Identification and Characterization of Golgin-84, a Novel Golgi Integral Membrane Protein with a Cytoplasmic Coiled-coil Domain*

The cytoplasmic face of the Golgi contains a variety of proteins with coiled-coil domains. We identified one such protein in a yeast two-hybrid screen, using as bait the peripheral Golgi phosphatidylinositol(4,5)P2 5-phosphatase OCRL1 that is implicated in a human disease, the oculocerebrorenal syndrome. The ∼2.8-kilobase mRNA is ubiquitously expressed and abundant in testis; it encodes a 731-amino acid protein with a predicted mass of 83 kDa. Antibodies against the sequence detect a novel ∼84-kDa Golgi protein we termed golgin-84. Golgin-84 is an integral membrane protein with a single transmembrane domain close to its C terminus. In vitro, the protein inserts post-translationally into microsomal membranes with an N-cytoplasmic and C-lumen orientation. Cross-linking indicates that golgin-84 forms dimers, consistent with the prediction of an ∼400-residue dimerizing coiled-coil domain in its N terminus. The dimerization potential is supported by a data base search that showed that the N-terminal 497 residues of golgin-84 contain a coiled-coil domain that when fused to the RET tyrosine kinase domain had the ability to activate it, forming the RET-IIoncogene. Data base searching also indicates golgin-84 is similar in structure and sequence to giantin, a membrane protein that tethers coatamer complex I vesicles to the Golgi.

The Golgi cisternae constitute a compartmentalized and polarized structure, allowing post-translational processing enzymes to act in succession. As a routing center, the Golgi receives proteins from the endoplasmic reticulum and routes them in bulk to the plasma membrane or specifically to the endosomes and lysosomes. In polarized cells, additional transport routes include targeting to the apical or basolateral plasma membranes or to storage granules (1,2). Protein sublocalization, however, is not precise, since the locations of many of the proteins overlap over two or more cisternae (3). All known resident Golgi proteins are peripheral or integral membrane proteins.
Coiled-coil-containing proteins with different functions (4) have been identified on the cytoplasmic face of the Golgi. Cterminally anchored membrane proteins, with cytoplasmic coiled-coil domains, bind transport vesicles to the cisternae membranes (5). One type tethers the vesicles to membranes, while another type, termed SNARES, ensure that specific vesicles dock to the membranes. Peripheral coiled-coil proteins on the Golgi surface are of the resident and transient types; they maintain Golgi structure and participate in budding, transport, tethering, and docking of vesicles. Additional coiled-coil proteins of unknown function have been identified as antigens in autoimmune diseases (6) and include golgins 95, 160, and 245. One autoantigen, giantin-376, (macrogolgin) has recently been shown to tether COPI 1 vesicles to the Golgi, thereby facilitating vesicle transport within the Golgi stack (7). Here we report the cloning and characterization of another, novel cytoplasmicassociated coiled-coil integral membrane Golgi protein termed golgin-84.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen-The yeast two-hybrid screen was performed on a human brain cDNA library (CLONTECH) using a Match-Maker kit (CLONTECH) and a ϳ4.3-kb human OCRL1 cDNA (full open reading frame) as bait. The human OCRL1 cDNA was subcloned into vector pGBT9 (GAL4 DNA binding domain) and the brain cDNA library into vector pGAD424 (GAL4 activating domain). The integrity of the OCRL1 construct was verified by in vitro transcription and translation. Approximately 1.5 million brain clones were screened. Screening and elimination of false positives were performed as described by the manufacturer.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF085199.
Sequence Analysis-Protein and cDNA sequences were analyzed by Lasergene (Macintosh) and PCGene (IBM). 2 Northern Blot Analysis-A mouse multiple tissue Northern blot (CLONTECH) containing 2 g of poly(A) ϩ RNA/lane was probed with radiolabeled 2.6-kb human golgin-84 cDNA (specific activity, ϳ5.7 ϫ 10 8 cpm/g). The blot was prehybridized with ExpressHyb TM hybridization solution (CLONTECH) for 1 h at 68°C and then hybridized in it for 2 h at 68°C. After hybridization, two 15-min room temperature washes of 2ϫ SSC, 0.05% SDS were performed followed by two 30-min washes at 60°C. The blot was then exposed for 4 days. After exposure, the blot was stripped, prehybridized, and hybridized with a ␤-actin probe (specific activity, ϳ4.45 ϫ 10 8 cpm/g) as described for golgin-84. Washing conditions were the same as for golgin-84, although an additional high stringency wash of 0.1ϫ SSC, 0.1% SDS was performed at 60°C for 1 h.
Antibody Production-Antibodies were raised in rabbit against denatured and solid phase renatured His 6 tag protein containing golgin-84 amino acids 499 -687. A 500-ml E. coli culture containing the golgin-84 pET construct (Novagen) was induced at A 595 ϳ0.5 with 2 mM isopropyl ␤-D-thiogalactopyranoside (final concentration) for ϳ4 -5 h. Cells were harvested by centrifugation and lysed overnight at 4°C in 6 M guanidine hydrochloride, 50 mM NaH 2 PO 4 , pH 8.0, and the lysate was cleared by centrifugation at 8000 rpm. The cleared supernatant was passed over a Ni 2ϩ resin column (pretreated as described by the manufacturer, Qiagen), and the column was then washed with ϳ25 bed volumes of lysis buffer. Protein was eluted from the column in a denatured form by the addition of 6 M guanidine hydrochloride, 50 mM NaH 2 PO 4 in aliquots of decreasing pH: pH 6.0 (2 ϫ 5 ml), pH 5.0 (3 ϫ 3 ml), and pH 3.5 (1 ϫ 10 ml). The denatured protein was then subjected to SDS-PAGE (nine gels, ϳ1 mg of antigen/gel), cut from the gels, and used for immunization. Solid phase renatured protein was also prepared on the column by a stepwise renaturation procedure; 25 bed volumes of MCAC (20 mM Tris⅐Cl, pH 7.9, 0.5 M NaCl, 10% glycerol, and 1 mM phenylmethylsulfonyl fluoride)/6 M guanidine hydrochloride (v/v) buffers mixed in ratios of 1:1, 3:1, and 7:1 were added to the column in the given order. Each renaturation buffer contained 20 mM imidazole. Solid phase renatured protein was eluted from the column with MCAC buffers containing increasing concentrations of imidazole, in increments of 20 mM. The NaCl concentration of the MCAC buffer was lowered to 0.15 M (instead of 0.5 M) immediately prior to and including the elution step in order to facilitate the immunization procedure. Eluted protein was concentrated by ultrafiltration in centricon filtration units (cut-off M r 3000). Eight milligrams of solid phase renatured golgin-84 protein was prepared at a concentration of 1.5 mg/ml. Antisera were raised in an accelerated program; three rabbits were each immunized at different sites with denatured and solid phase renatured golgin-84 (Zymed Laboratories Inc.). Antibodies were purified on a golgin-84 affinity column prepared with CNBr-Sepharose 4B (Amersham Pharmacia Biotech). The purified golgin-84 antibodies were eluted from the column with 3 M KSCN, pH 7.0, and dialyzed at 4°C overnight against 10 mM Tris, pH 7.0.
Protein Electrophoresis and Immunoblotting-Protein concentrations were determined with a Bio-Rad protein assay kit II. Cell and tissue lysates (30 g/lane) were separated on 6 -12% SDS-polyacrylamide gels and then electroblotted (110 V-h) onto polyvinylidene difluoride (Millipore Corp.) membrane. The integrity of the Western blots was analyzed by Coomassie Blue staining. Destained membranes were blocked with 3% gelatin (1 h) and then incubated with golgin-84 primary antibody (1:1000) in PBS containing 1% gelatin (room temperature for 1 h). Immunostained proteins were detected with an horseradish peroxidase-labeled secondary antibody (Amersham Pharmacia Biotech) and enhanced chemiluminescence (ECL) reagents (Amersham Pharmacia Biotech).
Light Microscopy-For light microscopy, cells were grown in four-or eight-chambered slides (Lab-Tek) using standard tissue culture techniques. Whole mouse testes were prepared by rapid freezing in OCT compound (Miles) followed by cryosectioning (10 m) and air drying the sections on slides. Cells and sections were fixed with 4% paraformaldehyde for 10 min. (room temperature) before permeabilization with either 0.2% Triton X-100 for 15 min or with 0.2% saponin (Sigma) added to all solutions.
For fluorochrome labeling, cells were fixed, blocked with 10% serum, and then probed with primary and secondary antibodies in PBS. Fluorescein isothiocyanate-and Texas Red-labeled secondary antibodies were used at their recommended working dilutions (Jackson Laboratories). After immunostaining, cells were PBS-washed and mounted with Immunofluor mounting media (ICN).
Horseradish peroxidase immunostaining was accomplished with Vectastain® Elite ABC (peroxidase) and VECTOR® VIP substrate kits (Vector Laboratories). The protocols were carried out essentially as described in the kits with the exception that additional blocking was performed with avidin and biotin (Vector Laboratories). Immunostained testis sections were also counterstained with methyl green (Vector Laboratories) prior to mounting with Permount®.
Immunostained samples were observed by fluorescence epiillumination and bright field microscopy on a Leica DMRBE microscope. Images were recorded on Kodak Ektachrome Slides (400 ASA) and then scanned and processed in Adobe Photoshop 4.
In Vitro Transcription and Translation-Golgin-84 number 16 cDNA was subcloned into Bluescript II SKϩ with its 5Ј-end oriented toward the T7 promoter. The construct DNA was digested with SpeI and phenol chloroform-extracted. Golgin-84 mRNA was then transcribed from the T7 promoter and capped using an mRNA capping kit (Stratagene). Capped golgin-84 mRNA was translated (0.2 g/25 l) for 2 h at 30°C in a rabbit reticulocyte lysate (Promega) containing [ 35 S]methionine (Amersham Pharmacia Biotech) label and canine pancreatic microsomes (Promega). After translation, the microsomes were collected by centrifugation (10 min, 12,000 ϫ g), washed three times with PBS, and incubated in 0.1 M Na 2 CO 3 , pH 11.0, at 4°C for 30 min to remove peripheral membrane proteins. The stripped microsomes were centrifuged at 4°C and resuspended in 1ϫ PBS.
In the proteinase protection study, CaCl 2 and Tris (pH 8.0) were added to the translation mix and put on ice for 15 min. before the proteinase K was added. The final concentrations were 10 mM CaCl 2 , 10 mM Tris (pH 8.0), and 0.2 g/l proteinase K. Proteinase K digestion was performed for 1 h at 4°C, after which the microsomes were pelleted (4°C, 10 min) and resuspended in a protease stop buffer (15 l of PBS, 5 l of 0.1 M phenylmethylsulfonyl fluoride, 5 l of 5ϫ sample buffer), and the resuspended microsomes were immediately boiled for 5 min. As a positive translocation control, ␤-lactamase was used to demonstrate the microsome membranes were functional and intact.
To analyze the mechanism of golgin-84 insertion into microsomal membranes, translation was terminated with a final concentration of 25 g/l cycloheximide. Microsomes were added to the terminated translation reactions and incubated for 2 h at 30°C before being PBSwashed and alkaline carbonate-stripped.
In cross-linking studies, PBS-washed (two times) golgin-84-containing microsomes were exposed to 0.01 and 0.10 mM dithiobis(sulfosuccinimidylpropionate) (DTSSP) in PBS (1.8 l of Promega microsome stock/50 l of DTSSP) for 30 min at room temperature. The crosslinking reactions were quenched with 50 mM Tris, pH 7.5 (final concentration). Cross-linked protein was analyzed by SDS-PAGE under nonreducing conditions, since the DTSSP cross-linker is thiol-cleavable.
After separation by SDS-PAGE, the in vitro translated products were electroblotted to polyvinylidene difluoride membrane and sprayed with En3Hance spray (DuPont) before being exposed to Kodak X-Omat AR film.
Enhanced Green Fluorescent Protein (EGFP)-tagged Golgin-84 Deletion Constructs, Plasmid Construction, and Expression-EGFP-tagged fusion proteins were expressed in cultured cells from cytomegalovirus promoters. Golgin-84 cDNA restriction fragments encoding the C-terminal 51 (BclI, XbaI) and 184 (BglII, XbaI) amino acid residues were subcloned into BglII, XbaI-digested pEGFP-C3 DNA (CLONTECH). Expression constructs were made with nonmethylated DNA and screened for in a methylase minus E. coli strain SCS110 (Stratagene). Once made, the constructs were transformed into E. coli DH5a. DNA was then prepared for transfection using Qiagen kits. Twenty-four hours prior to transfection, normal rat kidney (NRK) cells were seeded (ϳ1.6 ϫ 10 5 cells/ml) into chambered (four or eight chambers) slides (Lab-Tek) coated with poly-L-lysine (Sigma). Cells were transfected by exposing them to serum-free DMEM containing DNA and liposomes (1 g of DNA/5 l of LipofectAMINE (Life Technologies, Inc.) in 1 ml of medium), and after 5 h an equal volume of medium containing 20% fetal bovine serum was added. Transfected cells were analyzed by light microscopy ϳ20 h after the start of transfection. For immunostaining, cells were fixed for 10 min with 4% paraformaldehyde and permeabilized with 0.2% saponin. A pEGFP vector without insert was expressed in cells as a control.

Yeast Two-hybrid Screen
An adult human brain cDNA library was screened with a 4.3-kbp human OCRL1 cDNA (8) using a Matchmaker kit. Of ϳ1.5 million clones screened, two positives were identified. One clone was a novel 781-bp partial cDNA termed 13d. Based on its size and subcellular localization (see below) we term the 13d protein golgin-84. A search of the Unigene data base indicated that the golgin-84 gene maps to distal human chromosome 14q in the region D14S292-qter.

Golgin-84
mRNA Is Ubiquitously Expressed and Abundant in Testis-To analyze golgin-84 expression a mouse multiple tissue poly(A)ϩ mRNA blot was probed with a golgin-84 cDNA (Fig.  1). A single ϳ2.9-kb message was detected that was ubiquitously expressed and abundant in testis. Similar results were obtained with human poly(A) ϩ mRNA blots (message ϳ2.8 kb, data not shown). A 2.63-kbp golgin-84 cDNA (number 16) was then isolated from a human testis library with a 828-bp expressed sequence tag probe, 134426 (Fig. 2). The testis clone was found to lack ϳ0.2 kbp of sequence from its 3Ј-UTR when compared with expressed sequence tag AA405331, which matched 85 bp at the 3Ј-end of clone 16 and extended an additional 209 bp and included a poly(A) ϩ tail. The G ϩ C content was uniform across most of the cDNA but was increased significantly (68%) in the 5Ј-UTR (131 nucleotides). This high G ϩ C content is characteristic of housekeeping promoters and 5Ј-UTRs. Of all of the ATGs analyzed in the golgin-84 number 16 cDNA, the sequence flanking the 5Ј-most ATG (CCATCATGT) conformed best with the Kozak consensus (9, 10) as determined by a weighted matrix. An in frame stop codon occurred 21 nucleotides upstream of the first ATG. The open reading frame extending from this ATG was predicted by the Fickett algorithm (11) to have a 92% chance of coding and encoded a 731-amino acid protein with a mass of 83 kDa. In vitro transcription and translation of the golgin-84 cDNA yielded an ϳ84-kDa protein, which matches the protein size seen by Western blotting of tissues (described below). We conclude therefore, that the combined sequence contains the entire coding region and most if not all of the golgin-84 transcript.
A Ubiquitous 84-kDa Protein Abundant in Testis-To study the golgin-84 protein, antibodies were raised in rabbit against His 6 -tagged denatured and solid phase renatured antigen containing golgin-84 amino acids 498 -688 made in E. coli. Golgin-84 antibodies were purified from rabbit antisera with an affinity column. One affinity-purified antibody immunoprecipitated in vitro translated 35 S-labeled golgin-84 protein and detected an ϳ84-kDa protein on Western blots of human fibroblast lysates. The corresponding preimmune serum did neither, implying that the affinity-purified antibody recognizes both native and denatured golgin-84.
Golgin-84 protein expression was next analyzed in tissues. The affinity-purified golgin-84 antibody was used to probe a Western blot containing lysates (30 g/lane) from eight mouse tissues, the same tissues used for Northern blot analysis (Fig.  3A). Golgin-84 protein was abundant in testis and present at much lower levels in the other tissues, consistent with the Northern blot data. In heart and muscle, the antibody crossreacted with abundant myosin (data not shown).
To study golgin-84 expression further, mouse testis cryosections were histochemically immunostained with the affinitypurified golgin-84 antibody. Immunostained sections indicated that golgin-84 was abundant in the seminiferous tubules and Leydig cells (Fig. 3B). Not all seminiferous cells immunostained with the antibody; staining was greatly reduced or absent in the spermatozoa. The immunostaining pattern differed between seminiferous tubules, probably reflecting differences in the stage of spermatogenesis occurring in different tubules (12). The juxtanuclear immunostaining pattern and its absence/reduction from mature spermatozoa suggested that golgin-84 localized to the Golgi.
A Resident Golgi Protein-To determine if golgin-84 was sublocalized to the Golgi, NRK cells were double-labeled with the affinity-purified golgin-84 antibody and an antibody against the Golgi enzyme mannosidase II (medial-trans cisternae). Golgin-84 was detected by fluorescein isothiocyanate and mannosidase II with Texas Red-labeled secondary antibodies. The fluorescent signals from the two proteins overlapped considerably, indicating that golgin-84 is indeed a Golgi protein (Fig. 4A). Consistent with golgin-84 being a Golgi protein, anti-golgin-84 antibody labeled fragmented structures in mitotically dividing NRK cells (13); the mitotic status of the cells was established by double labeling with a ␤-tubulin antibody (data not shown).
To establish the resident time of golgin-84 in the Golgi, NRK cells were treated with 50 g of cycloheximide/l of culture media for 1, 2, 4, and 7 h. Cycloheximide inhibits protein synthesis without affecting protein traffic across the Golgi compartments, thus clearing the Golgi of proteins in transit. At 7 h, golgin-84 was still in the Golgi, implying that it is resident there (data not shown).
BFA-induced Redistribution Kinetics Is Similar to Known Golgi Integral Membrane Proteins-To confirm golgin-84's sublocalization as well as to examine its redistribution kinetics on Golgi perturbation, NRK cells were exposed to 10 g/ml brefel- din A (BFA) in tissue culture media (14). After a 1-h incubation with BFA, golgin-84 was transported to and distributed throughout the endoplasmic reticulum (Fig. 4B). Several cells still showed an intense staining immediately adjacent to their nuclei, suggesting that golgin-84 may be present in the trans-Golgi in addition to the earlier disassembling cis/middle cisternae (15). Shorter BFA exposure times of 2, 5, 15, and 30 min showed that golgin-84 began to redistribute to the endoplasmic reticulum (ER) at between ϳ2 and 5 min of BFA treatment (data not shown), similar to integral membrane Golgi proteins such as mannosidase II (data not shown), but much slower than rapid on-off proteins involved in coat formation such as ␤-COP (16, 17). Upon drug removal, the Golgi reformed, and golgin-84 returned to it. Cells recovered at different rates. By 1 h, many cells appeared almost fully recovered, with small amounts of golgin-84 remaining in the ER (Fig. 4B).

A C-terminally Anchored Type II Integral Membrane Protein That Post-translationally Inserts into Membranes-Golgin-84
was predicted to be a transmembrane protein with type II topology. Hydropathy profile analysis (18) predicted a single transmembrane domain (TMD) close to its C terminus at amino acid residues 699 -717 (Fig. 5A) with a slight hydrophobic moment on an Eisenberg hydrophobic moment plot (19). The hydrophobic moment suggests the golgin-84 TMD associates with one or more other TMDs in the membrane (19). Based on the charge flanking the golgin-84 TMD, the N-terminal 698 residues of the protein were predicted to be cytoplasmic, and the short 14-residue C terminus was predicted to be luminal (type II membrane protein). No signal sequence was detected with a von Heijne weighted matrix algorithm (20), suggesting To determine if golgin-84 is an integral membrane protein, its mRNA was translated into [ 35 S]methionine-labeled protein in the presence of canine pancreatic microsomes. After translation, the microsomes were washed, stripped to remove peripheral membrane proteins, and collected by centrifugation. The microsomes contained golgin-84, indicating that the protein had been translocated into them (Fig. 5B).
The membrane location and orientation of golgin-84 was then established by a protease protection assay. Microsomes containing the translocated protein were treated with proteinase K. Golgin-84 was digested by the protease to below 66.2 kDa, suggesting that the 698-amino acid N terminus was exposed on the surface of the microsomes, consistent with it being a type II membrane protein (Fig. 5, C and D). A control secretory protein ␤-lactamase was translocated into the microsome lumen and protected from digestion, indicating that the microsomes were functional and intact (Fig. 5C).
Golgin-84 lacks a predicted signal sequence and is thus expected to undergo post-translational insertion into membranes (21). To test this hypothesis, microsomes were added either prior to the start of translation or after translation had been terminated by the addition of cycloheximide (Fig. 5E). Microsomes that were exposed to translating or full-length golgin-84 protein (translation terminated by cycloheximide) for a period of 2 h at 30°C were washed with PBS, alkaline-extracted, and analyzed by SDS-PAGE. As expected, translating golgin-84 was translocated into the microsomes. Full-length golgin-84 protein, exposed to microsomes after termination of translation, was also translocated into the microsomal membranes, implying a post-translational mechanism of insertion.
The C-terminal 51 Residues Contain the Membrane Insertion Sequence and Golgi Retention Signal-In vitro transcription and translation studies indicated golgin-84 to be a C-terminally anchored membrane protein. To confirm that the protein is C-terminally anchored as well as to determine if its C terminus contains membrane insertion and Golgi retention sequences, the C-terminal 51-amino acid region of golgin-84 was tagged at its N terminus with EGFP and expressed in cultured NRK cells (Fig. 6).
EGFP, when expressed by itself, was present throughout the cells. In contrast, the tagged 51-residue C terminus (fusion protein I) was targeted to the Golgi. In low expressing cells, the fusion protein (I) was predominantly in the Golgi, while in high expressing cells it accumulated in the ER as well. The ER and Golgi were distinguished by their morphology as well as by immunostaining with antibodies against mannosidase II (Golgi) and protein-disulfide isomerase (ER). Once in the Golgi, fusion protein I was effectively retained there, and no fluorescent signal could be detected on the plasma membrane. The presence of fusion protein I in the Golgi and its absence from the plasma membrane indicate that the C-terminal 51 residues of golgin-84 contain membrane insertion and Golgi retention sequences. In addition, the absence of coiled-coil sequence in the C terminus implies that coiled-coil interactions are not required for Golgi retention, as suggested by an oligomerization-based model of Golgi retention (3). Finally, backup of fusion protein I in the ER of high expressing cells suggests that (a) the protein inserts into the ER membrane and is then transported to the Golgi, as has been reported for giantin, a Golgi protein of similar structure (22) and (b) export from the ER is inefficient because the fusion protein (I) lacks an ER export signal (23,24) or is detained by the ER quality control system (25, 26).
Since the 51-residue C terminus was inefficiently transported to the Golgi, a larger tagged construct encoding the last 184 residues of golgin-84 (fusion protein II) was made and expressed in NRK cells (Fig. 6). The larger fusion protein was transported efficiently to the Golgi in both low and high expressing cells. Endogenous golgin-84 could not be co-immunoprecipitated with fusion protein II, suggesting that it did not interact with the fusion protein and is not involved in transporting it to the Golgi (data not shown). A putative ER export signal present in fusion protein II but absent in I may be responsible for its more efficient transport to the Golgi. The sequence of amino acid residues DTE, located 37 residues N- FIG. 3. Golgin-84 is a ubiquitous ϳ84-kDa protein abundant in testis. A, an immunoblot of mouse tissue lysates labeled with affinity-purified golgin-84 polyclonal antibodies and detected by ECL. B, golgin-84 is abundantly expressed in seminiferous tubules and Leydig cells of testis. Immunostaining is restricted to a juxtanuclear region characteristic of the Golgi. Mouse testis cryosections were immunostained with affinity-purified golgin-84 antibodies and detected histochemically with a Vectastain ABC kit and Vector VIP substrate. The tissue sections were counterstained with methyl green (bar, 2.5 m).
terminal of the golgin-84 TMD, may ensure efficient recruitment of the protein into COPII vesicles, a requirement for transport to the Golgi (23, 24). Overexpression of fusion protein II in COS-7 cells caused Golgi fragmentation with no plasma membrane labeling, indicating that the Golgi retention mechanism was not saturated and therefore not receptor-dependent (data not shown).
Protein Dimerization-Golgin-84 contains a putative ϳ400amino acid coiled-coil domain in its cytoplasmic N terminus. Two leucine zippers were identified in the golgin-84 sequence (residues 227-248 and 301-322) by a prosite pattern search. Since leucine zippers are a specialized type of coiled-coil, a coils program (version 2.1) was used to analyze the extent of the coiled-coil domain (27,28). The coiled-coil domain, containing discontinuities within it, spans more than half the length of the protein; its boundaries were predicted at amino acid residues ϳ217 and 632 using a window of 28 and an arbitrary cut-off probability of Ͼ90%. The coiled-coil domain was also visible as internal heptad repeats on a dot matrix plot of golgin-84 against itself generated with a moving window of 30 and cut-off of 35% (data not shown).
Golgin-84 Is Similar to Coiled-coil-containing Golgi Proteins-To gain insight into golgin-84 function homology, searches were performed against available World Wide Web data bases such as GenBank TM . Golgin-84 shows significant sequence similarity to several coiled-coil-containing proteins, including Golgi proteins. The strongest similarity is to the RET-II oncogene (BLAST p value ϭ 6.2e Ϫ305 ), which it turns out consists of the first 497 amino acid residues of golgin-84 fused to residues 713-1114 of the RET protein (30). RET residues 713-1114 contain a tyrosine kinase domain that is activated by dimerization (31,32). Activation of the kinase domain in RET-II is probably achieved by constitutive dimerization induced by the golgin-84 coiled-coil sequence. An earlier RET-II report presented the partial golgin-84 protein sequence but neither characterized it nor suggested a mechanism by which it activates RET-II (33).
Golgin-84 shows sequence similarity to the coiled-coil-containing myosin family (BLAST p value Ͻ5.8e Ϫ16 ) and to several coiled-coil Golgi proteins (range of BLAST p values was 2.2e Ϫ14 to 4.6e Ϫ9 ). Golgin-84 is similar to six known Golgi proteins: giantin (macrogolgin), golgin-160, Golgi complex autoantigen-97, golgin-245, trans-Golgi p230, and cis-Golgi matrix protein GM130. Sequence similarities are associated primarily with the ϳ400-residue coiled-coil domain. The functions of most of these Golgi proteins are currently unknown, many having been identified as autoantigens. Golgin-84 is structurally similar to one of the autoantigens, known as giantin, a C-terminally anchored integral membrane protein with an extensive cytoplasmic coiled-coil domain (34,35). The golgin-84 coiled-coil domain and its immediate flanking sequence (residues 173-642) are 45% similar (22% identical) to the giantin coiled-coil sequence (residues 1322-1806). Recently, giantin was shown to tether COPI vesicles to the Golgi and has been proposed to play a role in stacking Golgi disks (7). DISCUSSION Golgin-84 is a novel C-terminally anchored Golgi protein with an extensive cytoplasmic coiled-coil domain. Golgin-84 was established as an integral membrane protein based on its FIG. 4. Golgin-84 is sublocalized to the Golgi. A, NRK cells were double-labeled with antibodies against golgin-84 and a Golgi marker protein mannosidase II. Golgin-84 was detected with fluorescein isothiocyanate-labeled secondary antibodies and mannosidase II with Texas Red-labeled secondary antibodies (bar, 2.5 m). B, golgin-84 is redistributed to the ER in BFA-treated NRK cells. NRK cells were exposed to 10 g/ml BFA in culture media for 1 h, after which time the BFA was removed and the cells were allowed to recover for an additional 1 h. After 1 h of BFA treatment, golgin-84 had been reabsorbed into and distributed throughout the endoplasmic reticulum. An intense juxtanuclear signal was also present in some cells, indicating that some of the protein may be associated with the trans-Golgi network. One hour after the removal of BFA, golgin-84 had regained its Golgi type distribution, although the recovery process was still incomplete (bar, 2.5 m). hydropathy profile and its inability to be extracted in vitro from microsomal membranes with a high pH buffer. Protease protection studies confirmed a predicted type II topology. The post-translational mode of golgin-84 insertion into membranes is consistent with its lack of a signal peptide and the presence of its TMD within the last 50 residues of the protein (21,36).
Golgi targeting sequence(s) were shown by EGFP fusion studies to be in the last 51 residues of the golgin-84 protein, suggesting its transmembrane domain may be involved in Golgi retention. A sequence comparison of Golgi and plasma membrane TMDs indicates that Golgi TMDs are on average 5 residues shorter than plasma membrane TMDs (37). Consistent with this observation, synthetic TMDs 17 leucines in length are retained in the Golgi, while those 23 leucines in length are transported to the plasma membrane (38). The Golgin-84 TMD is predicted to be 19 residues. A lipid-sorting model has been proposed to explain TMD retention by length. The model suggests post-Golgi membranes are thicker (sterols and sphingolipids), thereby preventing the shorter Golgi TMDs from moving forward (38). Oligomerization (kin selection) has also been suggested to play a role in the retention of some resident Golgi proteins (39).
The golgin-84 coiled-coil domain spans more than half the length of the protein, suggesting that the molecule is rod-like in shape. Discontinuities within the coiled-coil domain may create fixed bends or provide flexibility to the structure (40). The coiled-coil domain further suggests that, like myosin II, golgin-84 could form dimers, and we confirmed this by in vitro cross-linking studies. Golgin-84 dimerization may explain its ability to constitutively activate the RET-II oncogene. Included

FIG. 5. Golgin-84 is a C-terminally anchored type II integral membrane protein that inserts post-translationally into membranes.
A, golgin-84 was predicted to be an integral membrane protein with a single transmembrane domain located close to its C terminus (amino acids 700 -717). The hydropathy profile was plotted as the mean hydrophobic index against amino acid number, using a moving window of 19 amino acids with a 1-residue interval (18). B, golgin-84 was in vitro translated in the presence of [ 35 S]methionine and canine pancreatic microsomes. After translation, the microsomes were washed twice with PBS and extracted with an alkaline buffer to remove peripheral membrane proteins. Golgin-84 was retained with the microsomes, suggesting that it was translocated into them. C, the N terminus of translocated golgin-84 is exposed on the surface of the microsome. Microsome-translocated golgin-84 was digested by proteinase K to below 66.2 kDa, suggesting that the 700-amino acid N terminus was exposed to protease. A control protein, ␤-lactamase, was translocated and protected, indicating that the microsomes were functional and intact. D, schematic diagram of golgin-84 translocated into a microsome. E, golgin-84 lacks a signal sequence and undergoes post-translational insertion into membranes. When translated in the presence of microsomes, golgin-84 undergoes translocation via either a co-or post-translational route. Cycloheximide added before translation (bfT) inhibits translation of golgin-84. The addition of cycloheximide after translation (afT) but before microsome addition resulted in translocation, indicating that the fully translated golgin-84 protein is capable of inserting into membranes. Computer analysis did not identify a golgin-84 signal sequence, consistent with a post-translational mechanism of insertion.
in RET-II are the first ϳ274 residues of the golgin-84 coiled-coil domain with its two leucine zippers, which may activate RET-II by constitutive homodimerization. Constitutive leucine zipper-mediated homodimerization has been reported to activate another RET oncogene termed PTC1 in which a fragment of a coiled-coil protein H4 is fused to the RET tyrosine kinase domain (41,42). Unlike PTC1, which is rearranged with RET in 11-25% of papillary thyroid carcinomas, the RET-II rearrangement appears to be an artifact created during transfection of NIH3T3 cells with human colon cancer DNA, since the rearrangement was not detected in the original tumor DNA by Southern blot analysis (30).
The function of golgin-84 is unknown, but some clues are provided by its ubiquitous expression, Golgi sublocalization, and structure and sequence similarity to Golgi proteins of known function. Abundant expression in the testis indicates that it may serve a specialized role there such as in the formation of the acrosome. Alternatively, or in addition, its abundant testis expression may reflect the secretory potential of that organ.
Coiled-coil integral membrane proteins with structures similar to golgin-84 have been implicated in specific docking or in tethering (Velcro factor) of transport vesicles (5). Specific docking proteins termed SNARES ensure that transport vesicles fuse with their appropriate target membranes. A vesicle SNARE (v-SNARE) interacts with a specific target SNARE (t-SNARE) on the acceptor membrane. At first glance, golgin-84 does not appear to be a SNARE, since a prosite profile search indicates it lacks the t-SNARE homology domain reported previously (43). Golgin-84 does, however, show struc-FIG. 6. The C-terminal 51 residues of golgin-84 contain membrane insertion and Golgi retention sequences. A, schematic diagram of wild type and mutant EGFP-tagged golgin-84 proteins. Fusion protein I contains the last 51 amino acid residues of golgin-84, and fusion protein II contains the last 184. B, an EGFP tag expressed by itself appeared throughout NRK cells. Fusion protein I in contrast was targeted to the Golgi (mannosidase II immunostaining) and was not detected on the plasma membrane, indicating that the C-terminal 51 residues of golgin-84 contain both membrane insertion and Golgi retention sequences. A reticular and perinuclear staining pattern, also observed for fusion protein I, suggested that the protein was present in the ER. The ER localization was confirmed by immunostaining with a protein-disulfide isomerase antibody (data not shown). The presence of fusion protein I in the ER suggests that it inserts there but is inefficiently transported to the Golgi. The larger fusion protein II was transported efficiently to the Golgi, suggesting that an ER export signal occurs between residues 547 and 680 of the golgin-84 protein. Consistent with this suggestion, a DTE sequence was identified 37 residues N-terminal of the golgin-84 TMD (bar, 2.5 m). tural and sequence similarity to a larger 350-kDa tethering protein giantin (34,35). Like golgin-84, giantin is a type II C-terminally anchored membrane protein with an extensive cytoplasmic coiled-coil domain. Giantin also forms homodimers and lacks a t-SNARE domain. Giantin has been shown to tether COPI vesicles to Golgi membranes, thereby facilitating protein transport between Golgi cisternae (7). Being as much as four times the width of COPI vesicles, giantin is thought to increase the efficiency of intra-Golgi transport by ensuring COPI vesicles budding from one cisterna become attached to the next before budding is complete. In the tethering process, giantin becomes linked to a series of other proteins; giantin first binds p115 (7), p115 in turn associates with GM130 (44), and finally GM130 interacts with GRASP65 (45). The COPI docking process is inhibited during mitosis by GM130 phosphorylation, resulting in Golgi fragmentation. Golgi fragmentation during mitosis is essential to ensure the stochastic distribution of the organelle between daughter cells (13). Golgin-84, in view of its structure and sequence similarity to giantin, may play a similar role tethering transport vesicles to target membranes. Consistent with this speculation are electron micrographs of immunostained human fibroblasts that show golgin-84 on vesicles (ϳ80-nm diameter) in the perinuclear region. The identity of these vesicles remains to be established. 3 A role for giantin in maintaining Golgi structure has also been proposed. Giantin may link adjacent cisternae within the Golgi stack (35), or it may form a carpet of fibrous attachment proteins that guide transport vesicles within the Golgi stack (46). It has been suggested that giantin is attached to the Golgi cytoskeleton and plays a structural role. Supporting a cytoskeleton association, giantin localization is little affected when cells are extracted with 1% Triton X-100 in a microtubule-stabilizing buffer prior to fixation; in contrast, other Golgi membrane proteins such as galactosyltransferase were completely extracted (35). A similar detergent extraction experiment resulted in complete removal of golgin-84. In that experiment, mannosidase II was also mostly extracted, while the microtubule cytoskeleton, although slightly disrupted, remained intact. 3 Golgin-84, therefore, does not appear to associate with the cytoskeleton.
Taken together, the existing data suggest that golgin-84 plays a role in vesicle tethering/docking. Additional analysis will be needed to confirm such a role as well as to verify any interaction the protein has with OCRL1. If golgin-84 is proven to be involved in tethering/docking, then it should lead to a better understanding of Golgi-associated vesicle transport.