Effects of Arfaptin 1 on Guanine Nucleotide-dependent Activation of Phospholipase D and Cholera Toxin by ADP-ribosylation Factor*

Arfaptin 1, a ∼39-kDa protein based on the deduced amino acid sequence, had been initially identified in a yeast two-hybrid screen using dominant active ARF3 (Q71L) as bait with an HL-60 cDNA library. It was suggested that arfaptin 1 may be involved in Golgi functions, since the FLAG-tagged protein was associated with Golgi membranes when expressed in COS-7 cells and could be bound to Golgi in vitro in an ADP-ribosylation factor (ARF)- and GTPγS-dependent, brefeldin A-inhibited fashion. Arfaptin 2, found in the same two-hybrid screen as arfaptin 1, is 60% identical in amino acid sequence and may or may not have an analogous function. We now report some effects of arfaptin 1 on ARF activation of phospholipase D and cholera toxin ADP-ribosyltransferase. Arfaptin 1 inhibited activation of both enzymes in a concentration-dependent manner and was without effect in the absence of ARF. Two ARF1 mutants that activated the toxin, one lacking 13 N-terminal amino acids and the other, in which 73 residues at the N terminus were replaced with the analogous sequence from ARL1, were not inhibited by arfaptin, consistent with the conclusion that arfaptin interaction requires the N terminus of ARF. This region has also been implicated in phospholipase D activation, but whether the two proteins interact with the same structural elements in ARF remains to be determined. Arfaptin inhibition of the action of ARF5 and ARF6 was less than that of ARF1 and ARF3; its effects were less on nonmyristoylated than myristoylated ARFs. Arfaptin effects on guanine nucleotide binding by ARFs were minimal whether or not a purified ARF guanine nucleotide-exchange protein was present. These findings indicate that arfaptin acts as an inhibitor of ARF actions in vitro, raising the possibility that it has a similar rolein vivo.

Golgi, endosomal, and probably nuclear membranes. Six mammalian ARFs have been identified by cDNA cloning. Based on deduced amino acid sequence, phylogenetic analysis, and gene structure, they have been divided into three classes: class I, ARF1, -2, and -3; class II, ARF4 and -5; class III, ARF6 (1). By definition, all ARFs stimulate cholera toxin (CTA) ADP-ribosyltransferase activity. In addition, all ARFs activate a specific phospholipase D (PLD) that can serve as an effector in cellular signal transduction (2,3). Whereas, the domain necessary for CTA activation resides in the C-terminal portion of the ARF molecule, activation of PLD is a function of an N-terminal region (4). The identification and isolation of ARF GTPaseactivating proteins, termed ARF GAPs (5-7), and GEPs, or guanine nucleotide-exchange proteins (8 -10), have extended our understanding of ARF action and its regulation. ARF GAP accelerates the hydrolysis of GTP bound to ARF, yielding inactive ARF⅐GDP. ARF GEP catalyzes the replacement of bound GDP by GTP to produce active ARF⅐GTP. It has been recognized relatively recently that several proteins initially characterized in other contexts exhibit ARF GEP activity. These include proteins with Sec7 domains, such as p200 from bovine brain (11), yeast Gea1 and Gea2 (12), and human ARNO (13) and cytohesin 1 (14).
Arfaptin 1, a ϳ39-kDa protein based on the deduced amino acid sequence, was identified in a yeast two-hybrid screen using dominant active ARF3(Q71L) as bait with an HL-60 cDNA library (15). The recombinant arfaptin 2 bound tightly to both myristoylated and nonmyristoylated ARF1 and ARF3 but much less to ARF5 and ARF6. The native arfaptin, immunoprecipitated from an HL-60 cell lysate, behaved as a molecule of ϳ44 kDa on gel electrophoresis (15). The physiological role(s) of arfaptins remains to be defined. It was suggested that arfaptin 1 may be involved in Golgi functions based on observations that the FLAG-tagged protein was associated with Golgi membranes when expressed in COS-7 cells and could be bound to Golgi in vitro in an ARF-and GTP␥S-dependent, brefeldin A-inhibited fashion (15). Arfaptin 2, which has 60% amino acid identity to arfaptin 1, may or may not have an analogous function. Both of the proteins were phosphorylated to a limited extent by protein kinase C (15). A Rac-binding protein termed POR1, which may have a role in Rac-dependent membrane ruffling, is identical to the C-terminal 303 amino acids of arfaptin 2 (15).
To understand better the arfaptin-ARF interaction, we investigated some effects of arfaptin 1 on ARF activation of CTA and PLD, and on ARF guanine nucleotide exchange. To begin to define regions of the ARF molecule that influence its interaction with arfaptin, native and recombinant ARFs (with and * 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 U.S.C. Section 1734 solely to indicate this fact. § To whom correspondence should be addressed: NHLBI, National Institutes of Health, Rm. 5N-307, Bldg. 10, 10 Center Dr., MSC 1434, Bethesda, MD 20892-1434. Tel.: 301-496-4554; Fax: 301-402-1610. 1 The abbreviations used are: ARF, ADP-ribosylation factor; mARF or rARF, myristoylated or nonmyristoylated recombinant ARF, respectively; GDP␤S, guanosine 5Ј-[␤-thio]diphosphate; GTP␥S, guanosine 5Ј-[␥-thio]triphosphate; CTA, cholera toxin A subunit; GEP, guanine nucleotide-exchange protein partially purified from rat spleen cytosol (9); GAP, GTPase-activating protein; PLD, phospholipase D; GST, glutathione S-transferase. 2 For simplicity, arfaptin 1 is referred to as arfaptin in this report.
without myristoylation) were studied as well as chimeric molecules that include portions of the ARF amino acid sequence along with that from human ARL1, an ARF-like protein that is 56% identical in sequence to human ARF1 (4) and which, unlike ARFs, does not activate cholera toxin or rescue Saccharomyces cerevisiae with the lethal arf1 Ϫ /arf2 Ϫ double deletion (16).
Assay of ARF Activity-ARF activity was assayed by three methods, guanine nucleotide binding, CTA activation, and PLD activation. All data reported are means of values from duplicate assays and have been replicated twice or more except those in Table IV. To assay nucleotide binding, ARF was incubated at 37°C with or without arfaptin and either [ 35 S]GTP␥S or [ 3 H]GDP for the indicated time. Protein-bound nucleotide was collected on nitrocellulose for radioassay (10).
For assay of ARF activation of CTA, ARF was incubated with or without arfaptin for 40 min at 37°C with 10 M GTP␥S in a 50-l volume, then placed in an ice bath. After addition of CTA, [ 14 C]NAD and agmatine in a volume of 250 l, samples were incubated for 60 min at 30°C and [ 14 C]ADP-ribosylagmatine was collected for radioassay (8). The activity of partially purified ARF-dependent PLD was assayed by a published method (4). Briefly, mixed lipids with choline[methyl-3 H]dipalmitoyl phosphatidylcholine were added to PLD, ARF, and GTP␥S with or without arfaptin and incubated for 1 h at 37°C before addition of CHCl 3 /CH 3 OH/HCl, followed by centrifugation. [ 3 H]choline in the aqueous phase was quantified by liquid scintillation spectroscopy.

Effects of Arfaptin on Stimulation of Cholera Toxin ADPribosyltransferase Activity by Native, Recombinant, and Mutant ARFs-
In an effort to define the functions of arfaptin 1, its in vitro effects on some known ARF activities were first tested. Stimulation of cholera toxin ADP-ribosyltransferase activity by native ARF3 (8) or ARF5 (18) was assayed with or without arfaptin. As shown in Fig. 1, inhibition of ARF activity was dependent on, but not linearly proportional to, the concentration of arfaptin. The activities of 10 pmol of native ARF1 and ARF3 were similar and were similarly inhibited ϳ30% by 10 pmol and 60 -70% by 40 pmol of arfaptin (Table I). The activities of recombinant ARF proteins myristoylated or nonmyristoylated, synthesized in E. coli were variable and much lower than those of native proteins. Inhibition, which was dependent on arfaptin concentration, was apparently somewhat less for ARF5 and ARF6 than for ARF1 and ARF3 (Table I).
CTA activity was enhanced by ARL73/ARF, which contains the first 73 amino acids of ARL1 and the last 108 of ARF1, but arfaptin did not interfere with its activity (Table I). The considerably greater activation by rARF1⌬13 (ARF1 lacking the N-terminal 13 amino acids) likewise was not prevented by arfaptin. These data are consistent with the possibility that arfaptin interactions require the N terminus of ARF. ARF73ARL, a recombinant protein containing the first 73 amino acids of human ARF1 and the last 108 of human ARL1 did not activate CTA as previously reported (4).

Effects of Arfaptin on Stimulation of Phospholipase D by Recombinant and Mutant
ARFs-In the absence of added ARF, partially purified PLD from bovine brain exhibited a low level of activity, probably due to ARF contamination (Table II), a possibility consonant with its inhibition by arfaptin. The same recombinant ARF preparations used for the experiments in Table I stimulated PLD activity from 10-to 25-fold (Table II). In these assays, mARF3 and mARF5 were clearly more active than nonmyristoylated forms. The dramatic activation by mARF3 was virtually completely inhibited by arfaptin, which was less effective against nonmyristoylated rARF3 and also against the ARF5 and ARF6 preparations (Table II). ARF activation of PLD required the N-terminal domain of the ARF molecule as shown earlier (4). The chimeric ARF 73/ARL, containing the N-terminal 73 amino acids of ARF1, increased activity 14-fold and was clearly inhibited by arfaptin (Table II). ARL 73/ARF with the C-terminal region of ARF1 and the N-terminal 73 amino acids of ARL1 as reported (4) did not activate PLD nor did the mutant rARF1⌬13, which lacks 13 amino acids at the N terminus, confirming that the N terminus FIG. 1. Effect of arfaptin 1 on activation of CTA by ARF3 or ARF5. Native ARF3, 0.2 g, ϳ10 pmol (f) or ARF5, 0.35 g, ϳ17.5 pmol (q) was incubated (50 l final volume) for 40 min with the indicated amount of arfaptin 1 before assay of its effect on CTA activity. In the absence of arfaptin 1, CTA activity was increased 2.8 and 4.8 nmol of ADP-ribosylagmatine formed/hr by ARF3 and ARF5, respectively. Activity of CTA alone, which was not inhibited by arfaptin 1, was subtracted before calculation of ARF activity as percentage of that in the absence of arfaptin 1 ϭ 100%.

TABLE I Effect of arfaptin 1 on activation of CTA by ARFs and related proteins
The indicated amount of native or recombinant nonmyristoylated, myristoylated ARF, or related protein was incubated without or with 10 or 40 pmol of arfaptin in a volume of 50 l as described in Ref. 8, except that 72 g of cardiolipin replaced PS before assay of activated ARF by its effect on CTA ADP-ribosyltransferase activity. CTA activity without ARF or arfaptin (1.3 nmol/h) was subtracted from the total to calculate the increment due to ARF ϭ ARF activity. CTA activities in the absence of ARF with 10 and 40 pmol of arfaptin were, respectively, 1. 4  of the ARF molecule is important for PLD activation as well as for arfaptin inhibition of CTA activation by ARF (Table I).
Whether arfaptin and PLD interact with the same structural determinants in ARF remains to be determined.
Effect of Arfaptin on GTP␥S and GDP Binding to ARF3-Replacement of ARF-bound GDP by GTP to produce the active ARF⅐GTP (in the presence of 2 mM MgCl 2 ) is very slow in the absence of an ARF GEP. Although exchange can be accelerated by decreasing MgCl 2 to Ͻ1.0 mM (9), it is known that the higher MgCl 2 concentration enhances stability of ARF-nucleotide complexes and that at low MgCl 2 concentrations ARF is rather unstable. Using 0.7 or 3.5 mM MgCl 2 , the time course of [ 35 S]GTP␥S binding to ARF3 was determined with or without arfaptin (Fig. 2). At the lower MgCl 2 concentration, the initial rate of binding was 10-fold that at 3.5 mM MgCl 2 . Arfaptin increased the amount of [ 35 S]GTP␥S bound after longer incubations, perhaps by stabilizing ARF, but the initial rate of binding was not markedly accelerated. At 3.5 mM MgCl 2 [ 35 S]GTP␥S binding to ARF3 was much slower (Fig. 2), and arfaptin had little effect. Replacement of ARF-bound [ 3 H]GDP with GDP from the medium was assayed without and with arfaptin at two concentrations of MgCl 2 (Fig. 3). At 0.1 mM MgCl 2 [ 3 H]GDP binding to ARF3 was only 30% higher than it was at 3.4 mM MgCl 2 . Binding was slow, and the rate was approximately constant for 80 min. Replacement binding of GDP was Ͻ20% greater with arfaptin than without it at both MgCl 2 concentrations (Fig. 3).
Effect of MgCl 2 Concentration and Arfaptin on GTP␥S Binding to mARF5-Stable binding of GTP␥S to mARF5 required relatively high concentrations of MgCl 2 (Fig. 4). Binding was maximal with 2-3 mM MgCl 2 and much lower with 1 mM MgCl 2 (in the presence of 1 mM EDTA). Arfaptin 1 increased GTP␥S binding 4-fold with 1 mM MgCl 2 , but only ϳ40% with 2-3 mM MgCl 2 . The enhancement by arfaptin 1 (Figs. 2 and 4) (Table IV). These small differences could reflect slowed dissociation of nucleotides.
ARF3 was incubated with GTP with or without ARF GEP and/or arfaptin, then assayed for stimulation of CTA activity. Without ARF GEP, ARF1-3 stimulation of CTA activity was inhibited by addition of arfaptin in a concentration-dependent manner (Table V) as shown also in Fig. 1 and Table I. After incubation with ARF GEP, ARF1-3 increased CTA activity ϳ120% above the level induced by ARF1-3 alone, and its activity was likewise inhibited by arfaptin. ARF5 was not activated by ARF GEP and arfaptin inhibited its activity with or without GEP, as shown also in Table I and Fig. 1.

DISCUSSION
Arfaptin inhibited the actions of all classes of ARFs on CTA and PLD activities. However, ARF1 and ARF3 (class I) were more sensitive than ARF5 (class II) and ARF6 (class III) to its inhibitory effects (Tables I and II). Studies on the interaction of recombinant human ARF1, -3, -5, and -6 with immobilized GST-arfaptin fusion proteins previously revealed an extent of stable association decreasing in that order with negligible bind-  ing of ARF6 (15). Because of the demonstrated importance of the ARF N terminus in its association with arfaptins, the several differences among ARFs in amino acid sequences of this region (1) may provide clues to structural requirements for the interaction. ARF6, in fact, lacks 4 of the first 11 residues present in the other ARFs. Interaction of ARF with arfaptin was also demonstrated in a more physiological setting. Brefeldin A inhibited the GTP␥S-dependent association of arfaptin 1 and ARF from HL-60 cell cytosol with rat liver Golgi membranes (15), consistent with the demonstrated association of arfaptin with activated ARF and not ARF-GDP. When ARFdepleted cytosol was used, recombinant ARF3 was able to support arfaptin binding to Golgi. Arfaptin 1 was initially described as a widely distributed cytosolic protein target of activated ARF, which recruited it to Golgi membranes (15). A clone for a second protein also containing 341 amino acids (60% identical), termed arfaptin 2, was studied less extensively but has properties similar to those of arfaptin 1 (15).
Myristoylation was demonstrated as not essential for ARF3 interaction with arfaptin 1 in the yeast two-hybrid screen, although its influence was not directly assessed (15). In the functional studies reported here, no consistent effect of myristoylation on the inhibitory action of arfaptin was observed (Tables I and II). The activity of CTA was not inhibited by arfaptin in the absence of ARF, and the effect of arfaptin on the very low activity of PLD in the absence of ARF is probably due to ARF contamination of the PLD preparation (data not shown). Percentage inhibition of PLD activation by arfaptin was, in general, greater than that of CTA activation. It may be relevant that the N-terminal 13 amino acids of ARF that were required for PLD activation were also required for arfaptin inhibition of cholera toxin activation (Tables I and II). Interpretation of the difference is complicated, however, by the major differences in assay conditions. It is necessary also to take into account the differences in effects of specific phospholipids on GTP binding by individual ARFs that can influence markedly the rate at (and extent to) which each becomes activated. Differences in behavior of the chimeric proteins are seemingly more straightforward in determining the structural requirement for interaction of ARF and arfaptin. As expected, ARF73ARL did not activate CTA but did activate PLD (4), and its action on the latter was inhibited by arfaptin (Table II). Neither ARL73ARF nor rARF1⌬13 significantly activated PLD in these experiments as reported (4). Both did, however, activate CTA, and arfaptin had no effect on their action, consistent with the suggestion of Kanoh et al. (15) that the N terminus of ARF is important for its interaction with arfaptin.
No GEP-like activity of arfaptin was detected. Nor was there any evidence that arfaptin 1 interfered with the action of a GEP partially purified from rat spleen. Whether nucleotide binding or release was measured, the amount of GTP␥S or GDP bound to several ARF preparations was consistently slightly   Native ARF proteins were incubated for 10 min at 37°C without or with arfaptin 1 as indicated (total volume 40 l). After addition of 20 M GTP and 4 mM MgCl 2 (final concentrations) with or without GEP (final volume 50 l), incubation was continued for 40 min at 37°C. Activated ARF was then assayed by its effect on CTA ADP-ribosyltransferase activity and is reported as the increment in CTA activity due to addition of ARF. higher in the presence of arfaptin 1 than in its absence. This was true also when GEP was added to accelerate guanine nucleotide exchange. Since this effect of arfaptin 1 was greater with longer times of incubation, it could reflect stabilization of the ARF protein as a result of its association with arfaptin 1.
The effect was larger with GTP␥S than with GDP (Table IV), which would be consistent with the demonstrated preference of arfaptin 1 for interaction with activated ARF (15), and the possibility that the complex of activated ARF and arfaptin 1 is a functional entity. Because arfaptin was discovered using the yeast two-hybrid system with a presumably GTP-liganded ARF mutant as bait, ARF was apparently able to interact with arfaptin in these cells (15). In addition, arfaptin in HL-60 cell cytosol was recruited by ARF to Golgi membranes in a GTP␥S-dependent and brefeldin A-sensitive manner (15). It remained unclear, however, from the prior studies how its interaction with arfaptin would affect the ability of ARF to activate either PLD or cholera toxin. The experiments reported here established that, even though activation of PLD and cholera toxin involves two different domains of ARF (4), arfaptin interferes with both processes. Based on these data, it would appear that once ARF recruits arfaptin to a membrane, dissociation of ARF from arfaptin would need to occur prior to activation of phospholipase D.
The present findings demonstrate that arfaptin is a potent inhibitor of ARF actions on CTA and PLD in vitro, but has minimal effects on GEP-catalyzed guanine nucleotide binding to ARF. Among the questions arising from this work are how the association of ARF with arfaptin modifies its interaction with GAPs and the hydrolysis of bound GTP. ARF plays a major role in the regulation of vesicular trafficking through the Golgi, and it will be important to demonstrate whether and how this is modified by arfaptin. These and other questions are subjects of current study.