In vitro characterization of a tRNA editing activity in the mitochondria of Spizellomyces punctatus, a Chytridiomycete fungus.

In the chytridiomycete fungus, Spizellomyces punctatus, all eight of the mitochondrially encoded tRNAs are predicted to have one or more base pair mismatches at the first three positions of their aminoacyl acceptor stems. These tRNAs are edited post-transcriptionally by replacement of the 5'-nucleotide in each mismatched pair with a nucleotide that can form a standard Watson-Crick base pair with its counterpart in the 3'-half of the stem. The type of mitochondrial tRNA editing found in S. punctatus also occurs in Acanthamoeba castellanii, a distantly related amoeboid protist. Using an S. punctatus mitochondrial extract, we have developed an in vitro assay of tRNA editing in which nucleotides are incorporated into various tRNA substrates. Experiments employing synthetic transcripts revealed that the S. punctatus tRNA editing activity incorporates nucleotides on the 5'-side of substrate tRNAs, uses the 3'-sequence as a template for incorporation, and adds nucleotides in a 3'-to-5' direction. This activity can add nucleotides to a triphosphorylated 5'-end in the absence of ATP but requires ATP to add nucleotides to a monophosphorylated 5'-end; moreover, it functions independently of the state of tRNA 3' processing. These data parallel results obtained in a previous in vitro study of A. castellanii tRNA editing, suggesting that remarkably similar activities function in the mitochondria of these two organisms. The evolutionary origins of these activities are discussed.

RNA editing encompasses particular forms of RNA processing that change the primary sequence of an RNA molecule from that predicted by the gene sequence. The term "RNA editing" was coined to describe post-transcriptional U insertions within the mitochondrial cox2 transcript of two trypanosome species: insertions that correct a gene-predicted frameshift, thereby generating a functional mRNA (1). Editing has since been found to be an essential process in many diverse, predominantly organellar, systems (2). Many cases of mRNA editing have been identified, including uridylate insertion/deletion in trypanosome mitochondria (3,4), co-transcriptional insertion of nucleotides in myxomycete mitochondria (5,6), C-to-U and U-to-C editing in plant mitochondria and chloroplasts (7), Ato-I (8,9) and C-to-U (10) base deamination in animal nuclei, co-transcriptional nucleotide insertion in viruses (11), and the recently discovered nucleotide replacement editing in dinoflagellate mitochondria (12). In addition to mRNAs, ribosomal RNAs (13)(14)(15) and transfer RNAs (tRNAs) (16) can be substrates for editing.
Editing of tRNA has been described in the mitochondria of many organisms. In plant mitochondria, C-to-U editing repairs mismatches in tRNA stems (17)(18)(19)(20). C and U insertion occurs within myxomycete mitochondrial tRNAs, creating base pairs in stems as well as restoring the GUUC sequence in the T stem-loop (21). In marsupial mitochondria, C-to-U editing alters the anticodon of mitochondrial tRNA Asp , changing the decoding potential (22)(23)(24). Similarly, C-to-U editing changes the decoding properties of trypanosome mitochondrial tRNA Trp (25). Non-templated editing within the 3Ј-half of tRNA acceptor stems, including the discriminator position, has been described in the mitochondria of several animals (26 -28). Further, in the centipede Lithobius forficatus, nucleotides are added to the 3Ј-end of mitochondrial tRNAs, apparently using the 5Ј-half of the acceptor stem as a template (29). Intriguingly, editing similar to that found in L. forficatus repairs mismatches in mitochondrial tRNA acceptor stems in the protist Seculamonas ecuadoriensis, a jakobid flagellate (30).
An additional type of editing, and the first example of tRNA editing to be described, occurs in the mitochondria of the amoeboid protist Acanthamoeba castellanii (31). In this organism, 12 of 15 mtDNA 1 -encoded tRNAs are predicted to contain mismatches in one or more of the first three base pairs of the acceptor stem. These mismatches are such that they appear incompatible with canonical tRNA folding and function. Direct sequencing of tRNA 5Ј-ends (31) as well as a tRNA circularization/RT-PCR approach (32) have revealed that these mismatches are repaired in vivo by removing the 5Ј-nucleotide from each mismatched pair and replacing it with a nucleotide that has the potential to form a Watson-Crick base pair with its counterpart in the 3Ј-half of the stem. The A. castellanii editing activity is hypothesized to consist of at least two components: an endonuclease and/or 5Ј-to-3Ј exonuclease that removes mismatched nucleotides from tRNA 5Ј-ends, and a template-directed 3Ј-to-5Ј nucleotidyltransferase activity that restores acceptor stem base pairing (31,33). An in vitro assay developed to study this form of editing (33) has demonstrated nucleotide incorporation into various tRNA substrates. Nucleotide addition occurs at recessed tRNA 5Ј-ends, requires ATP when the 5Ј-terminus is monophosphorylated, and proceeds in a 3Ј-to-5Ј direction using the 3Ј-half of the acceptor stem as a template.
In Spizellomyces punctatus, a chytridiomycete fungus, all eight of the mtDNA-encoded tRNAs are predicted to contain mismatches at one or more of the first three acceptor stem positions (34). These mismatches are repaired in vivo, as in A. castellanii, by replacement of the 5Ј-nucleotide in the pair to form Watson-Crick base pairs, as shown by direct sequencing of tRNA 5Ј-ends (34) and tRNA circularization/RT-PCR (35). Other chytridiomycetes (but not all) have been shown to have this type of tRNA editing (35). Editing in S. punctatus is highly similar in effect to that in A. castellanii, a quite unexpected result considering the lack of a specific evolutionary relationship between chytridiomycete fungi and this amoeboid protist.
In the present study, we first demonstrate and then examine an enzymatic activity in S. punctatus mitochondrial extracts that incorporates nucleotides into tRNA molecules, an activity that has the properties expected of the in vivo mitochondrial tRNA editing activity. We compare the biochemistry of the S. punctatus and A. castellanii editing activities and discuss how they may have evolved.

MATERIALS AND METHODS
Culture Conditions and Preparation of Mitochondrial S100 Extracts-S. punctatus strain BR117 was kindly provided by B. F. Lang (Université de Montréal), maintained on agar plates, and grown in a liquid medium containing 0.5% yeast extract and 3% glycerol (adjusted to pH 5.8 with KH 2 PO 4 ). Cells were grown with hard shaking (ϳ100 rpm) in 100-ml cultures for ϳ48 h until the majority of the culture consisted of zoospores, which were found to be much less resistant to lysis than mature cells. Swimming zoospores were slowed by a 15-min incubation on ice, pelleted by centrifugation at 2000 ϫ g, and washed two times with 500 ml of phosphate-buffered saline (4.3 mM Na 2 HPO 4 , 1.4 mM KH 2 PO 4 , 137 mM NaCl, 2.7 mM KCl). Cells from 1 liter of culture (1-3 g of cells were routinely obtained) were mixed with 10 ml of homogenization buffer (0.35 M sucrose, 50 mM Tris (pH 8.0), 3 mM EDTA, 1 mM dithiothreitol, 0.1% bovine serum albumin) and re-suspended with a teflon homogenizer.
Cells were disrupted by passage through a French pressure cell at 1000 psi. If Ͼ80% cell breakage was not observed by microscopy, cells were passed through the cell a second time. Cell debris was removed by centrifugation at 440 ϫ g for 10 min, and the supernatant was centrifuged at 1780 ϫ g for a further 10 min. The second supernatant was then centrifuged at 9000 ϫ g for 20 min, yielding a pellet containing intact mitochondria. The supernatant from this centrifugation step was retained for the preparation of cytoplasmic RNA (see below). The pellet was resuspended in 10 ml of homogenization buffer, and residual cellular debris was removed by centrifugation at 1780 ϫ g for 10 min. The supernatant was then centrifuged at 9000 ϫ g for 20 min, and the mitochondrial pellet was washed twice by resuspension in 10 ml of homogenization buffer followed by centrifugation at 9000 ϫ g for 20 min.
Mitochondria were resuspended in 1 ml of sonication buffer (7.5% glycerol, 10 mM Tris, pH 8.0, 0.1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) and disrupted by sonication with a Fisher Scientific Sonic Dismembrator Model 100 at 1/3 maximum power for 3 ϫ 30 s, 1 ϫ 20 s, with 1-min rests between bursts. The sonicate was subjected to ultracentrifugation at 100,000 ϫ g for 1 h in a fixed angle rotor, and the supernatant (S100) was snap frozen in liquid nitrogen and stored at Ϫ70°C in 50-l aliquots. Alternatively, RNA was prepared from purified, intact mitochondria (see below). S100 fractions were active for nucleotide incorporation for more than a year under these conditions. For most experiments, the S100 was incubated on ice for 20 min with 1/5 volume of QAE-Sephadex A-50 strong anion exchange resin (ion exchange group: diethyl-(2-hydroxy-propyl)aminoethyl; stored in sonication buffer) prior to in vitro assays.
Preparation of Mitochondrial and Cytoplasmic RNA-Purified mitochondria were gently lysed in a solution containing 2% Triton X-100, 10 mM Tris-HCl, pH 8.5, 50 mM KCl, and 10 mM MgCl 2 , centrifuged at 9000 ϫ g for 10 min, and nucleic acids were prepared from both the resulting pellet (mito-P) and supernatant (mito-S) fractions. Cytoplasmic RNAs were prepared from the supernatant of the first 9000 ϫ g spin in the protocol for preparation of a mitochondrial S100 fraction (see above). RNA was isolated as described (36). To assess mitochondrial enrichment, 5 g of RNA from each fraction was treated with 2.5 units of DNase I for 30 min and resolved by agarose gel electrophoresis.
In Vitro Assay of Nucleotide Incorporation-Each assay contained 2.5 l of the mitochondrial S100 extract, 40 mM HEPES (pH 7.0 with NaOH), 15 mM MgCl 2 , 1 mM dithiothreitol, 100 M NTPs (when included), 10 Ci of [␣-32 P]NTP (when included), and exogenous tRNA substrate (when included; RNA was either 1 g of yeast soluble RNA, 1/20 of the product of a T7 transcription reaction, or a small portion of a 32 P-end-labeled T7 transcription product). Reactions (in a final volume of 50 l) were performed at room temperature (ϳ22°C) for 1 h (or less, as indicated in the text). Products of in vitro assays were extracted with phenol-cresol and precipitated with ethanol using linear polyacrylamide as carrier (37). Samples were electrophoresed through 10% polyacrylamide gels containing 7 M urea (36).
Preparation of Synthetic tRNA Substrates-Oligonucleotides used in this study are shown in Table I. The A. castellanii mitochondrial tRNA Leu (uag) gene cloned into the pT7Blue vector (see Ref. 33; referred to here as plasmid 1) was used to generate the PCR templates for T7 RNA polymerase transcription, yielding tRNA constructs 1 and 4 -9 (see Figs. 3 and 4A for sequences of constructs). The PCR templates for tRNA constructs 2 and 3 were generated by first introducing 5Ј or 3Ј extensions, respectively, into the cloned tRNA Leu in plasmid 1. This was accomplished by PCR using divergently oriented primers, to generate plasmid 2 (using oligonucleotides 1 and 2) and plasmid 3 (using oligonucleotides 3 and 4). PCR amplification was performed with Vent DNA polymerase in all cases. The linearized plasmids 2 and 3 thus produced were phosphorylated with T4 polynucleotide kinase and circularized with T4 DNA ligase. Plasmids 1, 2, and 3 were cloned in competent Escherichia coli cells (strain DH5-␣), and plasmid DNA was prepared using an alkaline lysis procedure. The sequence of the tRNA Leu region of the plasmids was confirmed by DNA sequencing.
Short DNA templates for in vitro transcription reactions with T7 RNA polymerase were produced using PCR with primers designed to incorporate specific sequences at the 5Ј-and 3Ј-termini of these templates. Plasmids were treated with the endonuclease HindIII (which cuts outside the tRNA Leu region in the multiple cloning site) prior to amplification in order to eliminate longer, unwanted products observed in PCR reactions using uncleaved plasmids. T7 templates for synthesis of tRNA constructs 1-9 were generated using the oligonucleotide combinations shown in Table II. Each T7 template contained a 6-nt 5Јleader, a T7 promoter, and the desired tRNA Leu sequence. The products of PCR were extracted with phenol-cresol, precipitated with ethanol, and 1/10 of each product was used as template for each 50 l of T7 in vitro transcription reaction. Products of in vitro transcription were treated with 2.5 units of DNase I, separated in 10% denaturing polyacrylamide gels, excised, and eluted. An approximately equal amount of T7 product (several micrograms) was generated for each construct, as estimated by ethidium bromide staining of polyacrylamide gels. End Analysis of RNA-Transfer RNAs labeled by the S100 extract in the presence of [␣-32 P]GTP were eluted from an homogenized gel slice by shaking overnight at 4°C in a 1:1 mixture of phenol-cresol:buffer [0.5 M NH 4 OAc, 10 mM Mg(OAc) 2 , 1.0 mM EDTA] (36). Extracted RNAs were precipitated with ethanol and treated with either nuclease P1, which generates nucleoside 5Ј-phosphates (pN), or RNase T2, which generates nucleoside 3Ј-phosphates (Np). The products of nuclease P1 digestion were separated by thin-layer chromatography (TLC) using polyethyleneimine cellulose plates (predeveloped with distilled H 2 O and allowed to dry) and 0.5 M (NH 4 ) 2 CO 3 as the solvent. The products of RNase T2 digestion were separated by TLC on cellulose plates (predipped in a 10-fold dilution of a saturated solution of (NH 4 ) 2 SO 4 and allowed to dry) using a 4:1 mixture of 95% ethanol:water as the solvent (38).
Partial sequences of the 3Ј-and 5Ј-portions of 3Ј-and 5Ј-end-labeled RNAs were determined by separation of the products of partial alkaline hydrolysis and partial digestion with RNase T1 (which cleaves 3Ј to G residues) in adjacent wells of a 20% polyacrylamide gel. Further 3Ј-end analysis was performed by TLC separation of an RNase T2 digest of 3Ј-end-labeled material (see above for TLC conditions).
Periodate Treatment of RNA-One-twentieth of the RNA product of a T7 transcription reaction or a small portion of a 32 P-end-labeled T7 transcription product was dissolved in a 40-l solution containing 150 mM NaOAc (pH 5.3) and 1 mM NaIO 4 , incubated on ice in the dark for 1 h, and precipitated with ethanol.

Development of an in Vitro
Assay of Nucleotide Incorporation-Intact mitochondria were recovered from S. punctatus zoospores by lysis under relatively mild conditions in a French pressure cell followed by differential centrifugation. This procedure resulted in significant mitochondrial enrichment, as estimated by the ratio of mitochondrial to cytoplasmic rRNAs in mitochondrial RNA profiles (Fig. 1). Purified mitochondria were disrupted by sonication, and an S100 fraction was obtained by ultracentrifugation of the sonicate.
In order to test for a tRNA editing activity in the S. punctatus S100 fraction, an in vitro assay developed to study A. castellanii 5Ј-tRNA editing (33) was employed. This assay monitors the incorporation of [␣-32 P]NTPs into tRNA molecules following incubation with a mitochondrial extract. As in A. castellanii, incubation of the S. punctatus mitochondrial S100 fraction in the presence of [␣-32 P]GTP resulted in the incorporation of label into tRNAs present in the S100 fraction ( Fig. 2A, lane 1). Yeast tRNAs were also found to be substrates for the labeling activity ( Fig. 2A, lane 2).
Nucleotide incorporation by the S. punctatus mitochondrial S100 fraction was initially observed using the assay conditions optimized for the A. castellanii editing activity (33). Optimal temperature and salt concentration for the S. punctatus activity were subsequently determined (not shown): the activity was found to function equally well from room temperature (ϳ22°C) to 37°C, with reduced activity at 42°C and loss of activity at 48°C. Added monovalent cations (in the form of KCl) were not required and did not affect labeling efficiency. Added divalent cation was required, with peak activity at 15 mM MgCl 2 and loss of activity at Ն30 mM MgCl 2 . The results in Fig. 2A were obtained using conditions optimized for the S. punctatus mito-  chondrial nucleotide incorporation activity.
Following optimization, endogenous tRNAs were removed from the S100 by incubation of the extract with QAE-Sephadex A-50 anion exchange resin, rendering this fraction dependent on addition of exogenous tRNA substrates (compare lanes 3 and 4, Fig. 2A). QAE-Sephadex-treated S100 fractions (hereinafter referred to as "the S100 extract") were used for all subsequent experiments.
When a nuclease P1 digest (which generates nucleoside 5Јphosphates, pN) of yeast tRNAs labeled with [␣-32 P]GTP was analyzed by thin-layer chromatography (TLC), the majority of the label was found to be incorporated as a nucleoside triphosphate (pppG), with a minority incorporated as a nucleoside monophosphate (pG) (Fig. 2B). This experiment indicated that nucleotide incorporation had occurred both internally (possibly at position 2) and at tRNA 5Ј-ends. The result also indicated that 5Ј-ends are left as triphosphates by the activity in vitro, as is also the case for the A. castellanii activity (33). Consistent with incorporation at tRNA positions 1 and 2, TLC analysis of an RNase T2 digest (which generates nucleoside 3Ј-phosphates, Np) of the same material revealed only a product that did not migrate from the origin (Fig. 2C). Because multiphosphorylated nucleotides are known to migrate little, if at all, from the origin in the TLC conditions used (33), it was possible that this product represented pppGp, the only product expected if all label was incorporated at positions 1 and 2 (Fig. 2D). Based on these results, the activity in the S. punctatus mitochondrial S100 extract was further characterized as a potential tRNA editing activity analogous to that described in A. castellanii.
Characterization of Nucleotide Incorporation using in Vitro Transcripts-To examine in more detail the mechanism of nucleotide incorporation into tRNA molecules, in vitro-transcribed tRNA constructs, based on an A. castellanii mitochondrial tRNA Leu (uag) scaffold (Fig. 3), were produced using T7 RNA polymerase. Specific sequence changes were engineered into the 5Ј-and 3Ј-side of the acceptor stem of this tRNA to create RNA constructs (Fig. 4A) that were then tested as substrates for the S. punctatus nucleotide incorporation activity. Transcripts were incubated with the S. punctatus mitochondrial S100 extract and either [␣-32 P]GTP or [␣-32 P]UTP, and the products were separated by denaturing polyacrylamide gel electrophoresis (Fig. 4B). The labeling results shown in Fig. 4B were consistent with the expected positions and identities of incorporated nucleotides (indicated in Fig. 4A).
Construct 1 was designed to direct G incorporation at positions 2 and 1 (Fig. 5A), templated by C71 and C72, respectively. Construct 1 was labeled by incubation with the S100 extract in the presence of [␣-32 P]GTP (Fig. 4B, lane a). TLC analysis of a nuclease P1 digest of the largest labeled band (indicated with an arrow in Fig. 4B, lane a) revealed that incorporation was 1 pG to 1 pppG (Fig. 5B). TLC analysis of an RNase T2 digest of the same material gave a product that did not migrate from the FIG. 3. Gene-predicted sequence and potential secondary structure of A. castellanii mitochondrial tRNA Leu (uag). A1xC72 and U2xC71 mismatches are both edited in vivo to G-C pairs (31,32). Standard numbering of nucleotides at tRNA 5Ј-and 3Ј-termini is indicated.

FIG. 4. Incorporation of label from [␣-32 P]GTP and [␣-32 P]UTP into in vitro-transcribed tRNA substrates.
A, tRNA constructs generated by in vitro transcription with T7 RNA polymerase. Only acceptor stem sequences are shown; curved lines represent the remainder of the tRNA sequence, shown in Fig. 3, which is identical for each construct. Arrows indicate nucleotides predicted to be incorporated after incubation with the S. punctatus mitochondrial S100 extract. B, autoradiograms of tRNA constructs 1-9 after incubation with the S100 extract, [␣- 32   A, acceptor stem sequence of construct 1. 32 P-labeled G residues are predicted to be incorporated at positions 1 and 2 after incubation with the S100 extract. Asterisks indicate the positions of 32 P. Following incubation of construct 1 with the S100 extract in the presence of [␣-32 P]GTP and unlabeled CTP, ATP, and UTP, the longest labeled product (indicated with an arrow in Fig. 4, lane a) was eluted from the gel and digested with either nuclease P1 (B) or RNase T2 (C), and the products were separated by one-dimensional TLC. The migration positions of nucleotide markers are indicated. origin (Fig. 5C), assumed to be pppGp (see above). These results indicated that nucleotide incorporation was occurring as expected at the 5Ј-end of this construct, specifically at positions 2 and 1 (see Fig. 2D).
Construct 2 (Fig. 4A) had a 13-nt 5Ј extension relative to construct 1. No radioactive products were observed after incubation of this RNA with [␣-32 P]GTP and the S100 extract (Fig.  4B, lane b). The absence of labeled products suggests that processing of tRNA 5Ј extensions occurs inefficiently or not at all in the S. punctatus mitochondrial extract with this particular substrate. In contrast, label from [␣-32 P]GTP was incorporated into construct 3 (Fig. 4B, lane c), which had a 15-nt 3Ј extension relative to construct 1 (Fig. 6A). This result suggests that the 5Ј-nucleotide incorporation activity is relatively insensitive to the state of 3Ј processing.
Many shorter labeled products of construct 3 in addition to the expected 96-nt (94 nt ϩ 2 added 32 P-labeled G residues ϭ 96 nt) full-length product were observed (Fig. 4B, lane c), suggesting that degradation of the 3Ј-end of this construct was occurring during incubation with the extract. To test this possibility, the construct was subjected to periodate oxidation, a treatment that breaks the ribose ring of the 3Ј-nucleotide in an RNA molecule, thereby converting the 2Ј,3Ј cis-diol to vicinal aldehydes. This treatment is expected to render the product a poor substrate for 3Ј-to-5Ј exonuclease activities. As anticipated, the amount of label in the shorter products was significantly reduced by periodate oxidation (Fig. 6B), with the majority of label now appearing in the expected full-length product. These results indicate (i) that the activity recognizes and incorporates 2 nucleotides at the 5Ј-end of construct 3 despite the absence of 3Ј processing; (ii) that the 3Ј extension does not act as a template for further nucleotide addition at the 5Ј-end past position 1; and (iii) that an unidentified 3Ј-to-5Ј exonuclease is active in the S100 extract.
Construct 4 has an additional G at position 2 and construct 5 has additional G nucleotides at positions 2 and 1, relative to construct 1 (Fig. 4A). Construct 4 was designed to incorporate one G only at position 1, whereas construct 5 was designed to require no additional nucleotides, having a completely base paired acceptor stem. As expected, construct 4 was found to be a very good substrate for the nucleotide incorporation activity (Fig. 4B, lane d). Construct 5, on the other hand, was a poor substrate for the activity (Fig. 4B, lane e). These results, together with those for construct 1, indicate that the activity adds nucleotides up to and including tRNA position 1, whether 1 or 2 nucleotides are missing from the 5Ј-end. When the 5Ј-terminal sequence of the added synthetic tRNA is complete, no nucleotides are incorporated.
cis-Templated 3Ј-to-5Ј Nucleotide Incorporation-Constructs 6 -9 (Fig. 4A) were designed to directly test whether the sequence on the 3Ј-side of the acceptor stem templates nucleotide addition to the 5Ј-side of the acceptor stem, and whether addition proceeds in a 3Ј-to-5Ј direction, as it does in the case of tRNA editing in A. castellanii mitochondria (33).
Construct 6 was designed to incorporate U at position 2 and G at position 1. Labeled products were observed after incubation with the extract in the presence of [␣-32 P]GTP and UTP (Fig. 4B, lane f). When UTP was not present during incubation, no labeled products were observed (Fig. 4B, lane g). These results support addition at position 2 followed by addition at position 1: i.e. sequential incorporation in a 3Ј-to-5Ј direction. When [␣-32 P]UTP was used, labeling occurred in both the presence and absence of GTP; however, the product formed in the absence of GTP (lane o) was shorter than in its presence (lane n), as expected if U addition at position 2 is required for G addition at position 1.
Construct 7 was designed to incorporate G at position 2 and U at position 1. Labeled products were observed with [␣-32 P]GTP in both the presence and absence of UTP, with the product in the absence of UTP (Fig. 4B, lane i) shorter than when UTP was present (lane h). When [␣-32 P]UTP was used, labeling occurred in the presence of GTP (lane p) but not in its absence (lane q), again supporting the inference of incorporation at position 2 followed by incorporation at position 1.
Construct 8 was designed to require the removal of a G at position 2 (which is involved in a G2xA71 mismatch) prior to addition of U at position 2 and G at position 1. However, no labeled products were observed for this construct following incubation with the extract and either [␣-32 P]GTP (Fig. 4B,  lanes j and k) or [␣-32 P]UTP (lanes r and s). These results, together with those obtained for construct 2, suggest that nucleotides are not efficiently removed from the 5Ј-ends of these synthetic tRNA substrates in this assay (see "Discussion"), and that the in vitro activity is only able to extend a 5Ј-terminus that forms a standard base pair with its partner nucleotide on the 3Ј-side of the acceptor stem.
Finally, construct 9 was designed to incorporate a U at position 1. This construct was not labeled with [␣-32 P]GTP (Fig. 4B, lanes l and m), but gave the same-sized labeled products with [␣-32 P]UTP in the presence (lane t) and absence (lane u) of GTP, as expected. These combined results strongly support a mechanism of nucleotide addition to tRNA 5Ј-ends, templated by the 3Ј-side of the acceptor stem and proceeding in a 3Ј-to-5Ј direction.
Nucleotide Requirements for Incorporation-To determine the nucleotide requirements of the incorporation activity, construct 1 (Fig. 4A) was incubated with [␣-32 P]GTP, the S100 extract and various combinations of CTP, ATP and UTP. The latter three NTPs were not required for incorporation of label from [␣-32 P]GTP into this construct (Fig. 7A, lanes b-i). The size and intensity of the labeled products, however, varied depending on which of these NTPs were included in the reaction mix. Through subsequent analysis, the observed size variation was attributed to an activity incorporating C residues at FIG. 6. Effect of periodate treatment on incorporation of label from [␣-32 P]GTP into tRNA construct 3. A, acceptor stem sequence of construct 3. 32 P-labeled G residues are predicted to be incorporated at positions 1 and 2 after incubation with the S100 extract. Asterisks indicate the positions of 32 P. The 15-nt 3Ј extension (relative to construct 1) is boxed. B, autoradiogram of construct 3, untreated (Ϫ) or pretreated (ϩ) with sodium periodate (NaIO 4 ) prior to incubation with the S100 extract in the presence of [␣-32 P]GTP and unlabeled CTP, ATP, and UTP. The predominant labeled band obtained with periodatetreated tRNA construct 3 (96 nt) is indicated with an arrow. tRNA 3Ј-ends (see below). The reason(s) for variation in labeling intensity remain(s) unknown, as further experiments did not indicate any role for CTP (see Fig. 9), UTP (not shown) or ATP (at least when the recessed tRNA 5Ј-end is triphosphorylated; see below) in this reaction.
The results shown in Fig. 7A were obtained with an untreated T7 transcript having a triphosphorylated 5Ј-end. When the 5Ј-end was converted from a triphosphate to a monophosphate by dephosphorylation with alkaline phosphatase followed by rephosphorylation using polynucleotide kinase, ATP was then required for incorporation of label from [␣-32 P]GTP (Fig. 7B, compare lanes l and m). No labeling was observed when the tRNA substrate had a 5Ј-hydroxyl terminus (not shown). These results indicate that an activated 5Ј-end is necessary for nucleotide incorporation, and that ATP can provide the necessary activation energy when the 5Ј-end of a tRNA substrate is monophosphorylated. It is likely that this activation energy comes in the form of an adenylylated tRNA intermediate (see "Discussion").
In order to determine how many nucleotides were incorporated into construct 1 after incubation with the extract and [␣-32 P]GTP, the T7 transcript was first 5Ј-end-labeled with polynucleotide kinase and [␥-32 P]ATP and then electrophoresed next to [␣-32 P]GTP-labeled material. Two labeled products in a 1:1 ratio (instead of the single expected 79-nt product) were observed in the 5Ј-end-labeled T7 transcription reaction (Fig.  7A, lane a). RNA sequencing (not shown) revealed that the shorter of the two T7 products (1-S) had the desired 79-nt sequence, whereas the longer product (1-L) had 1 additional nucleotide (a mixture of G, A, U, and C) at its 3Ј-end. Transcript 1-L is likely the result of the T7 RNA polymerase adding a random nucleotide at the 3Ј-end of a portion of its products, as has been documented previously (e.g. Ref. 41). Following incubation of construct 1 with the S100 extract and [␣-32 P]GTP, and using 1-S and 1-L as size standards, the largest labeled product (in Fig. 7A, lane b) was found to be 83 nucleotides in length.
Dissociation of 5Ј-and 3Ј-Nucleotide Incorporation Activities-5Ј-End-labeled 1-L and 1-S were separated from one another by polyacrylamide gel electrophoresis. The two products were shown individually to be substrates for the nucleotide incorporation activity (Fig. 8, A and B). After a sufficient period of incubation (ϳ30 min), the principal products with the two RNAs were single species, 83 nucleotides in length; i.e. the product after incubation of 1-L was three nucleotides longer (Fig. 8A) and that of 1-S four nucleotides longer (Fig. 8B) than the corresponding unincubated material, rather than the expected two nucleotides for each RNA. This result confirmed the observation that after incubation of unlabeled construct 1 with the S100 extract in the presence of [␣-32 P]GTP and CTP, the largest product was 83 nucleotides in length (Fig. 7A, lane f) rather than the expected 81 nucleotides. Further, in the absence of CTP, the principal labeled product was only 81 nucleotides in length (Fig. 7A, lane g).
Thus, it appeared that an activity additional to the 5Ј-nucleotide incorporation activity was present in the S100 extract, and that this activity was capable of incorporating C residues into tRNA substrates. A well known activity that catalyzes this type of reaction is ATP(CTP):tRNA nucleotidyltransferase, the ubiquitous enzyme that adds the 3Ј-terminal CCA sequence to tRNA. To test the idea of nucleotide addition to tRNA 3Ј-ends, 1-S (5Ј-end-labeled) was subjected to periodate oxidation and then incubated with the S100 extract and all four NTPs. After incubation, NaIO 4 -treated 1-S was 81 nucleotides in length; i.e. only two nucleotides longer than the unincubated material ( Fig. 8C; note that NaIO 4 -treated RNA species migrate slightly more slowly than the corresponding untreated RNAs). This result is consistent with the interpretation that periodate oxidation blocks the addition of two nucleotides to the 3Ј-end of this particular tRNA substrate.
To further confirm that CTP was the nucleotide whose addition had been blocked by the periodate treatment, 5Ј-end-labeled 1-S (untreated or treated with NaIO 4 ) was incubated with the S100 extract in the presence or absence of CTP. Whereas the size of the untreated product was two nucleotides shorter in the absence (81 nt) than in the presence (83 nt) of CTP (Fig. 9A, compare lanes b and c), the NaIO 4 -treated product remained the same size (81 nt) in the presence and absence of CTP (Fig. 9B, lanes g and h). The absence of CTP and either GTP or ATP resulted in no increase in the size of the labeled material (79 nt) following incubation with the S100 extract, whether or not the substrate had been treated with NaIO 4 prior to incubation (Fig. 9, A, lanes d and e; B, lanes i and j). These results, together with other data presented here, strongly support the idea that two G residues are added to the 5Ј-end of this tRNA in a reaction that requires ATP, and that two C residues are added to the 3Ј-end of the tRNA construct, likely by the mitochondrial ATP(CTP):tRNA nucleotidyltransferase. Why the full CCA sequence is not added to the 3Ј-end of the construct is not known; presumably in vitro conditions for the addition of the 3Ј-terminal A residue are suboptimal. DISCUSSION An in Vitro Assay of Mitochondrial tRNA Editing in S. punctatus-We have obtained a mitochondrial extract from S. punctatus, a chyridiomycete fungus, that supports specific and efficient incorporation of nucleotides into natural and in vitrotranscribed tRNAs. Incorporation is at the 5Ј-end of tRNA substrates at both internal and 5Ј-terminal positions (positions 2 and 1), templated by the 3Ј-half of the acceptor stem (positions 71 and 72) with nucleotide addition proceeding in a 3Јto-5Ј direction, leaving triphosphorylated 5Ј-ends in vitro. These features are consistent with this activity being the one that carries out in vivo editing of mitochondrial tRNAs. The development of an in vitro assay is an important step in the study of this putative editing activity, particularly as a tool to monitor the progress of any future enzyme purification.
The results obtained here are remarkably similar to those obtained with a previously reported in vitro assay of mitochondrial tRNA editing in the amoeboid protozoon, A. castellanii (33). The determination of mature tRNA acceptor stem sequences from S. punctatus and A. castellanii had demonstrated that these activities corrected very similar patterns of pre- dicted base pair mismatches in tRNA acceptor stems, by 5Јnucleotide replacement to produce standard Watson-Crick base pairs (31,32,34,35). The results in the present study indicate that these activities are not only similar in effect, but also in the biochemistry of the reactions they catalyze. The replacements mediated by the two activities are directed by the sequence of the 3Ј-half of the acceptor stem, both activities incorporate nucleotides in a 3Ј-to-5Ј direction, and both require a 5Ј-triphosphate to begin nucleotide incorporation (or a 5Ј-monophosphate plus ATP).
It will be of considerable interest to determine which enzymes are responsible for this form of 5Ј-tRNA editing, particularly as editing appears to involve a novel 3Ј-to 5Јnucleotidyltransferase. The sequences of these enzymes may reveal how the activity recognizes its substrates and carries out catalysis, by comparison with other enzymatic activities that carry out similar reactions. For example, histidine tRNA guanylyltransferase (HTGT) is the enzyme responsible for the non-templated, post-transcriptional addition of a G residue at the Ϫ1 position of tRNA His in all eukaryotes (42)(43)(44).
HTGT is the only activity other than the mitochondrial 5Ј-tRNA editing activities known to extend a polynucleotide chain in a 3Ј-to-5Ј direction via formation of a normal phosphodiester bond. To accomplish this extension, HTGT first activates the 5Ј-end of 5Ј-monophosphorylated tRNA His by adding an adenylate (AMP) moiety to the tRNA, forming a 5Ј-5Ј phosphoanhydride bond. The covalently bound AMP residue is subsequently displaced by the attack of the 3Јhydroxyl of the GTP that provides the incorporated GMP moiety. Activation of monophosphorylated 5Ј-ends by adenylylation is also employed by DNA and RNA ligases (45,46). Because the S. punctatus and A. castellanii mitochondrial activities catalyze a reaction similar to the one mediated by HTGT, both activities requiring ATP for nucleotide addition to monophosphorylated tRNA 5Ј-ends (likely via an adenylylated tRNA intermediate), it is possible that HTGT and the 5Ј-tRNA nucleotide incorporation activities have common ev-olutionary origins as well as biochemical mechanisms.
Absence of 5Ј-tRNA Processing Prior to Nucleotide Incorporation-Mitochondrial genomes are generally transcribed to produce several multigene transcripts from which tRNAs are processed. In all systems examined to date, the 5Ј-ends of mature tRNAs are generated by RNase P, an essential endonuclease that cleaves the phosphodiester bond between the nucleotides at positions 1 and Ϫ1 (in most cases), thereby removing 5Ј extensions and producing monophosphorylated 5Јends. Evidence for S. punctatus mitochondrial tRNAs being processed from large RNA transcripts has been presented (34). However, as all mtDNA-encoded tRNAs in S. punctatus undergo 5Ј editing, processing by RNase P between positions 1 and Ϫ1 would have to be followed by the further removal of nucleotides at the first three 5Ј-positions before the nucleotide incorporation activity described in the present study could restore proper acceptor stem base pairing. Alternatively, S. punctatus mitochondrial RNase P may have an altered cleavage site (i.e. between positions 4 and 3) thus removing the requirement for additional 5Ј nuclease events prior to nucleotide incorporation. In support of this idea, RNase P has been shown to cleave at sites other than between positions 1 and Ϫ1 in some systems. For example, RNase P cleaves 5Ј to position Ϫ1 in certain prokaryotes to generate the mature tRNA His (47).
In a recent analysis of cDNA sequences of acceptor stems produced by RT-PCR from circularized tRNAs in three chytridiomycetes (Monoblepharella15, Harpochytrium94 and Har-pochytrium105), partially and completely unedited tRNAs were found that support the existence of a 5Ј-to-3Ј exonuclease acting to remove 5Ј-nucleotides at tRNA positions 1-3, prior to 5Ј-nucleotide incorporation (35). In contrast, analysis of cDNAs of tRNA acceptor stems from S. punctatus mitochondria did not reveal any partially or completely unedited tRNAs (35). Similarly, no partially or completely unedited tRNAs were reported from cDNA studies of A. castellanii mitochondrial tRNAs (32). The possibility of an altered RNase P cleavage site in the latter two systems therefore remains an intriguing possibility.
Somewhat surprisingly, the in vitro system studied here yielded no strong evidence for the removal of 5Ј-nucleotides from tRNA substrates prior to nucleotide incorporation. For example, construct 2 (containing a 13-nt 5Ј extension; Fig.  4B, lane b), construct 5 (containing a fully base-paired acceptor stem; Fig. 4B, lane e) and construct 8 (containing a mismatch at base pair 2:71; Fig. 4B, lanes j, k, r, and s) were not labeled to a significant degree after incubation with [␣-32 P]GTP and the S. punctatus mitochondrial S100 extract. Further, radioactivity in 5Ј-end-labeled construct 1 is retained after incubation with the extract (Fig. 8). Suggestive evidence of a 5Ј nuclease activity was only obtained when yeast tRNAs were used as substrate: 32 P-labeled G residues were then found to be incorporated at tRNA positions 1 and 2 (Fig. 2, B and C). However, the incorporated nucleotides at these positions may constitute addition to degraded tRNA 5Ј-termini, and therefore this is not definitive evidence of FIG. 8. Autoradiogram of 5-end-labeled tRNAs after incubation (0 -60 min) with the S100 extract and unlabeled CTP, ATP, UTP, and GTP. The 5Ј-end-labeled substrates were the longer (1-L) (A) and shorter (1-S) (B) versions of tRNA construct 1 and 1-S pretreated with sodium periodate (NaIO 4 ) (C).
FIG. 9. Autoradiogram of 5-end-labeled tRNAs incubated for 30 min with the S100 extract and the NTPs indicated below each lane. Control (lanes a and f) indicates that the RNA was not incubated with the S100 extract. The 5Ј-end-labeled substrates were the shorter version of tRNA construct 1 (1-S) (A) and 1-S pretreated with sodium periodate (NaIO 4 ) (B). nuclease activity. Studies of A. castellanii mitochondrial tRNA editing in vitro have similarly yielded suggestive but not definitive evidence of 5Ј nuclease activities (33). These results may simply indicate that nuclease activities are much less efficient in comparison to the nucleotide incorporation activity under the conditions of the in vitro assays. Additional studies will be required to identify and characterize nuclease activities that process tRNA 5Ј-ends in S. punctatus, as well as in A. castellanii, mitochondria.
5Ј-Nucleotide Incorporation Is Independent of 3Ј-tRNA Processing-Two G residues were incorporated at the 5Ј-end of construct 1: (i) in the presence and absence of C residues added to the 3Ј-end (Fig. 7A), (ii) in the presence of a 15-nt 3Ј extension (Fig. 6B), and (iii) when the 3Ј-terminal residue (the discriminator nucleotide, position 73) had been chemically altered by periodate oxidation (Fig. 9B). The 5Ј-nucleotide incorporation activity therefore does not appear to be affected by the nature of the tRNA 3Ј-end, and 5Ј editing is likely independent of 3Ј processing. Interestingly, the results of Laforest et al. (35) also suggest that 5Ј-tRNA editing and the addition of the CCA sequence to tRNA 3Ј-ends are independent processes in the mitochondria of other chytridiomycetes. Further studies will be required to determine how substrates are recognized by the activity, and how the number of nucleotides to be incorporated is determined.
Evolution of 5Ј-tRNA Editing-The pronounced similarity of the S. punctatus and A. castellanii 5Ј-tRNA editing activities, in terms of both tRNA sequence changes observed in vivo and characteristics of activity seen in vitro, is surprising in light of the distant evolutionary relationship of these two organisms, as shown by phylogenies based on molecular data (e.g. Ref. 48). It is possible that this activity was present in the common ancestor of S. punctatus and A. castellanii but lost in the vast majority of descendant lineages. It is also possible that its components have been horizontally transferred between distantly related organisms. Additionally, it is possible that the activity has emerged independently in several lineages including Chytridiomycota, Amoebozoa (which includes A. castellanii and Dictyostelium discoideum, a cellular slime mold that likely has mitochondrial tRNA editing, based on an abundance of acceptor stem mismatches; Ref. 49, see also Ref. 35) and Heterolobosea (extensive mismatches have been identified in the mitochondrial tRNAs of the heterolobosean Naegleria gruberi; see Ref. 35). 2 An intriguing possibility is that this activity is derived from phylogenetically widespread but poorly characterized enzymes that are responsible for the maintenance of tRNA 5Ј-ends, much as the ATP(CTP):tRNA nucleotidyltransferase adds and maintains the 3Ј -CCA OH tail across the three domains of life. An activity of this sort would remain relatively cryptic in genomic and in vitro studies, as it would act only on 5Ј-degraded tRNAs to regenerate proper acceptor stem base pairing. Such an activity would, however, assume a much more prominent role if the sequence of the first three 5Ј-tRNA nucleotides diverged and Watson-Crick base pairing potential with the 3Ј-half of the stem was lost (as is seen in S. punctatus mtDNA-encoded tRNAs). An acceptor stem repair function would thus be rendered absolutely essential for the synthesis of functional RNAs. Identification of components of 5Ј-tRNA editing activities from various organisms will help to address this possibility, while at the same time clarifying the biochemistry and evolution of 5Ј-tRNA editing and its relationship to tRNA processing.