In vitro assembly and disassembly of coatomer.

Coatomer, a complex of seven proteins, is the major component of the non-clathrin (COP I) membrane coat. We report here the first system to reversibly disassemble and reassemble this complex in vitro. Coatomer disassembles at high salt concentrations and reassembles when returned to a more physiological buffer. Using this system, we show that alpha-, beta'-, and epsilon-COP interact directly and that gamma-COP interacts with zeta-COP. A partial complex comprising alpha-, beta'-, and epsilon-COP, obtained after coatomer disassembly, can bind to membranes in vitro. This binding is, at least in part, mediated by interactions with cytoplasmic KKXX motifs of proteins normally retained in or retrieved to the endoplasmic reticulum. Using coatomer disassembly and epitope-specific antibodies, we also demonstrate that the N- and C-terminal domains of beta-COP are buried within the native coatomer complex. These results provide the first insights into how the coatomer is structured.

icles bud from the plasma membrane and from the trans-Golgi network (Pearse and Robinson, 1990), non-clathrin-coated or COP I-coated vesicles are involved in transport between the intermediate compartment and the Golgi complex (Pepperkok et al., 1993) and in intra-Golgi transport (Ostermann et al., 1993), and COP II-coated vesicles participate in endoplasmic reticulum (ER) to Golgi transport (Barlowe et al., 1994).
Coatomer and ARF were originally identified using an in vitro intra-Golgi transport assay (Malhotra et al., 1989;Serafini et al., 1991). Good evidence now exists that these components function in transport to the Golgi complex in vivo. For example, mutations in the ␥-COP homologue of yeast coatomer, Sec21p, block transport between the ER and the Golgi complex (Hosobuchi et al., 1992). Microinjection of antibodies to ␤-COP into living mammalian cells blocks transport from the ER to the Golgi complex (Pepperkok et al., 1993). Interestingly, microinjected antibodies against ␤-COP also inhibit brefeldin A (BFA)-induced retrograde transport from the Golgi complex to the ER, 2 suggesting a role for coatomer in retrograde traffic. Indeed, it has recently been shown that coatomer is essential in yeast for retrieval from the Golgi complex of ER proteins bearing the KKXX signal .
The molecular weights of the coatomer subunits are similar to those of clathrin and the adaptors (Malhotra et al., 1989); ␤-COP has weak homology to ␤-adaptin (Duden et al., 1991), and more recently the cloning of -COP has shown that it has homology to the AP17 and AP20 adaptor subunits (Kuge et al., 1993), suggesting there may be similarities in structure and function between the two types of coat. In addition, the binding of coatomer and the AP1 adaptor to the Golgi complex and trans-Golgi network, respectively, is sensitive to brefeldin A (Robinson and Kreis, 1992) and requires ARF (Stamnes and Rothman, 1993), suggesting that similar mechanisms of coat assembly may occur. Interestingly, ␤Ј-COP contains five WD40 repeats, which are also found in trimeric G protein ␤-subunits (Stenbeck et al., 1993;Harrison-Lavoie et al., 1993). The recent finding that yeast ␣-COP also contains WD-40 repeats in its N terminus  suggests that these repeats might mediate interaction between the ␤Јand ␣-COP subunits.
We wanted to look at the intersubunit associations between individual COPs within the coatomer complex. To study interactions between the COPs, an in vitro system was established to reversibly disassemble and reassemble coatomer. Using this system, we show that ␣-, ␤Ј-, and ⑀-COP interact directly and that -COP interacts with ␥-COP. We also demonstrate that a partial complex comprising ␣-, ␤Ј-, and ⑀-COP, obtained after coatomer disassembly, can bind to membranes in vitro. In addition, using anti-peptide antibodies we show that the N-and C-terminal domains of ␤-COP are buried within the coatomer complex.

MATERIALS AND METHODS
Antibodies-Antibodies against synthetic peptides of ␤-COP have been prepared and affinity purified as described (Pepperkok et al., 1993). Anti-␤Ј-COP antibodies were raised in rabbits against the peptide KTDINLDEDILDD. A murine monoclonal antibody against ␤-COP (M3A5) was also used (Allan and Kreis, 1986). An anti-peptide antibody raised against ␥-COP was kindly donated by Drs. C. Harter and F. T. Wieland. Fab fragments were prepared according to the instructions of the manufacturer using papain immobilized on beads (Pierce).
Cell Culture-Vero cells (African green monkey kidney cells, ATCC CCL 81) were maintained as described earlier (Kreis and Lodish, 1986).
Metabolic Labeling of Vero Cells and Preparation of Cytosol-Cells were incubated in labeling medium (prepared by mixing 9 parts of methionine/cysteine-free minimal essential medium with 1 part of normal minimal essential medium) containing 30 Ci/ml [ 35 S]methionine/ cysteine (Pro-Mix, Amersham) for 18 -22 h at 37°C.
To prepare cytosol, cells were washed three times with ice-cold homogenization buffer, and the dishes were drained of excess liquid. Homogenization buffer was either HB-N (0.15 M NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA) if the cytosol was to be used immediately for immunoprecipitation or HB-P (0.1 M KPO 4 , pH 6.7, 5 mM MgCl 2 ) if the cytosol was to be used in experiments requiring dialysis. Cells from four 10-cm dishes were scraped into 1 ml of homogenization buffer containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 0.5 mM 1:10 phenanthroline, 2 M pepstatin A, 2 g/ml aprotinin, 0.5 g/ml leupeptin) and passed 20 times through a 25-gauge needle. The homogenate was centrifuged at 15,000 rpm for 15 min at 4°C in a microcentrifuge. The supernatant was collected and centrifuged again at 100,000 ϫ g for 1 h at 4°C, and the resulting high speed supernatant was used for experiments.
Immunoprecipitation-Cytosol, which had been dialyzed or frozen/ thawed was first centrifuged at 15,000 rpm in a microcentrifuge for 10 min at 4°C to remove aggregates. Triton X-100 was added from a 10% stock solution to give 0.5% final concentration. 20 l of protein A-Sepharose (Pharmacia, 50% slurry in immunoprecipitation buffer) was added, and the mixture was incubated for 2 h at 4°C on a rotating wheel. The beads were removed by centrifugation, and specific antibodies were added to the supernatant. After incubation overnight at 4°C, 20 l of protein A-Sepharose was added. For immunoprecipitations with Fab fragments, 20 g of goat anti-rabbit IgG was also added here. For immunoprecipitations with M3A5, 20 g of sheep anti-mouse IgG was added. After 2 h at 4°C, the beads were collected by centrifugation and washed five times with 1 ml of immunoprecipitation (IP) buffer and then once with 1 ml phosphate-buffered saline. IP buffer was IP-N (0.15 M NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA, 0.5% Triton X-100) for immunoprecipitations from cytosols prepared using HB-N, IP-P (0.1 M KPO 4 , pH 6.7, 5 mM MgCl 2 , 0.5% Triton X-100) for cytosols prepared using HB-P and cytosols dialyzed against RB-P, or IP-T (0.5 M Tris, pH 7.5, 2 mM EDTA, 0.5% Triton X-100) for cytosols dialyzed against DB-T. Proteins were eluted from the beads by boiling in sample buffer and were subjected to SDS-PAGE analysis.
Disassembly of Immunoprecipitated Coatomer-Coatomer was immunoprecipitated from [ 35 S]methionine-labeled Vero cytosol with 2 g of anti-EAGE polyclonal antibodies as described above. After the final wash in immunoprecipitation buffer, the beads were resuspended in 1 ml of IP-N buffer, DB-TT (0.5 M Tris, pH 7.5, 2 mM EDTA, 1 mM dithiothreitol, 0.5% Triton X-100), DB-MT (0.25 M MgCl 2 , 20 mM Tris, pH 7.5, 1 mM dithiothreitol, 0.5% Triton X-100), or DB-NT (1.0 M NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA, 1 mM dithiothreitol, 0.5% Triton X-100) and incubated for 1 h at 4°C on a rotating wheel. The beads were collected by centrifugation and washed twice with 1 ml of the same buffer and then once with 1 ml of phosphate-buffered saline. Proteins were eluted from the beads in sample buffer and subjected to SDS-PAGE analysis.
Sucrose Gradient Centrifugation and Gel Filtration of Disassembled Coatomer-Coatomer was immunoprecipitated from [ 35 S]methioninelabeled Vero cytosol with 5 g of anti-EAGE antibodies as described above. After the final wash in IP-N buffer, the beads were resuspended in 500 l of DB-TB (DB-TT minus Triton, with 0.25 mg/ml BSA), DB-MB (DB-MT minus Triton, with 0.25 mg/ml BSA), or DB-NB (DB-NT minus Triton, with 0.25 mg/ml BSA) and incubated for 4 h at 4°C on a rotating wheel. The beads were removed by centrifugation, and the supernatant was centrifuged at 15,000 rpm for 10 min at 4°C in a microcentrifuge to remove aggregates. 200 l of the supernatant was injected onto a 24-ml Superose 6 column (Pharmacia Biotech Inc.), equilibrated in DB-T, DB-M, or DB-N (DB-TT, DB-MT, DB-NT, respectively, minus Triton), and fractionated with a flow rate of 0.2 ml/min. After 10 ml of elution, 0.5-ml fractions were collected, 2.5 l of BSA (10 mg/ml) was added as a carrier, and proteins were precipitated with 10% trichloroacetic acid (see Duden and Franke, 1988) and analyzed by SDS-PAGE. The remaining 300 l of the supernatant was loaded on top of a 5-30% sucrose gradient (11.3 ml) poured in DB-T, DB-P, or DB-N. After centrifugation at 35,000 rpm for 18 h in a Kontron TST 41.14 rotor, 600-l fractions were collected from the top, BSA was added as a carrier, and proteins were precipitated with trichloroacetic acid and analyzed by SDS-PAGE.
Sucrose Gradient Centrifugation of Dialyzed Cytosol-Samples (300 l) were loaded on top of 5-30% sucrose gradients (11.3 ml) made in either DB-T (for Tris-dialyzed cytosol) or RB-P (for non-dialyzed cytosol), made in HB-P and cytosol dialyzed against RB-P, and centrifuged in a Kontron TST 41.14 rotor at 35,000 rpm for 18 h at 4°C. Fractions (600 l) were collected from the top, and proteins were precipitated with 10% trichloroacetic acid. Samples were analyzed by SDS-PAGE and immunoblotting.
Preparation of Rat Liver Cytosol and Golgi-enriched Membranes-To prepare rat liver cytosol, 30 g of frozen rat liver was thawed at 4°C, minced in 60 ml of homogenization buffer (20 mM Hepes-KOH, pH 7.0, 100 mM KCl, 2.5 mM MgCl 2 ) containing protease inhibitors, and homogenized with a Polytron using three 30-s bursts on setting 4, with a 2-min sedimentation and cooling period on ice between bursts. The resulting homogenate was centrifuged at 12,000 ϫ g for 15 min in a Haereus 18.50 rotor. The supernatant was collected and centrifuged at 100,000 ϫ g for 60 min in a Kontron TST 41.14 rotor. The resulting supernatant was collected and centrifuged for a further 120 min at 200,000 ϫ g in the TST 41.14 rotor. The 200,000 ϫ g supernatant (rat liver cytosol) was frozen in liquid nitrogen and stored at Ϫ80°C. Golgi-enriched membranes were prepared from rat liver using the method of Malhotra et al. (1989).
In Vitro Membrane Binding Assay-To measure binding of coatomer, incubations (120 l) were carried out for 10 min at 37°C in the presence of 0.2 M sucrose, 25 mM Hepes-KOH, pH 7.0, 25 mM KCl, 2.5 mM MgCl 2 , and an ATP regenerating system (50 M ATP, 2 mM creatine phosphate, 12.5 units/ml creatine kinase). 14 g of Golgi-enriched membranes and 160 g of non-dialyzed (3.3 mg/ml) or dialyzed (3.0 mg/ml) rat liver cytosol were added as indicated in the figure legends. Reactions were layered on top of 200 l of 15% sucrose (in 25 mM Hepes-KOH, pH 7.0, 25 mM KCl) and centrifuged at 4°C in a microcentrifuge for 30 min at 15,000 rpm. The membrane pellets were resuspended in 15 l of SDS sample buffer. Proteins were separated by SDS-PAGE, and ␤-COP was detected by immunoblotting with M3A5 at 1:1,000 dilution.
To measure binding of 35 S-labeled disassembled COPs, coatomer was immunoprecipitated from 35 S-labeled Vero cytosol with 5 g of EAGE antibody as described above. The immunoprecipitate was incubated in 250 l of DB-TB for 4 h at 4°C, and the beads were removed by centrifugation. The eluted COPs were centrifuged at 15,000 rpm in a microcentrifuge to remove aggregates and added to the binding assay. Incubations (200 l) were carried out for 10 min in the presence of 50 l of disassembled COPs, 0.2 M sucrose, 25 mM Hepes-KOH, pH 7.0, 25 mM KCl, 2.5 mM MgCl 2 , an ATP regenerating system (as above), and 40 g of the 10 -80-kDa fraction of rat liver cytosol. 14 g of Golgi-enriched membranes were added as indicated in the figure legends. For the WBP1 competition experiment, 25 l of disassembled COPs were preincubated for 2 h at 4°C in the presence of 25 mM Hepes-KOH, pH 7.0, 25 mM KCl, 2.5 mM MgCl 2 , an ATP regenerating system (as above), 40 g of the 10 -80-kDa fraction of rat liver cytosol, and either 30 g of GST-WBP1 or GST-WBP1-SS  as indicated in the legend to Fig. 3. 14-g membranes were then added, and incubations were carried out for 10 min at 37°C. Samples were layered onto sucrose cushions and processed as described above. Proteins were detected by fluorography. To analyze proteins in the supernatant, they were precipitated with 10% trichloroacetic acid and processed for SDS-PAGE. To immunoprecipitate membrane-bound proteins, the membrane pellet was solubilized in IP-N for 15 min on ice and then centrifuged for 10 min at 15,000 rpm. Proteins were then immunoprecipitated as described above.
In Vitro Binding to Dilysine Motifs-The GST-WBP1 and GST-WBP1-SS fusion proteins were purified from bacteria as described previously . Incubations (240 l) were carried out for 2 h at 4°C in the presence of 40 l of disassembled COPs (prepared as for the membrane binding experiments), 50 mM Hepes, pH 7.0, 90 mM KCl, 0.5% Triton X-100, and 30 g of GST-WBP1 or GST-WBP1-SS immobilized on 20 l of glutathione-Sepharose beads (Pharmacia). The beads were pelleted by centrifugation, and the proteins in the supernatant were precipitated with 10% trichloroacetic acid and processed for SDS-PAGE. The beads were washed 4 ϫ 1 ml with 50 mM Hepes, pH 7.0, 90 mM KCl, 0.5% Triton X-100 and then 1 ϫ 1 ml phosphate-buffered saline. Bound proteins were eluted by boiling in SDS-sample buffer and analyzed by SDS-PAGE.
Electrophoresis and Immunoblotting-Reduced proteins were separated on SDS-polyacrylamide gels according to Laemmli (1970). For the analysis of 35 S-labeled proteins, gels were stained with Coomassie Blue, destained, washed in water for 30 min, incubated in 1 M salicylate for 20 min, and vacuum dried. Fluorographs were taken on x-ray film (X-Omat-AR films, Eastman Kodak) at Ϫ70°C. Quantitation of radiolabeled proteins was performed with a ScanJet Plus (Hewlett Packard) using the Deskscan and NIH Image programs on the Macintosh. For immunoblotting, proteins separated by SDS-PAGE were transferred onto nitrocellulose filters. Filters were then incubated with primary antibodies followed by peroxidase-conjugated secondary antibodies (Cappel). Peroxidase labeling was detected by ECL (Amersham).

Coatomer Disassembles at High Salt Concentrations-To
study disassembly of coatomer in vitro, the complex was immunoprecipitated from 35 S-labeled Vero cytosol and then subjected to various treatments. Immunoprecipitation with rabbit polyclonal antibodies against the EAGE epitope of ␤-COP under native conditions gave a pattern characteristic of coatomer with bands corresponding to ␣-, ␤/␤Ј-, ␥-, ␦-, ⑀-, and -COP (Fig.  1A). If the immune complex was incubated in buffer containing 0.5 M Tris, 0.25 M MgCl 2 , or 1 M NaCl, only one major band at ϳ100 kDa remained bound to the beads (Fig. 1A). Note that with 0.5 M Tris there was a small amount of ␦-COP remaining bound to the beads. This was reproducible between experiments and represented 15% of the amount of ␦-COP present before incubation in 0.5 M Tris. Quantitation of the other COPs revealed that in each case less than 2% of the amount originally bound was present after treatment with Tris. Since ␤-, ␤Ј-, and ␥-COP migrate at ϳ100 kDa on a normal SDS-polyacrylamide gel, immunoblotting was performed to identify which of these COPs remained on the beads. After incubation with 0.5 M Tris, 0.25 M MgCl 2 , or 1 M NaCl, only ␤-COP was present (Fig. 1B). ␤-COP therefore dissociates from the other coatomer subunits when incubated in high salt concentrations. The finding that ϳ15% of ␦-COP remained on the beads in 0.5 M Tris while the other subunits were completely dissociated suggests that ␤-COP can interact directly with ␦-COP.
We next analyzed whether the other COPs had dissociated from each other after incubation in high salt concentrations. Immunoprecipitated coatomer was incubated with 0.5 M Tris, 0.25 M MgCl 2 , or 1 M NaCl; the beads were removed by centrifugation, and the COPs in the supernatant were analyzed by sucrose gradient centrifugation and gel filtration chromatography. After incubation with 0.5 M Tris or 1 M NaCl, ␣-COP comigrated with ⑀-COP and a 100-kDa COP on a sucrose gradient ( Fig. 2A). The three proteins peaked at fraction 6. There was another peak of 100-kDa COPs at fraction 4, and ␦-COP peaked at fraction 3. In buffer containing 0.5 M Tris, -COP comigrated with a 100-kDa COP at fraction 4 and had a second peak at fraction 2, whereas in buffer containing 1 M NaCl there was only one peak at fraction 2. 0.25 M MgCl 2 gave identical results to 1 M NaCl (data not shown). After gel filtration chromatography in buffer containing 0.5 M Tris, there was a peak of 100-kDa COPs at fractions 10 and 11 (Fig. 2B). ␦-COP peaked at fraction 12 while -COP migrated as two species, one at fractions 10 and 11 and one at fraction 16. There were faint bands for ␣and ⑀-COP between fractions 3 and 8. Both proteins were much less abundant than the other COPs, however, suggesting that they are degraded or stick to the column under these conditions. Gel filtration in buffer containing 1 M NaCl (or 0.25 M MgCl 2 , not shown) gave similar results to 0.5 M Tris except that -COP migrated as one peak at fraction 17, and strong bands were present for ␣-COP, ⑀-COP, and a 100-kDa COP at fractions 6 and 7.
To identify the 100-kDa COPs, proteins were separated on a 7% gel and immunoblotted with antibodies to ␥and ␤Ј-COP; note that ␤-COP remained attached to the beads (see Figs. 1 and 2B). ␥-COP migrated as one species peaking at fraction 4 on a sucrose gradient (Fig. 2C) and fraction 12 on a gel filtration column (Fig. 2D). ␤Ј-COP migrated as two species. The major species was at fraction 6 on sucrose gradient and fraction 7 on a gel filtration column, and the other (minor) species was at fraction 4 on a sucrose gradient and fraction 11 on a gel filtration column (Fig 2, C and D). ␤Ј-COP is therefore the 100-kDa subunit that cofractionates with ␣and ⑀-COP. It was also possible to co-immunoprecipitate ␣-, ␤Ј-, and ⑀-COP in buffer containing 0.5 M Tris using an antibody to ␤Ј-COP, confirming that these three proteins interact (data not shown). Thus, incubation with 0.5 M Tris dissociates the coatomer into smaller units, consisting of an ␣-, ␤Ј-, and ⑀-COP complex, a ␥and -COP complex, monomeric ␤-COP and ␦-COP, and some monomeric -COP and ␤Ј-COP. Incubation with 0.25 M MgCl 2 or 1 M NaCl dissociates coatomer into an ␣-, ␤Ј-, and ⑀-COP complex and monomeric ␤-COP, ␥-COP, ␦-COP, -COP, and some monomeric ␤Ј-COP. These results show that ␣-, ␤Ј-, and ⑀-COP interact directly in the coatomer complex and that ␥-COP interacts with -COP.
Membrane Binding of Disassembled Coatomer-We used an in vitro assay (Donaldson et al., 1991(Donaldson et al., , 1992aPalmer et al., 1993) to study the membrane binding activity of disassembled coatomer subunits. Coatomer that had been immunoprecipitated from 35 S-labeled Vero cytosol was incubated with 0.5 M Tris, and the beads were removed by centrifugation. The eluted COPs were then incubated with Golgi-enriched membranes in the presence of a 10 -80-kDa fraction of rat liver cytosol (as a source of ARF). ␣-COP, ⑀-COP, and a 100-kDa COP bound to the membranes (Fig. 3A). Note that ␦-COP and -COP did not bind. The 100-kDa COP was identified as ␤Ј-COP since an antibody raised against this subunit co-immunoprecipitated ␣-, ␤Ј-, and ⑀-COP from the membranes. This also showed that these three proteins were in a complex when bound to the membranes. In addition, separation of proteins on a 7% gel to resolve ␤Јfrom ␥-COP (␤-COP was not present in the assay since it is not eluted from the antibody with 0.5 M Tris) showed that only ␤Ј-COP, and not ␥-COP, was present in the membrane pellet (Fig. 3B). A partial coatomer complex comprising ␣-, ␤Ј-, and ⑀-COP can therefore bind to membranes.
Interestingly, membrane binding of the trimer, unlike the intact coatomer, was insensitive to GTP␥S (not shown). It has recently been demonstrated that a partial complex of coatomer, comprising ␣-, ␤Ј-, and ⑀-COP, which formed in cell extracts in 0.3 M NaCl, could bind to the KKXX retrieval motif of resident ER proteins in vitro . This suggested to us that membrane binding of the trimer may be through interactions with the KKXX motifs of certain membrane proteins. To investigate this further, the radiolabeled, disassembled COPs were incubated with beads containing either GST-WBP1, which has a functional KKXX ER retrieval signal, or GST-WBP1-SS, which does not have a functional retrieval signal . As shown in Fig. 3C, the trimer comprising ␣-, ␤Ј-, ⑀-COP bound specifically to the GST-WBP1 beads and not to the GST-WBP1-SS beads, in line with previous results . This also demonstrated that the trimer was not apparently denatured since it could specifically interact with the WBP1 KKXX motif. To study whether membrane binding of the trimer involved KKXX binding, a competition experiment was performed. The disassembled COPs were preincubated with either WBP1 or WBP1-SS, and then the membrane binding was performed. Trimer binding to membranes was reduced to 40% of the control amount when the COPs were incubated with GST-WBP1 (Fig. 3D). The WBP1-SS mutant had no effect on binding. This suggests that binding of the ␣-, ␤Ј-, and ⑀-COP partial complex to membranes is, at least in part, via interactions with KKXX motifs of certain membrane proteins.
Disassembly of Coatomer Is Reversible-Vero cytosol was dialyzed against buffer containing 0.5 M Tris to disassemble FIG. 2. Gel filtration and sucrose gradient analysis of disassembled coatomer. Immunoprecipitated 35 S-labeled coatomer was incubated in buffer containing 0.5 M Tris, pH 7.5 (DB-TB), or 1 M NaCl (DB-NB) as indicated for 4 h at 4°C. Eluted proteins were fractionated on a 5-30% sucrose gradient (A) or on a Superose 6 gel filtration column (B), subjected to SDS-PAGE on a 12% gel and detected by fluorography. Panel B, in lane B are the proteins remaining bound to the beads after elution in high salt (loaded 1/25th total for Tris, 1/10th total for NaCl). In lane L is the material loaded onto the column (1/20th total for Tris, 1/10th total for NaCl). Molecular mass standards are in kilodaltons. Panel C, proteins eluted with 0.5 M Tris (DB-TB) were fractionated on a 5-30% sucrose gradient. Panel D, proteins eluted with 1 M NaCl (DB-NB) were fractionated by gel filtration on a Superose 6 column. After gradient centrifugation (C) or gel filtration (D), proteins were subjected to SDS-PAGE on a 7% gel, transferred to nitrocellulose, and exposed to x-ray film. Filters were then probed with anti-␥-COP antibodies at 1:100 dilution and subsequently reprobed with anti-␤Ј-COP antibodies at 1:250 dilution. Blots were developed using ECL as described under "Materials and Methods." coatomer and then redialyzed against a more physiological buffer to see if reassembly occurred. Coatomer could be immunoprecipitated from non-dialyzed cytosol under native conditions with the anti-peptide antibodies EAGE (␤-COP) or KTDI (␤Ј-COP) (Fig. 4A). After dialysis against buffer containing 0.5 M Tris, only ␤or ␤Ј-COP was precipitated, showing that both of these proteins had completely dissociated from the other COPs. The complete dissociation of ␤Ј-COP was unexpected, since we had already demonstrated that it binds ␣-COP and ⑀-COP after disassembly of immunoprecipitated coatomer with 0.5 M Tris (see Fig. 2). This difference was not due to the longer time used for dialysis since incubation of immunoprecipitated coatomer in buffer containing 0.5 M Tris for 24 h gave the same result as shorter incubations (data not shown). Perhaps coatomer bound to the EAGE antibody adopts a different conformation to coatomer in cytosol, and this limits the extent of disassembly. Alternatively, cytosolic factors might be required for complete disassembly of coatomer. If cytosol that had first been dialyzed against 0.5 M Tris was then redialyzed against 0.1 M potassium phosphate (RB-P), both EAGE and KTDI again immunoprecipitated the entire coatomer complex showing that reassembly had occurred. Densitometry confirmed that the stoichiometry of subunits was the same for native and for reassembled coatomer (Table I).
Sucrose gradient centrifugation confirmed that disassembly with 0.5 M Tris is reversible. ␤-COP peaked at fraction 11 in non-dialyzed Vero cytosol, giving a sedimentation coefficient of ϳ13 S as previously reported (Duden et al., 1991) (Fig. 4B). After dialysis against buffer containing 0.5 M Tris, ␤-COP shifted to a lower density and peaked at fraction 5, corresponding to a sedimentation coefficient of ϳ6 S. Redialysis against 0.1 M potassium phosphate resulted in a ␤-COP distribution similar to that of non-dialyzed cytosol, showing that it had fully integrated into reassembled coatomer. Identical results were obtained with rat liver cytosol (data not shown).
Reassembled Coatomer Is Functional-The in vitro membrane binding assay was used to study the membrane binding activity of reassembled coatomer and to compare it to that of native coatomer. Incubation of non-dialyzed rat liver cytosol with Golgi-enriched membranes resulted in membrane binding of coatomer (Fig. 5). This binding was increased 2 to 3 fold by GTP␥S in agreement with previously reported results (Donaldson et al., 1991;Palmer et al., 1993). Rat liver cytosol, which had previously been dialyzed against buffer containing 0.5 M Tris and then 0.1 M potassium phosphate to disassemble and reassemble coatomer, gave a similar amount of binding to untreated cytosol. This was enhanced 2-fold by GTP␥S. Thus, reassembled coatomer is competent to bind membranes in a GTP-dependent manner, as described for native coatomer.
The N and C Termini of ␤-COP Are Sequestered within the Coatomer Complex-To identify domains of ␤-COP required for interactions with the other coatomer subunits, we took advantage of the large number of anti-peptide antibodies raised against ␤-COP and the finding that coatomer can be disassembled using relatively mild conditions. 35 S-Labeled Vero cytosol was subjected to immunoprecipitation under native conditions with eight antibodies raised against ␤-COP (see Fig. 7) (we initially used 24 different antibodies, but 16 of these did not immunoprecipitate under native conditions or in buffer containing 0.5 M Tris (data not shown)). Of the eight antibodies, seven failed to precipitate any specific proteins and only one, EAGE, immunoprecipitated coatomer (Fig. 6A). If the cytosol was first dialyzed against buffer FIG. 3. Membrane binding of disassembled coatomer. Panels A and B, 35 Slabeled COPs, obtained after disassembly of immunoprecipitated coatomer with 0.5 M Tris (DB-TB), were incubated in the presence or absence of Golgi-enriched membranes as described under "Materials and Methods," and membrane-bound COPs were detected by fluorography. Panel A, proteins were separated on a 12% gel. The left lane shows 20% of the total COPs added to the assay. The right lane shows the proteins immunoprecipitated from solubilized membranes with anti-␤Ј-COP antibodies. Panel B, membrane-bound proteins (pellet) and unbound proteins (supt) were separated on a 7% gel. Panel C, 35 S-labeled COPs were incubated with GST-WBP1 (1) or GST-WBP1-SS (2) beads, and the bound and unbound (supt) COPs were analyzed on a 12% gel. Panel D, 35 S-labeled COPs were preincubated with WBP1 or WBP1-SS fusion proteins for 2 h at 4°C and then incubated with membranes as described under "Materials and Methods." Membrane-bound COPs were separated on a 10% gel and detected by fluorography. Molecular mass standards are in kilodaltons.
containing 0.5 M Tris, all eight of the antibodies now immunoprecipitated ␤-COP. Note that, as expected, only ␤-COP and none of the other COPs was precipitated under these conditions since ␤-COP is dissociated from the other COPs. Similar results were obtained when the experiment was repeated with Fab fragments (Fig. 6B). The other coatomer subunits therefore appear to mask the epitopes of seven of these antibodies, and it is only when they are removed that the epitopes can be recognized. These epitopes may form interaction sites for binding to the other coatomer subunits. DISCUSSION We have developed an in vitro system that allows reversible disassembly of coatomer. Under defined conditions ␣-, ␤Ј-, and ⑀-COP and ␥and -COP remain associated, showing that these subunits interact directly in the coatomer complex. In addition, ␤-COP can interact directly with ␦-COP since a significant fraction of the ␦-COP present in immunoprecipitated coatomer remains associated with ␤-COP after treatment with 0.5 M Tris, while the other subunits are completely released. This interaction is weaker than that between ␥and -COP. The interaction between ␥and -COP is, in turn, weaker than that between ␣-, ␤Ј-, and ⑀-COP since ␥and -COP dissociate in buffer contain- FIG. 4. Disassembly of coatomer is reversible. Panel A, 35 S-labeled Vero cytosol was dialyzed against buffer containing 0.5 M Tris, pH 7.5 (DB-T), and then half of the dialysate was redialyzed against 0.1 M KPO 4 , pH 6.7, 5 mM MgCl 2 (RB-P). Non-dialyzed, Tris-dialyzed, and potassium phosphate-dialyzed samples were immunoprecipitated with antibodies to ␤-COP (1, EAGE) or ␤Ј-COP (2, KTDI) as indicated and analyzed by SDS-PAGE on a 10% gel. Proteins were detected by fluorography. Molecular mass standards are in kilodaltons. Panel B, Vero cytosol was dialyzed as described above and fractionated by sucrose gradient centrifugation. Proteins were subjected to SDS-PAGE on a 10% gel. ␤-COP was detected by immunoblotting with M3A5 antibodies at 1:1,000 dilution using ECL. The positions of marker proteins are indicated (A, BSA, 4.3 S; C, catalase, 11.15 S; T, thyroglobulin, 16.5 S).

TABLE I Stoichiometry of COPs in native and reassembled coatomer
Coatomer was immunoprecipitated using EAGE antibodies from nondialyzed cytosol (native) or cytosol that had been dialyzed against buffer containing 0.5 M Tris, pH 7.5, and then redialyzed against 0.1 M potassium phosphate (reassembled) (see Fig. 4A). The amount of each subunit was measured by densitometry as described under "Materials and Methods" and is expressed in arbitrary units relative to the amount of ␣-COP. ing 1 M NaCl or 0.25 M MgCl 2 , while the ␣-, ␤Ј-, and ⑀-COP complex does not.
The ␣-, ␤Ј-, and ⑀-COP partial complex could bind to Golgienriched membranes in vitro. This binding was insensitive to nucleotides and appeared to be due, at least in part, to interactions with membrane KKXX motifs. The competition with WBP1 was specific since the mutant containing two serines instead of lysines did not have any effect in the assay. However, the binding could only be reduced by about 60% by preincubation with the GST-WBP1 fusion protein. This could be explained by there being some KKXX motifs on the membranes with a much higher affinity for the trimer than for GST-WBP1. Alternatively, the trimer may interact with other membrane proteins (non-KKXX containing), or it may be sticky such that 40% can still bind to membranes even in the presence of an excess of WBP1. All three possibilities may have physiological relevance. One could imagine, for example, that coatomer first binds to membranes in an ARF-dependent manner, and during or after this step it undergoes a conformational change. The ␣-, ␤Ј-, and ⑀-COP subunits could then become available for interaction with cargo molecules bearing, for example, the KKXX motif or other as yet undiscovered signals; alternatively, this trimer might perhaps also interact with the actual lipids of the membrane. An important function of the other subunits within the coatomer complex may be the regulation of these activities. Further experiments are clearly necessary to increase our understanding of the mechanistic details involved in coatomer binding to membranes.
We initially chose to use 0.5 M Tris and 0.25 M MgCl 2 to dissociate coatomer because it had already been demonstrated that these conditions disassemble the coat of clathrin-coated vesicles (Woodward and Roth, 1978;Keen et al., 1979). The separate clathrin and adaptor complexes that comprise the clathrin coat do not disassemble under these conditions, however, and disassembly requires high concentrations of urea or guanidine hydrochloride (Ahle and Ungewickell, 1989;Scarmato and Kirchhausen, 1990;Prasad and Keen, 1991). The dissociated components of the clathrin (heavy and light chains) and adaptor (two 100-kDa, 50-kDa, and 19-kDa subunits) complexes reassemble when dialyzed out of non-denaturing conditions (Scarmato and Kirchhausen, 1990;Prasad and Keen, 1991). Similarly, we show here biochemically that dissociated coatomer subunits reassociate to form coatomer when they are dialyzed out of high salt concentrations. Thus, each set of coat protein complexes has the capacity to spontaneously reassemble in vitro. Reassembly of coatomer in vitro appears to be efficient since 100% of dissociated ␤-COP is integrated into coatomer during reassembly. Reassembled coatomer also appears to fold in a similar way to native coatomer, since both complexes have the same migration on a sucrose gradient. The fact that the subunit stoichiometry is the same for reassembled and native coatomer also supports the notion that the dissociated COPs refold properly during assembly. This is further supported by the finding that reassembled coatomer recruits onto membranes. Coatomer exists as a stable complex in vivo (Hara- Kuge et al., 1994), but it must assemble soon after syn-thesis of its individual subunits. Whether this assembly occurs spontaneously is not clear at the moment. It is, for example, conceivable that chaperonins might assist in folding of the COPs.
Anti-peptide antibodies have been raised against epitopes throughout the sequence of ␤-COP (Pepperkok et al., 1993). Immunoprecipitation experiments showed that only one of these antibodies, EAGE, could recognize the native coatomer complex. These data are in good agreement with results obtained from microinjection of antibodies into living cells (Pepperkok et al., 1993). Of the antibodies injected, only 110-12 and EAGE could bind to ␤-COP in vivo. The 110-12 and EAGE epitopes are in the central region of ␤-COP, while the other epitopes are either in the N-terminal 120 amino acids (KDLQ) or the C-terminal half (E1, M3A5, KESE, D1, KLVE, LGDK) of the protein (see Fig. 7). These parts of ␤-COP are therefore buried within the coatomer complex, rendering them inaccessible to antibodies. Only one of the ten anti-peptide antibodies raised against the N-terminal half of ␤-COP worked by immunoprecipitation in 0.5 M Tris (data not shown). This may be due to low antibody affinities or the possibility that the N-terminal half of ␤-COP has a complex folding. For much of the N terminus of ␤-COP, it is still therefore unclear whether it is sequestered in the coatomer complex, as is the C-terminal half of the protein. In addition, it is not known if those epitopes that are buried in the complex are contact sites for binding to other COPs, since antibody access can be sterically inhibited by COPs that may not bind to the epitope. Further experiments using mutated proteins should help identify contact sites between ␤-COP and other coatomer subunits.
The structure of the coatomer complex, the spatial arrangement of its subunits, and the functions of the individual COPs are poorly understood at the moment. The ability to disassemble and reassemble the coatomer complex should help to further elucidate the structure of coatomer and clarify the roles of the individual COPs within the complex. FIG. 7. Location of the anti-␤-COP antibody epitopes. Shown are the location of the epitopes of the anti-␤-COP antibodies used in this study. The epitope of antibody 110-12 is also shown. 110-12 did not work by immunoprecipitation (this study) but did recognize ␤-COP in living cells (Pepperkok et al., 1993). The epitope of the monoclonal antibody M3A5 has been mapped to residues 601-625 of ␤-COP using truncated forms of the protein (data not shown). The antibodies marked with an open arrow head only recognize ␤-COP after coatomer disassembly. The antibody marked with a filled arrow head (EAGE) recognizes ␤-COP in coatomer and after disassembly of the complex.