# 2007 The Authors Journal compilation # 2007 Blackwell Publishing Ltd Traffic 2007; 8: 582–593 Blackwell Munksgaard doi: 10.1111/j.1600-0854.2007.00554.x Multiple and Stepwise Interactions Between Coatomer and ADP-Ribosylation Factor-1 (Arf1)-GTP
Zhe Sun1,†, Frank Anderl1,†, Kathrin Fro¨ hlich1, suggesting that these complexes are comprised of Liyun Zhao1, Stefan Hanke2, Britta Bru¨ gger1, conserved structural elements (13). The structure of the Felix Wieland1,*, Julien Be´ thune1,* core complexes of adaptor protein (AP) complex 1 and 2 (AP-1 and AP-2) were solved by X-ray crystallography 1Biochemie-Zentrum der Universita¨t Heidelberg (BZH), (14,15). Two 100-kDa subunits are built of a trunk domain Im Neuenheimer Feld 328, D-69120 Heidelberg, connected via a flexible linker to a smaller appendage Germany domain. The trunk domains together with the smaller sub- 2 Max-Planck-Institute for Biochemistry, Am Klopferspitz units m and s form a densely packed core that contains the 18, D-82152 Martinsried, Germany sites for specific interactions with a donor membrane. The †These authors contributed equally to this work coatomer subunits b and g together with d and z are likely to *Corresponding author: Julien Be´thune, [email protected] or Felix Wieland, represent a similar core of coatomer. Structural homology [email protected] was corroborated for the appendage domains of a2and b-adaptin and g-COP by X-ray crystallography (16,17).
The small GTPase ADP-ribosylation factor-1 (Arf1) plays Formation of a vesicle includes binding to the donor mem- a key role in the formation of coat protein I (COP I)-coated brane of coat proteins and subsequent membrane deforma- vesicles. Upon recruitment to the donor Golgi membrane tion to sculpt out a bud. In the clathrin system, this is by interaction with dimeric p24 proteins, Arf1’s GDP is achieved by an Arf-GTP-dependent two-step process: bind- exchanged for GTP. Arf1-GTP then dissociates from p24, and together with other Golgi membrane proteins, it ing of an adaptor complex to the GTPase followed by recruits coatomer, the heptameric coat protein complex recruitment of clathrin heavy and light chains. In contrast, of COP I vesicles, from the cytosol. In this process, Arf1 COP I vesicles are formed by an Arf-GTP-dependent one- was shown to specifically interact with the coatomer b step recruitment of coatomer, implying that both adaptor and and g-COP subunits through its switch I region, and with clathrin functions are unified in a single coatomer complex. e-COP. Here, we mapped the interaction of the Arf1-GTP switch I region to the trunk domains of b and g-COP. Site- We previously used an in vitro translation technique to directed photolabeling at position 167 in the C-terminal introduce photolabile amino acids at defined positions in helix of Arf1 revealed a novel interaction with coatomer Arf1. Through the use of these photolabile Arf1 derivatives, via a putative longin domain of d-COP. Thus, coatomer is linked to the Golgi through multiple interfaces with we identified the coatomer subunits b-COP and g-COP as membrane-bound Arf1-GTP. These interactions are binding partners for the switch I region of Arf1-GTP located within the core, adaptor-like domain of coatomer, (18,19). In these functional binding assays with myristoy- indicating an organizational similarity between the COP I lated full-length Arf1 derivatives, Golgi enriched mem- coat and clathrin adaptor complexes. branes and coatomer, these GTP-dependent interactions were found after recruitment of coatomer to the mem- Key words: coat protein I, Golgi, membrane traffic, small branes, as well as in isolated COP I vesicles, indicating GTPase, vesicular transport their functional relevance. Another study described binding Received 19 September 2006, revised and accepted for of Arf1-GTP to b-COP and e-COP in a two-hybrid assay publication 8 February 2007, published online 26 March (20). In the clathrin system, GTP-dependent interaction of 2007 Arf1 through its switch I region with the adaptor com- plexes AP-1, 3 and 4, was attributed to the trunk domains of the larger adaptins. Coat protein I (COP I) vesicles mediate lipid and protein transport in the early secretory pathway of eukaryotic cells. In this study, we have extended our Arf1 photo-cross- Coatomer (1–3), the protein coat of these vesicles, is linking studies to map its interactions with coatomer on the recruited to Golgi membranes by the small, 20-kDa, Ras- Golgi membrane. We report that activated Arf1 binds the like GTPase Arf1 in its GTP-bound form (4,5). Inactivation trunk domains of b-COP and g-COP. In addition, we of Arf1 through GTP hydrolysis is sufficient to release describe two novel interactions between Arf1 and coat- coatomer from membranes and thus promotes vesicle omer: one involves a putative longin domain of d-COP and uncoating (6–8). Similarly, Arf1 plays a critical role in the the other the b0-COP subunit. These results underline recruitment of the clathrin adaptor complexes AP-1 (9), AP- structural and functional similarities between clathrin 3 (10) and AP-4 (11). Subunits of coatomer and clathrin adaptor complexes and coatomer and predict a potential adaptor complexes share weak sequence homology (12), interaction of adaptor m-chains with Arf1-GTP. In addition,
582 www.traffic.dk Arf1 and Coatomer Interactions they show differences that might relate to the mechanistic differences in vesicle formation that exist between the clathrin and the COP I systems.
Results
Expression of Arf1 proteins with site-directed photolabile amino acids Photolabile amino acid derivatives within polypeptides facilitate the identification of specific protein–protein inter- actions. Upon UV irradiation, highly reactive species are generated that covalently cross-link to binding partners within 3-A˚ distance (21). Until recently, generation of site- directed photolabile proteins was dependent on chemical synthesis or in vitro translation systems. Hence, photo- labile proteins were available in very limited quantities, precluding proteomic analysis and restricting identification of cross-link partners to Western blotting for known candidates. Novel binding partners could not be identified using these methods. Recently, an in vivo expression system was described that allows the production of large amounts of photolabile proteins in Escherichia coli using Figure 1: Expression of site-directed photolabile human Arf1 a benzophenone (Bp) amino acid derivative (22). We have proteins in E. coli. A) Sites within Arf1 where photolabile amino adapted this system to produce 16 site-directed photo- acid derivatives were introduced. SW1: switch I region, ISW: labile derivatives of full-length, myristoylated Arf1. Three of interswitch region, SW2: switch II region. The three labeled them could be cross-linked to putative binding partners and positions indicate the derivatives that yielded cross-linked inter- were characterized. In these derivatives, Bp residues action partners and are characterized in the present study. B) replaced amino acid positions I46 and I49 in the switch I Expression of wild-type Arf1 (WT) and photolabile derivatives region, and Y167 in the C-terminus (Figure 1A); each was (lane 2, 3) or was not (lane 1) induced at 278C by arabinose (Ara.) in the presence (lane 3) or absence (lanes 1 and 2) of derivative was obtained in up to microgram quantities p-benzoyl-L-phenylalanine (Bp). Expression products were ana- (Figure 1B). lyzed by Western blotting with an anti-Arf1 (C-terminus) antibody. Note that when produced in E. coli, Arf1 has a mixture of The photoactivable derivatives behave like myristoylated (myr.) and non-myristoylated (nonmyr.) forms (47). wild-type Arf1 For COP I-vesicles formation, Arf1 in its GDP bound form is transiently recruited by dimeric p23 protein to the Golgi Golgi membranes, photolabile Arf1-I46Bp, I49Bp and membrane (23–25). Thereafter, exchange of GDP for GTP Y167Bp were incubated with Golgi membranes and is catalyzed by a guanine nucleotide exchange factor coatomer in the presence of GDPbS or GTPgS, a non- (26,27), and the activated GTP-bound Arf1 is stably asso- hydrolyzable GTP analog. Membranes were then separated ciated with the membrane. As a result, about threefold from unbound material by centrifugation and irradiated at more Arf1-GTP is bound to Golgi membranes than Arf1- 366 nm to drive photo-cross-linking of Bp residues. Analysis GDP (28). by SDS–PAGE and Western blotting with antibodies against Arf1 and coatomer subunits revealed the known interac- To test their functionality, we assessed nucleotide- tions of Arf1 with b-COP and g-COP (Figure 3A and B, lane dependent binding of the Arf1 photolabile proteins to Golgi 4, as indicated). The lower band (Figure 3B, lane 4) repre- membranes. For all Arf1 species, replacement of GDP with sents Arf1 covalently linked to g2-COP, an isotype of g-COP GTP caused a threefold enhancement of Golgi membrane (29,30). Three additional Arf1 complexes were observed binding (Figure 2). This ratio is in very good agreement (Figure 3A and C, indicated as Arf1-I46 þ x, Arf1-I46 þ y, with the previous quantification of nucleotide-dependent Arf1-Y167 þ z) with apparent molecular masses of about binding of Arf1 to Golgi membranes using radiolabeled 150 kDa (Arf1 þ partner y), 130 kDa (Arf1 þ partner x) and Arf1 (28). Thus these photolabile Arf1 species were then 80 kDa (Arf1 þ partner z). used to map the binding sites on b-COP and g-COP and to search for additional binding partners. Arf1-GTP interacts with the trunk domains of b-COP and g-COP GTP-dependent interactions of Arf1 with coatomer Cross-linking and two-hybrid experiments have revealed To probe the interfaces of Arf1 with coatomer under that Arf1-GTP binds clathrin adaptor complexes through functional conditions, i.e. activated, GTP-loaded Arf1 on the trunk domains of AP-1, AP-3 (31) and AP-4 (11). Here,
Traffic 2007; 8: 582–593 583 Sun et al.
Figure 2: Photolabile Arf1 deriva- tives behave like wild-type Arf1 in a Golgi-binding experiment. Arf1-WT or photolabile derivatives were incu- bated with Golgi membranes in the presence of GDPbS (lane 1) or GTPgS (lane 2) and the membranes were then isolated by centrifugation. Membrane- bound Arf1 was detected by Western blotting with an anti-Arf1 (C-terminus) antibody. The signals detected for Golgi-bound Arf1-GDPbS and Arf1- GTPgS were quantified with the soft- ware UN-SCAN-IT gel (Silk Scientific) from three independent experiments. For wild-type and photoactive deriva- tives, the ratio between the two Arf1 forms is indicated on the right. The error bars represent the standard error of the mean. we have used limited proteolysis combined with Western with activated Arf1 through its trunk domain, and not its blotting to analyze the domains of b-COP and g-COP that appendage domain (Figure 5). interact with Arf1-GTP. In Figure 4, analysis of b-COP is depicted. After irradiation to induce covalent cross-linking, Golgi membranes were incubated with the polyspecific Novel interaction of Arf1-GTP with d-COP protease thermolysin and analyzed by Western blotting A novel product was observed upon irradiation of Arf1- with antibodies against Arf1 (Figure 4, upper panel). At low Y167Bp-GTP (Figure 3C, partner z). We used an immuno- protease concentration, the cross-link products seemed logic approach combined with mass spectrometry to unaffected (lane 4), whereas time-dependent degradation characterize the proteins cross-linked with this species of was observed at a higher protease concentration (lanes 6, Arf1-GTP. To this end, we upscaled the reaction and, after 8 and 10). A stable degradation product was observed with irradiation, Arf1-Y167Bp and cross-linked products were an apparent molecular mass of 70 kDa, a size that would isolated by immunoprecipitation with an anti-Arf1 antibody. be expected for a covalent adduct of the b-COP trunk Proteins were separated by SDS–PAGE and detected by domain with Arf1. Decoration with an antibody against the Coomassie staining (not shown). The region corresponding appendage domain of b-COP revealed, at high protease to 80 kDa was analyzed by mass spectrometry in both concentration, a protein fragment of 40 kDa, indicating irradiated and non-irradiated samples. As shown in a non-cross-linked appendage domain of b-COP (Figure 4, Table 1, the only proteins unique to the irradiated sample middle panel, lanes 5–10). In contrast, immunodetection are Arf1 and d-COP. The predicted apparent molecular with an antibody directed against the b-COP trunk domain mass of this pair (79 kDa) is consistent with their migration (Figure 4, lower panel) revealed a non-cross-linked trunk behavior on the gel. This previously unreported binding of domain at 55 kDa and an additional trunk domain of d-COP to Arf1 was further analyzed using recombinant higher apparent molecular mass, consistent with a cross- GST-tagged d-COP and ND17-Arf1, a truncated recombi- link to Arf1 (lanes 6, 8 and 10). This band co-migrates with nant version of Arf1 lacking the N-terminal a-helix and thus the Arf1 band in the upper panel of Figure 4, lanes 6, 8 and 10. soluble in both GDP- and GTP-bound forms. The two well- characterized ND17-Arf1 mutants Q71L and T31N were The interaction of g-COP with Arf1-I49Bp-GTP was ana- used: ND17-Arf1-Q71L is always bound to GTP as it is lyzed in a similar manner. Like b-COP, g-COP interacts unable to hydrolyze GTP and ND17-Arf1-T31N is locked in
584 Traffic 2007; 8: 582–593 Arf1 and Coatomer Interactions
Figure 3: GTP-dependent interactions between Arf1 and coatomer. Arf1 photolabile derivatives (Arf1-I46-Bp in A; Arf1-I49-Bp in B; Arf1-Y167-Bp in C) were incubated in the presence of Golgi membranes and coatomer, with GDPbS (lanes 1 and 2 in A–C) or GTPgS (lanes 3–6 in A; lanes 3–8 in B; lanes 3 and 4 in C). Samples were then centrifuged to isolate membranes which were then resuspended in reaction buffer and UV irradiated (lanes 2, 4 and 6 in A; lanes 2, 4, 6 and 8 in B; lanes 2 and 4 in C). Photo-cross-linked products were analyzed by Western blotting with anti-Arf1 antibodies (lanes 1–4 in A–C), anti-b-COP antibodies (lanes 5 and 6 in A), anti-g1-COP antibodies (lanes 5 and 6 in B), or the antibody g-r (which recognizes both g1-COP and g2-COP; lanes 7 and 8 in B). Previously characterized cross-linked products are indicated. Arf1-I46-Bp gives rise to two previously unknown adducts (labeled as Arf1-I46 þ x and Arf1-I46 þ y). Arf1-Y167-Bp can be cross-linked to an unknown partner labeled as Arf1-Y167 þ z. the GDP bound form. Immobilized GST-d-COP was able to which d-COP is proteolytically nicked to a stable fragment specifically pull down ND17-Arf1 but in contrast to the of 31 kDa. This d-COP fragment is still part of a coatomer cross-linking assay where d-COP is part of the coatomer complex, as it is present in a coatomer sample immuno- complex, the interaction of the isolated d-COP with Arf1 is precipitated with an antibody directed against another COP not nucleotide dependent (Figure 6). This indicates that subunit (Figure 7A, lanes 1 and 2). Therefore, rather than the binding site for Arf1 on d-COP is not restricted to Arf1- cytosol as a source of coatomer, samples of the purified GTP per se. coatomer complex were further investigated. Incubation of Arf1-Y167Bp with Golgi membranes and purified coat- Further mapping of the interaction between Arf1-Y167Bp omer, followed by irradiation, gave rise to two bands, and d-COP was facilitated by our observation that coat- one which corresponds to Arf1-GTP cross-linked to full- omer purifies in two forms: one that is intact and one in length d-COP (at 80 kDa), and another at 50 kDa
Traffic 2007; 8: 582–593 585 Sun et al.
Figure 4: Arf1 directly interacts through its SWI region with the trunk domains of b-COP. Binding of Arf1-I46-Bp to coatomer subunits and photo-cross-linking was per- formed as described in Figure 3. Samples without UV irradiation were treated as negative controls (lanes 1, 3, 5, 7 and 9). Samples were incubated with thermolysin as indicated, and analyzed by Western blotting with the anti-Arf1 (upper panel), anti-b-COP appendage domain (middle panel; anti-b-COP appendage) or anti-b-COP trunk domain antibodies (lower panel; anti-b-COP trunk).
Figure 5: Arf1 directly interacts through its SWI region with the trunk domains of g-COP. Same ex- periment as in Fig. 4 with Arf1-I49-Bp. The interaction between Arf1-I49-Bp and the g-COPI trunk domain is demon- strated by Western blotting with the anti-Arf1, g-r (recognizes trunk domains of g-COPs) and anti-g-COP appendage antibodies. Asterisks indicate bands that non-specifically cross-react with the antibodies.
586 Traffic 2007; 8: 582–593 Arf1 and Coatomer Interactions
Table 1: Proteins identified in the non-irradiated and the irradiated which showed a UV-dependent mass shift for b0-COP samples upon incubation of coatomer with Golgi membranes and Arf1-I46Bp-GTPgS (Figure 8). The small amount of cross- UV þUV linked material has thus far prevented us from mapping Ig gamma chain C-region Ig gamma chain C-region more precisely the interaction site on b0-COP. Ig gamma heavy chain Ig gamma heavy chain constant region – rabbit constant region – rabbit The second additional band observed by photo-cross- Ig kappa chain Ig kappa chain linking of Arf1-I46Bp corresponded to a product of 150 ADP-ribosylation factor kDa (Figure 3A, indicated as Arf1-I46 þ y). This material Coatomer delta subunit was electroeluted from the gel and resolved on a gel with a larger pore size for a further purification (see Materials and Methods section). Mass spectrometry analysis of this (Figure 7A, lanes 3–7). This band was recognized by both band yielded peptides corresponding to Arf1 and b-COP the anti-Arf1 antibody (Figure 7A, lane 4) and the antibody (Table 3). The apparent molecular mass of this covalent against d-COP (Figure 7A, lane 7). The d-COP fragment product suggests that b-COP is exclusively bound to two that gave rise to this band was analyzed by mass spec- Arf1-GTP molecules. This was corroborated by the fact trometry. As shown in Figure 7B, a comparison of full- that this species was also observed after denaturation of length d-COP with the sequence of its 31-kDa degradation cross-linked samples and immunoprecipitation with anti- product shows that Arf1 binds the N-terminus of d-COP. bodies against either b-COP or Arf1, but not with anti- bodies against other coatomer subunits. Furthermore, in Interactions of Arf1-GTP with b0-COP and b-COP the mapping experiment depicted in Figure 4, after limited Upon incubation with Golgi membranes and cytosol, and proteolysis, an additional band appeared at 90 kDa after subsequent UV irradiation, Arf1-I46Bp produced two addi- longer exposure time, in fair agreement with a trunk tional cross-linked products (running at 130 and 150 kDa) domain cross-linked to two Arf1-GTP molecules. Taken of apparent higher molecular mass than Arf1 cross-linked together, these observations indicate that more than one to b-COP (Figure 3A, indicated as Arf1-I46 þ x and Arf1-I46 molecule of Arf1-GTP can bind to coatomer. þ y). To identify the 130-kDa band (Arf1-I46 þ x), we followed the protocol described above for the Arf1/d–COP Contribution of I46, I49 and Y167 to the binding of cross-linked product, by upscaling the size of the reaction, Arf1 to coatomer immunoprecipitation with the antibody directed against To estimate the contribution to the binding to coatomer of Arf1, and detection by Coomassie staining after separation each of the three sites in Arf1, the amino acids at positions by SDS–PAGE. Analysis by mass spectrometry of the 46, 49 and 167 were exchanged for alanines, and the material running at 130 kDa region led to the identifica- resulting Arf1 variants analyzed in a coatomer recruitment tion of peptides from Arf1 and the b0-COP subunit of assay on Golgi membranes. As a result, single Arf1 mutants coatomer in the irradiated but not in the non-irradiated (Arf1-I46A, Arf1-I49A and Arf1-Y167A) showed a mildly sample (Table 2). This was subsequently confirmed by reduced ability to recruit coatomer to membrane (Figure 9), Western blotting analysis with an anti-b0-COP antibody, while an Arf1 variant with all three positions mutated to alanines showed a more pronounced reduction of coatomer binding. This suggests that all three interfaces occur syner- gistically during coatomer recruitment to Golgi membranes. As expected, none of the Arf1 variants could abolish coat- omer binding completely, indicating that these individual positions represent only parts of binding interfaces. These data do not, however, completely rule out the possibility of a sequential binding to coatomer of the three sites.
Discussion
The basic mechanisms of membrane deformation that lead to the formation of transport vesicles are still poorly Figure 6: Recombinant d-COP binds to Arf1 regardless of its understood. The goal of our investigations is to understand nucleotide state. Recombinant GST-tagged d-COP was coupled how coat proteins are recruited to and polymerize on to glutathione sepharose beads and incubated with ND17-Arf1- membranes to form COP I vesicles. Arf1 is a key player Q71L (GTP-bound) or ND17-Arf1-T31N (GDP bound) as indicated; empty beads were used as a negative control. After extensive in the recruitment of the coat protein complex coatomer washing the samples were analyzed by Western blot with anti- to Golgi membranes, and thus a detailed characterization bodies directed against d-COP or Arf1. Beads coupled to GST of this interaction will provide significant insight into the showed no binding to Arf1 (not shown). mechanism of coat assembly. In this study, we used
Traffic 2007; 8: 582–593 587 Sun et al.
Figure 7: Arf1-Y167-Bp interacts directly with an N-terminal 31-kD fragment of d-COP. A) Rabbit liver coatomer was purified by immunoprecipitation under native conditions with antibodies directed against b0-COP. The immunoprecipitated product was separated by SDS-PAGE and visualized by Coomassie staining (lane 1). Part of the sample was run simultaneously on SDS–PAGE and analyzed by Western blotting. The band labeled as d-COP and the band marked by an arrow correspond to the two bands that were immunoreactive with antibodies against d-COP (lane 2). Lanes 3–7: Arf1-Y167-Bp was incubated with GTPgS in the presence of Golgi membranes and rabbit liver coatomer. Samples were then centrifuged and the isolated membranes were UV irradiated (lane 4 and 7). Membrane-bound photo- cross-linked products were analyzed by Western blotting with anti-Arf1 antibodies (lanes 3 and 4) and anti-d-COP antibodies (lanes 5–7). In lane 5, a sample containing only Golgi membranes was loaded. The bands in lanes 5–7 indicated with asterisks represent cross-reacting Golgi membrane proteins. B) The bands marked by arrows in A) were identified by mass spectrometry as d-COP and an N-terminal fragment thereof. Identified peptides are highlighted in bold. The box indicates the anti-d-COP antibody epitope.
588 Traffic 2007; 8: 582–593 Arf1 and Coatomer Interactions
Table 2: Proteins identified in the non-irradiated and the irradiated Table 3: Proteins identified in the non-irradiated and the irradiated samples samples