Role of Coatomer and Phospholipids in GTPase-activating Protein-dependent Hydrolysis of GTP by ADP-ribosylation Factor-1*

The binding of the coat protein complex, coatomer, to the Golgi is mediated by the small GTPase ADP-ribosylation factor-1 (ARF1), whereas the dissociation of coatomer, requires GTP hydrolysis on ARF1, which depends on a GTPase-activating protein (GAP). Recent studies demonstrate that when GAP activity is assayed in a membrane-free environment by employing an amino-terminal truncation mutant of ARF1 (Δ17-ARF1) and a catalytic fragment of the ARF GTPase-activating protein GAP1, GTP hydrolysis is strongly stimulated by coatomer (Goldberg, J., (1999) Cell 96, 893–902). In this study, we investigated the role of coatomer in GTP hydrolysis on ARF1 both in solution and in a phospholipid environment. When GTP hydrolysis was assayed in solution using Δ17-ARF1, coatomer stimulated hydrolysis in the presence of the full-length GAP1 as well as with a Saccharomyces cerevisiae ARF GAP (Gcs1) but had no effect on hydrolysis in the presence of the phosphoinositide dependent GAP, ASAP1. Using wild-type myristoylated ARF1 loaded with GTP in the presence of phospholipid vesicles, GAP1 by itself stimulated GTP hydrolysis efficiently, and coatomer had no additional effect. Disruption of the phospholipid vesicles with detergent resulted in reduced GAP1 activity that was stimulated by coatomer, a pattern that resembled Δ17-ARF1 activity. Our findings suggest that in the biological membrane, the proximity between ARF1 and its GAP, which results from mutual binding to membrane phospholipids, may be sufficient for stimulation of ARF1 GTPase activity.

The binding of the coat protein complex, coatomer, to the Golgi is mediated by the small GTPase ADP-ribosylation factor-1 (ARF1), whereas the dissociation of coatomer, requires GTP hydrolysis on ARF1, which depends on a GTPase-activating protein (GAP). Recent studies demonstrate that when GAP activity is assayed in a membrane-free environment by employing an amino-terminal truncation mutant of ARF1 (⌬17-ARF1) and a catalytic fragment of the ARF GTPase-activating protein GAP1, GTP hydrolysis is strongly stimulated by coatomer (Goldberg, J., (1999) Cell 96, 893-902). In this study, we investigated the role of coatomer in GTP hydrolysis on ARF1 both in solution and in a phospholipid environment. When GTP hydrolysis was assayed in solution using ⌬17-ARF1, coatomer stimulated hydrolysis in the presence of the full-length GAP1 as well as with a Saccharomyces cerevisiae ARF GAP (Gcs1) but had no effect on hydrolysis in the presence of the phosphoinositide dependent GAP, ASAP1. Using wild-type myristoylated ARF1 loaded with GTP in the presence of phospholipid vesicles, GAP1 by itself stimulated GTP hydrolysis efficiently, and coatomer had no additional effect. Disruption of the phospholipid vesicles with detergent resulted in reduced GAP1 activity that was stimulated by coatomer, a pattern that resembled ⌬17-ARF1 activity. Our findings suggest that in the biological membrane, the proximity between ARF1 and its GAP, which results from mutual binding to membrane phospholipids, may be sufficient for stimulation of ARF1 GTPase activity. ARF 1 GTPases play a key role in the regulation of vesicular trafficking of proteins among different compartments of the eukaryotic cell. In the early secretory system, the ARF1 protein regulates the interaction of the coatomer coat complex with Golgi membranes (1,2). In the active GTP-bound form, ARF1 triggers the recruitment of coatomer (3)(4)(5) apparently by direct interaction with its ␤and ␥-subunits (6). The subsequent dissociation of coatomer depends on GTP hydrolysis on ARF1 (7,8). The cycles of GTP binding and hydrolysis on ARF1 are controlled by two sets of cytosolic regulatory proteins. Activa-tion of ARF1 is brought about by guanine nucleotide exchange proteins (9 -16) whereas GTP hydrolysis depends on GTPaseactivating proteins (GAPs). ARF GAPs are a family of proteins sharing a catalytic domain of 120 -140 amino acids that includes a Cys 4 -zinc finger motif. The first ARF GAP to be discovered (GAP1) is a 45-kDa protein that distributes between the cytosol and Golgi complex and functions in the regulation of membrane traffic through this organelle (17)(18)(19). Saccharomyces cerevisiae contains two proteins (Gcs1 and Glo3) that show high similarity to GAP1 and possess ARF GAP activity (20,21). The two yeast GAPs form an essential pair with a redundant function in the endoplasmic reticulum-Golgi shuttle. Recently, additional ARF GAPs belonging to two subfamilies were identified in mammalian cells. GIT1 is a 95-kDa protein from rat that interacts with GRK2 and regulates ␤ 2 -adrenergic receptor internalization (22) while its mouse homologues (Cat1/2) are involved in CDC42/Rac/Pak signaling (23). The second subfamily consists of large multidomain proteins represented by ASAP1 and PAP (24 -26). These proteins possess a pleckstrin homology domain and show phosphoinositide-dependent ARF GAP activity. Additionally, the proteins interact with nonreceptor tyrosine kinases through a proline-rich Src homology 3 (SH3)-binding domain. Despite distinct subcellular sites of action, all GAPs that have been described so far are highly active on ARF1 in vitro. Whether there is redundancy in the action of mammalian ARF GAPs in vivo remains to be established.
Recently, Goldberg (27) presented the crystal structure of a 130-amino acid catalytic fragment of GAP1 co-crystallized with the GDP-bound form of an ARF1 mutant lacking the first 17 amino acids (⌬17-ARF1). A unique feature of this structure is that switch I of ARF1, thought to comprise part of the "effector" site mediating coatomer interaction (6, 28), does not participate in GAP binding. This model was supported by biochemical data showing that coatomer dramatically stimulates GTP hydrolysis on ⌬17-ARF1, which suggests that both coatomer and GAP can simultaneously bind to ARF1. Interestingly, coatomer-dependent stimulation of GAP1 activity was inhibited by a coatomer-interacting peptide of one member of the p24 transmembrane Golgi proteins (29). Based on structural and functional observations, Goldberg suggested (27) that the "catalytic arginine finger" mechanism that is essential for catalysis of GTP hydrolysis by Ras and Rho GAPs may not operate in ARF GAP and that coatomer rather than GAP may contribute a catalytic residue for the GTPase reaction. The model of Goldberg was recently challenged by Mandiyan et al. (30) who reported that the crystal structure of a different ARF GAP, PAP␤, may not be compatible with the structure of the ARF1⅐GAP1 complex described by Goldberg (27).
Our laboratory in addition to others (17, 18, 20 -22, 24) has previously noted that ARF GAPs may display high activity in the absence of coatomer when the natural form of ARF1 (con-taining a myristoyl residue at its amino terminus) is employed. Unlike ⌬17-ARF1, which does not depend on phospholipids for GTP binding, the binding of the nucleotide to full-length ARF1 requires phospholipids. These act by interacting with and stabilizing the amphipathic amino-terminal peptide that becomes solvent-exposed in the GTP state (31)(32)(33). Consequently, GTP hydrolysis on myristoylated ARF1 has been assayed in the presence of phospholipid vesicles whose lipid composition may influence GAP activity (34). It was therefore of interest to investigate the relative contribution of coatomer and phospholipids to GAP activity using normal myristoylated ARF1 as substrate. We report that contrary to its effect on ⌬17-ARF1, coatomer does not affect GAP activity with myristoylated ARF1 bound to phospholipid vesicles. This as well as additional findings indicate that coatomer is not directly involved in the catalysis of GTP hydrolysis. We propose that both coatomer and phospholipids may facilitate GTP hydrolysis by bringing GAP into proximity with its substrate ARF1.
Expression and Purification of ARF GAP Mutants-GAP 1 mutants were generated by polymerase chain reaction and cloned into the pKM260 T7 polymerase-driven bacterial expression vector as described previously (18). All mutant GAPs were derived from constructs encoding the first 257 amino acids because this part of the protein in wildtype GAP1 retains full GAP catalytic activity, and moreover, longer constructs cannot be expressed in E. coli (18). Proteins were expressed in DE3 lysogens of strain BL21 by induction for 2.5 h at 37°C in the presence of 0.4 mM isopropyl-␤-D-thiogalactopyranoside. Bacterial pellets were extracted with 6 M guanidine hydrochloride, and proteins were purified by Ni 2ϩ -nitrilotriacetic acid chromatography according to the manufacturer's instructions using 6 M guanidine hydrochloride in 0.1 M sodium phosphate, 10 mM Tris, pH 8.0, throughout the purification. The Ni 2ϩ -nitrilotriacetic acid eluate was diluted with 6 M guanidine hydrochloride to a protein concentration of 3-5 mg/ml and was supplemented with 5 mM dithiothreitol. The eluate was dialyzed overnight against 50 mM NaCl, 25 mM Tris, pH 7.4, 1 mM dithiothreitol with one buffer change. The dialysate was cleared by centrifugation, and proteins were purified by Resource-Q anion exchange chromatography using a linear NaCl gradient in 25 mM Tris, pH 7.4, 1 mM dithiothreitol.
ARF GAP Assay-GTP hydrolysis on myristoylated ARF1 was assayed essentially as described previously (17). ARF1 was first loaded with [␣-32 P]GTP in the presence of phospholipid vesicles (0.4 m) containing 40% phosphatidylcholine, 30% phosphatidylethanolamine, and 30% phosphatidylserine prepared as described in Ref. 34 (17). Loading proceeded for 15 min at 30°C and was terminated by the addition of 2 mM MgCl 2 . Loading efficiency with respect to [␣-32 P]GTP was typically 60 -75%. GAP assays contained 40 nM [␣-32 P]GTP-loaded ARF1 (approximately 1:10 dilution of the loading reaction), 5 mM MgCl 2 , 25 mM MOPS, pH 7.4, 1 mM dithiothreitol, and 1 mM ATP with an ATP-regenerating system (5 mM phosphocreatine and 50 g/ml creatine phosphokinase) in a final volume of 10 l. Reactions were preincubated for 5 min at room temperature with or without coatomer and were initiated by the addition of different concentrations of GAP. Reactions proceeded for 15 min and were terminated by boiling for 30 s. GTP hydrolysis was determined by thin layer chromatography on polyethyleneimine cellulose (17), and data are pre-sented as the percentage of ARF-bound GTP that was converted to GDP.
Hydrolysis of GTP on ⌬17-ARF1 was assayed by a modification of the assay described by Goldberg (27). ⌬17-ARF1 was loaded with [␥-32 P]GTP in the presence of the guanine nucleotide exchange protein ARNO as described previously (27). GAP assays contained 400 nM GTP-bound ⌬17-ARF1, 5 mM MgCl 2 , 25 mM MOPS, pH 7.4, and 0.5 mM 5Ј-adenylyl-␤,␥-imidodiphosphate (AMP-PNP) in a final volume of 20 l. Preincubation with coatomer and subsequent incubation with GAP were carried out as described above, and reactions were terminated by the addition of 0.5 ml of cold charcoal suspension (5% charcoal in 50 mM NaH 2 PO 4 ). Following centrifugation, the amount of 32 P i in the supernatant was determined by scintillation counting. In both assays described above, the addition of coatomer without GAP had little effect (less than 5% of GTP hydrolyzed). Coatomer background values were subtracted from GAP activity in the presence of coatomer in all experiments except for those shown in Fig. 4. Experiments were repeated at least 3 times, and a representative experiment is presented.

RESULTS
Coatomer Differentially Affects GTP Hydrolysis on ⌬17-ARF Mediated by Different ARF GAPs-Goldberg (27) recently reported that coatomer stimulates GTP hydrolysis by up to 1000fold on a lipid-independent ARF1 mutant (⌬17-ARF1) in the presence of a fragment of GAP1 containing amino acids 6 -136. We investigated whether coatomer stimulation is restricted to GAP1 or can take place with additional members of the ARF GAP family. We employed the full-length GAP1 protein that was generated using a baculovirus expression system, S. cerevisiae Gcs1, and a catalytically active fragment of mammalian ASAP1. As shown in Fig. 1, coatomer stimulated GTP hydrolysis mediated by both GAP1 and Gcs1 on ⌬17-ARF1. Under the conditions employed (0.2 M coatomer, 0.5 M ⌬17-ARF1), the stimulation of GTP hydrolysis by coatomer (about 10-fold) was considerably lower than that reported by Goldberg (27). This difference was attributed to a large extent to the higher rate of GTP hydrolysis in the presence of GAPs alone in our experiments. The higher rate may have been caused by the fact that GAP1 and Gcs1 were employed as full-length proteins in contrast to the truncated GAP1 used by Goldberg, which may have only partial activity in the absence of coatomer (18).
We also tested the effect of coatomer with ASAP1, an ARF GAP that is distinct from GAP1 in structural features and cellular localization (plasma membrane for ASAP1 versus Golgi for GAP1 (18, 38)). Additionally, ASAP1 has a pleckstrin ho- mology domain, and its activity is stimulated by phosphoinositides that appear to act through an allosteric mechanism (39). As shown in Fig. 2A, using ⌬17-ARF1 as substrate, ASAP1 GAP activity was strongly stimulated by phosphoinositides. This is in agreement with recent findings by Kam et al. (39) who employed ⌬13-ARF1 as substrate. By contrast, coatomer had little effect on ASAP1 activity either in the absence or presence of phosphoinositides. A reciprocal pattern was observed with GAP1-(1-257) where coatomer but not phosphoinositides stimulated GTP hydrolysis on ⌬17-ARF1. Similar results were obtained with full-length GAP1/Gcs1 or in the presence of phosphatidylinositol 4Ј,5Ј-bisphosphate instead of the phosphoinositide mixture (data not shown).
Coatomer Does Not Stimulate GAP1-dependent Hydrolysis of GTP on Myristoylated ARF1-The experiments described above were carried out with ⌬17-ARF1, an ARF1 mutant that does not depend on lipids for GTP binding (40) and has therefore been employed to study the interaction of ARF1 with proteins and drugs in solution (27,32,40,41). However, the amino terminus of wild-type ARF1 with its attached myristate residue serves to anchor GTP-bound ARF1 to phospholipid bilayers, and it is this form of ARF1 that is likely to encounter GAP in the biological membrane. It was therefore of interest to test the effect of coatomer on GAP1-dependent hydrolysis of GTP on myristoylated ARF1. In the experiments presented in Figs. 3 and 4, myristoylated ARF1 was loaded with GTP in the presence of unilamellar liposomes containing a phospholipid mixture that is typical of biological membranes (40% phosphatidylcholine, 30% phosphatidylethanolamine, and 30% phosphatidylserine). In the absence of coatomer, GAP1 stimulated GTP hydrolysis on myristoylated ARF1 at concentrations that were lower by 2 orders of magnitude than those required to stimulate hydrolysis on ⌬17-ARF1 (compare Figs. 1 and 3). When GAP activity was assayed on myristoylated ARF1 over a broad range of GAP1 concentrations, there was no significant difference between the activities in the presence or absence of coatomer (Fig. 3). Moreover, coatomer had no effect on GTP hydrolysis on myristoylated ARF1 even at coatomer concentrations that were supra-optimal in the ⌬17-ARF1 assay (Fig. 4).
The difference in coatomer sensitivity between ⌬17-ARF1 and myristoylated ARF1 may be attributed to differences in conformation between the native and truncated proteins or to the presence of phospholipids in assays employing myristoylated ARF1. These phospholipids, which are required for GTP binding to myristoylated ARF1, may obscure the coatomer effect by promoting an efficient interaction between ARF1 and GAP. To distinguish between the above possibilities, we tested the effect of coatomer on myristoylated ARF1 in the presence of mixed detergent/phospholipid micelles. In the experiment presented in Fig. 5, ARF1 was loaded with GTP in the presence of DMPC and cholate, which were added at approximately equimolar concentrations. Under these conditions, coatomer caused a small but reproducible stimulation of GAP1-dependent GTP hydrolysis on myristoylated ARF1 (Fig. 5). When the ratio of detergent to phospholipid was increased with the ad- dition of a 4-fold excess of detergent to the assay (0.4% CHAPS), there was a dramatic decrease in GAP1 activity whereas the addition of coatomer caused an approximately 2-fold stimulation. Thus, under conditions where membrane structure is perturbed, GTP hydrolysis on myristoylated ARF1 can become responsive to coatomer.
GAP1 Contains an Essential Arginine Residue-The mechanism by which GAPs stimulate GTP hydrolysis on Ras and Rho GTPases was shown to involve an arginine residue in GAP that inserts into the GTP-binding pocket of the GTPase and assists in catalysis (42). Recently, structural and functional studies have suggested that in the case of ARF, coatomer rather than GAP might contribute the catalytic arginine (27) although another study (30) has brought this conclusion into question. All ARF GAPs described so far contain one invariant arginine residue in a position equivalent to Arg-50 in GAP1 (30,38). Replacement of Arg-50 of GAP1 with alanine, as well as conservative replacements with lysine or glutamine, completely abolished GAP activity on myristoylated ARF1 even at very high mutant concentrations (Fig. 6A). Similar findings were recently reported for PAP␤ and ASAP1 ARF GAPs (30,38). When activity was assayed using ⌬17-ARF1, the Arg-50 mutants were completely inactive in the absence or presence of coatomer (data not shown). These findings highlight the importance of the conserved Arg residue of ARF GAPs.
The Arg-50 mutants were highly soluble, had normal zinc content, and migrated in a Resource Q column similar to the wild-type protein, suggesting that the mutations do not adversely affect their conformation. It was thus of interest to test whether the mutants can compete with wild-type GAP1 for coatomer-dependent hydrolysis of GTP on ARF1. As shown in Fig. 6B, the R50K mutant had no effect on this reaction at mutant concentrations of up to 10-fold higher than the wildtype GAP concentration. The mutant also failed to inhibit GAP1 activity on myristoylated ARF1 (data not shown). A possible interpretation of these findings is discussed below.

DISCUSSION
The findings presented in this paper provide further insight into the role of coatomer in GTP hydrolysis on ARF1. We show that the effect of coatomer depends on both the form of ARF1 employed in the assay and on the phospholipid environment. In confirmation of the findings by Goldberg (27), we observed strong stimulation of GTP hydrolysis on the lipid-independent truncated ARF1 lacking the first 17 amino acids in the presence of both full-length mammalian GAP1 and yeast Gcs1. By contrast, GAP-dependent GTP hydrolysis on myristoylated ARF1 preloaded with GTP in the presence of phospholipid vesicles was coatomer-insensitive. The efficacy of ARF GAPs on myristoylated ARF1 was much higher than on ⌬17-ARF1 even when coatomer was used to stimulate GAP activity on ⌬17-ARF1 (compare Figs. 1 and 3). It is also noteworthy that the stimulation of GAP activity on ⌬17-ARF1 by coatomer in our experiments (ϳ10-fold) was much lower than that reported by Goldberg (up to 1000-fold) under similar conditions. This is probably because of differences in the GAP preparations. The full-length GAP1 and Gcs1 proteins we employed showed low but significant coatomer-independent activity whereas Goldberg employed a catalytic fragment of GAP1 (amino acids 6 -136), which showed negligible activity under comparable conditions. We have previously observed that truncated GAP1 mutants containing fewer than the first 146 amino acids displayed reduced activity on myristoylated ARF1 (18). It therefore appears that very short forms of GAP1 are incapable of effectively activating either ⌬17-ARF1 or myristoylated ARF1 whereas, for reasons that are not understood, coatomer appears to confer high activity on the short GAP1 mutants.
The fact that GAPs can be highly active with phospholipidbound ARF1 in the absence of coatomer argues against the previously proposed catalytic role of coatomer in GTP hydrolysis (27). Additionally, if coatomer acted through a catalytic mechanism such as the contribution of a catalytic residue to the ARF1 GTP-binding pocket (27), then one would predict that coatomer would stimulate the activity of all ARF GAPs. However, we found that ASAP1, which shows high similarity to GAP1 in its catalytic domain, displayed coatomer-insensitive activity (see Fig. 2A). Lastly, all ARF GAPs contain an invariant arginine residue, and this residue is essential for GAP1 activity either in the presence or absence of coatomer (Fig. 5) 6. A, an invariant arginine residue is required for GAP1 activity. GTP hydrolysis was assayed with the GAP1-(1-257) wild-type protein (WT) and its Arg-50 mutants using myristoylated ARF1 preloaded with GTP in the presence of DMPC and cholate. B, the Arg-50 mutant (R50K) does not inhibit coatomer-dependent GTP hydrolysis on ⌬17-ARF1. Activity was assayed in the presence of 0.2 M GAP1-(1-257) and different concentrations of the R50K mutant with or without 0.2 M coatomer. well as for the activity of two additional ARF GAPs (30,38). The observation that the gross structure of the arginine mutants is preserved (Refs. 30 and 38 and this study) argues that this residue in GAP plays a catalytic rather than structural role.
If coatomer does not act catalytically, then how does it stimulate GAP activity on ARF? The rate of GTP hydrolysis is determined by the kinetic parameters of the interaction between ARF and GAP, as well as by the turnover number of the GTPase reaction. Our previous studies (34) have shown that in the presence of phospholipid vesicles of varying compositions, GAP1 as well as Gcs1-dependent hydrolysis of GTP on myristoylated ARF1 correlates with the extent of binding of GAP to the vesicles. These observations suggest that phospholipid vesicles facilitate GAP activity because of a proximity effect brought about by the binding of both GAP and its substrate ARF1-GTP to the same vesicle. Coatomer may act in a similar manner by interacting not only with ARF1 but also with GAP, thus generating a tripartite complex with three pairs of protein-protein interactions. Such interaction could be restricted to only a few GAPs, because GAPs such as ASAP1 that do not depend on coatomer for activity ( Fig. 2A) may lack a coatomerbinding domain. The finding that coatomer does not facilitate GTP hydrolysis when ARF1 is bound to phospholipid vesicles suggests that GAP interacts with lipids more avidly than with coatomer. Upon disruption of the phospholipid vesicles by detergent (Fig. 5), GAP activity decreases and becomes coatomerresponsive. Apparently because of the small size of the mixed detergent/phospholipid micelles, simultaneous binding of ARF and GAP to the same micelle becomes unlikely. Thus, under these conditions, the system behaves in a manner that is qualitatively similar to the lipid-free ⌬17-ARF1 assay system. It is noteworthy that we did not observe an inhibition of coatomerdependent GAP activity by the GAP1 mutant R50K (Fig. 6B) and two other inactive mutants (D65A and W32A). 2 This absence of competition suggests that any interaction of GAP with coatomer must be of low affinity. Such low affinity interaction could nonetheless suffice for generating the stimulation of GAP activity by the proximity effect.
In addition to its modulation by coatomer and membrane lipids in vitro, GAP1 interacts in vivo with the Golgi receptor for endoplasmic reticulum proteins bearing the KDEL tag (43,44). Whereas coatomer and phospholipids affect the activity of carboxyl-terminal truncated GAP1 (34), the KDEL receptor interacts with GAP1 through the carboxyl-terminal segment (45). The understanding of how each of these factors contributes to GAP1 targeting and/or catalytic activity in vivo will be a subject of future studies.