Cloning of a novel family of mammalian GTP-binding proteins (RagA, RagBs, RagB1) with remote similarity to the Ras-related GTPases.

cDNA clones of two novel Ras-related GTP-binding proteins (RagA and RagB) were isolated from rat and human cDNA libraries. Their deduced amino acid sequences comprise four of the six known conserved GTP-binding motifs (PM1, −2, −3, G1), the remaining two (G2, G3) being strikingly different from those of the Ras family, and an unusually large C-terminal domain (100 amino acids) presumably unrelated to GTP binding. RagA and RagB differ by seven conservative amino acid substitutions (98% identity), and by 33 additional residues at the N terminus of RagB. In addition, two isoforms of RagB (RagBs and RagB1) were found that differed only by an insertion of 28 codons between the GTP-binding motifs PM2 and PM3, apparently generated by alternative mRNA splicing. Polymerase chain reaction amplification with specific primers indicated that both long and short form of RagB transcripts were present in adrenal gland, thymus, spleen, and kidney, whereas in brain, only the long form RagB1 was detected. A long splicing variant of RagA was not detected. Recombinant glutathione S-transferase (GST) fusion proteins of RagA and RagBs bound large amounts of radiolabeled GTPγS in a specific and saturable manner. In contrast, GTPγS binding of GST-RagB1 hardly exceeded that of recombinant GST. GTPγS bound to recombinant RagA, and RagBs was rapidly exchangeable for GTP, whereas no intrinsic GTPase activity was detected. A multiple sequence alignment indicated that RagA and RagB cannot be assigned to any of the known subfamilies of Ras-related GTPases but exhibit a 52% identity with a yeast protein (Gtr1) presumably involved in phosphate transport and/or cell growth. It is suggested that RagA and RagB are the mammalian homologues of Gtr1 and that they represent a novel subfamily of Ras-homologous GTP binding proteins.

Ras-homologous GTPases constitute a large family of signal transducers that alternate between an activated, GTP-binding, and an inactivated, GDP-binding state (Hall, 1990;Bourne et al., 1990;Boguski and McCormick, 1993). These proteins rep-resent cellular switches that are operated by GTP-exchange factors and factors stimulating their intrinsic GTPase activity. To date, five subfamilies of the Ras superfamily are known: Ras, Rho, Rab, Ran, and ARF 1 proteins. These subfamilies are not only characterized by common structural features but also by a similar function, e.g. regulation of growth (Ras) (Egan and Weinberg, 1993), cytoskeleton organization (Rho) (Aktories et al., 1992), or vesicle transport (Rab and ARF) (Novick and Brennwald, 1993;Kahn et al., 1993). All GTPases of the Ras superfamily have in common the presence of six conserved motifs involved in GTP/GDP binding, three of them as phosphate/magnesium binding sites (PM1-PM3), and the other three as guanine nucleotide binding sites (G1-G3) (Valencia et al., 1991). Therefore, the sequences of the least related GTPases comprise approximately 20 -30% identical amino acids, whereas the sequence similarity is considerably higher within subfamilies (e.g. Ͼ40% identity in the Rab family, Ͼ50% in the ARF family) (Kahn et al., 1992).
The application of the PCR method has greatly facilitated the identification of novel cDNAs belonging to larger families of homologous genes, e.g. Ras-homologous GTPases (Chavrier et al., 1992;Clark et al., 1993). Constructing degenerate oligonucleotide primers from the conserved PM1 and PM3 domains, we have previously employed this approach to cloning of novel members of the ARF family Cavenagh et al., 1994). In addition to the known ARF and several unknown ARF-like isoforms, we isolated a single clone of a cDNA fragment that appeared to encode a GTP-binding protein, but it did not resemble any of the known subfamilies. Here we report the isolation and characterization of full-length cDNAs encoding two novel GTP-binding proteins (RagA and RagB) that, on the basis of structural criteria, constitute a novel subfamily of mammalian Ras-homologous GTPases. RagB is unique among small GTP-binding proteins in that its mRNA is expressed as two splicing variants differing by an insertion of 28 codons within the coding region. Interestingly, this insertion is located between the PM2 motif and the ␤ 2 strand, which are presumably involved in nucleotide or effector binding.

MATERIALS AND METHODS
PCR Cloning-A domain of 34-40 codons of Ras-homologous GTPases was amplified with degenerate oligonucleotide primers matching the GTPbinding domains PM1 (amino acid sequence LDAAGKT) and PM3 (WD-TAGQE) of the ARF family as described previously  with cDNA from murine 3T3-L1 adipocytes (American Type Culture Collection, Rockville, MD) as the template. The reaction products were separated on 2% agarose gels and were cloned into pUC18 (Sureclone-kit, Pharmacia Biotech Inc.). Plasmid DNA was isolated from 200 different clones and was characterized by sequencing or Southern blotting. * This work was supported by Grant Jo 117/9-1 from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) X85183, X85184, X90529, and X90530.
Library Screening and DNA Sequencing-PCR products subcloned into pUC18 were isolated with restriction enzymes EcoRI and BamHI and used as probes to screen a rat fat cell g11 cDNA library (RL1011b, Clontech, Palo Alto, CA), a -Zap rat brain cDNA library (937502, Stratagene, La Jolla, CA), and a human -Zap fetal brain library (937226, Stratagene). Positive plaques were isolated, and inserts were subcloned into pBluescript (Stratagene). Deletions were generated by exonuclease digestion in both directions, and inserts were sequenced by the method of Sanger (T7 sequencing kit, Pharmacia).
PCR Amplification of Fragments Comprising the Alternatively Spliced Domain of Rag-Oligonucleotide primers were constructed that specifically matched the sequence of RagA or RagB in two domains flanking the 84-bp insertion of RagB (RagA-specific primers, 5Ј-ATT GCG CGC GAC ACC AGG and 5Ј-T GTC CTG GCC GCC ACA G; RagB-specific primers, 5Ј-ATT GCC AGA GAC ACA CGT and 5Ј-T GTC TTG TCC ACC ACA A). Total RNA from rat adrenal gland, brain, thymus, kidney, and spleen was reverse transcribed and used as template. The PCR was allowed to proceed for 34 cycles of 94°C (80 s), 50°C (90 s), and 72°C (90 s). Products were separated on agarose gels, transferred on to nylon membranes, and probed with both RagA and RagB-specific probes. Parallel samples of the PCR products were isolated from the gel, subcloned, and sequenced. In a separate experiment, primers matching the sequence of RagB were used to amplify a large cDNA fragment (bp 685-1765) comprising the insertion, most of the open reading frame and 170 bp of the 3Ј-untranslated region (forward primer, 5Ј-TTG CCA GAG ACA CAC CGT-3Ј; reverse primer, 5Ј-CCT CAT CTT CTA ACT CC-3Ј). Two fragments (1 and 1.1 kb) were isolated, subcloned, and sequenced.
Northern Blot Analysis-Rat tissues (brain, heart, soleus muscle, adipose cells, liver, kidney, spleen, testis, adrenal gland, ovary, thymus; intestine, and lung) were homogenized with a Polytron homogenizer in 4 M guanidine thiocyanate. The lysates were layered on a cesium chloride cushion (5.88 M) and centrifuged at 33,000 rpm (rotor SW40) for 22 h at 20°C. Pelleted RNA was dissolved with 300 l of sodium acetate/ Tris buffer and was neutralized by the addition of 50 l of 2 M potassium acetate (pH 5.5). First-strand cDNA was synthesized with reverse transcriptase (First-strand cDNA synthesis-kit, Pharmacia) by oligo(dT) priming. Samples of total RNA (20 g) were separated by electrophoresis on 1% agarose gels containing 6% formaldehyde and transferred on to nylon membranes (Hybond, Amersham-Buchler, Braunschweig, Germany). Before transfer, gels were stained with ethidium bromide in order to ascertain that equal amounts of total RNA had been separated. Probes were generated with the Klenow fragment of DNA polymerase I and [ 32 P]dCTP by random oligonucleotide priming (Feinberg and Vogelstein, 1983). The nylon membranes were hybridized at 42°C, and blots were washed twice with 0.12 M NaCl, 0.012 M sodium citrate, 0.1% SDS and once with 0.015 M NaCl, 0.0015 M sodium citrate, 0.1% SDS at 55°C.
Expression of a Recombinant GST-Rag Fusion Protein in Escherichia coli-Fragments of the cDNAs of RagA and RagB l (RagA in pBluescript digested with BamHI/KpnI or NaeI/KpnI; RagB l , digested with HaeIII or NlaIV) comprising the complete reading frames were subcloned into the expression vector pGEX-2TK (Pharmacia). A RagB s cDNA was prepared by deletion of 84 bp from RagB l by site-directed mutagenesis (Kunkel et al., 1987) and was subcloned into pGEX. E. coli (DH5␣) was transformed with pGEX-RagA, pGEX-RagB s , or pGEX-RagB l , and exponentially growing cultures were induced with 0.1 mM isopropyl ␤-Dthiogalactoside and were grown overnight at 28°C. Cells were lysed and centrifuged at 12,000 ϫ g for 10 min. The supernatant was incubated with glutathione-Sepharose beads (Pharmacia) for 3 h at 4°C. The beads were washed with PBS (140 mM NaCl, 2.7 mM KCl, 10.1 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3), and the fusion proteins were eluted with buffer containing 10 mM glutathione. Samples of the fusion proteins were separated by HPLC on a Sephasil C-18 reversed phase column (4 ϫ 250 mm, Pharmacia) in order to identify the bound nucleotides as described previously (Tucker et al., 1986).
Assay of GTP Binding and Exchange-Guanine nucleotide binding to the GST-Rag fusion proteins was assayed by a previously described procedure (Northup et al., 1982). Samples of 10 g of fusion protein were incubated with tracer [ 35 S]GTP␥S (300,000 cpm/sample) in a buffer containing 20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 10 mM MgCl 2 , 1 mM dithiothreitol, and 0.1% (w/v) Triton X-100 in a total volume of 100 l. Unlabeled GTP␥S was added as indicated, and the binding was allowed to proceed at 30°C for 1 h or as otherwise indicated. The incubation was terminated by the addition of 1 ml of ice-cold buffer containing 20 mM Tris (pH 8.0), 100 mM NaCl, and 25 mM MgCl 2 . The samples were filtered through nitrocellulose membranes (Sartorius GmbH, Göttingen, Germany; pore size, 0.2 m) and washed 4 times with 1 ml of fresh buffer. Radioactivity on the filters was determined by scintillation counting in a water-compatible scintillation mixture (Ready Protein, Beckman). The exchange of GTP was assayed by loading of recombinant fusion proteins with [ 35 S]GTP␥S for 60 min at 30°C, and addition of GTP to a final concentration of 1 mM. Aliquot samples containing approximately 4 g of protein were separated at the desired time points by filtration as described above.
Assay of GTPase Activity-GTPase activity was assayed by a previously described procedure (Tan et al., 1991). Samples of recombinant proteins (25 g) in Tris buffer (50 mM, pH 8.0) containing 2 mM EDTA, 1 mM dithiothreitol, 10 mM MgCl 2 , 0.5 mM GTP, and 500 g/ml bovine serum albumin were incubated with 25 nM [␣-32 P]GTP (3000 Ci/mM) in a total volume of 50 l at 30°C for 30 min. Free tracer GTP was removed by ultrafiltration in Microcon-10 tubes (Amicon, Witten, Germany), and the samples were again incubated at 30°C. At the desired time points, bound tracer was isolated by filtration on nitrocellulose, and the nucleotides were eluted with formic acid (2 M). Nucleotides were separated by thin layer chromatography on polyethyleneimine cellulose (1 M lithium chloride, 1 M formic acid) and detected by autoradiography.

RESULTS
Cloning of RagA and RagB-A cDNA fragment showing remote similarity with other members of the Ras superfamily (see Fig. 1, underlined sequence) was previously isolated in a PCR-based cloning approach designed to characterize ARF isotypes in 3T3-L1 cells . This fragment was used as a probe to screen a -Zap library from rat brain for full-length cDNA clones. Three clones were isolated and characterized. Two of them (sg2-13 and sg2-14, later designated RagA) were identical (1.6 kb), whereas the third (sg2-15, later designated RagB l ) was different in size (2.5 kb), intensity of hybridization, and pattern of restriction fragmentation.
Sequence of RagA-The nucleotide sequence of the clones sg2-13/sg2-14 contained a poly(A) tail, an open reading frame encoding a protein of 313 amino acids (Fig. 1), and a domain that was nearly identical (one mismatch) to the PCR product that had been used for screening of the library (Fig. 1, underlined). The translation start was assigned to the first AUG codon of the clone. The N-terminal region of the deduced amino acid sequence contains several structural motifs that are conserved within the Ras family (Valencia et al., 1991): the phosphate/magnesium binding motifs PM1 (14-GKSGSGKT), PM2 (42-T), PM3 (61-WDCGGQ), and the guanine-nucleotide binding motif G1 (31-Y). An additional putative guanine nucleotide binding motif G2 (127-HKMD) is strikingly different from those found in all other Ras homologues (NKXD). Moreover, a conserved G3 motif (usually TSA or TCA) is not readily identifiable, the most likely candidates being 156-ECAC and 162-TSI. Finally, it should be noted that the C-terminal domain of the protein, which is probably unrelated to GTP-binding, is considerably larger than that of all other known mammalian members of the Ras family. This domain contains a stretch of hydrophobic amino acids (259 -276) that was identified as a putative membrane associated helix by the RAOARGOS program (Rao and Argos, 1986). However, the motif did not fulfil the requirements of the HELIXMEM program (Eisenberg et al., 1984).
A human cDNA of RagA was isolated from a fetal brain cDNA library, and its missing 5Ј-end (132 bp) was completed with the rapid amplification of cDNA ends procedure. The nucleotide sequence of the open reading frame was 92.7% identical (not shown), and the deduced amino acid sequence was 100% identical with that of the rat sequence. There was no significant homology in the 5Ј-untranslated region. Thus, it appears unlikely that the open reading frame extends beyond the presumed translation start.
Sequence of RagB-The nucleotide sequence of the clone sg2-15 (Fig. 2) contained a poly(A) tail and an open reading frame of 374 codons. A comparison with RagA revealed that the sequence of the clone (later designated RagB l ) was very similar (84% within the reading frame) but differed in an insertion of 28 codons within the reading frame between the nucleotide binding motifs PM2 and PM3 (underlined in Fig. 2, also see alignment in Fig. 3a). In order to test the possibility that two isoforms of the RagB mRNA, with and without the insertion, existed, a large portion of the cDNA including 170 bp of the 3Ј-untranslated region (bp 685-1765) was amplified by PCR with specific primers. With reverse-transcribed cDNA from testis as template, two cDNA fragments (1 and 1.1 kb) were isolated and subcloned. As anticipated, the sequence of the larger PCR product corresponded with that of RagB l as isolated from the library (100% identity). The sequence of the shorter fragment was identical except for the 84-bp insertion as depicted in Fig. 2. Thus, there are indeed two isoforms of RagB, RagB l and RagB s , differing by an insertion of 28 codons between the PM2 and PM3 motifs (see Fig. 5). Because the nucleotide sequences of the two mRNA species are identical except for the insert, we assume that they are derived from the same gene, presumably by alternative mRNA splicing.
Based on the comparison with RagA, the translation start of RagB might be assigned to nucleotide 572. However, there is an alternative start codon in RagB at nucleotide 473 that extends the open reading frame by 33 codons. Therefore, a human cDNA clone was isolated and sequenced in order to compare the presumed translation starts. This clone represented the short isoform RagB s (nucleotide sequence not shown); the sequence of the 84-bp insertion was obtained by PCR cloning. As illustrated in Fig. 3a, the open reading frame of the human sequence is homologous to that of the extended rat sequence starting at nucleotide 473, with 27 identical amino acids and five conservative substitutions in the N terminus. Furthermore, the human sequence lacks the second start codon of the rat sequence (methionine 34; Fig. 3a), which corresponds with the translation start of RagA. Thus, it appears reasonable to conclude that the N terminus of RagB is 33 amino acids longer than that of RagA.
Except for the extended N terminus and the insertion in RagB l , the deduced amino acid sequences of RagA and RagB were nearly identical (7 conservative substitution, 97.8% identity). Because of this high similarity, the two isoforms were designated A and B (instead of 1 and 2) according to previously established guidelines (Kahn et al., 1992). The identity of the nucleotide sequences was 83.8% within the coding region, but only 51% or less in the 3Ј-untranslated regions and before the translation start.
Homology of Rag with Other Proteins-A data base search was performed in order to determine the degree of homology of RagA/B with other G-proteins and to identify other proteins with structural similarities. The closest related protein was Gtr1 (51.6% sequence identity), a putative GTP-binding protein from yeast that is encoded by an open reading frame near the TUB3 locus on chromosome XIII (Bun-ya et al., 1992). Fig.  3b depicts the alignment of the sequences. Their highest similarity is in the N-terminal half, which comprises the domains responsible for GTP-binding. Both sequences lack a readily identifiable G3 domain; based on the assumption that this domain, if present, should be identical in Rag and Gtr1, the best candidate would be the 162-TSI motif. Except for two isolated motifs, the similarity of the two proteins within the C-terminal domain is low. Thus, it might be predicted that the GTP-binding characteristics of Rag and Gtr1 are similar, but that there are profound differences in the functions conferred by the C termini.
The alignment of the first 200 amino acids of RagA/B with prototypes of the five subfamilies of Ras-homologous GTPases revealed that the similarity was low (17.5% identity with Ras, 17.6% with Rho1, 18% with Rab1, 17.5% with Ran, 21% with ARF1, 17% with ARL1; PALIGN program). However, as is illustrated by an alignment of RagB with Ras (Fig. 3c), the GTP-binding domain of Rag is homologous to Ras in order, distance, and structure of the conserved motifs of GTP binding. Based on this similarity, Rag can be assigned to the superfam-FIG. 1. Nucleotide and deduced amino acid sequence of rat RagA. The deduced amino acid sequence is shown in single letter code above the respective codons. The translation start was assigned to the first AUG codon of the clone. The sequence of the PCR product that was isolated from 3T3-L1 cells (mouse) and used for screening of the cDNA libraries is underlined; a single nucleotide differing from the rat sequence (321C 3 T) is shown below. The domains presumably involved in GTP binding are boxed.
ily of Ras-related GTPases. The alignment also illustrates the position of the insertion in RagB l , which disrupts the PM2 motif and the ␤ 2 strand.
The low structural similarity of Rag with other members of the Ras superfamily is illustrated by a multiple alignment and the resulting tree (Fig. 4). On the basis of this comparison, it is evident that the Rag isoforms cannot be assigned to any of the known subfamilies. Thus, the proteins appear to represent a novel subfamily of the GTP-binding proteins; we designated them Rag in order to emphasize both their unique structure and the remote similarity with the Ras, Rab, and Ran families.
Tissue Distribution of RagA and RagB Isoforms-In order to investigate the tissue distribution of the two isoforms of RagB, and to test the possibility that an additional isoform of RagA existed, PCR primers were constructed for amplification of the domain between PM1 and PM3; these primers were specific for either RagA or RagB. cDNA from various tissues (thymus, adrenal gland, spleen, brain, and kidney) was used as template. The products were separated, blotted onto nylon membranes, and hybridized with specific probes. As is illustrated in Fig. 5, a single product hybridizing with the specific probe was obtained with RagA-specific primers; its size corresponded with the product anticipated from the sequence (Fig. 1). With primers specific for RagB, two hybridizing products (109 and 193 bp) were obtained from thymus, adrenal gland, spleen, and kidney. In contrast, brain cDNA yielded exclusively the long PCR product. The identity of the PCR products was ascertained by cloning and sequencing; they corresponded with the sequence of RagB (Fig. 2) and differed indeed in the underlined insertion of 28 codons. Fig. 6, upper panel, illustrates a Northern blot of several rat tissues probed with a cDNA fragment of RagA. A single transcript with an approximate size of 1.8 kb was identified in most tissues with highest levels in adrenal gland; lower levels of mRNA were also detected in brain, skeletal muscle, fat cells, liver, spleen, testis, ovary, thymus, and lung. When a full-length cDNA was used as the probe (data not shown), additional, presumably nonspecific bands appeared; the pattern of tissue distribution of the 1.8-kb band was identical to that shown in Fig. 6. Three different probes derived from the RagB l cDNA were employed in the Northern blot analysis. Two of these probes (full-length cDNA and 3Ј-untranslated region) failed to detect any specific signal. With a cDNA fragment comprising a portion of the coding region, two very weakly hybridizing transcripts (2.5 and 3.8 kb) were detected against the background in brain, testis, adrenal gland, and thymus (Fig. 6, lower panel). Thus, the mRNA of RagB appeared much less abundant than that of RagA, and exhibited a somewhat different tissue distribution. GTP Binding of Recombinant RagA and RagB-In order to test whether the novel G-proteins do indeed bind GTP, both cDNAs (RagA and RagB l ) were subcloned into the expression vector pGEX. In addition, a vector comprising the RagB s cDNA was prepared from the cDNA of RagB l (clone sg2-15) by sitedirected mutagenesis. The resulting constructs encoding fusion proteins of GST and the respective Rag isoform were expressed in E. coli, and the recombinant proteins were purified by affinity adsorption on glutathione-Sepharose. HPLC separation of nucleotides bound to the native fusion proteins revealed that recombinant RagA and RagB s were loaded with GTP/GDP to a ratio of 4:1, whereas no bound nucleotides were detected on RagB l (data not shown). Binding of GTP was then assayed with [ 35 S]GTP␥S, and bound GTP was separated by filtration on nitrocellulose. As is illustrated in Fig. 7, upper panel, specific binding of tracer GTP␥S to the recombinant RagA and RagB s was measurable within 5 min after addition of the ligand, and approached an equilibrium that largely exceeded that of the GST-control. In contrast, binding of GTP␥S to recombinant RagB l was considerably lower than that of RagA and RagB s , hardly exceeding that of the GST control. This lack of specific GTP binding of RagB l was also observed with a second con- open gap cost, 10; unit gap cost, 10). Similarities were calculated from the matrix of the pairwise similarity scores. References for the known sequences: Ras, (Ruta et al., 1986); Rho, (Chavrier et al., 1990); Rab1, -2, and -4 (Touchot et al., 1987); Rab6, (Zahraoui et al., 1989); Ran, (Drivas et al., 1990); ARF1, (Bobak et al., 1989); ARL1, 4, ; Gtr1, (Bun-ya et al., 1992) all Rag isoforms are denoted by asterisks; amino acids identical in the human and rat RagB are designated by periods. Note that the amino acid sequences of human and rat RagA are 100% identical and that only the long isoforms of RagB (RagBs) are given. b, sequence alignment of RagA/B s with the yeast protein Gtr1. Gtr1 (Bun-ya et al., 1992) was identified as the closest relative of Rag by a data bank search (Swissprot, EMBL), and the deduced amino acid sequences of Rag and Gtr1 were aligned with the aid of the PALIGN program. Hyphens represent gaps introduced for optimal alignment (open gap cost, 10; unit gap cost, 2). Amino acid residues differing in RagB are given on top of the RagA sequence. The putative consensus domains for phosphate/magnesium binding (PM1-3) and guanine nucleotide binding (G1-3) are depicted on the top of the alignment. Identical amino acids are denoted by vertical lines. c, alignment of the GTP-binding domain (amino acids 34 -261) of RagB with Ras. The putative GTP-binding motif and the position and sequence of the 28-amino acid insertion in RagB l is given on top of the alignment. Identical amino acids are highlighted by vertical lines; conservative substitutions are depicted by colons. struct that differed in the fusion site at the N terminus (data not shown).
The lower panel of Fig. 7 illustrates experiments designed to study the nucleotide exchange on RagA and RagB. Fusion proteins were loaded with GTP␥S for 60 min, and nucleotide exchange was started by the addition of buffer with or without added GTP. On both fusion proteins, bound tracer GTP␥S decreased rapidly after the addition of unlabeled GTP. Thus, RagA and RagB s appear to exchange bound GTP␥S in the absence of any catalyzing factor.
GTPase activity of Rag was tested by loading of the protein with [␣-32 P]GTP. After an additional incubation period of 0 -60 min, the bound tracer nucleotides were separated by thin-layer chromatography. During the 60-min incubation period, there was no detectable decrease in the amount of bound tracer GTP, nor any formation of GDP. Thus, both RagA and RagB s appear to lack an intrinsic GTPase activity in the absence of an activating factor.

DISCUSSION
RagA and RagB are novel GTP-binding proteins with the striking feature that one isotype, RagB, was found as two mRNA species differing by an insertion of 28 codons within the reading frame. Surprisingly, the nucleotide sequence of ragB does not appear to contain the consensus sequences for splicing in the region of the insertion. However, we have repeatedly found both isoforms of RagB by PCR with cDNA from rat as well as from human tissues. The identity of the PCR products (109/193 bp and 996/1080 bp) was confirmed by sequencing; their sequences including portions of the 3Ј-untranslated region were identical except for the 84-bp insertion. Thus, it appears reasonable to assume that the two isoforms were derived from the same gene by alternative mRNA splicing. The splicing site near the PM2 domain of Rag (threonine 42) is not unexpected, because it corresponds exactly with well conserved exon borders of ras (McGrath et al., 1983) and ARF-2 (Serventi et al., 1993), and is only 12 codons downstream of the border between exon 2 and 3 in Rab isotypes (Wichmann et al., 1989).
Specific and saturable binding was observed with recombinant GST-fusion proteins of RagA. In contrast, preparations of GST-RagB l bound only minute amounts of GTP, which hardly exceeded those bound by GST preparations, but deletion of the 84-bp insert from the inactive RagB l generated a fully active, GTP-binding fusion protein of RagB s . Thus, it is concluded that the large difference in GTP binding between RagB s and RagB l is due to the insertion of 28 amino acids between PM2 and the ␤ 2 strand. Because the insertion might affect the orientation of PM2 (threonine 42), it appears plausible that it affects nucleotide binding, exchange, and/or hydrolysis. Furthermore, it was recently shown that the ␤ 2 strand in the Ras homologue Rap is involved in binding of the effector molecule Raf (Nassar et al., 1995). However, the possibility cannot be excluded that RagB l , hollow squares) were prepared as described, and samples (2 g of protein) were incubated with tracer GTP␥S. The bound tracer was assayed at the indicated times after separation by filtration on nitrocellulose membranes. Panels C and D, samples of the fusion proteins of RagA and RagB s were loaded with tracer GTP␥S for 60 min. Nucleotide exchange was initiated by the addition of buffer with (filled triangles) or without (hollow triangles) GTP (final concentration, 1 mM), and bound tracer was assayed at the indicated time points. the RagB l protein is not properly folded in the bacterial expression system or that it requires the extended N terminus for proper folding and GTP binding. Thus, although it is tempting to speculate that the 28-amino acid insertion may produce an important functional alteration, the exact nature of this alteration remains to be determined.
RagA and B differ slightly in the tissue distribution of their mRNA. According to the Northern blot analysis, ragA was expressed in most tissues, but it was most abundant in adrenal gland, ovary, and testis. Rag B appeared most abundant in brain and testis, but mRNA levels were much lower than those of RagA. Like RagA, RagB isoforms could be detected by PCR amplification in all tissues investigated. However, the alternative splicing of ragB was tissue-specific in that brain expressed only its long form. This tissue specificity is an indication that the alternatively spliced insertion in RagB may confer a functional modification of the GTP-binding protein.
Several structural criteria have previously been established for entry of a novel protein into the Ras superfamily: a size of 20 -29 kDa, the presence of the consensus motifs for GTP binding, the presence of other sequence motifs characteristic for one of the subfamilies, and an at least 30% identity with other members of the family (Valencia et al., 1991;Kahn et al., 1992). The novel gene product Rag comprises four of the known GTP-binding motifs (PM1-PM3 and G1). In addition, a G2 domain (HKMD) appears to be present but differs from that found in all other Ras homologues (NKXD) by a substitution of asparagine for histidine. A potential G3 domain, tentatively assigned to the TSI motif is unrelated to those found in other isotypes. It should be noted that G3 is the domain least conserved among other subfamilies and that it appears to be only indirectly involved in GTP binding (Valencia et al., 1991). The distances between the GTP-binding domains in Rag were essentially identical to those in other members of the Ras superfamily. Because of the presence of the GTP-binding domains in conserved order and distances, it appears reasonable to assume that the tertiary structure of the main part of Rag providing the interaction with GTP/GDP (amino acids 1-200) is similar to that of Ras (Pai et al., 1990) and ARF (Amor et al., 1994) and that its basic function is that of a GTP/GDP dependent switch. However, there are several striking differences to other members of the Ras-family. First, Rag is considerably bigger than most other known mammalian Ras homologues (calculated molecular mass 36.6 kDa (RagA), 40.2 kDa (RagB s ), or 43.2 kDa (RagB l )). This extraordinary size is due to an additional domain of approximately 100 amino acids at the C terminus; similar additional domains have been found in the yeast counterparts of Ras (Powers et al., 1984;Dhar et al., 1984) and in the ARF-related GTPase ARD (Mishima et al., 1993). Second, other characteristic motifs, e.g. the lipid modification motifs commonly encountered in Ras homologues except in Ran are absent. Finally, the overall amino acid identity of Rag to other members of the Ras family is low (highest to ARF1, 21%) and is restricted to the GTP binding motifs. In spite of these differences, we suggest that Rag is assigned to the superfamily of Ras-homologous GTPases. It is obvious, however, that the protein does not belong to any of the known subfamilies and may thus represent a sixth subfamily conferring a unique specialized function.
The deduced amino acid sequence of Rag is similar (51% identity) to that of a gene from yeast (GTR1), which has been identified as an open reading frame near the TUB3 locus on chromosome XIII (Bun-ya et al., 1992). The gene is located in close proximity (1.1 kb) of a phosphate transporter (PHO84), and its disruption produced phenotypes resembling those of PHO84 mutants. Thus, it was concluded that the function of the Gtr1 protein is related to that of the phosphate transporter PHO84. Furthermore, disruption of GTR1 resulted in moderately reduced growth at lower temperature. It should be noted, however, that our present findings raise the possibility that an isoform of GTR1 similar to ragB might have in part compensated for the disruption of GTR1. Because of the similarities between rag and GTR1, the possibility has to be considered that rag is the mammalian counterpart of GTR1. Further studies might therefore focus on an involvement of Rag in transport of phosphate or other ions. However, there are some distinct structural differences between Rag and Gtr1. Whereas the identity in the overall GTP-binding domain is 60%, the Cterminal 100 amino acids, which are probably not involved in GTP-binding show only a 35% identity. In contrast, the similarity between other yeast and mammalian GTPases with closely identical related functions is higher, e.g. 75% identity in the case of Ypt /Rab1 (Zahraoui et al., 1989). Thus, Gtr1 and Rag could have identical GTP-binding characteristics, but their functions might have diverged during evolution.