Guanine Nucleotide Exchange on ADP-ribosylation Factors Catalyzed by Cytohesin-1 and Its Sec7 Domain*

ADP-ribosylation factors (ARFs) are 20-kDa guanine nucleotide-binding proteins that require specific guanine nucleotide-exchange proteins (GEPs) to accelerate the conversion of inactive ARF-GDP to active ARF-GTP. Cytohesin-1, a 46-kDa ARF GEP, contains a central Sec7 domain of 188 amino acids similar in sequence to a region of the yeast Sec7 protein. Cytohesin-1 and its 22-kDa Sec7 domain (C-1 Sec7), synthesized in Escherichia coli, were assayed with recombinant non-myristoylated ARFs and related proteins to compare their GEP activities. Both were effective with native mammalian ARFs 1 and 3. Cytohesin-1 accelerated GTPγS (guanosine 5′-3-O-(thio)triphosphate) binding to recombinant human ARF1 (rARF1), yeast ARF3, and ARD1 (a 64-kDa guanine nucleotide-binding protein containing a C-terminal ARF domain). In contrast, C-1 Sec7 enhanced GTPγS binding to recombinant human ARFs 1, 5, and 6; yeast ARFs 1, 2, and 3; ARD1; two ARD1 mutants that contain the ARF domain; and Δ13ARF1, which lacks the N-terminal α-helix. Neither C-1 Sec7 nor cytohesin-1 increased GTPγS binding to human ARF-like ARL proteins 1, 2, and 3. Thus, ARLs, initially differentiated from ARFs because of their inability to activate cholera toxin, differ also in their failure to interact functionally with C-1 Sec7 or cytohesin-1. As C-1 Sec7 was much less substrate-specific than cytohesin-1, it appears that structure outside of the Sec7 domain is important for ARF specificity. Data obtained with mutant ARF constructs are all consistent with the conclusion that the ARF N terminus is an important determinant of cytohesin-1 specificity.

gene structure, and sequence identity (3,4). ARF structure is highly conserved among eukaryotic species, and non-mammalian ARFs of all three classes are known (4). The ability of ARFs to serve as activators of cholera toxin distinguishes them from a closely related family of ARF-like proteins or ARLs (5). An ARF sequence is also found in the 64-kDa ARD1 (6), which is a bifunctional protein that contains an N-terminal GTPase-activating protein domain (7).
The function of ARFs in vesicular transport has been studied most extensively in the endoplasmic reticulum and Golgi (8,9). It is widely agreed that vesicle formation is initiated when activated ARF molecules with GTP bound associate with the cytoplasmic surface of a membrane. ARF-GTP interacts with a specific component (␤COP) of the so-called coatomer protein complex (10). Accumulation of multiple molecules of ARF plus coatomer results in membrane deformation, budding, and vesicle release after bilayer fusion at the base of the bud (1). Since ARF activation of phospholipase D was first reported (11,12), actions of this enzyme in both vesicle formation (13) and fusion (14) have been postulated. ARF function depends on the regulated alternation between active GTP-bound and inactive GDPbound states. Inactivation by hydrolysis of bound GTP requires the interaction of ARF-GTP with a GTPase-activating protein (15)(16)(17)(18). Activation of ARF-GDP results from the replacement of bound GDP by GTP, which is accelerated by guanine nucleotide-exchange proteins or GEPs (19,20).
All ARF GEPs of known structure contain Sec7 domains. Sec7 is a ϳ230-kDa yeast protein identified during characterization of a temperature-sensitive secretion-defective mutant (32). The Sec7 domains of ARNO (25) and GRP1 (27) catalyze guanine nucleotide exchange on ARFs, and crystal structures of the Sec7 domain of ARNO were recently reported (33,34). The Sec7 domain comprises 10 ␣-helices, including two highly conserved sequence motifs that appear to be essential for GEP activity. Although several ARF GEPs have been cloned and their activities demonstrated, very little is known about the determinants of substrate specificity. We, therefore, compared the activities of full-length cytohesin-1 (46-kDa) and its 22-kDa Sec7 domain (C-1 Sec7) with several ARF and ARL protein substrates. As reported here, cytohesin-1 accelerated GTP␥S binding by rARF1, yeast ARF3, and ARD1 only, whereas C-1 Sec7 was effective with a much broader range of ARFs and related proteins. The N terminus of ARF1 was shown to be important for its effective interaction with cytohesin-1.
Preparation of Cytohesin-1 and C-1 Sec7-Recombinant cytohesin-1 (B2-1) was prepared as described by Meacci et al. (26). To prepare the Sec7 domain (Arg 62 -Asp 249 ) as a His 6 fusion protein, B2-1 cDNA was digested with restriction enzymes KpnI and HindIII. The product was subcloned in plasmid pQE30, which was then integrated into the Escherichia coli M15 host strain carrying the pREP4 repressor plasmid. Single colonies, selected by resistance to ampicillin and kanamycin, were added to 5 ml of Luria-Bertani medium containing ampicillin (100 g/ml) and kanamycin (25 g/ml), incubated for 3-4 h at 37°C, and transferred to 100 ml of Luria-Bertani medium containing the same antibiotics. After reaching an A 600 of 0.5, and incubation with 2 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C, cells were harvested for purification of C-1 Sec7 (26). GEP activities of cytohesin-1 and C-1 Sec7 were regularly assayed with a partially purified mixture of native bovine brain ARFs (chiefly ARFs 1 and 3) as basis for comparing activities with other substrates.
[ 35 S]GTP␥S-Binding Assay-Procedures for assay of GTP␥S binding and GEP activity have been published (28). In brief, assays (50 l) containing ARF (1 g unless otherwise indicated) in Buffer A, with 40 g of bovine serum albumin, 10 g of PS, 0.1 g each of aprotinin, leupeptin, and soybean and lima bean trypsin inhibitors, 0.5 mM 4-(2aminoethyl) benzenesulfonyl fluoride hydrochloride, cytohesin-1, or C-1 Sec7 (1 g unless otherwise specified), and 4 M [ 35 S]GTP␥S (2.5 ϫ 10 6 cpm) were incubated for the indicated time at 37°C. Samples were then cooled on ice and transferred to nitrocellulose filters followed by an 8-ml wash of the incubation tube, and 4-ml wash of the filter with buffer B (25 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 , 100 mM NaCl, 1 mM dithiothreitol). Scintillation counting fluid was added to dried filters before radioassay of bound [ 35 S]GTP␥S. Most data are reported as means Ϯ S.E. of values from triplicate assays in a representative experiment. All observations have been replicated at least twice with different preparations of recombinant proteins.
Release of Bound Guanine Nucleotides from Recombinant Proteins-hARF (usually 50 pmol) was incubated for 40 min at 37°C in reaction mixture (50 l) containing 20 mM Tris-HCl (pH 8.0), 1 mM NaN 3 , 10 mM dithiothreitol, 200 mM sucrose, 10 g of PS, 40 g of bovine serum albumin, protease inhibitors as used in binding assays, 1 mM EDTA, 40 mM NaCl, and 4 M [ 35 S]GTP␥S (2.5 ϫ 10 6 cpm), or [ 3 H]GDP (1.3 ϫ 10 6 cpm). After addition of 1 mM amounts of the indicated non-radiolabeled nucleotide and 3 mM MgCl 2 (final concentration in a total volume of 100 l), incubation was continued at 37°C for the indicated time, with or without cytohesin-1 (1 g) or C-1 Sec7 (1 g) to determine nucleotide release. Bound [ 35 S]GTP␥S or [3H]GDP was quantified as described for the binding assay. K off represents the means Ϯ half the range from two independent experiments with triplicate incubations. ARFs 1,5, and 6 -The purified recombinant ARF proteins (non-myristoylated) were functional in their ability to stimulate cholera toxin ADP-ribosyltransferase activity. hARF1 and hARF5 required GTP for activity, whereas hARF6 was similarly active with GDP and GTP, presumably due to the presence of tightly bound nucleotide, which did not exchange with that in the medium (data not shown) as had been found previously (42). The rate of GTP␥S binding to hARF1 was proportional to the amount of GEP below ϳ2.2 pmol and was maximal with ϳ23 pmol of C-1 Sec7 or 11 pmol of cytohesin-1 (Fig. 1). GEP activity of both these molecules is relatively low, and data demonstrating GTP␥S binding in excess of the amount of GEP present are shown only in Figs. 3, 5, and 8. (Such data for cytohesin-1 are also found in Ref. 26.) Cytohesin-1 increased GTP␥S binding by hARF1 in a time-dependent manner, but had little or no effect on hARF5 or hARF6 (Figs. 2 and 3). hARF1 appeared to be the preferred substrate for cytohesin-1 with K cat /K m of 6.7 Ϯ 3.2 ϫ 10 4 min Ϫ1 /M. C-1 Sec7 increased GTP␥S binding to 1 M hARF1, 5, and 6 by 8-, 4-, and 3-fold, respectively (Fig. 4). The K cat /K m values for ARFs 1, 5, and 6 were, respectively, 2.2 Ϯ 1.2 ϫ 10 5 , 5.4 Ϯ 1.6 ϫ 10 4 , and 3.3 Ϯ 1.1 ϫ 10 4 min Ϫ1 /M (Fig. 5). C-1 Sec7 was less substrate-specific than cytohesin-1 and also more efficient in catalyzing nucleotide exchange on hARF1, consistent with the conclusion that structural elements outside of the Sec7 domain contribute to substrate specificity and serve also to restrain catalytic activity.
Effect of PS on GTP␥S Binding to Native and Recombinant hARF-Nucleotide binding to ARFs, with and without GEPs, is known to be influenced by phospholipids and detergents. In the absence of PS, cytohesin-1 did not alter GTP␥S binding by  (Table I). GTP␥S binding to 1 M hARF1 was significantly less than to 0.4 M native ARF, despite its higher concentration.
Effects of Cytohesin-1 or C-1 Sec7 on GTP␥S Binding by Other ARF-related Molecules-The initial rate of GTP␥S binding to ⌬13ARF1, which lacks the first 13 amino acids and the N-terminal myristate, was markedly increased by C-1 Sec7 (Fig. 8). Cytohesin-1 was, however, without effect (Fig. 8).
ARD1 is a 64-kDa protein that contains an 18-kDa C-terminal ARF domain (p3), which is 60% identical to the corresponding sequence of ARF1. The mutant ⌬387ARD1 which lacks the first 387 amino acids of ARD1, corresponds to the ARF domain plus a 15-amino acid N-terminal extension. GTP␥S binding to ARD1, ⌬387ARD1, and p3 was similarly accelerated by C-1  Sec7, although the effects were smaller than that on hARF1 (Table II, Experiment 1). Cytohesin-1 had effects similar to those of C-1 Sec7 on ARD1 and ⌬387ARD1 but was without effect on p3, which lacks sequence corresponding to the ARF N terminus that appears to be important for interaction with cytohesin-1.
C-1 Sec7 increased GTP␥S binding to yARFs 1 and 2 approximately 9-fold and to yARF3 4-fold (Table II, Experiment 2). Cytohesin-1, which had no effect with yARFs 1 and 2, did significantly increase binding to yARF3, albeit to a lesser degree than did C-1 Sec7 (Table II, Experiment 2). Although ARLs (ARF-like) proteins are very similar to ARFs in size and structure, neither C-1 Sec7 nor cytohesin-1 accelerated GTP␥S binding to hARLs 1, 2, or 3 (Table II, Experiment 3). DISCUSSION The data reported here establish that C-1 Sec7 can act as a GEP for a variety of ARF-related proteins. Determinants of the much more restricted substrate specificity of cytohesin-1 must, therefore, lie outside of the Sec7 domain. C-1 Sec7 accelerated GTP␥S binding to hARFs 1, 5, and 6; yARFs 1, 2, and 3; ARD1 and its ARF domain; and ⌬13ARF1 (lacking the first 13 amino acids). Cytohesin-1, on the other hand, was active only with hARF1 (or native class I ARFs), yARF3, ARD1, and ⌬387ARD1. Its ability to serve as a GEP for yARF3 was somewhat surprising, as yARF3 is in toto more similar to ARF6 (60% identical) than to any other of the mammalian ARFs (40). At the N terminus, however, it is not. In the first 12 positions, there are eight differences, including four residues missing in ARF6. Thus, the N terminus of hARF6 differs markedly from that of yARF3, as it does also from the five other mammalian ARFs. hARF1 differs also from yARF3 in 8 of the first 12 positions, but some of those are conservative replacements, and there are no "missing" amino acids so that the N-terminal structure of yARF3 overall probably resembles that of ARF1 more than it does that of hARF6. The failure of ARLs to serve as substrates for cytohesin-1 or its Sec7 domain parallels their relative inability to activate cholera toxin. It may be related to differences in structure of the switch 2 regions (4 of 11 residues different in ARF1 and ARL1), which in ARF is believed to interact with the ARNO Sec7 domain (33,34).
Although there was no release of bound GTP␥S from hARF6 without or with cytohesin-1 or C-1 Sec7, the latter did accelerate release of bound [ 3 H]GDP, consistent with its ability to accelerate GTP␥S binding by hARF6 (as well as several other ARF-related molecules). The mechanism of action of the ARF GEPs remains to be established. Certain GEPs for other GTPases of the Ras superfamily, however, are believed to act by stabilizing the protein in a nucleotide-free state and the association of ARNO with a nucleotide-free ARF mutant was shown directly by gel filtration (43). The intrinsic rate of GDP release from ARF1 has been reported to range over 2 orders of magnitude depending on the specific phospholipid present, whereas rates of release of GTP␥S varied less than 20% (44). In the presence of PIP 2 for example, the GDP off-rate was 20 times that for GTP␥S (44). Preferential interaction with the nucleotide-free conformation of hARF6 could be consistent with the effect of C-1 Sec7 on release of bound GDP, but not GTP␥S.
Cytohesin-1 failed to accelerate GTP␥S binding by the ARF domain of ARD1 (p3), which lacks sequence corresponding to the first 15 amino acids of ARF1, and thus resembles ⌬13ARF1 (6, 7). It did, however, function with ⌬387ARD1, which corresponds to p3 with an N-terminal extension of 15 amino acids, demonstrating again the importance of the ARF N terminus in the functional interactions with cytohesin-1. Both of these ARFs with N-terminal deletions were, however, substrates for C-1 Sec7. hARF6, although it does not have such a large deletion, may be another example of an ARF N terminus unsuitable for cytohesin-1 interaction, but functional with C-1 Sec7.
Other workers (43) demonstrated the importance of the ARF N terminus in nucleotide exchange catalyzed by ARNO, which is 82% identical in amino acid sequence to cytohesin-1. Two groups recently reported crystal structures of the ARNO Sec7 domain (33,34). Paris et al. (43) had earlier investigated the interactions of intact ARNO or its Sec7 domain with myristoylated ARF1 or ⌬17ARF1 (lacking the first 17 amino acids as well as the N-terminal myristate) and the effects of several phospholipids on them. Although GTP␥S binding to myristoylated ARF1 in the presence of ARNO was markedly accelerated by the addition of phospholipid vesicles containing PIP 2 , the phospholipids had no effect on ARNO-catalyzed GTP␥S binding to ⌬17ARF1. In addition, in the absence of phospholipid, the activities of intact ARNO and its Sec7 domain toward ⌬17ARF1 were identical (43). Thus, PIP 2 binding to the PH domain of ARNO does not modify its catalytic activity but serves to promote its association with membranes. Similarly, Paris et al. (43) found that ARF1-GDP was a poor substrate for ARNO in the absence of PIP 2 and concluded that binding of PIP 2 by both the PH domain of ARNO and the terminal ␣-helices of ARF is needed to concentrate the two proteins at a membrane surface, thereby facilitating their interaction. Conformational change at the N terminus resulting from interaction of the myristate and/or the ␣-helix with membrane lipids probably contributes to the release of bound GDP (intrinsic or GEP-catalyzed) and thereby GTP binding.
The effect of PS, which was required for cytohesin-1 activity and also enhanced that of C-1 Sec7, is an example of the importance of acidic phospholipids in the action of these ARF  GEPs. It differs, however, from the effect of phosphatidylinositol 3,4,5-triphosphate on GRP1, which was not seen with the Sec7 domain alone, i.e. it apparently required specifically the PH domain (45). PH domains from GRP1 and cytohesin-1, but not those from IRS-1 or SOS, specifically bound phosphatidylinositol 3,4,5-trisphosphate (27). Klarlund et al. (45) suggested that GRP1 and cytohesin-1 might serve to link signaling pathways that involve activation of phosphatidylinositol 3-kinase to ARF function in vesicular trafficking or in cell adhesion, based on the reported interaction of cytohesin-1 with ␤ 2 integrin (31).
Frank et al. (46) concluded that ARNO is a GEP for ARF6, whereas, in the experiments reported here, the effect of cytohesin-1 on ARF6 nucleotide exchange was considerably less than that on ARF1. The conditions used for assays in the two studies were, however, very different. Specific detergents or phospholipids can have major effects; we used PS, and Frank et al. used azolectin. They compared ARNO activities with 1 M myristoylated ARF1 and ARF6 (although total concentrations of the two ARFs were quite different because the recombinant ARF1 was 20% myristoylated and the ARF6 50%). All of the recombinant ARFs used in our studies were non-myristoylated, which could, of course, have contributed to the low activity of cytohesin; hARF1 was clearly a less effective substrate than the native ARF 1/3 mixture for both cytohesin-1 and C-1 Sec7 (Table I).
Because of the dramatic effects of brefeldin A on Golgi structure and function, a GEP that activates ARF for Golgi transport would be expected to be inhibited by BFA. Thus, when BFA sensitivity was apparently lost with purification of an ARF GEP from bovine brain, we had rationalized it as perhaps due to removal of a component that mediated BFA inhibition (28). In retrospect, it seems more likely that the ϳ55-kDa purified GEP (28) is a member of the cytohesin-1/ARNO/GRP1 family of BFA-insensitive GEPs and does not function in anterograde endoplasmic reticulum-Golgi transport. These GEPs, with C-terminal PH domains, may participate in events involving the plasma membrane, as suggested by Frank et al. (46) or in other of the numerous pathways of bi-directional intracellular membrane trafficking that are even less well understood. Localization of ARF6 with plasma membrane and tubulovesicular structures in the cell periphery had been known and subsequent studies clearly implicated it in endocytosis and regulation of the actin cytoskeleton (47). Demonstration of the similar distribution of ARNO (46), as well as its lack of sensitivity to BFA, is consistent, of course, with the suggested role as a GEP for ARF6. Elucidation at a molecular level of the physiological function of individual ARF proteins remains the goal.
The development of in vitro assays of GEP activity that will accurately reflect ARF specificity is of major importance, along with improved detection of individual ARFs in intact cells.