Direct and GTP-dependent Interaction of ADP-ribosylation Factor 1 with Clathrin Adaptor Protein AP-1 on Immature Secretory Granules*

ADP-ribosylation factor 1 (ARF1) mediates clathrin coat formation on PC12 immature secretory granules (ISGs). We have used two approaches to investigate whether ARF1 interacts directly with the clathrin adaptor protein, AP-1. Using an in vitrorecruitment assay and co-immunoprecipitation, we could isolate an AP-1·ARF1 complex. Then we used a site-directed photocross-linking approach to determine the components that act downstream of ARF1 in clathrin coat formation on ISGs. Myristoylated ARF1, with a photolabile phenylalanine analogue incorporated into its putative effector domain (switch 1), showed a specific, GTP-dependent interaction with both the γ- and β−adaptin subunits of AP-1 on ISGs. These experiments provide evidence for a direct interaction of ARF1 with AP-1. On mature secretory granules myristoylated ARF1 does not bind, and hence clathrin coat formation cannot be initiated, supporting the hypothesis that molecules involved in coat recruitment are removed during ISG maturation.

The assembly of clathrin coats on the cytoplasmic face of intracellular membranes occurs at specific sites. Ultimately, the site of coat assembly becomes a transport vesicle as the membrane is deformed by the rearrangement of the clathrin triskelion, driven by interaction with other coat constituents (for recent review see Refs. 1 and 2). The mechanism underlying assembly of the clathrin coat at specific sites involves a complex set of protein-protein and protein-lipid interactions that is not fully understood. At all intracellular sites where clathrin assembles, the minimum machinery is the adaptor proteins (APs) 1 and clathrin. Four types of adaptors, AP-1 through AP-4 (1,3,4) have been identified to date. All adaptors are heterotetrameric complexes of two large subunits with a molecular mass of ϳ100 kDa, a medium subunit (ϳ50 kDa) and small subunit (ϳ20 kDa). The first identified and most studied adaptors are AP-1 and AP-2. AP-1 consists of the large subunits ␥ and ␤1, medium chain 1, and small chain 1 and is found primarily on the Golgi complex (5) and immature secre-tory granules (6), whereas AP-2, which consists of the ␣, ␤2, 2, and 2 subunits, is localized mainly to the plasma membrane and endosomes (7). The localization of AP-1 and AP-2 to different compartments can, in part, be attributed to their ability to bind to different sequences in the cytoplasmic domains of trans-membrane receptors (for recent reviews see Refs. 8 and 9).
ARF1 recruits AP-1 to both ISG membranes (6) and the trans-Golgi network (10,11). ARFs are a family of small molecular mass (ϳ20 kDa) GTP-binding proteins that associate with lipid bilayers and membranes via a myristoylated, hydrophobic amino terminus (12,13). ARF in the GDP-bound form is cytosolic. Interaction of ARF with its exchange factor allows GDP to be exchanged for GTP, and this causes a major structural change in the switch 1 domain of ARF1 (14), thus exposing the myristic acid and promoting interaction of ARF with the membrane. ARF1 has been identified as a component of Golgi-derived nonclathrin coated (COP-coated) vesicles (15) and is required for the formation of COPI vesicles (16). However, unlike COPcoated vesicles, ARF is not present in clathrin-coated vesicles in sufficient quantities to be a stoichiometric component of the coat, although it has been shown to recruit AP-1 in a stoichiometric manner (17). Furthermore, ARF1 in the presence of GTP␥S generates high affinity binding sites for AP-1 that are stable even in the presence of high concentrations of Tris (17). To investigate whether ARF1 interacts with AP-1, we used an in vitro recruitment assay with [ 35 S]myrARF1, followed by co-immunoprecipitation with antibodies directed against AP-1 and discovered an AP-1⅐ARF1 complex on ISG membranes.
A second approach using site-specific cross-linking was employed to start to understand the nature of the ARF1 interaction with AP-1 on ISG membranes. Recent work using this approach has demonstrated that the switch 1 domain of ARF1 interacts directly with COPI (18). ARF1, with amino acid Ile-46 replaced with the photoactivatable analogue of phenylalanine ((Tmd)Phe) (19), cross-linked to both ␤and ␥-COP (20). Using the ARF1-(Tmd)Phe-46 mutant, we performed cross-linking experiments with ISGs as the acceptor compartment for ARF and AP-1. We have found that on ISG membranes ARF1 can be cross-linked to AP-1 in a GTP-dependent manner via the ␥and ␤1-adaptin subunits of AP-1, and this interaction is strictly dependent on the presence of ISG membranes. Mature secretory granules (MSGs) or liposomes do not facilitate a direct interaction between ARF and AP-1. Our data show that the switch 1 domain of ARF1, encompassing Ile-46, interacts with AP-1 as well as COPI.

EXPERIMENTAL PROCEDURES
ISG Preparation-ISGs and MSGs were prepared from PC12 cells by velocity and equilibrium sucrose gradient centrifugation, as described previously (6). Each gradient generated 3 ml of ISGs and 2 ml of MSGs, 1 ⁄24 of which was used per reaction.
PC12 ISG ␥-Adaptin Binding Assay-The binding assay was performed as described elsewhere (6). Briefly, 125 l of ISGs were incubated in binding buffer with purified AP-1 and recombinant myrARF1 in the presence or absence of 100 M GTP␥S in a total volume of 250 l. After incubation at 37°C for 30 min the samples were diluted 4-fold in binding buffer, centrifuged (100, 000 g for 1 h), and analyzed by immunoblotting with subsequent incubations of the monoclonal antibody (mAb) anti-␥-adaptin, 100/3 (Sigma), rabbit anti-mouse antiserum, and 125 I-protein A (Amersham Pharmacia Biotech) as described (6). The radioactivity bound to the ISGs was quantified using a PhosphorImager system (Molecular Dynamics). All assays were done at least in duplicate.
In Vitro Transcription and Translation-The human ARF1 cDNA (pET-11d vector) was in vitro transcribed with T7 RNA polymerase (Promega) and translated using the rabbit reticulocyte lysate based Flexi-lysate kit (Promega) in the presence of [ 35 S]methionine (Amersham Pharmacia Biotech) according to the manufacturer's protocol. ARF1 was labeled during the in vitro translation reaction with 1 Ci/l [ 3 H]myristic acid (Amersham Pharmacia Biotech) instead of [ 35 S]methionine as described (28).
PC12 ISG Binding Assay Using [ 35 S]ARF and Subsequent Immunoprecipitation-25 l of ISGs were incubated with 5 l of in vitro translated [ 35 S]ARF with 100 M GTP␥S or GDP␤S in a total of 50 l at 37°C for 30 min. The ISGs were sedimented, solubilized, and immunoprecipitated under native conditions (see below) with either preimmune serum or the ␥-adaptin antiserum, STO-25 (6). Samples were analyzed by SDS-PAGE, quantified using a PhosphorImager system, and then subjected to autoradiography.
Synthesis of Site-specific Photolabile ARF Mutants-The cDNA encoding human ARF1, inserted into the pET-11d plasmid, with amino acid Ile-46 replaced by the amber stop codon (TAG), was obtained from Felix Wieland (Biochemie-Zentrum Heidelberg, Heidelberg, Germany) (20). For the samples without membranes, a volume of 25 l was used; these were not centrifuged but irradiated directly. Samples were irradiated as described (18) and then either trichloroacetic acid precipitated and analyzed by SDS-PAGE and autoradiography or immunoprecipitated.
Antibodies and Immunoprecipitation-Antibodies used include: rabbit polyclonal antiserum anti-␥-AP, STO-25 (6); rabbit polyclonal antisera against ␤-COP and ␥-COP (a gift from F. T. Wieland); mAb anti-␥-AP, 100/3 (Sigma); mAb anti-␤-AP (a gift from T. Kirchhausen, Harvard Medical School, Cambridge, MA) (29); mAb anti-␣-AP, 100/2 (Sigma); and mAb anti-ARF (a gift of R. Kahn). For immunoprecipitation under native conditions, cross-linked ISG pellets were solubilized in 250 l of 20 mM Tris-HCl, pH 7.5, 2 mM EDTA, 0.15 M NaCl, 0.5% (w/v) Triton X-100 (TNTE buffer). For denaturing conditions, the samples were boiled for 3 min in the presence of 1% SDS and then diluted as for the native conditions except that the Triton X-100 concentration was increased to 0.9%. Rabbit anti-mouse antiserum was added to all incubations containing mAbs. The immune complexes were bound by protein A-Sepharose CL-4B (Amersham Pharmacia Biotech). After washing in TNTE buffer and once with 20 mM Tris-HCl, pH 7.5, samples were solubilized in sample buffer and analyzed by Western blotting and autoradiography.
Removal of AP-1 and COPI from Lysate-We used a method previously shown to deplete 80% of COPI from rat liver cytosol (30). 110 l of the translation mix containing in vitro translated [ 35 S]ARF-(Tmd) Phe-46 was placed over a 40-l 0.5 M sucrose cushion (in binding buffer) and centrifuged at 300,000 ϫ g using a TLA-100 rotor (Beckman) for 2 h at 4°C.

RESULTS
Purified Myristoylated ARF1 Regulates AP-1 Binding to the ISGs-It has previously been shown, using a cell-free assay system and partially purified components, that ARF1 is required for ␥-adaptin binding to PC12 ISGs (6). We have reproduced the ARF1-dependent AP-1 recruitment to ISGs using purified components (Fig. 1). Quantification relies on the use of the mAb 100/3 that recognizes bovine but not rat ␥-adaptin; hence only the bound exogenous bovine ␥-adaptin is detected. Both purified bovine AP-1 and purified recombinant myrARF1 were titrated into the binding assay (Fig. 1A). We used bovine AP-1 because we could obtain large amounts of tissue for the purification. Our previous experiments have shown that bovine AP-1 binds to ISGs as efficiently as rat AP-1 (6). Human, bovine, and mouse ARF1 are 99% identical; thus it was assumed that the human ARF1 would bind as efficiently to rat ISGs as the endogenous rat ARF1. Increasing the myrARF1 concentration up to 20 g/ml resulted in an increase in the ␥-adaptin bound to the ISGs at all concentrations of AP-1 used except 1 g/ml. At 1 g/ml AP-1, the ␥-adaptin binding was saturated at 4 g/ml myrARF1, this may be due to limiting amounts of AP-1. In the absence of added ARF1 there is a basal level of ␥-adaptin binding (Fig. 1B); this is likely to be due to endogenous ARF1 already bound to the ISGs and is similar to the level observed in the presence of ARF1 but in the absence of any nucleotide. The addition of GTP␥S increases ␥-adaptin binding to ISGs 3-4-fold but only in the presence of ARF1 (Fig.  1B). We did not observe GTP-dependent binding of nonmyristoylated ARF1 to the membranes (results not shown). Hence, the only cytosolic component required for purified AP-1 to bind to the ISGs is GTP-bound myrARF1.
In Vitro Translated [ 35 S]myrARF1 Interacts with AP-1 on the ISGs in a GTP-dependent Manner-To monitor ARF1 binding to the ISGs, we wanted to generate [ 35 S]myrARF1 by in vitro translation using rabbit reticulocyte lysate. To demonstrate that the ARF1 was myristoylated during the translation reaction, ARF1 mRNA was in vitro translated in the presence of [ 3 H]myristic acid. This resulted in a 3 H-labeled ϳ20-kDa band whose migration was identical to purified recombinant myrARF1 ( Fig. 2A, lane 1). Others have also observed N-myristoyltransferase activity (28,32,33) in the rabbit reticulocyte lysate. In vitro translation of ARF1 mRNA in the presence of [ 35 S]methionine resulted in a broader ϳ20-kDa band (Fig. 2A, lane 2) that, after a lower exposure time, is visible as a doublet (not shown) corresponding to myristoylated and nonmyristoylated [ 35 S]ARF1. These results confirm that ARF1 is myristoylated during the in vitro translation reaction.
To determine whether AP-1 and ARF1 form a complex, ISG binding assays were performed in the presence of [ 35 S]ARF1. After incubation of [ 35 S]ARF1 with the ISGs in the presence of GTP␥S or GDP␥S, immunoprecipitation of ␥-AP was performed with polyclonal, STO-25 to determine whether the [ 35 S]myrARF1 was interacting with AP-1 (Fig. 2, B and C). The myristoylated form of [ 35 S]ARF1 was specifically co-immunoprecipitated with AP-1 in the presence of GTP␥S (Fig. 2B). The amount of [ 35 S]myrARF1 coimmunoprecipitated in the presence of GTP␥S was twice that in the presence of GDP␤S and 4-fold more than preimmune controls.
Approximately 1% of the total membrane bound [ 35 S]myrARF1 added to the assay could be co-immunoprecipitated with anti-␥ϪAP antibodies. These results demonstrate that AP-1 and myrARF1 do form a complex on the ISG membrane.

Membrane-bound [ 35 S]ARF-(Tmd)Phe-46
Cross-links to 100 -120-kDa Coat Proteins-To investigate whether ARF1 interacts directly with AP-1, we decided to use site-directed photocross-linking, as used previously by others to demonstrate a direct interaction of the nonclathrin coat, COPI and ARF1 (18).  Fig. 3A, upper panel, lanes 3 and 7). Addition of AP-1 appeared to increase the intensity of the 120-kDa bands (Fig. 3A, lane 7). This was consistent with cross-linking of ARF (ϳ20 kDa) to one or both of the large subunits of AP-1 (ϳ100 kDa). The photocross-linked products are specific because they were not observed in the presence of GDP␤S (Fig. 3A, upper panel, lanes 1 and 5), or in the absence of irradiation (Fig. 3A, upper panel, lanes 2 and 6). Displacement of the membrane-bound [ 35 S]ARF-(Tmd)Phe-46 using excess unlabeled recombinant wild type myrARF1 (Fig. 3A, lower  panel, lanes 4 and 8) abolished the photocross-linking (Fig. 3A,  upper panel, lanes 4 and 8). Importantly, the cross-linked products were not observed in the absence of ISG membranes either in the presence or absence of added AP-1 (Fig. 3A, upper panel,  lanes 9 and 10), demonstrating that only membrane-bound myrARF1 interacts with the 100 -120-kDa proteins and that ARF1⅐AP-1 does not interact in solution.
Immunoprecipitation of the specific cross-linked products confirmed an interaction of myrARF1 with AP-1. Immunoprecipitation of the whole AP-1 complex with the ␥-AP polyclonal, STO-25 (6) under native conditions resulted in a 35 S-labeled doublet (Fig. 3B, lane 3), whereas immunoprecipitation of the dissociated subunits under denaturing conditions showed a single band of lower intensity corresponding to ␥-AP (Fig. 3B,  lanes 4). Because the pattern of cross-linked products derived from [ 35 S]ARF-(Tmd)Phe-46 bound to ISGs were similar to that observed with Golgi membranes in the presence of coatomer (20), antisera to ␤and ␥-COP were used to characterize the other cross-linked products observed with the ISG membranes. Similar to the result obtained for Golgi membranes (20), anti-␤-COP immunoprecipitated a band of 120 kDa (Fig. 3B, lane 5), and the anti-␥-COP immunoprecipitated two bands of approximately 120 and 140 kDa (Fig. 3B, lane 6).

Membrane-bound [ 35 S]ARF-(Tmd)Phe-46 Can Directly
Interact with ␥and ␤1-AP-To increase the efficiency of crosslinking of [ 35 S]ARF1-(Tmd)Phe-46 to AP-1 on the ISGs, purified bovine AP-1 was titrated into the photocross-linking binding assay (Fig. 4A, lanes 1-3). The cross-linking pattern showed an increase in the intensity of the labeled 120-kDa bands with increasing AP-1 (Fig. 4A, top panel, lanes 1-3). Immunoprecipitation of the whole AP-1 complex with 100/3 confirmed an increase in the efficiency of cross-linking of [ 35 S]ARF1-(Tmd)Phe-46 to AP-1 (Fig. 4A, middle panel, lanes  1-3). Increasing the [ 35 S]ARF1-(Tmd)Phe-46 caused a further increase in the intensity of the cross-linked products (Fig. 4A,  top panel, lanes 6 and 7), which is due to increased crosslinking to AP-1 (Fig. 4A, middle panel, lanes 6 and 7). Because the mAb 100/3 only recognizes bovine and not rat AP-1, it was surprising that, in the absence of exogenous bovine AP-1, 100/3 could immunoprecipitate a cross-linked product (Fig. 4A, middle panel, lane 1). This suggested that there might be AP-1 in the rabbit reticulocyte lysate used to generate the [ 35 S]ARF1-(Tmd)Phe-46. Western blot analysis demonstrated that indeed this was the case (data not shown). Using purified bovine AP-1 as a standard and assuming equal cross-reactivity of 100/3 with rabbit AP-1 it was estimated that the rabbit reticulocyte lysate contains ϳ16 g/ml AP-1, giving a final concentration of ϳ1 g/ml rabbit AP-1 in the binding assay. AP-2 and COPI were also detected in the lysate using polyclonal anti-␤-COP and  1-4 and 9) or presence (lanes 5-8 and 10) of 4 g/ml purified AP-1 or with 40 g/ml recombinant myrARF1 (lanes 4 and 8). After incubation at 37°C for 30 min, the membranes were sedimented and irradiated (except samples lanes 2 and 6) and trichloroacetic acid precipitated. Then 9 ⁄10 were analyzed by 7.5% SDS-PAGE and autoradiography (upper panel) to visualize the cross-linked products, and 1 ⁄10 was analyzed by 15% SDS-PAGE and autoradiography (lower panel) to visualize the uncross-linked [ 35  anti-␥-COP and may explain the source of ␤and ␥-COP crosslinked to ARF1 on the ISG membranes. Thus, the rabbit reticulocyte lysate acts like a cytosol, providing the adaptor proteins and coatomer for recruitment to ISGs. Increasing the amount of [ 35 S]ARF1-(Tmd)Phe-46 resulted in an increased binding (Fig. 4A, bottom panel, lanes 4 -7), whereas increasing AP-1 had no effect on the amount of [ 35 S]ARF1-(Tmd)Phe-46 bound to the membrane (Fig. 4A, bottom panel, lanes 1-3). This suggests that AP-1 does not influence [ 35  Immunoprecipitation with STO-25 under native and denaturing conditions (Fig. 3B, lane 3 and 4) implied that both ␥and ␤1-AP may interact with ARF. The photocross-linking assay in the absence of added bovine AP-1 contains endogenous rat AP-1 from the PC12 ISGs and exogenous rabbit AP-1 from the rabbit reticulocyte lysate. Anti-␥-AP antiserum, STO-25 recognizes rat and, to a lesser extent, rabbit ␥-AP, whereas anti-␥-AP antibody 100/3 recognizes rabbit and not rat ␥-AP. Quantative immunoprecipitation with STO-25 (Fig. 4B, lane 3) and 100/3 (Fig. 4C, lanes 3 and 5) under denaturing conditions demonstrate that both rat and rabbit ␥-AP directly bind to [ 35 S]ARF1-(Tmd)Phe-46. Immunoprecipitation of the whole AP-1 complex with the same antisera under native conditions revealed bands of greater intensity than that immunoprecipitated under denaturing conditions, again suggesting crosslinking of [ 35 S]ARF1-(Tmd)Phe-46 to both ␤1and ␥-AP. This was confirmed by immunoprecipitation of the dissociated AP-1 subunits with the anti-␤-AP (Fig. 4, C, lane 6 and D, lanes 3  and 5). Because the anti-␤-AP antibody also recognizes both ␤1-and ␤2-AP, we cannot rule out from these immunoprecipitations that ␤2-AP also cross-links with myrARF1. However, [ 35 S]ARF1-(Tmd)Phe-46 does not interact with ␣-AP of AP-2 because no products were immunoprecipitated with anti-␣-AP under native or denaturing conditions (data not shown); therefore, it is unlikely that [ 35 S]ARF1-(Tmd)Phe-46 interacts with ␤2-AP. Although ARF1 has been shown to recruit AP-2 to endosomal membranes in a GTP␥S-dependent manner (34), we failed to detect a GTP-dependent association of AP-2 to ISGs, 2 consistent with the cross-linking data.
To make the cross-linking assay dependent on exogenous AP-1, we removed the high molecular mass coat proteins in the rabbit reticulocyte lysate, after in vitro translation of [ 35 S]ARF1-(Tmd)Phe-46, by high speed centrifugation. Most of the AP-1 and COPI remained in the cushion and was absent from the supernatant (Fig. 5A, top and middle panels, respectively), whereas most of the [ 35 S]ARF1-(Tmd)Phe-46 remained in the supernatant (Fig. 5A, bottom panel). Greater than 90% of AP-1 and COPI were removed. Some [ 35 S]ARF1-(Tmd)Phe-46 was lost in the sucrose cushion (Fig. 5A, bottom panel, lane 2), and as a result, for the cross-linking assay, the amount of [ 35 S]ARF1-(Tmd)Phe-46 was normalized to that in the crude translation mix. The [ 35 S]ARF1-(Tmd)Phe-46 in the depleted lysate bound to the ISGs in a GTP␥S-dependent manner (Fig.  5B, bottom panel, compare lanes 4 and 5). However, the amount bound was significantly lower than that observed for the crude in vitro translation mix (Fig. 5B, bottom panel,  was either kept constant (see "Experimental Procedures") and incubated in the presence of ISG membranes with increasing amounts of purified bovine AP-1 (lanes 1-3), or the AP-1 concentration was kept constant (4 g/ml) and the in vitro translated [ 35 S]ARF-(Tmd)Phe-46 (lanes 4 -7) was varied. All samples contained GTP␥S. After sedimentation and irradiation, the products (upper panel) were subjected to 7.5% SDS-PAGE and autoradiography, and the region of the specific photocross-linked products is shown (X). Alternatively, cross-linked samples were immunoprecipitated with the mAb anti- 46 added to the reaction was the same. The myrARF1 bound to the ISGs was still able to interact directly with exogenous bovine AP-1 as indicated by an increase in the cross-linking products with increasing bovine AP-1 (Fig. 5B, top panel, lanes  6 -8), and by immunoprecipitation of bovine AP-1 cross-linked to [ 35 S]ARF1-(Tmd)Phe-46 with mAb, 100/3 (Fig. 5B, middle  panel, lanes 5-7). As expected, in the absence of bovine AP-1, no cross-linked AP-1 was immunoprecipitated with 100/3 (Fig. 5B, middle panel, lane 5). However, cross-linking of [ 35 S]ARF1-(Tmd)Phe-46 to AP-1 was still detected in the absence of exogenous bovine AP-1 (Fig. 5B, top panel, lane 5). Because the PC12 ISGs used in the assay have AP-1 containing clathrin coats (6), the cross-linked product in Fig. 5B, top panel, (6). No GTP-dependent binding of [ 35 S]ARF1-(Tmd)Phe-46 was observed when using MSGs in the binding assay compared with ISGs (Fig. 6, lower panel), even using the crude in vitro translation mix containing the rabbit coat proteins. As a result, no cross-linking of [ 35 S]ARF1-(Tmd) Phe-46 to AP-1 or COPI was observed on MSGs compared with ISGs (Fig. 6, upper panel, compare lanes 2 and 5). In addition, we did not observe [ 35 S]ARF1-(Tmd)Phe-46 binding or crosslinking to AP-1 on liposomes using the crude in vitro translation mix or even after the addition of ARF-depleted bovine brain cytosol (18) prior to photocross-linking (results not shown). This is in contrast to previous results showing recruitment of AP-1 and clathrin onto liposomes (from soyabean lipids) using ARF-GTP and cytosolic factor(s) (35). Hence, the recruitment of myrARF1 and the AP-1/myrARF1 interaction is specific to ISGs, confirming our previous data that there are molecules on the ISGs that allow coat formation that are removed or inactivated during secretory granule maturation. DISCUSSION myrARF1 is required for the recruitment of AP-1 to ISGs (6) and ultimately for clathrin-coated vesicle formation during ISG maturation. Previous work showed that ARF1 is a limiting factor in the GTP-stimulated recruitment of AP-1 onto isolated Golgi membranes, and although ARF1 is not a stoichiometric component of the clathrin coat, it appears to function in a stoichiometric manner to generate high affinity binding sites for AP-1 (17). We have extended these findings by demonstrat- ing a direct interaction between AP-1 and myrARF1 in the GTP-bound form. We show that AP-1 and ARF1 form a complex using co-immunoprecipitation of [ 35 S]myrARF1 with ␥-adaptin from solubilized ISG membranes. We could estimate that 1% of the membrane bound myrARF1 is in a complex with AP-1. We exploited a site-directed photocross-linking approach to demonstrate that myrARF1 in the GTP form interacts directly with both the ␥and ␤1 subunits of AP-1. The interaction of ARF1, via Ile-46, which is in the switch 1 domain, the putative effector region of ARF1 (14), with AP-1 is consistent with data of Liang and Kornfeld (36); using chimeras of mammalian and yeast ARF they showed residues 35-94 were essential for AP-1 recruitment to the Golgi. Because the reticulocyte lysate contains significant amounts of AP-1, COPI, and AP-2 (data not shown), we removed the endogenous coat proteins from the lysate using high speed centrifugation and demonstrated that interaction of myrARF1 and AP-1 was dependent on the addition of exogenous AP-1. In addition, we could not detect an interaction of [ 35 S]ARF-(Tmd)Phe-46 and AP-1 in solution, either in the cross-linking assay in the absence of ISGs or by co-sedimentation of ARF1 with AP-1 during the centrifugation.
The photocross-linking method is qualitative and does not allow a quantification of the stoichiometry between ARF and AP-1. Cross-link yields are typically Ͻ1%, because the reactive carbene predominantly interacts with the water molecules (19). The photoprobe is highly specific, because a protein of very close proximity would be required to compete with the abundant water molecules. Because the rabbit reticulocyte lysate used for the in vitro translation also contains ϳ1 g/ml of unlabeled ARF proteins, as determined with the pan-ARF antibody, the specific activity of the [ 35 S]ARF-(Tmd)Phe-46 is further reduced. We are unable to detect by immunoblotting, with antibodies to AP-1 or COPI, the population of coat proteins that were photo cross-linked to myrARF1. Thus, we are unable to quantify the extent of the interaction between myrARF1 and AP-1 or COPI using the photocross-link approach.
Using the reticulocyte lysate, in the presence of GTP␥S, we observed cross-linking of COPI to [ 35 S]ARF-(Tmd)Phe-46, which was dependent upon membranes. The amount of ␤-COPI cross-linked is at least as great as AP-1. ␤-COP is present, but not enriched, in the ISG fraction. Although Golgi contamination in the ISG fractions has been excluded with assays for trans-Golgi network 38 (37), sialyltransferase assays and labeling with [ 35 S]sulfate (38), endosomal membrane contaminants maybe present in the ISG fraction (39), which may explain the cross-linking of ␥and ␤-COPI. Alternatively, the observations of Martínez-Mená rguez et al. (40), showing low but significant labeling of ISGs with COPI in the exocrine pancreas by immunoelectron microscopy, suggest that COPI binding sites may be present on ISG membranes. However, clathrin coats were far more abundant than the COPI coats, suggesting that COPImediated transport is a minor pathway and may be a remnant of the retrograde machinery from the Golgi complex.
Interaction of both AP-1 and COPI with the switch 1 domain of ARF1 suggests that common structural features exist between these multisubunit coat protein complexes. A phylogenetic analysis by Schledzewski et al. (41) provides evidence that the clathrin adaptors and F-COP (a subcomplex of COPI containing the subunits ␤, ␥, ␦, and ) have arisen from coordinated gene duplications of common ancestral genes and predicts that ␤-COP is homologous to ␤1-adaptin and ␥-COP to ␥-adaptin. The structural similarities between the adaptors and COPI could explain the cross-linking data of Zhao et al. (20) showing cross-linking of [ 35 S]ARF-(Tmd)Phe-46 to ␤and ␥-COP on Golgi membranes and the results presented here.
How can the switch 1 domain of ARF1 interact with two subunits of AP-1 or COPI? There is increasing evidence to suggest that ARF oligomerizes; crystals of ARF-GDP were found in a dimeric form (42,43), and cross-linking of ARF with itself has been detected using [ 35 S]ARF-(Tmd)Phe-49 (20). Furthermore, others have determined that there are multiple copies of membrane associated ARF-GTP per coatomer (15,45,46). Thus, it is possible that dimers of myrARF1 exist, where the switch I region of one molecule of ARF interacts with the ␥or ␤-subunit of one AP-1 or COPI and the other ARF molecule with the other large subunit. ISG maturation involves homotypic fusion and changes in the size and dense core (47) of the secretory granule. Unlike MSGs, which lack a clathrin coat, ISGs are partially coated vesicles (48,49), with AP-1 containing clathrin coats (50). Clathrin coats have been seen enveloping small vesicular structures on ISGs (49), leading to the hypothesis that clathrincoated vesicles serve to remove components of the secretory granule. We show that myrARF1 does not bind to MSGs and, using photocross-link, that AP-1 and myrARF1 do not interact on the MSG, suggesting that there are proteins on the ISG, required for myrARF1 binding and subsequent AP-1/ARF1 interaction, which cannot be supplied by the reticulocyte lysate or bovine brain cytosol. These proteins may be transmembrane proteins such as mannose-6-phosphate receptor or furin, both of which are present on ISGs but absent from MSGs (37,39). In support of this others have shown that mannose-6-phosphate receptors can stabilize AP-1 and ARF1 interaction on trans-Golgi network membranes (44), and a bivalent interaction of coatomer with membrane-bound ARF and cytoplasmic tails of cargo or putative cargo receptors has been demonstrated (45).