Characterization of a GDP Dissociation Inhibitory Region of ADP-ribosylation Factor Domain Protein ARD1*

ADP-ribosylation factors (ARFs) are ∼20-kDa guanine nucleotide-binding proteins initially identified by their ability to stimulate cholera toxin ADP-ribosyltransferase activity and later recognized as critical components in intracellular vesicular transport and phospholipase D activation. ARF domain protein 1 (ARD1) is a member of the ARF family that differs from other ARFs by the presence of a 46-kDa amino-terminal extension. We previously reported that this extension acts as a GTPase-activating protein for the ARF domain of ARD1 (Vitale, N., Moss, J., and Vaughan, M. (1996)Proc. Natl. Acad. Sci. U. S. A. 93, 1941–1944). Both GTP binding and GTP hydrolysis are necessary for physiological function of guanine nucleotide-binding proteins, and the rates of GDP/GTP exchange and GTPase activity are critical in the activation/deactivation cycle. Dissociation of GDP from the ARF domain of ARD1 was faster than from ARD1 itself (both proteins synthesized in Escherichia coli). Using deletion mutations, it was demonstrated that the 15 amino acids directly preceding the ARF domain were responsible for decreasing the rate of GDP dissociation but not guanosine 5-[γ-thio]triphosphate dissociation. By site-specific mutagenesis it was shown that hydrophobic amino acids in this region were particularly important in stabilizing the GDP-bound form of ARD1. It is suggested that, like the amino-terminal segment of ARF, the equivalent region in ARD1, located between the GTPase-activating protein and ARF domains, may act as a GDP dissociation inhibitor.

Although roles of G proteins 1 are enormously diverse, these GTPases all operate by a fundamentally similar mechanism (1). When GTP occupies its guanine nucleotide-binding site, the G protein can interact with and modify the activity of a downstream target protein. Hydrolysis of GTP causes the dissociation of the G protein-target complex and thus 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 recently reported that the ARF domain of ARD1 binds specifically GDP and GTP, whereas the amino-terminal domain does not (17). Using recombinant proteins, it was shown that the amino-terminal p5 domain of ARD1 stimulates hydrolysis of GTP bound to the ARF domain p3 and consequently appears to be the GAP component of this bifunctional protein (18). The stimulatory effect of the p5 domain on the GTPase activity of p3 was specific, because GTP hydrolysis by other members of the ARF family was not increased (16). The presence of an intrinsic GAP domain is an apparently unique phenomenon for monomeric guanine nucleotide-binding proteins, because GTPase activity of other members of the Ras superfamily is increased by a separate GAP molecule, which interacts with the effector region of the protein (17). Using chimeric proteins, we demonstrated that the GAP domain of ARD1 similarly interacts with the effector region of the ARF domain of ARD1 and thereby stimulates GTP hydrolysis (17).
Nucleotide hydrolysis and product dissociation are both regulated steps in the GTPase cycle. Because their rates determine when and for how long the G protein is active, it is important to precisely understand these mechanisms. We reported that GDP␤S dissociation from p3 was faster than from ARD1 (18). Accordingly, in the presence of certain phospholipids (i.e. brain phosphatidylcholine plus phosphatidylinositol bisphosphate plus phosphatidylethanolamine or phosphatidylserine) more GTP was bound to p3 than to ARD1 (17). These results suggested that p5 may influence GDP dissociation as well as GTP hydrolysis. We report here that 15 amino acids from the carboxyl-terminal domain of p5 positioned before the ARF domain p3 inhibit GDP release but not GTP␥S release. Results of site-directed mutagenesis suggest that hydrophobic amino acids within this region are critical in reducing the rate of GDP dissociation. This region therefore appears to have a function similar to that of the amino-terminal segment of ARF proteins.

EXPERIMENTAL PROCEDURES
Materials-Bovine thrombin was purchased from Sigma. TLC plates were purchased from VWR Scientific, and glutathione-Sepharose beads * 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  were from Pharmacia Biotech Inc. PCR reagents and restriction enzymes, unless otherwise indicated, were from Boehringer Mannheim. The sources of other materials have been published (7,(17)(18)(19).
Preparation of Recombinant Fusion Proteins (p3, p5, and p8)-For large scale production of fusion proteins (18), 10 ml of overnight culture of the 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 (final concentration, 0.5 mM). After incubation for an additional 3 h, bacteria were collected by centrifugation (Sorvall GSA, 6000 rpm, 4°C, 10 min) and stored at Ϫ20°C. Bacterial pellets were dispersed in 10 ml of cold phosphate-buffered saline (pH 7.4) with 20 g/ml trypsin inhibitor, 5 g/ml each leupeptin and aprotinin, 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, 16000 rpm, 4°C, 20 min). Fusion proteins purified on glutathione-Sepharose were ϳ90% pure as estimated by silver staining after SDS-polyacrylamide gel electrophoresis (18). After cleavage by bovine thrombin, glutathione S-transferase was removed with glutathione-Sepharose beads and thrombin with benzamidine-Sepharose 6B (21). Proteins were further purified 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-polyacrylamide gel electrophoresis was Ͼ98%. Amounts of purified proteins were estimated by a dye binding assay (22) and by SDS-polyacrylamide gel electrophoresis using bovine serum albumin as standard.
Construction and Expression of Mutated Forms of N⌬387ARD1-Two mutated fragments of N⌬387p8 with deletions at the 5Ј end were prepared by PCR using pGEX5G/LIC/N⌬387p8 as template and the following forward primers 5Ј-AGCATCCCTGTCACTCCGCGGATG-GATAATCGAGTTCAC-3Ј and 5Ј-ACTTTTACAAAGGATAATCCGCG-GATGATTGGACCAAAAATG-3Ј (differences from the original clone are underlined) and the reverse primer 5Ј-GAATTCCCGGGGATC-CAACTGCG-3Ј (italicized sequence is a BamHI restriction site). The forward primers introduced a SacII restriction site (italicized sequences) and an initiation codon in-frame (bold sequences) in the deleted fragments N⌬393p8 and N⌬398p8. The PCR fragments were cloned in-frame in SacII-and BamHI-digested pGEX5G/LIC expression vector. Ultra competent cells (Stratagene) were transformed with plasmids pGEX5G/LIC/N⌬393p8 or pGEX5G/LIC/N⌬398p8. Sequences of mutated N⌬387 fragments were confirmed by automated sequencing (Applied Biosystems, 373 DNA Sequencer) using the primers 5Ј-TTAT-ACGACTCACTATAGGG-3Ј, 5Ј-ATGATTGTAGAGTTGTCTT-3Ј, and 5Ј-GCTAGTTATTGCTCAGCGG-3Ј. Fusion proteins were expressed and purified as described for ARD1 proteins.

Release of [ 35 S]GDP␤S or [ 3 H]GDP from Recombinant
Proteins-Samples (300 pmol) were incubated for 30 min at 30°C in 450 l of 20 mM Tris (pH 8.0), 10 mM dithiothreitol, 2.5 mM EDTA, with 0.3 mg/ml bovine serum albumin and 1 mg/ml cardiolipin and then incubated for 40 min in the same medium plus 10 mM MgCl 2 and 3 M of [ 35 S]GDP␤S (2 ϫ 10 7 cpm; total volume, 500 l) or [ 3 H]GDP (1.5 ϫ 10 7 cpm; total volume, 500 l) before the addition of p5 or deleted/mutated-forms (or water) in 100 l of 10 mM Tris (pH 8.0), 2.5 mM EDTA. After incubation for 5 min at 30°C, samples (60 l) were transferred to nitrocellulose filters, which were washed five times with 1 ml of 25 mM Tris (pH 8.0)/5 mM MgCl 2 /100 mM NaCl before radioassay in a liquid scintillation counter (19). To calculate zero time values for dissociation curves, radioactivity bound to filters in the absence of protein was subtracted from the total with proteins. The remainders of the mixtures were immediately diluted with an equal volume of reaction buffer containing 2 mM of GDP␤S or GDP. Samples (120 l) were taken after 5, 15, 30, 45, 60, 75, 90, 105, and 120 min at 30°C for quantification of bound radioactivity, as described for the zero time samples. Release of [ 35 S]GTP␥S was measured as described previously (18). Data are presented as the means Ϯ S.E. of values from triplicate determinations. Errors bars smaller than the symbols are not shown.
GTP␥S Binding Assay-GTP␥S binding to purified recombinant ARD proteins was assessed using a rapid filtration technique as described previously (17).

RESULTS AND DISCUSSION
Although there was no significant difference between p8 (ARD1) and p3 (ARF domain of ARD1) in the amount of [␤-35 S]GDP bound at zero time (Fig. 1B), [␤-35 S]GDP release (in the presence of large excess of [␤S]GDP) was much slower from p8 than from p3 as seen earlier (18), consistent with the notion that the ARD1 amino-terminal extension p5 may act to inhibit nucleotide release. To define more precisely the region of p5 involved, we generated four mutants of ARD1 with aminoterminal deletions (Fig. 1A). GDP[␤-35 S] dissociation from the four mutants (N⌬88p8, N⌬200p8, N⌬304p8, N⌬387p8) approximated that of the intact p8 (Fig. 1B). The shortest deletion mutant protein N⌬387p8, which is equivalent to p3 with 15 amino acids from ARD1 added at the amino terminus, retained the dissociation behavior of ARD1. It was concluded that these 15 residues are apparently involved in stabilizing the inactive GDP form of the ARF domain of ARD1.
When synthesized separately in Escherichia coli and incubated together, the non-ARF domain of ARD1 p5 reduced the rate of GDP␤S release from the ARF domain p3 (Fig. 2B) to approximately that of ARD1 (18). Physical interaction between p5 and p3, expressed separately in E. coli, had also been demonstrated by gel filtration and using glutathione S-transferase fusion proteins (18). In agreement with the hypothesis that the 15 carboxyl-terminal amino acids of p5 act as an inhibitor of GDP␤S dissociation from p3, three carboxyl-terminal deletion mutants of p5 ( Fig. 2A) had no effect on GDP release from p3 (Fig. 2B). Surprisingly, however, three amino-terminal deletion mutants of p5 ( Fig. 2A) were also ineffective in reducing GDP␤S dissociation from p3. In large excess (up to 10 times p3), N⌬88p5 or N⌬200p5 had no effect on GDP␤S release from the ARF domain of ARD1 (data not shown). There are a number of possible reasons for this apparent discrepancy. A domain involved in interaction with p3 may be located in the amino terminus of p5. Incorrect folding of the deleted mutants may also be an explanation. We favor yet another explanation in which the functional interaction between p3 and the part of p5 that slows nucleotide dissociation requires very precise alignment of the two domains to allow the latter to stabilize the GDP form of p3. Consistent with this possibility is the finding that the deletion mutants p5N⌬88 and p5N⌬200 can physically interact with p3 and also that addition of N⌬200p5 and C⌬191p5 together with p3 did not reproduce the effect of p5 on GDP␤S release from p3 (data not shown).
Earlier reports suggested a role for the amino terminus of ARF in regulating nucleotide dissociation (24 -26), based on analysis of nucleotide binding and release from amino-terminal deletion mutants of ARF1. We suggest that the carboxyl terminus of p5 corresponding to the amino terminus of ARF may also play this role in ARD1. In favor of this hypothesis are the facts that p5 increased GDP␤S release from a chimeric ARF1 protein with an amino-terminal deletion (N⌬15ARF1(39 -45p3)), which possesses the binding site for p5, but not from the chimeric protein with the amino terminus intact (ARF1(39 -45p3)) (17), suggesting that the carboxyl end of p5 and the amino terminus of ARF may have analogous locations in the molecules and therefore perform analogous functions in regulating nucleotide dissociation.
GDP␤S release from N⌬393p8 was significantly faster than that from ARD1 or N⌬387p8 but also significantly slower than that from p3 (Fig. 3). These results suggested that amino acids 1-6 of the 15 carboxyl-terminal residues of p5 participate in stabilizing the GDP␤S-bound form of the ARF domain of ARD1 but also that residues 7-15 have a similar function. Accordingly, removal of 11 residues dramatically affected GDP␤S release, because rates of dissociation from N⌬398p8 and p3 were similar (Fig. 3).
GDP␤S or GDP dissociated faster from p3 than from p8 or N⌬387p8 (Figs. 1B, 3, and 4). However, GTP␥S dissociated slightly faster from p8 or N⌬387p8 than from p3 (Fig. 4). In fact, dissociation curves from N⌬387p8 for both GDP and GTP␥S approximated those from p8 or ARD1. These results suggested that the 15 residues preceding p3 may specifically inhibit GDP dissociation and slightly increase GTP␥S dissociation. Note that identical results were obtained with [ 3 H]GDP or [ 35 S]GDP␤S (Figs. 1, 3, and 4).
Partial deletion of amino acid sequence can have subtle but adverse effects on overall protein structure. Therefore, to reduce the possibility that the observed differences between N⌬387p8, N⌬393p8, and N⌬398p8 in GDP release could arise from slight structural changes that could perturb the threedimensional organization of the ARF domain, mutant proteins were constructed with single amino acid replacements that should cause minimal disturbance of global protein structure. Specifically, hydrophobic amino acids between residues 1 and 15 were individually changed to arginine, asparagine was replaced by leucine, proline was replaced by glycine, and lysine was replaced by alanine (Table I).
Whereas the mutation (V2R) had no effect on GDP␤S release, GDP␤S dissociated significantly faster from N⌬387-(F4R)p8 than from N⌬387p8 (Fig. 5A). In fact, GDP␤S dissociation from N⌬387(F4R)p8 approximated that from N⌬393p8 (Figs. 3 and 5A), suggesting that the accelerated GDP␤S release from N⌬393p8 (relative to N⌬387p8) might be due to removal of Phe 4 . Replacement of two other hydrophobic residues, Val 10 and Ile 12 , significantly increased the rate of GDP␤S dissociation, whereas replacement of Asn 8 with the hydrophobic L (present in ARF1; Table I)

TABLE I
Effect of site-specific mutation of N⌬387p8 on GDP␤S release rate ARF1, N⌬387p8, and site-specific mutant amino-terminal sequences are aligned. Mutations from the original ARD1 sequence are indicated in bold, and ␣-helical sequences are underlined. ␣-Helical structure for the amino terminus of ND387p8 and mutants was estimated by the method of Garnier et al. (20). The sequence adopting an ␣-helical conformation in ARF1 crystal is underlined (27,29). Average hydrophobicity was calculated using an helical wheel representation with the software PC/GENE. Binding in the presence of cardiolipin (1 mg/ml) was determined as described in the legend to Fig. 5. The dissociation rates (in presence of large excess of GDP␤S), k off (min Ϫ1 ) were determined by fitting the data from In parentheses, the rate relative to that of N⌬387p8 ϭ 100 is indicated. b ARF1 GDP␤S dissociation rate in the presence of cardiolipin appears slightly lower than that reported for GDP (25) in the presence of dimyristoyl-phosphatidylcholine/cholate (0.020 versus 0.027). (Fig. 5B). However, substitution of Asn 8 by Leu, as well as that of Ile 12 by Arg, did not modify the GTP␥S dissociation rate (data not shown). The mutation K15A had no effect on GDP␤S release (Fig. 5B). Interestingly, the mutant N⌬387(P14G) also exhibited an increased rate of GDP␤S dissociation, suggesting that this proline may contribute to correct positioning of hydrophobic residues. These results, therefore, are consistent with the view that hydrophobic residues in this region of ARD1, as in the amino terminus of ARF, stabilize the GDP-bound form of the ARF domain, probably by interacting with the hydrophobic core in the tertiary structure of ARF (27). Despite significant differences between the corresponding sequences in ARD1 and ARF1, hydrophobic residues are numerous in both (Table  I). Other ARF proteins share the same characteristic ␣-helical amino-terminal extension. It is notable that the ARF6 aminoterminal region is four amino acids shorter than ARF1, ARF2, and ARF3 but does contain the critical hydrophobic residues. It would be interesting to compare the rates of GDP release among different ARFs to understand more precisely how this domain stabilizes the GDP-bound form of the ARF motif.
We used functional assays to monitor conformational integrity of the deletion and site-specific mutants. 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, N⌬387p8, and single site-specific mutants of N⌬387p8, N⌬393p8, N⌬398p8, and p3 were observed (Table II), suggesting no difference in folding of the guanine nucleotide-binding site of these recombinant proteins. Activation of cholera toxin-catalyzed ADP-ribosylagmatine formation by ARFs requires binding of GTP followed by interaction with the bacterial toxin. The site of interaction with cholera toxin has been localized in the carboxyl-terminal region of ARF (24,26,28), and the ability of ARD1 mutants to activate CTA should represent a good indicator of conformational integrity. Removal of 387 residues from the amino terminus of ARD1 reduced CTA activation by about 28% (Table II), suggesting that the large amino-terminal extension contributes to maintain the native conformation of the protein. Accordingly, further deletion of the amino terminus reduced the ability of N⌬393p8, N⌬398p8, and p3 to activate CTA by ϳ 39, 62, and 63%, respectively (Table II). Because GTP␥S binding was not affected (Table II), we assume that the amino-terminal extension of ARD1, specifically the segment that influences nucleotide release, is important for correct folding of the region of the protein that interacts with cholera toxin.
In the crystal structure of ARF1, Phe 5 , Leu 8 , and Phe 9 reside in the hydrophobic core of the molecule (27,29). We suggest that residues Phe 391 , Val 397 , and Ile 399 in ARD1 play the same role as Phe 5 , Leu 8 , and Phe 9 in ARF1. Because single mutation of Phe 4 , Val 10 , or Ile 12 in N⌬387p8 significantly affected GDP␤S dissociation (Figs. 5, A and B), we generated a triple mutant N⌬387(F4R,V10R,I12R)p8 in which critical hydrophobic amino acids were replaced by arginines. Whereas GTP␥S dissociation rate was not modified (data not shown), the rate of GDP␤S release was faster than those from single-mutants and even faster than that from p3 ( Fig. 5C and Table I). At zero time, the triple mutant protein also had ϳ38% more GDP bound than did N⌬387p8; GDP[␤-35 S] was bound to 3.93 Ϯ 0.04% of the former and 2.84 Ϯ 0.02% of the latter (Fig. 5C). Accordingly, we found that N⌬387(F4R,V10R,I12R)p8 bound more GTP␥S (ϳ35%) and was a better activator (ϳ30%) of cholera toxin-catalyzed ADP-ribosylation than was N⌬387p8 (Table II).
Lower hydrophobicity of the amino-terminal sequences of N⌬387p8 was apparently correlated with a greater rate of GDP release (k off , Table I). We cannot, however, completely rule out the possibility that disruption of the ␣-helical structure (Table  I) also affected GDP release from ARD1, although the mutant protein N⌬387p8(V2R), which had virtually no amino-terminal ␣-helical structure, differed very little from N⌬387p8 in rate of GDP release (Table I). These results suggest that hydrophobic amino acids in this region are crucial to stabilize the GDPbound form of the ARF domain of ARD1, because replacement by the positively charged arginine increased GDP␤S dissociation. The most important difference between ARFs and the other ϳ20-kDa GTPases is precisely the presence of an aminoterminal ␣-helix (␣A) and the connecting loop (L1). Both of these sequences have no counterpart in Ras family members (30) but do appear in the sequences of G protein ␣-subunits (31). In accord with our findings is the report that charged residues constitute the site that provides the signal directly from rhodopsin to transducin to activate GDP release (32).
The amino-terminal segment is critical to ARF1 functions both in vivo and in vitro (reviewed in Ref. 3). Therefore, involvement of the amino-terminal ␣-helix of ARF1 in membrane affinity, nucleotide-induced conformational change, and effector interaction has been suggested (33,34). Our results are consistent with the view that an equivalent region may play the same role in ARD1. Thus, as in ARF1, residues 389 -402 in  S]GTP␥S binding to the indicated protein (70 pmol) was assessed by the rapid filtration technique in the presence of 3 M GTP␥S and of cardiolipin (1 mg/ml). In the absence of cardiolipin, GTP␥S binding was very low (ϳ0.040 pmol Ϯ 0.008) and not significantly different among the proteins. Binding in the absence of protein has been subtracted. After the protein (70 pmol) was incubated with 100 M GTP␥S and cardiolipin (1 mg/ml), ARF stimulation of cholera toxin-catalyzed ADPribosylagmatine formation was assayed for 60 min at 30°C. ARF activity is the difference between cholera toxin-catalyzed formation of [ 14 C]ADP-ribosylagmatine without and with ARD protein (nmol/h). In the absence of cardiolipin, ARF activity was less then 0.005 nmol/h. ARD1 appear to function as a GDP dissociation inhibitory region. In addition, in ARD1, there is a proline in position 388 that may demarcate this region from the rest of p5 and perhaps contribute to the functional alignment with p3.
To our knowledge, no ARF guanine nucleotide dissociation inhibitor proteins have been reported. Although this could be due to technical failure, we might propose that the aminoterminal segment of ARF and the equivalent region in ARD1 can serve as intrinsic modulators of guanine nucleotide dissociation and thus association. Further study of these regions in the association of ARD1 and ARF with vesicular membranes should reveal whether they affect this property, as well as nucleotide dissociation, and act thereby as intrinsic guanine nucleotide dissociation inhibitor components. Such guanine nucleotide dissociation inhibitor regions may represent sites of interaction for ARF GEPs. Indeed, in ARF proteins and in ARD1, the guanine nucleotide dissociation inhibitor site is analogous to a site in G␣ subunits of heterotrimeric G-proteins (31). It is thus tempting to speculate that ARF GEPs increase GDP dissociation by a mechanism similar to that of the receptors with seven membrane-spanning helices. Positively charged amino acids in the third intracellular loop are directly responsible for acceleration by these receptors of guanine nucleotide exchange on heterotrimeric G proteins (35). We are currently investigating the possibility that the amino termini of the ARFs are involved in the activity of ARF GEPs.