Identification and characterization of a novel Golgi protein, GCP60, that interacts with the integral membrane protein giantin.

We demonstrated previously that the integral membrane protein giantin has the Golgi localization signal at the COOH-terminal cytoplasmic domain (Misumi, Y., Sohda, M., Tashiro, A., Sato, H., and Ikehara, Y. (2001) J. Biol. Chem. 276, 6867-6873). In the present study, using this domain as bait in the yeast two-hybrid screening system, we identified a novel protein interacting with giantin. The 3.6-kilobase mRNA encoding a 528-amino acid protein of 60 kDa designated GCP60 was ubiquitously expressed and was especially abundant in the testis and ovary. Immunofluorescence and immunoelectron microscopy confirmed that GCP60 was co-localized with giantin in the Golgi complex. GCP60 was found to be a peripheral protein associated with the Golgi membrane, where a COOH-terminal domain of GCP60 interacts with the COOH-terminal cytoplasmic domain of giantin. Overexpression of the COOH-terminal domain of GCP60 caused disassembly of the Golgi structure and blocked protein transport from the endoplasmic reticulum to the Golgi. Taken together, these results suggest that GCP60 is involved in the maintenance of the Golgi structure by interacting with giantin, affecting protein transport between the endoplasmic reticulum and the Golgi.

The Golgi complex is comprised of structurally distinct subcompartments and plays a key role in sorting and modification of proteins exported from the endoplasmic reticulum (ER) 1 (1)(2)(3). Protein transport from the ER to the Golgi and through the Golgi compartments is mediated by small vesicles and requires a number of soluble and membrane proteins (1,4,5). These include coat protein complex (COPI and COPII) and the GTPases ARF and Sar1p for vesicle budding, N-ethylmaleim-ide-sensitive fusion protein, soluble N-ethylmaleimide-sensitive fusion protein attachment protein (SNAP), and SNAP receptors for vesicle docking and fusion. In addition, p115 is known to function in tethering of vesicles with membranes before their complete docking (6,7). Available evidence indicates that the vesicular transport system functions not only in the delivery of proteins to their final destinations but also in maintenance of the Golgi structure. Brefeldin A blocks the budding and formation of transport vesicles by inhibiting a GDP/GTP exchange of ARF (8,9). The inhibition of vesicular transport causes redistribution of Golgi components to the ER, resulting in disassembly of the Golgi stack (10 -13). The Golgi structure is reformed immediately after removal of the drug. These observations suggest that the Golgi structure is maintained by vesicular transport-dependent supply of the resident Golgi components and their balanced recycling between the ER and Golgi.
Despite these findings, the mechanism for vesicular transport alone may not be sufficient to explain the formation of the unique Golgi structure with the ordered stacking, close apposition, and constant spacing of stacked cisternae in mammalian cells. There must be other components that would be responsible for the formation or structural arrangement of the subcompartments in the Golgi complex. Attempts on this line, mostly using antibodies from patients with autoimmune diseases, have identified an increasing number of new Golgiassociated proteins. These proteins, named the golgin family, have a long coiled-coil domain and include cytoplasmic peripheral proteins such as golgin-245/p230 (14,15), GCP170/golgin-160 (16,17), GM130/golgin-95 (16,18), and golgin-97 (19). Also included in the family are integral membrane proteins such as golgin-67 (20), golgin-84 (21), and giantin/GCP372 (22,23). Although the functions of these Golgi proteins remain largely unknown, GM130 was found to be involved in tethering of transport vesicles to the Golgi membrane by interacting with p115 (6,7). In addition, recent studies have shown that giantin is also present in COPI vesicles and interacts with p115, contributing to the formation of a tethering bridge before vesicle docking (6,24,25).
Giantin contains a large NH 2 -terminal cytoplasmic domain (Ͼ370 kDa) and a COOH-terminal membrane-anchoring domain without a luminal extension (22,23,26). The newly synthesized protein is initially inserted into the ER membrane and then transported to the Golgi (27). Our previous study has demonstrated that giantin contains the Golgi localization signal at the COOH-terminal cytoplasmic domain of about 100 residues adjacent to the membrane anchor domain (28). This suggests that giantin is localized to the Golgi membrane by a protein-protein interaction rather than by the protein-lipid interaction that is proposed for other Golgi membrane proteins * This work was supported in part by grants from the Ministry of Education, Science, Sports and Culture of Japan, the Japan Science and Technology Corporation (CREST), and the Central Research Institute of Fukuoka University. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the DDBJ/GenBank TM /EBI Data Bank with accession number(s) AB043587.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screening, cDNA Cloning, and Northern Blotting-The COOH-terminal cytoplasmic domain (positions 2804 -3162) of rat giantin (26,28) was fused to the LexA DNA-binding domain and used as the bait construct. EGY48 harboring the ␤-galactosidase reporter plasmid was transformed with the bait plasmid and the HeLa cell cDNA library generated in pB42AD vector (29) (CLONTECH). Screening was performed according to the manufacturer's instructions. ␤-Galactosidase activities expressed by the specific interaction between the bait and prey constructs were measured using o-nitrophenyl-galactoside as a substrate (30). Of about 10 6 clones screened, 1 positive cDNA clone (2.6 kilobase pairs) was obtained and designated B8, which was found to contain an incomplete open reading frame. A cDNA clone (3.6 kilobase pairs) containing a complete open reading frame (GCP60) was isolated from the ZAPII cDNA library of HeLa cells by the plaque hybridization method (17) using 32 P-labeled B8 cDNA as a probe. Northern blotting was carried out using a human multiple tissue Northern blot (CLONTECH) containing poly(A) ϩ RNA (2 g/lane), which was hybridized with 32 P-labeled B8 cDNA or human ␤-actin cDNA as a probe (17).
Preparation of Anti-GCP60 Antibodies-A cDNA encoding a chimeric protein of B8 fused to the COOH terminus of glutathione S-transferase (GST) was constructed in expression vector pGEX4T-1 (17) and designated GST-B8. The recombinant protein was expressed in and purified from bacteria and injected into rabbits to raise anti-GST-B8 antibodies as described previously (17). The anti-GST antibody was removed from the antiserum by passing it through GST-coupled Sepharose 4B beads, and the remaining antibody was used as anti-GCP60, unless otherwise indicated. In some experiments, we used rabbit anti-GCP60N, which was raised against the NH 2 -terminal region (positions 1-175) of GCP60 and prepared by essentially the same procedures as described above.
Construction and Transfection of Expression Plasmids-cDNAs encoding the entire coding region of GCP60 (wild type) and B8 were ligated to a position downstream of the sequence encoding the Met-FLAG tag in pSG5 expression vector so that the products could be recognized by the monoclonal anti-FLAG antibody M2 (28,31). The COOH-terminally deleted mutant M1 was prepared by introducing a termination codon into an appropriate site of the wild type plasmid. The NH 2 -terminally deleted mutants D1, D2, and D3 were obtained by amplification of B8 cDNA by polymerase chain reaction with a set of 25-mer primers synthesized for the indicated positions shown in Fig.  7A. The mutant cDNA sequences (D1ϪD3) obtained were verified and ligated in frame to pEGFP-1 for GFP-fused proteins. The plasmid for N-acetylglucosaminyltransferase fused with GFP (NAGFP) (32) was kindly donated by Dr. D. T. Shima (Imperial Cancer Research Fund, London, United Kingdom). The constructs in the pSG5 expression vector were transfected into COS-1 cells or HeLa cells with the FuGENE6 transfection reagent (Roche Diagnostics Corp., Indianapolis, IN). All constructs were also ligated to pB42AD for the yeast two-hybrid assay.
Cell Fractionation and Immunoblotting-A postnuclear supernatant fraction was prepared from HeLa cells or transfected COS-1 cells and centrifuged at 105,000 ϫ g for 1 h into membrane and cytosol fractions as described previously (31). A Golgi-enriched fraction was prepared by flotation of the postnuclear supernatant fraction in a sucrose density gradient as described previously (28,31). Each sample was analyzed by SDS-PAGE (5% or 10% gels) followed by immunoblotting with the indicated primary and secondary antibodies (at a 1:1000 dilution for each antibody). The immunoreactive proteins were visualized using the ECL kit (17).
Overlay Assay-A Golgi-enriched fraction and immunoprecipitates with the indicated antibody were separated by SDS-PAGE (5% gels) and transferred to an Immobilon membrane (Millipore Corp., Bedford, MA). The membrane was incubated with GST or GST-B8 (2 mg/ml) at 25°C for 1 h, then incubated with goat anti-GST antibody (1:1000) (Amersham Pharmacia Biotech) for 1 h, and finally incubated with horseradish peroxidase-conjugated anti-goat IgG (1:1000) (Santa Cruz Biotechnology Inc., Santa Cruz, CA). The immunocomplex was visualized with the ECL kit.
Analysis of Protein Transport-Protein transport from the ER to the Golgi was examined using the temperature-sensitive glycoprotein of vesicular stomatitis virus (VSV-G/ts045). COS-1 cells were transfected with the pSG5 plasmid containing the cDNA construct of VSV-G/ts045 fused to green fluorescent protein (GFP-VSV-G/ts045) (34). After incubation at 39.5°C for 12 h, cells were transferred to 32°C and incubated for an additional 2 h. The cells were fixed and observed for the GFP image. When indicated, FLAG-tagged GCP60 constructs (wild type, B8, and D3) were transfected into COS-1 cells together with the GFP-VSV-G/ts045 plasmid. Cells were fixed, stained with anti-FLAG antibody, and observed for each construct of GCP60 and the GFP image of VSV-G/ts045.
Yeast Two-hybrid Assay-The assay was performed according to the method of Golemis et al. (29). EGY48 harboring the ␤-galactosidase reporter plasmid was transformed with pLexA/giantin 2804 -3161 and pB42AD/GCP60 or its deletion mutants. Transformants were plated on agar plates with synthetic media containing dextrose lacking Ura, His, and Trp and incubated for 2 days at 30°C. Surviving transformants were streaked on a plate containing galactose and 5-bromo-4-chloro-3indolyl-␤-D-galactopyranoside without Ura, His, Trp, and Leu and incubated for 3 days at 30°C.

Isolation of Clone B8 Enocoding a Giantin-interacting Pro-
tein Fragment-We demonstrated previously that giantin has a Golgi localization signal in the COOH-terminal region (positions 2804 -3162), which contains three coiled-coil domains, I, II, and III (Fig. 1A, top) (28). The entire region containing the three domains was inserted into the LexA fusion vector and used as bait in the yeast two-hybrid screening system for identification of a protein that interacts with giantin. We screened the HeLa cell cDNA library in the pB42AD fusion vector and obtained a 2.6-kilobase pair cDNA clone, B8, which was not a full-length cDNA but encoded a partial sequence with 211 amino acids (B8p). The interaction of B8p with each domain of giantin was analyzed by the two-hybrid system, demonstrating that B8p specifically interacts with domain III of giantin, although not as efficiently as it does with the entire COOHterminal region (Fig. 1A).
We further examined the in vitro interaction of B8 with giantin by an overlay assay. A Golgi fraction and immunoprecipitates obtained with the indicated antibodies were separated by SDS-PAGE and overlaid with GST or GST-B8p. The bound GST-B8p was detected as a single band corresponding to giantin in the Golgi fraction (Fig. 1B, lanes 2 and 5) and as two bands in the giantin immunoprecipitate (lanes 3 and 6). The faster-migrating band of the two bands (lane 3) appears to be a degradation product containing the COOH-terminal region of giantin, whereas another major band (about 180 kDa) detected on the immunoblot (lane 6) may be the NH 2 -terminal counterpart with which GST-B8p does not interact. GST-B8p, however, was found not to interact with GM130 (lanes 4 and 7), a Golgi matrix protein (18). These results indicate that the B8 product specifically interacts with giantin both in vivo and in vitro.
Cloning and Structural Characteristics of GCP60 -We obtained a 3.6-kilobase pair cDNA clone from the HeLa cell library by the plaque hybridization method using B8 cDNA as a probe. The cDNA contained a complete open reading frame encoding a protein of 528 amino acid residues with a calculated mass of 60 kDa ( Fig. 2A) that we designated GCP60 (Golgi complex-associated protein of 60 kDa). The B8p sequence was found in the COOH-terminal half of GCP60 starting with the Pro 328 residue. GCP60 contains the following characteristic domains ( Fig. 2A and Drosophila melanogaster (40.6% identity) (Fig. 2B), although these homologues have not been identified at the protein level. No related proteins were found in Saccharomyces cerevisiae and plants. Fig. 2C shows an alignment of the ACB domain sequence of GCP60 with the corresponding sequences of acyl-CoA-binding protein and ⌬ 3 -cis-⌬ 2 -trans-enoyl-CoA isomerase (37), demonstrating that the GCP60 domain has a 33-43% identity with these proteins. Hydropathy analysis predicts that GCP60 does not contain a hydrophobic domain that could participate in membrane localization (data not shown). A long hydrophilic region corresponding to the CAR domain (positions 182-240) is also predicted to have a high probability of coiled-coil formation (38). The highly charged coiled-coil structure is thought to form a homo-and/or hetero-oligomerization of proteins stabilized by ionic interaction (39).
Expression of B8 and GCP60 -For expression of B8 and GCP60, a GFP-fused B8 cDNA (GFP-B8) and a FLAG-tagged GCP60 cDNA were prepared (Fig. 3A) and transfected into COS-1 cells. The GFP-B8 chimera was detected by the GFP image, whereas GCP60 was detected by immunofluorescence staining with anti-FLAG antibody (Fig. 3B). Both the expressed proteins were localized to a juxtanuclear region corresponding to the Golgi complex, where giantin was co-localized as detected with anti-giantin. Cells expressing FLAG-tagged GCP60 were labeled with [ 35 S]methionine and subjected to immunoprecipitation with anti-FLAG antibody. GCP60 expressed in the transfected cells was heavily labeled and had the same molecular mass (65 kDa) as the endogenous protein from control cells (Fig. 3C, lanes 1 and 2). GCP60 was found to be phosphorylated because 32 P i was incorporated into the protein (data not shown). In addition, when the 35 S-labeled protein was treated with alkaline phosphatase, its molecular mass was slightly reduced (Fig. 3C, lane 3). Thus, the apparent molecular mass of GCP60, which is slightly higher than the calculated mass (60 kDa), may be due in part to its phosphorylation.
Ubiquitous Expression of GCP60 -The GCP60 mRNA level was examined by Northern blotting of poly(A) ϩ RNA from various human tissues with 32 P-labeled B8 cDNA (Fig. 4A). A single 3.7-kilobase message was detected in all the tissues examined and was especially abundant in testis and ovary, indicating that GCP60 is ubiquitously expressed (Fig. 4A). The ubiquitous expression of GCP60 was also examined in cultured cells from different species by immunofluorescence microscopy and immunoblotting, for which polyclonal antibodies raised against recombinant human GCP60 were used. The antibodies were found to cross-react well with the antigen in cells derived from other species including monkey COS-1 cells, Chinese hamster ovary CHO-K1 cells, and rat NRK cells, confirming its localization in the juxtanuclear region (Fig. 4B) and a single band in SDS-PAGE (Fig. 4C).
Localization of GCP60 -Immunofluorescence microscopy confirmed that the endogenous GCP60 in HeLa cells was colocalized with giantin to the juxtanuclear region (Fig. 5A). The intracellular localization of GCP60 was examined in more detail by immunoelectron microscopy. As shown in Fig. 5B, immunoreactive gold particles were detected on the characteristic stack and related structures of the Golgi complex. The subcel-  (28). The initial clone B8 was obtained by yeast two-hybrid screening for which the entire sequence containing domains I-III (2804 -3162) was used as the bait construct. Yeast two-hybrid analysis was performed with four pLexA constructs containing the indicated domain of giantin (used as the bait) and pB42AD-B8 (as the prey). ␤-Galactosidase activity obtained with pLexA/I-III (2804 -3162) is taken as 1, and relative values of enzyme activity with the other bait constructs are shown as the means Ϯ S.D. (n ϭ 4). B, in vitro interaction of B8 with giantin. A Golgi fraction (lanes 1, 2, and 5) and immunoprecipitates with anti-giantin (lanes 3 and 6) or anti-GM130 antibody (lanes 4 and 7) prepared from HeLa cells were separated by SDS-PAGE (5% gels) and transferred onto the Immobilon membrane. For the overlay assay, the membrane was incubated with GST (lane 1) or GST-B8 (lanes 2-4) and then with incubated with anti-GST antibody. The membrane was also subjected to immunoblotting with anti-giantin (lanes 5 and 6) or anti-GM130 antibody (lane 7). The interacted protein complexes were visualized as described under "Experimental Procedures." lular distribution of GCP60 was also examined by cell fractionation and immunoblotting (Fig. 5C). When a postnuclear supernatant fraction was separated into membranes and cytosol, 30 -40% of the total GCP60 was recovered in the membrane fraction. The majority of the membrane-associated GCP60 was recovered in the Golgi-enriched fraction. When the Golgi membranes were treated with 1 M KCl and separated into membranes and supernatant by centrifugation at 105,000 ϫ g for 1 h, GCP60 was completely dissociated from the membranes into the soluble fraction. In addition, membrane-associated GCP60 was found to be completely digested with a mixture of trypsin and chymostrypsin in the absence of a detergent (data not shown). The results indicate that GCP60, although primarily present as a soluble protein in the cytosol, is associated with the cytoplasmic face of Golgi membranes.
Effect of Drugs on the Localization of GCP60 -We then examined the effect of Golgi-perturbing drugs on the localization of GCP60. HeLa cells were treated with nocodazole for complete depolymerization of microtubules and concomitant Golgi fragmentation, under which GCP60 was detected on punctate structures scattered in the cytoplasm and remained co-localized with giantin (Fig. 6A). When cells were incubated with brefeldin A, GCP60 localized in the Golgi was diffusely dispersed into the cytoplasm, as already detected 5 min after the treatment (Fig. 6B, c). This was in contrast to the behavior of giantin, the majority of which was still retained in the Golgi region at 5 min after treatment (Fig. 6B, d). After 30 min, GCP60 was found to be concentrated in the nucleus in addition to the diffuse cytoplasmic distribution, a pattern that clearly differed from the ER-like staining pattern of giantin (Fig. 6B, e and f). The response of GCP60 to brefeldin A was not as rapid as that of COPI components such as ␤-COP (9) but was much faster than that of Golgi integral membrane proteins including giantin. Upon removal of the drug, GCP60 was relocalized to the Golgi with a recovery rate similar to that of giantin (data not shown).
The Domain of GCP60 Essential for Its Golgi Localization-To determine which domain of GCP60 is required for Golgi localization, we constructed expression plasmids encoding various deletion mutants of GCP60 (Fig. 7A). The FLAGtagged or GFP-fused mutants were expressed in COS-1 cells by transfection, and their intracellular localization was examined by fluorescence microscopy (Fig. 7B). Mutant M1, a truncated GCP60 that lacks the B8 sequence, was distributed diffusely throughout the entire cell including the nucleus, confirming that the NH 2 -terminal half containing the characteristic domains is not involved in the Golgi localization of GCP60. In contrast, mutant D1 lacking the NH 2 -terminal 45 residues of the B8 sequence was detected at the perinuclear region and colocalized with endogenous GCP60. However, mutant D2 (derived from D1 with a deletion of an additional 20 NH 2 -terminal residues) was no longer localized to the Golgi region. In addition, mutant D3 derived from B8 by deleting the COOH-terminal 10 residues failed to localize to the Golgi. Taken together, these results indicate that the amino acid sequence between residues 373 and 528 is required for the Golgi localization of GCP60. We further examined the interaction of these mutants with giantin by the yeast two-hybrid assay, confirming that only mutants B8 and D1 specifically interacted with giantin (Fig. 7C). Thus, it is likely that the giantin-binding domain functions as the Golgi localization domain.
Involvement of GCP60 in Structural Maintenance of the Golgi-To address the function of GCP60, we transfected FLAG-tagged deletion constructs of GCP60 with (B8) or without (M1) the giantin-binding domain into COS-1 cells (Fig. 8A) and examined the effect of their overexpression on the Golgi structure (Fig. 8B). When wild type GCP60 and mutants B8 and M1 were expressed at a moderate level, no significant change was observed for the Golgi structure, as monitored by localization of the endogenous giantin (Fig. 8B, aϪf). In addition, overexpression of M1 lacking the giantin-binding domain had no significant effect on the Golgi structure (Fig. 8B, i and  j). In contrast, a drastic change in the Golgi structure was caused by overexpression of wild type GCP60 and B8, as revealed by a diffuse cytoplasmic distribution of endogenous giantin (Fig. 8B, g, h, k, and l). The drastic change in the Golgi structure was further examined by the distribution of another Golgi marker protein, for which NAGFP was co-expressed. The NAGFP expressed was localized to the Golgi in the presence of M1 (Fig. 8B, o and p), whereas it was diffusely distributed throughout the cytoplasm by overexpression of the wild type (panels m and n) and B8 (panels q and r). Such a dominant negative effect of the wild type and B8 suggests that GCP60 is involved in the structural maintenance of the Golgi by direct interaction with giantin.
Involvement of GCP60 in Protein Transport to the Golgi-We then examined the effect of overexpression of wild type and mutant GCP60 on protein transport from the ER to the Golgi, using a temperature-sensitive and GFP-tagged VSV-G protein (GFP-VSV-G/ts045). When COS-1 cells were transfected with the plasmid expressing GFP-VSV-G/ts045 and cultured at 39.5°C for 12 h, the GFP-tagged G protein remained in the ER. A temperature shift to 32°C allowed the protein to be concentrated predominantly in the Golgi region, although a small amount of the protein reached the cell surface under these conditions (Fig. 9A). In the presence of overexpressed wild type GCP60 and mutant B8, however, the GFP-tagged G protein retained in the ER at the nonpermissive temperature was not transported to the Golgi and cell surface even at the permissive temperature of 32°C (Fig. 9B, cϪf). In contrast, overexpressed mutant M1 exerted no inhibitory effect on the temperature-dependent transport of the GFP-tagged G protein to the Golgi and cell surface (Fig. 9B, g and h). These results suggest that GCP60 is also involved in protein transport from the ER to the Golgi. Thus, it is likely that the giantin-binding domain is essential for GCP60 to exert its function on protein transport as well as on maintenance of the Golgi structure.

DISCUSSION
In this study, using the yeast two-hybrid screening system with the Golgi localization domain of giantin as bait, we identified the novel 60-kDa protein GCP60 that is localized to the Golgi complex. Data base searches show that GCP60 has a significant similarity in amino acid sequence to unannotated gene products of C. elegans (GenBank TM accession number Z9559.2) and D. melanogaster (GenBank TM accession number AFF49009.1), with an identity of 47.5% and 40.6%, respectively. However, no homologous sequence is found in yeast and plants, indicating that GCP60 is absent in yeast and plants. This is unusual because most vesicular transport factors are evolutionarily conserved (40). It is possible that there is a functional homologue in yeast that has diverged in sequence. However, considering that the yeast counterpart of giantin also has not been identified, it is likely that the giantin-GCP60 complex is part of a system that might represent one of the additional levels of complexity intrinsic to higher eukaryotic cells.
GCP60 is predicted to contain a coiled-coil domain but not a hydrophobic domain that can participate in membrane localization, indicating that the protein is primarily a soluble cytoplasmic protein. There are many other coiled-coil proteins that are peripherally associated with the cytoplasmic face of the Golgi membrane, including GM130/golgin-95 (16, 18), golgin-97 (19), golgin-160/GCP170 (16, 17), and golgin-245/p230 (14,15). Golgin-97, golgin-245/p230, and several other proteins have been found to contain a conserved COOH-terminal sequence (of about 50 amino acids) that is designated the GRIP domain (41), the Rab6-interacting domain (42), or the Golgi localization domain (43). The domain contains a consensus sequence including a tyrosine-based motif, which functions in targeting these proteins to the Golgi, possibly by interacting with Rab6 on the Golgi membrane (42). The consensus GRIP domain, however, is not found in the corresponding COOHterminal sequence of GCP60. Instead, all the data presented here support the conclusion that GCP60 is localized to the Golgi by interaction with giantin, for which the COOH-terminal sequence (positions 373-528) of GCP60 serves as the giantin-binding domain. In fact, the giantin-binding domain contains two regions (positions 367-459 and 477-528) that are highly conserved between human GCP60 and its homologues from C. elegans and D. melanogaster.
Giantin was originally identified as a Golgi membrane protein anchored at the COOH terminus, with most of its mass projecting into the cytoplasm (22,23,26,44). In addition to residing on Golgi membranes, giantin is also incorporated into budding COPI vesicles and implicated to function as a p115 receptor for tethering the vesicles to Golgi membranes (6,24). A tethering model proposed by Warren and colleagues (6,24) hypothesizes that giantin in the COPI vesicles is cross-bridged by p115 to GM130 (45), which is tightly associated with GRASP65 on Golgi membranes (46). Another model suggests that giantin and GM130 are not cross-bridged by p115 but that each protein interacting with p115 functions at distinct steps of transport (47). However, despite this discrepancy, it is agreed that p115 interacts with the NH 2 -terminal domain of giantin (6,47). In addition, it has been confirmed that the corresponding NH 2 -terminal peptide blocks cell-free Golgi reassembly (6) or inhibits protein transport from the ER to Golgi (47). Taken together, these findings indicate that giantin is indeed involved in protein transport between the ER and Golgi and in the formation of the Golgi stack structure. The Golgi structure with stacked cisternae is a characteristic feature in mammalian cells that is clearly different from the yeast Golgi, which are seen as single, isolated cisternae generally not arranged into parallel stacks (48). Such a difference in the Golgi structure suggests the presence of unique factors in mammalian cells, but not in yeast, for which giantin and GCP60 may be included as candidates.
As opposed to p115, GCP60 binds to the COOH-terminal cytoplasmic domain III of giantin, which is essential for Golgi localization and contains a coiled-coil domain (28). The giantinbinding domain of GCP60 maps to the COOH-terminal sequence (positions 373-528) that does not contain a coiled-coil domain, indicating that the interaction between giantin and GCP60 is probably not due to the coiled-coil structure. This is FIG. 7. Essential domain of GCP60 for its Golgi localization. A, schematic representation of wild type (WT) and deleted mutants (M1, B8, and D1ϪD3) of GCP60. Characteristic domains of GCP60 are also shown at the top: proline-rich (PR), ACB region, CAR domain, and glutamine-rich domain (QR). B, FLAGtagged (WT, M1, and B8) or GFP-fused (D1ϪD3) proteins were expressed in COS-1 cells by transfection with the indicated plasmids. After cells were fixed, the expressed proteins were detected by immunostaining with anti-FLAG antibody (top row) or directly by GFP image (middle row). The same cells shown in the middle row were also stained for the endogenous GCP60 with the anti-NH 2 -terminal domain of GCP60 (GCP60N) in combination with rhodamine-conjugated anti-rabbit IgG (bottom row). C, yeast two-hybrid analysis. EGY48 harboring the ␤-galactosidase reporter plasmid was transformed with pLexA/giantin 2804 -3162 and pB42AD/GCP60 (WT) or the indicated mutants. The surviving transformants were stained for ␤-galactosidase activity expressed by the specific interaction between giantin 2804 -3162 and the indicated GCP60 mutants. in contrast to a relationship proposed for syntaxin-6 and its interacting protein , FIG (49, 50). Syntaxin-6 has a cytoplasmic H2 domain adjacent to the membrane anchor that is predicted to form an ␣-helical structure and play a major role in Golgi localization (51). The FIG protein interacts with syntaxin-6 for its Golgi retention, for which the COOH-terminal coiled-coil domain of the two domains contained in FIG is required (49).
When construct B8 or D1 encoding the giantin-binding domain of GCP60 was overexpressed, the Golgi proteins giantin and NAGFP were dispersed throughout the cell, and transport of VSV-G from the ER to the Golgi was remarkably inhibited (Figs. 8 and 9). These results suggest that giantin with the Golgi-targeting domain occupied by the deletion mutant B8p is no longer localized to the Golgi, and this delocalization results in the disruption of the Golgi structure and the blockade of protein transport between the ER and the Golgi. Although overexpression of the NH 2 -terminal half (M1) alone had no significant effect on Golgi structure and transport, it is likely that some part of the NH 2 -terminal sequence is also required for GCP60 to form a functional complex with giantin in vivo.
The COOH-terminal half sequence of GCP60 contains several characteristic domains including a proline-rich domain and an ACB domain. The proline-rich domain contains the consensus sequence Pro-Pro-Leu-Pro for binding with the WW domain (52), which is found in a number of cytoskeletal, regulatory, and signaling molecules. These include dystrophin, a cytoskeletal protein that stabilizes membranes and generates contractile force (53), and IQGAP1, a GTPase-mediated cytoskeleton-regulating Golgi protein (54). In addition, proteinprotein interactions via the proline-rich motif were reported in a vesicular transport system with dynamin (55) and in cytoskeletal architecture with ␤III spectrin (56). The second characteristic domain is an ACB domain that has a similar amino acid sequence to the lipid-binding domain of cytoplasmic acyl-CoA-binding proteins (35,37). In a cell-free vesicular transport assay, acyl-CoA esters including palmitoyl-CoA were shown to be essential for either the budding or the fusion process (57,58). However, a direct target molecule for acyl-CoA involved in the process has not been identified. Although two essential tyrosine residues in the lipid-binding motif (35) are substituted with valine and phenylalanine in the ACB domain of GCP60, it is of interest to examine whether GCP60 is able to bind acyl-  d, f, h, j, and l). The top two rows (a-f) show cells with moderate expression of each construct, whereas the middle two rows (g-l) show cells with overexpression of the introduced proteins. In the bottom two rows (m-r), cells with overexpression of each construct of GCP60 were immunostained with anti-FLAG (m, o, and q) or directly observed by the GFP image for NAGFP (n, p, and r) that was coexpressed in each cell by transfection of the NAGFP plasmid.

FIG. 9. Effect of overexpression of wild type and mutant GCP60 on transport of VSV-G protein.
A, the chimera GFP-VSV-G/ts045 was expressed in COS-1 cells, which were incubated at 39.5°C for 12 h (a) and then incubated at 32°C for 2 h (b). The expressed protein was detected by the GFP image. B, GFP-VSV-G/ts045 and FLAG-tagged wild type (WT) or truncated mutants (B8 and D3) of GCP60 were co-expressed in COS-1 cells. The cells were incubated at 39.5°C for 12 h and then incubated for 2 h at 39.5°C (a and b) or 32°C (cϪh). Cells were fixed and immunostained with anti-FLAG antibody and TRITC-conjugated secondary antibody (b, d, f, and h). GFP-VSV-G/ts045 co-expressed in each cell was detected by the GFP image (a, c, e, and g). Note that the indicated GCP60 proteins were overexpressed in each cell.
CoA, and whether acyl-CoA is required for the formation of the functional complex between GCP60 and giantin.
Previous observations highlighted the important role of the NH 2 -terminal domain of giantin in the tethering of vesicles to the Golgi membrane, involving other factors such as p115, GM130, and GRASP65 (6,24). In the present study, together with a previous one (28), we have shown that the COOHterminal cytoplasmic domain of giantin also plays an essential role in its localization to the Golgi by interacting with GCP60. Further functional analysis of GCP60 will shed light on the mechanism by which giantin is involved in protein transport between the ER and Golgi and the maintenance of the Golgi structure.