Specific Interaction of Golgi Coatomer Protein a -COP with Phosphatidylinositol 3,4,5-Trisphosphate*

The phosphoinositide binding selectivity of Golgi coatomer COPI polypeptides was examined using photoaffinity analogs of the soluble inositol polyphosphates Ins(1,4,5)P 3 , Ins(1,3,4,5)P 4 , and InsP 6 , and of the poly- phosphoinositides PtdIns(3,4,5)P 3 , PtdIns(4,5)P 2 , and PtdIns(3,4)P 2 . Highly selective Ins(1,3,4,5)P 4 -displace- able photocovalent modification of the a -COP subunit wasobservedwitha p -benzoyldihydrocinnamide(BZDC)-containing probe, [ 3 H]BZDC-Ins(1,3,4,5)P 4 . A more highly phosphorylated probe, [ 3 H]BZDC-InsP 6 probe labeled six of the seven subunits, with only b , b * , d , and e -COP showing competitive displacement by excess InsP 6 . Im- portantly, [ 3 H]BZDC- triester -PtdIns(3,4,5)P 3 , the lipid with the same phosphorylation pattern as Ins(1,3,4,5)P 4 , showed specific, PtdIns(3,4,5)P

The conversion of the inactive, soluble ARF-GDP to active, membrane-associated ARF-GTP form requires a GTP exchange factor and exposes the buried N-myristoyl group necessary for membrane localization (9 -11). Both the stimulation of phospholipase D by ARF and the activation of ARF GTPase activity (12) that is important in uncoating of the vesicles require PtdIns(4,5,)P 2 (13,14). In addition to phospholipase D, essential roles for phosphatidylinositol transfer protein, free diacylglycerol, and protein kinase C have been recently incorporated into models for the budding and vesicle scission processes (15,16).
The detailed roles of COPI polypeptides in the directional transport of membrane proteins require further investigation (17). COPI proteins were first proposed to mediate non-selective transport of proteins from the endoplasmic reticulum (ER) through the Golgi complex to the cell surface (8,18). Evidence in support of the anterograde role, in which COPI-coated vesicles, phospholipase D, and a novel p24 family protein are involved in moving newly translated proteins from the ER to the Golgi, was recently summarized (19). In an alternative model, COPI may be involved principally in selective retrograde transport of membrane proteins from the Golgi complex to the ER by selective association with a dilysine retrieval motif; a separate set of coat proteins, COPII, was identified that seems to function only in the anterograde vesicular transport of cargo from the ER to the Golgi (20). The recent demonstration that yeast COPII directs the formation of vesicles from the ER, and that these vesicles capture both cargo and necessary components of the molecular secretory apparatus further supports the importance of the COPII in anterograde traffic. The most recent data suggests that COPI may be involved in traffic in both directions, as two distinct populations of Golgiassociated COPI-coated vesicles were found in pancreatic cells, each with a different cargo implicating an intended transport direction for the vesicle population (21).
The polypeptide subunits of COPI have been characterized, and selected aspects of their individual roles are known. First, ␣-COP is the clathrin-like subunit. It is localized to coated transport vesicles and coated buds of Golgi membranes derived from Chinese hamster ovary cells (22). The gene for ␣-COP has been cloned and characterized (22) from Saccharomyces cerevisiae. Disruption of this gene in yeast has been found to be lethal. ␣-COP has also been shown to be required for ER localization of dilysine-tagged proteins (23). A novel human gene, HEP-COP, has been isolated, the product of which is highly homologous to yeast ␣-COP (24,25). ␤-COP, which is homologous to the ␤Ј subunit of assembly protein 2 (AP-2) (26), has been reported to be essential for transport of protein from the ER to Golgi in vitro (27). The binding of ␤-COP with Golgi membranes has been shown to be enhanced by non-hydrolyz-able GTP analogs and AlF 4 Ϫ . ␤Ј-COP is homologous to ␤ subunits of heterotrimeric G-proteins. ␥-COP binds to dilysine motifs of membrane proteins (28) and is related to Sec 21, a secretory mutant of the yeast S. cerevisiae (29). A single point mutation in ⑀-COP has been shown to result in temperaturesensitive lethal defects in membrane transport (30). -COP has sequence homology to AP-17 and AP-20 subunits (31). The B complex (␣-COP, ␤Ј-COP, and ⑀-COP) has been found to interact directly and bind to membranes (32), while ␥-COP and -COP form a second subcomplex (33).
Direct binding studies using purified coatomer isolated from bovine liver cytosol show that coatomer specifically binds both Ins(1,3,4,5)P 4 and InsP 6 with affinities of 0.1 and 0.2 nM, respectively (34). The degree of phosphorylation of the inositol polyphosphates (InsP n ) has been proposed to dictate the order of binding to coatomer (InsP 6 Ͼ InsP 5 Ͼ InsP 4 ) (35). Since dissociation of the COPI polypeptides can only be accomplished under conditions that would not permit measurement of InsP n binding, the affinities of individual COPI subunits for InsP n are unknown. Moreover, binding of coatomer complexes to phosphatidylinositol polyphosphates (PtdInsP n s) remains unreported. To address the InsP n and PtdInsP n binding specificities of individual COPI subunits, and to obtain evidence to support the roles of these high affinity interactions in vesicular trafficking, we employed a photoaffinity labeling approach with benzophenone-containing InsP n and PtdInsP n analogs (36). The benzophenone photophore allows handling in ambient light, activation at wavelengths Ͼ320 nm, and covalent labeling of active site residues in hydrophobic regions of proteins with high efficiency (37,38). Photoaffinity analogs of soluble inositol polyphosphates Ins(1,4,5)P 3 , Ins(1,3,4,5)P 4 , and InsP 6 (39,40) and the lipid-containing polyphosphoinositides PtdIns-(3,4,5)P 3 (41), PtdIns(4,5)P 2 (41), and PtdIns(3,4)P 2 (42) are used herein to determine the polyphosphoinositide selectivity for the COPI subunits in Golgi coatomer (39).
Purification and Photoaffinity Labeling of Golgi Coatomer-Coatomer was purified from bovine liver cytosol as described (45) to give material of about 60% purity after Mono Q chromatography. An aliquot of this partially purified Golgi coatomer (6 g, 0.3 M) was incubated with 30 l of buffer A (150 mM KCl, 10% glycerol, 0.5 mM dithiothreitol, 25 mM HEPES/KOH, pH 8.9) and either [ 3 H]BZDC-InsP n or [ 3 H]BZDC-PtdInsP n (0.5 Ci, 0.28 M). The specificity and affinity of binding were determined by adding an aliquot of an aqueous solution or suspension of the corresponding unlabeled PtdInsP n or InsP n (0.28 mM) directly into the incubation mixture. Samples were equilibrated at 4°C for 10 -15 min in a 96-well plate. The wells in the plate were then aligned with the axis of a UV light source with minimum distance maintained between the bottom of the wells and the bulb. The samples were photolysed for 45 min at 4°C (360 nm at 1900 W/cm 2 ) (32). Following irradiation, sample buffer (5 ϫ) was added to the samples and the proteins were separated by SDS-PAGE (10% Laemmli gels), stained with Coomassie Blue; the gel was processed for fluorography as described (32), and exposed to XAR-5 x-ray film for 7-14 days at Ϫ80°C. There was no covalent incorporation of the photolabel in the absence of UV irradiation. Fluorograms were digitized on a UMAX-UC 840 scanner using Adobe Photoshop (Macintosh version 2.0.1). The densities of the bands on the fluorogram were determined by NIH IMAGE 1.59 to calculate relative incorporation.
Disassembly of Golgi Coatomer (33)-Protein (30 g, 0.05 M) was incubated in 1 ml of buffer B (1 M NaCl, 20 mM Tris, pH 7.5, 2 mM EDTA, 1 mM dithiothreitol, 0.5% Triton X-100) and gently shaken and incubated for 1 h at 4°C. This mixture was applied to a gel filtration high performance liquid chromatography column (TSK-GEL, G3000SW, 7.5 mm inner diameter), flow rate 0.5 ml/min, 30-min run, isocratic elution using buffer C (25 mM Tris, pH 7.46, 0.4 M NaCl). The high performance liquid chromatography system was calibrated using molecular weight protein markers, and 1-min fractions were collected. The 16-min fraction, corresponding to mass 315 kDa, was photolabeled as described above.
Immunoprecipitation of Golgi Coatomer Subunits (46)-Coatomer (6 g, 0.3 M) was photoaffinity labeled with [ 3 H]BZDC-PtdIns(3,4,5)P 3 (0.5 Ci, 0.28 M) as described above. The labeled mixture was then incubated for 2 h at 4°C with the indicated antibodies (3 and 4 l of anti-␣-COP antisera and 1 and 2 l of anti-␤-COP antisera, respectively) in 250 l of immunoprecipitation buffer (IP buffer: 20 mM Tris-HCl, pH 7.5, 2 mM EDTA, 0.15 M NaCl, 0.5% Triton X-100). Protein A-Sepharose beads (20 l) (Pharmacia) were then added to the mixture and incubated for another 2 h at 4°C. The beads were washed once with IP buffer, and then incubated overnight in IP buffer containing 1 M NaCl. Subsequent wash steps were performed in IP buffer (3 ϫ) and finally in phosphate-buffered saline, pH 7.4. Sample buffer (5 ϫ) was added to the washed pellets and the immunoprecipitate was separated by SDS-PAGE (10% Laemmli gels). The gel was stained with Coomassie Blue, and processed for fluorography as described above.

RESULTS
Synthesis of Photoaffinity Labels-Five photoaffinity labels ( Fig. 1) were used to study the coatomer subunit specificity and selectivity for different InsP n and PtdInsP n probes. Each InsP n and PtdInsP n photoaffinity probe was prepared from the corresponding 1-or 2-O-(-aminoalkyl)-InsP n or -PtdInsP n derivative with the heterobifunctional reagent [ 3 H]BZDC-NHS ester (38) and had the same nominal specific activity of 42.5 Ci/ mmol. This ensured that levels of radioactivity in each probe corresponded to equivalent concentrations, thereby allowing direct comparison of relative efficiencies of photoattachment.
Photoaffinity Labeling of Golgi Coatomer-Selectivity of the probes for coatomer subunits was determined by photoaffinity labeling experiments employing [ 3 H]BZDC-InsP n or [ 3 H]BZDC-PtdInsP n probes. Specific binding was determined by competitive displacement of photocovalent modification in the presence of a 1000-fold excess of unlabeled (Ptd)InsP n . The specificity could be approximated by the difference between the total binding (no competitor) and binding in the presence of the competing ligand. Since coatomer has been demonstrated to bind Ins(1,3,4,5)P 4 and InsP 6 with subnanomolar affinities (0.1 and 0.2 nM, respectively (34)), we initially employed the Coatomer has also been shown to have high affinity binding interactions with InsP 6 . To study the polypeptides involved in this binding, we employed the P-2-tethered [ 3 H]BZDC-InsP 6 probe. Fig. 3 shows the labeling obtained with this probe. [ 3 H]BZDC-InsP 6 exhibited intense labeling of most proteins in this partially purified preparation. Interestingly, addition of competitors InsP 6 and Ins(1,3,4,5)P 4 (0.28 mM) competitively displaced the labeling from ␤-COP, ␤Ј-COP, ␦-COP, and ⑀-COP, indicating specific labeling of these subunits. However, labeling of the ␣-COP and ␥-COP resisted competitive displacement, suggesting that the soluble probe might not be sufficient to displace a combined electrostatic-hydrophobic interaction afforded by the photoaffinity analog.
Evidence from our laboratories using photoaffinity labeling (39,40) has implicated highly selective binding of PtdInsP n derivatives to a number of proteins important in the budding and fusion of lipid bilayers, as well as those known to be recruited to polyphosphoinositide-rich membranes. Thus, coato-mer was labeled with a series of [ 3 H]BZDC-triester-PtdInsP n (n ϭ 2 or 3) probes (40) that have diacylglycerol moieties for membrane anchoring, the correct phosphorylation pattern on the D-myo-inositol ring, and a photoactivatable group that can covalently modify proteins recruited to the membrane surface. Fig. 4 shows that each of the PtdInsP n probes, [ 3 H]BZDCtriester-PtdIns(3,4,5)P 3 , -PtdIns(3,4)P 2 , and -PtdIns(4,5)P 2 ( Fig. 1), labeled only ␣-COP with high subunit and PtdInsP n selectivity.
To verify that the labeled 170-kDa band was indeed ␣-COP and not an unrelated protein of molecular weight similar to ␣-COP, immunoprecipitation experiments were performed with the anti-␣-COP and anti-␤-COP subunit specific antibodies. Fig. 6 shows the SDS-10% PAGE gel and corresponding fluorogram of the immunoprecipitation of the photoaffinitylabeled protein. The anti-␤-COP antisera immunoprecipitated the ␤-subunit at 110 kDa, but no radiolabel was present. (The rigorous washing conditions used for the immunoprecipitated, protein A-Sepharose bound protein, including an overnight incubation in 1 M NaCl, effectively dissociates all the COPI subunits and would remove any cargo proteins.) In contrast, the anti-␣-COP antibody immunoprecipitated the 170-kDa ␣-COP subunit (as seen on the SDS-PAGE gel). Moreover, this subunit contained the [ 3 H]BZDC-triester-PtdIns(3,4,5)P 3 label, providing conclusive evidence for the interaction of this subunit with PtdIns(3,4,5)P 3 . Negligible amounts of radioactivity were detected in the wash buffers, in contrast to the label recovered in the high salt washes of the anti-␤-COP-precipitated proteins. Interestingly, the anti-␣-COP antibody used was developed against the C terminus of the ␣-subunit, strongly suggesting that the PtdIns(3,4,5)P 3 -binding site on ␣-COP is not contained within the epitope for antibody recognition on this subunit.
To examine the possibility that PtdIns(3,4,5)P 3 binding might be involved in the ARF-mediated recruitment of coatomer, two experiments were performed. Brefeldin A (BFA), which has been shown to decrease ␤-COP binding to membranes (48), had no effect on the [ 3 H]BZDC-triester-PtdIns-(3,4,5)P 3 photolabeling of ␣-COP (Fig. 7A). This fungal metabolite has been shown to prevent the assembly of coatomer onto the membrane by inhibiting the GTP-dependent interaction of ARF with the Golgi membrane (49). Addition of up to 200 M BFA did not affect the covalent modification of ␣-COP by the PtdIns(3,4,5)P 3 photoaffinity probe. Similarly, addition of up to 10 mM GTP was found to have no effect on this binding. These results further support a specific interaction between PtdIns-(3,4,5)P 3 and ␣-COP.
Finally, the effects of salts on this interaction were investigated (Fig. 7B), since the high affinity InsP 6 -coatomer interaction had been reported to exhibit salt dependence (34). Addition of up to 500 mM CaCl 2 had no effect on the labeling of ␣-COP by [ 3 H]BZDC-triester-PtdIns(3,4,5)P 3 . However, presence of greater than 300 mM KCl inhibited labeling. Interestingly, addition of both 5 mM GTP and 150 mM KCl was also found to inhibit the labeling.  0.10, 150, and 500 mM, respectively). Lanes e-g show the effects of increasing amounts of KCl in buffer A (300, 500, and 1000 mM, respectively).

PtdIns(3,4,5)P 3 Binds Golgi Coatomer
their ligand-binding sites, and verification of their PtdInsP n and InsP n selectivity (36). The advantages of benzophenone over the classical arylazide photochemistry include improved chemically stability of ligands and adducts, stability in ambient light, low background from nonspecific labeling, and the efficient C-H insertion of the triplet diradicaloid intermediate formed by irradiation at 360 nm (39). Herein we report an application of this photochemical technique to study the subunit specificity of these benzophenone-tethered InsP n and Ptd-InsP n probes with Golgi coatomer COPI polypeptides.
The probes employed in this study ( Fig. 1) are more hydrophobic than the endogenous ligands Ins(1,3,4,5)P 4 and InsP 6 due to the presence of the photoactivatable BZDC moiety and the aminopropyl phosphate ester. Indeed, BZDC-Ins(1,3,4,5)P 4 is a reasonable structural surrogate for the inositol phospholipid PtdIns(3,4,5)P 3 (40), as previously observed for the PtdIns(3,4,5)P 3 -binding protein centaurin-␣ (47). Analogously, [ 3 H]BZDC-Ins(1,4,5)P 3 has been used as a probe to study the PtdIns(4,5)P 2 -binding sites of the pleckstrin homology domain of recombinant phospholipase C␦ 1 isozyme (50) and of recombinant human profilin I. 3 Coatomer has been shown to be similar to AP-2 and cardiac AP-3 in that all three proteins formed K ϩ channels when incorporated into planar lipid bilayers (34,51), and each exhibited high affinity binding to certain InsP n s. PtdIns(3,4,5)P 3 has been shown to be a high affinity ligand for AP-2 (52) and AP-3 (53). Phosphorylated phosphatidylinositol may cooperate with membrane proteins in the recruitment of cytosolic proteins for certain vesicle coats (50). It has been postulated that the binding of a coat protein to the head group of a phospholipid may orient the coat protein and facilitate side-to-side association through homophilic-heterophilic interaction with other proteins to generate the coat (54). To test this hypothesis, we examined PtdInsP n -coatomer interactions using photoaffinity labeling.
A human phosphatidylinositol (PI)-specific 3-kinase activity has been implicated in non-clathrin-mediated Golgi membrane traffic (55,56). This PI 3-kinase complex has been related to the yeast Vps34p-Vps15p protein sorting. Our data thus reflect the potential role of Golgi coatomer as a ligand for PtdInsP n s and emphasize the potential role of a PI 3-kinase on its recruitment to membranes. Coatomer bound to the products (PtdIns(3,4,5)P 3 and PtdIns(3,4)P 2 ) of a PI 3-kinase with higher affinity than a potential substrate PtdIns(4,5)P 2 . Also, the substrate PtdIns(4,5)P 2 was unable to displace the binding of the product PtdIns(3,4,5)P 3 .
The phosphoinositide products of PI 3-kinase have pivotal roles in regulation of protein trafficking, cell survival, cell growth, actin rearrangement, and cell adhesion (57). Indeed, the actions of a variety of proteins implicated in membrane trafficking and in exo-and endocytosis are modulated by interactions with PtdInsP n s (54). For example, PtdIns(3,4,5)P 3 binds specifically and saturably to soluble AP-2, and this binding inhibits the clathrin binding and assembly activities of this heterotetrameric protein (52). Similarly, the brain-derived assembly protein AP-3 (a.k.a. AP-180) also showed preferential binding to and functional regulation by PtdIns(3,4,5)P 3 (53). In the synaptic vesicle cycle, synaptotagmin I acts as a bimodal calcium-regulated switch, binding with high affinity to PtdIns(3,4,5)P 3 -containing liposomes at [Ca 2ϩ ] below 1 M, but preferentially to PtdIns(4,5)P 2 -containing liposomes at calcium concentrations above 10 M. In addition, phospholipase D is activated by polyphosphoinositides (13) and has been shown to mediate ARF-dependent formation of Golgi-coated vesicles (14). Ktistakis and co-workers (14) have demonstrated that purified coatomer binds selectively to artificial lipid vesicles that contain phosphatidic acid and PtdIns(4,5)P 2 .
We investigated the effects of salt concentration on the photoaffinity labeling of the COPI polypeptides, since binding of InsP n s to coatomer was previously reported to be highest at pH 8.9 with 140 mM KCl (34) and decreased with increased salt concentrations. In corroboration of these results, no photoaffinity labeling was observed at or below pH 7.5 (data not shown). Moreover, no labeling was observed above 300 mM KCl, while up to 500 mM CaCl 2 had no apparent effect on labeling. High (millimolar) GTP concentrations were reported to block the K ϩ channel activity on coatomer (34) but had little effect on its InsP n binding. The results herein reflect on a similar behavior for the interaction of PtdInsP n s with ␣-COP. In addition, the inability of BFA or GTP to interfere with the PtdInsP n -␣-COP interaction suggests that separate, non-allosterically regulated binding sites are involved. Thus, the PtdIns(3,4,5)P 3 -␣-COP interaction appears to be independent of ARF binding and the coatomer recruitment process.
The inability of the chromatographically isolated B complex of ␣-COP, ␤Ј-COP, and ⑀-COP complex to bind PtdIns(3,4,5)P 3 suggests that the PtdIns(3,4,5)P 3 -␣-COP binding may involve a more complex set of protein-protein interactions. Thus, conformational changes due to subunit interactions may be required to permit PtdInsP n binding to ␣-COP. Alternatively, the observed failure of the B complex to undergo photoaffinity labeling could be an artifact of a non-reversible effect resulting from the buffer conditions required for subunit dissociation. However, the physiological significance of this dissociated B complex is not clear, despite reports of its binding to membranes (33).
In conclusion, the data presented offer the first evidence for a specific interaction of one, and only one, polypeptide subunit of Golgi coatomer, ␣-COP, with the polyphosphoinositide PtdIns(3,4,5)P 3 . Moreover, these data also demonstrate the specificity of interactions of the soluble inositol polyphosphates Ins(1,3,4,5)P 4 and InsP 6 with individual coatomer polypeptides. This result offers a new perspective on the potential role of PI 3-kinase in non-clathrin-mediated Golgi membrane traffic. Moreover, while the 3-phosphate on the inositol ring plays a critical role in defining this interaction, the 5-phosphate is also required for maximal binding activity. In addition, the importance of a hydrophobic group, either the aminopropyl BZDC group or the diacylglycerol moiety, suggests that the Ptd(3,4,5)InsP 3 -␣-COP interaction might be important for the recruitment (or its inhibition by conformational change) of coatomer to membranes during the budding and/or coating process. To date, ␣-COP has not been implicated as an active participant in COPI recruitment and vesicle coating; ␤-COP has been more fully examined for its importance for coat assembly and in the budding reaction (27). Our results suggest that a re-examination of protein-polyphosphoinositide, as well as protein-protein interactions, will further illuminate the complex process of vesicular trafficking. Finally, the results reported herein represent an application of a new class of Ptd (3,4,5)InsP n photoaffinity probes that sample the interface between the charged phosphoinositide head group and the lipid bilayer. Additional examples of the uses of these [ 3 H]BZDCtriester-PtdInsP n probes for characterization of other protein targets will be presented in due course.