Processing and Editing of Overlapping tRNAs in Human Mitochondria*

Overlapping tRNA genes in mitochondria of many metazoans introduce a problem for the processing of such polycistronic primary transcripts. Using runoff transcripts and an S100 extract from HeLa cell mitochondria, the processing of the human mitochondrial tRNATyr/tRNACys precursor (carrying an overlap of one base) was investigated: tRNACys is released in its complete form carrying the overlapping residue at the first position, whereas tRNATyr lacks that nucleotide at the discriminator position. Partial deletion of tRNACys or complete replacement by a non-tRNA-like sequence does not alter the processing reaction and indicates that the upstream tRNATyr alone is recognized by a 3′-endonuclease activity. The truncated 3′-end of this tRNATyr is then completed in an editing reaction that incorporates the missing residue. The processing of this tRNA overlap seems to be species-specific, because an overlapping tRNA precursor (tRNASer(AGY)/tRNALeu(CUN)) from opossum mitochondria is not recognized by the human extract. Because processing activities for overlapping and nonoverlapping tRNA precursors could not be separated, it seems that one general activity is responsible for the 3′-end processing of mitochondrial tRNAs and that this activity coevolved with the particular overlap between tRNATyr and tRNACys in human mitochondria, being unable to recognize overlaps between other tRNAs.

In the mitochondrial genomes of many animal species, some tRNA genes overlap by one to six nucleotides with downstream genes on the same strand (1)(2)(3)(4)(5). An example is the human gene for tRNA Tyr , which shares an A residue with the downstream gene for tRNA Cys such that this nucleotide potentially represents not only the first base of the tRNA Cys but also the discriminator nucleotide of tRNA Tyr . These overlaps introduce a problem for the processing of polycistronic precursors in mitochondria during which the tRNAs are released by cleavage reactions at their 5Ј-and 3Ј-ends (6): from an individual RNA molecule, only one functional tRNA can apparently be created, whereas the other one lacks the shared nucleotide(s). An alternative processing reaction, in which one of the tRNAs is correctly processed in some transcripts and the other in other transcripts, could potentially produce both complete transcripts.
To approach the problem of how the cleavage of the precursor occurs, a processing system using a human mitochondrial extract and the tRNA Tyr /tRNA Cys precursor from human mito-chondria was established. This system leads to the release of a complete downstream tRNA and a 3Ј-truncated upstream tRNA. Furthermore, tRNA deletion and replacement experiments indicate that the cleavage reaction represents the activity of a 3Ј-endonuclease recognizing solely the upstream tRNA Tyr and exclude the participation of RNase P. The released tRNA Tyr must then be completed posttranscriptionally by an editing mechanism. Using an in vitro transcript that corresponds to this 3Ј-truncated tRNA, the editing reaction could be observed in vitro.
In addition, the opossum mitochondrial tRNA Ser (AGY)/ tRNA Leu (CUN) precursor, which also overlaps by one A residue, was not recognized by the human extract. Therefore, it seems that the described mitochondrial processing activity is highly specific in that it does not accept overlapping mitochondrial tRNA precursors found in other species.

PCR 1 and Primers
Wild type and mutated templates for in vitro transcription were prepared by PCR amplification (standard conditions) of the corresponding mitochondrial genes using the following primers synthesized on an ABI DNA/RNA synthesizer. T7

In Vitro Mutagenesis
The site-directed mutagenesis was carried out using the PCR technique as described (7).

In Vitro Transcription and RNA Purification
Transcription was carried out in a volume of 30 l according to the manufacturer (New England Biolabs).
Radioactively labeled transcripts or reaction products were purified by denaturing PAA gel electrophoresis. The bands were cut out with a sterile blade, and the RNA was eluted by incubation in 500 mM ammonium acetate, pH 5.7, 0.1 mM EDTA, 0.1% SDS at 4°C overnight (8) and concentrated by ethanol precipitation. * This work was financially supported by 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.
‡ To whom correspondence should be addressed.

Preparation of Human and Yeast Mitochondrial S100 Extract
Mitochondria were prepared from frozen HeLa cells (Computer Cell Culture Center, Mons, Belgium) or cultivated yeast cells by differential centrifugation and subsequent sucrose step gradient ultracentrifugation (9,10). For extraction, the mitochondria were incubated on ice for 5 min in 6 mM HEPES (pH 7.9), 30 mM KCl, 0.5 mM DTT, 0.2% Nonidet P-40. The lysate was centrifuged at 20,000 ϫ g at 4°C for 20 min, followed by a centrifugation of the supernatant at 100,000 ϫ g at 4°C for 1 h. The supernatant (S100) was dialyzed twice for 8 h against 2 liters of 20 mM HEPES (pH 7.9), 1 mM DTT, 1 mM EDTA, 20% glycerol, and Complete TM protease inhibitor mixture (Boehringer Mannheim; applied according to the manufacturer's instructions). The dialyzed extract was stored at Ϫ80°C until use.
Fractionation of the S100 extract was carried out on an Amersham Pharmacia Biotech fast protein liquid chromatography system with a high S cation exchange column (Bio-Rad) using 20 mM HEPES (pH 7.9), 1 mM DTT as buffer. A linear NaCl gradient (0 -0.8 M over 10 column volumes) was used in the elution procedure. The obtained fractions were desalted using Millipore Ultrafree 5000 microconcentrators; washed twice using 1 volume of 20 mM HEPES (pH 7.9), 1 mM DTT, 1 mM EDTA; adjusted to 40% glycerol; and stored at Ϫ80°C.
The purity of the mitochondrial preparation was assayed by Northern blot analyses for nuclear (U6) and cytoplasmic (tRNA Leu ) marker RNAs and showed no detectable hybridization signal in the mitochondrial fraction, whereas nuclear and cytoplasmic fractions showed strong signals. Additionally, a marker enzyme test for cytoplasm (lactate dehydrogenase) 2 was carried out and revealed 5.6% specific activity in the mitochondrial fraction in comparison to the cytoplasm, indicating a high purity of the mitochondria.

In Vitro Assays
Processing-1 pmol of 32 P-labeled tRNA precursor transcript was incubated with 0.5-2 g of protein of the mitochondrial S100 extract from HeLa cells or with an aliquot of the individual fast protein liquid chromatography fractions in the presence of 30 mM HEPES-KOH (pH 7.6), 6 mM MgCl 2 , 30 mM KCl, 2 mM DTT, and 30 units of RNase inhibitor (Boehringer Mannheim) in a total volume of 20 l for 20 min at 25°C.
Editing-5-10 pmol of 33 P-labeled tRNA Tyr -1 were incubated with 0.5 g of protein of the mitochondrial S100 extract from HeLa or yeast cells in the presence of 1 mM NTPs (Amersham Pharmacia Biotech), 30 mM HEPES-KOH (pH 7.6), 6 mM MgCl 2 , 30 mM KCl, 2 mM DTT, and 30 units of RNase inhibitor (Boehringer Mannheim) in a total volume of 20 l for 6 h at 30°C. The reactions were terminated by phenol-chloroform extraction. The products were ethanol-precipitated, separated by electrophoresis on a 10% polyacrylamide gel containing 8 M urea (11), and visualized by autoradiography.

tRNA Tagging
Ligation of tRNAs to an EcoRI linearized pUC 18 plasmid and subsequent cDNA synthesis was performed as described (12).

RNA Bridging
Bridge-mediated RNA ligation was done as described (13). Chain termination sequencing was carried out on an Amersham Pharmacia Biotech automated laser fluorescence sequencer.

RESULTS
Processing of the Human Mitochondrial tRNA Tyr /tRNA Cys Precursor-A 5Ј-end-labeled in vitro transcript consisting of the overlapping tRNAs for tyrosine and cysteine carrying neither leader nor trailer sequences (Fig. 1A) was incubated with a mitochondrial S100 protein extract from HeLa cells for 20 min. The reaction products were separated by electrophoresis on a denaturing 10% polyacrylamide gel, and the 5Ј-labeled products containing the tRNA Tyr were visualized by autoradiography (Fig. 1D). The processing product appears as one single band with an apparent size of 65 nucleotides, corresponding to the length of a truncated tRNA Tyr . This result excludes an alternative processing, which would lead to two products differing by one nucleotide in length.
In control experiments, 2% SDS was added to the assay, and the S100 extract was pretreated with 20 g of proteinase K for 30 min at 37°C, or heated to 94°C for 5 min prior to the incubation with the RNA substrate. Every individual treatment led to a complete inactivation of the extract, indicating a proteinaceous activity (data not shown).
Processing of a Heterologous Overlapping Bicistronic tRNA Precursor-In order to test whether the human mitochondrial processing machinery is able to recognize other (heterologous) overlapping tRNA molecules in a precursor transcript, a 5Јend-labeled in vitro transcript of the opossum mitochondrial tRNA genes for serine and leucine was offered. Similar to the human tRNA Tyr /tRNA Cys precursor, these tRNAs overlap by one A residue in the primary transcript (Fig. 1B). The result of the incubation of this transcript in the presence of the human mitochondrial S100 protein extract is shown in Fig. 1E: besides background degradation of the substrate RNA, no reaction product is observed (lane 1). One possible explanation for this is that a precursor containing tRNAs for serine and leucine is generally not processed by the mitochondrial S100 extract. To address this question, the experiment was repeated with the human mitochondrial tRNA Ser (AGY)/tRNA Leu (CUN) precursor. This transcript is highly similar to the opossum RNA in primary and secondary structure ( Fig. 1C) but contains no overlapping region between the two tRNAs (both tRNAs have their own A residue at their 3Ј-and 5Ј-ends, respectively). A 5Ј-end-labeled version of this precursor was again tested as a substrate for the human S100 protein extract. Although the processing extract does not recognize the wild type opossum overlapping tRNA Ser /tRNA Leu precursor (lane 1), it is able to cleave the human nonoverlapping transcript and leads to a release of the labeled tRNA Ser (Fig. 1E, lane 3). Therefore, the extract is able to process tRNA Ser /tRNA Leu precursors.
Furthermore, an opossum transcript carrying an insertion of one additional A residue at the border of the tRNA Ser and tRNA Leu sequences (which removes the overlap, Fig. 1B, (2)) was constructed. Interestingly, this transcript carrying the A insertion, and therefore mimicking the situation in the human transcript, is cleaved to release the tRNA Ser (Fig. 1E, lane 2). The different migration positions of the released tRNA Ser versions in lanes 2 and 3 represent the length difference between the opossum tRNA Ser (58 bases, lane 2) and the human tRNA Ser (59 bases, lane 3). Whether these cleavage reactions represent a recognition of the downstream located tRNA Leu and therefore an RNase P activity or a recognition of the tRNA Ser (upstream) by a 3Ј-endonuclease is not clear at this point (both reactions would lead to the same processing products).
To analyze whether both the overlapping tRNA Tyr /tRNA Cys precursor and the nonoverlapping tRNA Ser /tRNA Leu transcripts are cleaved by the same activity, the S100 protein extract was fractionated on a cation exchange chromatography using a High S column and a linear NaCl gradient ranging from 0 to 0.80 M. The resulting 20 fractions were tested for the processing activity using the human overlapping tRNA Tyr / tRNA Cys transcript and the human nonoverlapping tRNA Ser / tRNA Leu precursor (Fig. 2). Fractions eluting between 0.24 and 0.60 M NaCl were able to process both overlapping and nonoverlapping precursor transcripts. Furthermore, the fraction eluting at 0.48 M NaCl (showing the highest processing activity) was used for competition studies with the overlapping (tRNA Tyr /tRNA Cys ) and nonoverlapping (tRNA Ser /tRNA Leu ) precursors. Both transcripts showed a similar efficiency in competing with a radioactively labeled overlapping tRNA Tyr / tRNA Cys precursor (data not shown).
Is RNase P the Responsible Nuclease?-The cleavage between the two tRNA structures in the precursor molecule can in principle be achieved either by recognition of the upstream tRNA Tyr , the downstream tRNA Cys sequence (RNase P), or both tRNA structures.
To be cleaved by RNase P, a tRNA precursor needs to have at least a part of the T loop and the 3Ј-half of the acceptor stem that serves as an external guide sequence (16,17), as has been shown for E. coli, human, and Xenopus laevis RNase P (16 -20). In order to test whether in our transcript the downstream tRNA Cys determines the cleavage (representing a 5Ј-cut by RNase P), an in vitro transcript was designed that contained a complete tRNA Tyr and a 3Ј-truncated form of the tRNA Cys (consisting of 31 bases and terminating immediately 3Ј of the anticodon). This transcript was incubated as a 5Ј-end-labeled molecule in the presence of the S100 protein extract (Fig. 3A). As a control, the complete precursor tRNA Tyr /tRNA Cys was also tested. Interestingly, both incubations led to products migrating at identical positions in the polyacrylamide gel, indicating identical processing reactions. Therefore, the enzymatic activity probably recognizes the upstream tRNA and possibly ele- ments in the 5Ј-half of the downstream tRNA. To distinguish between these two possibilities, the truncated tRNA Cys was completely replaced by the first 30 bases of the mitochondrial mRNA for coxII. This sequence has no similarities to a tRNA structure and carries presumably no recognition elements for a tRNA processing enzyme. The construct was tested as a 5Ј-endlabeled transcript in an overlapping (the A residue of the AUG start codon in coxII represented also the discriminator base of the tRNA Tyr ) as well as in a nonoverlapping situation (Fig. 3B). The released tRNA molecules migrate at a position identical to that of the processing product of the tRNA Tyr /tRNA Cys precursor. These results indicate that the recognition elements for the processing activity are restricted to the upstream tRNA Tyr and that the downstream sequences do not influence the processing reaction.
Determination of the Cleavage Position-In order to get information about the cleavage position on nucleotide level in the human tRNA Tyr /tRNA Cys precursor, an RNA molecule carrying a single 32 P-labeled 5Ј-phosphate group at the overlapping nucleotide position was constructed using a bridging technique (13) and verified by reverse transcription PCR and sequencing (data not shown).
The processing reaction was performed using this precursor transcript, and the resulting products were separated on a denaturing 10% polyacrylamide gel (Fig. 4). Because both tRNA Tyr and tRNA Cys have lengths of 66 nucleotides (including the overlapping nucleotide), it was not possible to distinguish between the two molecules simply by their migration position. Therefore, the product was isolated and subjected to a nearest neighbor analysis by RNase T2 digestion (14) (Fig. 5). This FIG. 2. Activity profile of mitochondrial S100 protein fractions. The fractions were obtained by cation exchange chromatography using a linear NaCl gradient (0 -0.80 M). Human tRNA Tyr /tRNA-Cys transcript (human Y/C) was tested as an overlapping precursor, whereas the human tRNA Ser /tRNA Leu RNA (human S/L) was offered as a nonoverlapping substrate. The maximum of each curve was set to 100%, and the other values were scaled accordingly. For both tRNA precursors, the activity appeared in identical fractions (0.24 -0.60 M NaCl), showing two peak activities. The overlapping A residue is shown in boldface. Bottom, processing of full-length or 3Ј-truncated tRNA Tyr /tRNA Cys precursors. Both versions give rise to processing products (arrow) of identical migration properties. M, tRNA Tyr /tRNA Cys nuclease P1 marker; mock, incubation of the transcripts without S100. B, top, structure of the tRNA Tyr precursor having the downstream tRNA Cys replaced by 30 nucleotides of the coxII mRNA in an overlapping or nonoverlapping situation (A insertion, indicated by the arrow). The overlapping A residue is shown in boldface. Bottom, processing of the tRNA Tyr /CoxII constructs in comparison to the original overlapping tRNA Tyr /tRNA Cys precursor. Both versions give rise to processing products (arrow) having migration properties identical to those of the tRNA Tyr /tRNA Cys transcript; mock, incubation of the transcripts without S100. enzyme leads to the release of nucleotide 3Ј-monophosphates (Fig. 5A, Np) and thus allows the 5Ј-neighboring base of the labeled nucleotide to be identified. Fig. 5A shows the reaction products expected for the possible processing products. If the tRNA Tyr is released as a complete molecule, it should carry the overlapping labeled base at its 3Ј-end. The T2 digest should therefore lead to a labeled cytosine 3Ј-monophosphate (an identical result is expected for the digestion of the precursor RNA), whereas a complete tRNA Cys should carry the overlapping base at its 5Ј-end and lead to a 5Ј-labeled adenosine 3Ј,5Ј-diphosphate (as would the RNase T2 digestion of the 5Ј-labeled tRNA-Cys used in the bridging experiment). The resulting mononucleotides were separated by thin layer chromatography, and the labeled bases were identified by autoradiography. Fig. 5B shows that the labeled nucleotide migrates at a position identical to the digestion product of the 5Ј-end-labeled tRNA Cys . It therefore indicates that the band in Fig. 4 corresponds to a complete tRNA Cys and that the upstream located tRNA Tyr is released in a truncated form lacking the discriminator base.
In order to further analyze the 5Ј-end of the processed tRNA-Cys obtained in the in vitro reaction as well as isolated from HeLa cell mitochondria, the 5Ј-end (carrying a phosphate group) and the 3Ј-end (carrying a 3Ј-OH group) of the tRNA were ligated by T4 RNA ligase, resulting in a circularized tRNA molecule (2). The acceptor stem of the tRNA was reverse transcribed, amplified by PCR, and cloned, and individual clones (representing the 5Ј-and 3Ј-termini of individual tRNA molecules) were analyzed by chain termination sequencing. All sequences showed that the tRNA Cys carries an A residue at the first position, indicating that this tRNA is found as a complete molecule not only under the described in vitro conditions ( Fig.  6) but also in vivo (data not shown).
The Truncated tRNA Tyr Is Completed by an Editing Reaction-Because the released tRNA Tyr is truncated at its 3Ј-end, the acceptor stem has to be completed in order to render the tRNA molecule functional. In vivo analysis revealed that this is indeed the case (data not shown), as shown also for overlapping tRNA transcripts in other organisms (2)(3)(4)(5): an A residue is incorporated at the discriminator position, and the CCA terminus is added. To investigate whether this incorporation activity is also present in the mitochondrial S100 protein extract, an in vitro transcript corresponding to a tRNA Tyr lacking the discriminator base (Fig. 7A; the 3Ј-end was verified by ligation of the transcript to DNA, cloning, and sequencing according to Ref. 12; data not shown) was incubated as a 33 P-5Ј-end-labeled version with the extract in the absence or presence of NTPs in FIG. 4. In vitro processing of a tRNA Tyr /tRNA Cys precursor carrying a labeled overlapping A residue. A tRNA Tyr /tRNA Cys (Y/C) precursor carrying a 32 P label at the overlapping position (marked by an asterisk) was incubated with S100 for 20 min (20Ј); the resulting product (indicated by the arrow) was isolated and subjected to further analysis. M, tRNA Tyr /tRNA Cys nuclease P1 marker; mock, incubation of the transcript without S100.

FIG. 5. Nearest neighbor analysis by RNase T2 digest.
A, schematic drawing of possible resulting products: digestion of the bridging product (tRNA Tyr /tRNA Cys (Y/C) precursor) leads to a radioactively labeled cytidine-3Ј-monophosphate. This is also expected for a tRNA Tyr carrying the overlapping labeled A residue at its 3Ј-end, whereas a complete tRNA Cys would give rise to a labeled adenosine-3Ј,5Ј-diphosphate. B, TLC analysis of T2 digested RNAs (solvent system: isopropanol:concentrated HCl:water, 70:15:15). Lane 1, bridging product (tRNA Tyr /tRNA Cys ) leads to cytidine-3Ј-monophosphate; lane 2, 5Ј-endlabeled tRNA Cys gives rise to adenosine-3Ј,5Ј-diphosphate. Lane 3, processing product (shown in Fig. 4): the resulting signal shows migration properties matching those of lane 2 (adenosine-3Ј,5Ј-diphosphate), indicating that the processing product is identical to tRNA Cys carrying the labeled overlapping nucleotide at the 5Ј-end. C, control lanes: bridging product digested by nuclease P1, leading to labeled adenosine-5Ј-monophosphate.
FIG. 6. Sequence comparison between cDNA clones of the released in vitro tRNA Cys and the corresponding genomic sequence. The genomic sequence shows the overlapping region between the genes for tRNA Tyr and tRNA Cys . The shared A residue is shown in boldface. In the aligned sequences of the cDNA clones, only the 5Ј-parts of the acceptor stems are shown. Identical positions are indicated by dots. All clones carry the overlapping A residue at position 1, indicating that this tRNA is released in a complete form. equimolar amounts (Fig. 7B). This truncated tRNA Tyr -1 carried a diagnostic mutation at position 52 (G52A), indicated in Fig.  7A by the asterisk (this mutation did not interfere with the analyzed reactions and allowed us to distinguish between substrate and endogenous tRNA Tyr found in the protein extract (data not shown)). Fig. 7B shows that in the presence of NTPs, the 3Ј-end is elongated by incorporation of four bases, leading to a fulllength product migrating at a position identical to that of the complete tRNA Tyr (tRNA(Tyr)ϩA) carrying a CCA end. In order to determine the nature of the incorporated nucleotides, the shifted bands of tRNA Tyr -1 were isolated, ligated to doublestranded DNA (12), amplified by reverse transcription PCR, and cloned, and individual clones were sequenced. This analysis revealed that CTP (7 of 14 clones, 50%) and ATP (4 of 14 clones, 29%) are readily accepted by the extract and incorporated at the discriminator position, whereas UTP (2 of 14 clones, 14%) and GTP (1 of 14 clones, 7%) are incorporated at a lower efficiency. An explanation for the misincorporation of a C residue at the discriminator position might be the high activity of the terminal nucleotidyl transferase in the extract, which adds the CCA terminus to the 3Ј-ends of tRNAs: in a control assay, a transcript representing the complete tRNA Tyr (carrying the discriminator base at the 3Ј-end) was used (tRNA(Tyr)ϩA). The observed base incorporation represents the addition of the CCA terminus and indicates that the terminal nucleotidyl transferase is highly active in the S100 extract.
In contrast to the addition of four bases by the human S100 extract, a yeast mitochondrial S100 protein extract incorporated only three residues at the 3Ј-end of the truncated tRNA-Tyr (as it did with the complete tRNA Tyr ϩA), corresponding probably to the CCA end (Fig. 7C). This indicates that the yeast extract is not able to incorporate the discriminator base in addition to the CCA terminus, a result that correlates with the fact that in yeast mitochondria, overlapping tRNA genes were not identified. DISCUSSION In animal mitochondria, many tRNA genes are transcribed as multicistronic precursors consisting of individual tRNA molecules arranged in tandem with no or only a few bases in between (6). In some of these primary transcripts, one tRNA not only abuts the next tRNA or protein coding sequence but overlaps by one to several bases (1,2). While nonoverlapping tRNA precursors are expected to be processed either by 5Ј-or 3Ј-cleavage (or both where spacers exist) (9, 21), overlapping precursors introduce a problem.
In principle, the release of complete versions of two overlapping tRNA molecules can be achieved in two ways. One possibility is an alternative processing reaction that leads to the formation of only one functional tRNA per primary transcript. Alternatively, it is conceivable that the processing takes place at one position in all transcripts, leading to the release of only one complete tRNA, whereas the other one is truncated. In the latter case, it is not clear how the primary transcript is recognized by the processing enzyme(s). It is imaginable that only one (either the upstream or the downstream) tRNA sequence in the precursor molecule serves as a recognition element. Another possibility is that both tRNAs are needed for a proper processing.
The established in vitro system described in this work served to investigate the processing reaction in more detail. In the analyzed reactions, the release of a truncated upstream and a complete downstream tRNA was detected. Because these findings are compatible with the in vivo results (data not shown), the in vitro system seems to contain the enzymatic activity that is responsible for the processing reaction. The fact that an overlapping precursor consisting of tRNA Tyr and only 31 bases of the downstream overlapping tRNA Cys is processed in the same way as the full construct (two complete tRNAs) indicates that the cleavage reaction is catalyzed by an enzyme that recognizes primarily the upstream tRNA. Furthermore, replacing the complete downstream tRNA Cys by a non-tRNA-like sequence (the 5Ј-part of coxII mRNA) did not interfere with the processing reaction and did not lead to a shift of the cleavage position, demonstrating that the upstream tRNA, exclusively, is recognized by the processing activity.
A 3Ј-truncated version of the tRNA Tyr /tRNA Cys precursor (carrying 36 nucleotides of the tRNA Cys ) very similar to the one used in Fig. 3A was tested in an in vitro system by Rossmanith et al. (9). In their system, the tRNA Tyr seemed to be liberated as a complete tRNA carrying the overlapping A residue at the FIG. 7. Base incorporation and completion of the 3-end of the truncated tRNA Tyr . A 33 P-5Ј-end-labeled in vitro transcript representing tRNA Tyr -1 (lacking the discriminator base) (A) was incubated in the presence of mitochondrial S100 extract and NTPs (B). The nucleotide marked by the asterisk in A represents an introduced diagnostic mutation that allows us to distinguish the substrate tRNA from the endogenous tRNA Tyr in the extract. The mutation did not affect the processing or base incorporation reactions (data not shown). B, although the tRNA is partially degraded in the presence of the extract, an incorporation of nucleotides can be observed that completes not only the tRNA Tyr ϩA, but also the 3Ј-truncated version (tRNA Tyr -1) to a full-length molecule carrying a discriminator base and the CCA end. The control experiment using the complete tRNA Tyr (tRNA(Tyr)ϩA, ending with the discriminator position) shows a high terminal nucleotidyl transferase activity in the protein extract. C, comparison of nucleotide incorporation by the human and the yeast mitochondrial S100 extracts. Whereas the yeast extract adds only three bases (corresponding to the CCA terminus) to the 3Ј-end of the tRNA Tyr -1 (lacking the discriminator base) (A), the human S100 incorporates an additional nucleotide. This indicates that only the human extract has the ability to incorporate a base at the discriminator position in addition to the CCA terminus.
3Ј-end. A reason for this discrepancy might be the different methods of analysis of the tRNA 3Ј-ends. Whereas in the described experiments, the analysis was carried out by direct characterization of the 3Ј-terminal base, Rossmanith et al. (9) compared the migration position of the released 3Ј-part of the tRNA Cys to a partially hydrolyzed RNA (alkaline RNA ladder) and to a tRNA precursor cleaved by nuclease T1. Because alkaline hydrolysis and T1 cleavage both produce 2Ј,3Ј-cyclic phosphate and 5Ј-hydroxyl groups, the migration properties of the resulting RNA fragments differ from those of the processing products, which have 5Ј-phosphate and 3Ј-hydroxyl groups. Furthermore, interpretation of migration properties, particularly of short RNA molecules, can be difficult because partial base pairing, as well as the base composition, may influence the electrophoretic mobility. Another explanation for the different results might be that the construct used by Rossmanith et al. (9) carried five additional bases at the 3Ј-end and that the incubation conditions were slightly different.
The in vivo and in vitro analyses presented show that the downstream tRNA is released as a full version carrying the overlapping part at its 5Ј-end (2,3,5), whereas all transcripts of the upstream tRNA analyzed to date are either missing the overlap (a likely intermediate) or have been completed posttranscriptionally by an enzymatic activity that is not yet characterized. In the described in vitro system, additionally to the CCA end, a single A residue is incorporated at the truncated 3Ј-end of tRNA Tyr , a reaction that delivers the discriminator base to the tRNA molecule. The efficiency of this fill-in editing mechanism is rather low, which corresponds to the in vivo situation (only 44% of the isolated and analyzed tRNA Tyr carried a completed 3Ј-terminus, whereas 56% were truncated; data not shown). Whereas in vivo, only A is found at the discriminator position of the tRNA Tyr , in the in vitro system, not only ATP but also CTP is readily incorporated. Because the used S100 extracts contain a highly active terminal nucleotidyl transferase, as shown in Fig. 7, B and C, it is conceivable that this activity in the human extract also acts on a tRNA substrate lacking the discriminator position at the 3Ј-terminus (tRNA Tyr -1), a reaction that was observed using the yeast mitochondrial extract (Fig. 7C): the gel shows that the human mitochondrial extract incorporates four nucleotides, whereas the yeast extract adds only three bases to the 3Ј-end of the tRNA Tyr -1 and is therefore obviously not able to incorporate an additional base besides the CCA end. An explanation for this difference might be the fact that the yeast mitochondrial genome does not contain overlapping tRNA genes. Therefore, a discriminator nucleotide incorporating activity is not needed.
The apparent gap in the bands shifted by the human extract indicates that the base incorporation at the discriminator position is the rate-limiting step in the completion reaction of the 3Ј-end of the tRNA. Because GTP and UTP are also incorporated at this position (although at a low efficiency), it is also possible that the fill in activity has a high error rate in the in vitro system. However, a misincorporation of CTP (or G and U residues) at the discriminator position was not observed in vivo, which might be due to either a higher fidelity of the system in the whole cell or to a excision reaction that removes the misincorporated C residue. A scenario is possible in which tRNA Tyr molecules possessing an A residue as a discriminator followed by the CCA terminus become protected against exonucleolytic activity by aminoacylation, whereas those variants carrying a misincorporated base at this position are not aminoacylated and underlie the excision reaction to get a new chance for the addition of the correct discriminator nucleotide.
A similar but probably unrelated base incorporation system has been described for mitochondrial tRNAs in the slime mold Acanthamoeba castellanii, in which mismatches in the acceptor stem of tRNA molecules are corrected by nucleotide conversions at the 5Ј-end (22). Whereas in the latter case, the 3Ј-part of the acceptor stem could function as a template, in the system described in this work, an unpaired nucleotide has to be incorporated at the 3Ј-end. Therefore, the activity cannot be guided by a simple template intrinsic to the tRNA. A possible candidate for such a base-incorporating activity is the poly(A) polymerase, which in a template-independent way adds adenosine residues to the 3Ј-end of mitochondrial RNAs (6). In chicken mitochondria, the tRNA Tyr (which overlaps by one nucleotide with the sequence for tRNA Cys ) was found to carry up to 13 extra A residues at the 3Ј-end as a possible reaction intermediate. On the basis of these observations, it has been suggested that the poly(A) polymerase might be involved in the addition of the poly(A) tail to this tRNA (4), although the addition of more than one A residue was never observed in the described in vitro system. A recent publication describes another enzyme that is possibly involved (23): in the nucleus and cytoplasm of human cells, several small RNAs, including U2, U6, 7SL, and 5S, carry a single not encoded A residue at the 3Ј-end. Sinha et al. (23) developed an in vitro system that is capable of adenylation of some of the transcripts. Whereas the function of this adenylation is not known, the fill-in reaction described here obviously restores the functionality of the truncated mitochondrial tRNA. A further indication for the involvement of one of these enzymes might be the fact that only tRNAs carrying adenosine as a discriminator base seem to overlap in their sequence with downstream genes (4).
A third candidate for a template independent base incorporating activity is the terminal nucleotidyl transferase that adds the CCA end to the 3Ј-end of tRNAs. It is known that this enzyme is able to repair partially degraded CCA ends at tRNA molecules (24). Because the acceptor stem of the released human tRNA Tyr carries two C residues (CC) at the 3Ј-end, it is conceivable that this structure mimics an incomplete CCA end, which could be recognized and repaired by the terminal nucleotidyl transferase. Because the tRNA would then carry a discriminator base, it would represent a conventional substrate for addition of an unpaired CCA terminus by the CCA-adding enzyme. Therefore, this scenario would imply the repair of one and the addition of another CCA end to an individual tRNA molecule. Whether this corresponds to the actual situation remains to be clarified. However, because only the tRNA Tyr carries this additional CC(A) sequence at the 3Ј terminus, whereas other overlapping tRNAs like tRNA Ser have a different base composition, this scenario would represent a special case.
In opossum mitochondria, the genes for the tRNA Ser (AGY) and tRNA Leu (CUN) overlap by one adenosine residue (25). Because a general processing machinery is believed to recognize all of the individual mitochondrial tRNA precursors, one would expect that a system that recognizes two overlapping tRNAs in a homologous assay (human mt proteins and human mt tRNA precursor) would also be able to handle a foreign tRNA precursor consisting of two different tRNAs having a similar overlapping nucleotide. However, the nuclease responsible for processing the human tRNA Tyr /tRNA Cys does not recognize the opossum transcript, although nonoverlapping tRNA Ser /tRNA Leu precursors (artificial and natural) are recognized and processed. Because both overlapping and nonoverlapping RNA substrates are cleaved by the same protein fractions obtained from the cation exchange chromatography and showed no different efficiencies in competition experiments, it seems that one individual activity exists in human mitochondria that is able to process conventional tRNA precursors lack-ing overlapping bases and seems to have coevolved with the overlapping situation found in the tRNA Tyr /tRNA Cys precursor; this activity is therefore specific for this individual overlap. A necessity for such a coevolution might be that the two overlapping tRNA acceptor stems in the precursor molecule acquire a three-dimensional structure that hinders RNase P in getting access to the cleavage position and renders such a precursor dependent on a 3Ј-endonuclease. Furthermore, a coevolution of this activity might explain why the overlapping opossum mt tRNA Ser /tRNA Leu precursor is not cleaved by the human extract: it is probably not accessible for RNase P and relies on an analogously coevolving 3Ј-endonuclease.
Interestingly, the observed activity releases a truncated and thus nonfunctional tRNA molecule, whereas all eukaryotic 3Јprocessing endonucleases known so far cleave tRNAs such that a complete molecule carrying the discriminator base is produced (24,26,27). Because the activity recognizes only the upstream tRNA Tyr , it is conceivable that this RNA carries features that guide the processing enzyme to the observed position, one nucleotide upstream of the discriminator position. A mammalian nuclear tRNA 3Ј-processing endonuclease can be directed to new cleavage sites downstream of the discriminator position by varying the number of base pairs in the acceptor and T-stem. 3 Although the human mt tRNA Tyr has no abnormal number of base pairs in these stems, it is possible that individual bases or base pairs can influence the cleavage site. It remains to be clarified which features of tRNA Tyr are responsible for guiding the endonuclease to the observed cleavage position.