The amino terminus of ADP-ribosylation factor (ARF) 1 is essential for interaction with Gs and ARF GTPase-activating protein.

The role of the amino terminus in the actions of ADP-ribosylation factor 1 (ARF1) was examined by comparing wild type ARF1, a 13-residue NH2-terminal deletion mutant ([delta 13]ARF1), and a 17-residue NH2-terminal deletion mutant ([delta 17]ARF1). The amino-terminal 13 residues of ARF1 are required for cofactor activity in the ADP-ribosylation by cholera toxin when Gs is the substrate. This is in marked contrast to the finding that cofactor activity is the same for wild type and [delta 13]ARF1 when agmatine is substrate (Hong, J.-X., Haun, R. S., Tsai, S.-C., Moss, J., and Vaughan, M. (1994) J. Biol. Chem. 269, 9743-9745). These data support the conclusion that ARF1 interacts with both cholera toxin and Gs and that the amino terminus of ARF1 is required specifically for binding Gs. Surprisingly, this result also clearly revealed that the two principal assays for ARF activity, cofactor activity for cholera toxin using either Gs or agmatine as substrates, used for over 10 years in different laboratories, can yield quite different results. While both NH2-terminal deletion mutants failed to support the ADP-ribosylation of Gs by cholera toxin, [delta 13]ARF1, but not [delta 17]ARF1, inhibited the activity of the wild type protein. The GTPase activity of [delta 13]ARF1 was activated to a small extent by ARF GTPase-activating protein (GAP), whereas that of [delta 17]ARF1 was unaffected. We conclude that residues 14-17 are involved in the interaction of ARF with both cholera toxin and ARF GAP. The co-purifying nucleotides, nucleotide exchange kinetics, and dependence of exchange on phospholipids for the mutant proteins were all different from the wild type ARF1 proteins. The importance of monitoring the nucleotide binding to ARF proteins under the conditions used in the ARF assay and expressing ARF activities as specific activities, normalized to GTP binding sites, particularly when comparisons between different proteins or preparations are made, is discussed.

conclusion that ARFl interacts with both cholera toxin and G, and that the amino terminus of ARFl is required specifically for binding G,. Surprisingly, this result also clearly revealed that the two principal assays for ARF activity, cofactor activity for cholera toxin using either G, or agmatine as substrates, used for over 10 years in different laboratories, can yield quite different results.
While both NH,-terminal deletion mutants failed to support the ADP-ribosylation of G, by cholera toxin, [A13]ARFl, but not [A17]ARF1, inhibited the activity of the wild type protein. The GTPase activity of [A13]ARF1 was activated to a small extent by ARF GTPase-activating protein (GAP), whereas that of [A17lARF1 was unaffected. W e conclude that residues 14-17 are involved in the interaction of ARF with both cholera toxin and ARF GAP.
The co-purifying nucleotides, nucleotide exchange kinetics, and dependence of exchange on phospholipids for the mutant proteins were all different from the wild type ARFl proteins. The importance of monitoring the nucleotide binding to ARF proteins under the conditions used in the ARF assay and expressing ARF activities as specific activities, normalized to GTP binding sites, particularly when comparisons between different proteins or preparations are made, is discussed.
ADP-ribosylation factors (ARF)' are a family of GTP-binding proteins that include the ARFs and the structurally related ARF-like (ARL) proteins (1,2). Several members of the ARF family are ubiquitous and highly conserved in both structure * The costs of publication of this article were defrayed in part by the "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to payment of page charges. This article must therefore be hereby marked indicate this fact.
$ To whom correspondence should be addressed: Laboratory of Biological chemistry, Bldg. 37, Rm. and function in eukaryotes (2). They are essential in Saccharomyces cerevisiae (ARF1 and ARF2) (3) and in Drosophila melanogaster (arflike) (4). ARF proteins have been implicated as regulators of a number of steps in both the exocytic and endocytic transport pathways (5,6) and as an activator of phospholipase D activity (7,8).
ARF was originally purified as the protein cofactor for cholera toxin-catalyzed ADP-ribosylation of G,, the G-protein that regulates activation of adenylate cyclase and is the major cellular substrate for cholera toxin. This covalent modification of G, results in the persistent activation of adenylate cyclase, leading to the imbalances in salt and fluid retention associated with cholera. By reconstituting this reaction with purified components, ARF was found to be a required cofactor and was first purified from membranes using the activation of G, as a quantitative assay for ARF (9,10). The purified protein allowed the cloning of the first ARF cDNA (11). Parallel studies originally focusing on the mechanism of action of cholera toxin led to the development of an ARF assay which employs the arginine analogue, agmatine, as substrate for cholera toxin (12). Using this assay ARF was first purified from soluble sources and led to the cloning of bovine ARF3 (13). ARFl and ARF3 proteins are 96% identical and apparently indistinguishable in these two assays. As these assays each quantify cholera toxin cofactor activity and have yielded very similar results over the past 10 years it has been assumed that they are interchangeable. However, this assumption has never been examined directly.
ARFl is a ubiquitous and abundantly expressed, N-myristoylated, GTP-binding protein that co-purifies with tightly bound GDP (10). The binding of GTP is required for expression of ARF activity, using either G, or agmatine as ADP-ribose acceptors. The binding of GTP, or the slowly hydrolyzable GTP analog GTP+, to ARF in vitro is highly dependent on phospholipids (10,14) and is sensitive to the concentrations of divalent metals and detergents. Under conditions which promote the formation of ARF.GTP, GTP stimulates the ADP-ribosylation of G, by cholera toxin more than 50-fold. Thus, stimulation of the ADP-ribosylation of G, by cholera toxin is a highly specific and quantitative biochemical assay for the activated protein, ARF-GTP. However, because of the enormous differences in nucleotide binding observed under different conditions the use of specific activities, i.e. normalizing cofactor activity to ARF.GTP determined under identical assay conditions, is particularly important when comparing different ARF proteins or preparations. Modification of agmatine by cholera toxin is also dependent on ARF and GTP but fewer studies are available which correlate the binding of the activating nucleotide with cofactor activity.
Among the distinguishing structural features of ARF proteins is a 14-16-amino acid extension at the amino terminus, evident after alignment with other Ras superfamily members. A mutant lacking the first 17 amino acids, termed [Al7lARF1, lost both dependence on phospholipids for GTP binding and all activity in the cofactor assay, using G, as substrate (15). Con-29490 versely, replacing the amino terminus of Drosophila Arll with the 17 amino-terminal residues of ARFl was sufficient to bestow both ARF activity and increased dependence on phospholipids for GTP binding on the chimera (15). Thus both loss of function and gain of function were tightly linked to the presence of the amino terminus.
The conclusion that the amino terminus of ARFl is an essential functional domain has been challenged recently by Hong et aE. (16). Another NH,-terminal mutant, in which the 13 NH,-terminal amino acids were deleted, [Al3]ARF1, was found to stimulate the cholera toxin-catalyzed transfer of ADP-ribose to agmatine to a similar extent (5-10 fold over cholera toxin alone) as did ARF1. Further, this activity was independent of added GTP. These data were interpreted as evidence that the amino terminus of ARF is not critical for activity (16). These data have motivated us to re-examine the role of the NH, terminus in ARF activities by directly comparing the two NH,-terminal deletion mutants and wild type protein in ARF and related assays. One of two explanations for the different conclusions seemed likely. Either residues 14-17 are critical for ARF activity or the assays used by the different laboratories are different in some fundamental way. Surprisingly, we conclude that these two assays of cofactor activity yield very different results as a result of the different cholera toxin substrates utilized. We also conclude that residues 14-17 are not critical for ARF (G,) activity but are implicated in the binding of ARF.GTP to both ARF GAP and cholera toxin.
and bovine brain G,, (18) were prepared as described elsewhere. reading frame was subcloned into the pET3C vector before transfection into BL21(DE3) cells (14). Bacteria containing a vector directing expression of [A13lARF1 were also generously provided by Dr. Martha Vaughan (16). Both proteins were purified as described for ARFl (19) with yields of 25 and 40 mg/liter of bacterial culture, respectively. Thermal cycle DNA sequencing was performed using the Promega kit (fmol Sequencing System, Promega Corp., Madison, WI) to confirm sequences of each mutant. As expected, the [A13lARF1 proteins behaved identically in all respects.
HPLC-Proteins were heated to 95 "C for 10 min and the released nucleotide was separated from the denatured protein by centrifugation in a Centricon 30 filtration apparatus. Filtrates prepared from 2-20 nmol of protein were fractionated using a Whatman analytical 10-pm Partisil SAX column equilibrated in 7 mM ammonium phosphate, pH 3.8 (buffer A) at a flow rate of 2 mllmin. The chromatograms were developed in a gradient from 100% buffer A to 100% 250 mM ammonium phosphate, pH 4.75,500 m~ KC1, over 30 min and monitored by absorption at 254 nm. To verify peak positions and to estimate recoveries, filtrates of samples to which known amounts of GTP and GDP were added prior to sample preparation were also fractionated. Between 0.5 and 1.0 mol of nucleotide was recoveredmol of ARF in each instance.
ARF GAP-The ARF GAP assay was performed as previously described (21) in the presence of l mg/ml phosphoinositides, 10-80 n~ ARF.GTP as substrate, and 10-87 pg/ml of partially purified ARF GAP from bovine brain (21). ARF GAP specific activity (ARF.GTP hydrolyzed min/mg GAP/ml) was calculated as previously described (21). ARF Assay-The ARF/G, assay was performed as described previously, using 0.5 PM cholera toxin, 0.25 p~ recombinant G,,, and the indicated concentrations ofARF proteins (10). This assay measures the ability of ARF to stimulate the cholera toxin catalyzed covalent incorporation of ADP-ribose from [a3'P1NAD into G,,. G,, subunits are required in the ARF assay and are supplied as a preparation from bovine brain membranes containing a small amount ofARF as a contaminant. ADP-ribose incorporation in the absence of added recombinant ARF was subtracted as background. Parallel GTPyS binding to ARF was performed using the same conditions but in the absence of G,. ARF activity was then normalized to the amount of GTPyS bound under the same conditions and specific activity was expressed as fmol ADP-ribose-G,J pmol GTPyS bound20 min.

RESULTS AND DISCUSSION
To test the hypothesis that residues 2-13 or 14-17 are critical determinants of ARF activity we compared wild type, [Al31ARFl, and [A17]ARFl in the ARF cofactor assay, using G, as substrate. Neither [A13]ARFl nor [Al7]ARFl had any activity in this assay. A time course for ADP-ribosylation of G, is shown in Fig. 1, and specific activities (normalized toARF.GTP formed) are presented in Table I. In the presence of 3 p~ ARF1, ADP-ribosylation increased with time over a 20 min incubation (Fig. 1). In contrast, in the presence of [A131ARFl, tested at concentrations ranging from 2.5 to 38 VM, no incorporation of ADP-ribose into G, was detected ( Fig. 1 and Table I). Because [A13]ARF1 bound up to 15-fold more GTP than did ARF1, as much as 100-fold more [Al31ARFl.GTP than ARF1,GTP was included in these reactions yet the mutant was inactive. Rather, activity in the presence of [A13lARF1 was less than that seen in the absence of added ARF (Fig. 1). This is thought to result from inhibition by [A13]ARF1 of the small amount of ARF contaminating the GPy preparation. Consistent with this interpretation, 2 VM tA131ARFl.GTP and 6 PM lAl31ARFl.GTP inhibited activity of 0.12 VM wild type ARFl by 95 and loo%, respectively (not shown). [Al71ARFl also had no detectable activity (see Table I) (151, and up to 7 p~ lA17lAFtFl did not significantly inhibit activity of the wild type protein. The differences in the two deletion mutants in the ability to inhibit wild type ARFl activity is discussed further below. We conclude that the 13 NH,-terminal amino residues ofARFl are essential for activity in this assay. The amino terminus has previously been shown to affect nucleotide exchange (15). Under the conditions of the ARF assay, which include the presence of DMPCkholate mixed micelles, [Al3]AEtF1 and [A17]ARFl bound l mol of GTPySlmol protein (Table I and Fig. 2). If phospholipid was excluded from the binding mixture, the stoichiometry was still 0.7-0.85 mol of GTP@ bound/mol ARF. In contrast, in the presence of phos-  In the absence of phospholipid, the wild type protein bound 5-10-fold less GTPyS (-0.01 mol/mol; Fig. 2, range of three experiments). The two mutants had similar GDP dissociation rates (Table I) which were more than 10-fold faster than the GDP dissociation rate from ARF1. These effects of the amino terminus to decrease stoichiometry of GTPyS binding and increase dependence on phospholipids are evidence that the amino terminus is involved in conformational changes that accompany GTP binding. The rapid and complete exchange of guanine nucleotide on [A13]ARF1, described above and seen in Fig. 2, are in marked contrast with the conclusion reached in Hong et al. (16) that nucleotides that purify with [A13lARF1 do not exchange. As a direct test, GTPyS binding under the conditions of the agmatine assay used by Hong et al. (including 0.003% SDS and 5 mM MgC1,; 16) was also determined (Fig. 2, filled squares). Binding to [A131ARFl was linear with time and reached a stoichiometry of only 0.15 mol of GTPySlmol protein after 1 h a t 30 "C. No binding to the wild type protein was detected under these conditions. Thus, the conditions used by Hong et al. (16) favored binding of GTP or GTPyS on [A13lARF1 relative to ARF1, although exchange was slowed relative to the other condition tested, that including DMPC and cholate.
We also analyzed the nucleotides that had co-purified with the different purified, recombinant proteins (Fig. 3). More than 85% of the nucleotide bound to ARFl was GDP. In contrast, 93 and 80% of the nucleotide bound to [Al31ARFl (Fig. 3) and [Al7lARF1 (not shown), respectively, was GTP. The differences in co-purifying nucleotides likely reflect differences in nucleotide affinities found for the different proteins as all three proteins lack intrinsic GTPase activity. The wild type protein has a greater affinity for GDP than for GTP, with GTP dissociation rates 10-fold greater than GDP dissociation rates. In contrast, GDP dissociates 10-fold faster than does GTP+ from LAl71ARFl (Table I and unpublished observation). Thus, NH, truncations promote the formation of ARF.GTP.
The differences in bound nucleotides in the different ARF preparations and slow nucleotide exchange rates measured under conditions used in the agmatine assay likely explain the lack of dependence of activity on added GTP, reported in Hong et al. (16). We estimate that the amount of ARF1.GTP and [A13]ARFl.GTP present in the assay employed by Hong et al. (16) differ by 100-1000 fold. The need to express ARF activities as specific activities, normalized to the active protein (ARF.GTP), when comparing different proteins or preparations is clear from these results.
In spite of the differences in exchange rates and co-purifying nucleotides we reach the startling conclusion that the cofactor assays yield different results when testing the [A13]ARFl mutant, depending on the substrate for cholera toxin employed. The clear qualitative difference is that this mutant is active in the ARF/agmatine assay but not in the ARF/G, assay. Having considered the effects of bound nucleotides and nucleotide exchange on specific activities we are left with only the differences in substrates as the likely explanation for this marked difference. Thus, we conclude that the amino terminus ofARFl is required for interaction with G, but is not required for the binding of cholera toxin.
The conclusion that G, and cholera toxin have distinct binding sites on ARF is further supported by the observation that [A131ARFl, but not [A171ARFl, inhibits the ADP-ribosylation of G, by ARFl (see above). This observation is explained if residues 14-17 of ARFl are involved in the binding of ARF to cholera toxin. The [A13lARF1 mutant can bind to the toxin but not G,, thus blocking the productive interaction of wild type ARFl with toxin and G,. This reasoning leads to the prediction that LA17IARF1 lacks activity as cofactor for cholera toxin even when agmatine is the substrate.
These same mutants were used to test whether the amino terminus was required for interactions of ARF with ARF GAP. ARFs are unique among GTP-binding proteins in having no intrinsic GTPase activity (10). GTP hydrolysis by ARF requires the presence of a GTPase activating protein, ARF GAP (21). Like the wild type protein, neither [A13]ARF1 (Fig. 4) nor LA17IARF1 (not shown) had detectable GTPase activity. Both mutants were found to be very poor substrates for ARF GAP (Table I). GAP stimulated GTPase activity of [A13]ARF1 was 1.2% that of ARFl ( Fig. 4 and Table I) and ARF GAP did not significantly activate the GTPase activity of [A17]ARF1. The addition of up to 24 p~ [Al3lARF.GTP had no effect on GAPstimulated ARFl GTPase activity. Therefore, the small stimulation of [A13lARF1 GTPase activity was likely the result of a low affinity interaction with ARF GAP. These data demonstrate that the amino terminus is required for interaction of ARFl with ARF GAP but also identify residues 14-17 as playing a small but detectable role in the interaction.
In summary, the NH,-terminal 13 residues of ARFl are required for activity in the ARFIG, assay and for interaction with G,. We speculate that residues 14-17 are involved in the interaction of ARFl with both cholera toxin and ARF GAP, based on differences in activities observed between the [A13]ARF1 and [A171ARF1 mutants. These results also demonstrate the need to monitor both the nucleotides bound and the rates of nucleotide exchange on ARF proteins when making quantitative comparisons between different ARF proteins or preparations. The dramatic effects of a variety of agents (most notably phospholipids, detergents, and divalent metals) on the rates of nucleotide exchange can lead to differences between preparations of greater than 1000-fold in the amount of active protein present and can confound even qualitative comparisons. The different cofactor assays have been used previously in attempts to define the site of ARF binding. It was originally suggested that activated ARF binds G, (lo), but the fact that it increases cholera toxin activity in the absence of G, led to the conclusion that ARF binds the toxin (12). An attractive feature of the above results is the conclusion that both previous conclusions appear to have been correct. It will be very interesting to see ifARF plays a similar "coupling" role in promoting coatomer association with membranes or in activation of phospholipase D.