Different domains of the AP-1 adaptor complex are required for Golgi membrane binding and clathrin recruitment.

The assembly of clathrin-coated buds on the Golgi requires the recruitment of the heterotetrameric AP-1 adaptor complex, which is dependent on both guanine nucleotides and the small GTP-binding protein ADP-ribosylation factor (ARF). Here, we have investigated the structural domains of the AP-1 complex necessary for ARF-mediated translocation of the adaptor complex onto Golgi membranes and the subsequent recruitment of clathrin onto the membrane. Controlled proteolysis of purified AP-1, derived from bovine adrenal coated vesicles, was used to generate AP-1 core fragments composed of the amino-terminal trunk regions of the β1 and γ subunits and associated μ1 and σ1 subunits, and lacking either the β1 subunit carboxyl-terminal appendage or both β1 and γ subunit appendages. On addition of these truncated fragments to AP-1-depleted adrenal cytosol, both types of core fragments were efficiently recruited onto Golgi membranes in the presence of GTPγS. Recruitment of both core fragments was inhibited by the fungal metabolite brefeldin A, indicative of an ARF-dependent process. Limited tryptic digestion of recruited, intact cytosolic AP-1 resulted in the quantitative release of the globular carboxyl-terminal appendage domains of the β1 and γ subunits. The adaptor core complex remained associated with the Golgi membranes. Recruitment of cytosolic clathrin onto the Golgi membranes was strictly dependent on the presence of intact AP-1. Tryptic removal of the β1 subunit appendage prevented subsequent clathrin recruitment. We conclude that the structural determinants required for the ARF-mediated binding of cytosolic AP-1 onto Golgi membranes are contained within the adaptor core, and that the carboxyl-terminal appendage domains of the β1 and γ subunits do not play any role in this process. Subsequent recruitment of cytosolic clathrin, however, requires an intact β1 subunit.

The assembly of clathrin-coated buds on the Golgi requires the recruitment of the heterotetrameric AP-1 adaptor complex, which is dependent on both guanine nucleotides and the small GTP-binding protein ADPribosylation factor (ARF). Here, we have investigated the structural domains of the AP-1 complex necessary for ARF-mediated translocation of the adaptor complex onto Golgi membranes and the subsequent recruitment of clathrin onto the membrane. Controlled proteolysis of purified AP-1, derived from bovine adrenal coated vesicles, was used to generate AP-1 core fragments composed of the amino-terminal trunk regions of the In and y subunits and associated,..,l and u1 subunits, and lacking either the 131 subunit carboxyl-terminal appendage or both 131 and y subunit appendages. On addition of these truncated fragments to AP-1-depleted adrenal cytosol, both types of core fragments were efficiently recruited onto Golgi membranes in the presence of GTPyS. Recruitment of both core fragments was inhibited by the fungal metabolite brefeldin A, indicative of an ARF.dependent process. Limited tryptic digestion of recruited, intact cytosolic AP-1 resulted in the quantitative release of the globular carboxyl-terminal append. age domains of the /31 and y subunits. The adaptor core complex remained associated with the Golgi memo branes. Recruitment of cytosolic clathrin onto the Golgi membranes was strictly dependent on the presence of intact AP-l. Tryptic removal of the /31 subunit append. age prevented subsequent clathrin recruitment. We conclude that the structural determinants required for the ARF-mediated binding of cytosolic AP-1 onto Golgi membranes are contained within the adaptor core, and that the carboxyl-terminal appendage domains of the /31 and y subunits do not play any role in this process. Subsequent recruitment of cytosolic clathrin, however, requires an intact /31 subunit.
The clathrin-coated vesicle exhibits a well defined structural organization, and the major protein components have been purified and characterized (1)(2)(3)(4). In addition to the principal * This work was supported in part by a Monsanto Company/Washington University Research grant and National Institutes of Health Grant CA 08759 (to S. K.) and in part by start-up funds from the Dept. of Pathology at Washington University (to E. U.). 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. §  coat protein clathrin, two major adaptor complexes have been identified. One type is restricted to the trans-Golgi network (TGN)l and Golgi-derived coated vesicles and termed AP-l. The second, localized to the plasma membrane and endocytic clathrin-coated vesicles, is termed AP-2. Adaptors are heterotetrameric complexes, composed of one related and one unique -100-kDa subunit, and two additional components, the -50-kDa IL and -20-kDa a subunits. Thus, the Golgi AP-1 adaptor complex is composed of 'Y, 131, ILl, and a1 subunits, and the related AP-2 is composed of 0:, 132, 1L2, and a2 subunits. The 131 and 132 subunits are very similar in structure (5)(6)(7)(8), while the 0: and ' Y subunits are the most distantly related adaptor subunits (4,9). Electron microscopy has revealed that the AP-2 adaptor has a large globular core domain and two exposed globular head or appendage domains 2 (10). Controlled proteolysis of both AP-1 and AP-2 (11)(12)(13)(14) has shown that the core is composed of the amino-terminal trunk regions of either the 0: and 132 or the 131 and ' Y subunits as well as the intact IL and a subunits, while the appendages represent the 30-40-kDa carboxyl-terminal regions of the large subunits.
Adaptors are believed to facilitate clathrin-coated vesicle formation by combining a clathrin binding domain and a membrane association domain within an oligomeric protein complex. When it became apparent that the 131 and 132 subunits were highly related, it was suggested that these subunits might contain a common domain capable of binding to clathrin (5). Indeed, the purified 132 subunit interacts directly with preformed clathrin cages (15), and, more recently, it has been shown that recombinant l3-type subunits alone can induce clathrin polymerization (16). Further attempts to define the clathrin binding site more precisely indicated that proteolytically generated cores can bind to preformed clathrin cages (12), suggesting that the clathrin binding site is located in the amino-terminal trunk domain. However, in buffer conditions that prevent adaptor self-association, tryptic removal of the 13 subunit appendage results in the release of both appendage and core fragments from clathrin-coated vesicles (14). This suggests that both the trunk and appendage domains of the l3-type subunits are required for high affinity clathrin binding (14,16).
Before clathrin binding and polymerization can occur, however, the adaptors must associate stably with the appropriate intracellular membrane compartment. According to the simplest model for the interaction of adaptors with membranes (1,3,4,17,18), it was assumed that they attach directly to the cytoplasmic portions of selected receptors. Indeed, there is evidence that adaptors can interact directly, albeit weakly, with the cytoplasmic regions of certain proteins sorted into clathrincoated vesicles (17)(18)(19)(20)(21)(22). This simple model, while attractive, failed to explain the restricted localization of AP-l within the cell and why AP-l does not bind to cytoplasmically oriented trafficking motifs in receptors transiting through other intracellular compartments (2)(3)(4). Furthermore, the dramatic effect of brefeldin A (BFA) on the intracellular localization of AP-l also suggested that clathrin coat formation is highly regulated (23,24).
Recently, the recruitment of cytosolic AP-l onto purified Golgi membranes was reconstituted in vitro (25,26). Adaptor binding was found to be dependent on GTP and antagonized by BFA. ARF, a small GTP-binding protein, was identified as the GTP-requiring component (25,26). Coatomer-coated vesicle formation also proceeds by the initial recruitment of ARF (27,28), explaining the sensitivity of both clathrin-and coatomercoated vesicle formation to BFA. However, since no clear ARF specificity has yet been discerned (25,26), we proposed that ARF may be required to ensure the selective interaction of AP-l with a specific docking protein in the context of the TGN membrane. While our model overcomes the limitations of the direct association model, the structural determinants on the AP-l complex that are recognized by the putative docking protein remain to be identified. The role of the a and 'Y subunit appendages has been probed using a chimera-based approach (29). Swapping the a and 'Y subunit head and/or hinge regions did not appear to affect either the assembly or targeting of the AP-l or AP-2 chimeric complexes. Here, we have taken a different approach to directly assess the structural features of the AP-l heterotetramer required for Golgi membrane association and subsequent clathrin recruitment. Defined tryptic fragments were produced by controlled proteolysis (8,14), and this has enabled us to follow the ARF-dependent recruitment ofthe different fragments onto Golgi membranes. We have found the translocation of the AP-l core complex to be indistinguishable from intact AP-l. This precludes the globular carboxyl terminus and hinge regions from controlling this process. In contrast, clathrin binding to Golgi membranes was found to be strictly dependent on an intact /31 subunit.

EXPERIMENTAL PROCEDURES
Materials-[a-32 P1GTP (>3,000 Cilmmol) was obtained from ICN. GTPyS was purchased from Boehringer Mannheim. Aprotinin, ATP, benzamidine, BFA, dithiothreitol, leupeptin, soybean trypsin inhibitor, Triton X-100, and Tween 20 were from Sigma. BFA was stored as a 5 mg/ml stock solution in ethanol at -20°C. Sprague-Dawley rats were purchased from SASCO, and the frozen bovine adrenal glands were obtained from PelFreez. BA 83 nitrocellulose membranes were obtained from Schleicher and Schuell, and Sepharose 4B, CNBr-activated Sepharose 4B, and the M r markers for electrophoresis were obtained from Pharmacia Biotech Inc. Tosylphenylalanyl chloromethyl ketone-treated trypsin was purchased either from Cooper Biomedical or Worthington. The ECL reagents for chemiluminescent detection were obtained from Amersham. All other reagents were the highest grade available.
Antibodies-The anti-13 subunit antibody mAb 100/1 (5), the anti-a subunit antibodies mAb 100/2 (5) and mAb AP.6 (30), the anti-y subunit antibody mAb 100/3 (5), the polyclonal anti-clathrin light chain antibody R461 (5), the anti-clathrin heavy chain antibody mAb TD.l (31), and the polyclonal anti-a-mannosidase II serum (32) were prepared and used as described previously (26). Antiserum AE/l was raised against a synthetic peptide corresponding to the carboxyl-terminal dodecapeptide of the murine y subunit of AP-l (9). Peptide conjugation, immunization, and screening was as detailed previously (33). AE/l was further affinity-purified from immune serum on a column of the peptide (AEVNN-FPPQSWQ) coupled to CNBr-activated Sepharose 4B. Peptide-specific antibodies were eluted with 100 mM glycine HCI, pH 2.5, followed by 100 mM triethylamine, pH 11.5 (34). Antiserum GD/l was similarly raised against a dodecapeptide (GDLLNLDLGPPV), a conserved sequence in the hinge region of all known mammalian (3-type subunits. Affinity-purified GD/l was prepared similarly on a peptide-Sepharose 4B column. The rhodamine-conjugated anti-rabbit IgG and fluoresceinconjugated anti-mouse IgG secondary antibodies were purchased from Dakopatts. Subcellular Fractionation-Rat liver Golgi membranes and cytosol were prepared as described previously (26). For the preparation of bovine adrenal cytosol, frozen adrenal glands were rapidly thawed at 37°C in a homogenization buffer of25 mM Hepes-KOH, pH 7.4, 250 mM sucrose, 1 mM EDTA supplemented with 5 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride. Once thawed, the tissue was immediately transferred to homogenization buffer on ice, finely minced, and homogenized in a Potter-Elvehjem homogenizer containing approximately 2 volumes of homogenization buffer with 5 mM benzamidine, 1 mM phenylmethylsulfonyl fluoride, 0.1 trypsin inhibitory unitlml aprotinin, and 5 p,glmlleupeptin. This and all subsequent procedures were performed at 0-4°e. The crude homogenate was centrifuged sequentially at 3,000 x g for 10 min, 10,000 x g for 20 min, and 100,000 x g for 60 min to yield a high speed cytosolic supernatant fraction. The cytosol was stored in aliquots at -80°C after rapid freezing in dry ice.
Adrenal or rat liver cytosol was depleted of clathrin by gel filtration on Sepharose 4B. 50-60 ml of cytosol was rapidly thawed, centrifuged at 150,000 x g for 60 min to remove aggregated material, and loaded at 1 mllmin onto a Sepharose 4B column (5.0 x 70 em) equilibrated in 25 mM Hepes-KOH, pH 7.3,125 mM potassium acetate, 5 mM magnesium acetate, and 0.05% sodium azide. Fractions of20 ml were collected, and the elution positions of clathrin and AP-l and AP-2 adaptors was determined on immunoblots using mAb TD.l and mAb 100/1. The clathrin triskelia eluted well ahead of the adaptors, and fractions after the clathrin peak were pooled, concentrated by precipitation with 60% ammonium sulfate, and dialyzed against 25 mM Hepes-KOH, pH 7.0, 125 mM potassium acetate, 5 mM magnesium acetate, and 1 mM dithiothreitol (Buffer C). After centrifugation at 12,000 x g for 15 min, the clathrin-depleted cytosol was aliquotted, quick frozen in dry ice, and stored at -80°C. The clathrin-containing fractions were pooled separately and concentrated approximately lO-fold by ultrafiltration using an Amicon YM 30 filter. After centrifugation at 12,000 x g for 15 min, the clathrin-enriched fraction was stored at 4°C.
Endogenous AP-l was removed from the clathrin-depleted adrenal cytosol by immunodepletion (26). Briefly, cytosol was passed over an immobilized mAb 100/3 column equilibrated in buffer C on ice, and the loading eftluent was reapplied to the column several times, before similar passages over a second mAb 100/3 column equilibrated in buffer C on ice. The resulting AP-l-depleted cytosol was analyzed on immunoblots using affinity-purified AE/l, and the extent of depletion was more than 95%. Clathrin-coated vesicles and coated vesicle-derived AP-l were purified from frozen bovine adrenal glands as described in detail previously (8). Protein concentrations were estimated using either the Coomassie Blue method (35) with bovine serum albumin as a standard or, for the purified AP-l preparations, by A 28o , using an E}":;= value of 6.0 (14).
Golgi Binding Assay-The assay was a modification of our binding assay utilizing rat liver cytosol (26). For most assays, bovine adrenal cytosol replaced the rat cytosol, and Golgi membranes and GTPyS were added to final concentrations of 50 p,g/ml and 100 p,M, respectively. Reactions were terminated by chilling on ice followed by centrifugation, and the Golgi membrane-containing pellets were prepared for immunoblotting as described (26). Routinely, pellets were resuspended in 20 p,1 of 1 x SDS sample buffer, and 10 p,1 was loaded per lane. When aliquots of the assay supernatant were also examined, an aliquot corresponding to 1/80 of each supernatant was loaded. For the two-stage clathrin binding assay, the first stage incubation contained clathrindepleted rat liver cytosol, Golgi membranes, and GTPyS. Membranes were recovered by centrifugation and digested with trypsin as described below. The trypsinized membranes were then added to a second stage incubation containing 0.1 mg/ml c1athrin-enriched Sepharose 4B pool and 20 p,g/ml BFA in buffer C. After incubation at 37°C for 15 min, the tubes were returned to ice, and the pellets were prepared for analysis as outlined above.
The asterisk denotes addition after 10 min of incubation at 37°C. All reactions were terminated after 20 min, and the membrane fractions were analyzed on immunoblots with mAb 100/3 (upper portion) or by [0l_3 2 PjGTP overlay (lower portion). The -80-kDa band detected with mAb 100/3 represents an endogenous proteolytic degradation product of the y subunit of the adrenal AP-l. In the lower panel, the -22-to -30-kDa GTP-binding proteins represent an uncharacterized group of low molecular mas GTP-binding proteins while the band migrating slightly ahead of the 20-kDa standard is ARF, based on both co-migration with purified ARF and sensitivity to BFA. The upper and lowel' portions originate from a single blot, and the position of the molecular size markers is indicated on the left.
to the time of addition of BFA ( Fig. 1, lanes d-g). These results are analogous to those obtained using rat liver cytosol (26) and reiterate the sensitivity of both ARF and AP-1 membrane association to BFA.
Proteolysis of Golgi-associated AP-l Adaptor Complexes-The conformation of the adrenal AP-1 recruited onto the Golgi in the presence of GTPyS was probed by treating the resulting membranes with trypsin. The digestion conditions were similar to those used for controlled digestion of purified AP-1 (8,14) and AP-1 within clathrin-coated vesicles (14). In both those instances, the globular carboxyl-terminal appendage of the f31 subunit was rapidly cleaved from the intact AP-1 complex, followed by much slower cleavage and release of the carboxyl-terminal y subunit appendage. The aminoterminal trunks of both the f31 and y subunits, as well as the iLl and 0'1 subunits, collectively comprising the AP-1 core fragment, proved relatively resistant to trypsin. We found that the AP-1 recruited from adrenal cytosol onto the Golgi membrane fraction was digested into analogous appendage and trunk tryptic fragments (Fig. 2, A and B). As noted previously (8,14), the f31 subunit was significantly more sensitive to trypsin than the y subunit. Extensive tryptic degradation of a-mannosidase II, a lumenal Golgi marker, was observed only in the presence of Triton X-lOO (Fig. 2C,  lane f). This demonstrated that during the assay and subsequent proteolysis, the Golgi membranes remained correctly oriented and sealed, and that AP-1 was recruited onto the cytoplasmic face of these membranes. Furthermore, membrane association of AP-1 in the absence of clathrin did not appear to result in major conformational changes as reflected by gross alterations in proteolytic ensitivity.
To ascertain whether the AP-1 appendages, the core, or both remained membrane-associated, the Golgi fraction was sedimented following digestion, and the resulting supernatant and pellet fractions were analyzed (Fig. 3). On untreated membranes or membranes that contained excess trypsin inhibitor prior to addition of trypsin, the recruited AP-1 remained membrane-associated during the 10-min digestion at 37°C (lanes b and h). This is in contrast to endogenous Golgi-associated AP-1, potassium acetate, 2.5 m~l magnesium acetate, 1 mg/ml gluco e (buffer E) and then frozen on dry ice. The cell were thawed, and the endogenous cyto 01 wa removed by dipping the coverslip into a beaker of buffer E 10 times at room temperature. Gel-filtered whole rat liver cyto 01, at a concentration of -5 mg/ml in buffer D, was supplemented with 100 p.~1 GTPyS and coated ve icle-derived adrenal AP-1 or proteolytic fragment to give a final adaptor concentration of 50-70 p.g/ml. After addition of the cytosol, the cell were warmed to 37°C for 15 min and then wa hed by dipping the coverslip into a beaker of buffer E 10 time . The cells were then fixed for 5 min at room temperature with 3.7% formaldehyde in PBS and double-labeled with R461 and mAb 100/3 as described in detail previou Iy (5).
Controlled Tryptic Digestion-Tryptic digestion of Golgi membrane wa carried out in buffer C at a membrane concentration of 1 mg/ml. Triton X-100 and trypsin were added on ice a indicated in the figure, and dige tion was performed at 37°C for 10 min, followed by chilling and addition of excess oybean tryp in inhibitor. For some experiments, the membrane were then collected by centrifugation at 15,000 x g for 10 min at 4°C, the supernatant fractions were concentrated by precipitation with methanol/chloroform (36), and both supernatant and pellet fractions were analyzed by immunoblotting. Proteolysis of the purified adrenal AP-1 was performed in 25 mM Hepes-KOH, pH 7.0, 125 mM potassium acetate fOI' 1 h at 30°C using a protein/trypsin ratio of 1:30 for the quantitative removal of the 131 subunit appendage. The reaction was stopped by adding excess soybean trypsin inhibitor, and the sample was u ed for the binding studie without furthet· purification. For the preparation of appendagele s AP-1 cores, the digestion was carried out at 30°C using a trypsin to substrate ratio of 1:20. Residual AP-1 containing an intact y subunit, which we have found difficult to cleave quantitatively with tryp in (14), was removed by immunoad orption on a mAb 100/3 column. The unbound material was concentrated to approximately 0.5-1.0 mg/ml in a Centricon 30 before use in the binding as ays. The polypeptide compo ition of aJl the fragments used in this study was verified on Coomassie Blue-stained gels.
Electrophoresis and Immunoblotting-Discontinuous SDS-polyacrylamide gel electrophoresis was carried out as described previously (26).
Proteins were tran ferred onto nitrocellulose membranes at no v for 70 min in a buffer of 15.6 mM Tris, 120 mM glycine, pH -8.3, equilibrated to 4°C. Following transfer, the membranes were blocked and probed with antibodies as described previou Iy (26).

RESULTS
The goal of this study was to identify the structural regions of the heterotetrameric AP-1 adaptor complex which are involved in the ARF-dependent recruitment onto Golgi membranes and the subsequent recruitment of cytosolic clathrin. To distinguish between the endogenous rat liver Golgi-associated AP-1 and exogenously added cytosolic AP-1, using a speciesspecific mAb directed against the y subunit, we had to modify our existing AP-1 binding assay by substituting bovine adrenal gland cytosol for rat liver cytosol. To limit our analysis to the initial events in clathrin-coated vesicle assembly, clathrin-depleted adrenal cytosol (26) has been used for several of the experiments described below.
Recruitment of Bouine Adrenal AP-l onto Golgi Membranes-Recruitment of AP-1 onto Golgi membranes is preceded by the recruitment of ARF (25,26). While only very low levels of ARF are found on purified Golgi membranes (37, 38) (Fig. 1, lane a), the level increased markedly upon incubation with clathrin-depleted adrenal cytosol in the presence of GTPyS (lane c). Several other uncharacterized low molecular weight GTP-binding proteins were also recruited onto the Golgi membrane. The accumulation of ARF was accompanied by recruitment of cytosolic AP-1 onto the Golgi membrane fraction. When recruitment was followed after pretreatment with BFA, decreased levels of both ARF and AP-1 were observed in the membranes, and the extent of recruitment varied according which dissociates from the membranes on incubation at 37°C (26). The stability of the recruited exogenous AP-1 is probably a consequence of the GTPyS, which we believe traps the AP-1 irreversibly on the membranes. In the presence of low concentrations of trypsin, the bulk of both the generated {31 trunk and the undigested intact y subunit pelleted with the Golgi membranes (Fig. 3, panels A and B, lane d). However, all of the -40-kDa {31 subunit appendage fragment formed was observed exclusively in the supernatant fractions (panel A, lane c). The -40-kDa adrenal {31 subunit appendage fragment is smaller than the -44-kDa brain AP-1 {31 subunit appendage (14) since a 14-amino acid neuron-specific insert is absent (7). At higher trypsin concentrations, all the detectable 32-kDa y subunit appendage fragment generated was also found only in the supernatants (panel B, lane e).
Recruitment of Purified AP-l and AP-l Core Fragments-While the above data indicate that the carboxyl-terminal appendage domains of the {31 or y subunits are not required to retain the recruited AP-1 core complex on the Golgi membrane in the presence of GTPyS, they do not reveal whether these regions are required for the initial recognition and recruitment of the AP-1 complex onto the Golgi. We therefore used purified AP-1, derived from adrenal clathrin-coated vesicles, to prepare defined tryptic fragments lacking either the (31 subunit ap- pendage domain or both the {31 and y subunit appendages. The recruitment of these fragments was then followed in a modified binding assay in which the endogenous AP-1 was first removed from the adrenal cytosol by immunodepletion, and then either intact AP-1 or the tryptic fragments were added back to this depleted cytosol. The results of a typical experiment are presented in Fig. 4. From a comparison of the supernatant fractions from incubations containing either control or AP-1-depleted cytosol, it is apparent that the immunodepletion had quantitatively removed AP-1 (Fig. 4, panels A and B, lanes band c compared to  lanes d and e). Consequently, no recruitment of adrenal AP-1, as judged by the presence of an exogenous y subunit, was evident in the pellet fractions (panel D, lane e) of incubations containing the depleted adrenal cytosol. However, when probed with either mAb 100/1 (panel C) or GD/l (panel E), two closely migrating {3-type subunits were observed on the Golgi membranes incubated with the AP-1-depleted cytosol and GTPyS (panels C and E, lane e). The slower migrating {3 form corresponds to the {31 subunit of residual endogenous rat liver Golgiassociated AP-1, which has not completely dissociated from the membrane during the course of the assay (26). The partner y subunit can be seen in panel F (lanes a and e). Because this y subunit was detected with the AE/l antibody (panel F) but not the species-specific mAb 100/3 (panel D), this membrane-associated AP-1 must be of rat origin. Note also that more of this endogenous AP-1 was observed on the membranes incubated together with the AP-1-depleted adrenal cytosol than on the Golgi membranes incubated without cytosol (panels C, E, and F, lane e compared to lane a). This demonstrates that in the presence of AP-1-depleted cytosol, the endogenous rat AP-1 that had dissociated from the Golgi can rebind to the mem-  -k) were prepared together with 20 fLg/ml Golgi membranes and 50 fLg/ml BFA as indicated in the figure. All the tubes contained 100 fLM GTP')'S, added either at the beginning of the assay, or, when BFA was present, after 10 min at 37°C. Reactions were terminated aftel' 20 min at 37°C, and the Golgi membrane pellets were recovered by centrifugation. AIiquots corresponding to 1/80 of each upernatant (panels A and B) and 1/2 of each pellet (panels C-F) fraction were analyzed on duplicate immunoblots using either anti-(3 subunit mAb 100/1 (panels A and C) or GD/1 (panel E) or anti-')' subunit mAb 100/3 (panels Band D) or AE/1 (panel F).
branes in an ARF-dependent manner. The lower f3-type band remains to be identified definitively. Additional experiments have shown that no a subunit had been recruited, excluding the possibility that the lower band was a 132 subunit. The unidentified band might be a degradation product of the endogenous Golgi-associated 131 subunit or, less likely, a free monomeric 13 subunit.
When purified coated vesicle-derived AP-1 was added into the depleted cytosol to levels approximating 1 to 5 times the estimated endogenous concentration, the level of exogenous AP-1 recruited onto the membrane was similar to that of the cytosolic form of AP-1 (Fig. 4, panels C and D, lane g compared  to lane c). In both cases, however, the majority of the AP-1 adaptors remained in the supernatants (panels A and B), which is consistent with our previous observation that another cytosolie component limits the extent of AP-1 recruitment (26). When the truncated AP-1 tryptic fragment, lacking the 131 subunit appendage, was added to the depleted cytosol together with GTPyS, the resulting Golgi membranes showed that the -60-kDa 131 trunk was membrane- associated (panel C, lane i). On a duplicate blot probed either with mAb 100/3 (panel D) or AE/1 (panel F), the intact y subunit was also seen to be membrane-bound (lane i), indicating that the tryptic 131 fragment was indeed part of a heterotetrameric complex. Again, the bulk of the added fragments remained in the supernatant fraction (panels A and C). BFA effectively antagonized the translocation of the f31-appendageless core in the presence of GTPyS (panels C-F, lane k). Analysis of the blots with GD/1, an antiserum raised against a conserved peptide epitope in the 13 subunit hinge, that recognizes the 131 appendage (Fig. 3), revealed that the free 131 subunit appendage had not been recruited onto the membrane (panel E, lane i). Thus, the translocation of AP-1 onto the TGN appeared to proceed normally in the absence of the 131 subunit head and hinge regions.
Subcellular Localization of Recruited Exogenous AP-I-Independent confirmation of the specific targeting of the exogenous AP-1 adaptors onto the TGN was obtained using a cellbased recruitment assay (23,24). Permeabilized NIH 3T3 fibroblasts were incubated at 37°C with whole rat liver cytosol that had been supplemented with exogenous adrenal coated vesicle-derived AP-l. Mter washing, the cells were prepared for indirect immunofluorescence microscopy. The distribution of The effect of removing the /31 appendage on the binding of the exogenous AP-1 complex added to rat liver cytosol with GTPl'S is shown in Fig. 5f. Strong perinuclear staining was observed, and the recruitment of the trypsinized AP-1 complex was indistinguishable from the intact coated vesicle-derived AP-l. Taken together, our data indicate that tryptic removal of the /31 subunit appendage had no effect on AP-1 recruitment onto the Golgi membrane fraction.
Having established that the globular carboxyl-terminal and intact hinge regions of the /31 subunit were not required for TGN recruitment, we next examined any role played by the l' subunit appendage in this process. A tryptic AP-1 core fragment, lacking both /31 and l' subunit appendages, was generated by trypsinization and removal of remaining intact l' subunit-containing adaptors by immunoadsorption on mAb 100/3 and then added to the AP-1-depleted cytosol. The appendageless AP-1 core prepared in this way remains as an assembled heterotetramer (14), but since removal of the l' subunit appendage abolishes reactivity with both mAb 100/3 and AE/1, we were only able to follow the fate of the /3 subunit trunks. Nevertheless, our experiments showed that the appendageless AP-1 core was recruited onto the Golgi membranes as efficiently as the intact AP-1 complex (Fig. 6, panel B, lane g  compared to lane c). Again, recruitment of the core fragment was inhibited by BFA (panel B, lane i). Similar levels of recruitment of the different cores were observed over a range of concentrations (Figs. 4 and 6), so major differences in affinity between intact AP-1 and the core fragments appears unlikely. Taken together, these experiments show that neither appendage domain is essential for either recruitment or membrane   Fig. 5a. A dispersed peripheral punctate staining pattern with a juxtaposed concentrated perinuclear staining was typical and represents cell surface-and TGN-derived clathrin-coated pits and vesicles, respectively (2,5,39). The same cells, double-labeled with the species-specific mAb 100/3, exhibited barely discernible perinuclear staining (Fig. 5b). While the overall clathrin staining pattern was not altered by adding 100 J.l.M GTPl'S to the donor cytosol (Fig. 5c), a dramatic enhancement of AP-1 staining in the perinuclear Golgi region was evident (Fig. 5d).
Since little exogenous AP-1 was detected in the perinuclear area in the absence of GTPl'S, and little of the recruited AP-1 was associated with the peripheral clathrin-coated regions, direct binding of exogenous AP-1 to vacant sites on preassembled clathrin lattices appears unlikely. We have also performed additional experiments using the cc subunit-specific mAb AP.6 and found no evidence for recruitment of the AP-2 adaptor complex onto the TGN (data not shown). This suggests that authentic recruitment of the intact AP-1 onto the TGN was being observed. Clathrin recruitment is prevented following tryptic removal of the (31 subunit appendage domain of Golgi-associated AP-l. AP-1-containing Golgi membranes were prepared using clathrin-depleted rat liver cytosol and GTPyS, recovered by centrifugation and then incubated at 37 ·C with either excess soybean trypsin inhibitor and 2 JLg/ml trypsin (lanes b-d) or 2 JLg/ml trypsin (lanes e-g). After 10 min, the membranes were again recovered and then used for the second step reaction. Tubes containing untreated Golgi (lane a), mock-digested Golgi (lanes b and d), or trypsinized Golgi (lanes e and g) membranes and 0.1 mg/ml Sepharose 4B clathrin-enriched fraction (lanes a, c, d, (, and g) were prepared on ice. 20 JLg/ml BFA was added to each tube followed by incubation at 37 ·C for 15 min. The membranes were pelleted and resolved on duplicate 12.5% SDS-polyacrylamide gels, transferred to nitrocellulose, and analyzed by immunoblotting with a mixture of anti-clathrin mAb TD.1 and anti-13 subunit mAb 100/1 (panel A) or anti-y subunit AE/l (panel B).

DISCUSSION
Adaptors are thought to playa pivotal role in the assembly of clathrin-coated vesicles by determining the intracellular site of coat nucleation and by promoting assembly of the growing lattice. To perform this regulatory role, recruited cytosolic adaptors must establish multiple contacts with components within the membrane and with the cytosolic clathrin, and it is likely that this is the underlying reason for the structural complexity of the adaptor heterotetramer. In this study, we have delineated the regions required for the ARF-dependent membrane association of AP-1 and the subsequent recruitment of cytosolic clathrin triskelia onto the TG . We found that an AP-1 adaptor core, lacking both the {31 and y subunit appendages, was fully competent to bind to Golgi membranes. Binding of the cores was dependent on ARF·GTP and strongly inhibited by BFA, as also observed for intact AP-1 complex (25,26). This suggests that the structural determinants required for the ARF-dependent recruitment of AP-1 onto the Golgi membranes reside within the core fragment.
While this work was in progress, the results of a study examining the intracellular distribution of chimeric adaptor complexes in which the a and y subunit heads were either switched or deleted were published (29). Immunofluorescence analysis of cells transfected with these various constructs showed that the chimeric or headless adaptors associated with the appropriate intracellular membranes in spite of the modifications. It was concluded that the y subunit appendage has, at most, a passive role in determining the intracellular targeting of AP-1 (29). Our work now provides independent biochemical confirmation of this conclusion and, in addition, establishes that the appendage of the {31 subunit is not required for the ARF-dependent membrane a sociation of AP-l.
Although we have not yet established whether GTPyS inhibits the complete assembly of a coated vesicle or the budding process, clearly the analog does not interfere with the recruitment of cytosolic clathrin tri kelia. Our observation that an intact {31 subunit is required for this recruitment of clathrin is a strong argument for the selective association of this ubunit with clathrin. This idea was initially proposed when it was determined that the {3-type subunits of AP-1 and AP-2 were highly related, both by peptide mapping and immunological criteria (5) and by sequencing (6)(7)(8). Our data are also con istent with the finding that under conditions that reduce aggregation of adaptors, cleavage of the {31 subunit in the hinge region was sufficient to release the truncated AP-1 core, together with the appendage, from purified coated vesicles (14). Furthermore, a {31-like subunit, purified from bacteria expressing the recombinant protein, induces clathrin coat assembly in vitro (16). Together, these observations all implicate the {31 subunit in clathrin association.
The {32 subunit of AP-2 also interacts directly with clathrin (14)(15)(16). Weaker interactions between the AP-2 adaptor core and clathrin cages have been reported, but these interactions failed to induce clathrin assembly (11)(12)(13)(14). The a subunit of AP-2 also appears to bind to clathrin (41,42), and it has been proposed that this is the first contact that is established between these two molecules (41). By contrast, we found that clathrin could not be recruited onto Golgi-associated AP-1Iacking only the {31 appendage. Therefore, if the y subunit of AP-1 binds to cytosolic clathrin triskelia first, the affinity of this initial interaction is not sufficient to retain clathrin on the TG in the absence of an intact {31 subunit. Thus, while it is possible that the adaptor cores contain a second clathrin binding site, which is independent of the one that involves the {3 appendage, it seems to playa secondary role in the early recruitment step of clathrin onto TGN-associated AP-l.
Our data on the requirement of an intact {31 subunit for the recruitment of clathrin differ from the results of a recent study showing that clathrin could assemble onto elastase-generated, plasma membrane-associated, AP-2 cores (43). The most obvious difference between the two systems is the adaptor population being studied. AP-1 and AP-2 adaptors might utilize different domains for clathrin recruitment. However, given the high degree of sequence and structural homology between the {31, {32, and {31-like subunits (5)(6)(7)(8), this appears unlikely. The differences may rather be related to the fact that, under our experimental conditions, we have analyzed de novo formation of a coated bud on the TGN, while the AP-2 binding sy tem is more likely to reflect rebinding of clathrin to pre-existing coated pit structures that had been previously disrupted by nonphysiological manipulation of pH and salt conditions. This makes these two systems difficult to compare.
At present, the available evidence supports the notion that AP-1 initially associates with the Golgi by binding to a TGNspecific docking protein in an ARF-regulated manner (23)(24)(25)(26). The results of our controlled proteolysis experiments indicate that this association of AP-1 with the putative docking protein occurs through the core domain. We have also shown that the {31 subunit is involved in clathrin recruitment. What then is the function of the y subunit appendage? The considerable amino acid sequence divergence between the appendages of the a, {3, and y subunits has prompted the suggestion that the appendages might encompass a receptor binding domain (6, 7, 12, 13, Clathrin recruitment follows and requires an intact /31 subunit. For clarity, the orientation of the /31 subunit is shown to indicate the interaction with clathrin. Our data do not preclude that the /3 subunit trunk might also playa role in AP-1 membrane association. The J.L1 and (71 subunits are positioned within the AP-1 core, reflecting the resistance of these subunits to proteolysis. A sorting motif within the cytoplasmic domain of a receptor to be included in the growing coated bud is indicated in black. Proteins are not drawn to scale, and although the putative docking protein is drawn as a transmembrane protein, no direct evidence for this is available at present. 29). This hypothesis is appealing because the proline-and glycine-rich hinge regions, separating the globular heads from the core complex, may be flexible and accommodate sorting motifs located at varied positions within the cytoplasmic domains of different receptors (6,7). One attractive possibility is that the 'Y subunit appendage interacts with the cytoplasmic domains of membrane receptors which are preferentially sorted into Golgi-derived clathrin-coated vesicles. In fact, the analogous subunit of the AP-2 adaptor, the a subunit, can bind to plasma membranes after dissociation from the heterotetrameric complex (20). The binding is partially inhibited by peptides corresponding to the cytoplasmic domains of two receptors endocytosed within clathrin-coated vesicles (20). One interpretation of the partial nature of the inhibition is that the ex subunit recognizes both a plasma membrane docking protein (43,44) and sorting motifs on select membrane proteins, and only the latter interaction can be inhibited by receptor peptides (20). Because AP-2 cores can compete with intact AP-2 for plasma membrane binding (43), it seems reasonable to propose that like AP-1, AP-2 associates with a specific docking protein on the plasma membrane via the core domain.
The experiments presented here allow us to expand our model for the association of AP-1 with the cytoplasmic surface of the Golgi (26). We have deduced that when AP-1 is recruited onto the TGN, the bound adaptor would lie parallel to the plane of the membrane, associated with the docking protein via the core domain (Fig. 10). Note that our experiments do not preclude a role for the Jd or the if1 subunit in the docking process. Anchored by a docking protein, a growing adaptor-containing coat structure might trap select TGN membrane proteins through multiple weak interactions between the 'Y subunit appendages and the cytoplasmic portions of the selected proteins. These interactions would restrict the lateral diffusion of the cargo, resulting in their concentration and preferential inclusion within a coated pit. At some point during the assembly of the coated vesicle, AP-1 may be released from the putative docking protein, perhaps regulated by nucleotide hydroly-sis of ARF·GTP. Several lines of evidence, including the data shown here, support the notion that the clathrin-AP-1 interaction would be maintained by the f31 subunit (14)(15)(16). Polygonally arranged c1athrin would remain membrane-associated due to multiple, cooperative low affinity adaptor-receptor interactions, docking protein-AP-1 interactions at the growing edges of the lattice, and perhaps adaptor-adaptor interactions, as exemplified by AP-2.
Upon budding and removal of the clathrin coat by the uncoating ATPase Hsc70 (3,45,46), the interaction between individual adaptors and receptors may not be sufficiently strong to maintain adaptors on the membrane. If this were the case, upon removal of the clathrin, the remaining coat components would be released from the budded vesicle surface without necessitating additional uncoating factors. However, the multivalency of the proposed adaptor-membrane association may result in only slow release of the adaptors (47,48) and could be regulated by specific cytoplasmic factors (49). We have not yet examined what requirements the clathrin has to meet in order to be recruited by membrane-bound AP-1 and if there is any difference between cytosolic and coated vesicle-derived clathrin as has been observed for AP-2-mediated clathrin recruitment (50). These and other questions concerning the role of light chains and clathrin domains can now be studied using our membrane binding assay.