rgr Oncogene: Activation by Elimination of Translational Controls and Mislocalization¹

Small GTPases function as critical relays in the transduction of extra- and intracellular signals by cycling between inactive GDP and active GTP-bound states. The regulation of the GTPases is achieved through the opposing effects of GEFs⁵ and GTPase-activating proteins. Guanine nucleotide dissociation inhibitors also play a role in mediating the control of GTPases. The prototype of the small GTPases is the Ras subfamily, which includes Ras, R-Ras, Ral (A and B), Rap, and TC21 (see Ref. 1 for review). RAS proteins are activated essentially upon ligand engagement of membrane receptors, and activated Ras can switch on the differentiation, proliferation, senescence, and apoptosis pathways. These pleiotropic effects rely on the ability of RAS-GTP to associate with a number of effectors. RAL proteins are downstream elements in RAS-mediated signaling (1), due to the ability of RAS-GTP to recruit RAL-GEF to the plasma membrane.

In addition to the RAS-dependent event, RAL activation can also be achieved by RAS-independent mechanisms (2). RAL has been localized both at the plasma membrane and in the membranes of endocytic and exocytic vesicles (3). The identification of new members of the RAL-GEF family that lack the RAS-GTP interacting domain (4, 5, 6) strongly suggests that RAL proteins have two different sets of functions. One might be dependent on RAS activation, which would lead to the activation of the RAL molecules located in the plasma membrane through RAL-GEF and via the RAS-GTP interacting domain. The Ras-independent actions of Ral take place in the endosomes and are modulated by RAL-GEFs lacking the RAS-GTP-binding domain.

rgr is an oncogene that was isolated in our laboratory by its ability to produce tumors in the nude mice assay (4). This protein was part of the RSC fusion product derived from a 7,12-dimethylbenz(a)anthracene-induced rabbit squamous cell carcinoma. rgr appeared to be deleted at the 5′-end and fused to another gene, rabbit hr23A, the orthologue of rad23, a Saccharomyces cerevisiae nucleotide excision repair gene. Although the oncogenic form was isolated fused to another gene, only the rgr part is oncogenic (4). rgr belongs to the GEF family and has significant homology to RALGDS (7). Similar to RALGDS, RGR has been shown to have specific exchange activity for RAL, although unlike RALGDS, it lacks the RAS-interacting domain present in the COOH-terminal end of other family members (4). Recently, we have observed that the human orthologue of rgr, hrgr, is found to be frequently altered in a subset of human T-cell malignancies (8), providing an additional impetus to study this oncogene.

Previously, the signal transduction pathways induced by RGR were dissected in the processes of eliciting proliferation, cell transformation, and gene expression (9). In an attempt to elucidate the mechanism that RGR uses to elicit these cellular effects, the different properties of this oncogene were investigated. To test the possibility that structural differences in the oncogenic form of RGR were responsible for its tumorigenic properties, the first goal of this work was the identification and characterization of the wild-type form of the rabbit rgr gene. Because it was observed previously that RGR activates different GTPases, we also investigated the possibility that regions in RGR other than the guanine nucleotide exchange domain were involved in these effects. In addition, the RGR binding properties to different small GTPases were analyzed. Because an effective interaction of RGR with RAS in vitro was detected, we hypothesized that this interaction also may occur in vivo. Therefore, the subcellular localization of the wild-type and the oncogenic form of RGR was determined. These results have indicated that the transforming ability of rgr is a consequence of its overexpression, which, in turn, is due to the elimination of translational regulatory elements. This overexpression results in mislocalization of RGR and, subsequently, RAS activation.

Mouse monoclonal anti-FLAG antibody was obtained from Sigma, and the mouse monoclonal anti-RAL antibody was obtained from Transduction Laboratories. Rabbit polyclonal anti-p-MAPK and anti-MAPK antibodies were obtained from Promega. The guinea pig polyclonal anti-RGR antibody was generated as described previously (9). Oregon Green goat antirabbit antibody, BODIPY FL goat antimouse antibody, and Texas Red-conjugated donkey anti-guinea pig antibody were obtained from Molecular Probes.

RK13 cells, 293T cells, and the Phoenix cell line, a 293T cell-derived retroviral producer line, were grown in DMEM supplemented with 10% FCS, penicillin (50 units/ml), and streptomycin (50 μg/ml; Life Technologies, Inc.) at 37°C. NIH3T3 cells were grown in the same medium, except that it contained 10% calf serum.

Total RNA from different tissues and the RK13 cell line was extracted with Trizol (Life Technologies, Inc.), and poly(A)⁺ RNA was isolated using the Oligotex mRNA kit (Qiagen). The reverse transcription reaction was performed using the rgr-specific primer 1646–1621 (5′-GGACGACTCCCTTCTGTTGTTTCCGC-3′). The primers used to analyze the expression of endogenous rgr in different tissues were the 648–673 primer and the 1099–1074 primer (5′-TGCTGTGCTCATATGGTGACTTCAGC-3′ and 5′-AGGCAGCTCCTCTGCAGTCCCAGCCC-3′, respectively). The primers used to analyze expression in the transfected clones to assess the inhibitory effect of the 5′-UTR were RGR-F1, 907–884, β-F, and β-R (5′-ATGGCCTGGCCTAACATCCAT-3′, 5′-GTGCCTGGCTGCAGGCTCCGCAGG-3′, 5′-GTGGGCCGCTCTAGGCACCAA-3′, and 5′-CTCTTTGATGTCACGCACGATTTC-3′, respectively).

The cDNA cloning of the 5′-end of the rabbit rgr cDNA was accomplished using a 5′-RACE system and the Marathon cDNA Amplification Kit (Clontech), and the cDNA was obtained in the reverse transcription reaction from rabbit spleen mRNA (see above). The rgr-specific primers used for the first and the nested PCR reactions were primers 611–589 (5′-CTGGGACCAGGGATTCCTCCAGC-3′) and 579–556 (5′-GTGCCTGGCTGCAGGCTCCGCAGG-3′), respectively. The RACE products were cloned into pCR2.1 (Invitrogen) and sequenced.

For amplification of the complete rgr cDNA sequence, reverse transcription was performed with the rgr-specific primer 2041–2022 (5′-GGCCTCAGCTTCACAACAGG-3′). The first PCR was performed with primers 1–29 (5′-CCCTTCTCTTATCAGCACGGCAGAGCTGG-3′) and 2010–1986 (5′-CCCAGGCACCAGGCTCACAATGAGC-3′). In the nested PCR reaction, the primers used were 40–61 (5′-ATGGCCTGGCCTAACATCCATC-3′) and 1985–1959 (5′-CAGATCTGCCCCTGAGGTCCCAGGC-3′).

The full-length rgr cDNA was tagged with the FLAG sequence at the 3′-end by PCR, using primers 40–72 (mentioned above) and Rgr-FLAG (5′-TCACTTGTC GTCATCGTCTTTGTAGTCCATTGGCCTTGAGGCTGCTGTGGGCTC-3′). To construct the mutant F1ΔuORF-Rgr, the sequence that contains the first two ATGs was eliminated by PCR performed on the full-length rgr cDNA, using the forward primer 88–114 (5′-CAACCACACTCAGGGACGCAGAGTCAG-3′) and Rgr-FLAG as the reverse primer. The mutant F1→2-Rgr was created by amplifying the full-length rgr as two fragments. The first pair of primers were 40–72 and 566–536 (SalI; 5′-GGCTCCGCAGGGGTCGACCTCCGCCAGGTG-3′), and the pair of primers for the remaining rgr sequence were 536–566 (SalI; 5′-CACCTGGCGGAGGTCGA CCCCTGCGGAGCC-3′) as forward primer and Rgr-FLAG as reverse primer. This strategy, using the primer as reverse of the first fragment and its corresponding forward primer for the following fragment, introduced a SalI site and deleted a single bp. These two fragments of the full-length rgr were ligated to each other by the new SalI site. The construct 713-Rgr was generated using primers 703–727 (5′-CCATAGCAACATGGCTTCCTTCCTG-3′) and Rgr-FLAG. RSC-Rgr was obtained using a primer that introduces an ATG at the 5′-end of the fusion site to initiate the translation of the oncoprotein previously fused to the HR23A protein (forward primer RSCRalATG, 5′-ACTGCGAGATGCAGCTTCCATCTCTGCACTGCTGCT-3′) and Rgr-FLAG as the reverse primer. In all cases, the PCR products were subcloned into pCR2.1 vector (Invitrogen) and subsequently excised with EcoRI. They then were introduced into the EcoRI site of SP73 vector (Promega), downstream of the SP6 promoter for the in vitro transcription/translation coupled reaction. Some of these cDNAs were also subcloned into the EcoRI site of the retroviral vector MSCV-Puromycin for the transfection of 293T cells and posterior infection of NIH3T3 cells or into the EcoRI site of the previously described pMEXneo vector (9) to perform the focus formation assay. For the luciferase experiments, PCR amplification was performed using the forward primers for some of the constructs described above, and the common reverse primer was 731–703 (NcoI; 5′-GTCCCAGGAAGGAAGCCATGGTGCTATGG-3′). The PCR products were cloned into pCR2.1 vector, excised with HindIII and NcoI, and introduced into the TK.Luc.pGL3 vector (10) by using those two restriction sites. The stable stem loop (11) was inserted into the 5′-terminal end of the leader by using the HindIII site.

To delete NH₂-terminal regions of the oncogenic RGR, PCR reactions were performed using the 703–727 (see above) and ATG-1024–1049 (5′-CCCATGGCACAATCAGCACCAGGGC-3′) primers as forward primers for the 713-Rgr and the 1027-Rgr constructs, respectively. The PCR products were cloned into pCR2.1 vector, excised with EcoRI, and introduced into the MSCV-Puromycin retroviral vector.

The cDNAs with mutations inside the catalytic domain were generated using the Site-Directed Mutagenesis Kit (Stratagene), following the instructions of the manufacturer. The forward primers used to inactivate the RGR catalytic domain in four different constructs were as follows: 5′-CCAGGACGATGCCAGCAGAGCTCACACTTCTGGATGC-3′ (D1); 5′-GCCCGTGTGGTGGTGCCCGGGATCAGGTGGCC-3′ (D2A); 5′-GGAGTGCCAGACCCCCGGGCACGTTTCCTCGGCC-3′ (D2B), and 5′-CAGAAGGGAGTCGACCTCTGTCTTGGCACCTACCTG-3′ (D3).

Finally, rgr was amplified by PCR with the forward primer BamHI-Rgr (5′-GGATCCATCTCTGCACTGCTGCTCCC-3′) and fused in frame to the GFP tag using pEGFPC3 vector (Clontech).

Each cDNA was transcribed and translated using the in vitro TnT-coupled rabbit reticulocyte lysate system (Promega), following the recommendations of the manufacturer. Samples from the in vitro translation reactions were resolved on a 10% polyacrylamide gel. In addition, gels were treated with Intensify (DuPont) before autoradiography.

NIH3T3 and 293T cells were transfected by the calcium phosphate precipitation standard method (12). Clones exhibiting expression were selected and maintained in DMEM/10% calf serum supplemented with the selection antibiotic. When the Phoenix cells were transfected with retroviral vectors, the culture medium was used to infect NIH3T3 cells in the presence of Polybrene (5 μg/ml), and pools of antibiotic-selected infected cells were used to carry out further analyses. For the focus formation assay, cells were split 24 h after the transfection and maintained in DMEM plus 5% calf serum for 2 weeks.

The luciferase experiments were performed by transfecting 293T cells as described previously (9). Transfection efficiencies were normalized using the pRL-CMV vector (Promega) as an internal control.

Protein and RNA were extracted and subjected to Western blot or Northern blot analysis, respectively, as described previously (9).

The activation of RAL was detected essentially as described previously (9), using the GST-tagged form of the RALBP-RAL-binding domain (13, 14).

GST fusion proteins were expressed and purified as described previously (6). Briefly, bacteria were lysed by sonication, and the lysate was cleared by centrifugation and incubated with glutathione-agarose beads (Sigma). Beads were washed with 20 mm Tris-HCl (pH 8.0), 20% glycerol, 1 mm DTT, 50 mm NaCl, 1 mm phenylmethylsulfonyl fluoride, 5 mm MgCl₂, and 10 μm GDP.

The coimmunoprecipitation of epitope-tagged RGR with the small GTPase fusion proteins was performed as described previously (6). In brief, GTPase-bound beads were incubated with RAS buffer containing 10 mm EDTA for 10 min at room temperature. 293T cells were transfected with 10 μg of pCMV-FLAG-RGR by the calcium phosphate precipitation method. After 48 h, the cells were lysed, and the lysate was then added to the pre-equilibrated GTPase-bound beads. We always use 15 μg of GST-GTPase, so these assays are performed in excess of GST-GTPase, and therefore there is no need for a loading control for this molecule. This slurry was incubated for 2 h at 4°C, and then the beads were washed. The amount of Rgr in the pre-precipitate is one-tenth of the amount added to the experimental reaction. The fraction of RGR bound to the GTPases was detected with an anti-FLAG antibody (Sigma).

RK13 cells were plated into 4-well trays. The next day, they were fixed with methanol for 3.5 min at −20°C, blocked with 1% milk in PBS, and stained with the anti-RGR antibody followed by the TRITC-conjugated anti-guinea pig secondary antibody.

RGR was isolated as part of a fusion protein, RSC, between HR23A and RGR. Whereas the expression of the endogenous normal hr23A mRNA is easily detectable in different rabbit tissues, rgr mRNA expression was barely detected after hybridization of a RT-PCR blotted gel (4).

Therefore, the first goal was to find an adequate source of rgr mRNA for further analysis. Because the previous data indicated that this mRNA was poorly expressed, we needed to design a strategy to improve the probability of detecting the rgr mRNA. A RT-PCR reaction was performed on poly(A)⁺ RNA from different rabbit tissues with rgr-specific primers for both the reverse transcription and the PCR.

The primer used for the reverse transcription was designed to specifically use mRNA as a template, and not genomic DNA, because the primer straddles an intron. To further rule out the spurious amplification of genomic DNA, the primers for PCR were designed to distinguish between these two possibilities by the size of the generated band.

Following this approach, a band of around 450 bp was detected that was due to the amplification of cDNA in the lanes corresponding to spleen, testis, and thymus (Fig. 1). A band was also observed in the RK13 cell line, a rabbit kidney epithelial cell line. The sequencing of these bands corroborates that the rgr mRNA is present in these tissues and the cell line, although at very low levels. Although other bands of different molecular weights than expected could also be detected in a few of the sample tissues, the sequencing of these bands indicated that they were products spuriously amplified and unrelated to rgr.

As an initial step toward the characterization of the rabbit rgr gene, RACE on mRNA from rabbit spleen was performed to obtain the 5′ unknown region of the normal endogenous rgr cDNA. Using this approach, and after two rounds of amplification, two transcription start sites were identified. The longest isolated transcript had an additional 500 bp, extending the overall length of the 5′ leader sequence to 710 nucleotides (Fig. 2 A). In both transcripts, there is a segment of 209 bp flanked by sequences that are present in the 5′-end of the oncogenic form of rgr. This result indicates that during the fusion process, abnormal splicing of this 209-bp sequence occurred.

Once the complete sequence of rgr was known, two rounds of PCR were performed to confirm the existence of the mRNA. By using forward primers that encompass the 5′-end of the longest transcript and reverse primers that match sequences of the 3′-UTR of the oncogenic form, a PCR product of around 2 kb was amplified, cloned, and sequenced. In this sequence, the stop codons that were present in the oncogenic form were also present. This finding corroborates the fact that normal RGR lacks the RBD, and therefore it should not become activated by RAS-GTP.

Next, the organization of the genomic region that corresponds to the oncogenic rgr cDNA was determined by PCR amplification of genomic rabbit DNA, using exon-specific primers. The genomic structure and exon-intron organization of rabbit rgr is shown in Fig. 2 B. The rabbit rgr gene spans approximately 7 kb and consists of nine exons and eight introns. The AUG that initiates the translation of RGR is at the beginning of the second exon. This methionine also was present in the oncogenic form of RGR, but in this case it was preceded by ∼80 additional amino acids. These extra amino acids are the consequence of the fusion of the rgr gene with the hr23A gene and result in the protein expression of a sequence that belongs to the 5′-UTR of the normal rgr gene. This was possible due to abnormal splicing that occurred in the oncogenic mRNA. The last exon of rgr consists of 300 bp and contains the stop codon (TGA) and 3′-UTR.

The 5′ leader region of rgr mRNA has 712 bp and is carried almost on its entirety on a single exon. It contains eight AUG codons upstream of the AUG that initiates the translation of RGR, several of which are followed by short ORFs. It has been extensively reported that long leader sequences have translational inhibitory effects on the synthesis of proteins (15, 16). However, the most prominent feature of this leader is that the first upstream AUG codon, located 40 nucleotides from the 5′-end and in a Kozak consensus sequence, initiates a 208-codon ORF that terminates 49 nucleotides before the RGR ORF (uORF). In a cap-dependent translational mechanism, three possibilities for the expression of RGR can be taken into account: (a) the reinitiation of the ribosomal machinery; (b) a leaky scanning mechanism; and (c) a ribosomal shunting. Alternatively, the translation of the main ORF could be reached by a cap-independent translation process (for review, see Ref. 17).

To confirm that the RGR protein actually is translated, we analyzed the products of the in vitro translation. For in vitro translation reactions, equal amounts of chimeric cDNA were transcribed and translated in the rabbit reticulocyte coupled system. Translation products were labeled by including [³⁵S]methionine in the reactions and visualized after SDS-PAGE and autoradiography. As shown in Fig. 3, A and B, the translation of the Fl-Rgr construct (rgr mRNA) leads to the synthesis of two proteins. The smaller band corresponds to the protein initiated at the uORF, whereas the larger band presents the size predicted from the RGR ORF sequence (45 kDa). As expected, the 713-Rgr construct produces the endogenous RGR protein, slightly smaller than the oncogenic RGR protein (translated from the construct RSC-Rgr), which resulted from the fusion (Fig. 3,B), and its apparent molecular mass is identical to the product of a chimeric mRNA that lacks all of the upstream leader sequence (Fig. 3 A). Although this is not a quantitative analysis, it can be observed that the levels of the proteins that are encoded by transcripts lacking the uORF are much higher than the levels of those coded by the complete mRNA. This observation underscores the repressive effect of this long leader sequence on the translation of the RGR protein.

To further evaluate the role of the uORF on the repression of RGR expression, the first two upstream AUG codons were deleted to avoid the translation of the uORF (F1ΔuORF-Rgr). One of the models that could explain this repression postulates that the act of translating an uORF reduces downstream gene expression because ribosome reinitiation is inefficient. To address this issue, we mutated the 3′-end of the uORF to make a fusion protein between this uORF and the RGR (F1→2-Rgr), thus eliminating the contribution of reinitiating ribosomes to RGR expression (Fig. 3,A). When translated in vitro, F1ΔuORF-Rgr encoded only the RGR protein; however, the construct, F1→2-Rgr, encoded more than one protein, including RGR (Fig. 3,B). To quantify the effect of these leader modifications, the constructs were fused to the luciferase coding region and then transfected into 293T cells with them. As shown in Fig. 3 C, the wild-type leader sequence has a clear repressive effect on the translation of the reporter gene (F1-Luc). This repression was partially diminished in the absence of the translation of the uORF (F1ΔuORF-luc), indicating that the synthesis of the upstream protein has a negative effect on the translation of RGR. Conversely, when the reinitiation of the ribosome was impaired, the reduction in the translation of the reporter protein was more pronounced (F1→2-luc). This result demonstrated that the ribosome reinitiation mechanism partially mediates RGR translation.

Although the cap-dependent mechanism plays a role in RGR translation, the possible contribution of a cap-independent mechanism was investigated, via binding to an IRES. Therefore, introducing an inverted repeat structure near the 5′-end of the RNA blocked scanning ribosomes (S1-F1-luc). Interestingly, the luciferase activity decreased, although it was not completely abolished. These data point to an independence, at least in part, from ribosomes loaded at the 5′-end of the mRNA.

Finally, rgr mRNA was tested to analyze translation in vivo. To address this issue, NIH3T3 cells were infected with retroviral supernatants obtained from 293T cells transfected with retroviral vectors carrying the complete rgr mRNA. Western blot analysis demonstrated that the low levels of translation of the normal RGR in these cells was due to the presence of the 5′ leader sequence (Fig. 3 D).

The endogenous RGR protein is almost identical to the oncogenic form of RGR isolated previously as a fusion protein, except for a few amino acids that are present at the 5′-end of the oncogenic form and not in the normal form of the protein. This fact pointed to overexpression as the most likely mechanism that renders rgr oncogenic. To test this hypothesis, NIH3T3 cells were transfected with either the plasmid that carries the oncogene or one that contains the oncogene with the 5′-UTR from the wild-type rgr transcript in its 5′-end, using the empty vector as control. Foci formation was assessed to determine the oncogenic potential of the two constructs. As a control, some of the plates were selected with neomycin, clones were grown up, and RNA was extracted to determine expression of the constructs. As shown in Fig. 4,A, the RGR-transformed cells were able to induce foci efficiently, as expected. In contrast, the construct containing the rgr wild-type 5′-UTR was unable to produce foci. When RNA from colonies containing each construct was screened for expression by RT-PCR, the construct that did not elicit any foci was expressed at the RNA level (Fig. 4 B), confirming that the 5′-UTR was most likely inhibiting translation.

The activation of the RAS-MAPK-ERK cascade induced by RGR (9) could be due to a direct interaction of RGR with RAS through the guanine exchange domain or to an indirect mechanism through an adaptor protein. In this indirect model, RGR could use domains other than the catalytic region to bind to this adaptor protein.

To define the domain of RGR involved in RAS activation, two expression plasmids were constructed encoding NH₂-terminally deleted derivatives of the oncogenic form, 713-Rgr and 1027-Rgr. The first of these constructs carries the DNA sequence starting at the first methionine translated in the normal RGR, which is shorter than the oncogenic form (see above), whereas the 1027-Rgr plasmid encodes a protein that starts a few amino acids before the catalytic domain.

To investigate the biochemical properties of these NH₂-terminally deleted proteins, 293T cells were transfected with retroviral vectors carrying the rgr NH₂-terminally deleted forms, and the viral supernatants of these cultures were used to infect NIH3T3 cells. After the infection, these cells were cultured and maintained in selection media, and the expression of the mRNA for the different rgr truncations was analyzed by Northern blot. Because we had previously observed that MAPK activation by the oncogenic RGR is mediated by RAS (9), extracts were prepared from the cells, and the p-MAPK levels were analyzed by Western blot. As shown in Fig. 5 A, 1027-Rgr is able to induce the phosphorylation of MAPK, although it lacks almost all of the NH₂-terminal sequence up to the catalytic domain.

Four point mutants, D1, D2A, D2B, and D3, were constructed in which the mutations were designed to inactivate the CDC25 catalytic domain. One point mutation in the COOH-terminal end of RGR was also generated to disrupt a consensus phosphorylation site for the cyclic AMP-dependent protein kinase (KKPTA). This site is also present in other members of the family, but its significance is not yet known.

The analysis of the free RAL-GTP levels in pools of NIH3T3 cells that constitutively expressed these mutant proteins confirmed that the catalytic domain was no longer able to activate RAL, whereas the mutation in the COOH-terminal end did not affect the activation of RAL by RGR (Fig. 5 B).

Next, the effect of these mutants on the activation of the RAS-mediated pathway was investigated. Whereas the mutation in the COOH-terminal end of the RGR protein does not affect the induction of MAPK, all of the mutations that disrupt the guanine exchange factor domain render RGR unable to activate this pathway. Instead, these mutants displayed levels of p-MAPK similar to those of the empty vector-infected cells (Fig. 5 B).

Finally, as an alternative approach to investigate the ability of the truncation and point mutation RGR derivatives to induce cellular transformation, a focus assay was performed. As shown in Fig. 5 C, only the catalytic domain is necessary for RGR-induced transformation. This result correlates with the activation of the RAL- and RAS-dependent pathways by the different forms of RGR and is consistent with a direct mechanism of activation of RAS by RGR, although we cannot completely rule out an indirect mechanism of Ras activation by rgr.

To further investigate the hypothesis of a direct mechanism of interaction between RAS and RAL, we proceeded to identify the small GTPases that could be substrates for the oncogenic form of RGR by assessing their ability to bind a panel of nucleotide-free GTPases. GEFs bind with high affinity to RAS proteins in the absence of Mg²⁺/GTP or GDP and stabilize the nucleotide-free state (6). Therefore, epitope-tagged RGR protein expressed in 293T cells was incubated with the GST-GTPase fusion proteins that had been stripped of GDP in the presence of EDTA. As shown in Fig. 6, both of the nucleotide-free RAL isoforms were able to bind the oncogenic form of RGR but also bound H-RAS, albeit with less affinity. A faint band could also be detected in the lane corresponding to RAP1, but RGR was unable to bind to TC21, a small GTPase of the same family.

In the search for distinctive functions for this new form of RAL-GEF, and given its lack of a RBD, we hypothesized that this might result in a different subcellular localization from other members of the RAL-GEF family.

RALGDS family members seem to be activated by RAS, upon targeting RAL to the plasma membrane. However, this small GTPase has been reported to localize and function not only at the plasma membrane, but also in the Golgi. Because endogenous RGR lacks the RBD, its recruitment to the plasma membrane by RAS is very unlikely. Therefore, the subcellular localization of the endogenous RGR was examined in the RK13 cell line by immunofluorescence using a previously described RGR polyclonal antibody (9).

RK13 cells were fixed, permeabilized, and stained for RGR. As shown in Fig. 7, endogenous RGR was clearly observed in the perinuclear region of these cells, with a multivesicular distribution. This punctate localization is highly suggestive of the endoplasmic reticulum, Golgi apparatus, and/or mitochondria.

Although the endogenous RGR was detected in endomembranes, the results presented above demonstrate that the oncogenic RGR induces the activation of the MAPKs and that it is indeed able to directly interact with RAS by using its catalytic domain, leading to the activation of this GTPase in vivo. It has been described that H-RAS and N-RAS transit to the Golgi (18). However, activation of their downstream effectors at the plasma membrane is thought to be the predominant route for many pathways. To analyze the putative location of the interaction between the oncogenic RGR and RAS detected by the in vitro experiments, the NH₂ terminus of the oncogenic RGR was tagged with GFP, and then NIH3T3 cells were transfected with this construct, and the clones were selected with antibiotic. Subsequently, the levels of the fusion protein mRNA in different clones were analyzed by Northern blot (Fig. 8,C), and the subcellular localization in living cells was observed by epifluorescence microscopy. In Fig. 8 A, the cell line that exhibits low expression of GFP-RGR presents an area of fluorescence adjacent to the nucleus (Golgi). However, in the cell lines that overexpress the fusion protein, the most prominent feature was the extension of the fluorescence to the plasma membrane.

When the activation of the MAPK pathway was analyzed, this pathway was activated only in the cell lines that overexpress GFP-RGR (Fig. 8 B). These data support the idea that RGR needs to be overexpressed and then mislocalized to activate RAS and to become oncogenic.

The rgr oncogene, initially reported as part of the RSC fusion protein, was isolated in our laboratory from chemically initiated tumorigenesis in rabbit skin. Further analysis was based on the ability of RGR to induce tumor formation in nude mice. In addition, RGR promoted the transformation of NIH3T3 cells, and this property was dependent on the activation of the mitogen-activated protein/ERK kinase (MEK)-MAPK pathway because rgr-induced transformation is correlated with an increase of activated ERK (9). Here we report that endogenous rgr mRNA is expressed in rabbit tissues such as spleen, thymus, and testis. This expression pattern suggests that the normal function of RGR is important in cells of hematopoietic origin. The restricted localization of the rgr mRNA in these tissues (spleen, thymus, and testis) contrasts with the ubiquitous expression of other members of the family, such as RalGDS, rlf, and rgl. The pattern of expression in rabbit coincides with our recent observation that the alteration of human rgr is involved in human T-cell malignancies (8) and is consistent with a specific role for RGR in T-cell function.

The identification cDNA and isolation of the full-length wild-type rgr cDNA demonstrated that in the fusion process between hr23A and rgr, the rgr gene was fused to hr23A by an upstream region to the 5′-UTR rgr sequence. This fusion resulted in the abnormal splicing of a segment located in the 5′ leader region; as a consequence, the oncogenic RGR is a slightly larger version than its normal counterpart, and it contains an additional NH₂-terminal end.

The isolation of the wild-type, endogenous rgr cDNA demonstrated that it contained a long 5′ leader sequence. This leader sequence contains 712 nucleotides and 8 upstream AUG codons. Furthermore, in vitro translation of rgr showed that this transcript encodes two proteins: the first one by the leader sequence, whose function, if any, is thus far unknown; and the second one, which corresponds to RGR. It has been extensively reported that most mRNAs encoding oncoproteins and growth factors related to cell proliferation possess a long leader sequence with one or more AUG triplets upstream of the translation initiation codon (19, 20). This configuration poses a major barrier to 5′ cap-dependent translation initiation and ribosomal scanning. These features suggest that they are involved in translational control (21). Unlike transcriptional control, regulation at the level of translation allows for an immediate and rapid cellular response. The experiments in which the 5′-end of the rgr cDNA was fused to the luciferase cDNA confirmed that the 5′ leader sequence exerts a potent inhibitory effect on the translation of RGR and indicated that ribosomal reinitiation occurred. However, other cap-dependent mechanisms cannot be ruled out, such as leaky scanning or ribosomal shunting. More interesting is the fact that RGR translation takes place partially when the scanning ribosomes were blocked at the 5′-end. The IRES are defined solely by functional criteria and cannot yet be predicted by the presence of characteristic RNA sequence or structural motifs (22). Therefore, additional experiments would be necessary to analyze the independent translation of the second cistron.

An IRES in ODC mRNA whose translation was considered to be cap dependent because the first segment represses translation in vitro and in vivo, presenting a strong homology with the picornavirus 5′-UTRs, has been described previously (23). The picornavirus IRES contains two conserved elements: a UUUC sequence in an extended pyrimidine-rich stretch; and an AUG-containing element located at a conserved distance (approximately 20 nucleotides) downstream from the first element. These features also were present in the sequence of the mammalian ODC 5′-UTR, and this IRES functions during G₂-M, a cell cycle phase in which there is a strong inhibition of protein synthesis. In the case of rgr mRNA, there is also a pyrimidine-rich sequence, 23 nucleotides upstream of the AUG, that initiates the synthesis of RGR. This sequence resembles those viral and cellular IRES, and we hypothesize that this also might function at G₂-M. This could lead to the activation of cytocentrin, a RAL effector involved in the assembly and function of the mitotic apparatus (24). As it occurs with ODC synthesis, the endogenous RGR could be cap-dependently translated during the G₁-S-phase, at a time when there is a general augmentation of protein synthesis necessary for the entry into the cell cycle. In addition, it could also be cap-independently translated during mitotic spindle formation and chromatin condensation and thus be involved in the activation of cytocentrin.

The 5′-UTR of the wild-type rgr gene has been shown to inhibit the transforming ability of the oncogene (Fig. 4,A), whereas the construct is expressed at the RNA level (Fig. 4 B). This evidence lends strong support for the inhibitory translational control exerted by the 5′-UTR of the rgr gene. It also is consistent with the results obtained in human tumor cell lines with the altered hrgr, where high expression of truncated forms of the gene is observed (8).

The isolation of the entire rgr cDNA confirmed that this GEF does not have a detectable RBD, a feature that is also present in two other members of the RAL-GEF family (5, 6). The endogenous RGR also lacks a REM, an NH₂-terminal domain found in most GEFs that seems to play a role in maintaining the conformation of the catalytic domain (25, 26). Interestingly, the two other members of the RAL-GEF family that lack the RBD also are defective in the REM domain.

To investigate the possibility that other modules in the RGR protein are involved in its transforming properties, the NH₂-terminal region was deleted up to the catalytic domain. Although it has been reported that this truncation renders other GEFs inactive (26), in our case, the NH₂-terminally deleted RGR was able to induce foci formation at the same level as the oncogenic protein. In RALGDS, the NH₂-terminal region is involved in the mechanism of the protein kinase C action (27). Because the RGR NH₂-terminal region also presents consensus sites for protein kinase C phosphorylation, it is possible that such regulation also occurs in the RGR protein. However, the CDC25 domain mutation narrowed down the catalytic domain of RGR as the domain in which the transforming properties of the oncogene reside.

We have previously described the ability of RGR to induce RAS activation. Although in vitro experiments performed with RGR obtained from bacteria extracts failed to induce RAS-GTP dissociating activity (4), the in vitro experiments presented with RGR of eukaryotic origin confirmed the ability of RGR to bind to a less restricted spectrum of GTPases. The explanation for the original negative exchange activity published for Rgr is the use on those in vitro assays of a prokaryotic fusion protein with very low GEF activity, and we have seen in the pull-down assays that the GEF activity of Rgr for Ras is lower than that for Ral. RAL-GEF proteins that have the RBD seem to be very specific with respect to the substrate GTPase. In contrast, it has recently been reported that AND34, a RAL-GEF that lacks the RBD, as does RGR, also exhibits a somewhat unusual spectrum of activation for GTPases (5). Moreover, other investigators were also able to detect a faint interaction between RALGPS, the other RAL-GEF that lacks the RBD, and Rap1A (6). It is possible that different domains that are present in the RAL-GEF family members could modulate substrate specificity. For instance, the REM domain also could render these proteins more specific for the RAL GTPase. The efficiency with which the oncogenic RGR and the different GTPases interact might be determined by their subcellular localization, by their expression levels, and by the presence of other proteins in this particular cell compartment.

It has been reported that RAL is present in the plasma membrane along with RAS (28). However, the majority of the protein is in intracellular vesicles (3). RGR lacks the RBD, and therefore it was very unlikely that the endogenous protein could be recruited to the plasma membrane, where it would be more easily activated by RAS. The immunofluorescence analysis of the subcellular localization of the normal, endogenous protein indicated that this GEF is present in the endomembrane network. This localization supports the hypothesis that, when acting under physiological conditions, RGR might activate the pool of RAL in the vesicles in a RAS-independent manner. In agreement with this idea, it has been reported that RAL and RalBP1, a RAL effector highly similar to cytocentrin (24), regulate the endocytosis of epidermal growth factor and insulin receptors (29).

When the GFP-RGR fusion protein is overexpressed in NIH3T3 cells, fluorescence appears not only in the perinuclear region but also in the cytosol and in the plasma membrane. We have demonstrated here that RGR is able to interact with RAS, and this provides the molecular underpinnings for our previous observation that RGR is able to activate RAS (9). RGR also was shown to go to the plasma membrane when overexpressed, and this is consistent with facilitating the interaction between these two proteins, leading to the activation of the RAS downstream pathways.

In summary, we report that the DNA rearrangement that led to formation of the RSC fusion protein abrogated the strongly negative regulatory element that is the 5′ leader sequence of the endogenous rgr. The translational regulation, which restricts expression of rgr in organs such as the spleen and thymus as well as the subcellular localization of RGR in the vesicles, suggests a dual function that would take place in different phases of the cell cycle. On one hand, RGR would be translated by a cap-dependent mechanism during G₁-S phase, and in this phase it might be involved in endocytosis. Conversely, when the cell enters the G₂-M phase, the translational regulation of RGR would occur by ribosomal binding to an IRES in a cap-independent model. RAL activation at this time could lead to the proper assembly of the mitotic spindle. These hypothesized functions for RGR will be interesting avenues to explore in search of blocking strategies for this oncogene now that we have shown its involvement in human lymphomas (8).

In this report we have demonstrated that: (a) the catalytic domain is essential for the activation of the MAPK pathway, (b) RGR is able to interact directly with RAL and RAS, and (c) overexpression of the oncogene, due to the abrogation of its normal translational controls, leads to its mislocalization to the plasma membrane. This information is crucial to understanding the mechanism of activation of rgr and to explain its transforming ability when overexpressed.

We are indebted to R. Cuesta, I. Novoa, R. Schneider, and G. Kreibich for helpful discussions. We also thank H. Yano for assistance with the confocal microscopy and L. Martello-Roonie for critical reading of the manuscript.