PRA1 inhibits the extraction of membrane-bound rab GTPase by GDI1.

Rab is a family of small Ras-like GTPases regulating intracellular vesicle transport. We have previously reported that prenylated Rab acceptor or PRA1 interacts with Rab GTPases and vesicle-associated membrane protein (VAMP2). Structural prediction programs suggest that PRA1, with its two extensive hydrophobic domains, is likely to be an integral membrane protein. However, subcellular fractionation and immunocytochemical analyses indicated that PRA1 is localized both in the cytosol and tightly associated with the membrane compartment. The membrane-bound form can be partially extracted with physiological buffer and urea, suggesting that PRA1 is an extrinsic membrane protein. Deletion of the carboxyl-terminal domain resulted in a protein that behaved as an integral membrane protein, indicating that this domain plays an essential role in maintaining PRA1 in a soluble state. PRA1 can also bind weakly to GDP dissociation inhibitor (GDI), a protein involved in the solubilization of membrane-bound Rab GTPases. Addition of PRA1 inhibited the extraction of membrane-bound Rab3A by GDI, suggesting that membrane localization of Rab GTPases is dependent on the opposing action of PRA1 and GDI. The binding of Rab and VAMP2 to PRA1 is mutually exclusive such that Rab3A can displace VAMP2 in a preformed VAMP2-PRA1 complex.

With over 30 known isoforms, Rab GTPases are ideal molecular switches in regulating the orderly progression of vesicles through membranous compartments of the eukaryotic exocytic and endocytic pathways (1)(2)(3)(4). The activated GTP-bound form is localized to the membrane, whereas the inactive GDP-bound form is bound to the cytosolic carrier GDP dissociation inhibitor or GDI. 1 Proteins that regulate the cycling between the GDPand GTP-bound forms are essential for Rab function. Translocation of the Rab GTPase between membrane and cytosol is dependent on GDI, which extracts the Rab GTPase from the membrane as well as participates in the delivery of the GTPase to the correct membrane compartment (5)(6)(7). The Rab GTPase is recruited to the membrane compartment following dissocia-tion from GDI, and GDP-GTP exchange catalyzed by a guanine nucleotide exchange factor. A number of proteins are known to possess nucleotide exchange activity. These include rabex-5, which acts specifically on Rab5 (8), and a Rab3A-specific guanine nucleotide exchange protein (9). Mss4, a small cytosolic protein that interacts with a subset of Rab GTPases (10), appears to be involved in this process by transiently binding to the guanine nucleotide-free state (11). By analogy to Ras, the GTP-bound form of Rab serves to recruit and activate effector proteins that eventually mediate the various aspects of vesicle trafficking from transport, docking to eventual fusion. Proteins known to interact with the activated form of Rab include rabphilin-3A (12), rabaptin-5 (13), RIM (14), and rabkinesin-6 (15). The precise mechanism by which these proteins mediate vesicle trafficking remains unclear, but mutation or inactivation of these proteins has been shown to disrupt vesicle trafficking. Subsequent hydrolysis of the bound GTP is thought to terminate the Rab signaling cycle. A GTPase activating protein, of which a Rab3A-specific (16) and the yeast Ypt/Sec-specific (17) forms have recently been identified, stimulates the low intrinsic GTPase activity of Rab. The resulting GDP-bound form Rab is subsequently extracted from the membrane by GDI, which maintains Rab in the cytosol to complete the cycle.
Rab mutations altering GTP binding, hydrolysis, or interactions with effector molecules are known to block exocytic and endocytic vesicle transport (18 -22). In vitro reconstitution experiments in yeast indicate that the docking reaction is sensitive to the GTPase, Ypt1p, which in turn regulates Uso1p-dependent tethering of donor vesicles to target membrane (23). This suggests that Ypt1p mediates vesicle tethering prior to membrane fusion. However, mutation of the synaptic vesiclespecific Rab3A affects synaptic vesicle fusion (24). Transgenic mice exhibited an increase in the number of exocytic events upon calcium-triggered release without a change in the release probability. Thus, the site of action of the Rab GTPases remains unclear and raises the possibility that there are differences in the mechanism of action of the various Rab GTPases.
Vesicle fusion requires an ordered series of molecular events leading to effective interaction between v-and t-SNARE complexes. Studies on yeast vacuolar fusion indicate that formation of the SNARE complexes requires prior priming by NSF and ␣-SNAP, followed by docking and fusion (25,26). In the yeast ER-to-Golgi transport, docking requires a functional Rab GTPase and is independent of the SNARE proteins (23,27).
Most studies indicate that the interaction between SNARE proteins alone appears sufficient for vesicle fusion (28,29), although other proteins such as Munc18/n-Sec1/rbSec1 (30 -32) and tomosyn (33) play a regulatory role in this process. Together these studies indicate that Rab GTPase plays a regulatory role in SNARE-mediated vesicle fusion.
We have recently isolated a protein, PRA1, through a yeast two-hybrid screen that interacts with both prenylated Rab GTPases and the v-SNARE VAMP2 (34). This dual Rab and VAMP2 binding property of PRA1 suggests that it may link the function of these two families of vesicle trafficking proteins. PRA1 is a 21-kDa protein that is conserved from yeast to human (35,36). It contains two extensive hydrophobic domains and is predicted to be a type III membrane protein. However, we report here that PRA1 is not an integral membrane protein, but is present in both the cytosol and membrane compartment. The membrane-bound form can be partially extracted with low salt buffer and urea. Deletion of the carboxyl-terminal 21 amino acids resulted in a truncated protein that behaved as an integral membrane protein. Moreover, the wild-type PRA1 but not the truncated PRA1-(1-164) binds weakly to GDI1. We also show that it can inhibit the extraction of GDP-bound Rab from the membrane by GDI1. Finally, binding of VAMP2 to PRA1 can be displaced by Rab3A, indicating that the three proteins are unlikely to exist as a stable ternary complex.

MATERIALS AND METHODS
Cloning and Purification of PRA1, GDI1, and Rab3A-The cloning and purification of PRA1 from bacteria were as described previously (34). Both recombinant wild-type and truncated PRA1-(1-164) contained a 6xHis followed by a HA tag at the amino terminus. The rat GDI1 was subcloned into Bluescript (Stratagene) vector containing the hemagglutinin (HA) tag by blunt-end ligation onto the EcoRV site. GDI1 was PCR-amplified using the following oligonucleotides: 5Ј-CAT GCC ATG GAT GAG GAA TAC G-3Ј and 5Ј-GGG AAG CTT CAC TGA TCG GCT TCT CC-3Ј. The HA-tagged GDI1 was then isolated as a XhoI/BamHI fragment and subcloned into the EcoRI site of pGEX-2T (Novagen) vector by blunt-end ligation. The purification of GDI1 were as described by the manufacturer for GST fusion proteins. Briefly, the cells were homogenized in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM DTT, 2 mM PMSF. After centrifugation at 10,000 ϫ g for 20 min, the supernatant was loaded onto a glutathione-agarose column. The column was washed with Wash Buffer 1 (25 mM Tris-HCl, pH 7.5, 500 mM NaCl, 10% glycerol, 1 mM DTT) followed by Wash Buffer 2 (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10% glycerol, 1 mM DTT). The fusion protein was then eluted with 10 mM glutathione in Wash Buffer 2.
The rat Rab3A was subcloned as a BamHI/HindIII fragment into pRSET-A (Invitrogen). The NdeI to HindIII fragment was then isolated and blunt-end ligated into the EcoRI site of pAAR6, between the constitutive yeast ADH1 promoter and terminator. The BamHI fragment was then transferred to pRS424 plasmid and transformed into yeast Y190 strain by the LiCl method (37). To purify the 6xHis-tagged Rab3A, the transformed yeast was grown to saturation in 200 ml of Trp dropout medium. The cells were harvested at 3,000 ϫ g for 5 min, transferred to 1 liter of YPD, and grown at 30°C for 8 h. The cells were then collected by centrifugation at 3,000 ϫ g for 5 min, and homogenized with acid-washed glass beads (0.5 mm diameter) in lysis buffer (10 mM Tris-HCl, pH 7.5, 1 mM ␤-mercaptoethanol, 2 mM PMSF) at 1 ml/g of cell pellet. The cell debris and glass beads were removed by centrifugation at 3,000 ϫ g for 5 min. Triton X-100 was added to the supernatant at 1% final concentration and incubated at 4°C for 1 h. The insoluble material was then removed by centrifugation at 100,000 ϫ g for 1 h. The supernatant was loaded onto 1 ml of Ni 2ϩ -NTA beads (Qiagen) and washed with 40 volumes of wash buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM ␤-mercaptoethanol, 0.05% Triton X-100, 10% glycerol). The 6xHistagged Rab3A was eluted with 300 mM imidazole in the above wash buffer.
Expression of PRA1 in Mammalian Cells-The HA-tagged wild-type and truncated PRA1-(1-164) was blunt-end ligated into the XbaI site of the Sindbis virus expression vector, pSinRep5 (Invitrogen). Production of RNA transcripts, transfection into baby hamster kidney (BHK) cells, and collection of pseudovirions were as described by the manufacturer. BHK cells infected with the pseudovirions were processed for immunocytochemistry or homogenization 4 -6 h after infection. The cells were fixed in 4% paraformaldehyde (EM Sciences) in PBS for 30 -60 min and washed with PBS supplemented with 100 mM glycine. The cells were incubated with Blocking Buffer (1% bovine serum albumin, 2% normal goat serum, 10 mM NaN 3 , and 0.4% saponin in PBS) for 30 -60 min. Monoclonal anti-HA (Roche) were diluted in Blocking Buffer and incubated with the cells at room temperature for 1-2 h. The cells were then washed with PBS, and fluorescein-or rhodamine-labeled secondary antibodies (Chemicon) were used to detect the bound primary antibodies. The coverslips were mounted with SlowFade anti-quench solution (Molecular Probes). Images were captured with a Bio-Rad MRC-1024MP laser confocal microscope. Subcellular fractionation of pseudovirion-infected cells was modified from previously described procedure (38). A 70% confluent 60-mm dish of BHK cells was infected with pseudovirions at 10:1 multiplicity of infection for 4 -6 h. The cells were harvested and homogenized by sonication in 1 ml of 10 mM HEPES-NaOH, pH 7.4, 150 mM NaCl, 0.1 mM MgCl 2 , 1 mM EGTA, 2 mM PMSF. The cell debris was removed by centrifugation at 12,500 ϫ g for 10 min. The resulting homogenate was subjected to sequential centrifugation at 30,000 ϫ g for 30 min followed by 150,000 ϫ g for 1 h. The same fractionation procedure was used for PC12 cells and rat brain.
Since the Sindbis expression system often results in extremely high levels of expression, the HA-tagged PRA1 and PRA1-(1-164) were subcloned into the bicistronic pIRESpuro (CLONTECH) for comparison. The PRA1 sequences from the pQE constructs (34) were PCR-amplified using the oligonucleotides 5Ј-CCA TCG ATA TGT ACC CAT ACG ATG TTC CA-3Ј and 5Ј-CGG AAT TCT GAG GTC ATT ACT GGA TCT ATC-3Ј. The resulting PCR fragments were restriction-digested and subcloned into the ClaI and EcoRI sites of pIRESpuro. Chinese hamster ovary cells were transfected with the bicistronic pIRESpuro constructs of wild-type PRA1 and truncated PRA1-(1-164) using LipofectAMINE (Life Technologies, Inc.). Briefly, 1 ϫ 10 5 Chinese hamster ovary cells were seeded on coverslips overnight. Plasmid DNA (0.5 g) was mixed with 1 l of LipofectAMINE in a 100-l final volume of Opti-MEM (Life Technologies, Inc.). The lipid-DNA mixture was added to the PBSwashed cells containing 200 l of Opti-MEM and incubated at 37°C for 3-5 h. The medium was replaced with ␣-minimal essential medium supplemented with 5% fetal bovine serum and antibiotics. The cells were processed for immunocytochemistry after 36 -48 h, as described above. Rabbit anti-calnexin (Stressgen) or anti-mannosidase II (generously provided by Dr. M. Farquhar) was used in conjunction with monoclonal anti-HA in double immunofluorescent labeling reaction.
For subcellular fractionation, BHK cells were seeded at 3 ϫ 10 5 cells/35-mm dish and transfected with 1-1.5 g of wild-type HA-PRA1 or truncated HA-PRA1-(1-164) using 4 l of LipofectAMINE (Life Technologies, Inc.). Cells were harvested after 48 h and resuspended in Buffer A (20 mM Tris-HCl, pH 7.5, 1 mM EGTA, and 1 mM EDTA). The cell suspension was lysed by sonication, and the crude extracts were centrifuged at 5,000 ϫ g for 5 min to remove the cell debris. The resulting supernatant was submitted to a 200,000 ϫ g spin for 30 min at 4°C to collect the membranes. The resulting membrane pellet was resuspended in 240 l of Buffer A, and 40 l was added to 760 l of each of the following: Buffer A, 1 M NaCl in Buffer A, 1 M urea in Buffer A, 0.1 M Na 2 CO 3 and 1% Triton X-100 in Buffer A. After incubation on ice for 30 min, the samples were centrifuged at 200,000 ϫ g for 1 h at 4°C. The supernatants were collected and the pellets resuspended in 200 l of 10 mM Tris-HCl, pH 7.5, 1 mM EDTA. Both supernatant and pellet fractions were precipitated with 10% trichloroacetic acid. The resulting precipitates were solubilized in denaturing buffer and subjected to Western immunoblot analysis. Both PRA1 isoforms were detected with anti-HA antibody.
Preparation of PC12 Microsomal Membranes and Cytosol-PC12 cells from two 100-mm dishes were harvested with 5 ml of 5 mM EDTA in PBS. The cells were homogenized by sonication in 1 ml of TBS (10 mM Tris-HCl, pH 7.5, 150 mM NaCl) with 1 mM PMSF. The cell debris was removed by centrifugation at 10,000 ϫ g for 10 min. The resulting supernatant was spun at 150,000 ϫ g for 60 min. The supernatant was discarded, and the membrane pellet was resuspended by brief sonication in 0.2 ml of 20 mM HEPES-KOH, pH 7.5, 300 mM sucrose, 1 mM PMSF. Protein concentration was determined by Bradford assay (Bio-Rad).
To determine cytosolic proteins that might bind to recombinant PRA1, supernatant after 150,000 ϫ g centrifugation from two 100-mm dishes of PC12 was added to 30 pmol of recombinant His-tagged PRA1 in a 1-ml final volume. A buffer blank was used as a control. The samples were supplemented with MgCl 2 and GDP at 1 and 0.2 mM, respectively. After incubation at 30°C for 30 min, 40 l of 50% slurry of Ni 2ϩ -NTA was added and incubated with gentle agitation at 4°C for 30 -60 min. The Ni 2ϩ -NTA beads were then washed three times with ice-cold HBS (10 mM HEPES-KOH, pH 7.4, 150 mM NaCl). The proteins were eluted from the beads with SDS sample buffer and separated on SDS-polyacrylamide gel. After transfer to nitrocellulose membrane, the blot was processed with anti-GDI (Zymed Laboratories Inc.), followed by the appropriate secondary antibodies and ECL (Amersham Pharmacia Biotech).
Extraction of Rab3A by GDI1-PC12 cells were homogenized as described above. After the 10,000 ϫ g centrifugation step, the mem-branes were isolated from the resulting supernatant by centrifugation at 200,000 ϫ g for 30 min. The resulting pellet was suspended in 0.2 ml of TS (10 mM Tris-HCl, pH 7.5, 300 mM sucrose). A typical assay contained 25 g of PC12 membranes, recombinant GDI1 and PRA1 in 10 mM Tris-HCl, pH 7.5, 1.5 mM MgCl 2 , 250 M GDP in a 200-l final volume. After incubation at 37°C for 20 min, the samples were chilled on ice and subjected to centrifugation at 250,000 ϫ g for 30 min. The supernatant and pellet were then subjected to Western immunoblot analysis using anti-Rab3A (Santa Cruz).
GDI1 Binding Assay-The binding of GDI1 to full-length and truncated PRA1-(1-164) was done using bacterially expressed proteins in a pull-down assay. GDI1 was purified on glutathione-Sepharose as described above. To demonstrate binding of PRA1 to GDI1, 40 pmol of recombinant PRA1 was added to either GST or GST-HA-GDI1 fusion protein in 200 l of TBS (25 mM Tris-HCl, pH 7.5, and 125 mM NaCl). After incubation at 37°C for 10 min, the GST fusion proteins were recovered with glutathione-Sepharose and washed four times with TBS. Both recombinant GDI1 and PRA1 were detected with anti-HA antibodies (Roche Molecular Biochemicals). To demonstrate binding of GDI1 to PRA1, we used thrombin-cleaved product since this gave more efficient binding. The GST-HA-GDI1 fusion protein was cleaved with 3 milliunits of thrombin/pmol of protein at 37°C for 2 h in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2.5 mM CaCl 2 and 288 M ␤-mercaptoethanol, and stopped by the addition of 2 mM PMSF. The GDI1 was subsequently incubated with 40 pmol of Ni 2ϩ -NTA-purified wild-type PRA1 or truncated PRA1-(1-164) at 37°C for 30 min in 25 mM Tris-HCl, pH 7.5, and 125 mM NaCl. The resulting protein complex was pulled down using Ni 2ϩ -NTA-agarose beads and washed four times with 25 mM Tris-HCl, pH 7.5, 125 mM NaCl, 1 mM MgCl 2 , and 0.03% CHAPS. Denaturing loading buffer was added to the beads and proteins subject to Western immunoblot analysis. The GDI1 and PRA1 were detected with anti-GDI1 and anti-HA antibodies, respectively.
In Vitro Competition Assay-Purified recombinant Rab3A isolated from yeast (100 pmol) was incubated in 200 l of Buffer A (25 mM HEPES-KOH, pH 7.5, 140 mM KAc, 1 mM DTT, 0.1% gelatin, 0.01% Triton X-100, 10% glycerol) containing 10 mM EDTA at 37°C for 5 min to remove the bound guanine nucleotide. To replace it with the desired guanine nucleotide, MgCl 2 and GDP were added at 15 mM and 250 M final concentration, respectively. After incubation at 37°C for 10 min, the sample was chilled on ice and used in the competition assay. In a typical assay, recombinant PRA1 was pre-bound to GST-VAMP2 in Buffer A at room temperature for 60 min. The GST fusion protein was recovered with 40 l of 50% glutathione-Sepharose slurry. The beads were then washed three times in Buffer A to remove any unbound proteins. Recombinant Rab3A was then added to the beads in 200 l of Buffer A and incubated at room temperature with gentle agitation for 60 min. The beads were then collected by centrifugation, and proteins in the unbound fraction were precipitated with 10% trichloroacetic acid. The beads were washed three times with Buffer A, and proteins bound to the beads were eluted with SDS loading buffer at 65°C for 10 min. Both bound and unbound fractions were subjected to Western immunoblot with anti-HA monoclonal antibodies.

PRA1 Appears as a Predominantly Cytosolic Protein-To
determine the subcellular localization of PRA1, the full-length and truncated PRA1-(1-164) were transfected into BHK cells and analyzed by immunocytochemistry. The truncated PRA1-(1-164) encompasses amino acids 1-164, including the two hydrophobic domains. We have shown that this carboxyl-terminal deletion abrogated binding to both Rab and VAMP2 (34). Both constructs were subcloned with a HA epitope at the amino terminus into the Sindbis viral expression vector pSinRep5, and the resulting pseudovirions were used to infect BHK cells. The cells were processed for immunocytochemistry or immunoblot at 4 -6 h after infection. This early time point was chosen to avoid possible cytotoxic effect of the Sindbis pseudovirion. Using a multiplicity of infection of 10, approximately 5-10% of the cells remained uninfected, and 25% showed a low level of expression based on staining intensity. The latter showed a punctate staining pattern that was mainly localized at the perinuclear region and a weakly diffused cytosolic staining (Fig. 1A). The majority of the cells exhibited a higher level of expression. These cells showed a more pronounced cytosolic staining while the perinuclear punctate staining pattern remained evident (Fig. 1C). The diffused staining pattern suggested that PRA1 might exist as a cytosolic protein. This was surprising in light of the fact that PRA1 contained two extensive hydrophobic domains. Both hydrophobicity and structural prediction programs have suggested that PRA1 is a type III integral membrane protein with the amino-and carboxyl termini localized to the cytoplasmic side of the membrane. In contrast to the wild-type PRA1, cells expressing the truncated PRA1- (1-164) showed a reticular staining pattern in both low and high expressing cells (Fig. 1, B and D). There was no evidence of a diffused cytosolic staining in the truncated PRA1-(1-164)-infected cells. The truncated protein was also evident in the perinuclear structures, and plasma membrane. Thus, deletion of the carboxyl-terminal 21 residues appears to have caused the protein to remain tightly associated with intracellular membranes.
To exclude the possibility that the pattern observed may be due to overexpression of the protein by the Sindbis virus expression system, we compared the expression of the PRA1 constructs using the bicistronic vector pIRESpuro. The cells were transfected with LipofectAMINE and processed after 2 days. We observed no significant difference in the cellular distribution of the wild-type and truncated PRA1 between the two systems. To further identify the PRA1-associated intracellular organelle, the transfected cells were co-stained with mannosidase II and calnexin, a known Golgi complex (39) and ER (40) marker, respectively. The wild-type PRA1 was found to co-localize with the Golgi marker mannosidase II (Fig. 2, A and  B) and not with the ER marker calnexin (Fig. 2, C and D). In contrast, the reticular pattern of the truncated PRA1-(1-164) showed extensive co-localization with calnexin (Fig. 2, G and  H) with little or no co-localization with mannosidase II (Fig. 2,  E and F). These results clearly indicate that the wild-type PRA1 is preferentially associated with the Golgi complex, whereas most of the truncated PRA1-(1-164) remains tightly associated with the ER structures.
The appearance of the transfected wild-type PRA1 suggests that the protein resides in the cytosol and preferentially associates with the Golgi membranes. To test this, we performed a subcellular fractionation on rat brain and PC12 cells. The results indicated that a substantial fraction of the protein is indeed present in the cytosol after high speed centrifugation (Fig. 3A). The distribution of PRA1 was estimated to be 85% in the cytosol with 15% tightly associated with the membrane. Rab3A was present in both cytosolic and membrane fraction. In contrast, VAMP2, which is an integral membrane protein, was found only in the membrane fraction. This distribution was also evident in PC12 cells (Fig. 3B).
To test whether PRA1 is indeed an extrinsic membrane protein, crude homogenates from cells transfected with either the HA-tagged wild-type or truncated PRA1-(1-164) were extracted with 1 M NaCl, 1 M urea, 0.1 M Na 2 CO 3 , or 1% Triton X-100. Extrinsic membrane proteins are susceptible to extraction by high salt, urea, or Na 2 CO 3 whereas integral membrane proteins tend to be resistant to these reagents and can only be solubilized with detergent. A substantial fraction of PRA1 was extracted with 1 M NaCl and urea (Fig. 4A), suggesting that PRA1 is an extrinsic membrane protein. However, there remains a fraction that is resistant to extraction even with Na 2 CO 3 . In contrast, the truncated PRA1-(1-164) was resistant to extraction with NaCl, urea, and alkaline Na 2 CO 3 . The nonionic detergent, Triton X-100, was able to solubilize both  D and H). The membrane-bound wild-type PRA1 co-localized with the Golgi marker mannosidase II (panels A and B) but not with the ER marker calnexin (panels C and D). In contrast, the truncated PRA1-(1-164) extensively co-localized with calnexin (panels G and H) and showed little or no co-localization with mannosidase II (panels E and F).

FIG. 3. Subcellular fractionation of rat brain (A) and PC12 cells (B).
The tissue and cells were processed as described under "Materials and Methods." Each lane contained 35 g of total protein. Lanes denoted with C represent the crude homogenate. S and P refer to the supernatant and pellet fractions, respectively. The numbers denote the centrifugal g force (in thousands).
full-length and truncated PRA1. To further show that the wildtype PRA1 is indeed an extrinsic membrane protein, we examined whether the reagents above can effectively remove the membrane-bound form of the protein. Microsomal membranes were collected from BHK cells transfected with the wild-type HA-tagged PRA1 and washed to remove the cytosolic PRA1. The membranes were then extracted with the various solutions, and recovered by high speed centrifugation a second time. As shown in Fig. 4B, the membrane-bound PRA1 was partially extracted with Tris-buffered saline and urea. It appeared to be insensitive to salts, as 1 M NaCl and alkaline Na 2 CO 3 were not as effective in removing the membranebound PRA1. These results indicate that PRA1 is both a cytosolic and membrane-associated protein. The fraction that is membrane-bound remains tightly associated with the membrane such that it is resistant to high salt and Na 2 CO 3 extraction, but can be effectively removed by dissolving the membrane with detergent. This unusual property indicates that PRA1 may exist in the cytosol and membrane compartment similar to the Rab GTPases. It also suggests that the carboxylterminal domain play an essential role in maintaining PRA1 in a soluble state. Since deletion of this domain caused the trun-cated protein to behave as an integral membrane protein, it suggests that its deletion has unmasked one or both of the hydrophobic domains, which may act as transmembrane domains in anchoring the truncated protein to the membrane.
GDI1 Co-precipitates with PRA1-We next explored cytosolic factors that might interact with PRA1 by affinity chromatography. Cytosol from PC12 cells was mixed with recombinant 6xHis-tagged PRA1, which was subsequently recovered by Ni 2ϩ -NTA beads. Protein complexes eluted from the beads were analyzed by Western immunoblot with anti-GDI1 antibodies. The results revealed the presence of GDI1 co-precipitating with the recombinant PRA1 (Fig. 5A). Only a small fraction of the total cellular GDI1 was associated with the recombinant PRA1 but none was found to co-precipitate with the Ni 2ϩ -NTA beads in its absence. This may reflect a low avidity in the interaction between the two proteins under our binding and washing conditions. Alternatively, since most of the cytosolic GDI1 exists as a complex with a Rab GTPase, it is possible that the interaction is limited to free or unbound GDI1. This appears to be the case as we were unable to detect the presence of a stable PRA1-Rab3A-GDI1 trimeric complex. To verify this interaction, we examined the binding of these two proteins in vitro. Recombinant 6xHis-, HA-tagged PRA1 was added to bacterially expressed GST or GST-HA-GDI1 in a pull-down assay. The GST-tagged proteins were then recovered with glutathione-Sepharose. PRA1 was readily detected with the GST-HA-GDI1 fusion protein but not with the GST control, indicating that PRA1 can form a complex with GDI1 (Fig. 5B). Conversely, we examined whether GDI1 can be recovered with recombinant PRA1. The GST-HA-GDI1 fusion protein was cleaved with thrombin to remove the GST moiety, which we have subsequently discovered to be more efficient in binding to PRA1 (data not shown). This was added to either the fulllength or truncated 6xHis-, HA-tagged PRA1, and Ni 2ϩ -NTA beads were then used to recover the recombinant PRA1. As shown in Fig. 5C, recombinant GDI1 was recovered along with the full-length PRA1 in the Ni 2ϩ -NTA beads. No GDI1 was recovered with the Ni 2ϩ -NTA beads in the absence of PRA1 or in the presence of equal amount of truncated PRA1-(1-164). Thus, it would appear that PRA1 could interact directly with GDI1 in the absence of Rab, and that deletion of the carboxylterminal domain also abolished this activity.
PRA1 Inhibits the GDI1-mediated Extraction of Rab3A from the Membrane-Removal of GDP-bound Rab from the membrane is dependent on GDI, which is also involved in inhibiting the dissociation of GDP from the Rab GTPases, and in delivering Rab to the correct membrane compartment (7,41,42). Since PRA1 binds weakly to GDI1, we sought to examine whether this can influence the cycling of Rab GTPases. We first confirmed that our recombinant GST-HA-GDI1 fusion protein could indeed extract membrane-bound Rab GTPases in an in vitro extraction assay. Purified recombinant GDI1 was added to PC12 microsomal membranes in the presence of GDP and Mg 2ϩ , conditions known to facilitate extraction of Rab GTPases by GDI1. After incubation at 37°C for 20 min, the membranes were chilled on ice and recovered by high speed centrifugation. Both supernatant and pellet fractions were analyzed by Western immunoblot for the presence of Rab3A. Extraction of Rab3A from the PC12 membranes would result in its appearance as a soluble Rab3A-GDI1 complex in the supernatant fraction. We first determined the amount of recombinant GDI1 required to effectively extract the membrane-bound Rab3A. As shown in Fig. 6A, addition of recombinant GDI1 resulted in the extraction of membrane-bound Rab3A in a concentration dependent manner. Approximately 10 nM GDI1 was required to extract 50% of the total Rab3A in 25 g of PC12 microsomal  PRA1-(1-164). A, extraction of the crude homogenate with 20 volumes of 1 M NaCl, 1 M urea, 0.1 M sodium carbonate, or 1% Triton X-100. Lanes S and P refer to the 100,000 ϫ g supernatant and pellet, respectively. B, extraction of membrane-bound wild-type PRA1. Membranes were collected from cells transfected with HA-tagged PRA1 by centrifugation at 100,000 ϫ g for 1 h. The membranes were resuspended in 20 mM Tris-HCl, pH 7.5, 1 mM EGTA, and 1 mM EDTA (Buffer), and extracted with 20 volumes of Buffer, 1 M NaCl, 1 M urea, 0.1 M sodium carbonate, or 1% Triton X-100 at 4°C for 30 min. The membranes were then recovered by high speed centrifugation. PRA1 in the resulting supernatant (S) and pellet (P) fractions were identified by Western immunoblot using anti-HA antibodies. membranes (Fig. 7A). Increasing the amount of recombinant GDI1 to 30 nM extracted nearly all the membrane-bound Rab3A. A higher amount of GDI1 was required to extract Rab1A, which may reflect either a difference in efficacy or a greater amount of Rab1A in the membrane (data not shown). To test whether recombinant PRA1 can influence the extraction of Rab3A by GDI1, enough recombinant GDI1 was added to effectively remove approximately 50% of the membranebound Rab3A. Under these conditions, addition of increasing amounts of recombinant PRA1 resulted in inhibition of GDI1 extraction of membrane-bound Rab3A (Figs. 6B and 7B). There was a concomitant increase in Rab3A remaining in the membrane pellet with increasing amount of recombinant PRA1. In contrast, addition of recombinant truncated PRA1-(1-164), which was unable to bind to Rab or GDI1, failed to inhibit the extraction of Rab3A by GDI1 (Fig. 7C). Thus, the presence of excess amount of PRA1 exerted an inhibitory effect on the extraction of Rab3A by GDI1 resulting in membrane retention of the Rab GTPase.
Binding of PRA1 to VAMP2 Can be Displaced by Rab GTPase-We have shown previously that PRA1 can interact with Rab GTPases and VAMP2 (34). Since PRA1 can be found associated with the membrane, this raises the possibility that all three proteins might form a transient oligomeric complex on the membrane during the docking process or during recycling after vesicle fusion. Alternatively, binding of PRA1 to VAMP2 may preclude its binding to the Rab GTPase or vice versa. To test these two possibilities, recombinant proteins were used in an in vitro competition assay. VAMP2 was expressed as a GST fusion protein, as described previously (34). To determine the binding capacity of VAMP2, increasing amount of recombinant PRA1 was added until saturation was obtained. Under our binding conditions, 1.2 pmol of PRA1 will saturate 1.8 pmol of VAMP2 in a 200-l reaction volume, suggesting that binding of PRA1 to VAMP2 is likely to be at a 1:1 ratio. To determine whether Rab3A has any effect on PRA1-VAMP2 binding, increasing amounts of recombinant 6xHis-tagged Rab3A purified from yeast were added to the pre-bound PRA1-VAMP2 complex isolated on glutathione-agarose. After incubation at room temperature, the glutathione-agarose beads were recovered by brief centrifugation, washed, and proteins bound as well as that in the supernatant were analyzed by Western immunoblot. As shown in Fig. 8, the amount of PRA1 bound to VAMP2 decreased with increasing amount of Rab3A added. This was accompanied by an increase in the amount of unbound PRA1 appearing in the supernatant. We observed no significant difference between GTP-or GDP-bound Rab3A in dissociating the PRA1-VAMP2 complex. Furthermore, no detectable Rab3A was found to co-precipitate with the PRA1-VAMP2 complex, indicating that all three proteins are unlikely to form a stable oligomeric complex (data not shown). We also observed no direct interaction between Rab3A and VAMP2 in our in vitro binding assay. Thus, we conclude that Rab3A can dissociate a preformed PRA1-VAMP2 complex. The large amount of recombinant Rab3A required to dissociate the PRA1-VAMP2 complex may either reflect the low avidity between Rab3A and PRA1 or that our recombinant Rab3A may not be fully prenylated or a combination of both. Nevertheless, the results imply that either Rab3A and VAMP2 share a common binding site on PRA1 or that binding of the Rab to PRA1 causes a conformational change that ultimately decreased its affinity for VAMP2. DISCUSSION PRA1 was identified through a yeast two-hybrid screen and shown to interact with prenylated Rab GTPases and VAMP2 (34). Structural prediction programs suggest that PRA1 is most likely a type III membrane protein because of its two extensive hydrophobic domains. However, subcellular fractionation and immunocytochemical analyses revealed that PRA1 is present both in the cytosol and membrane. A significant fraction of the protein was found in the high speed supernatant by differential centrifugation analysis. This is consistent with the diffused cytosolic staining pattern observed in wild-type PRA1 transfected BHK cells. However, the transfected cells also exhibited a punctate staining pattern indicative of a membrane-associated protein. The membrane-bound wild-type PRA1 was preferentially associated with the Golgi complex based on extensive co-localization with mannosidase II, a Golgi membrane protein. Partial extraction of the membrane-bound PRA1 with buffer and urea supported the conclusion that the protein is indeed peripherally associated with the membrane. There was a substantial fraction of wild-type PRA1 that is tightly associated with the membrane such that it is resistant to high salt and carbonate extraction. This may reflect a tight association of PRA1 with residual cytoskeletal proteins that sediment with the high speed membrane pellet. There clearly is a link between Rab and cytoskeletal elements. For example, rabphilin-3A, which specifically binds to the GTP-bound form of Rab3A, has been shown to interact with ␣-actinin, an actin bundling protein (43). Moreover, rabkinesin-6, which has a kinesin-like ATPase motor, an extended coiled-coil domain, and a Rab6specific binding domain, is likely to recruit the Rab GTPase to the cytoskeletal element (15). In contrast to the subcellular distribution of wild-type PRA1, deletion of the carboxyl-terminal 21 amino acids resulted in a truncated protein that clearly behaved as an integral membrane protein. It can only be extracted by dissolving the membrane with detergent. Immuno-  (n ϭ 3). B, addition of recombinant PRA1 to the reaction with 10 nM GDI1 inhibited the removal of Rab3A from the membrane. Each sample contains 25 g of PC12 membranes and 10 nM GDI1. Increasing amount of recombinant PRA1 was added to the samples and the distribution of Rab3A in the cytosol (q) and membrane (E) was determined by Western immunoblot. Each point represents the mean Ϯ S.E. (n ϭ 6). C, effect of full-length (q) and truncated PRA1-(1-164) (E) on GDI1-mediated removal of Rab3A from PC12 membranes. The assay conditions were as described in panel B except only the cytosolic Rab3A expressed as a percentage of the total was plotted. Each point represents the mean Ϯ S.E. (n ϭ 4).
FIG. 8. Displacement of PRA1 from VAMP2 by Rab3A. PRA1 was pre-bound to GST-VAMP2, and the complex was recovered using glutathione-Sepharose. The beads were washed to remove nonspecifically bound proteins. Recombinant Rab3A was pre-bound with GDP and added to the washed glutathione-Sepharose beads. Proteins in the unbound and bound fractions were recovered and analyzed by Western immunoblot using anti-HA antibodies. cytochemical analysis also revealed extensive co-localization of the truncated protein with calnexin, an ER marker. This suggests that deletion of the carboxyl-terminal domain resulted in insertion of the truncated PRA1-(1-164) into the ER membrane with the two hydrophobic domains presumably acting as membrane-spanning domains. Thus, the carboxyl-terminal domain of PRA1 is indispensable in maintaining either the protein's structural integrity or solubility.
The presence of PRA1 in the cytosol and membrane indicates that the protein can exist in these two compartments. This is not unique as other proteins known to interact with the Rab GTPases, such as GDI and Mss4/Dss4 (42,44), also exhibit this property. We also found that a small fraction of the cytosolic GDI1 can co-precipitate with recombinant PRA1. This small amount most likely represents free GDI1 because most, if not all, of the cytosolic GDI1 exists as a Rab-GDI complex. This interaction between PRA1 and GDI1 was confirmed by in vitro binding assay using recombinant GDI1, which does not contain bound Rab GTPase. Moreover, only the full-length PRA1 but not the truncated PRA1-(1-164) was able to bind to GDI1. We were unable to detect the presence of a stable complex consisting of Rab, PRA1, and GDI1. Since the addition of recombinant GDI1 failed to have a significant influence on the subcellular distribution of PRA1 (data not shown), we believe that other cytosolic factors are responsible for the maintenance of PRA1 in the cytosol. Hence, we favor the notion that the interaction between PRA1 and GDI1 might be involved in the recruitment of cytosolic GDI1 to the membrane.
Proper functioning of Rab involves reversible membrane translocation and cycling to the cytosol with the latter dependent upon GDI (42). We have shown that purified recombinant GDI1 can effectively remove Rab3A from PC12 membranes, and that this activity can be blocked by the addition of recombinant PRA1. The truncated PRA1-(1-164) was ineffective, which is consistent with its inability to interact with GDI1 or Rab GTPases. It is unlikely that PRA1 plays a role in the solubilization of Rab GTPases, since addition of either the recombinant wild-type or truncated PRA1 to the membranes in the absence of GDI1 showed no effect on the solubilization of Rab3A (data not shown). Thus, the membrane and cytosolic distribution of Rab depends on the opposing action of PRA1 and GDI1 with PRA1 favoring membrane retention of Rab and GDI1 maintaining Rab soluble in the cytosol. Although membrane detachment is not obligatory for Rab function in yeast (45), it nevertheless is required and that the loss of protein such as GDI1 in X-linked mental retardation can clearly disrupt membrane trafficking (46). Once translocated to the membrane, the binding of PRA1 to Rab3A and VAMP2 is mutually exclusive such that Rab3A can displace VAMP2 from PRA1. We observed no difference between GDP-and GTP-bound Rab3A in displacing the PRA1-VAMP2 interaction (data not shown). Since GDP-bound Rab3A is effectively bound to GDI1 and remains in the cytosol in vivo, it is reasonable to assume that it is the GTP-bound Rab3A that is responsible for dissociating the PRA1-VAMP2 interaction in the membrane. Taken together, these results suggest that Rab and VAMP2 may share a common binding site in PRA1, and that the binding of PRA1 to either Rab or VAMP2 may influence the availability of these proteins to interact with potential effector molecules. The fact that PRA1 can bind to multiple proteins is not unusual given that proteins such as Rab3A can interact with a host of proteins in a guanine nucleotide-dependent manner (9,12,14,16,47). Similarly, syntaxin is known to interact with a number of proteins with binding to one protein influencing its binding to another (30 -33, 48). Such sequential binding may be a general mean by which these proteins are regulated.
Because both PRA1 and GDI1 are known to bind to a number of Rab GTPases, our results suggest that these two proteins may ultimately determine the level of membrane association of Rab GTPases in general. It is not clear whether membrane association of PRA1 occurs before or after Rab GDP/GTP exchange. However, it is likely that PRA1 remains with the Rab GTPases since it showed little preference for either the GDP-or GTP-bound form. The fact that the binding of PRA1 to Rab and VAMP2 is mutually exclusive suggests that PRA1 may influence the availability of these two proteins to interact with other regulatory proteins at the membrane. Upon completion of vesicle fusion, recycling of GDP-bound Rab to the cytosol may require prior dissociation of the PRA1-Rab complex, a step that might be facilitated by the availability of VAMP2 at the membrane. The interaction between PRA1 and GDI1 may also serve to recruit GDI to the membrane thereby facilitating the removal of membrane-bound Rab. Although a number of questions remain to be resolved, this model may provide a framework for future study directed at dissecting the regulation of the Rab GTPases.