Novel G Proteins, Rag C and Rag D, Interact with GTP-binding Proteins, Rag A and Rag B*

Rag A/Gtr1p are G proteins and are known to be involved in the RCC1-Ran pathway. We employed the two-hybrid method using Rag A as the bait to identify proteins binding to Rag A, and we isolated two novel human G proteins, Rag C and Rag D. Rag C demonstrates homology with Rag D (81.1% identity) and with Gtr2p ofSaccharomyces cerevisiae (46.1% identity), and it belongs to the Rag A subfamily of the Ras family. Rag C and Rag D contain conserved GTP-binding motifs (PM-1, -2, and -3) in their N-terminal regions. Recombinant glutathione S-transferase fusion protein of Rag C efficiently bound to both [3H]GTP and [3H]GDP. Rag A was associated with both Rag C and Rag D in their C-terminal regions where a potential leucine zipper motif and a coiled-coil structure were found. Rag C and D were associated with both the GDP and GTP forms of Rag A. Both Rag C and Rag D changed their subcellular localization, depending on the nucleotide-bound state of Rag A. In a similar way, the disruption of S. cerevisiae GTR1 resulted in a change in the localization of Gtr2p.

G proteins are a superfamily of regulatory GTP hydrolases and are composed of a large number of proteins. These include Ras family proteins, hetrotrimeric G protein ␣ subunits, and elongation factors TU and G, among others (1). Ras-like small G proteins such as Ras, Rab, Rho, ARF, and Ran are monomeric and bind to the guanine nucleotides, GTP or GDP, to function as molecular switches while also playing crucial roles in cell growth, differentiation, and protein traffic between different compartments within the cells (2,3). Ras is a key regulator of cell growth and is an essential component of the signal transduction pathways initiated by receptor tyrosine kinase (4). The Rho family members consist of Rho, Rac, and Cdc42 subtypes that control the actin cytoskeleton and that play a role in the regulation of transcription (5). ADP-ribosylation factors play a role in the vesicular trafficking pathway (6). The Rab subfamily plays a role in secretory and endocytic pathways and is located within a distinct cellular compartment (7).
RCC1 catalyzes guanine nucleotide exchange on Ran (30) and is located inside the nucleus, bound to chromatin (31). The concentration of GTP within the cell is ϳ30 times higher than the concentration of GDP, thus resulting in the preferential production of the GTP form of Ran by RCC1 within the nucleus (8). In the cytoplasm, the GTP of Ran is hydrolyzed to GDP through the aid of RanGAP1, which is located within the cytoplasm, thus producing a difference in the concentration of the GTP form of Ran between the nucleus and the cytoplasm. The loss of RCC1 resulted in hamster (tsBN2) and yeast pleiotropic phenotypes, such as premature chromosome condensation, lack of chromosome condensation, the suppression of the receptorless mating process, chromosome instability, an abnormal mRNA metabolism, and mRNA export defects (reviewed in Ref. 28).
Mutation of GTR1 suppressed the prp20 mutation of Saccharomyces cerevisiae RCC1, thus suggesting that the function of GTR1 is related to that of the RCC1-Ran (PRP20/MTR1/ SRM1-GSP1) system (32). Rag A is a human homologue of GTR1 as shown by a high sequence similarity and by the fact that the mutation of Rag A suppressed the prp20/mtr1 mutation of S. cerevisiae (33). Among the RCC1-Ran system proteins, Gtr1p/Rag A is another subfamily of Ras-like small G proteins (34). Gtr1p is located within both the cytoplasm and the nucleus and has been reported to play a role in cell growth (35,36). Rag A was originally isolated by polymerase chain reaction (PCR) 1 methods during the search for novel G proteins (37) and also independently by two-hybrid screening using adenovirus E3 14.7 kDa as the bait and thus was shown to be involved in apoptosis (38). Rag A and Gtr1p were shown to belong to the sixth subfamily of the Ras-like small GTPase superfamily (34,37). Rag A and Rag B differ by seven conserv-ative amino acid substitutions (98% identity) and by 33 additional residues at the N terminus of Rag B (37). There are three alternative splice variants in Rag B, such as Rag B s , Rag B l , and Rag B n (33,37).
We herein identify two novel Rag A-interacting proteins, Rag C and Rag D. Rag C had GTP-binding motifs in its N-terminal region and was also shown to bind significantly to [ 3

H]GTP and [ 3 H]GDP in vitro.
Both Rag A and Rag C had a mutual binding region in their C-terminal regions downstream of the GTPbinding region. These structural features led us to propose that Rag C and Rag D belong to the Rag A subfamily of the Ras-like small G protein superfamily. Rag A and Rag B bound to both Rag C and Rag D. Rag A was also colocalized with Rag C and Rag D in BHK21 cells, thus suggesting that Rag A may form a heterodimer with Rag C or with Rag D.

EXPERIMENTAL PROCEDURES
Cell Culture and Transient Transfection-BHK21 cells and HeLa cells were grown at 37.5°C in Dulbecco's modified Eagle's medium containing 10% calf serum, penicillin (100 units/ml), and streptomycin (100 g/ml) in a humidified atmosphere of 10% CO 2 , 90% air. Cells were washed with TD buffer (25 mM Tris-HCl, pH 7.4, 136.8 mM NaCl, 5 mM KCl, and 0.7 mM Na 2 HPO 4 ). BHK21 cells (2 ϫ 10 5 cells) were transfected with DNA-lipid complex of 1 g each of various vectors and 7 (for 35-mm dish immunofluorescence) or 15 l (for 60-mm dish immunoprecipitation) of LipofectAMINE TM Reagent (Life Technologies, Inc.) for 4 h in the absence of serum and antibiotics, as recommended by the supplier, and incubated at 37.5°C for 48 h as described (39).
Yeast Growth Media, Two-hybrid Screening-S. cerevisiae cells were grown in the following media: YPD (2% glucose, 2% peptone, and 1% yeast extract), SD-Leu, Trp, His (2% glucose and 0.67% yeast nitrogen base without amino acids, supplemented with all essential amino acids except for leucine, tryptophan, and histidine), and GE-Leu, Trp (3% glycerol, 2% ethanol, and 0.67% yeast nitrogen base without amino acids, supplemented with all essential amino acids except for leucine and tryptophan). Amino acids were added at a final concentration of 20 -50 g/ml. The solid media contained 2% agar in addition to the components described above.
Galactosidase Assay-For quantitative ␤-galactosidase assay, yeast cells harboring a two-hybrid vector encoding GAL4BD and GAL4AD fusion proteins were grown in 2 ml of SD-Leu, Trp medium overnight. Then 0.4 ml of the yeast culture was mixed with 15 ml of GE-Leu, Trp medium in a small flask. The yeast culture was grown until the A 600 was 0.8 -1.0 at 30°C. Then 15 ml of culture were centrifuged, washed once with water, and resuspended into 0.2 ml of lysis buffer (0.1 M Tris-HCl, pH 8.0, 20% glycerol, and 1 mM phenylmethylsulfonyl fluoride). The cell suspension was mixed with 0.4 mg of glass beads (Sigma) and vortex-mixed for 30 s, 6 times. The cell suspension was mixed with 0.2 ml of the lysis buffer. Cell lysate was obtained by centrifugation at 10,000 ϫ g for 5 min. 0.1 ml of the cell lysate was incubated with 0.9 ml of Z-buffer (100 mM Na 2 PO 4 , pH 7.0, 10 mM KCl, 1 mM MgSO 4 , and 5 mM 2-mercaptoethanol) and 0.2 ml of -nitrophenyl-␤-D-galactopyranoside (4 mg/ml) at 28°C as described previously (34). When the reaction solutions turned yellow, the reaction was stopped with 0.8 ml of 1 M Na 2 CO 3 . One unit of ␤-galactosidase activity was calculated as 1 nmol of -nitrophenyl-␤-D-galactopyranoside cleaved per min/mg of protein at 28°C. Yeast colonies grown on the plates with synthetic medium lacking tryptophan and leucine were also analyzed by ␤-galactosidase filter assay using 5-bromo-4-chloro-3-indolyl ␤-D-galactoside (X-gal) as described elsewhere (41,42).
Purification and Nucleotide-binding Assay of GST-Rag C and GST Proteins-E. coli BL21 harboring a GST plasmid was grown in 750 ml of LB medium, treated with isopropyl ␤-D-thiogalactoside (final concentration, 0.2 mM) for 4 h at 30°C as described (20). Cells were dispersed in lysis solution at a ratio of 1:5 (cell volume:lysis solution (1ϫ PBS, 2 mM EDTA, 0.1% ␤-mercaptoethanol, 0.2 mM phenylmethylsulfonyl fluoride, and 10 g/ml aprotinin)) and sonicated for 5 min three times on ice (Sonicator TM , Heat system-Ultrasonics Inc., microtip, 40% cycle, output control 4). After centrifugation at 10,000 ϫ g for 30 min at 4°C, 10 ml of the supernatant (50 mg/ml protein concentration) were mixed with 1 ml of 50% (v/v) slurry of glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) and rotated for 30 min at 4°C. The beads were washed four times with the lysis buffer.
GST protein beads that were washed with nucleotide-binding buffer A (25 mM Tris-HCl, pH 7.5, 50 mM NaCl, 5 mM EDTA, 1 mM DTT, and 1 mM CHAPS) or buffer B (20 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 10 mM MgCl 2 , 1 mM DTT, and 0.1% Triton X) were incubated with various concentrations of [ 3 H]GDP or [ 3 H]GTP for 30 min at 30°C in a total volume of 50 l of binding buffer for the binding assay in Fig.  2, a-c, as described (34). The reaction was stopped by the addition of 1 ml of the ice-cold stop buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 25 mM MgCl 2 ). Nucleotides bound to GST protein beads were washed with 1 ml of stop buffer four times and then transferred into Clear-sol (Nakarai Chemicals, Kyoto, Japan). Radioactivity was counted by a scintillation counter.
For Fig. 2, d-f, GST fusion proteins were eluted from glutathione-Sepharose 4B beads with 2 ml of buffer (50 mM Tris, pH 8, 10 mM reduced glutathione) three times (total, 6 ml), concentrated with Ultrafree-15 centrifugal filter device-Biomax 10K (Millipore Corp, Bedford, MA), and suspended in PBS to a final concentration of 1 mg/ml GST proteins. GST-Rag C and GST were incubated with various concentrations of [ 35 S]GTP␥S, [ 3 H]GDP, or [ 3 H]GTP for 30 min at 30°C in 50 l of binding buffer for the binding assay. The reaction was stopped by the addition of 2 ml of the ice-cold stop buffer. Nucleotides bound to the protein were trapped by filtering the reaction mixture through nitrocellulose filters. Filters were washed three times with 25 ml of the same ice-cold stop buffer. Radioactivity was counted by a scintillation counter. Ten percent of GST-Rag C (0.3 nM) bound to [ 35 S]GTP␥S (0.03 nM) maximally. Nucleotide off-rates (dissociation) were found to follow the first-order kinetics. 4 nM [ 35 S]GTP␥S or 0.68 mM [ 3 H]GDP were incubated with 250 nM Rag C for on-rate experiments. The on-rate (association) and off-rate (dissociation) experiments were done according to procedures described previously (43,44). The GTPase activity was assayed essentially based on a procedure described previously (37). Briefly, GST-Rag C (2 M) or GST (2 M) was mixed with [␣-32 P]GTP (3,000 Ci/mmol, 0.6 nM) in a 50-l buffer (50 mM Tris-HCl, pH 8.0, 2 mM EDTA, 1 mM DTT, 10 mM MgCl 2 , and 500 g/ml bovine serum albumin) at 30°C for 2 or 4 h. Reaction buffer (5 l) was mixed with 2 M formic acid (5 l) and separated by thin layer chromatography on polyethyleneimine cellulose in buffer (1 M lithium chloride, 1 M formic acid).
Recombinant DNA-Fusion proteins with GAL4 DNA-binding domain (GAL4BD) were constructed in pAS1 or pAS404 (34), and those with the GAL4 activation-binding domain (GAL4AD) were constructed in pACT2. The NcoI/BamHI fragment of human Rag A cDNA from T7-RagA-pcDEB⌬ (33) was inserted into the NcoI/BamHI sites of the pAS404 vector (34). Deletion cDNA fragments of Rag A, Rag C, and Rag D in Fig. 5 were polymerase chain reaction-amplified (PCR) by KOD polymerase (Toyobo Co. Ltd., Osaka, Japan) as recommended by the supplier using various synthetic oligonucleotide primers and were subcloned into pAS1 or pACT2 vectors. Each construct was checked by sequencing. Synthetic oligonucleotides were purchased from Hokkaido System Science Ltd. (Sapporo, Japan). Rag C cDNA, which was amplified by PCR with synthetic oligonucleotides introducing a BamHI cleavage site at the ATG and a BglII at the stop codon at the C terminus, was inserted into the BamHI site of pGEX-KG. NcoI/SalI fragments of the C-terminal regions of Rag A and Rag C were inserted into the NcoI/XhoI sites of pET28a. BamHI/BglII fragments of Rag C and Rag A were inserted into the BglII site of pEGFP-C1 (CLONTECH Laboratories Inc., Palo Alto, CA). pCDNA-HA Rag C and Rag D were constructed by inserting XhoI-BglII fragments of TKS-HA1 Rag C and Rag D into pCDNA3.1/Hygro(Ϫ) (Invitrogen Corp., Carlsbad, CA). T7-Rag A-pc-DEB⌬, T7-Rag A gtr1-11 -pcDEB⌬, T7-Rag A Q66L -pcDEB⌬, and T7-Rag B-pcDEB⌬ were described previously (33). A gtr1-11 mutation, a mutation of residue 21 from serine to leucine (hereafter referred to as T21L) in Rag A, corresponds to a dominant negative form of Ran, T24N (GDP form) (45). The Q66L (residue 66 from glutamine to leucine) mutant of Rag A corresponds to the dominant positive form of Ran Q69L (GTP form) (24). MYC-GTR2, pL275 (pRS404 containing Myc-GTR2), was constructed by inserting Myc tag DNA into GTR2 at the N-terminal region, which was subcloned into the SalI/SacI site of pRS404.
Immunoprecipitation, Immunoblotting, and Antibodies-1 ϫ 10 6 cells that were transfected with various vectors were lysed in 1 ml of lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Nonidet P-40, 5 mM EDTA, 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl fluoride, 10 g/ml each of aprotinin, leupeptin, pepstatin, and antipain) and clarified by centrifugation at 10,000 ϫ g for 10 min at 4°C. Cell extracts (500 g) were incubated with 1 l of anti-HA antibodies or anti-T7 antibodies for 1 h at 4°C and then were incubated with 50 l of protein G-Sepharose 4 Fast Flow (Amersham Pharmacia Biotech) (50% v/v) for an additional hour. Immunoprecipitates were collected by centrifugation at 10,000 ϫ g for 2 min at 4°C, washed five times with 1 ml of the lysis buffer at 4°C, and were then suspended in 50 l of the sample buffer containing 62.5 mM Tris-HCl, pH 6.8, 10 mM 2-mercaptoethanol, 3% (w/v) SDS, and 20% glycerol before being boiled. Protein samples were electrophoresed in a 10 or 12.5% SDS-polyacrylamide slab gel and analyzed by immunoblotting. Immunoblotting was performed as described (46) using an ECL kit (Amersham Pharmacia Biotech) as recommended by the supplier. Rabbit anti-HA antibody (catalogue number 561) was purchased from the MBL Institute (Nagano, Japan). Mouse anti-T7 antibody (catalogue number 69522) was purchased from Novagen Inc. (Madison, WI). Rabbit anti-Myc antibody (A-14) was purchased from Santa Cruz Biotechnology Inc. (Santa Cruz, CA).
In Vitro Binding Assay-For the in vitro binding assay, the C-terminal portions of Rag A and Rag C were synthesized in vitro. The Rag A and Rag C cDNAs in the pET28a vector (1 g) were incubated with 20 Ci of [ 35 S]methionine (PerkinElmer Life Sciences) and 40 l of a quick master mix of TNT R T7 Quick-coupled Transcription/Translation Systems (Promega Corp., Madison, WI) for 90 min at 30°C as recommended by the supplier. The resultant extract was diluted to 500 l with the immunoprecipitation buffer. Either GST (20 g), GST-Rag A (20 g), or GST-Rag C (20 g), which were bound to the glutathione-Sepharose 4B beads, was mixed with the 35 S-labeled proteins. After incubation at 4°C for 30 min, the beads were spun down, washed 4 times, and suspended in 50 l of the sample buffer. Bound proteins were run on SDS-polyacrylamide gel and analyzed using the Fuji BAS2000 Image Analyzer (Fuji Photo Film Co. Ltd., Japan).
Immunofluorescence-2 ϫ 10 5 cells, which were seeded on coverslips in 35-mm dish, were transfected with 1 g of DNAs using 7 l of LipofectAMINE Reagent TM as described above. The transfected cells on coverslips were fixed with 1 ml of cold methanol/acetone (1:1) for 5 min at Ϫ20°C and washed three times with 1 ml of PBS buffer (Sigma). Fixed cells were incubated with blocking buffer (PBS containing 3% bovine serum albumin, 0.2% Tween 20, and 10% normal goat serum) for 1 h at room temperature and then were processed to immunostaining by primary antibody and then by either FITC-or Texas Red-conjugated secondary antibodies as described previously (39). Cells on coverslips were mounted on Vectashield (Vector Laboratory Inc., Burlingame, CA). Digital imaging of stained cells was obtained using the Olympus laser-scanning microscope LSM-GB200 system as described (33).

Identification of Novel Proteins, Rag C and Rag D, That
Interact with Rag A-Human Rag A is a new member of the Ras-like small G protein superfamily and is involved in cell growth and the RCC1-Ran system. To elucidate the function of Rag A further, it may be helpful to identify interacting proteins that have identifiable enzymatic functions and/or specific sub-cellular localizations. We performed yeast two-hybrid screening using human Rag A as the bait from human Burkitt's lymphoma cDNA library. We picked up and analyzed 100 samples of human cDNA that were histidine-and 3-aminotriazolepositive and found two positive genes, Rag C and Rag D, the number of cDNA clones being 54 and 22, respectively. ␤-Galactosidase filter assay showed that clones expressing Rag A fused with the GAL4 DNA-binding domain and those expressing Rag C or Rag D fused with the GAL4-transactivation domain had significant ␤-galactosidase activity (Fig. 4a, clones 1 and 4). As controls, colonies expressing Rag A fused with the GAL4 DNA binding domain and vector pACT2 fused with the GAL4-transactivation domain were examined, but these did not show any ␤-galactosidase activity (Fig. 4a, clone 7).
Sequence analysis showed Rag C and Rag D to be similar proteins with a homology of 81.1% regarding the level of amino acids (Fig. 1a). Whereas N-and C-terminal sequences were not conserved, the middle part of both proteins was highly conserved (more than 90% identity from 61 to 369 amino acid residues of Rag C and from 62 to 370 amino acid residues of Rag D). A homology search using GenBank TM revealed Rag C to have a similarity to Schizosaccharomyces pombe Gtr2p (spGtr2p) (63.3% identity) and to S. cerevisiae Gtr2p (46.1% identity).
Although the similarity between Rag C and Rag A was low (24.4% identity) (Fig. 1b), Rag C did have considerable sequence similarity to Rag A regarding the structural motifs that are conserved within the Ras family (48) as follows: phosphate/ magnesium-binding motifs PM1 (GXXXXGKS), PM2 (Thr), and PM3 (WDXXGQ) in the other members of the Ras family were 68-GLRRSGKS, 96-T, and 115-WDFPGQ in Rag A, respectively, as indicated in Fig. 1, a and b. The guanine nucleotidebinding motifs (G1, G2, and G3) are strikingly different from those found in other members of the Ras family. Since 31tyrosine of Rag A was proposed to be the G1 motif (37), the G1 motif of Rag C could be 85-methionine. The conserved asparagine residue in the G2 motif (NKXD) in other members of the Ras family is histidine (178-HKVD), whereas that of Rag A is 127-HKMD. The G3 motif of Rag C could be 215-TSI, whereas that of Rag A is 162-TSI. Structurally, the five polypeptide loops that form the guanine nucleotide-binding sites are highly conserved and encompass PM1, PM2, PM3, G2, and G3 motifs in Ras-like small G proteins. Since Rag C and D have conserved PM1, PM2, PM3, G2, and G3, Rag C and Rag D may in fact be G proteins and may bind to the guanine nucleotides.
The C-terminal domains of Rag C and Rag D, similar to those of Rag A, are larger than those of other known mammalian Ras family members (37) and are probably unrelated to guanine nucleotide binding. In the C-terminal region of Rag A and Rag C, a potential leucine zipper motif that was implicated in dimer formation was found just downstream of the nucleotide-binding region, as shown by black circles in Fig. 1b. Moreover, in the same region, a coiled-coil structure that was also implicated in protein-protein interaction was predicted in both Rag A and Rag C (Fig. 5) using a computer software package (COILS program) (49). Although cysteine residues are reported to function in anchorage to the membrane in the C-terminal region of many GTP-binding proteins, Rag C and Rag D similar to Rag A do not have any cysteine residues in their C-terminal portion. These similar sequence features led us to assume that Rag C and Rag D belong to the Rag A subfamily of the Ras superfamily.
Rag C Is Another Guanine Nucleotide-binding Protein-Deduced amino acid sequence examination as described above strongly suggested that Rag C and Rag D are in fact G proteins. To confirm this prediction, we tested whether Rag C has the ability to bind to the guanine nucleotides, GTP and GDP, in vitro. Rag C cDNA was subcloned into the pGEX-KG vector. GST and GST-Rag C fusion proteins were produced in E. coli BL21 cells and purified using glutathione-Sepharose 4B beads, as described under "Experimental Procedures." GST-Rag C (560 pmol) and GST (560 pmol) proteins, which were bound to glutathione-Sepharose 4B beads, were mixed with various amounts of [ 3 H]GTP and [ 3 H]GDP in vitro at 30°C. Thirty minutes later, the mixture was washed four times with the stop buffer, and the remaining radioactivity was counted by a liquid  (Fig. 2b). Moreover, the binding of GST-Rag C to [ 3 H]GTP was inhibited by nonradioactive GTP and GDP in a dose-dependent manner, thus implying that Rag C is a G protein (Fig. 2c). The binding of GST-Rag C to [ 3 H]GTP persisted even after an excessive amount of ATP was mixed in the binding reaction as a control, thus demonstrating that Rag C bound specifically to guanine nucleotide (Fig. 2c).
To characterize Rag C protein first as a G protein, guanine nucleotide on-rates and off-rates were studied (Fig. 2, d and  e). Since [ 35 S]GTP␥S did not bind to Rag C in the absence of Mg 2ϩ , an assay was performed in the presence of 10 mM MgCl 2 . The on-rates of GTP were calculated to be about 8.3 ϫ 10 5 s Ϫ1 M Ϫ1 . The off-rates of GTP were 1.3 ϫ 10 Ϫ4 s Ϫ1 . The dissociation constant (K d ) for GTP of Rag C was thus 1.6 ϫ 10 Ϫ10 M. Mg 2ϩ had no effect on GDP binding. Saturation binding occurred in 1 min, when Rag C (250 nM) was mixed with [ 3 H]GDP (675 nM). The on-rates of GDP were 4.9 ϫ 10 4 s Ϫ1 M Ϫ1 . The off-rates of GDP were 1.4 ϫ 10 Ϫ2 s Ϫ1 . When comparing the off-rates between GTP␥S and GDP from Rag C, Rag C released GDP very quickly. The dissociation constant (K d ) for GDP of Rag C was 2.9 ϫ 10 Ϫ7 M. It thus seems that Rag C binds to GTP predominantly and releases GDP quickly. Fig. 2f showed that GST-Rag C had a weak intrinsic GTPase activity when GST-Rag C (1.25 M) was incubated with [␣-32 P]GTP (0.6 nM) for an assay.  4. The association of wild-type, T21L, and Q66L forms of Rag A with Rag C and Rag D. a, extracts from cultures of S. cerevisiae colonies harboring two human genes, one is Rag A wild-type, T21L (gtr1-11) or the Q66L form in the pAS404 vector, and the other is Rag C or Rag D in the pACT2 vector, as shown in the figure were obtained. Their ␤-galactosidase activities were measured as described under "Experimental Procedures" and are shown as means of duplicate values with standard deviations. A ␤-galactosidase filter assay was also performed as described (41,42) and is shown on the right of the figure. Two independent colonies of each clone were picked up. The experiment was done at least twice, and representative results are shown. nt, not tested. Ectopically Expressed Rag C and Rag D Proteins Are Associated with Rag A and Rag B in BHK21 Cells-To ascertain whether human Rag C or Rag D are associated with human Rag A in mammalian cells, HA tag Rag C or HA tag Rag D plasmids were cotransfected into hamster BHK21 cells with either T7-tag Rag A or T7-tag Rag B by LipofectAMINE methods (Fig. 3, upper, middle and lower panels, lanes 1-4). The cell lysate from transfected cells was prepared, and T7-Rag A and T7-Rag B proteins were immunoprecipitated by an anti-T7 tag antibody. Immunoprecipitates were then run on SDS-poly-acrylamide gel, transferred onto a nylon filter, and blotted with an anti-HA tag antibody to detect HA-Rag C or HA-Rag D. As shown in Fig. 3, upper panel, lanes 1-4, ectopically expressed Rag C and Rag D were associated with Rag A and Rag B in BHK21 cells. As controls, BHK21 cells were transfected with each of T7-Rag A, T7-Rag B, HA-Rag C, and HA-Rag D plasmids (Fig. 3, upper, middle, and lower panels, lanes 5-8). When the cell lysates were immunoprecipitated with an anti-T7 antibody, none of the control lanes detected Rag C or Rag D (Fig.  3, upper panel, lanes 5-8). Similar amounts of Rag A and Rag FIG. 5. Identification of the interaction region of Rag A with Rag C and Rag D. a, serial deletion clones of Rag A were constructed in the pAS1 vector as described under "Experimental Procedures." The resultant plasmids were transfected into Y190 harboring pACT2-Rag C. Extracts from Y190 harboring serial deletion clones of Rag A and Rag C were prepared as in Fig. 4 and were processed to measure their ␤-galactosidase activities. The data are shown on the right of the figure as means of duplicate values with standard deviations. As a control, ␤-galactosidase activities of the extract prepared from Y190 harboring pAS404-Rag A and pACT2 are shown in parentheses. All experiments were done at least twice. A schematic view of nucleotide-binding and Rag C-binding regions is shown at the top of the figure. A possible coiled-coil region and the leucine zipper region are shown. The prediction of a coiled-coil region was performed using the COILS program (49) and MacStripe 2.0b1. b, serial deletion clones of Rag C were constructed in the pACT2 vector. The resultant plasmids were transfected into Y190 harboring pAS404-Rag A. The representative ␤-galactosidase activities of extracts of the Y190 strains are shown on the right of the figure with standard deviations. c, the C-terminal region of Rag D corresponding to the minimal binding region of Rag C in pACT2 was transfected into Y190 harboring pAS404-Rag A. The representative ␤-galactosidase activities of extracts of the Y190 strains are shown on the right of the figure with standard deviations. d, in vitro association of C-terminal portions of Rag A and Rag C with Rag C and Rag A, respectively. C-terminal portions of Rag A and Rag C proteins were synthesized with [ 35 S]methionine in vitro as described under "Experimental Procedures." Either GST (20 g), GST-Rag A (20 g), or GST-Rag C (20 g), which were bound to the glutathione-Sepharose 4B beads, were mixed with the 35 S-labeled proteins. After incubation at 4°C for 30 min, the beads were spun down and washed. Bound proteins were run on SDS-polyacrylamide gel and analyzed using the Fuji Image Analyzer. Lane 1, in vitro synthesized C-terminal portion of Rag C (230 -399 amino acids residue (a.a.)); lane 2, in vitro synthesized C-terminal portion of Rag A (161-308 a.a.); lane 3, GST that was bound to glutathione-Sepharose 4B beads (GST-beads) was mixed with in vitro synthesized Rag C (230 -399 a.a.); lane 4, GST-Rag A that was bound to glutathione-Sepharose 4B beads was mixed with in vitro synthesized Rag C (230 -399 a.a.); lane 5, GST-beads were mixed with in vitro synthesized Rag A (161-308 a.a.); lane 6, GST-Rag C that was bound to glutathione-Sepharose 4B beads was mixed with in vitro synthesized Rag A (161-308 a.a.) B were detected in the immunoprecipitates from the lysate of transfected cells (Fig. 3, middle panel, lanes 1-6). Similar amounts of Rag C and Rag D were also detected (Fig. 3, lower  panel, lanes 1-4, and 7 and 8). No protein band was detected in the immunoprecipitates from the lysate of BHK21 cells (Fig. 3, upper, middle, and lower panels, lane 9).
It is known that Ras-like small G proteins adopt two different conformations, GDP form and GTP form. We then examined whether Rag C and Rag D can interact with the T21L mutant (GDP form) or Q66L mutant (GTP form) of Rag A. The T21L mutant of Rag A (mutation at 21st threonine to leucine), corresponding to a dominant negative form of Ran, T24N (GDP form), and the Q66L mutant of Rag A (mutation at 66th glutamine to leucine), corresponding to a dominant positive form of Ran, Q69L (GTP form), (33,34) in pAS404 were used to transform the Y190 strain harboring Rag C or Rag D in the pACT2 vector. The control Y190 strain harboring pAS404-Rag A and pACT2 showed fewer than 6.0 ␤-galactosidase units. Thus, when more than 10 units of ␤-galactosidase activities were obtained, we assumed that Rag A had interacted with the target proteins in S. cerevisiae. We observed more than 100 galactosidase units in pAS404-Rag A-pACT2-Rag C and pAS404-Rag A-pACT2-Rag D pairs (Fig. 4a, clones 1 and 4), which indicated that these interactions were very strong. In addition to wild-type Rag A, both mutants (T21L and Q66L) of Rag A were bound to Rag C and Rag D to a similar degree (Fig.  4a, clones 2, 3, 5, and 6).
To confirm further that the T21L and Q66L mutants of Rag A could form a complex with Rag C or Rag D, HA-Rag C or HA-Rag D were cotransfected into BHK21 cells with T7-Rag A, T7-Rag A gtr1-11 , or T7-Rag A Q66L plasmids. HA-Rag C and HA-Rag D were immunoprecipitated with the anti-HA tag antibody and blotted with the anti T7-tag antibody to detect various forms of Rag A. As shown in Fig. 4b (lanes 1-6), Rag C and Rag D were associated with a wild-type and also with the gtr1-11 (T21L) (GDP) and Q66L (GTP) forms of Rag A in BHK21 cells. As a control, T7-Rag A Q66L alone was transfected into BHK21 cells. Rag A was not immunoprecipitated from the cell lysate lacking Rag C or Rag D (Fig. 4b, lane 7).
C-terminal Regions of Rag A and Rag C Are Mutual Binding Regions-Rag C has a long C-terminal region where the potential leucine zipper motif and the coiled-coil structure were found, as described above. Since the leucine zipper motif and the coiled-coil structure are implicated in the protein-protein interaction (reviewed in Ref. 50), it is possible that interaction between Rag A and Rag C occurs within their C-terminal region. To confirm this possibility, a set of deletion cDNA clones of Rag A in pAS1 and of Rag C in pACT2 was constructed using PCR amplification, and these were then used to transform the Y190 clone expressing either whole Rag C in pACT2 or whole Rag A in pAS404, respectively. A ␤-galactosidase assay of extracts from yeast Y190 clones was performed to ascertain whether the long C-terminal regions of Rag A and Rag C were involved in mutual interaction. The Y190 clone harboring the region encompassing 161-308 amino acid residues of Rag A and whole Rag C had 1903 galactosidase units (Fig. 5a). Thus, this C-terminal region of Rag A could be the minimal region responsible for the association with Rag C. The Y190 clone harboring the C-terminal region of Rag C (230 -350 amino acids residues) in pACT2 and whole Rag A in pAS404 had 2746 galactosidase units. Thus, this C-terminal region of Rag C, which contained a leucine zipper motif and a coiled-coil structure, could be the minimal region responsible for the association with Rag A (Fig. 5b). When we compared the amino acid sequences ranging from 230 to 350 amino acids between Rag C and Rag D, 92% amino acid identity was found, thus suggesting that Rag D was also associated with Rag A in this C-terminal region. As expected, the Y190 clone harboring the C-terminal region of Rag D (230 -350 amino acids residues) and whole Rag A in pAS404 had 200 galactosidase units. This C-terminal region of Rag D was shown to be responsible for the association with Rag A (Fig. 5c).
To confirm further that the C-terminal regions of Rag A and Rag C are responsible for mutual binding, the C-terminal portions of Rag C (230 -399 a.a.) and Rag A (161-308 a.a.) were synthesized in vitro using the rabbit reticulocyte lysate system (Fig. 5d, lanes 1 and 2). Next, either GST, GST-Rag A, or GST-Rag C, which were bound to the glutathione-Sepharose 4B beads, were mixed with the 35 S-labeled proteins. GST-Rag C efficiently pulled down the in vitro synthesized C-terminal portion of Rag A protein (Fig. 5d, lane 4). GST-Rag A efficiently pulled down the in vitro synthesized Cterminal portion of Rag C (Fig. 5d, lane 6). Thus, the Cterminal portion of Rag C or Rag A was shown to be responsible for mutual binding. Control GST was not able to pull down these proteins (Fig. 5d, lanes 3 and 5).
The above results imply that Rag A and Rag C have a nucleotide-binding domain in their N-terminal regions and a mutual binding domain in their C-terminal regions, which is unique in the Rag A/C subfamily among the Ras-like small G proteins.
Subcellular Localization of Rag C and Rag D Is Influenced by the Localization of Rag A-We next examined the subcellular localization of Rag C and Rag D by the ectopic expression of HA-or EGFP-Rag C and Rag D in the BHK21 cells and sub- sequently by confocal microscopic observations (Fig. 6, a and b). When transfected cells were fixed with methanol/acetone (1:1), Rag C was found to be localized in both the cytoplasm and the nucleus (Fig. 6a), although the cytoplasmic staining was stronger than the nuclear staining, and these findings were similar to those for Rag A staining (33). Rag D was mainly located in the cytoplasm, with some in the nuclear speckles. To confirm further their subcellular localization, we employed EGFP fusion constructs of Rag C and Rag D to allow us to observe these proteins in live cells by confocal microscopy. Rag C and Rag D were consistently found to be located in both the cytoplasm and the nucleus, although cytoplasmic staining was stronger than nuclear staining (Fig. 6b).
In Figs. 4 -6, Rag A was shown to be associated with Rag C and Rag D. It is thus natural to assume that Rag A is colocalized with Rag C and Rag D in mammalian cells. When we cotransfected Rag A with either Rag C or Rag D in BHK21 cells, Rag C was shown to be colocalized with the wild-type and with the T21L (GDP) and Q66L (GTP) forms of Rag A in BHK21 cells (Fig. 7a, panels 3, 6, and 9). Rag C was localized in both the nucleus and the cytoplasm when coexpressed with the wild-type and the Q66L form of Rag A. We previously showed that the T21L form of Rag A was distributed in discrete nuclear speckles, being localized side by side with SC35 (33), as shown in Fig. 7a, panel 4. Interestingly, Rag C was found in the same nuclear speckles when coexpressed with the T21L form of Rag A (Fig. 7a, panels 5 and 6). Rag C changed its subcellular localization depending on the changes in Rag A localization. When Rag D was coexpressed with the wild-type and the Q66L form of Rag A, Rag D was colocalized in the nucleus and the cytoplasm with Rag A (Fig. 7b, panels 1-3, and 7-9). When Rag D was coexpressed with the T21L form of Rag A, Rag D was colocalized with the T21L form of Rag A (Fig. 7b, panels 4 -6), thus showing that Rag D also changed its localization dependent upon the changes in Rag A localization.
Gtr2p, as shown in Fig. 1a. To ascertain whether Gtr1p also influences Gtr2p localization in S. cerevisiae, the localization of Myc-tagged Gtr2p driven by its own GTR2 promoter in the yeast cells was studied in wild-type cells and GTR1 disruptant cells through the immunostaining of cells with an anti-Myc antibody. Although Gtr2p is localized in both the cytoplasm and the nucleus (34) (Fig. 8, panel 4), disruption of GTR1 resulted in the accumulation of Gtr2p within the nucleus (Fig. 8, panel 7). Thus, the subcellular localization of Gtr2p also changed through the removal of Gtr1p, implying that Gtr1p influences Gtr2p with regard to its subcellular localization.
The location of the nucleus was demonstrated by staining the cells with DAPI (Fig. 8, panels 2, 5, and 8). Phase contrast images of yeast cells are shown in Fig. 8, panels 3, 6, and 9. As a control, cells without Myc-tagged Gtr2p did not show any staining (Fig. 8, panel 1). DISCUSSION We herein demonstrated that the G protein, Rag A, is associated with the novel G proteins, Rag C and Rag D. Human Rag C bound to guanine nucleotides efficiently and specifically, demonstrating that Rag C is in fact a G protein.
Ras-like small G proteins are present as monomeric proteins by nature, although Ras-like small G proteins transiently interact with various effectors. Rag A-Rag C association is the first example of a Ras-like small G protein forming a stable heterodimer. It is well known that stable heterodimers and trimers are present in other members of the G protein family, such as trimeric G proteins and ␣and ␤-tubulin, which form trimers and dimers, respectively. Trimeric G proteins act as switches that regulate information-processing circuits (reviewed in Ref. 51). G␣ is a GTP-binding protein in trimeric G protein and is associated with the non-G proteins, G␤ and G␥, whereas ␣and ␤-tubulin are both G proteins. It appears that ␣and ␤-tubulin GTPases have separate functions; the ␣-tubulin GTPase is important for heterodimer formation, whereas the ␤-tubulin GTPase is important for microtubule assembly. Although the ␤-tubulin-bound GTP is rapidly hydrolyzed to GDP after microtubule assembly, the ␣-tubulin-bound GTP is nonexchangeable and is never hydrolyzed (reviewed in Ref. 52), due to strong association of the effector region of ␣-tubulin with ␤-tubulin. The association between Rag A and Rag C occurred in the C-terminal regions that were downstream of the effector region.
In S. cerevisiae, Gtr1p and Gtr2p are associated with each other (34) and are homologous to Rag A and Rag B and to Rag C and Rag D, respectively. To date, homologous genes in Caenorhabditis elegans are reported in each of GTR1 (GenBank TM accession number CET24F1-1) and GTR2 (CEY24A-a, probably incomplete due to the lack of a C-terminal region). As a result, GTR1 and GTR2 DNA may become duplicated in higher eukaryotes during evolution. Schurmann et al. (37) suggested that Rag A was a member of the superfamily of Ras-like G proteins, which did not belong to any of the previously known subfamilies, but which did represent a sixth subfamily conferring a unique specialized function. Moreover, Rag A/Gtr1p and Gtr2p were shown to belong to a novel subfamily of small G proteins by phylogenetic tree analysis (34). It appears very likely that Rag C and Rag D are also members of the Rag A/Gtr1p subfamily, because G2 and G3 motifs that are different from other G proteins are highly conserved between Rag The cells were fixed with 4% paraformaldehyde solution and immunostained with the anti-Myc antibody (panels 1, 4, and 7) as described (32). DNA was stained with DAPI (panels 2, 5, and 8). The phase images of each strain are shown in panels 3, 6, and 9. The upper panels show a parental wild-type NBW5 strain (panels 1-3). The middle panels show a GTR2 disruptant (gtr2-1⌬) harboring MYC-GTR2 (panels 4 -6). The lower panels show a GTR1 and GTR2 double disruptant (gtr1-1⌬, gtr2-1⌬) harboring MYC-GTR2 (panels 7-9). Bar, 5 m.
In all G proteins, GTP is bound as a complex with Mg 2ϩ . Since many G proteins are unstable in the absence of bound nucleotides, nucleotide affinities are estimated from the rates of nucleotide dissociation (53). The off-rates of Ras are in the order of 10 Ϫ4 and 10 Ϫ5 s Ϫ1 for GDP and GTP, respectively, in the presence of Mg 2ϩ (54). Mg 2ϩ was also required for GTP␥S binding to Rag C. On the other hand, Mg 2ϩ was not required for GDP binding to Rag C. The off-rate of GDP from Rag C was in the order of 10 Ϫ2 s Ϫ1 , which was different from Ras but which was similar to those of the trimeric G proteins G i ␣, G o ␣, and G s ␣. Moreover, the GDP complexes of these trimeric G proteins have little affinity for Mg 2ϩ . As a result, Rag C seems to have a similar biochemical character to these trimeric G proteins. Rag C had intrinsic GTPase activity, although the GTPase activity was low. We were not able to measure the intrinsic rate of GTP hydrolysis correctly, because Rag C rapidly released GDP.
We have herein shown that Rag C and Rag A have a substantially higher sequence similarity to spGtr2p (63.3% identity) and spGtr1p (66.4% identity), respectively. As a result, the C-terminal region of spGtr2p can also be expected to be involved in the binding to spGtr1p. The Y190 strain harboring the spGTR2 fragment ranging from 161 to 314 amino acids in pACT2 and the spGTR1 fragment in pAS1 was constructed. As expected, the strain was able to grow on a selection plate (SD, ϪLeu, ϪTrp, ϪHis, ϩ3-aminotriazole) (data not shown), thus demonstrating that the C-terminal region of spGtr2p is responsible for binding to spGtr1p.
Although Gtr1p also interacted with both Gtr2p and Gtr1p (34), we were not able to detect any Rag A interaction with itself (data not shown). This is probably due to the weak homology between Rag A and Gtr1p in the C-terminal region (40% identity). Furthermore, the potential leucine zipper sequences of Rag A (VX 6 LX 6 IX 6 LX 6 L) are different from those of Gtr1p (MX 6 LX 6 MX 6 LX 6 L). Although Rag A seems to contain two coiled-coil structures in its C-terminal region, Gtr1p contains one coiled-coil structure. It is possible that these structural differences are the reason why Rag A does not form a homodimer.
Rag A changed its subcellular localization depending on the nucleotide-bound forms, T21L (GDP) and Q66L (GTP), and it seems probable that it shuttles between the nucleus and the cytoplasm (33). Whereas wild-type Rag A was soluble in the presence of 0.1% Nonidet P-40 in vivo, the T21L form of Rag A (GDP) was largely insoluble (data not shown), thus suggesting that a conformational change of Rag A resulted in a change in its biochemical character, which then influences its subcellular localization. As shown in Fig. 7, a and b, Rag C and Rag D changed their localization depending on the changes in Rag A localization. Moreover, the removal of Gtr1p resulted in the nuclear localization of Gtr2p (Fig. 8). Rag A/Gtr1p may influence Rag C and Rag D/Gtr2p regarding their subcellular localization, thus suggesting that Rag A/Gtr2p is associated with Rag C and Rag D/Gtr2p. The GTP concentration in the nucleus is about 30 times higher than the GDP concentration, which may result in the preferential production of the GTP form of Rag A within the nucleus. The GTP form of Rag A/Gtr1p is thought to have an ability to bind to and transport Rag C/Gtr2p from the nucleus to the cytoplasm through heterodimer formation.
We previously showed that the T21L mutant form of Rag A was partially colocalized with the splicing factor SC35 in the nuclear speckle (33). Since Rag C was colocalized with the T21L mutation of Rag A when coexpressed, the function of Rag C may thus be related with either transcription or splicing. The prp20/mtr1 mutation of RCC1 was suppressed by the gtr1-11 mutation (GDP form) of Rag A/Gtr1p. It should be noted that the RCC1-Ran system is involved in transcription/splicing (55). At present, we do not know the molecular mechanism behind the suppression of the prp20/mtr1 mutation of RCC1 by the gtr1-11 mutation of Rag A/Gtr1p. Disruption of GTR2 was able to rescue the prp20/mtr1 mutation as well as the gtr1-11 mutation (34). It is possible that an overexpression of the GDP form of Rag A/Gtr1p results in the nuclear accumulation of Rag C/Gtr2p. The inactivation of GTR2 either by a mutation of GTR1/Rag A or by gene disruption of GTR2 is necessary for the suppression of RCC1 mutation.