Trypanosoma brucei TBRGG1, a Mitochondrial Oligo(U)-binding Protein That Co-localizes with an in VitroRNA Editing Activity*

We report the characterization of aTrypanosoma brucei 75-kDa protein of the RGG (Arg-Gly-Gly) type, termed TBRGG1. Dicistronic and monocistronic transcripts of theTBRGG1 gene were produced by both alternative splicing and polyadenylation. TBRGG1 was found in two or three forms that differ in their electrophoretic mobility on SDS-polyacrylamide gel electrophoresis gels, one of which was more abundant in the procyclic form of the parasite. TBRGG1 was localized to the mitochondrion and appeared to be more abundant in bloodstream intermediate and stumpy forms in which the mitochondrion reactivates and during the procyclic stage, which possesses a fully functional mitochondrion. This protein was characterized to display oligo(U) binding characteristics and was found to co-localize with an in vitro RNA editing activity in a sedimentation analysis. TBRGG1 most likely corresponds to the 83-kDa oligo(U)-binding protein previously identified by UV cross-linking of guide RNA to mitochondrial lysates (Leegwater, P., Speijer, D., and Benne, R. (1995) Eur. J. Biochem.227, 780–786).

Trypanosomes are primitive eukaryotes whose parasitic life cycle involves the differentiation into several successive adaptive forms in different hosts and environments. The main developmental stages are the bloodstream form in the mammalian host and the procyclic form in the tsetse fly vector. Each of these stages is characterized by a major surface protein, the variant surface glycoprotein in the bloodstream form and procyclin in the procyclic form (1,2). Another major difference between these forms is the energy metabolism. Bloodstream forms possess an inactive mitochondrion and respire through the catabolism of glucose in specialized organelles termed glycosomes, whereas procyclic forms utilize a fully functional mitochondrion for oxidative phosphorylation, and amino acids probably serve as the major carbon source in vivo (3). In the bloodstream, the mitochondrion is reactivated when the trypanosomes differentiate from the proliferative slender form into the quiescent stumpy form through several intermediate stages.
The nuclear genome of these organisms appears to be organized in long polycistronic transcription units and probably contains only a few promoters (4 -7). The expression of many genes analyzed so far appears to be stage-specific and reflects the developmental stage of the parasite. Interestingly, genes belonging to the same transcription unit are often differentially stage-regulated, indicating that post-transcriptional processes operating at the levels of RNA maturation, stability, and translation are primarily responsible for controlling cellular differentiation (4 -7). The primary polycistronic transcripts of trypanosomes are rapidly processed into mature mRNAs by trans-splicing and polyadenylation. These processing events appear to be coupled, the choice of a polyadenylation site being apparently dictated by the position of the downstream splice site, probably through the scanning of transcripts by a multifactorial complex encompassing both processing activities, which is recruited to the RNA by the binding to a polypyrimidine stretch (7,8). Several observations indicate that transsplicing and polyadenylation can be stage-regulated. Alternative splicing of transcripts of the ESAG6 and TBA1 genes differs between bloodstream and procyclic forms of the parasite (9,10). Similar results have been obtained regarding the polyadenylation of transcripts for a high mobility group protein (11). In all these processes, it is believed that RNA-binding proteins play a major role, but presently nothing is known about these proteins.
Mitochondrial transcripts of trypanosomes are also subjected to a developmentally regulated post-transcriptional RNA processing, which involves mechanisms of a nature totally different from that of nuclear genes. These transcripts are rendered functional by a complex editing mechanism involving endonuclease(s), terminal uridylyl transferase or uridylyl exonuclease, and RNA ligase activities, which leads to the insertion and/or deletion of uridylate residues templated by short polyuridylated RNAs termed guide RNAs, all encoded in the mitochondrial genome (12)(13)(14)(15)(16)(17). RNA editing allows the expression of several mitochondrial cryptogenes, which encode subunits of the enzymes of a respiratory-redox chain coupled to phosphorylation of ADP, which is only active in procyclic forms.
Here we report the characterization of an RNA-binding protein specific for poly(U), which selectively accumulates in the mitochondrion when this organelle is active. Alternative processing of the transcripts and stage-regulated protein modification suggest that complex controls regulate the expression of TBRGG1. It is likely that this protein is involved in mitochondrial RNA processing, either in rRNA synthesis and maturation or in RNA editing.

Trypanosomes
Trypanosoma brucei pleomorphic bloodstream forms were from the AnTat 1.1 clone. Procyclic forms were derived from these cells by in vitro cultivation (18).

DNA Cloning and Analysis
A cDNA library was constructed in gt10 from poly(A) ϩ mRNA isolated from bloodstream forms using the Amersham Pharmacia Bio-tech cDNA synthesis and cloning kit according to the instructions of the manufacturer. For in vitro transcription and translation, the entire 2.9-kb 1 cDNA containing the TBRGG1 open reading frame (ORF) and a 0.95-kb KpnI-SalI fragment containing only the beginning of this ORF (and thus the RGG repeats) were cloned in pGEM3Z-digested, respectively, with EcoRI or KpnI ϩ SalI. This gave rise to pGEM2.9 and pGEMRGG, respectively. For bacterial expression, a MscI-PvuII fragment from pGEM2.9 containing the whole TBRGG1 ORF was cloned in SmaI-digested pGEX4T1, giving pGEX2.9. In addition, a 720-bp MscI-SalI fragment from pGEMRGG containing the sequence encoding the RGG repeats was cloned in SmaI ϩ SalI-digested pGEX4T1, giving pGEXRGG. Southern and Northern blot analysis were conducted as described previously (18).

Production of Specific Antibodies
The TBRGG1 protein and a polypeptide of 238 amino acids (amino acids 30 to 268) containing the RGG repeats (RGG peptide) were expressed in E. coli from pGEX2.9 and pGEXRGG. The two GST fusion proteins were purified in large amounts using the Bio-Rad Cell Prep SDS-PAGE apparatus. The fractions containing the purified recombinant proteins were applied on a column of AG11A8 resin to remove the SDS. The RGG peptide was cleaved from the GST moiety using thrombin and repurified using the same procedure. The TBRGG1-GST fusion and the RGG peptide were then used to immunize rabbits and chickens. Chicken egg yolks delipidified with chloroform as well as rabbit antisera were used for subsequent analyses.

In Vitro Synthesis of TBRGG1 Polypeptides
The TBRGG1 protein and the RGG peptide were synthesized in reticulocyte lysate in the presence of 35 S-labeled methionine from pGEM2.9 and pGEMRGG, respectively, using the Promega in vitro transcription and translation systems.
DNA and RNA Binding Assays 2 l of reticulocyte lysate containing the labeled protein were mixed with 25 l of agarose beads bound to oligoribonucleotides (poly(A), poly(C), poly(G), and poly(U), purchased from Amersham or Sigma) or single-stranded salmon sperm DNA (purchased from Life Technologies, Inc.), in 0.5 ml of binding buffer (2.5 mM MgCl 2 , 0.5% Triton X-100, 10 mM Tris-HCl, pH 7.4, with 250, 500, 750, or 1000 mM NaCl). The binding was allowed to proceed for 10 min at 4°C with mild agitation. The beads were then pelleted and subjected to 6 washes with 1 ml of binding buffer. Sample buffer was added to the beads, followed by boiling for 5 min and SDS-PAGE analysis. After electrophoresis, the gel was subjected to fluorography, dried, and exposed. UV cross-linking assays were conducted as described previously (19).

Subcellular Fractionation Procedures
Renografin Gradient-Mitochondrial vesicle isolation was performed in a 20 -35% renografin gradient, essentially as described (20). Renografin gradients were divided into 14 fractions of 2 ml. Before Western blotting, 0.02% laurylmaltoside was added to the fractions followed by dialysis against 5 mM EDTA, 25 mM Tris, pH 7.6. Proteins were precipitated with 10% trichloroacetic acid and resuspended in SDS-PAGE layer mix at a concentration of approximately 3 g/ml.
Glycerol Gradient-Procyclic T. brucei IsTaR 1 (21) was grown at 27°C in SDM-79 medium supplemented with 10% (v/v) heat-inactivated bovine fetal calf serum (22). The preparation of mitochondrial vesicles followed the protocol of Harris et al. (23). Detergent lysates of the vesicle preparations were prepared as described by Göringer et al. In panel A, the underlined stretches are 5 repeats containing the conserved RGG motif found in a class of RNA-binding proteins (32). Panel B shows an alignment of the 5 repeats, with indication of the RGG and related motifs (boxed) and of the aromatic residues (black squares). Panel C shows the amphipatic helix contained in the N-terminal sequence, which could be a mitochondrial targeting signal (33). Ⅺ, hydrophobic. for the solubilization. Protein concentrations were determined according to Bradford (25). Mitochondrial lysates (10 mg) in 20 mM Hepes, pH 8.3, 30 mM KCl, 10 mM magnesium acetate, 5 mM CaCl 2 , 0.5 mM dithiothreitol, 1% (v/v) Triton X-100, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 10 g/ml bovine trypsin inhibitor were fractionated by density centrifugation in linear 10 -35% (v/v) glycerol gradients (26). Centrifugation was performed in a Beckman SW 41 rotor at 38,000 rpm for 5 h at 4°C. Twelve 1-ml fractions were collected from the top of the gradient. The fractions were tested for their in vitro RNA editing activity as described by Seiwert et al. (13).

Affinity Purification of anti-TBRGG1 Antibodies and Immunoblotting
Affinity purification of the antibodies was performed as described (27). Immobilized full-length GST-TBRGG1 fusion protein was used as the antigen. For immunoblotting, protein extracts were separated by SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked with 5% (w/v) bovine serum albumin in phosphate-buffered saline and probed with the antibodies.

Western Blotting
The protein samples were run in 10% Laemmli SDS-PAGE. Western blotting was performed by the semi-dry method for 2 h limited at 250 mA onto nitrocellulose. This was followed by standard detection procedures. The antisera used were rabbit and chicken anti-TBRGG1, a rabbit anti-ATPase raised against the mitochondrial ATP synthase of Crithidia fasciculata and showing cross-reactivity predominantly with the ␤-subunit from T. brucei, and a rabbit anti-glucosephosphate isomerase raised against glucosephosphate isomerase of T. brucei, which in procyclic forms is 50% glycosomal and 50% cytoplasmic (a gift of Prof. F. R. Opperdoes, Institute for Cellular Pathology, Brussels). Pre-immune sera served as negative controls. Secondary incubations were done with the appropriate antibodies coupled to alkaline phosphatase. Detection of alkaline phosphatase activity was performed with 4-nitro blue tetrazolium chloride and 5-bromo-4-chloro-3-indolyl-phosphate, both from Boehringer Mannheim, according to the manufacturer.
Chicken anti-TBRGG1 was used at a 1:50 dilution, and the fluorescein isothiocyanate-coupled anti-chicken antibody (Sigma) was diluted 320fold. Images were taken on a Zeiss Axioscop microscope coupled to a CCD camera and processed by a ISIS 3 software.

Electron Microscopy Immunochemistry
Pleomorphic bloodstream forms and a purified mitochondrial fraction from procyclic culture forms were fixed and embedded in Lowicryl as described previously (29). Ultrathin sections were floated overnight at 4°C on rabbit anti-TBRGG1 immune and pre-immune serum diluted 100-fold. After careful washes, the sections were incubated for 1 h with Fab fragments of goat anti-rabbit IgGs conjugated to 10-nm gold particles (EM.GFAR10, BioCell, Cardiff, UK) diluted 100-fold.

Cloning of a T. brucei Gene for a RNA-binding Protein (TBRGG1)-
The cloning of the TBRGG1 gene was initially designed to characterize the trypanosome homologue of nucleolin from higher eukaryotes. Nucleolin is an RNA-binding protein involved in the synthesis and processing of polI transcripts (30). We were interested in the identification of a trypanosome homologue of this protein because the transcription of the variant surface glycoprotein and procyclin genes appears to be mediated by polymerase I (31), and RNA-binding proteins are likely to represent key control elements for the expression of these genes (2, 4 -7). Therefore, we used a probe from hamster nucleolin to screen a T. brucei bloodstream form cDNA library cloned in a bacteriophage vector. Several positive clones were obtained, which were found to contain the same sequence, at least partially. The nucleotide sequence of the largest cDNA is shown in Fig. 1. This cDNA (3.4 kb) was full-sized, since at the 5Ј extremity it contained the end of the common spliced leader (termed mini-exon), and it was polyadenylated at the 3Ј extremity. Interestingly, this cDNA was found to contain two successive ORFs. The largest one encoded a putative 75-kDa protein showing five typical repeated motifs found in a class of RNA-binding proteins termed RGG, which consist of closely spaced RGG tripeptides interspersed with phenylalanine and/or aromatic amino acids (32) ( Fig. 2A and B); hence the trypanosome gene was termed TBRGG1, for T. brucei RGG gene 1. The region of the RGG repeats is the only one showing significant homology with hamster nucleolin (30) and was probably responsible for the selection of these clones. The first ORF encoded a putative polypeptide of about 15 kDa, showing no significant homology with sequences from data bases.
Different fragments from the cDNA were used as probes to analyze the nuclear genes by hybridization. As shown in Fig. 3, the different probes confirmed the arrangement of the two successive ORFs. Probe 1, which is specific to ORF1, hybridized to the same 2.2-kb SacI fragment as probe 2, which is specific to the beginning of ORF2 (SS ϩ S digests in panels 1 and 2). In the same digests, probe 3 recognized two different fragments of 2.4 and 0.8 kb (see map in Fig. 3). Similarly, as expected from  5. The TBRGG1 protein is present in two to three forms depending on the developmental stage of the parasite. Western blots of 50 g of total protein extracts from procyclic (p) and bloodstream long slender (b) forms of T. brucei were incubated with immune (i) or pre-immune (pi) anti-TBRGG1 antibodies. The arrow points to the procyclic-specific band. the map, the same 1.6-kb PstI fragment was recognized by probes 1 and 2, whereas KpnI allowed an identification of the specific fragments hybridizing to each of the two probes. This analysis showed that although the ORF1 region is unique, a few sequences distantly related to the region encompassing TBRGG1 exist elsewhere in the genome (see, for instance, the weakly hybridizing bands at the top of the SS ϩ S and K lanes in panel 2 or the upper KpnI fragment in panel 3).
Alternative Splicing and Polyadenylation of TBRGG1 Transcripts-To evaluate the significance of the existence of two ORFs in the TBRGG1 cDNA, we used probes specific to either of these ORFs to screen the cDNA library and analyze the pattern of steady-state transcripts. The diagrams at the bottom of Fig. 4 show the structure of different full-sized cDNAs ob- tained by this screening. The 3.4-kb cDNA was described above, whereas the 2.9-and 0.7-kb cDNAs appeared to result, respectively, from alternative splicing and alternative polyadenylation of the same primary transcript, generating mature mRNAs for ORF1 and ORF2, respectively. The respective 5Ј limit of the 2.9-kb cDNA and 3Ј limit of the 0.7-kb cDNA are indicated in Fig. 1 (black and open arrowheads, respectively). From the location of these limits, it was clear that the two alternative processings are mutually exclusive. The different transcripts cloned as cDNAs were detected in steady-state RNA from both bloodstream and procyclic forms. The Northern blot analysis presented in Fig. 4 indicates that the 3.4-kb species was a minor transcript slightly more abundant in bloodstream than in procyclic forms, whereas the 0.7-kb transcript was more abundant and slightly enriched in procyclic forms. The 2.9-kb transcript was the major species and was present in roughly equal amounts in both forms of the parasite. Apart from a possible alternative transcript of ORF1 (about 2 kb, see panel 1), there was no evidence for other processing products, indicating that the generation of the stable mRNAs for ORF1 and 2 resulted from mutually exclusive processing of the same primary transcripts. These data also indicated that the DNA fragments faintly cross-hybridizing with the TBRGG1 probe 2 in Fig. 3 did not give rise to detectable levels of additional TBRGG1-specific RNAs.
Stage-regulated Modification of the Protein-The entire TBRGG1 gene and the region of the RGG repeats were cloned in an E. coli expression vector, and the recombinant proteins were purified by SDS-PAGE from cells transfected with the appropriate vectors. Antibodies were raised against these polypeptides in rabbits and chickens. Western blot analyses were performed using these antibodies, and similar results were obtained irrespective of the nature of the immunogen or the source of antibodies. As shown in Fig. 5, the TBRGG1 protein was present in both bloodstream and procyclic forms as a doublet or triplet of bands, depending on the antibody and cell extract, and migrated with the relative molecular mass predicted from the nucleotide sequence of the cDNA (about 75 kDa). With both antibodies, the upper band was more visible in the procyclic form (arrow). As the TBRGG1 mRNA appears to be unique (see above), the multiplicity of protein bands may be explained by either modification or proteolysis or by crossdetection of related proteins.
We attempted to determine whether TBRGG1 was modified by phosphorylation (data not shown). First, TBRGG1 was im-munoprecipitated from trypanosomes grown in the presence of 32 P-orthophosphate and analyzed by SDS-PAGE. No labeling was observed although TBRGG1 was readily immunoprecipitable in control experiments. Second, treatment with phosphatases did not interfere with the pattern of migration in SDS-PAGE.
Since other RGG proteins contain the modified amino acid N G ,N G -dimethylarginine (35), we evaluated the possible involvement of this modification in the observed pattern of bands. The methylation of arginines is inhibited by adenosine dialdehyde (35). However, after incubation of both bloodstream and procyclic forms for three generation times (respectively, 24 and 50 h) in the presence of sub-toxic concentrations of this compound (10 M), no change of the electrophoretic pattern was observed (not shown).
It is worth noting that the electrophoretic mobility of the upper band corresponded to that of the in vitro translation product of the TBRGG1 mRNA (data not shown). Therefore, it is possible that the lower bands arise by proteolytic cleavage. In support of this hypothesis, we observed that a second band can appear from the in vitro synthesized protein after prolonged storage.
Subcellular Localization-The use of the chicken anti-TBRGG1 antibodies allowed the demonstration that TBRGG1 is mitochondrial in intermediate and stumpy bloodstream forms. A preferential accumulation of the protein in the mitochondrion of these developmental forms is clearly demonstrated by immunofluorescence (Fig. 6, SS) and electron microscopy (Fig. 7, panel b). In bloodstream slender forms, this preferential localization was not observed (Fig. 6, SLD). In procyclic forms, the protein also appeared to be present in the mitochondrion (Fig. 6, P).
The mitochondrial localization in procyclic forms was verified by renografin (floating density) gradient analysis. Renografin gradients are routinely used to purify trypanosomal mitochondrial vesicles (see Ref. 20), which are easily identifiable because of the presence of specific cytochromes. Fig. 8A shows a Western blot comparing equal amounts of total and mitochondrial proteins from procyclic forms probed with both anti-TBRGG1 and anti-mitochondrial ATPase antibodies. The bands, respectively, detected with each of these antibodies alone are indicated by the arrows. The TBRGG1 signal in the mitochondrial fraction was about 10 times stronger than that obtained with the total protein fraction, which is similar to the enrichment observed for the mitochondrial ATPase, providing further evidence for the mitochondrial location of the protein.
To check that this result was not due to contamination of the mitochondrial fraction with the glycosomal fraction, the complete renografin gradient was tested in Western blotting assays with different antibodies. As can be seen in Fig. 8B, the TBRGG1 protein colocalized with the mitochondrial ATPase and not with glycosomal glucosephosphate isomerase. The presence of the TBRGG1 protein in the mitochondrial vesicles was directly observed by electron microscopy. Fig. 7a allowed the verification of the purity of the mitochondrial fraction and showed that the labeling is associated with fuzzy material inside the vesicles.
Taken together, these results indicated that, at least at the developmental stages where the mitochondrion is active, TBRGG1 is present in the mitochondrion. Interestingly, the amino acid sequence at the N terminus of TBRGG1 predicted an amphipatic helix compatible with a targeting signal for mitochondrial import (Fig. 2C) (33).
Selective Binding to Poly(U)-The amino acid sequence of the TBRGG1 protein strongly suggested a RNA binding capacity. To check this hypothesis, the protein was synthesized in a   FIG. 8. Localization of TBRGG1, as determined by renografin gradient fractionation. In panel A, a Western blot of 1 g of total (T) or mitochondrial (M) protein from T. brucei procyclic forms was incubated first with chicken anti-TBRGG1 antibodies then with rabbit anti-mitochondrial ATPase. The bands, respectively, detected with each of these antibodies alone are indicated with arrows. In panel B, approximately 35 g of protein from renografin gradient fractions was analyzed by Western blotting using a chicken anti-TBRGG1 antiserum, a rabbit anti-ATPase antiserum, and a rabbit anti-glucosephosphate isomerase (PGI) antiserum. The glucosephosphate isomerase band migrates at an apparent molecular mass of approximately 65 kDa. Procyclin was detectable in fraction 12 to 14 only (results not shown). N D , refractive index.
reticulocyte lysate system in the presence of [ 35 S]methionine, then used in affinity binding assays with a variety of synthetic oligonucleotides and nucleic acids coupled to agarose beads. Electrophoretic analysis of the proteins attached to the beads indicated that at low concentrations of salt, TBRGG1 binds to all oligoribonucleotides except poly(C) and does not bind to single-stranded DNA (Fig. 9A). The binding to poly(U) and poly(A), but not to poly(G), was not affected by concentrations of NaCl up to 1 M, showing that TBRGG1 has a preferential affinity for these two oligoribonucleotides. However, the stronger binding to poly(U) clearly indicated that TBRGG1 is a oligo(U)-binding protein.
To evaluate the involvement of the RGG domain in RNA binding, a fragment of 238 amino acids containing the RGG repeats of TBRGG1 was subjected to the same assays, giving similar results as those obtained with the entire protein (Fig.  9B). Therefore, the RGG domain of TBRGG1 accounts for the binding to poly(U).
TBRGG1 Is Most Likely the 83-kDa Oligo(U)-binding Protein, One of the Two Major Guide RNA-binding Proteins of T. brucei-The RGG domain of TBRGG1 and a mitochondrial lysate of T. brucei were incubated with 32 P-radiolabeled oligo(U), then cross-linked by UV irradiation under conditions previously described (30). Western blots of these labeled materials were probed with the anti-TBRGG1 antibodies and autoradiographed to visualize the bound oligo(U). As expected, the RGG domain was associated to a strong oligo(U) labeling (Fig.  10, lane 2 in panels A and B). The TBRGG1 protein present in the mitochondrial vesicles was found to precisely comigrate with a major oligo(U)-binding protein with an apparent size of 83 kDa under these conditions, which was described previously as the largest T. brucei protein cross-linkable to synthetic guide RNAs provided they are equipped with a U-tail and to oligo(U) (19) (Fig. 10, lane 1 in panels A and B). This last result was not related to the relative abundance of proteins in the mitochondrial extract, since the labeling did not coincide with a major band as seen after silver staining (not shown). Mixing the RGG domain with the mitochondrial lysate led to a decrease in the intensity of the 83-kDa band, indicating that the RGG domain of TBRGG1 can compete with the endogenous 83-kDa protein for limiting amounts of oligo(U) (data not shown).
TBRGG1 Co-localizes with an in Vitro RNA Editing Activity-The mitochondrial localization as well as the oligo(U)-binding properties of TBRGG1 tempted us to test whether the protein is potentially involved in the RNA editing process. The reaction is likely performed within a high molecular mass ribonucleoprotein complex (26,36,37), which probably involves several protein components in addition to gRNAs and preedited mRNAs. A potential participation of TBRGG1 in this process would predict that some or even the entire mitochondrial TBRGG1 population might co-localize with an in vitro RNA editing activity in a sedimentation analysis. To test this hypothesis, we isolated mitochondrial vesicles from procyclicstage trypanosomes (23). The vesicles were lysed (24), and the lysates were separated in glycerol density gradients and fractionated (26). Aliquots of the different fractions were separated by SDS-PAGE and analyzed by immunoblotting using affinitypurified anti-TBRGG1 antibody (Fig. 11A). The majority of the protein was detected in fractions 5-9 equivalent to an apparent S-value range of 35-40 S. Quantitating the specific abundance of TBRGG1 in the different fractions resulted in a distribution as shown in Fig. 11B. Fractions 6 -8 represented the peak of the separation and contained approximately 75% of the total amount of TBRGG1. Thus, the protein showed a sedimentation behavior suggestive of being associated with or assembled into high molecular mass complexes.
To analyze in vitro RNA editing in the different gradient fractions, we chose to monitor a gRNA-dependent U-deletion activity (13). Educt in the reaction is a synthetic, radiolabeled pre-mRNA derived from the first editing site of the T. brucei ATPase 6 (A6) mRNA. This molecule is edited by the addition of mitochondrial protein extracts and a specific gRNA molecule that directs the deletion of exactly 4 uridylate residues. Typically, the reaction results in the conversion of approximately Ϯ10% educt RNA into the editing product. In addition to monitoring editing product formation, the assay resolves a 3Ј pre-mRNA cleavage intermediate as well as chimeric gRNA/mRNA side products (13). Uridylate deletion activity was assayed in the first 10 fractions from the top of the gradient. The activity peaked in fractions 6 to 8 (Fig. 11C) and, thus, precisely overlapped with the gradient distribution of TBRGG1 protein.

DISCUSSION
The TBRGG1 protein is the second RNA-binding protein of the RGG type described in T. brucei. The first one, termed Nopp44/46 (38), is exclusively nucleolar and, thus, is probably involved in nuclear rRNA synthesis and/or processing. Interestingly, Nopp44/46, like TBRGG1, preferentially binds to poly(U). As in both cases, the RGG domain was found to be entirely responsible for RNA binding; the selective affinity for poly(U) in the two proteins is due to the RGG repeats only.
The transcription and processing pattern of TBRGG1 is probably the most complex reported to date in T. brucei. It provides the first example of steady-state accumulation of dicistronic transcripts and involves alternative splicing and polyadenylation of the individual mRNAs for each of the two ORFs. The location of the splice and polyadenylation sites of these mRNAs indicated that their maturation is mutually exclusive. This arrangement opens the way for mutually exclusive control of the expression of ORF1 and TBRGG1. However, in the major developmental stages analyzed, bloodstream and culture procyclic forms, both mRNAs and the polypeptides they encode were produced in relatively large amounts. As both trans-splicing and polyadenylation can exhibit stage specificity (9 -11), is possible that an exclusive accumulation of either ORF1 or TBRGG1 occurs at a minor stage of the parasite life cycle or depends on particular environmental conditions. This issue could be clarified with the determination of the function of the two proteins.
In addition to the complexity of transcript processing, the TBRGG1 protein was found to exist in several forms, one of which is preferentially found in the procyclic stage of the parasite life cycle. We were unable to ascribe this pattern to differential protein phosphorylation or arginine methylation. Proteolytic cleavage might be responsible, although this remains to be proven. Finally, we cannot exclude cross-detection of other proteins of the RGG type, although one of these, Nopp44/46 (44 -46 kDa) (38), is clearly not recognized by the anti-TBRGG1 antibodies. Interestingly, in the Nopp44/46 case also, two to three bands were detected (38). The heterogeneity of TBRGG1 does not seem to be required for sub-cellular targeting, since both major forms appeared to be present in the mitochondrion, as demonstrated by their co-sedimentation all along the renografin gradient.
The role of TBRGG1 is obviously related to the function of the mitochondrion, since this protein is targeted to this organelle as soon as it is reactivated in intermediate bloodstream forms, and it is found in the active mitochondrion of the procyclic form. In a sedimentation analysis of mitochondrial ly-sates, we were not able to identify free TBRGG1 protein. Instead, the entire TBRGG1 population we found in a high molecular mass region of the glycerol gradient. Based on the nonstringent lysis conditions of the mitochondrial vesicles, we feel it is unlikely that this represents high molecular mass aggregates of TBRGG1. In contrast, we favor the interpretation that the protein is assembled into or associated with mitochondrial complexes of a high structural complexity. Intriguingly, the same gradient fractions that contained TBRGG1 protein also exhibited an in vitro RNA editing activity. Although a co-localization per se is not a conclusive argument for a functional relationship, several characteristics of TBRGG1 support such a participation. First, RNA editing is likely mediated by a complex ribonucleoprotein complex (26,36,37). This complex contains pre-edited mRNAs and gRNAs that in a transient interaction direct the processing reaction. Thus, some of the FIG. 10. Detection of TBRGG1 in proteins UV-cross-linked to oligo(U). Aliquots of mitochondrial lysate of T. brucei (lanes 1) or the RGG domain of TBRGG1 fused to GST (lanes 2) were incubated with radiolabeled oligo(U) and UV cross-linked as described in Leegwater et al. (19). Western blots of the labeled extracts were probed with anti-TBRGG1 antibodies (panel A) and autoradiographed (panel B). The proteins of 83, 64, 55, and 42 kDa have previously been identified as oligo(U)-binding proteins in the T. brucei mitochondrial lysate (19).
FIG. 11. Co-localization of TBRGG1 and RNA editing activity. Mitochondrial lysates were separated in 10 -35% (v/v) glycerol gradients and fractionated into 12 1-ml fractions. A, aliquots (3 l) from fractions 1 (top of the gradient) to 11 were separated in 10% (w/v) SDS-containing polyacrylamide gels, transferred onto nitrocellulose membranes, and probed for the presence of TBRGG1 using affinitypurified anti-TBRGG1 antibodies. B, graphical representation of the specific abundance of TBRGG1 in each gradient fraction. The data were derived from a densitometric analysis of the fluorograph shown in A in relation to the total protein concentration of the different fractions. Apparent sedimentation coefficients were determined using the following markers: 5S rRNA from Escherichia coli, thyroglobulin (19.5 S), E. coli 23 S rRNA, E. coli 30 S ribosomal subunits, and E. coli 50 S ribosomal subunits. C, aliquots (12 l) of fractions 1 to 10 were tested for their in vitro uridylate deletion activity and analyzed by denaturing PAGE (9% [w/v]) followed by autoradiography. The electrophoretic mobilities of the editing educt, the U-deletion product (lacking four uridylate residues), the 3Ј cleavage product as well as the main gRNA/mRNA chimeric side product are given in the margin on the right. An asterisk indicates the radioactive phosphate residue at the 3Ј-ends of the different RNAs. -EAF represents a control sample that was treated identically throughout the experiment without the addition of editing-active glycerol gradient fraction. protein components of the editing ribonucleoprotein must contain recognition motifs for the RNA moieties of the complex, such as the RGG box in TBRGG1. Second, the preferential binding to poly(U) suggests an interaction with gRNAs that has been shown to contain 3Ј oligo(U) extensions. In accordance with this suggestion, in T. brucei, U-binding proteins are found in guide RNA-containing particles in nondenaturing gels, and some of them co-sediment with guide RNAs in glycerol gradients (36,39). Among the oligo(U)-binding proteins identified by cross-linking in T. brucei, two major components exhibited molecular weights of 83/90 and 21/26 kDa (19,33,39,40). Whereas the 21/26-kDa species was characterized by gene cloning and found to be an arginine-rich guide RNA-binding protein (33), the 83/90-kDa species appeared to exhibit the binding properties of a C. fasciculata 65-kDa guide RNA-binding protein (19). We show that at least part of the oligo(U) binding activity present in the 83/90-kDa region must be ascribed to TBRGG1. Therefore, a role in guide RNA binding can obviously be proposed for TBRGG1.
Another possible function of TBRGG1 is the involvement in mitochondrial rRNA synthesis and/or processing. This is suggested by the fact that most RNA-binding proteins of the RGG type known to date, including nucleolin of higher eukaryotes and Nopp44/46 of T. brucei, are nucleolar and, therefore, presumably interact with rRNA. Interestingly, the mitochondrial 9S and 12S rRNAs of T. brucei are modified by posttranscriptional addition of a 3Ј-polyuridine tail (41). The preferential binding of TBRGG1 to poly(U) would also be compatible with a role in this process.