Dominant Inhibitory Mutants of ARFl Block Endoplasmic Reticulum to Golgi Transport and Trigger Disassembly of the Golgi Apparatus*

Using three different trans dominant mutants of bovine ARFl affecting GDP exchange or GTP hydrolysis we demonstrate the central role of ARFl in controlling vesicular traffic from the endoplasmic reticulum (ER) to the Golgi apparatus and between successive Golgi compartments. Overexpression of ARFl(Q71L1, a mutant likely to be restricted to the GTP-bound form, resulted in the accumulation of vesicular stomatitis virus glyco- protein in pre-Golgi intermediates, inhibited transport between successive Golgi compartments, and led to a striking association of &COP with pre-Golgi intermediates and the Golgi stack. In contrast, ARFl(T31N1, a mu- tant which is likely to have a preferential affinity for GDP compared to the wild-type protein, inhibited ex- port from the ER and triggered a brefeldin A-like phenotype, resulting in the redistribution of B-COP from Golgi membranes to the cytosol and the collapse of the Golgi into the ER. This mutant, which may efficiently sequester an ARF-specific guanine nucleotide-exchange protein (ARF-GEF), suggests that ARF and ARF-GEF are essential for export from the ER. These results are dis- cussed in the context of the GDP and GTP-bound forms of ARF in controlling both membrane structure and ve- sicular traffic through the early secretory pathway.

of ARF in controlling both membrane structure and vesicular traffic through the early secretory pathway.
Multiple GTP-binding proteins belonging to the rab (reviewed in Zerial et al. (1993)), ARF (Stearns et al., 1990a;Waters et ul., 1991a;Kahn et ul., 1992;Balch et ul., 1992), andSARl (reviewed in F'ryer et al., 1992) protein families control vesicular traffic through the exocytic pathway of eukaryotic cells. These small GTP-binding proteins are likely to function as molecular switches, controlling the assembly and disassembly of protein complexes involved in vesicle budding, targeting andor fusion. The composition of these complexes or their specific roles in vesicular traffic remain to be elucidated.
The ARF gene family now includes at least 6 members of a closely related group with homologies ranging from 80 to 95% Tsuchiya et al., 1991). ARF was originally discovered as the cofactor required for the ADP-ribosylation of the heterotrimeric G-protein G, (Kahn and Gilman, 1986;Kahn et al., 1988). More recent evidence suggests that ARF proteins play a critical role in vesicular traffic in yeast (Stearns et al., 1990b) and in mammalian cells  ence Organization, National Institutes of Health Grants GM 42336 and * This work was supported by grants from the Human Frontier Sci-CA-MFPPG (to W. E. B.), and the Deutsche Forschungsgemeinschaft (to C. D.). This work was also supported a grant from the Lucille P. Markey Charitable Trust. This is paper 8297-CB from the The Scripps Research Institute. 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. 1992; Taylor et al., 1992). AFW is a component of nonclathrincoated vesicles which accumulate in the presence of GTP@ (Waters et al., 1991b;Taylor et al., 1992) and appears to be essential in the recruitment of coatomer, a coat protein complex containing a-, p-, y-, and &COP proteins, to Golgi membranes (Donaldson et al., 1992a;Palmer et al., 1993;Orci et al., 1993). ARF not only participates in the recruitment of nonclathrin coat complexes, but has recently been demonstrated to facilitate the binding of p-adaptin, an adaptor protein associated with clathrin-coated vesicles (Robinson and Kreis, 1992;Stamnes and Rothman, 1993).
In addition to its role in vesicular traffic, ARF plays a key role in the maintenance of Golgi structure. Brefeldin A (BFA) triggers the reversible collapse of the cidmedial Golgi compartments to the ER (Donaldson et al., 1990;Lippincott-Schwartz et al., 1990;Graham et al., 1993;reviewed in Lippincott-Schwartz, 1993). While the target for BFA remains to be directly demonstrated, several lines of evidence strongly suggest that it inhibits the activity of a n ARF-specific guanine nucleotide-exchange protein (ARF-GEF) (Donaldson et al., 1992b;Helms and Rothman, 19921, indicating that the ability to recruit and maintain a population of ARF on Golgi membranes through the activity of GEF is fundamental to Golgi structure/ function . While in vitro assays which reconstitute ER to Golgi  and intra-Golgi transport Helms et al., 1993;Palmer et al., 1993) strongly implicate a role for ARF in vesicular traffic, less attention has been given to its role in vivo. ARF proteins, like all members of the ras superfamily, contain highly conserved guanine nucleotide-binding domains involved in GDP exchange and hydrolysis of GTP (Bourne et al. 1990. Amino acid substitutions in each of these domains lead to inactive GDP or activated GTP-bound forms. These mutants frequently have a trans dominant negative phenotype and have proven to be exceptionally powerful probes for elucidation of GTP-binding protein structure and function. The most stringently characterized protein in this regard has been ras (reviewed in Barbacid, 1987;Lowy and Willumsen, 1993), although considerable progress has been made in understanding the function of the small GTP-binding proteins SEC4 and YPTl in yeast (Schmitt et al., 1986(Schmitt et al., , 1988Walworth et al., 1989), and rabl (Tisdale et al., 1992), rab4 (van der Sluijs et al., 19921, and rab5 (Bucci et al., 1992) in mammalian cells through analysis of mutant function in vivo and in vitro (reviewed in Zerial(1993) and Pryer et al. (1992)).

1437
In the present paper, we address the specific role of ARFl in vesicular traffic between the ER and the Golgi, and control of Golgi structure in vivo. We utilize a transient expression system in conjunction with a series of trans dominant mutant forms of ARFl to analyze their effects on protein transport. This approach has a distinct advantage over previous pharmacological approaches (which have utilized more general reagents such as GTPyS, AlF4, or BFA) to explore EWGolgi structure/function. It allows us to identify the specific functional role(s) of the different GDP and GTP-bound states of ARFl in the control of membrane function.
We provide evidence that a substitution which is likely to convert ARFl to a form with a preferential affinity for GDP (ARFl(T31N)), inhibits export from the ER, triggers the release of p-COP to the cytoplasm, and promotes the disassembly and collapse of the Golgi compartment into the ER. Using both morphological and biochemical criteria, we find that this mutation in all respects mirrors the effects of BFA. Given the likelihood that ARFl(T31N) sequesters the function of a n ARFspecific GEF, these results provide the direct evidence that ARF-GEF is essential in ER to Golgi traffic, as well as Golgi membrane function and structure. In contrast, a substitution which is likely to constitutively lock ARF in the GTP-bound form (ARFl(Q71L)) does not inhibit export from the ER. Rather, this mutant leads to the accumulation of vesicular stomatitis virus glycoprotein (VSV-G) in pre-Golgi transport vesicular carriers and promotes binding of p-COP to pre-Golgi intermediates and Golgi membranes. These results provide evidence for the essential role of ARF and p-COP in mediating the organization and function of the Golgi stack in vesicular traffic, and suggest that both ARF and p-COP are essential for export from the ER in vivo.

EXPERIMENTAL PROCEDURES
M~terials-Tran~~S-label (specific activity > 1,000 Ci/mmol) was purchased from ICN Biomedicals, Inc. (Imine, CA) and [a-32P]GTP (specific activity 3,000 Ci/mmol) from DuPont-New England Nuclear Research. Endoglycosidase H (endo H) was obtained from Boehringer Mannheim, fluorescein isothiocyanate-conjugated Lens culinaris lectin from EY Laboratories (San Mateo, CAI, Texas Red goat a-mouse antibody from Molecular Probes, Inc., and fluorescein isothiocyanate goat a-rabbit immunoglobulins from Organon Teknika Corp. (West Chester, PA). All other reagents were obtained from Sigma unless otherwise indicated. The recombinant expression plasmid pOW12 (Weiss et al., 1989) carrying the bovine ARFl coding sequence under control of the T7 promoter was kindly provided by R. Kahn, National Cancer Institute (Bethesda, MD). The PAR-G (pET3a) plasmids encoding VSV-G and ts045-VSV-G were obtained from J. Rose, Yale University (New Haven, CT) and C. Machamer, Johns Hopkins University (Baltimore, MD). The polyclonal rabbit a-VSV-G antibody was raised against a 15-amino acid COOHterminal peptide of the cytoplasmic tail of VSV-G (our laboratory). All other antibodies used in this study were generous gifts from the following laboratories: the polyclonal antibody against p-COP from T. Kreis, University of Geneva (Geneva, Switzerland), the monoclonal antibody against p53 from H. P. Hauri, Biocenter (Basel, Switzerland), monoclonal antibodies recognizing the cytoplasmic tail (P5D4) or a lumenal domain of VSV-G protein (8G5) from K. Howell, University of Denver (Denver, CO) and from B. Wattenberg, Upjohn (Kalamazoo, MI), respectively.
The 3"fragments were generated using oligonucleotides complementary to the above mutagenic primers in combination with the 3'-end anti-sense oligonucleotide primer 5'-CCC CTC GGA TCC TCA TTT CTG G'IT CCG G-3', which was in addition used to introduce a BurnHI restriction site (underlined in the sequence) at the 3'-end of the ARF mutant genes. In the second fusogenic PCR reaction the appropriate pairs of overlapping fragments were then combined with the 5'-end oligonucleotide primer and the 3'-end anti-sense oligonucleotide primer to generate full-length mutant sequences. The G2A mutation was generated by PCR amplification using the mutagenic 5'-end oligonucleotide primer 5°C CCC TCT AGA CAT ATG GCG AAT ATC TTT G-3' and the 3'-end antisense oligonucleotide primer described above. All final PCR products were subsequently treated with T4 DNA polymerase and after solution from an agarose gel subcloned into the EcoRI restriction site of pBluescript SK(+). Mutations were confirmed by sequencing of the entire coding sequence by the chain termination method (Sanger et d . , 1977). All ARF mutant sequences were then introduced into the pET3a vector (Stratagene) as NdeIfBarnHI fragments for expression from the T7 promoter as described previously (Studier et al., 1990;Koshravi-Far et al., 1991;Tisdale et al., 1992). The G 2 M 3 1 N and G2NQ71L double mutations were generated using the pET3a constructs of ARFl(T31N) and ARFl(Q71L) as a template, respectively. The G2A mutations were introduced by PCR amplification using the mutagenic 5'-end oligonucleotide primer and the 3'-end anti-sense oligonucleotide primer and double mutants were subsequently subcloned as described above. Mutations were confirmed by sequencing of the entire coding sequence by the chain termination method (Sanger et al., 1977).
Dansfection Procedure-HeLa cells were maintained in Dulbecco's modified Eagle's medium (DMEM, Life Technologies Inc.) supplemented with 10% fetal bovine serum and 100 unitdml each of penicillin and streptomycin in a 5% humidified CO, incubator.
Infection with recombinant vaccinia virus and transfection of HeLa cells were performed as described by Tisdale et al. (1992). Briefly, HeLa cells (1.5 x 109 were plated in 60-mm cell culture dishes the day prior to transfection. Before infection the cells were washed twice with DMEM and then infected with vaccinia T7 RNA polymerase recombinant virus (vTF7-3; Fuerst et al., 1986) a t a multiplicity of 10 plaqueforming unitdcell in 1 ml of DMEM for 30 min with intermittent rocking in a 37 "C CO, incubator. The infection media was removed and the cells co-transfected with 7 pg of recombinant PET plasmid DNA encoding ARF wild-type or a particular mutant construct, and 7 pg of an expression vector carrying VSV-G (PAR-G). Both plasmid DNAs had been premixed for 15 min a t room temperature in 2 ml of DMEM and 35 p1 of transfection reagent (TransfectACE) prepared as described (Rose et al., 1991). Two hours post-transfection the inoculum was removed and saved. The cella were detached from the culture dish with PBS containing 5 n m EDTA (PBSBDTA), washed once with DMEM, and resuspended in the transfection media from the parental dish and 4 ml of DMEM, 4% fetal calf serum. One-third of the cells (5 x lo5) was then replated on a 35-mm tissue culture dish and used for pulse-labeling experiments. Two-thirds of the cells (1 x 10") were replated in a 60-mm dish and used for Western blot analysis to determine the expression level for each individual experiment. Cells were incubated for an additional 3 h to allow expression of VSV-G protein and ARF proteins.
To study the time course of inhibition two 10-cm dishes of HeLa cells (3 x 106/10-cm dish) were infected with recombinant vaccinia virus at a multiplicity of 10 plaque-forming unitdcell in 2 ml of DMEM and either transfected with 15 pg of PAR-G alone or co-transfected with PAR-G and the particular ARF expression plasmid, 15 pg each in 4 ml of DMEM as described above. Two h post-transfection the inoculum was removed and saved and the cells were detached from the culture dishes with PBSBDTA. The cells were combined, washed in DMEM, and resuspended in the transfection media from the parental dishes and twice the volume DMEM, 4% fetal calf serum. For each indicated time point 5 x lo5 cells were replated in 35-mm dishes for pulse-labeling experiments and 1 x 10" cells in 60-mm dishes to determine the level of expression by Western blot analysis. Cells were further incubated for the total times as indicated under "Results." ?).ansport Analysis-Transport analysis was performed as described earlier (Tisdale et al., 1992). Briefly, after post-transfection (see above) the inoculum was removed and the cells incubated for 15 min in methionine-deficient MEM medium. Cells were radiolabeled for 10 min with 20 pCi of Tran35S-label and then chased for 1.5 h in DMEM, 10% fetal calf serum. After chase, cells were rapidly transferred to ice to terminate transport, washed once with PBS, and lysed in lysis buffer (50 n m Tris-HC1, pH 8.0, 150 m NaCl, 1% Triton X-100, 1 m phenylmethylsulfonyl fluoride). VSV-G protein was subsequently immunoprecipitated using a monoclonal antibody to VSV-G (clone 8G5) and

TABLE I
A R F l mutant proteins generated for this study The three GTP-binding domains highly conserved between members of the ras superfamily of small GTP binding proteins are indicated as GI, GZ, and GB. The acceptor site of post-translational myristoylation is marked by myr. Amino acid residues within these domains subjected to mutagenesis are indicated by large bold letters highlighted by shaded boxes. The corresponding amino acid substitutions are shown in boxed letters for each single mutant protein. Pansorbin (Calbiochem Corp.). For digestion with endo H, immunoprecipitates were washed three times with 10 m~ Tris-HC1, pH 6.8,150 m~ NaCl, 0.1% Triton X-100 and heated to 95 "C in 10 m~ Tris-HC1, pH 6.8, 1% SDS for 5 min. After centrifugation half of each sample was treated with endo H overnight. Samples were finally analyzed by SDS-PAGE and autoradiography. Autoradiograms were quantitated by scanning densitometry (GS 300 transmission scanning densitometer, Hoeffer Scientific Instruments, San Francisco, CA).
Indirect Immunofluorescence-HeLa cells (5 x lo5) were plated on 10-mm coverslips in 35-mm tissue culture dishes 5 h prior to transfection. Cells were infected and co-transfected as described above with 5 pg of a plasmid encoding the temperature-sensitive mutant form of VSV-G (pARtsO45-G) and 5 pg of pET3a plasmid DNAencoding ARF proteins. Cells were incubated at 40 "C for 4 h followed by incubation at 40 or 32 "C for 1.5 h as indicated under "Results." To terminate transport, cells were rapidly transferred to ice, washed three times with PBS, and fixed in 2% paraformaldehyde in PBS for 10 min at room temperature. After fixation, cells were washed three times with PBS and permeabilized with 0.05% saponin in PBS supplemented with 5% normal goat serum for 10 min at room temperature (or-VSV-G, u-p53) or, alternatively, with ice-cold anhydrous methanol for 1 min on ice (or,p-COP). Permeabilized cells were again washed three times with PBS and incubated for 30 min at room temperature with the appropriate first antibody diluted in saponinA'BShorma1 goat serum. ARer rinsing the cells three times with PBS, primary antibodies were visualized by incubating the coverslips with Texas Red goat a-mouse antibody or fluorescein isothiocyanate goat a-rabbit IgGs in saponinPBShorma1 goat serum for 30 min at room temperature. For labeling Golgi compartments, cells were co-stained with fluorescein isothiocyanate-conjugated L. culinuris lectin. For double immunofluorescence, the appropriate combination of primary and secondary antibody or L. culinuris lectin was simultaneously added to the cells. The cells were washed extensively and coverslips were mounted in Aqua-Polymount (Polysciences, Warrington, PA1 and viewed using a Axiover microscope at x 630 (Carl Zeiss, Oberkochen, Germany). The images were photographed with T-MAX 400 Asa film (Kodak, Rochester, NY).
Preparation of Antibodies and Immunoblot-Recombinant bovine A R F l protein was bacterially overexpressed and purified from a soluble pool as described by Weiss et al. (1989). Fractions with highly enriched ARFl protein (more than 70%) were used as antigen to generate polyclonal antibodies from rabbits by standard procedures. From these antisera-specific a-ARF1 antibodies were affinity purified in small scales using recombinant ARFl wild-type protein.
The level of expression ofARF proteins in HeLa cells was determined by Western blot analysis. HeLa cells from a 60-mm tissue culture dish infected and co-transfected as outlined above, were washed with PBS and detached with PBWDTA. After washing, an aliquot of the cell suspension was taken to determine protein concentration and cells were pelleted and lysed in Laemmli sample buffer at 95 "C for 5 min. Proteins of whole lysates (100 pg of protein) were separated by 12.5% SDS-PAGE and electrophoretically transferred to nitrocellulose filters which were then blocked for 1 h in buffer A (50 m~ Tris-HC1, pH 7.4, 150 m~ NaCl, 0.25% Tween 20) with 5% milk. Membranes were incubated for 1 h at room temperature with affinity purified rabbit u-ARF1 antibodies diluted in blocking buffer. Filters were then washed with buffer A, two times with buffer B (0.2% SDS, 0.9% NaCl, 0.5% Triton X-loo), and again with buffer A for 10 min each. Staining was performed with horseradish peroxidase-conjugated goat a-rabbit immunoglobulins (Pierce Chemical Co.) and the ECL-Western blotting detection system (Amersham Corp.) according to the suppliers recommendations. The amount of ARFl protein was estimated by scanning densitometry.
GTP Ouerlay-ARF1 wild-type and A R F l mutant proteins were overexpressed in exponentially growing Escherichia coli cells (strain BLal(DE3)pLysS) transformed with the particular pET3a-ARF construct by incubation with 1 m~ isopropyl-P-D-thiogalactopyranoside (IF'l'G). After 3 h of induction cells were harvested and lysed in Laemmli sample buffer. The relative amounts of overexpressed recombinant ARFl proteins were estimated by Western blot analysis. The GTP overlay experiment was performed as described by Wagner et al. (1992). Protein lysates containing equivalent amounts of ARF proteins were separated by 12.5% SDS-PAGE. The gel was then soaked in 50 m~ Tris-HC1, pH 7.5, 20% (v/v) glycerol for 30 min and transferred onto nitrocellulose filters using a combination of 10 m~ NaHC03 and 3 m~ NaZCO3, pH 9.8, as a transfer buffer. After transfer, the filters were rinsed twice for 10 min each in binding buffer (50 m~ NaH2P0,, pH 7.5, 10 p~ MgCl2, 2 m~ dithiothreitol, 0.3% (v/v) Tween 20, 4 p~ ATP) and then incubated for 2 h in binding buffer containing 1 pCi/ml [CI-~~PIGTP (13 pmol) alone or in combination with an equal amount of unlabeled GTP (13 pmol) or 1000-fold excess of unlabeled GDP (13 nmol), respectively. The filters were subsequently washed 6 times for 3 min each in binding buffer and, after drying, exposed to x-ray film (Kodak, Rochester, NY).

Generation and Characterization of Bovine ARFl Mutants-
To address the potential role of ARF proteins in the regulation of vesicular transport between the endoplasmic reticulum and the Golgi apparatus, we performed site-directed mutagenesis within domains likely to be essential for ARF function. A series of mutants were generated in three guanine nucleotide binding motifs which are highly conserved between members of the ras superfamily ( Table I) (Bourne et al., 1990. The generation of analogous mutations in other members of the GTPbinding protein family has previously documented the potency of this approach to study the diverse cellular functions of these proteins (Der et al., 1988;Schmitt et al., 1986Schmitt et al., , 1988Walworth et al., 1989Bucci et al., 1992van der Sluijs et al., 1992;Tisdale et al., 1992). In particular, intensive mutational and structural analyses of H-ras have assigned residues in each of these domains (Table I: GI, G2, and GS) with defined functions in the binding and hydrolysis of GTP (reviewed in Barbacid (1987), Bourne et al. (1990,1991), Pai et al. (1990), and Wittinghofer and Pai (1991)). A S17N substitution in the GI domain of H-ras alters the guanine nucleotide binding affinity, resulting in a form of ras which is largely restricted to the GDP-bound form (Feig and Cooper, 1988 Table I).
In addition, and in contrast to most other small GTP-binding proteins which are prenylated or palmitoylated at the carboxyl terminus (reviewed in Der and Cox (1991)), ARF proteins are myristoylated a t their amino termini on a glycine residue a t position 2 (Table I). Myristoylation has been shown to be important for the reversible interaction of ARF with membranes  and to be critical for its biological function in vitro Palmer et al., 1993). The substitution of this residue by alanine should prevent membrane attachment and should yield a cytosolic, nonfunctional protein.
The analogous mutations in bovine ARFl which were generated for the present study are summarized in Table I. To test the biochemical properties of the three GTP binding mutants of ARFl we performed a qualitative GTP overlay technique using ARF bound to nitrocellulose filters (Wagner et al., 1992). As expected, bacterially expressed ARFl(T31N) and ARFl(N1261) mutant proteins showed only very weak binding of [a-32P]GTP when immobilized on nitrocellulose membranes (data not shown). No binding was detected in the presence of 1000-fold molar excess of GDP (Fig. 1). An identical result was observed forARFl(Dl29N) (data not shown). In contrast, both wild-type and mutant ARFl(Q71L) recombinant proteins showed strong [a-32PlGTP binding in the absence (data not shown) or the presence of a large excess of unlabeled GDP (1000-fold) (Fig. l), suggesting a high affinity for GTP over GDP. These results indicate that the predicted phenotypes based on guanine nucleotide binding of other small GTP-binding proteins is applicable to ARF1, one of the most distant relatives of the ras superfamily (Valencia et al., 1991).

Dansient Expression ofARFl Wild-type and Mutant Proteins
in HeLa Cells-In order to examine the effects of the generated mutant forms of bovine ARFl on vesicular transport from the ER to the Golgi apparatus in mammalian cells, we used a recombinant T7 vaccinia virus system (Fuerst et al., 1986) to transiently express the wild-type and mutant forms in HeLa cells. This approach has previously been applied to document the inhibitory effects of trans dominant negative mutant forms of rabl and rab2 on the transport of VSV-G from the ER to the Golgi stack (Tisdale et al., 1992).2 VSV-G has been used extensively as a marker protein to study the biochemical and molecular basis for transport through the exocytic pathway in vivo (Tisdale et al., 1992)2 and in vitro (Beckers et al., 1987 et Davidson and Balch, 1993h3 VSV-G is a type I transmembrane protein containing two N-linked carbohydrate chains. Vectorial transport of VSV-G from the ER through sequential cis, medial, and trans Golgi compartments can be measured by the processing of the two oligosaccharide chains from the high mannose (Man9) endo H-sensitive form found in the ER and pre-Golgi intermediates (S in Fig. 2 A ) to endo H-resistant forms found in the Golgi stack. These processing intermediates can be readily distinguished by SDS-PAGE (Schwaninger et al., 1991;Davidson and Balch, 1993). The first endo H-resistant (R1) form corresponds to the transport of VSV-G to the early cis Golgi compartment where one or both of the oligosaccharides chains becomes endo H resistant by the action of resident al,2-mannosidases and glycosyltransferases (R1 in Fig. 2, A and B ) (Schwaninger et al., 1991;Tisdale et al., 1992;Davidson and Balch, 1993). Subsequent transport of VSV-G to the trans Golgi network results in the appearance of the fully processed form containing two complex, endo H-resistant oligosaccharides (RT in Fig. 2, A and B ) (Schwaninger et al., 1991;Tisdale et al., 1992;Davidson and Balch, 1993). The appearance of sequential processing intermediates allows us to directly examine the effects of mutant constructs on both ER to Golgi and intra-Golgi transport within the same experiment (Tisdale et al., 1992;Davidson and Balch, 1993h2 HeLa cells infected with the recombinant vaccinia virus (fl'F7-3) were co-transfected with an expression vector canying either wild-type or mutant ARFl genes and a vector encoding VSV-G. Cells were incubated for 3-6 h to allow sufficient time for protein expression (Tisdale et aL, 1992). Transfected cells were subsequently pulse-labeled with [35S]methionine for 10 min, followed by a chase in the presence of unlabeled methionine for 1.5 h. Using this protocol, between 30 and 60% of the total cells are transfected. The efficiency of co-transfection of VSV-G with plasmids encoding ARF constructs is close to 100% (Tisdale et al., 1992h2 The expression of wild-type and mutant ARFl proteins was monitored by Western blot analysis (Fig. 2C) using an affinity purified polyclonal antibody generated against bacterially expressed recombinant ARFl wildtype protein. Overexpression in this system can be observed as early as 3 4 h post-transfection ( Fig. 3; Tisdale et al., 1992). In general, expression levels of the wild-type and mutant proteins by 5-6 h of post-transfection varied between 2 and 15-fold over endogenous levels depending on the vector and the transfection efficiency within each particular experiment.
As shown in Fig. 2, A and B, expression of the ARFl wildtype protein in vivo partially (12%) inhibited transport of VSV-G to the %form when expressed at a level 2-5-fold that of the endogenous pool. A correlation between the extent of inhibition of transport and the levels of expression of wild-type ARFl is shown in Fig. 3A. Expression could be detected between 3 and 4 h post-transfection with ARFl(wi1d-type) protein continuing to accumulate over the 6-h time course. Transport to the Rp form was inhibited by only 16% even when the expression level was -12-fold the endogenous pool (Fig. 3A ). Although no variation in the steady state level of the cidmedial Golgi R1 intermediate form could be detected throughout the time course, reflecting its rapid transit through these compartments, at 6 h post-transfection -36% of VSV-G remained in the endo H-sensitive form (S in Fig. 3  myristoylated at the amino terminus) had no effect on transport (Fig. 2, A and B). In this case, VSV-G protein was efficiently processed to the endo H-resistant R1 and & forms in a manner comparable to the control lacking any recombinant ARFl plasmid (Fig. 2, A and B). Mutant Forms of Bovine ARFl Defective in GTP-binding Potently Inhibit ER to Golgi und Intra-Golgi Dunsport-In contrast to the rather weak inhibition observed by overexpression of the wild-type ARFl protein, mutant forms defective in GTP binding or hydrolysis resulted in striking inhibition of VSV-G transport. In this experiment the levels of expression of the different mutants varied from 2 to 5-fold (Fig. 2B 1. Overexpression of the GDP-bound ARFl(T31N) mutant, the GTP-bound form ARFl(Q71N1, and a mutant form with a high guanine nucleotide-exchange rate (ARFl(N1261)) efficiently (>99%) inhibited maturation of VSV-G to the terminally processed & form (Fig. 2B). In all cases, a significant fraction of the total VSV-G remained in the endo H-sensitive, pre-Golgi S form (Fig.  2 B ) . However, even at high levels of expression, partial processing of VSV-G to the R1 form (20-30%; Figs. 2 and Fig. 3, C and D) was observed for both the ARFl(T31N) and ARFl(N1261) mutants. In contrast, the ARFl(Q71L) mutant protein efficiently prevented processing of VSV-G to the Rl form (&lo%, Figs. 2, A and B, and 3B ). The lack of substantial accumulation of VSV-G in the R1 form in the presence of each of the three mutants indicates that both ER to Golgi and intra-Golgi transport are substantially impaired, with transport through sequential Golgi compartments being particularly sensitive.
Analysis of the time course of inhibition indicates that low levels of each of the mutant proteins potently inhibit ER to wt 2A

31N 'lL 12" 129N
Golgi and intra-Golgi transport (Fig. 3, B-D). Unlike the wildtype ARFl protein for which only partial inhibition of transport to the & form (16%) was observed even when the protein was These results are in agreement with previous studies in which we demonstrated that the addition of recombinant myristoylated wild-type ARFl partially inhibited VSV-G transport from the ER to the Golgi  and intra-Golgi transport  in vitro. As a control, overexpression of the unprocessed ARFl(G2A) mutant (which cannot be present a t a concentration 15-fold over the endogenous pool after 6 h of transfection (Fig. 3A), complete (>99%) inhibition of transport to the & form was observed after a 2-5-fold overexpression by all three mutants after only 3 h of expression (Fig.  3, B-D). Inhibition of transport of VSV-G from the ER to the cis Golgi R, form by the ARFl(T31N) and ARFl(Q71L) mutants showed maximal effects by 3-4 h post-transfection when the expression levels were 5-fold or less than the endogenous pool (Fig. 3, B and C ) . Although the onset of the ARFl(N1261) mutant phenotype was also quite rapid (61% inhibition of transport by 3 h), the level of inhibition increased after 4-6 h of expression ( t o 75% inhibition) (Fig. 30).
Only partial inhibition of VSV-G processing to the & form was observed when VSV-G was co-transfected with the ARFl(D129N) mutant (Fig. 2B ), consistent with the less potent phenotype of the equivalent mutation in H-ras (Der et al., 1988). In this case, -18% of the total VSV-G could be detected in the mature endo H-resistant & form compared to complete inhibition in the presence of the other mutants (Fig. 2, A and B 1. To determine whether myristoylation was critical for the function of the various ARFl mutants, double mutants were constructed. For this purpose, the myristoyl acceptor site was mutated to Ala in the T31N and Q71L mutants to form ARFl(GBAIT31N)) and ARFl(GPNQ71L). In contrast to the striking effects of the myristoylated forms (Figs. 2 and 31, no inhibition of transport could be detected at levels of expression 2-5-fold the endogenous pool (data not shown). These results demonstrate that the effects observed on VSV-G transport are a specific consequence of co-expression of VSV-G with the different trans dominant ARF mutants, and that myristoylation is critical for membrane association.
The data described above indicate that the GDP, GTP and "activated"' mutant forms of ARFl either directly or indirectly perturb both ER to Golgi and intra-Golgi vesicular traffic. Given that the onset of inhibition of intra-Golgi transport (processing of VSV-G to the form) compared to ER to Golgi transport (processing to the R1 form) was more rapid, complete, and required a lower level of expression, these results suggest that each of these mutants may have a compound effect on both Golgi function in vesicular traffic and possibly Golgi structure.

Morphological Analysis of VSV-G Dunsport in the Presence of ARFl
Mutants-To determine the effects of the different mutants on ER and Golgi structure and function, we examined the morphological effects of the ARFl mutants on transport in vivo. As demonstrated above and as established previously (Tisdale et al., 1992),2 co-expression of VSV-G with different plasmids occurs with >90% efficiency, allowing us to directly analyze the effects of the different ARFl mutant constructs on the transport of VSV-G and Golgi structure in individual cells using indirect immunofluorescence (Tisdale et al., 1992L2 For this purpose, HeLa cells were transfected with an expression vector carrying a temperature-sensitive form of VSV-G (ts045) which fails to exit the ER when cells are incubated at the restrictive temperature (39.5 "C). Under these conditions, ts045-VSV-G shows a reticular distribution characteristic of the ER (Fig. 4u). Shift of the cells to the permissive temperature (32 "C) 4 h after transfection at the restrictive temperature results in the synchronized release of ts045-VSV-G from the ER and transport to the compact, juxta-nuclear Golgi stack during an ensuing 90min incubation (Fig. 4c, arrows). In this case, ts045-VSV-G overlaps with the distribution of L. culinaris lectin (Fig. 4, compare c (VSV-G) to d (L. culinaris lectin), a protein which binds with high-affinity to N-linked oligosaccharides containing terminal N-acetylglucosamine, serving as a strong marker for cidmedial Golgi compartments (Yamamoto et al., 1982). The lectin will also recognize, albeit with reduced affinity, N-linked oligosaccharides with terminal Man residues associated with pre-Golgi compartments (Osawa and Tsuji, 1987).
When cells were co-transfected with the GTP-bound mutant form ARFl(Q71L) for 4 h and then shifted to the permissive temperature for 90 min, ts045-VSV-G did not enter the Golgi stack (Fig. 4, compare e (VSV-G) to f (L. culinuris lectin)).
Rather, in >90% of co-transfected cells ts045-VSV-G protein accumulated in punctate structures throughout the cytoplasm alone ( a d ) or co-transfected with both PAR ts045-G and the ARFl(Q71L) expression construct (e-h) as described under "Experimental Procedures." Cells were incubated for 4 h at the restrictive temperature (39.5 "C) and then shifled to the permissive temperature for 1.5 h to allow release of ts045-VSV-G from the ER and transport to the Golgi apparatus. Cells were fixed, saponin permeabilized, and the distribution of tsO45-VSV-G (a, c, e, andg), Golgi membranes (L. culinaris lectin) (6, d, and f ) . and p53 ( h ) were viewed by double immunofluorescence microscopy as described under "Experimental Procedures." a and 6, distribution of ts045-VSV-G ( a ) and Golgi membranes (6) after 5.5 h continuous incubation at the restrictive temperature (39.5 "C). c and d, distribution of ts045-VSV-G (c) and Golgi membranes ( d ) aRer incubation for 1.5 h at the permissive temperature (32 "C). e, fi g, and h, overlapping distribution of ts045-VSV-G (e andg), Golgi membranes (f), and p53 ( h ) in cells overexpressing the ARFl (Q71L) mutant. Arrowheads indicate the juxta-nuclear Golgi region. Small arrows show punctate structures in which ts045-VSV-G and p53 colocalize (e, g, h ) or the distribution of the Golgi stack ( b d ) . Large bold arrows denote co-transfected cells. (Fig. 4e, arrows), and in punctate structures within and around the compact, juxta-nuclear Golgi stack (Fig. 4e, arrowhead). To address the possibility that this distribution of ts045-VSV-G might reflect a restricted transport to pre-Golgi intermediates, we examined whether the ts045-VSV-G found in punctate structures in the presence ofARFl(Q71L) overlapped with p53 (Schweizer et al., 1988(Schweizer et al., , 1990. p53 preferentially resides in pre-Golgi transport intermediates composed of clusters of vesicles and tubular elements which are distributed throughout the cytoplasm and enriched in the juxta-nuclear Golgi region (Schweizer et al., 1990). As shown in Fig. 4 (compare g (VSV-G) to h (p53)), ts045-VSV-G and p53 show a striking overlap both in terms of peripheral punctate staining as well as in punctate staining in the juxta-nuclear Golgi region. These results suggest that ts045-VSV-G is efficiently transported to pre-Golgi intermediates, but fails to enter the cis Golgi compartment. They are also consistent with the ability of the ARFl(Q71L) mutant protein to inhibit the processing of VSV-G to the Golgi associated R1 (and &) forms at early time points of transient expression (Figs. 2, A and B , 3B). Since the juxtanuclear Golgi stack typically remained intact (Fig. 4f), overexpression of the GTP-bound form of AFtF appears to have no effect on Golgi structure, at least as detectable using indirect immunofluorescence.
A different morphological phenotype was obtained by overexpression of the ARFl(T31N) and ARFl(N1261) mutants. In the case of the AFtFl(TBlN), ts045-VSV-G failed to mature to the position normally observed for the juxta-nuclear Golgi stack, being retained in a more uniform distribution characteristic of the ER and perinuclear ER (Fig. 5, a (VSV-G) and b (L.  culinaris lectin)). In this figure the large bold arrows indicate cells co-transfected with VSV-G and the mutant plasmid; the small arrows denote the position of the Golgi stack revealed by lectin staining. In some cells we also observed partial maturation of VSV-G to punctate intermediates which overlapped with p53 (Fig. 5, c (VSV-G) and d (~531, arrows). However, the extent of maturation of VSV-G to punctate, pre-Golgi intermediates was, qualitatively, considerably less than that observed in the presence of the ARFl(Q71L) mutant (Fig. 4, g and h). In the presence of ARFl(T31N1, -60% of the cells on the coverslip showed a distribution reflecting complete retention in the ER, -3040% showed partial (weak) maturation to dispersed pre-Golgi intermediates and only 10-15% of transfected cells expressing ts045-VSV-G showed transport to the juxta-nuclear Golgi position. These results are likely to reflect the differential levels of expression of the ARFl mutant forms in individual cells as a consequence of variation in the efficiencies of transfection, but emphasize the effects of this mutant on inhibition of export from the ER.
A similar distribution of ts045-VSV-G to that of cells overexpressing ARFl(T31N) was observed in the presence of the ARFl(N1261) mutant. The prominent morphological phenotype of this mutant (>60% of transfected cells examined) was also the apparent inhibition of export from the ER (Fig. 5e). However, we noted that in most (>70-80%) of the cells expressing either the ARFl(N1261) or the ARFl(T31N) mutants, the strong juxta-nuclear lectin staining pattern normally observed in nontransfected cells (for example Fig. 5, b and f, small arrows) was either absent or considerably more diffuse in transfected cells (Fig. 5, b and large arrow), suggesting that the expression of these mutants may lead to disassembly of the Golgi stack.
The morphological effects of the T31N and N126I mutants were very similar to those obtained when cells are treated with the drug BFA, which causes the inhibition of ER to Golgi transport, collapse of the Golgi into the ER, and release of the coatomer subunit p-COP from Golgi membranes (Donaldson et al., 1990(Donaldson et al., ,1991Lippincott-Schwartz, 1993). To address the possibility that these mutants were indeed mimicking the effects of BFA, we first examined the effects of BFA on the redistribution of lectin-positive Golgi membranes and p-COP in HeLa cells. In these experiments we utilized ts045 VSV-G to identify cells co-expressing the respective mutant protein (Fig. 6, a, c, e,   g, and i, large bold arrows). For clarity, cells were kept at the restrictive temperature throughout the duration of the experiment to retain VSV-G in the ER. Similar to results observed in other cell lines, BFA caused the compact, juxta-nuclear Golgi stack to disperse (data not shown) and the redistribution of p C 0 P to the cytosol (Fig. 6, compare b (-BFA, small arrows) to d (+BFA)) in both transfected and nontransfected cells.
An identical result to BFA was observed in cells co-transfected with either the ARFl(T31N) or ARFl(N1261) mutants. In this case, p-COP was completely dispersed to a diffuse cytoplasmic staining pattern (Fig. 6, h (T31N) a n d j (N12611, large bold arrows). However, we have observed that in cells expressing theARFl(N1261) mutant, that p-COP was frequently found in punctate aggregates scattered throughout the cytoplasm (Fig. 6j, see cell with large bold arrow). These structures did not overlap with compartments of the secretory pathway (data not shown), reminiscent of the aggregation of p-COP when cells are depleted of ATP (Hendricks et al., 1993). In contrast, when the distribution of PCOP in cells expressing the ARFl(Q71L) mutant was examined, PCOP was strongly associated with the compact, juxta-nuclear Golgi stack in >90% of the transfected cells examined (Fig. Sf, compare p-COP distribution (small arrows) in transfected cells (large bold arrow) to nontransfected cells). In these cells, p-COP could also be detected in punctate structures which overlap with p53 containing pre-Golgi intermediates ( Fig. 6L small arrowheads) (data not shown). Cells expressing the Q71L mutant which were shifted to the permissive temperature showed strong overlap of p-COP with pre-Golgi intermediates containing VSV-G (data not shown).

BFA Diggers Partial Processing of VSV-G to the RI Form in
HeLa Cells-It is now well recognized that BFA triggers the delivery of Golgi-associated al,2-mannosidases and glycosyltransferases normally residing in the cidmedial Golgi compartments to the ER (Lippincott-Schwartz et al., 1989;Doms et al., 1989;Lippincott-Schwartz, 1993). To examine whether the residual processing of VSV-G to the R1 form in cells expressing theARFl(T31N) or ARFl(N1261) mutants (Fig. 2 B ) was possi- Structure   FIG. 7. VSV-G protein is processed to the endo €I-resistant R1 form in BFA-treated cells. HeLa cells infected with recombinant vaccinia virus were transfected with a plasmid encoding VSV-G protein (PAR-G) for 5.5 h as described under "Experimental Procedures." Cells were radiolabeled and chased in the absence (ctl ) or presence of BFA (10 pg/ml final concentration). After chase, cells were transferred to ice and the processing of VSV-G protein was analyzed as described in the legend to Fig. 2 and under "Experimental Procedures." bly due to the retrograde delivery of these cidmedial Golgi enzymes to the ER, we examined whether VSV-G in transfected HeLa cells could be processed to the R1 form in the presence of BFA. As shown in Fig. 7, in the presence of BFA -25-30% of the VSV-G accumulated in the endo H-resistant R1 form. In contrast, no processing was detected to the % form (Fig. 7). Thus, the morphological and biochemical phenotypes observed with the ARFl(T31N) and ARFl(N1261), but not the ARFl(Q71L1, mutants are indistinguishable from the effects of BFA on vesicular traffic and Golgi structure. These results suggest that the T31N and N1261 ARF mutants may inhibit VSV-G protein vesicular traffic by preventing the recruitment of ARF and p-COP to the ER and Golgi membranes, resulting in cessation of vesicle formation, and, similar to BFA, disassembly of the Golgi apparatus.

DISCUSSION
Cumulative evidence now strongly implicates ARF as a molecular switch in the regulation of vesicular traffic through early compartments of the exocytic pathway (reviewed in Lippincott-Schwartz (1993)). In the present studies, we applied a novel in vivo model system to analyze the relationships between VSV-G transport from the ER through sequential Golgi compartments and ARF function. The function of ARF was dissected on the hypothesis that different substitutions in three highly conserved guanine nucleotide-binding domains found within all members of the ras superfamily would alter the steady-state "switch" status of the protein, thus enabling us to generate unique trans dominant phenotypes blocked in the GDP, GTP, or dominant negative configurations. These, in turn, would be anticipated to block the normal guanine nucleotide cycle necessary for ARF function. Fig. 8 summarizes our results on the role of ARF in protein transport, on Golgi structure, and on the binding of p-COP to membranes. These results provide the first in vivo demonstration that ARF and its cognate exchange protein are essential for membrane structure and vesicular traffic.

Q71L Enhances p-COP Association with Membranes and
Promotes Vesicle Formation-The substitution of Q for L at position 71 in ARFl is analogous to the transforming H-ras Q6lL mutation in the second highly conserved guanine nucleotide-binding region which is essential for the function of all  (Table I). In the case of ras, this substitution reduces its intrinsic GTP hydrolytic activity significantly, rendering the protein constitutively active (GTP-bound form) (Der et al.,  1986). An identical biochemical phenotype was observed for ARFl(Q71L1, suggesting that this protein is likely to be constitutively in the GTP-bound form. In the presence of the ARFl(Q71L) mutant, although VSV-G failed to be processed to the early Golgi R1 form, it was efficiently exported from the ER to punctate structures which co-localized with p53, a marker for pre-Golgi intermediates. Even more striking was the ability of this mutant to significantly enhance the binding of p-COP to the Golgi stack and to ER to Golgi transport intermediates.
The effects of the ARFl(Q71L) mutant on vesicular traffic is similar to the effects of GTP-yS on ER to Golgi and intra-Golgi transport in vitro (Beckers et al., 1989;Schwaninger et al., 1992). Incubation of either permeabilized cells (Beckers et al., 1990;Schwaninger et al., 1992)3 or isolated Golgi stacks (Melancon et al., 1987) in the presence of GTP-yS inhibits transport. The protein conferring GTP-yS sensitivity to transport between Golgi compartments has been recently demonstrated to be ARF (Taylor et al., 1992). In the presence of GTPyS, the PCOP containing coatomer complex and ARF are strongly recruited to the surface of the Golgi stack. As a consequence, Golgi compartments become decorated with numerous vesicle budding profiles and inactive, coated carrier vesicles (Melancon et al., 1987). More recently, the association of coatomer to the Golgi membrane in the absence of GTPyS has also been shown to require the participation of ARF (Donaldson et al., 1992a;Orci et al., 1993;Palmer et al., 1993) and an ARF-specific GEF (Helms and Fbthman, 1992;Donaldson et al., 1992b).
Our results provide more direct insight into the mechanisms of the GTP-bound form of ARF in the recruitment and function of p-COP in ER to Golgi transport. First, like GTP@, which is exchanged onto the endogenous ARF pool via GEF to generate the inhibitory phenotype, the Q71L may be recruited by GEF in the GDP-bound form, but be subsequently unable to hydrolyze GTP in a time frame compatible with vesicular traffic. Alternatively, during synthesis ARF is likely to be loaded with GTP due to the fact that the steady-state guanine nucleotide pool in cells is largely GTP. In this instance, ARF(Q71L) may bypass the requirement for GEF, being directly recruited to the membrane surface either via specific receptors which recognize the "active" GTP-bound form, or simply through an initial more nonspecific stable association with the lipid bilayer as a consequence of its active conformation . In either case, a n excess ofARFl(Q71L) on the membrane would provide a stable, saturating pool of the GTP-bound form of ARF to recruit coat components and initiate vesicle formation, suggesting that GTP hydrolysis by ARF per se is not required for budding. A prediction of this interpretation is that the distri-bution of p-COP in cells transfected with the ARFl(Q71L) mu-rab family are dominant lethal (Schmitt et al., 1986) or are tant will be insensitive to BFA, a result which we have ob-defective in transport between exocytic compartments in vivo ~e r v e d .~ However, since these carrier vesicles are defective in (Walworth et al., 1991;Tisdale et al., 1992) and in vitro.3 Altargeting or fusion to a downstream acceptor compartment, though the ARFl(N1261) mutant protein exhibited dominant these results suggest that GTP hydrolysis promoted by a mem-negative activity, its biochemical and morphological phenobrane-associated ARF-specific GAP is essential for a down-types were not like those observed for the Q71L mutant, but stream event either prior to or during vesicle docking andor were similar to those observed for the T31N mutant leading to fusion. Inhibition of targeting and hsion could also be due to retention of VSV-G in the ER, collapse of the Golgi, and partial the inability of ER to Golgi carrier vesicles to fuse with the cis processing of VSV-G to the R1 form. However, ARFl(N1261) Golgi compartment due to an unusual accumulation of coat differed from ARFl(T31N) in causing the accumulation of proteins on the acceptor Golgi compartment in the presence of PCOP in small dots which were distributed throughout the the ARFl(Q71L) mutant. Given the ability of the 2 Ala substi-cytoplasm and which may represent insoluble, p-COP containtution to prevent ARFl(Q71L) inhibition, myristoylation is ing aggregates. clearly essential for association of ARFl(GTP) with the mem-There are a number of explanations currently being enterbrane in vivo.
tained to account for the phenotype of this mutant. One possi-T31N Inhibits Export of VSV-G from the ER and Prevents bility is that it is recruited to membranes via an ARF-GEF. Coat Assembly-The substitution of N for T at position 31 is However, due to its high exchange rate, it may inhibit GEF analogous to the growth inhibitory S17N mutant of H-ras (Feig function and may be unable to establish or maintain a stable and Cooper, 1988). This ras mutant shows preferential affinity association between p-COP and other transport components on for GDP over GTP and is thought to be preferentially con-the membrane surface. As a consequence, coats fail to assemble strained to the GDP-bound inactive form. This mutant has and a BFA-like effect ensues. Alternatively, the mutant could been proposed to antagonize normal H-ras function by seques-bind directly to the soluble pool of p-COP, leading to aggregatering the endogenous pool of ras-GEF (Farnsworth et al., tion and creating a p-COP "null" phenotype. 1991). Analysis of GTP binding to ARFl(T31N) suggests a simi-Multiple GTP-binding Proteins Are Required for Export from lar biochemical phenotype. Given the ras paradigm, we propose the ER-The combined results from the above experiments sugthat the ARFl(T3 1N) similarly prevents the endogenous pool of gest that A R F l plays a critical role in the recruitment of p-COP ARFl in the cell from interacting with an ARF1-GEF. Such a containing coat complexes from the cytosol in vivo which are mutant would be anticipated to efficiently prevent coat assem-involved in initiating vesicular traffic from the ER and the bly and export of VSV-G from the ER (Fig. 8). Although this maintenance of Golgi function and structure. These results are mutant has a preferential affinity for GDP, it, like H-ras (Feig consistent with our recent observations that antibodies specific and Cooper, 19881, can still bind GTP, suggesting that even a for p-COP also inhibit the recruitment of a P-COPIrabl conlow level of exchange is insficient to maintain function of the taining pre-coat complex and block vesicle budding from the ER vesicular traffic pathway. (Peter et al., 1993). In addition, we now have evidence that a In contrast to the effect of theARFl(Q71L) mutation on Golgi homologue to yeast SARl (Nakano and Muramatsu, 1989; Oka structure or the distribution of p-COP, the ARFl(T31N) mutant et al., 1991;d'Enfert et al., 1991;Barlowe et al., 1993) is also triggered collapse of the Golgi into the ER, redistribution of required for vesicle budding from the ER in mammalian cells.5 PCOP to the cytosol, and promoted processing of VSV-G to the Thus, at least 3 GTP-binding proteins, ARF1, rabl, and SARI early Golgi-like forms (Fig. 8). Since the predicted phenotype of are required for export from the ER. The individual functions of ARFl(T31N) is to "inactivate" an ARF-specific GEF, our results these three molecular switches must be ultimately integrated are consistent with the recent suggestion that BFA may also to choreograph the recruitment and concentration of cargo durinactivate either directly and indirectly an ARF-specific GEF ing export from the ER3 and coat assemblyIdisassemb1y during (Donaldson et al., 1992b;Helms and Rothman, 1992), and em-subsequent downstream budding, targeting, and fusion events. phasize an essential role for ARF and ARF-GEF in the dynamic Experiments to elucidate their sequential roles in each of these turnover of p-COP on both the ER and the Golgi compartments. processes are currently in progress. Since the 2 Ala substitution also neutralizedARFl(T31N) function, myristoylation appears to be essential for interaction be-Acknowze&ment-we thank R. Kahn for generously SuPPlying the tween ARF and its specific GEF in vivo. Moreover, these results suggest that either the retrograde transport pathway does not