). ARF proteins also activate phospholipase D (5,6). Guanine nucleotide binding to ARFs, like that to other monomeric G proteins, appears to be governed by guanine nucleotide-exchange proteins (GEPs) and GTPase-activating proteins (GAPs) (2). ARF GEPs (7, 8; for review, see Ref. 2) and GAPs (9 -12) have been purified and cloned. The deduced amino acid sequence of ARF1 GAP from rat liver has a zinc finger motif near the amino terminus, which was required for GAP activity (11). The GAP appeared to be recruited to the Golgi by an ARF1-dependent mechanism (11).
Although the roles of G proteins are extremely diverse, they all operate by a fundamentally similar mechanism (13). When GTP occupies the guanine nucleotide-binding site, the G protein can interact with and modify the activity of a downstream target protein. Hydrolysis of GTP causes dissociation of the G protein-target complex and terminates the "active state" of the G protein. Cells regulate the ratio of active and inactive G proteins by modulating the rates of GDP release and GTP hydrolysis (GTPase activity).
It was reported that dissociation of GDP from the ARF domain of ARD1 was faster than from ARD1 itself (14). Using ARD1 truncations, the 15 amino acids immediately preceding the ARF domain were shown to be responsible for decreasing the rate of GDP, but not GTP, dissociation (15). By site-specific mutagenesis it was shown that hydrophobic residues in this region were particularly important in stabilizing the GDPbound form of ARD1. Therefore, it was suggested that, like the amino-terminal segment of ARF, the equivalent region of ARD1 may act as a GDP dissociation inhibitor.
Until recently, it was believed that GTP hydrolysis by monomeric G proteins was stimulated by separate GAPs, whereas the presence of an intrinsic GAP-like domain in G␣ was responsible for GTPase activity in the heterotrimeric G proteins (16). Some effector proteins that are regulated by heterotrimeric G proteins also act as GAPs for their G protein regulators. The GAP activities of the effectors, such as phospholipase C-␤ (17) and the cGMP phosphodiesterase ␥ subunit (18), may allow effector-specific modulation of responses. A relatively new class of GAPs for heterotrimeric G proteins includes the RGS (regulator of G protein signaling) family (19). They can contribute to desensitization induced by a prolonged signal or act as long term attenuators of signal amplitude, presumably by stimulating GTP hydrolysis (for review, see Ref. 20).
ARFs are ϳ20-kDa proteins that exhibit no detectable GTPase activity (21). Like ARFs, the 18-kDa ARF domain (p3) of the 64-kDa ARD1 binds specifically GDP and GTP and lacks detectable GTPase activity (22). Using recombinant proteins, it was shown that the 46-kDa amino-terminal domain of ARD1 (p5) stimulates hydrolysis of GTP bound to p3, and conse-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  quently it appears to be the GAP component of this bifunctional protein (14). The stimulatory effect of the p5 domain on the GTPase activity of p3 was specific, as GTP hydrolysis by other members of the ARF family was not increased (12). Based on these and prior data on ARD1 (14,15), it appeared that p5 may control GTP hydrolysis as well as GDP dissociation.
We reported that functional and physical interactions between p3 and p5 required two negatively charged amino acids in the "effector" region of p3 (22). We report here that these residues probably interact with two positively charged amino acids in the amino-terminal extension (p5). Using affinity-purified antibodies and truncated mutants of ARD1, we show here that the amino terminus of ARD1 is not required for these interactions. By site-specific mutagenesis, we demonstrate further that in p5 an intact zinc finger motif, two arginines, and a sequence that resembles a consensus motif present in Rho/Rac GAPs are required for GAP activity.

EXPERIMENTAL PROCEDURES
Materials-Bovine thrombin was purchased from Sigma, TLC plates from VWR Scientific, and GSH-Sepharose beads from Pharmacia Biotech Inc. Polymerase chain reaction reagents and restriction enzymes, unless otherwise indicated, were from Boehringer Mannheim. Sources of other materials have been published (7,14,15,22,23).
Preparation of Recombinant Fusion Proteins (p3, p5, and p8)-For large scale production of fusion proteins (14), 10 ml of overnight culture of transformed bacteria were added to a flask with 1 liter of LB broth and ampicillin, 100 g/ml, followed by incubation at 37°C with shaking. When the culture reached an A 600 of 0.6, 500 l of 1 M isopropyl-␤-D-thiogalactopyranoside were added (0.5 mM final concentration). After incubation for an additional 3 h, bacteria were collected by centrifugation (Sorvall GSA, 6,000 rpm, 4°C, 10 min) and stored at Ϫ20°C. Bacterial pellets were dispersed in 10 ml of cold phosphatebuffered saline, pH 7.4, with trypsin inhibitor, 20 mg/ml, leupeptin and aprotinin, each 5 mg/ml, and 0.5 mM phenylmethylsulfonyl fluoride. Lysozyme (20 mg in 10 ml) was added. After 30 min at 4°C, cells were disrupted by sonication and centrifuged (Sorvall SS34, 16,000 rpm, 4°C, 20 min). Fusion proteins, purified on glutathione-Sepharose, were ϳ90% pure as estimated by silver staining after SDS-PAGE (12). After cleavage by bovine thrombin, GST was removed with glutathione-Sepharose beads and thrombin with benzamidine-Sepharose 6B (24). Proteins were purified further by gel filtration through Ultrogel AcA 54 and then Ultrogel AcA 34 before storage in small portions at Ϫ20°C. Purity, estimated by silver staining after SDS-PAGE, was Ͼ98%. Amounts of purified proteins were estimated by a dye-binding assay (25) and by SDS-PAGE using bovine serum albumin as standard. ARF1(39 -45p3) was synthesized as published (22).
Construction and Expression of Mutated p5-For site-directed mutagenesis of p5, a modification of the unique site-elimination mutagen-esis procedure of Deng and Nickoloff (26) was used. 25 pmol of a 5Ј-phosphorylated selection primer and 25 pmol of a 5Ј-phosphorylated mutagenic primer were annealed simultaneously to 750 ng of p5-pGEX5G/LIC in 20 ml of 10 mM Tris acetate, pH 7.5, 10 mM magnesium acetate, 50 mM potassium acetate by heating for 5 min at 100°C and cooling 5 min on ice, followed by incubation at room temperature for 30 min. The selection primer 5Ј-CTGTGACTGGTGACGCGTCAACCAAG-TC-3Ј changed a ScaI restriction site in the Amp r gene of pT7 into a MluI restriction site (underlined). Mutagenic primers (see Fig. 6 and Tables I-III) introduced the desired mutations. Primers were extended with T7 DNA polymerase, and the new strands were ligated with T4 DNA ligase for 1 h at 37°C (final volume 30 l). Plasmids were then digested for 2 h at 37°C with 20 units of ScaI (final volume 60 l). 4-l samples were used to transform 90 ml of Epicurian Coli XL1-Bluecompetent cells (Stratagene). Plasmids were purified with Miniprep Wizard (Promega) from bacteria grown overnight in 2 ml of 2 ϫ YT broth with ampicillin, 100 mg/ml. Samples (500 ng) of the plasmids were digested with 20 units of ScaI for 3 h at 37°C. 4-l samples were used to transform 40 l of XLmutS-competent cells (Stratagene). Colonies were screened selectively by digestion with MluI, and the presence of the mutations was confirmed by automated sequencing (Applied Biosystems, 373 DNA Sequencer) using the primers 5Ј-TTA-TACGACTCACTATAGGG-3Ј, 5Ј-ATGATTGTAGAGTTGTCTT-3Ј, and 5Ј-GCTAGTTATTGCTCAGCGG-3Ј. Large scale production of mutated p5 proteins was carried out as described for ARD1.
Assay of GTPase Activity-Samples were incubated for 30 min at 30°C in 20 mM Tris, pH 8.0, 10 mM dithiothreitol, 2.5 mM EDTA with bovine serum albumin, 0.3 mg/ml, and cardiolipin, 1 mg/ml, then for 40 min at 30°C in the same medium with 0.5 M [␣-32 P]GTP (3,000 Ci/mmol) and 10 mM MgCl 2 (total volume 120 l). After addition of p5 or mutant proteins (40 l), incubation at room temperature was continued for 1 h (final volume 160 l) before proteins with bound nucleotides were collected on nitrocellulose (23). Bound nucleotides were eluted in 250 l of 2 M formic acid, of which 3-4-l samples were analyzed by TLC on polyethyleneimine-cellulose plates (14), and 240 l was used for radioassay to quantify total 32 P-nucleotide. TLC plates were subjected to autoradiography at Ϫ80°C for 18 -28 h. Total amounts of labeled nucleotides (GTP ϩ GDP) bound to p3, p8, or p3 after incubation with p5, whether quantified by radioassay of the formic acid solution, by counting total radioactivity on the filter, or by Phos-phorImaging (Molecular Dynamics) after TLC, were not significantly different under any condition (14), except as mentioned. An increase in bound GDP was always correlated with a decrease in bound GTP (22).
Assay of Cholera Toxin-catalyzed ADP-ribosylagmatine Forma-tion-p3 or ARD1 was incubated for 30 min at 30°C in 40 l of 20 mM Tris, pH 8.0, 10 mM dithiothreitol, 2.5 mM EDTA with bovine serum albumin, 0.3 mg/ml, and cardiolipin, 1 mg/ml, before addition of 20 l of solution to yield final concentrations of 100 M GTP␥S or GTP and 10 mM MgCl 2 . Where indicated, p5 or mutant protein was then added for 30 min. Components needed to quantify ARD stimulation of cholera toxin-catalyzed ADP-ribosylagmatine formation were then added in 70 l to yield final concentrations of 50 mM potassium phosphate, pH 7.5, 6 mM MgCl 2 , 20 mM dithiothreitol, ovalbumin, 0.3 mg/ml, 0.2 mM [adenine-14 C]NAD (0.05 Ci), 20 mM agmatine, cardiolipin, 1 mg/ml, and 100 M GTP␥S or GTP with 0.5 g of cholera toxin (29). After incubation at 30°C for 1 h, samples (70 l) were transferred to columns of AG 1-X2 equilibrated with water and eluted with five 1-ml volumes of water (29). The eluate, containing [ 14 C]ADP-ribosylagmatine, was collected for radioassay.

RESULTS
Identification of the GAP Domain of ARD1-Incubation of p3 or p5 with affinity-purified polyclonal antibodies raised against recombinant p3 or p5, respectively, markedly reduced, in a concentration-dependent manner, the ability of p5 to stimulate hydrolysis of GTP bound to p3 (Fig. 1), whereas the antibodies did not affect GTP binding (data not shown). 30 g of either antibody completely blocked p5-stimulated GTPase activity (Fig. 1), whereas up to 50 g of an anti-GST antibody had no effect (data not shown). On the other hand, 30 g of anti-p3 or anti-p5 antibodies reduced the intrinsic GTPase activity of ARD1 only 8.8 Ϯ 1.2% and 9.6 Ϯ 0.9%, respectively (data not shown). 30 g of anti-p3 or anti-p5 antibodies reduced hydrol-ysis of GTP bound to p3 by only 26.5 Ϯ 2.3% and 32.3 Ϯ 1.9%, respectively, when added to p3 simultaneously with p5 (data not shown). These results indicated that anti-p3 or anti-p5 antibodies inhibited GTP hydrolysis more effectively when the two domains of ARD1 were present in separate proteins than when covalently linked in recombinant ARD1. Based on these data, the two antibodies may decrease GTP hydrolysis by decreasing the ability of the two domains to interact.
To characterize more precisely the GAP and interaction sites on p3 and p5, we prepared two polyclonal antibodies against undecapeptides corresponding to the amino-and carboxyl-terminal sequences. The affinity-purified carboxyl-terminal antibody only slightly reduced (ϳ20%) the amount of GTP bound to p3 (data not shown), perhaps by affecting the structure of the GTP binding pocket of the ARF domain. 30 g of carboxylterminal antibody reduced GTP hydrolysis by about 25% (Fig.  1), suggesting that when antibody was bound to p3, the affinity between the two domains of ARD1 was reduced, or the rate of GTP hydrolysis was decreased directly. 30 g of the affinitypurified amino-terminal antibody affected neither GTP binding (data not shown) nor GTP hydrolysis (Fig. 1), suggesting that the amino terminus of p5 might not be involved in the GAP activity.
We synthesized four mutants of ARD1 with amino-terminal deletion ( Fig. 2A) and used functional assays to monitor their conformational integrity. Binding of GTP␥S to ARF requires a strict positioning of residues involved in the nucleotide binding pocket and is responsible for the conformational switch that activates ARF proteins. No significant differences in GTP␥S binding among ARD1 and amino-terminal deleted mutants were observed (15). The ARF domain of ARD1 (p3) exhibited no detectable GTPase activity (14), whereas 35-40% of GTP bound to ARD1 (p8) was hydrolyzed in 1 h at room temperature (Fig. 2B). The ARD1 mutant lacking 88 amino acids at the amino terminus (N⌬88p8) retained GTPase activity (Fig. 2B). Deletion of 200, 304, or 387 residues from the amino terminus completely prevented GTP hydrolysis (Fig. 2B), whereas binding of [␣- 32  All members of the ARF family, in the presence of GTP or a nonhydrolyzable analog, serve as allosteric activators of CTA (2,29). The site of interaction with the toxin has been localized to the carboxyl-terminal region of ARF (30 -32). Removal of up to 304 amino-terminal residues from ARD1 did not affect CTA activation, whereas removal of 387 amino acids reduced it by about 28% (15 and Fig. 2C), suggesting that the sequence preceding the ARF domain contributes to its native conformation. Activation of CTA by p3 was similar to GTP and GTP␥S, although, as we have reported (15), it was less than that by ARD1. As expected, stimulation of CTA by ARD1 (p8) was less with GTP than with GTP␥S (Fig. 2C), presumably because of its ability to hydrolyze GTP but not GTP␥S. Similarly, the ability of N⌬88p8 to activate CTA in presence of GTP was much less than in presence of GTP␥S (Fig. 2C). Mutant proteins with larger deletions of the amino terminus activated CTA with the same potency in the presence of GTP and GTP␥S (Fig. 2C), consistent with an absence of significant GTPase activity.
Four additional amino-terminal deletion mutants of ARD1 (p8) were synthesized to identify more precisely the GAP site in p5. In GTP␥S binding and CTA activation, N⌬101p8, N⌬124p8, N⌬146p8, and N⌬161p8 did not differ significantly from p8 (data not shown). Removal of 101 amino acids from the aminoterminal end reduced GAP activity only 8.5 Ϯ 2.6%, whereas removal of 23, 45, or 60 additional residues decreased GAP activity by 49.6 Ϯ 2.3, 97.5 Ϯ 1.2, and 99.1 Ϯ 0.9%, respectively (Fig. 3), consistent with a GAP site localized to a region downstream of residue 101.
We had reported that addition of the amino-terminal domain FIG. 1. Effect of affinity-purified antibodies on p5-stimulated GTPase activity of p3. ARD1 contains an amino-terminal GAP domain (p5) and a carboxyl-terminal, GTP-binding ARF domain (p3). Molecular masses of p3, p5, and p8 expressed as recombinant proteins are indicated. 55 pmol of p3 (ϳ1 g) was incubated with [␣-32 P]GTP for 40 min at 30°C in 60 l of 20 mM Tris, pH 8.0, 10 mM dithiothreitol, 2.5 mM EDTA with bovine serum albumin, 0.3 mg/ml, and cardiolipin, 1 mg/ml and then for 30 min at 4°C with the indicated amount of affinity-purified antibodies raised against recombinant p3 or the undecapeptide corresponding to the carboxyl terminus of p3 (CtARD1) (200 l, final volume). 110 pmol of p5 (ϳ5 g) was incubated (30 min at 4°C) with the indicated amount of affinity-purified antibody raised against recombinant p5 or the undecapeptide corresponding to the amino terminus of p5 (NtARD1) before addition to p3 with [␣-32 P]GTP bound. GTP hydrolysis during the next 60 min at room temperature is expressed as the increase in GDP bound to p3 relative to that during incubation with p5 without antibody (ϭ100%) based on PhosphorImager quantification. Data are means of duplicate values Ϯ one-half the range. Error bars smaller than symbols are not shown. Each experiment was repeated at least once. of ARD1 (p5) increased hydrolysis of GTP bound to p3 in a concentration-dependent manner (14), with the maximal effect at a ratio of 2 mol of p5/mol of p3 (22). We synthesized three amino-terminal and three carboxyl-terminal deletion mutants of p5 (Fig. 4A). Consistent with the results obtained with amino-terminal deletion mutants of ARD1, N⌬88p5 stimulated hydrolysis of GTP bound to p3 (Fig. 4B) and decreased the activation of CTA by p3 in presence of GTP but not GTP␥S (Fig.  4C). Further deletion of the amino terminus completely abolished GAP activity of mutant proteins, as N⌬200p5 and N⌬304p5 did not hydrolyze GTP bound to p3 (Fig. 4B), and neither mutant reduced CTA activation in presence of GTP (Fig. 4C). In large excess (10 ϫ p3), N⌬200p5 and N⌬304p5 had no effect on hydrolysis of GTP bound to p3 (data not shown). Removal of 69 amino acids from the carboxyl terminus did not affect GAP activity (Fig. 4, B and C), although the larger carboxyl-terminal deletions of p5 entirely prevented GTP hydrolysis (Fig. 4, B and C). In large excess (10 ϫ p3), neither C⌬191p5 nor C⌬293p5 stimulated hydrolysis of GTP bound to p3 (data not shown). Altogether, these results indicate that the first 101 and the last 69 amino acids of p5 were not required for hydrolysis of GTP bound to p3. Thus, the GAP domain of ARD1 can be localized to residues 101-333.

Identification of the Interaction Domain between p3 and p5-
Physical interaction between p3 and p5 mutant proteins was evaluated using recombinant GST fusion proteins with p5 or mutant p5 bound to GSH-Sepharose beads that were then incubated with the ARF domain (p3) of ARD1. Proteins associated with the beads or interacting with them were eluted with GSH and separated by SDS-PAGE. We reported earlier that under those conditions p3 interacted with GST-p5 but not with GST (14). N⌬88GST-p5 and N⌬200GST-p5 both clearly interacted physically with p3, whereas N⌬304GST-p5 did not (Fig. 5). C⌬69GST-p5, but not C⌬191GST-p5 and C⌬293GST-p5, also associated with p3 (Fig. 5). Therefore, removal of the first 200 or the last 69 amino acids of p5 did not prevent physical interaction with p3, suggesting that the interaction domain may be located between residues 200 and 333. Since the mutant N⌬200GST-p5 was able to interact with p3 ( Fig. 5) but did not stimulate hydrolysis of GTP bound to p3 (Fig. 4, B and C), it appears that residues critical for GAP activity may be located between amino acids 101 and 200 in p5.
We demonstrated that p5 interacted functionally with the ARF domain of ARD1 but not with other ARF proteins (12). A small sequence of seven amino acids ( 426 QDEFMQP 432 ) located in the effector region, which differs in other ARFs, was demonstrated to be critical for functional and physical interaction between the two domains of ARD1 (22). Two negatively charged residues, Asp 427 and Glu 428 , as well as Pro 432 , appeared crucial for those interactions (22). To identify the positively charged residues that interact with Asp and Glu, we mutated a cluster of basic amino acids between residues 200 and 250. Three mutant proteins (K210G/H211A)p5, (H214A/ K215G/H216G)p5, and (R249A/K250G)p5, were synthesized as GST fusion proteins and used to evaluate physical interaction with p3. (K210G/H211A)GST-p5 and (H214A/K215G/H216G)-GST-p5 interacted with p3, whereas (R249A/K250G)GST-p5 did not (Fig. 6A), suggesting that Arg 249 and Lys 250 might be the residues that interact with negatively charged residues from the effector region of the ARF domain. Moreover, (R249A/K250G)p5 did not stimulate GTP hydrolysis by p3-GTP, whereas (K210G/ H211A)p5 and (H214A/K215G/H216G)p5 mutants were as efficient as the non-mutated p5 in promoting GTP hydrolysis (Fig.  6B). These results indicated that mutations of two amino acids that abolished physical interaction also prevented GAP activity, suggesting that association of basic residues from p5 with acidic residues from p3 may be required for GTP hydrolysis.
To define more precisely the role of that interaction site, FIG. 3. Intrinsic GTPase activity of amino-terminal deletion mutants of ARD1. Deletion of 101, 124, 146, or 161 amino acids from the amino terminus of ARD1 yielded N⌬101p8, N⌬124p8, N⌬146p8, and N⌬161p8, respectively. 55 pmol of ARD1 or mutated protein with [␣-32 P]GTP bound was incubated for 60 min at room temperature before separation of bound nucleotides by TLC. GTPase activity is expressed as the increase in bound GDP relative to the increase of GDP bound to p8 (ϭ100%) based on PhosphorImager quantification. Data are means of duplicates Ϯ one-half the range in one experiment representative of two with two independent protein preparations.

FIG. 2. Effect of amino-terminal deletions on the intrinsic
GTPase activity of ARD1. Panel A, deletion of 88, 200, 304, or 387 amino acids from the amino terminus of ARD1 yielded N⌬88p8, N⌬200p8, N⌬304p8, and N⌬387p8. Panel B, 55 pmol of p3, ARD1 (p8) or mutated ARD1 with [␣-32 P]GTP bound was incubated for 60 min at room temperature before bound nucleotides were separated by TLC. Positions of standard GTP and GDP are indicated on the left. Data are duplicate assays representative of at least three different protein preparations. Panel C, after the protein (70 pmol) was incubated with 100 M GTP or GTP␥S, stimulation of cholera toxin-catalyzed ADP-ribosylagmatine formation was assayed for 60 min at 30°C. ARD activity is the difference between CTA-catalyzed formation of [ 14 C]ADP-ribosylagmatine without and with ARD1 protein (nmol/h). Data are means of quadruplicates Ϯ one-half the range. These findings were replicated twice with two independent preparations of proteins.
synthetic peptides corresponding to the two interacting domains in p3 and p5 were used as competitors. A tridecapeptide corresponding to the effector region in p3 dramatically reduced p5-induced hydrolysis of GTP bound to p3, whereas a tridecapeptide corresponding to the equivalent region in ARF1 had no effect (Fig. 7A). A dodecapeptide corresponding to the region 245-256 (containing Arg 249 and Lys 250 ) also prevented p5 stimulation of GTP hydrolysis by p3, whereas a peptide with the same residues in random order (Rp5p) had no effect (Fig. 7A). The values of the mean inhibitory doses, ID 50 , were 9 M for the peptide p3 and 12 M for the p5. These results indicated that the two peptides prevented GTP hydrolysis, probably by competing for the interaction sites of the two domains of ARD1.
We reported recently that a chimeric protein ARF1(39 -45p3) in which amino acids 39 LGEIVTT 45 , in the effector region of ARF1, had been replaced with QDEFMQP (the sequence in p3) bound to p5 and increased its GTPase activity (22). Peptides from p3 and p5 inhibited p5-induced hydrolysis of GTP bound to ARF1(39 -45p3), with values of ID 50 virtually identical to those that are inhibitory with p3, whereas ARF1 and Rp5 peptides had no effect (Fig. 7B).
Like the anti-p3 and anti-p5 antibodies (Fig. 1), the p3 and p5 peptides had much smaller effects on the GTPase of ARD1 than they did on that of p5 plus p3 (Fig. 7C). It was therefore assumed that accessibility of the interaction site to antibodies and peptides is relatively limited when the two domains are in the conformation of the intact molecule, although 50 M p5 peptide did inhibit the GTPase activity of ARD1 by 38 Ϯ 0.9%, whereas Rp5p peptide had no effect (Fig. 7C).
Identification of Critical Residues in the GAP Domain of ARD1-Deletion of amino acids can have subtle but adverse effects on overall protein structure, sometimes with structural Panel C, after 70 pmol of p3 was incubated with 100 M GTP or GTP␥S for 30 min at 30°C and then with 30 l (140 pmol) of p5, mutated protein, or water (control) for 20 min at room temperature, ARD stimulation of cholera toxin-catalyzed ADP-ribosylagmatine formation was assayed for 60 min at 30°C as described in Fig. 2. Toxin activity without added p3 in each condition was subtracted. Data are means of values from quadruplicate assays Ϯ one-half the range in one experiment representative of two with two different protein preparations. changes in domains of the protein which are (in the linear sequence) far from the deletion. Although we could not observe any difference in overall protein structure of the p8/p5 deletion mutants from that of the wild type p8/p5 proteins, subtle adverse changes in structure cannot be ruled out completely. Therefore, to reduce the possibility that the observed differences among N⌬101p8, N⌬124p8, and N⌬146p8 in GTPase activity resulted from subtle perturbation of their three-dimensional organization, mutant proteins were constructed with single amino acid replacements, which should cause minimal disturbance of global protein structure.
The putative GAP domain, residues 101-200, contains several amino acids potentially important for the GTPase activity. A cluster of cysteines is predicted to form a zinc finger structure CX 2 CX 4 CX 2 C (where X is any amino acid) in the GAP region. To probe the role of these cysteines each was replaced with alanine (Table I). A possible role for the zinc finger structure was supported by the finding that replacement of Cys 139 , Cys 142 , Cys 147 , or Cys 150 with alanine, which is expected to prevent the formation of the zinc finger (33), resulted in a complete loss of GAP activity, whereas mutation of Cys 178 and Cys 190 had no effect ( Table I). All of the mutants interacted physically with p3 ( Table I), suggesting that the single mutations did not affect folding of the proteins.
Numerous proteins that enhance the GTPase activity of monomeric G proteins have been identified. Rho/Rac GAPs share three consensus sequences (34). There is limited similarity between the second consensus sequence of Rho/Rac GAPs, which is KXXXXXLPXPL (where X is any amino acid), and residues 158 -168 (KTLAKHRRVPL) of ARD1. Replacement of Lys 158 by Ala completely abolished GAP activity, whereas substitution of Gly for Lys 162 (which is not in the consensus sequence) did not affect GTPase activity (Table II). Moreover, replacement of Pro 167 and Leu 168 by two glycines prevented GTP hydrolysis (Table II). All three mutants (K158G)GST-p5, (K162G)GST-p5, and (P167G/L168G)GST-p5, were able to interact physically with p3, as well as GST-p5 (Table II), suggesting no major differences in folding. The results are consistent with a role for this motif in the GAP activity of p5.
Despite the biological and medical importance of signal transduction via monomeric G proteins, their mechanism of GTP hydrolysis remains controversial. For Ras, it is speculated that a significant fraction of the GAP-activated GTPase activity arises from an additional interaction of the ␤-␥ bridge oxygen of GTP with an arginine side chain that is provided in trans by GAP (for review, see Ref. 35). Single replacement of any of the four arginines present in the GAP region of p5 had no effect on the ability of GST fusion proteins to interact physically with p3 (Table III). Replacement of Arg 164 or Arg 165 , however, almost completely prevented GTP hydrolysis, whereas GAP activity of (R101G)p5 and (R126G)p5 was unchanged (Table III). Together these results indicate that Arg 164 and Arg 165 are critical residues for GAP activity and may therefore participate in the removal of the phosphoryl group of GTP bound to the ARF domain of ARD1. DISCUSSION Crystal structures of Ras (36,37) and G␣ (38 -41) proteins in their GTP-and GDP-bound forms have been solved. Ras and G␣ likely hydrolyze GTP by similar catalytic mechanisms. Nonetheless, by themselves, monomeric G proteins hydrolyze GTP at a rate about 100-fold lower than heterotrimeric G proteins (13). In the presence of Ras GAP, however, Ras hydrolyzes GTP at least 100-fold faster than G␣ s (43). One explanation of this difference is that Ras GAP resembles the so-called "helical domain" that is present in G␣ s but absent in Ras (16) and that both Ras GAP and the helical domain introduce into the catalytic cleft an arginine residue that helps to stabilize the transition state (35).
ARD1 exhibits significantly greater GTPase activity than other members of the Ras family (14). Although its GTP-binding domain (p3) has no GTPase activity, addition of the aminoterminal extension (p5) promoted hydrolysis of GTP bound to p3. Deletion of 101 and 69 amino acids from, respectively, the amino and carboxyl termini of p5 did not prevent physical and functional interactions with p3, thus demarcating a minimal domain required for GAP activity. The smallest GAP domain of ARD1 (232 residues) is comparable in size to the minimal catalytic domains of the Ras GAPs, p120 GAP , and neurofibromin, respectively, 272 and 229 amino acids (44).
The unusual GTPase activity of ARD1 made possible the identification of a region specifically involved in both functional and physical interaction between the GTP binding and the GAP domains of ARD1. Specific mutations of amino acids in the effector region of the ARF domain of ARD1 provided evidence for a function of two negatively charged residues (Asp 427 and Glu 428 ), as well as of Pro 432 , which presumably creates a curve in the ␤-sheet structure which could place charged residues in correct position for interaction with the GAP domain (22). Our data show that these residues might form salt bridges with Arg 249 and Lys 250 in p5. Accordingly, it has been demonstrated that the Ras/GAP association is based on interaction between positively charged Arg and Lys, conserved in GAPs, and negatively charged residues in the effector region of Ras (45). It has also been suggested that hydrophobic residues in the effector domain of ARD1 (Phe 429 and Met 430 ) could be involved in the interaction with p5 (22). We postulate that they might interact with Leu 251 and Val 252 , which directly follow the two positively charged amino acids in p5, as expected. GTPase activity always required physical interaction between the two domains of ARD1, and binding of p5 to the effector domain appeared necessary for GTP hydrolysis. The peptide p5, corresponding to the region of interaction of p5 with p3, effectively prevented association of the two proteins and GTP hydrolysis. The peptide also significantly reduced (ϳ40%) the intrinsic GTP hydrolysis by ARD1 and could be useful to assess the importance of the intrinsic GTPase activity of ARD1 in its biological activity.
Despite very little amino acid identity, the minimal GAP domain of ARD1 does exhibit similarities to GAPs characterized previously. Indeed, a zinc finger motif has been identified are tridecapeptides corresponding to the effector regions of ARF1 and p3, respectively. p5p is a dodecapeptide corresponding to amino acids 245-256 of ARD1, and Rp5p is a dodecapeptide with the same residues in random sequence. GTPase activity is expressed as the increase in bound GDP relative to that without peptide (ϭ100%), based on Phos-phorImager quantification. Data are means of values from three experiments performed in duplicate Ϯ one-half the range. Error bars smaller than symbols are not shown.
in the recently cloned mammalian ARF GAP (11) as well as in the yeast ARF GAP Gcs1 (46). Replacement of cysteines that are expected to form a zinc finger structure (33) resulted in a complete loss of GAP activity in ARD1 and in ARF GAP (11). The 139 CX 2 CX 4 CX 2 C 150 motif in the GAP domain of ARD1 also resembles a ferrodoxin signature (iron-sulfur) domain. However, up to 10 M, zinc or iron sulfate had little effect on GTP binding or GTPase activity of ARD1. 2 The exact function of the domain is not known, and the importance of metal binding to zinc finger motifs in ARF GAPs remains to be determined. A clue may be provided by the recent demonstration that Rab GEPs from mammals and yeast, respectively, Mss4 and Dss4, also have a critical zinc binding motif, which may bind to the GTPase at the region that surrounds its effector domain (47). ARD1 also contains a second potential zinc-binding domain ( 31 CX 2 CX 16 CXHX 2 CX 2 CX 12 CX 3 75 ) near the amino terminus, the function of which remains unknown but seems unlikely to involve GAP activity, as its deletion did not affect GTP hydrolysis.
The region sharing partial identity with the second consensus sequence of Rho/Rac GAPs also appeared to play an important role in p5 GAP activity. The crystal structure of p50rhoGAP shows that residues conserved among members of the Rho GAP family, which are confined to one face of the protein, are likely involved in binding to G proteins and enhancing GTPase activity (48). We speculate that Lys 158 , Pro 167 , and Leu 168 may play an equivalent role in the GAP domain of ARD1. Replacement of either of the two arginines, located precisely in this domain, prevented GTP hydrolysis. It is conceivable that they both contribute to catalysis as has been suggested for Arg 789 and Arg 903 in Ras GAP (49) and for Arg 201 in G␣ s (16). The crystal structure of Ras associated with the GAP domain of Ras GAP confirmed that Arg 789 of GAP-334 is positioned in the active site of Ras to neutralize developing charges in the transition state, whereas Arg 903 stabilized the arginine finger motif (50).
Ras GAP contacts the GTP-binding pocket and the effector domain of Ras, a loop that undergoes significant conformational change upon GTP hydrolysis (49,50). In G␣, the helical domain interacts with the GTP-binding pocket, but not with "switch" regions that undergo conformational change during GTP hydrolysis. Hence, an RGS protein could accelerate GTP hydrolysis of G␣ by binding to one or more of the switch elements and/or by introducing additional arginine(s) to the catalytic center. As the GTPase activity of ARD1 is much lower than that of the Ras⅐Ras GAP complex, it is possible that, like heterotrimeric G proteins, ARD1 has an RGS-like protein that stimulates GTP hydrolysis. A recently purified ARF GAP (12), as well as p5 expressed separately in Escherichia coli, however, failed to increase intrinsic GTPase activity of ARD1. 2 Further  TABLE I studies will be required to identify partners of ARD1 involved in its alternation between GDP-and GTP-bound forms and to demonstrate the role of the intrinsic GTPase activity in the intracellular function of ARD1. In Ras and G␣, GTP is hydrolyzed by in-line attack of its ␥ phosphate by a nucleophilic water molecule (35). A glutamine residue (Gln 61 in Ras, Gln 204 in G␣ i1 , and Gln 71 in ARF1) located in the amino terminus of switch II seems to abstract a proton from this attacking water molecule in all G proteins (35). In ARD1, the equivalent residue is Lys 458 , which might also explain the extremely low rate of GTP hydrolysis by the ARF domain (p3) and the relatively modest rate by ARD1 itself relative to that of the Ras⅐GAP complex.
ARFs are possible sites at which phospholipids may function in membrane traffic. The interaction of ARF1 with three different GAPs (9,10,12), and phospholipase D (5, 6) has been shown to be PIP 2 -dependent. Two lipid-binding sites on ARF1 have been identified, and GAP activity depended on occupancy of both sites (9,51). The effect of PIP 2 on nucleotide dissociation from ARF1 has been taken as an evidence of PIP 2 binding to ARF1 (51). Furthermore, the crystal structure of ARF1 has revealed that basic amino acids in positions 10, 15, 16, 59, 178, and 181 form a solvent-exposed patch of positive charges (52,53), which is reminiscent of a pleckstrin-homology domain. Four of these residues were critical for PIP 2 binding (54). PIP 2 also accelerated dissociation of GTP and GDP from p3 or p8(1), but was not required for GTP hydrolysis induced by p5 (14). Interestingly, three of the four positively charged residues that were implicated in PIP 2 -dependent GAP binding (54) are not present in ARD1. Phospholipids, however, are known to affect GTP binding to ARFs (10) as well as to ARD1 (22).
The unusual intrinsic GTPase activity of ARD1 may result from the covalent attachment of a GAP-like domain to the GTPase core of an ARF protein, by exon shuffling during evolution (55). The mechanism by which GAPs accelerate the GTPase reaction of monomeric G proteins has been a matter of considerable debate. Our data seem to favor the arginine finger hypothesis (35,50) in which arginines are expected to stabilize the transition state in GTP hydrolysis. The GAP site of ARD1, between amino acids 101 and 333, can be divided into a region important for physical association with the ARF domain (residues 200 -333) and a domain directly involved in stimulation of GTP hydrolysis (residues 101-200). The latter contains a zinc finger motif reminiscent of one found in ARF GAP and a region that resembles the second consensus sequence in Rho/ Rac⅐GAPs. The function of the amino-terminal 101 residues, as well as that of amino acids 333-387, remains to be determined, whereas the hydrophobic ␣-helical structure (residues 387-402) preceding the ARF domain has been demonstrated to have a GDP dissociation inhibitor-like effect (15). Crystal structures of Ras⅐GAP (36,37,49) and RGS4 (42) as well as those of G␣ subunits (38 -41), have revealed that GAPs, RGS, and GAPlike structures contain exclusively helical secondary structure elements. It will be interesting to learn whether the GAP domain of ARD1 also has that structure. Structural information about ARD1 in GDP-and GTP-bound forms will surely be helpful in understanding the interaction between the GTPbinding and GAP domains, as well as alterations associated with the GDP-GTP transition.