Identification and Characterization of Mammalian Mitochondrial tRNA nucleotidyltransferases*

The CCA-adding enzyme (ATP:tRNA adenylyltrans-ferase or CTP:tRNA cytidylyltransferase (EC 2.7.7.25)) generates the conserved CCA sequence responsible for the attachment of amino acid at the 3 (cid:1) terminus of tRNA molecules. It was shown that enzymes from various organisms strictly recognize the elbow region of tRNA formed by the conserved D- and T-loops. However, most of the mammalian mitochondrial (mt) tRNAs lack consensus sequences in both D- and T-loops. To character-ize the mammalian mt CCA-adding enzymes, we have partially purified the enzyme from bovine liver mitochondria and determined cDNA sequences from human and mouse dbESTs by mass spectrometric analysis. The identified sequences contained typical amino-terminal peptides for mitochondrial protein import and had characteristics of the class II nucleotidyltransferase superfamily that includes eukaryotic and eubacterial CCA-adding enzymes. The human recombinant enzyme was overexpressed in Escherichia coli , and its CCA-add-ing activity was characterized using several m M m M 0.1 m M CTP, 0.1 m M [ (cid:1) - 32 P]ATP, 1 m M dithiothreitol, 0.05% bovine serum albumin, 0.1 A 260 unit of substrate RNA, and appropriate amounts of the enzyme fraction. We used a crude yeast tRNA mixture containing 3 (cid:1) -truncated tRNAs and 5 S rRNA obtained from overgrown cell cultures as the substrate for the purification of bovine mt CCA-adding enzyme. The enzyme activity was detected by a gel electrophore- sis assay or filter assay. The gel electrophoresis assay was used during the initial steps of enzyme preparation (Steps 2 and 3 (see below)) to distinguish the activity of the CCA-adding enzyme from that of poly(A) polymerase. After the reaction, the tRNA substrate was extracted by phenol from the reaction mixture, mixed with gel-loading solution, and then applied to a 10% polyacrylamide gel containing 7 M urea. After electrophoresis, the gel was dried and exposed to an imaging plate. The radioactivity was analyzed by an image analyzer (BAS-1000, Fuji Pho- tosystem). The filter assay was performed as follows. Aliquots (5 (cid:3) l) were spotted on a cellulose filter (Whatman No. 3MM) and acid-precip- itated with 5% trichloroacetic acid (10 min), followed by two trichloroacetic acid washes (10 min) and an ethanol wash (5 min). The level of radioactivity in the spots was visualized and quantified by the image analyzer.

The CCA-adding enzyme adds and repairs the conserved CCA sequence of the 3Ј terminus of tRNA using CTP and ATP as substrates. The CCA terminus of the tRNA molecule is the attachment site for the amino acid, and most of the aminoacyl-tRNA synthetases and elongation factor Tu require this sequence to be present to function (1)(2)(3). Furthermore, it has been shown that the CCA sequence is necessary for the exact positioning of the peptidyl-tRNA at the P site and aminoacyl-tRNA at the A site on the large ribosomal subunit to facilitate peptide bond formation (4 -6). In certain organisms such as eukaryotes, some archaea, and many eubacteria, the tRNA genes do not encode the CCA sequence; therefore its addition is an essential step for tRNA maturation (7).
CCA-adding enzymes belong to the nucleotidyltransferase superfamily, which is divided into two classes (8). Class I contains the archaeal CCA-adding enzyme and eukaryotic poly(A) polymerases, whereas class II contains eubacterial and eukaryotic CCA-adding enzymes and eubacterial poly(A) polymerases. The class II CCA-adding enzymes exhibit significant homology in the over 25-kDa region including the active site, which commonly has DXD and RRD motifs (8), whereas class I enzymes do not show a significant homology with class II enzymes around the active site.
The addition of CCA by the CCA-adding enzymes does not require any nucleic acid template unlike other DNA or RNA polymerases. Several mechanisms for CCA addition have been hypothesized from the results of biochemical experiments (9 -11). We proposed the "dead-end AMP incorporation hypothesis" based on the finding that the class II enzymes have significantly high affinity for ATP compared with CTP and that AMP incorporation clearly terminated poly(C) polymerization (12).
The CCA sequence is not encoded on organellar tRNA genes (7,13). In mammals, all 22 mitochondrial (mt) tRNA genes present in the mitochondrial genome are encoded without CCA sequences (14). After primary transcription, mt tRNAs are cleaved from the precursor RNA, and their CCA termini are synthesized by mt CCA-adding enzyme (15). It has been shown that the CCA-adding enzymes from E. coli, yeast, and Sulfolobus shibatae recognize the elbow region of tRNA formed by Dand T-loops (9,16,17). T-loop sequence is highly conserved in cytoplasmic tRNAs and is supposed to be especially important for CCA addition. In contrast to cytoplasmic tRNAs, the majority of mt tRNAs have no consensus sequences in either T-or D-loops, and some of them lack the entire loops (7,18). Therefore, we are intrigued as to how the mt CCA-adding enzyme recognizes mt tRNAs that have no conserved T-loop sequences.
Although CCA-adding activity in mitochondria has been detected in various organisms (15, 19 -22), purification of the mt CCA-adding enzymes and determination of their amino acid sequences have not yet been reported. The sole exception is the yeast mitochondrial CCA-adding enzyme, which is encoded by the same gene as the cytoplasmic enzyme, but carries an amino-terminal mitochondrial import sequence derived from an alternative start codon (20). Because yeast mt tRNAs have canonical T-loop with conserved sequence, the yeast mt CCAadding enzyme can be assumed to be similar to the bacterial enzymes in terms of tRNA recognition.
To investigate how mammalian mt CCA-adding enzymes recognize mt tRNAs, we have partially purified the enzyme from bovine liver mitochondria and determined cDNA sequences of human and mouse mitochondrial counterparts by mass spectrometric analysis coupled with dbEST search. We have expressed the human mt CCA-adding enzyme in E. coli cells and compared its activity with that of the E. coli CCAadding enzyme using mt tRNAs as the substrate. The results clearly showed that the human mt CCA-adding enzyme efficiently repaired the mt tRNAs whereas the E. coli enzyme had a much lower efficiency, although both enzymes work equally well on cytoplasmic tRNAs. These findings strongly suggest that the mt CCA-adding enzymes have co-evolved with the altered shapes of the mt tRNAs.

Assays of CCA-adding Activity for Purification of Bovine mt CCA-adding Enzyme
The assays were carried out at 37°C for 30 min in reaction mixtures (10 l) containing 50 mM Tris-HCl (pH 8.5), 10 mM MgCl 2 , 100 mM KCl, 0.1 mM CTP, 0.1 mM [␣-32 P]ATP, 1 mM dithiothreitol, 0.05% bovine serum albumin, 0.1 A 260 unit of substrate RNA, and appropriate amounts of the enzyme fraction. We used a crude yeast tRNA mixture containing 3Ј-truncated tRNAs and 5 S rRNA obtained from overgrown cell cultures as the substrate for the purification of bovine mt CCAadding enzyme. The enzyme activity was detected by a gel electrophoresis assay or filter assay. The gel electrophoresis assay was used during the initial steps of enzyme preparation (Steps 2 and 3 (see below)) to distinguish the activity of the CCA-adding enzyme from that of poly(A) polymerase. After the reaction, the tRNA substrate was extracted by phenol from the reaction mixture, mixed with gel-loading solution, and then applied to a 10% polyacrylamide gel containing 7 M urea. After electrophoresis, the gel was dried and exposed to an imaging plate. The radioactivity was analyzed by an image analyzer (BAS-1000, Fuji Photosystem). The filter assay was performed as follows. Aliquots (5 l) were spotted on a cellulose filter (Whatman No. 3MM) and acid-precipitated with 5% trichloroacetic acid (10 min), followed by two trichloroacetic acid washes (10 min) and an ethanol wash (5 min). The level of radioactivity in the spots was visualized and quantified by the image analyzer.

Purification of Bovine Liver mt CCA-adding Enzyme
All procedures were generally performed in a cold room (4°C).
Step 1-Mitochondria were obtained from a fresh bovine liver following established methods (23). Mitoplasts were prepared by digitonin treatment of the mitochondria. Mitochondrial S-100 fraction was prepared as described (23).
Step 2-Fresh S-100 fraction (3000 mg) was applied to an anion exchange column (DEAE-Sepharose fast flow or Q-Sepharose fast flow) equilibrated with Buffer TG containing 30 mM KCl (TG.03) and developed with a linear gradient from 30 to 300 mM KCl in Buffer TG at a flow rate of 1.5 ml/min. Fractions containing CCA-adding activity were pooled and dialyzed against Buffer HG.
Step 3-The pooled fractions were applied to a SP-Sepharose fast flow column (1.6 ϫ 10 cm) equilibrated with HG.02 and developed with a linear gradient (20 -300 mM KCl in Buffer HG) at a flow rate of 1.5 ml/min. Factions with activity were pooled and diluted with Buffer HG to decrease the concentration of KCl to less than 0.1 M.
Step 4 -The sample was then applied to a Hi Trap Blue column (1 ml) equilibrated with HG.1 and developed with a linear gradient from 100 to 1000 mM KCl in Buffer HG at a flow rate of 0.5 ml/min. Fractions with activity were pooled and dialyzed against Buffer PG.
Step 5-The sample was applied to a hydroxyapatite column (Bio-Scale CHT2-I (2 ml)) equilibrated with PG.1 and developed with a linear gradient from 100 to 300 mM potassium phosphate in Buffer PG at a flow rate of 0.5 ml/min. The pooled fractions with high activity were subjected to 12.5% SDS-PAGE (24) to show the protein band representing the mt CCA-adding enzyme.

In-gel Digestion of mt CCA-adding Enzyme for Peptide Mass Mapping
Following SDS-PAGE, candidate bands were excised, then soaked in a buffer containing 0.2 M NH 4 HCO 3 with 50% acetonitrile, and incubated at 30°C for 30 min to remove SDS from the gels. To ensure the complete removal of SDS, the step was repeated twice. The gel pieces were dried completely in vacuo, then rehydrated with trypsin digestion solution (0.2 M NH 4 HCO 3 , 15 ng/l L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin (Pierce)), and incubated for 12 h at 30°C. The peptides were extracted from the gel by incubation with 200 l of 0.1% trifluoroacetic acid in 60% acetonitrile (3 ϫ 20 min) with shaking. The collected fractions were combined, and the peptides were dried and dissolved in 40 l of 0.1% formic acid for peptide mass mapping by LC/MS/MS.

Mass Spectrometry and Protein Identification
A Finnigan LCQ ion trap mass spectrometer (ThermoFinnigan) equipped with an electrospray ionization source and Magic 2002 HPLC system (Michrom BioResource) was used for peptide analysis. The LC/MS/MS analysis and protein identification were performed as described (25,26). The data set of the peptide product ions generated from the triple play analysis was used to search the human and mouse EST data bases using the SEQUEST search program (27,28).

Amino-terminal Sequencing of the Bovine mt CCA-adding Enzyme
The partially purified fraction containing the CCA-adding enzyme was separated by SDS-PAGE and blotted onto the polyvinylidene difluoride membrane. The band corresponding to the enzyme, which was identified by mass spectrometric analysis, was excised and subjected to a gas-phase protein Sequencer (PPSQ-21, Shimadzu) using Edman degradation to determine the amino-terminal sequence.

cDNA Sequencing of the 5Ј Region of the Bovine mt CCA-adding Enzyme
The degenerate primers for 5Ј-RACE were designed from the human cDNA sequence of the mt CCA-adding enzyme found from the EST data base. The cDNA primer for first strand cDNA synthesis is 5Ј-TCTGT-CAGACTCTTCAGTCCTTCTG-3Ј, and the nested primer for PCR amplification of dC-tailed cDNA is 5Ј-TGAAAAGTGACTGGAATTCGG-GAGA-3Ј. 5Ј-RACE was performed according to the manufacturer's instructions (Life Technologies, Inc.). The 5Ј-RACE product was then cloned into a pCR-TOPO ® vector (Invitrogen). Plasmid DNA was isolated from positive clones, and the 5Ј region of the bovine cDNA was sequenced.

Construction of a Plasmid Expressing the Human mt CCA-adding Enzyme
The full-length coding region of the human mt CCA-adding enzyme cDNA was amplified from human total RNA using primers that were designed to bind outside the coding region; sense and antisense primers were 5Ј-GGAGTAGGCTGTGCCTTCTGAAGCAG-3Ј and 5Ј-TGCTTTT-TAGTAGCCATCAGTTTTAG-3Ј, respectively. Nested primers for the second round of PCR were upstream primer with a NdeI site (5Ј-AAAAGGGGCATATGTTCACAATGAAGTTGCAG-3Ј) and downstream primer with a XhoI site (5Ј-CCTTTCTCGAGGGTCTTCTTTATGTA-ACTC-3Ј). The first codon of the mature enzyme (CTA) in the upstream primer was changed to ATG to create an initiation codon (underlined). The PCR product was cloned into the corresponding sites of the pET-29a-c(ϩ) vector (Novagen) to obtain the expression vector, pET-CCAmt.

Expression and Purification of the Recombinant Human Enzyme
The transformant of E. coli strain BL21(DE3) carrying pET-CCAmt was cultured in LB broth (100 g/ml ampicillin) at 37°C to an A 600 value of 0.6, and expression was induced by incubation with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C. Cells harvested from 1 liter of LB broth were resuspended in 20 ml of HT buffer (50 mM Hepes-KOH (pH 7.6), 100 mM KCl, 10 mM MgCl 2 , and 7 mM ␤-mercap-toethanol) containing 0.2 mM PMSF, 0.03% (w/v) egg white lysozyme, and 0.1% Triton X-100 and disrupted by an 8-min sonication (repeated 1-s bursts following 4-s cooling periods) at 100 watts at 0°C. The homogenate was cleared by centrifugation at 100,000 ϫ g for 60 min. The supernatant fraction (S100) was loaded onto a nickel-charged Hi-Trap chelating column (5 ml) (Amersham Pharmacia Biotech). After the non-bound proteins were washed off, the recombinant protein was eluted with a 60-ml linear gradient from 0 to 350 mM imidazole in HT buffer. The recombinant mt CCA enzyme was eluted in a fraction containing about 200 mM imidazole and then dialyzed against HT buffer. Glycerol was added to the enzyme to a final concentration of 30%, frozen quickly with liquid nitrogen, and stored at Ϫ70°C.

Preparation of the Recombinant E. coli CCA-adding Enzyme
E. coli CCA-adding enzyme was overexpressed and purified as described by Tomari et al. (12).

Characterization of Recombinant Human mt CCA-adding Enzyme
The assays were performed under the same conditions described under "Assays of CCA-adding Activity for Purification of Bovine mt CCA-adding Enzyme," and the activity was detected by gel electrophoresis. Substrate tRNAs were prepared as follows. Yeast tRNA Phe -DC (3Ј-CA-truncated) was synthesized by the run-off transcription using T7 RNA polymerase as described by Milligan and Uhlenbeck (29). Native mitochondrial tRNAs that had been confirmed to contain 3Ј-truncated forms were kindly provided by Takeo Suzuki (University of Tokyo). In vitro transcribed Ascaris suum mitochondrial tRNA Met -DC was from Dr. Takashi Ohtsuki and Sarin Chimnaronk (University of Tokyo).

Determination of K m and k cat Values
The initial rates of the CCA-adding activity were determined using different concentrations of 3Ј-truncated tRNAs ranging from 1 to 20 M. Results were used in Lineweaver-Burk plots to determine the apparent K m and k cat values. Yeast tRNA Phe -DC was prepared as described (29). To uniformly truncate the native bovine mitochondrial tRNA Ala , the CCA end was first completed with the human mt CCA-adding enzyme, and then 3Ј-terminal A was removed by periodate oxidation and dephosphorylation (30).

Purification of the Bovine mt CCA-adding Enzyme-The
CCA-adding enzyme from bovine liver mitochondria was purified as described under "Experimental Procedures." The activity of the CCA-adding enzyme in the protein fractions was easily detected by [␣-32 P]AMP incorporation into the 3Ј-truncated tRNA mixture that acted as the substrate. However, in the presence of other nucleotidyltransferase enzymes such as poly(A) polymerase, it was difficult to differentiate between the activities of the two enzymes. To ensure specificity of the assay system, we performed a control experiment using denaturing gel electrophoresis of the tRNA mixture after the reaction to enable discrimination of the two activities; only the tRNA fraction was labeled with [␣-32 P]AMP by the E. coli CCA-adding enzyme, whereas 5 S rRNA present in the mixture was labeled by E. coli poly(A) polymerase (Fig. 1B, lanes 2 and 3). As shown in Fig. 1, A and B, using the gel electrophoresis assay we were According to the mass spectrometric analysis, MC1 and MC2 were related to human acyl-CoA dehydrogenase (short chain-specific precursor) and human aminomethyltransferase, respectively. MC3 was identified as the human mt CCA-adding enzyme (see Fig. 3). The elution profile of other bands did not correspond to CCA-adding activity. Incidentally, the most intense band with Ͼ100 kDa was related to rat carbamoyl-phosphate synthase.
able to distinguish the two enzyme activities present in the protein fractions separated by anion exchange chromatography in Step 2 (see "Experimental Procedures"). CCA-adding activity was isolated in different fractions (Fractions 29 -35) to those with poly(A) polymerizing activity (Fractions 25-31), indicating that two enzyme activities in bovine mitochondria could be distinctively measured by this assay. For further purification, we adopted a simplified method using cellulose filters (filter assay) to detect activity instead of gel electrophoresis because most of the poly(A) polymerase activity was eliminated during the first three steps. We concentrated the fraction containing the CCA-adding activity with three more steps of column chro- The numbers indicate the identified peptides in the human mt CCA-adding enzyme, the location and sequences of which are shown in C. B, the CID spectrum for peptide number 12 (NLGLFIVK) is shown as an example of peptide sequence confirmation. The parent ion is the single-charged ion of the peptide indicated by an arrow. Fragment ions derived from the parent ion were identified as a, b, or y type ions and deaminated ions of b (indicated by b*) according to nomenclature found in the literature (44). C, protein sequence of the human mt CCA-adding enzyme. The boxed region indicates the corresponding EST sequence of AW582962, which was initially retrieved by a SEQUEST search. The conserved DXD and RRD motifs are indicated in the black background. The sequences in red are identified peptides numbered with arrows. The amino-terminal sequence of the bovine counterpart is shown in green letters. The amino-terminal tryptic peptide (MFTMK) derived from the bovine sequence was able to be detected as peptide number 1. Signal sequence for mitochondrial protein import is shown in magenta.

matography. Hi Trap Blue in
Step 4 was particularly effective in the purification of the enzyme (data not shown).
The active fractions after Step 5 were separated by SDS-PAGE, and about 10 bands were observed following Coomassie Brilliant Blue staining (Fig. 1C). We selected three bands (MC1, MC2, and MC3) as candidates for the mitochondrial CCA-adding enzyme, considering that the quantities of these bands in the fractions correlated well with the activity profile.
Peptide Analysis and cDNA Cloning-The three protein bands were excised from the gel, digested by trypsin, and subjected to mass spectrometric analysis (LC/MS/MS), and the results were searched against human and mouse data bases. Protein bands MC1 and MC2 were identified as known mt metabolic proteins (shown in the legend to Fig. 1). The analysis of protein MC3 with 47 kDa clearly hit to the human EST sequence AW582962. This EST has high sequence homology to enzymes belonging to class II of the nucleotidyltransferase superfamily that includes eukaryotic and eubacterial CCAadding enzymes. The EST also possesses conserved DXD and RRD motifs within the amino-terminal domain, responsible for nucleotide incorporation activity (Fig. 2C). The complete cDNA sequences encoding the proteins in human and mouse were obtained by assembling many related EST sequences retrieved by a BLAST search, and gaps and sequencing errors were corrected (Fig. 2C). Many peptide ions in the tryptic digest of MC3 were identified in the complete cDNA sequence (Fig. 2, A  and C). Each peptide sequence was confirmed by assignment of a CID spectrum (Fig. 2B).
The cDNA sequence encoding the amino-terminal region of the bovine enzyme was determined using 5Ј-RACE. Because bovine and human mt CCA-adding enzymes have a TGA codon 15 nucleotides upstream from the start of the predicted ORF, we could determine the initial ATG codon of the enzyme with confidence (Figs. 2C and 3). The sequence of the amino terminus of the mature enzyme from bovine mitochondria was determined to be NH 2 -MFTMXLQXPEFQSLF by peptide sequencing (Figs. 2C and 3). This also revealed that the aminoterminal 26-amino acid peptide of the precursor protein was removed after importation into the mitochondria (Figs. 2C and  3). The calculated molecular mass of the putative human ma- ture protein is 47 kDa, which is the same size as the purified bovine enzyme as determined by SDS-PAGE (Fig. 1C). We noted in the data base the presence of the protein sequence, CGI-47, which covers a large region of the human CCA-adding enzyme, compiled by Lai et al. (31). This sequence was derived from comparative gene identification in silico using Caenorhabditis elegans proteins as queries. However, this sequence lacks the amino-terminal region of the human CCA-adding enzyme that we have shown to be present. Also, in the CGI-47 sequence, an internal methionine at position 30 has been incorrectly ascribed as the putative initiation methionine (Figs.  2C and 3).
Gene Organization and Sequence Alignment-During the assembly of the ESTs, we found a splice variant of the human cDNA that shared the amino-terminal region. This prompted us to search the human genome draft sequence (32,33) using the cDNA sequences as queries. We found that the gene encoding the human mt CCA-adding enzyme was located at chromosome 3p25.1 and was composed of seven exons (Fig. 4). In addition, two pseudogenes were identified on chromosomes 1 and 22, respectively. It is conceivable that the observed splice variant was produced by an incorrect splicing event between exons 1 and 2 as shown in Fig. 4. A small protein containing exon 1 of the gene encoding the CCA-adding enzyme could be translated from the splice variant. However, we could not detect a similar splice variant in the mouse, suggesting that this variation in the splicing process is a human-specific event.
The sequences of mammalian mt CCA-adding enzymes were aligned with other paralogues from bacteria and Drosophila, which were retrieved by a BLAST search using the human counterpart as a query (Fig. 3). Analysis showed that the mammalian enzymes had higher homology to the Drosophila enzyme than to the bacterial counterparts. These enzymes commonly have conserved DXD and RRD motifs in their amino-terminal regions, which are characteristic for members of the class II nucleotidyltransferase superfamily. High sequence homology was observed among the aminoterminal regions, whereas there was low homology in the carboxyl-terminal regions.
Overexpression of the Recombinant Human mt CCA-adding Enzyme in E. coli-The human mt CCA-adding enzyme was amplified by reverse transcription-PCR using primers designed from the mature sequence and cloned into the pET29a vector. According to the amino-terminal position of the matured bovine enzyme, leucine at position 27 should be the corresponding amino terminus of the human enzyme. We therefore replaced the Leu at position 27 with Met to create an initiation codon to express the matured human mt enzyme. The human enzyme, with a carboxyl-terminal His-tag, was expressed in E. coli and isolated with high purity (Fig. 5A). About 20 mg of the recombinant protein was purified from 1 liter of culture. The activity of the expressed enzyme was confirmed by incorporation of both ATP and CTP into the 3Ј-CA-truncated tRNA (yeast tRNA Phe -DC) under the standard assay conditions (Fig. 5B).
Characterization of Human mt CCA-adding Enzyme-It has been reported that the CCA-adding enzyme strictly recognizes the elbow region of tRNA formed by D-and T-loops (9,16,17), and the conserved sequence in the T-loop is supposed to be of particular importance for recognition (Fig. 6C, yeast tRNA Phe ). Because most of the mammalian mt tRNAs have no consensus sequence in the T-loop, it was of great interest to examine the substrate specificity of the mammalian mt CCA-adding enzyme. To compare the activities of the human mt and E. coli CCA-adding enzymes, we chose four species of bovine mitochondrial tRNAs specific to Ser (GCU), Ala, Pro, and Thr, which were purified from bovine liver, as substrates (Fig. 6C). Each of these tRNAs has non-conserved T-loops with different lengths composed of eight, four, seven, and eleven nucleotides, respectively. In addition, mt tRNA Ser (GCU) has an unusual secondary structure in that it lacks the entire D-arm. Yeast tRNA Phe was used on behalf of cytoplasmic tRNA as a positive control. Results showed that the mt CCA-adding enzyme is able to repair both mt and yeast tRNAs with high efficiency, whereas the E. coli enzyme could repair only yeast tRNA (Fig.  6A). This finding indicates that E. coli enzyme requires the presence of the conserved T-loop sequence in the tRNA for recognition.
The kinetic parameters of CCA repair by the mt and E. coli enzymes were measured using mt tRNA Ala and yeast tRNA Phe as substrates (Table I). In the case of the human mt enzyme, the K m value with mt tRNA Ala as the substrate was close to that obtained using yeast tRNA Phe , whereas the K m value for the E. coli enzyme with mt tRNA Ala as the substrate was 19 times higher than that for yeast tRNA Phe . This result suggests that the E. coli enzyme cannot efficiently recognize mt tRNA Ala without a conserved T-loop sequence because of weak interaction with the substrate, whereas the human mt enzyme can recognize a wide variety of tRNAs independent of their T-loop sequences.
We also investigated whether there was any recognition of non-conserved T-loops by the human mt CCA-adding enzyme. Because nematode mitochondrial tRNAs lack the T-arm, we carried out assays of CCA addition to A. suum mt tRNA Met as substrate using both human mt and E. coli enzymes. As shown in Fig. 6B, both enzymes had low CCA addition activity on the nematode tRNA lacking the T-arm (upper panel) compared with that to yeast tRNA Phe (bottom panel), indicating that the T-arm is required for efficient addition of CCA by the human mt enzyme. However, the nematode tRNA was repaired by increased amounts of the human mt CCA-adding enzyme but not with increased amounts of the E. coli enzyme (Fig. 6B,  middle panel). This suggests that the human mt CCA-adding enzyme loosely recognizes the structure of the non-conserved T-arm but not the specific sequence, which is different from the E. coli enzyme.

DISCUSSION
We could identify two distinct nucleotidyltransferase activities in bovine mitochondrial extract using a yeast tRNA mixture as the substrate. The activity of the putative poly(A) polymerase ranged over the fractions and was weaker than the activity of the CCA-adding enzyme (Fig. 1A). Although we tried to concentrate the poly(A) polymerizing activity by three additional column chromatography steps, we lost the activity on the way, indicating that mt poly(A) polymerase was unstable. On the other hand, as the CCA-adding activity was stable and easy to concentrate, we successfully purified the corresponding pro-tein and identified its gene from an EST data base using mass spectrometric analysis. The analysis of the amino-terminal region of human mt CCA-adding enzyme by PSORT (34) showed that the probability score of importation into the mitochondrial matrix was relatively high (0.719), indicating that the gene product is actually localized in mitochondria and that the amino-terminal peptide functions as a mitochondrial protein import signal.
Evidence for a mammalian cytoplasmic form of the CCAadding enzyme has not been found so far, although Deutscher's group partially purified a cytoplasmic CCA-adding activity FIG. 6. Characterization of the mammalian mt CCA-adding enzyme. A, the substrate specificity of human mt and E. coli CCA-adding enzymes. The CCA-repairing efficiency of these two enzymes was compared using 3Ј-CA-truncated yeast tRNA Phe -DC and four bovine native mt tRNAs (tRNA Ser GCU, mt tRNA Pro , mt tRNA Ala , and mt tRNA Thr ), which partially contain 3Ј-truncated tRNAs as substrates. B, addition of CCA to A. suum tRNA Met -DC (5 g) by human mt and E. coli CCA-adding enzymes. Top panel shows CCA addition to A. suum tRNA Met -DC by using 0.2 pmol (1ϫ) of human mt (right) and E. coli (left) enzymes. Middle panel shows results using 2 pmol (10ϫ) of each enzyme. Bottom panel shows CCA addition to yeast tRNA Phe -DC as a control. C, primary sequences of tRNA substrates used in this study and comparison of their T-loops. All tRNA sequences are shown without modified nucleosides. from rabbit and rat livers (22,35). In this study, we failed to find any homologous sequence of nucleotidyltransferases other than the human mt CCA-adding enzyme in human EST and genomic data base, despite the recent rapid increase in the amount of genome information available. Furthermore, we observed that the human mt enzyme catalyzes the addition of CCA to both cytoplasmic and mt tRNAs with lower specificity. It is possible that one gene may encode both cytoplasmic and mitochondrial enzymes as reported in yeast (20, 36 -38). In the yeast S. cerevisiae, the CCA1 gene has three ATG codons at the 5Јend of the open reading frame. The first ATG is an initiation codon for the mitochondrial enzyme, whereas the second and third ATGs direct production of cytoplasmic counterparts. Alternative use of multiple ATG codons may lead to the localization of the mammalian CCA-adding enzymes to different parts of the cell. Another possibility is that alternative splicing creates cytoplasmic and mitochondrial forms. It has been reported that human cytoplasmic and mitochondrial lysyl-tRNA synthetases are produced by alternative splicing in the aminoterminal region of the corresponding gene (39). However, in human and mouse mt CCA-adding enzymes, only one in-frame ATG is likely to be used for translation (Fig. 3), and no productive splicing variants can be detected. Thus, if one gene codes for both mitochondrial and cytoplasmic proteins, the distribution mechanisms may be different from those for yeast CCA1 and human lysyl-tRNA synthetase.
As observed in this study, the human mt CCA-adding enzyme can efficiently repair mt tRNAs with non-conserved Tloop and unusual structures (Fig. 5B). In addition, the human mt enzyme could potentially recognize nematode mt tRNA without the T-arm. This suggests that the human mt enzyme does not strictly recognize the conserved T-loop sequence that is supposed to be one of the major recognition elements for most of the CCA-adding enzymes from other organisms. This result implies that the mammalian mt enzymes have evolved to recognize mt tRNAs with a wide variety of T-loops and secondary structures. This low substrate specificity toward tRNA is one of the characteristic features of other mitochondrial enzymes, such as mt aminoacyl-tRNA synthetases (40) or elongation factors (41,42).
In the future, we will address the molecular mechanism explaining the loose recognition of tRNA by the human CCAadding enzyme. According to the study of E. coli poly(A) polymerase, which is a member of the class II nucleotidyltransferase superfamily, the C terminus of the enzyme is the substrate RNA binding domain (43). This would also be the case with the CCA-adding enzymes. The carboxyl-terminal regions of the mammalian mt CCA-adding enzymes have low homology with those of their bacterial counterparts (Fig. 3). This observation suggests that the low substrate specificity of human mt enzymes may have derived from the sequence diversity of the carboxyl-terminal region responsible for tRNA recognition.
Further study is required to clarify the molecular mechanism for tRNA recognition by the CCA-adding enzyme.