Dual Mode Recognition of Two Isoacceptor tRNAs by Mammalian Mitochondrial Seryl-tRNA Synthetase*

Animal mitochondrial translation systems contain two serine tRNAs, corresponding to the codons AGY (Y (cid:1) U and C) and UCN (N (cid:1) U, C, A, and G), each possessing an unusual secondary structure; tRNA GCUSer (for AGY) lacks the entire D arm, whereas tRNA UGASer (for UCN) has an unusual cloverleaf configuration. We previously demonstrated that a single bovine mitochondrial seryl-tRNA synthetase (mt SerRS) recognizes these topologi-cally distinct isoacceptors having no common sequence or structure. Recombinant mt SerRS clearly footprinted at the T (cid:2) C loop of each isoacceptor, and kinetic studies revealed that mt SerRS specifically recognized the T (cid:2) C loop sequence in each isoacceptor. However, in the case of tRNA UGASer , T (cid:2) C loop-D loop interaction was further required for recognition, suggesting that mt SerRS recognizes the two substrates by distinct mechanisms. mt SerRS could slightly but significantly misacylate mitochondrial tRNA Gln , which has the same T (cid:2) C loop sequence as tRNA UGASer , implying that the fidelity of mito- chondrial translation is maintained by kinetic discrimination of tRNAs in the network of aminoacyl-tRNA HT m M Hepes-KOH (pH 7.6), m M KCl, 10 m M MgCl , and 7 m M (cid:2) -mercaptoethanol) containing 0.2 m M phenylmeth- anesulfonyl fluoride, 0.03% (w/v) egg white lysozyme, and Triton X-100 and disrupted by sonication (repeated and was by centrifugation (cid:2) (S100) was onto a nickel-charged chelating (5 ml). After washing out nonbound proteins, the recombinant protein was eluted with a linear gradient from to 350 m M imidazole in HT buffer. mt was eluted in a fraction containing 200 m M imidaz- ole. Protein concentrations were determined a Bio-Rad protein assay kit pooled Ser -mt as performed as by the gel was stained with Coomassie Brilliant Blue and toluidine blue to analyze the components of the mt Ser complex. To determine the K d value by gel retardation assay, each mt tRNA Ser was labeled 32 P at the end. mt tRNA Ser was added to each labeled tRNA as a carrier tRNA. Assays were performed by using five different con- centrations of tRNAs Ser ranging from 0.25 to 5 (cid:1) M (0.25, 1.0, 3, 4, and The aminoacylation reaction was performed as described by Yokogawa et al. (18). The initial rates of aminoacylation were ascertained by using six different tRNA concentrations between 0.1 and 2 (cid:1) M , determined, respectively, according to the amount of each tRNA recovered. In each case, mt SerRS was used at a fixed concentra- tion optimized according to the activity of each tRNA and ranging from 7.15 n M to 14.3 (cid:1) M . Kinetic experiments gave reasonable Hanes-Woolf plots for determining the kinetic parameters of K m and k cat . All values are the averages of three independent determinations, which varied less than 15%. The aminoacylation reaction for the minihelix tRNAs Ser and non-cognate mt for Gln, Glu, and Tyr was carried out in 60 (cid:1) l of a reaction mixture containing 18 of and 9.5 (cid:1) g of mt SerRS. At each point, a 10- (cid:1) l aliquot was withdrawn. We calculated the quantity of each tRNA per A 260 unit according to its length, namely 1800 pmol for tRNA UGASer and its variants, 2000 pmol for tRNA GCUSer and 3600 pmol for the two minihelix tRNAs. Other conditions were the same as those used to determine the kinetic parameters.

The fidelity of protein synthesis relies on the specific attachment of amino acids to their cognate tRNA species. This process is catalyzed by aminoacyl-tRNA synthetase (ARS), 1 each species of which discriminates with high selectivity among the many structurally similar tRNAs and amino acids (1,2). To avoid misacylation of tRNAs from any of the 19 non-cognate groups, tRNAs possess identity elements within their sequence or tertiary structure that are strictly recognized only by the cognate synthetase. These identity elements are most commonly located in the anticodon and in the acceptor stem, particularly the discriminator base at position 73 (2,3). However, in the case of the serine tRNA of Escherichia coli, several biochemical experiments have revealed that neither the anticodon stem/loop nor the discriminator base is involved in recognition (4,5); instead, the E. coli tRNA Ser identity elements are located in the characteristic long extra arm (4 -7). These findings conform well with analyses of the crystallographic structures of seryl-tRNA synthetase (SerRS)-tRNA Ser complexes from E. coli and Thermus thermophilus (8 -10), which indicate that the N-terminal long helical domain of SerRS plays an important role in recognizing the long extra arm and the T⌿C loop of tRNA Ser . In eukaryotic systems, cytoplasmic tRNA Ser also has a long variable arm, and biochemical studies of Saccharomyces cerevisiae and human tRNAs Ser have revealed that it contains the major identity element of tRNA Ser (11)(12)(13)(14). The recognition mechanism of SerRS thus appears to be evolutionarily conserved in both prokaryote and eukaryotic cytoplasm.
The mammalian mitochondrial (mt) translation system utilizes two tRNA Ser species, one specific for codon AGY and the other for UCN. Neither of these tRNAs has a long extra arm as a recognition site for cytoplasmic SerRS (15). In addition, each possesses an unusual secondary structure; tRNA GCU Ser (for codon AGY) lacks the entire D arm (16), whereas tRNA UGA Ser (for codon UCN) has an unusual cloverleaf configuration with an extended anticodon stem (17). We previously demonstrated that the single mt SerRS recognizes these distinct isoacceptors with almost the same activity (18). Additionally, inspection of the primary sequences of several mt SerRSs revealed differences between mammalian mt SerRS and its prokaryotic counterpart in the N-terminal domain responsible for tRNA recognition, which are in line with structural and recognition differences between the extra arms of mammalian mt and prokaryotic tRNAs Ser . Because no other tRNA investigated to date recognizes structurally different tRNA isoacceptors, it is supposed that the recognition mechanism of mammalian mt SerRS differs considerably from that of any other ARS.
Elucidating the mystery of how mammalian mt SerRS recognizes and discriminates two isoacceptors with no common structure and sequence from non-cognate tRNAs will not only extend our knowledge of the recognition mechanism of ARS but also shed light on hidden aspects of the mammalian mt translation system.
To investigate the recognition mechanism of mammalian mt SerRS, we recombinantly expressed bovine mt SerRS in E. coli and performed a series of biochemical experiments using this enzyme and several tRNA variants. On the basis of the results, we report here the unique recognition mechanism of mammalian mt SerRS.

EXPERIMENTAL PROCEDURES
Materials-Phenylmethanesulfonyl fluoride was purchased from Sigma; [ 14 C]L-serine (5.59 GBq/mmol), [ 32 P]pCp, and a HiTrap chelating column were from Amersham Pharmacia Biotech; the vector pET-19b was from Novagen; nucleotide-specific RNases T 1 and U 2 were from Amersham Pharmacia Biotech and Seikagaku Kogyo (Tokyo), respectively; vectors pUC18 and pUC19 were from Takara; an anion-exchange tip was from Qiagen; and a QuikChange site-directed mutagenesis kit was from Stratagene. Recombinant mt LeuRS was overproduced from an expression vector kindly provided by Dr. L. L. Spremulli (University of North Carolina, Chapel Hill, NC). Native bovine mt tRNAs were purified from bovine mitochondria by selective hybridization using a solid phase DNA probe as described by Wakita et al. (19).
Construction of Expression Plasmid-cDNA for bovine mt SerRS without an N-terminal peptide for mitochondrial importation was amplified by PCR using synthetic primers; a forward primer corresponding to the N-terminal end of mature mt SerRS (atataccatgggccatcatcatcatcatcatcatggcagcgacgacgacgacaaggcaacggagaggcaggatcg) possessing an NcoI site, and a reverse primer for the C-terminal end (cagccggatcctcagctcgaggcaggctgg) carrying a BamHI site. The PCR product was cloned into pET-19b to construct an expression vector for mature bovine mt SerRS with a hexahistidine tag in the N-terminal region.
Expression and Purification of mt SerRS-E. coli BL21 (DE3) was used as a host for expression of the recombinant mt SerRS. The culture conditions for overproducing cells were optimized to maximum the expression of soluble enzyme. The transformant was cultured in LB broth (100 g/ml ampicillin) at 37°C to an A 600 value of 0.6, and then induced by 10 M isopropyl-1-thio-␤-D-galactopyranoside for 20 h at 28°C. Cells harvested from 3 liters of LB broth were resuspended in 40 ml of HT buffer (50 mM Hepes-KOH (pH 7.6), 100 mM KCl, 10 mM MgCl 2 , and 7 mM ␤-mercaptoethanol) containing 0.2 mM phenylmethanesulfonyl fluoride, 0.03% (w/v) egg white lysozyme, and 0.1% Triton X-100 and disrupted by 12-min sonication (repeated 1-s bursts after 4-s cooling periods) at 100 watts and 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 HiTrap chelating column (5 ml). After washing out nonbound proteins, the recombinant protein was eluted with a 60-ml linear gradient from 0 to 350 mM imidazole in HT buffer. mt SerRS was eluted in a fraction containing ϳ200 mM imidazole. Protein concentrations were determined with a Bio-Rad protein assay kit using bovine serum albumin as a standard. Glycerol was added to pooled mt SerRS fractions at a final concentration of 30%, frozen quickly with liquid nitrogen, and stored at Ϫ70°C.
Native PAGE and Gel Retardation Assay-The mt tRNA Ser -mt SerRS complex was formed as described by Yokogawa et al. (18). Native PAGE was performed as described by Hornung et al. (20); the gel was stained with Coomassie Brilliant Blue and toluidine blue to analyze the components of the mt tRNA Ser -mt SerRS complex. To determine the K d value by gel retardation assay, each mt tRNA Ser was labeled with 32 P at the 5Ј end. Nonradioactive mt tRNA Ser was added to each labeled tRNA as a carrier tRNA. Assays were performed by using five different concentrations of tRNAs Ser ranging from 0.25 to 5 M (0.25, 1.0, 3, 4, and 5 M) for tRNA GCU Ser or from 0.25 to 3 M (0.25, 0.5, 1.0, 2.0, and 3.0 M) for tRNA UGA Ser with a fixed concentration (0.44 M) of mt SerRS. The relative radioactivity of the RNA band was quantified by a BAS-1000 imaging system (Fuji Photo Film). The amounts of tRNA Ser -mt SerRS complexes were calculated by subtracting the free tRNA counts from that of the whole count. The K d value for each native tRNA Ser was obtained by Scatchard plots.
tRNA Footprinting with Ethylnitrosourea-tRNA footprinting using ethylnitrosourea was performed as described by Vlassov et al. (21) with slight modification. mt tRNA Ser was alkylated with or without mt SerRS at room temperature for 4 h in a reaction mixture containing 5Ј end-labeled mt tRNA Ser (ϳ50,000 cpm/tube) with 1.5 M cold carrier tRNA Ser , 50 mM sodium cacodylate (pH 8.0), 10 mM MgCl 2 , and 0.11 volume of saturated ethylnitrosourea (ethanol solution). The recombinant mt SerRS and non-cognate mt LeuRS were added to the reaction mixture at final concentrations of 1-5 and 5 M, respectively. The following experiment conducted after the alkylation was described previously (21). Electrophoresis using 12% acrylamide, 7 M urea gel was performed to analyze the footprint site of tRNA Ser in the presence of mt SerRS. The band position was assigned by comparison with partial RNase T 1 and/or U 2 digests. The gel was exposed to an imaging plate, followed by analysis using a Fuji BAS-1000 imaging analyzer.
Preparation of in Vitro Transcribed tRNAs-Mitochondrial tRNA variant genes were constructed on pUC18 under a T7 class III promoter as described previously (22). The transcriptional template harboring the T7 promoter and tRNA gene terminating at its discriminator base (position 73) was amplified from the constructed plasmid by PCR and transcribed to in vitro according to the literature (23), except that the reaction mixture contained E. coli CCA, adding at 8 g/1 ml. After the reaction, a small part of the mixture was electrophoresed on 12% long denaturing polyacrylamide gel (40 cm ϫ 20 cm ϫ 0.35 mm) to check the purity of the synthesized tRNA. All tRNA variants with the complete CCA sequence at the 3Ј end were synthesized with more than 70% purity. The transcribed tRNA was extracted by phenol/chloroform treatment, applied onto an anion-exchange tip (Qiagen), and then the complete tRNA band was cut out from the 12% denaturing gel to separate it from the adjacent minor bands. In this way, highly purified tRNA was obtained. The 3Ј end of each tRNA was confirmed by thin layer chromatography analysis of the 3Ј-terminal adenosine, which was labeled with [ 32 P]pCp (data not shown). In addition, the RNA sequences of some variants were verified by Donis-Keller's enzymatic digestion method (data not shown) (24). We finally succeeded in recovering about 0.2 A 260 unit of pure tRNA/1-ml reaction mixture.
Preparation of tRNA Variants by in Vivo Expression-To prepare variants of mt tRNA GCU Ser in vivo, the plasmid vector pUC19 carrying the tRNA gene with the T7 promoter and terminator was constructed as reported previously (25). The plasmids for the other tRNA GCU Ser derivatives used were prepared with a QuikChange site-directed mutagenesis kit (Stratagene). tRNA variants were expressed in E. coli strain BL21 (DE3) according to Hayashi et al. (25). Expressed tRNA was extracted from the cells as described (26) and purified by electrophoresis on 8% polyacrylamide gel containing 7 M urea and 20% formamide.
Aminoacylation Assay-The aminoacylation reaction was performed as described by Yokogawa et al. (18). The initial rates of aminoacylation were ascertained by using six different tRNA concentrations between 0.1 and 2 M, determined, respectively, according to the amount of each tRNA recovered. In each case, mt SerRS was used at a fixed concentration optimized according to the activity of each tRNA and ranging from 7.15 nM to 14.3 M. Kinetic experiments gave reasonable Hanes-Woolf plots for determining the kinetic parameters of K m and k cat . All values are the averages of three independent determinations, which varied less than 15%. The aminoacylation reaction for the minihelix tRNAs Ser and non-cognate mt tRNAs for Gln, Glu, and Tyr was carried out in 60 l of a reaction mixture containing 18 pmol of tRNA substrate and 9.5 g of mt SerRS. At each point, a 10-l aliquot was withdrawn. We calculated the quantity of each tRNA per A 260 unit according to its length, namely 1800 pmol for tRNA UGA Ser and its variants, 2000 pmol for tRNA GCU Ser and 3600 pmol for the two minihelix tRNAs. Other conditions were the same as those used to determine the kinetic parameters.

Characterization of Recombinant mt SerRS-Because it is
almost impossible to obtain sufficient native mt SerRS from bovine liver to carry out a detailed investigation of the recognition mechanism, recombinant mt SerRS with an N-terminal histidine tag was overexpressed in E. coli cells. The optimized culture conditions for maximum expression of the soluble protein are described under "Experimental Procedures." As shown in Fig. 1, mt SerRS was purified almost homogeneously by nickel-chelating chromatography. The yield was 8.4 mg from 1 liter of the cell culture.
The overall structure of the recombinant mt SerRS was examined by measuring the circular dichroism (CD) in the far-UV spectrum, which exhibited CD troughs at 209 and 222 nm, typically indicative of a high ␣-helical content (data not shown). E. coli SerRS has a similar CD spectrum. According to the online COILS algorithm for the prediction of coiled coils from amino acid sequences (dot.imgen.bcm.tmc.edu:9331/seq-search/struc-predict.html) (29), mt SerRS has a single long ␣-helix in the Nterminal region (residues 57-87), which could explain the CD spectral observation that mt SerRS has a high helical content similar to that of E. coli SerRS. In the crystal structure, E. coli SerRS is known to have two long anti-parallel ␣-helices in the N-terminal domain (residues 28 -53 and 71-103) (27,28), which were clearly predicted by the COILS algorithm.
In addition, the apparent molecular mass of mt SerRS in solution was estimated by gel filtration column chromatography. The enzyme was eluted mainly at 33 min, corresponding to a molecular mass of 109 kDa, with minor peak at 37 min corresponding to 57 kDa (data not shown). As the molecular mass of the enzyme has been calculated from its gene sequence to be 56.3 kDa (18), mammalian mt SerRS seems to take a homodimeric form, as is the case with prokaryotic SerRS (9,27,28,30).
To establish whether the recombinant mt SerRS could specifically recognize the two mt tRNAs Ser , the serylation activity kinetic parameters were determined using the native mt tRNA GCU Ser and tRNA UGA Ser as substrates. As shown in Table I, the parameters of the recombinant mt SerRS were almost identical to those of the native mt SerRS. A gel retardation assay was performed to ascertain the complex formation of mt SerRS with each isoacceptor. The assay showed that mt SerRS specifically formed binary complexes with the two serine isoacceptors, whereas no complex was formed with non-cognate mt tRNA UAA Leu ( Fig. 2A). The dissociation constant (K d ) for each tRNA substrate was determined by means of Scatchard plots (Fig. 2B). The larger K d value of tRNA GCU Ser (1 M) is indicative of a lower affinity toward mt SerRS. The difference in the dissociation constants seems to reflect the different K m values of the two substrates (Table I). In addition, it appears that the x-intercept of each Scatchard plot approaches 1.8, suggesting that the two tRNAs Ser bind into one dimeric form of mt SerRS, although further structural analysis is necessary to clarify the stoichiometry of the mt SerRS-tRNA Ser complex.
Because the foregoing observations strongly suggested that the recombinant mt SerRS retained the original characteristics of the native enzyme, the subsequent experiments were performed with this recombinant enzyme.
mt SerRS Contact Sites on the Two Serine tRNAs-tRNA footprinting was carried out with ethylnitrosourea for both mt tRNAs Ser in the presence of mt SerRS, which strongly protected mt tRNA GCU Ser against alkylation by ethylnitrosourea at two specific regions, i.e. phosphate positions 57-58 and 64 -67 (Fig.  3A). mt tRNA UGA Ser was also protected at similar phosphate positions (55-59 and 65-67). No protection was observed in the presence of human mt leucyl-tRNA synthetase (mt LeuRS), which was used as a negative control for each mt tRNA (lanes 5 and 6 in Fig. 3, A and B, respectively). Moreover, no strong protection was found in the 5Ј region of mt tRNAs Ser (data not shown). These results indicate that the probable contact sites on both tRNAs are on the T⌿C loop and at the bottom of the acceptor stem. As shown in Fig. 3C, mammalian mt SerRS contacts both tRNAs at similar positions though they have different topologies.
The crystal structure of T. thermophilus SerRS reveals that Arg 195 contacts positions 66 -67 in the acceptor helix of tRNA Ser , which is the same location as one of the mt SerRS contact sites. Because Arg 195 is conserved in mammalian mt SerRS (18), it can be speculated that the contact site at phosphate positions 64 -67 may not be required for specific recognition but would be involved in the essential interaction needed to arrange the CCA terminus at the catalytic center of the enzyme.
tRNA UGA Ser Identity Elements for mt SerRS-Although tRNA footprinting clearly demonstrated that mammalian mt SerRS contacts both isoacceptors similarly on the T⌿C loop and at the junction between the T⌿C and acceptor stems, it was still unclear how mt SerRS specifically recognizes two tRNAs having no common structure or sequence and identifies them among the 22 mt tRNAs. To clarify the unique recognition mechanism of mt SerRS, the identity elements of the respective isoacceptors needed to be determined.
To examine the recognition elements of mt tRNA UGA Ser , we constructed several variants carrying a mutation(s), mainly at bases in the T⌿C loop, which is one of the mt SerRS contact sites and is predicted to be involved in tertiary interactions according to a previously proposed model (31). The mutational variations are shown in Fig. 4A, and the kinetic parameters are summarized in Table II  with the canonical sequence was shown to have a K m value 4 times higher than that of the native tRNA from bovine liver. According to Hayashi et al. (32), replacement of the two terminal base pairs should have no effect on serylation activity, in which case the relatively high K m value is apparently attributable to the lack of modified bases. We first examined the variants with mutations in the anticodon at positions 34 -36 and at the discriminator base (position 73), which are recognition elements in most tRNAs. Replacement of the anticodon sequence or the discriminator base caused no significant reduction in serylation activity. As it was revealed that the A73G mutation could be recognized more efficiently by the enzyme, mt SerRS apparently prefers G73 as the discriminator base, which is in agreement with a previous observation that a G73A mutation in tRNA GCU Ser caused an ϳ3fold reduction in activity (33). It can thus be postulated that tRNA UGA Ser possesses the less efficient A73 as a discriminator base to balance the serine-accepting activities of the two isoacceptors in vivo (Table I).
Next, to evaluate the effects of tertiary interactions in mt tRNA UGA Ser on serylation activity, point mutations were introduced at bases involved in four possible interactions (31): A15⅐U59, G18⅐U55, G19⅐C56, and U54⅐A58 (Fig. 4A). The A15⅐U59 interaction was shown not to be involved in the activity because the individual mutations A15U and U59A as well as the double mutation A15U,U59A all had relatively little effect in reducing the serylation activity (Table II). The variants G18U and U55G had k cat /K m values that were respectively reduced to 15 and 38% that of the wild type, which probably resulted from destabilization of D loop-T⌿C loop interaction induced by the mutations. Severely reduced serylation activity was induced by the mutations G19C and C56G, the relative  (Table II). However, the variant carrying both mutations (G19C and C56G) completely recovered its activity, suggesting that the Watson-Crick base pair G19⅐C56 supporting D loop-T⌿C loop interaction is indispensable for recognition.
U54⅐A58 interaction in the T⌿C loop was also revealed to be a strong recognition element for serylation. The U54A and A58U mutants exhibited significantly reduced serylation activity. Unlike the case of the G19⅐C56 tertiary base pair, no restoration in the activity was observed in the double mutant U54A,A58U, indicating that mt SerRS probably recognizes the U54⅐A58 interaction not only structurally but also sequencespecifically in the T⌿C loop.
When compared with the canonical cloverleaf structure, mammalian mt tRNA UGA Ser lacks six conserved residues, at positions 8, 16, 17, 21, 47, and 48, but contains one extra base pair (⌿26a⅐A43a) in the anticodon stem so as to form a characteristic pseudo-cloverleaf structure (31). On the other hand, chicken mt tRNA UGA Ser possesses the canonical cloverleaf struc-ture (34). Further, it has already been demonstrated that the chicken tRNA can be serylated by the native bovine mt SerRS. 2 To investigate the effect of the unusual cloverleaf structure of bovine mt tRNA UGA Ser on its serylation activity, we designed a variant mt tRNA UGA Ser possessing a canonical cloverleaf structure based on the chicken mt tRNA Ser sequence. All the residues of the chicken mt tRNA Ser apart from the D loop sequence were substituted for those of bovine mt tRNA UGA Ser (Fig. 4C). As this variant showed no significant change in serylation activity, it seems that the unusual cloverleaf structure of mammalian mt tRNA UGA Ser does not function as a key element for discrimination by mammalian mt SerRS.
Tiranti   this mutation etiologically induces deafness by altering the structure of the T⌿C loop in mt tRNA UGA Ser (35). In light of this, we constructed a variant based on bovine mt tRNA UGA Ser containing the same mutation ( Fig. 4A; G46.1 insertion) and examined its serylation activity. The mutation had little effect on the k cat /K m value, indicating that the deafness induced by the C7472 insertion is not caused by a defect in the serine-accepting activity.
From the foregoing, it is concluded that mammalian mt SerRS recognizes the mt tRNA UGA Ser T⌿C loop sequence-specifically and requires the D loop-T⌿C loop interaction sustained by the G19⅐C56 tertiary base pair for efficient aminoacylation, whereas the characteristic pseudo-cloverleaf structure per se is hardly involved in the serylation reaction.
tRNA GCU Ser Identity Elements for mt SerRS-Ueda et al. (33) previously examined the recognition sites of mt tRNA GCU Ser using partially purified enzyme, but the activity was too weak for the kinetic parameters of several tRNA variants with low serineaccepting activity to be determined. Therefore, we investigated the identity elements for mt tRNA GCU Ser using the recombinant enzyme with full activity. tRNA variants for mt tRNA GCU Ser were prepared by the in vivo expression system developed by Hayashi et al. (25), which allows variants to be easily purified in large quantities. Five A⅐U base pairs in the acceptor stem were replaced with G⅐C pairs (Fig. 4B) for efficient expression as described (25). tRNA GCU Ser expressed with the canonical sequence had a K m value 6 times higher than that of native tRNA (Table III). However, because Ueda et al. (33) previously revealed that the identity elements for mt tRNA GCU Ser are not located in the acceptor stem, we prepared a series of tRNA GCU Ser variants using the same expression system. The variations are shown in Fig. 4B, and the kinetic parameters are summarized in Table III. We first examined the serylation activity of the tRNA variant lacking a bulge (A43) in the anticodon stem, which is a conserved structure among mammalian mt tRNAs GCU Ser and is considered to be involved in tertiary interaction with the T⌿C loop (16). This deletion hardly reduced the serylation activity. de Bruijn and Klug (36) proposed a tertiary structural model of human and bovine mt tRNAs GCU Ser with four possible tertiary interactions, which are indicated by broken lines in Fig. 4B. Point mutations were introduced at the bases involved to evaluate the effects of these postulated tertiary interactions on serylation activity. The interactions U8⅐A60 A , and A9⅐U59 were revealed to have no relation to the activity because the double mutation U8A,A9U did not substantially reduce serylation (Table III). On the other hand, the double mutation U59A,A60 A U severely reduced the activity, indicating that mt SerRS recognizes U59 and A60 A base-specifically. Similarly, the U10⅐A57 tertiary interaction was not involved in serylation because the U10A variant showed only a slight decrease in activity. To examine the U54⅐A58 interaction in the T⌿C loop, point mutations were introduced at positions 54 and 58. The U54A mutation reduced the k cat /K m value by fourfifths, but this was relatively little compared with the drastic loss of activity by 2 orders of magnitude in the A58U variant (Table III). Because no activity was restored by introducing the double mutation U54A,A58U, which actually resulted in an even more severe decrease in the k cat /K m value, it can be assumed that the mt SerRS recognizes A58 base-specifically. With regard to the individual effects of the two point mutations on serylation activity, the theoretical activity derived by multiplying the k cat /K m values of the respective mutations should be in good agreement with the observed value. Because the theoretical value for the double mutation (0.0039) did in fact correspond well with the experimental value (0.0059), U54A and A58U were shown to be mutations that individually reduced the serylation activity. Finally, a point mutation was introduced at position 57, which is one of the strong footprint sites; this A57U mutation markedly reduced the activity.
From the foregoing, it can be concluded that mammalian mt SerRS recognizes the T⌿C loop sequence of mt tRNA GCU Ser but, unlike the case of mt tRNA UGA Ser recognition, it does not recognize its tertiary structure. The major determinants, A57 and A58, were shown to be the sites footprinted by the enzyme (Fig.  3A).
Dual Mode Recognition of the Two Isoacceptors by mt SerRS-Our findings on the identity elements of the two structurally dissimilar isoaccepting tRNAs raised the possibility that mammalian mt SerRS recognizes each tRNA by two distinct mechanisms. To examine this notion, we constructed mt tRNA UGA Ser variants in which the D arm and/or T⌿C loop were substituted by the D arm and T⌿C loop configurations of mt tRNA GCU Ser (Fig. 4A). As shown in Table II, the variant with the D arm substitution (i.e. lacking a D arm, as is the case in mt tRNA GCU Ser ) had a seriously decreased k cat /K m value, whereas the activity was increased 3-fold by substituting the T⌿C loop. The third variant, carrying both the D arm and T⌿C loop substitutions, showed similar serylation activity to that of the canonical transcript. These findings clearly indicated that the D loop-T⌿C loop tertiary interactions are required for the serylation of mt tRNA UGA Ser , whereas only the sequence of the T⌿C loop is important for the activity of mt tRNA GCU Ser . Aminoacylation of E. coli tRNA Ala by alanyl-tRNA synthetase (AlaRS) depends on a G3⅐U70 wobble base pair in the acceptor stem (37). Understanding such a unique identity determinant was greatly advanced by the demonstration that the minihelix variant, composed only of the amino acid acceptor-T⌿C helix, could be a good substrate for AlaRS (38,39). There- fore, to confirm the dual mode recognition of the two tRNA Ser species, we similarly prepared two minihelix tRNAs Ser based on the two isoacceptor tRNAs by connecting the acceptor stem and the T⌿C stem (Fig. 5A). As expected, the tRNA GCU Ser minihelix exhibited significant serylation activity (Fig. 5B, and Table III), whereas the tRNA UGA Ser minihelix showed none (Fig.  5B, and Table II).
The above results clearly demonstrated that mammalian mt SerRS recognizes the T⌿C loop of tRNA GCU Ser well in a sequencespecific manner. At the same time, the necessity of dissimilar tertiary interactions for the recognition of the two mt tRNAs Ser by mt SerRS was underscored by this experiment. Thus, there are both common and dissimilar features in the recognition mechanisms of the single enzyme for the two substrates.
Misacylation of mt tRNA Gln by mt SerRS-Because the T⌿C loop sequence of each serine tRNA was shown to be important for serylation, we searched mt tRNA sequences to discover whether a T⌿C loop sequence similar to that of either of the two serine tRNAs exists among the 20 non-cognate mt tRNAs. Although no other bovine mt tRNA has a sequence resembling the T⌿C loop of mt tRNA GCU Ser (40), three bovine mt tRNAs (for Gln, Glu, and Tyr) have T⌿C loop sequences the same as or similar to that of mt tRNA UGA Ser ; the mt tRNA Gln sequence is identical, that of mt tRNA Glu shows one nucleotide change (C56A), and tRNA Tyr has two changes (C56U/U59C). Therefore, using native mt tRNAs from bovine liver, we examined the mt SerRS misacylation activity for each of these tRNAs. The results are shown in Fig. 6 and Table IV. No activity was observed for the two tRNAs with T⌿C loop sequences similar to that of mt tRNA UGA Ser (mt tRNA Glu and mt tRNA Tyr ). To check the negative effects of these small changes in the T⌿C loop sequence on serylation activity, we constructed two tRNA UGA Ser variants, C56A and C56U,U59C, with T⌿C loop sequences respectively corresponding to those of mt tRNA Glu and mt tRNA Tyr (Fig. 4A). These mutants showed 60-fold or more reductions in their k cat /K m values (Table II), indicating that the variant T⌿C loop nucleotides in the mt tRNAs for Glu and Tyr function as strong negative determinants in preventing misacylation by mt SerRS, thereby maintaining the fidelity of translation, i.e. mt SerRS seems to reject these non-cognate mt tRNAs on the basis of their T⌿C loop sequences. mt tRNA Gln , whose T⌿C loop sequence is identical to that of mt tRNA UGA Ser , was found to be misacylated slightly but significantly by mt SerRS (Fig. 6), although its k cat /K m value was 4 orders of magnitude smaller compared with that of mt tRNA UGA Ser (Table  IV). This finding is a novel instance that appears to exist in the aminoacylation network to maintain a high level of translational fidelity. Further experiments were carried out to elucidate this notion.
Discrimination of mt tRNA UGA Ser from Non-cognate mt tRNA Gln by mt SerRS-As noted above, although mt tRNA Gln can be misacylated by mt SerRS, the activity is kept at a low level even though mt tRNA Gln possesses the same T⌿C loop sequence as mt tRNA UGA Ser . To exclude the possible influence of modified nucleotides, unmodified mt tRNA Gln was prepared by  Table IV. in vitro transcription. The data in Table IV show that this mt tRNA Gln transcript still had serylation activity 10-fold higher than that of the native tRNA. The misacylation activity of the unmodified mt tRNA Gln decreased ϳ42-fold when compared with the activity of unmodified mt tRNA UGA Ser . We next investigated the sequence elements embedded in mt tRNA Gln to reduce misacylation by mt SerRS. We focused on the G⅐U base pairs in the T⌿C stems of mt tRNA Gln and mt tRNA UGA Ser , of which the former has one and the latter has two. It has been reported that a G⅐U base pair in an RNA duplex distorts the conformation of the phosphodiester backbone (41). We therefore postulated that the number and/or location of the wobble base pairs in the T⌿C stem may have an effect on the recognition of the T⌿C loop by mt SerRS. To investigate this possibility, several tRNAs UGA Ser variants with mutations in their T⌿C stems were constructed by altering the number and/or location of the G⅐U base pairs. As shown in Table V, substituting either or both of the mt tRNA UGA Ser G⅐U base pairs with G⅐C (v1, v2, and v3) had no significant effect on serylation. However, the variant with a single G⅐U base pair at the same position as that in mt tRNA Gln (v4) showed a 15-fold reduction in serylation activity. This result suggests that a single base pair, G49⅐U64, in T⌿C stem of mt tRNA Gln acts as one of the negative determinants for serylation by mt SerRS. DISCUSSION In general, isoacceptor tRNAs possess common sequences or structures as identity elements for recognition by a cognate aminoacyl-tRNA synthetase and to exclude recognition by noncognate synthetases. Thus, determining common features of isoacceptors provides an important clue for predicting possible identity elements. In fact, a number of identity elements of various tRNAs so far studied correspond well to common sequences or structures, particularly the anticodon sequence, discriminator base, and acceptor stem (15). However, in the case of mammalian mitochondrial serine tRNAs, it is difficult to extract common features of the two isoacceptors, mt tRNA GCU Ser and mt tRNA UGA Ser . In addition to there being no apparent consensus sequence between these two tRNAs, structurally each has a distinct topology. One speculation had been that two species of SerRS might exist, one for each tRNA, as in the case of two threonyl-tRNA synthetases in yeast mitochondria (42). However, we demonstrated recently (18) that a single enzyme specifically recognizes the two isoacceptors. How, then, does this mt SerRS recognize and discriminate these two dissimilar tRNAs from non-cognate species?
The footprinting experiment with the mt SerRS-mt tRNA Ser complex revealed that the enzyme contacts both tRNAs at nearly the same position; the T⌿C loop and the bottom of the acceptor stem (Fig. 3). The series of mutation studies clearly demonstrated that one of the contact sites, the T⌿C loop sequence, is important for recognition in both cases. Herein can be found commonality in the recognition of the two dissimilar tRNAs by mt SerRS. However, we next noted that mt SerRS recognizes the T⌿C loop in each tRNA by distinctive mechanisms, because T⌿C loop-D loop tertiary interaction is required for the recognition of only mt tRNA UGA Ser . The different modes of recognition for the two substrates was accentuated by the fact that the minihelix variant of mt tRNA GCU Ser was a good substrate for mt SerRS, whereas that of mt tRNA UGA Ser was not (Fig. 5). Crystallographic studies have revealed that E. coli SerRS has a long helical arm consisting of an antiparallel, twostranded coiled coil in the N-terminal domain (27,28). In the crystal structure of T. thermophilus SerRS complexed with tRNA Ser , the N-terminal helical arm is tightly buried in the gulf between the T⌿C arm and the long extra arm of tRNA Ser with several backbone contacts, whereas the C-terminal region of another subunit is responsible for the catalytic activity (9). However, during the process of mitochondrial evolution in the eukaryotic cell, prokaryotic tRNAs Ser would have lost their characteristic extra arms, whereas the D arm was lost only in the case of mt tRNA GCU Ser . Accordingly, the N-terminal region of mt SerRS might have been obliged to recognize the T⌿C loop of mt tRNA Ser instead of the missing extra arm, because the catalytic core in the C-terminal region is well conserved in mt SerRS (18). Despite low homology in the N-terminal domain between mt SerRS and prokaryotic enzymes, a single, long helical arm was clearly predicted in the N-terminal region of bovine mt SerRS both by the CD spectrum of the recombinant enzyme and computational analysis of the helical structure using the COILS algorithm. It can be speculated that the predicted N-terminal helical arm of mt SerRS may also be responsible for mt tRNA recognition through interaction with the T⌿C loop. However, an investigation of the tertiary structure of the mt SerRS-mt tRNA Ser complex is required to elucidate how the N-terminal arm recognizes the T⌿C loop of each tRNA. The molecular basis of the precise mechanism of the dual mode recognition will be clarified by such a structural study, which is now in progress in our laboratory.
The error rate of translation has been estimated to be 10 Ϫ4 to 10 Ϫ5 (43), which is of the same order as misacylation (44), indicating that the fidelity of translation was maintained by the accuracy of aminoacylation. In Candida species, the codon CUG is known to be translated as serine instead of leucine by the tRNA CAG Ser responsible for this non-universal decoding (45,46). We previously reported that this tRNA CAG Ser can be aminoacylated not only with serine, but also with leucine to some extent in vitro as well as in vivo (47). This was the first report, and a unique instance, of a single tRNA in a natural organism being acceptable to more than one species of amino acid. In the present study, we unexpectedly found that mt tRNA Gln can be slightly but significantly misacylated by mt SerRS (Table IV  and Fig. 6), because mt tRNA Gln has a T⌿C loop sequence identical to that of mt tRNA UGA Ser . Although ambiguous specificity of mt aminoacyl-tRNA synthetase has been reported in unilateral aminoacylation between heterologous tRNAs (48), it was a surprise that mt SerRS cannot discriminate a cognate tRNA strictly even inside the mitochondrion. The fact that the recognition of mt tRNA Gln by mt SerRS in vitro is 3700 times lower than that of mt tRNA UGA Ser suggests that kinetic discrimination arising from competition between mt SerRS and mt GlnRS occurs in the mitochondrion to maintain the fidelity of mt translation. It can be speculated that a reduction in the number of tRNA species might impair the discriminatory ability of mt ARS. Our finding raises the possibility that other mammalian mt aminoacyl-tRNA synthetases may also misacylate non-cognate mt tRNAs, with the fidelity of mt translation being maintained by the kinetic discrimination of mt tRNAs in the network of ARSs. Further studies will certainly clarify this notion and, as a result, deepen our understanding of the mammalian mitochondrial translation system.