Decoding Mechanism of Non-universal Genetic Codes in Loligo bleekeri Mitochondria*

Background: Non-universal genetic codes are frequently found in animal mitochondrial decoding systems. Results: Five mitochondrial tRNAs from Loligo bleekeri were isolated and analyzed by mass spectrometry. Conclusion: AUA codon is deciphered as Met by CAU anticodon of tRNAMet(AUR), UGA codon is deciphered as Trp by τm5UCA anticodon of tRNATrp(UGR). Significance: AUA codon is deciphered by unmodified CAU anticodon through non-canonical A-C pairing. Non-universal genetic codes are frequently found in animal mitochondrial decoding systems. In squid mitochondria, four codons deviate from the universal genetic code, namely AUA, UGA, and AGA/AGG (AGR) for Met, Trp, and Ser, respectively. To understand the molecular basis for establishing the non-universal genetic code, we isolated and analyzed five mitochondrial tRNAs from a squid, Loligo bleekeri. Primary structures of the isolated tRNAs, including their post-transcriptional modifications, were analyzed by mass spectrometry. tRNAMet(AUR) possessed an unmodified cytidine at the first position of the anticodon, suggesting that the AUA codon is deciphered by CAU anticodon via non-canonical A-C pairing. We identified 5-taurinomethyluridine (τm5U) at the first position of the anticodon in tRNATrp(UGR). τm5U enables tRNATrp to decipher UGR codons as Trp. In addition, 5-taurinomethyl-2-thiouridine (τm5s2U) was found in mitochondrial tRNAs for Leu(UUR) and Lys in L. bleekeri. This is the first discovery of τm5U and τm5s2U in molluscan mitochondrial tRNAs.

Non-universal genetic codes are frequently found in animal mitochondrial decoding systems. In squid mitochondria, four codons deviate from the universal genetic code, namely AUA, UGA, and AGA/AGG (AGR) for Met, Trp, and Ser, respectively. To understand the molecular basis for establishing the non-universal genetic code, we isolated and analyzed five mitochondrial tRNAs from a squid, Loligo bleekeri. Primary structures of the isolated tRNAs, including their post-transcriptional modifications, were analyzed by mass spectrometry. tRNA Met (AUR) possessed an unmodified cytidine at the first position of the anticodon, suggesting that the AUA codon is deciphered by CAU anticodon via non-canonical A-C pairing. We identified 5-taurinomethyluridine (m 5 U) at the first position of the anticodon in tRNA Trp (UGR). m 5 U enables tRNA Trp to decipher UGR codons as Trp. In addition, 5-taurinomethyl-2-thiouridine (m 5 s 2 U) was found in mitochondrial tRNAs for Leu(UUR) and Lys in L. bleekeri. This is the first discovery of m 5 U and m 5 s 2 U in molluscan mitochondrial tRNAs.
Variation in the genetic code is a characteristic feature of mitochondrial decoding systems. In animal mitochondria, six codons deviate from the universal genetic code (1). The AUA codon (for Ile) specifies Met in most metazoan mitochondria. The AAA codon (for Lys) changes to assign Asn in echinoderm and some platyhelminth mitochondria. The AGA and AGG (AGR) codons (for Arg) are used for Ser in most invertebrate mitochondria, for Gly in tunicate mitochondria, and as a stop codon in vertebrate mitochondria. The UGA stop codon is used for Trp in all animal mitochondria, and the UAA stop codon is used for Tyr in nematode mitochondria.
To clarify the molecular basis for deciphering the non-universal genetic codes in animal mitochondria, we have been carrying out a project to isolate and analyze individual mitochondrial (mt) 3 tRNAs responsible for the non-universal genetic codes in a variety of animal phyla. We have proposed a hypothesis that the acquisition of wobble modifications is a prerequisite for genetic code alteration in mitochondria (2). In bovine mitochondria, we identified and determined the chemical structure of 5-formylcytidine (f 5 C) at the anticodon first (wobble) position of tRNA Met (3). f 5 C is required to decipher the AUA codon as Met in mitochondria of mammals (3,4), nematode (Ascaris suum) (5) and insect (Drosophila melanogaster) (6). In addition, we discovered 5-taurinomethyluridine (m 5 U) and its 2-thio derivative (m 5 s 2 U) at the wobble position of human and bovine mt tRNAs responsible for NNR codons (7,8). Because an unmodified U at the wobble position enables tRNAs to decode any of the four codons in family boxes according to the mitochondrial four-way wobble rule (1,9), the wobble U in tRNAs responsible for NNR codons must be modified. In fact, m 5 U is used to assign the UGA codon to Trp in mammals and vertebrates (2,4). In ascidian (tunicate) mitochondria (10), m 5 (s 2 )U is required to decipher all four non-universal genetic codes (AUA for Met, UGA for Trp, and AGR for Gly).
Four non-universal genetic codes exist in the mitochondrial decoding system of the squid Loligo bleekeri, namely, AUA, UGA, and AGA/AGG (AGR) for Met, Trp, and Ser, respectively (Table 1) (11). We previously attempted to elucidate the underlying molecular mechanism that deciphers these non-universal codons by isolating and analyzing mt tRNAs from L. bleekeri. A gene encoding a tRNA Ser containing a GCT anticodon is encoded in the mtDNA of L. bleekeri. To decode the non-universal AGR codons in addition to the universal AGY codons as Ser, the GCU anticodon of this tRNA must be modified posttranscriptionally. Indeed, we identified 7-methylguanosine (m 7 G) at the wobble position of the mt tRNA Ser isolated from L. bleekeri liver (12), indicating that all four AGN codons are deciphered by the single tRNA Ser bearing the m 7 GCU anticodon. The unique four-way wobbling mediated by m 7 G was originally detected in the starfish mitochondrial decoding system (13). In the case of the AUA codon, our group used a postlabeling method to identify a partial f 5 C modification at the wobble position of squid mt tRNA Met (14), implying that the AUA codon is deciphered by the f 5 CAU anticodon, as observed in mammals, nematode, and insect. However, evidence for this modification relied upon the detection of weak spot with a high background on a two-dimensional thin-layer chromatogram (two-dimensional TLC). To define conclusively the mechanism for deciphering the AUA codon, the frequency of the f 5 C modification needs to be reliably estimated. There have been few studies over the past decade on squid mitochondrial decoding systems because of technical limitations in the isolation and analysis of mt tRNAs. By using the efficient RNA isolation technique based on the reciprocal circulating chromatography (15) and highly sensitive mass spectrometric methods for analyzing RNA molecules (16), we successfully analyzed the post-transcriptional modifications of L. bleekeri mt tRNAs that are responsible for the non-universal genetic codes. Moreover, we discuss our results in the wider context of the evolutionary reorganization of decoding systems in animal mitochondria.

EXPERIMENTAL PROCEDURES
Isolation of mt tRNAs-The procedure for isolation of mt tRNAs from edible muscle of L. bleekeri was largely the same as that used for ascidian mt tRNAs reported previously (10). After separation by DEAE-cellulose column chromatography, 2400 A 260 units of the crude tRNA fraction were isolated from 280 g of squid muscle. Total tRNA (700 A 260 units) was obtained from 1200 A 260 units of the crude tRNA fraction by removing contaminating polysaccharides with TRIzol-LS (Invitrogen) followed by rinsing with chloroform. For the isolation of individual mt tRNAs, the following 5Ј-EC amino-modified DNA probes (Sigma-Aldrich) were designed using Raccess software (17) and used for reciprocal circulating chromatography (RCC) (15): 5Ј-GGGGTATGAACCCAACAGCTTATTTTTTAGCTTAC-3Ј for tRNA Met (AUR), 5Ј-TTGAAAGCCTTCAGTTTAACTTAA-CTTAAAATCTT-3Ј for tRNA Trp (UGR), 5Ј-AAAGGTAATTA-GGAATAAAATAAAGCTGCTAACTT-3Ј for tRNA Ser (AGN), 5Ј-AAATTCTATGCACTGATCTGCCATCTTAAT-3Ј for tRNA Leu (UUR), and 5Ј-TCTAGTGCTTACTCATTCGGCCAC-TTAATA-3Ј for tRNA Lys (AAR). The tDNA sequences were obtained from the NCBI database of L. bleekeri mtDNA (accession number NC_002507) (18,19). The DNA probes were covalently immobilized on NHS-activated Sepharose 4 Fast Flow (GE Healthcare) and packed into respective tip columns for the RCC instrument. The five tRNA species were simultaneously isolated from 700 A 260 units of total tRNA by RCC as described (15).
Mass Spectrometric Analysis of mt tRNAs-Each purified tRNA was digested with RNase T 1 and then analyzed by capillary liquid chromatography (LC)/nanoelectrospray ionization mass spectrometry (MS) as described (16). A linear ion traporbitrap hybrid mass spectrometer (LTQ Orbitrap XL; Thermo Fisher Scientific) equipped with a custom-made nanospray ion source and a splitless nanoHPLC system (DiNa; KYA Technologies) was employed in this study. Modified bases were assigned by comparing observed m/z values of RNase T 1 -digested fragments to the calculated values of these fragments. The bases were allocated on the basis of the sequences of the fragments, which were inferred from the collision-induced dissociation (CID) spectra and assisted by referring to evolutionary conservation of each modified base in tRNAs. N 2 -methylguanosine (m 2 G) can be distinguished from 1-methylguanosine (m 1 G) by examining the 3Ј-terminal structure generated by RNase T 1 . When RNase T 1 cleaves tRNA on the 3Ј side of m 2 G, a 3Ј-phosphate is produced. On the other hand, when the 3Ј side of m 1 G is cleaved by RNase T 1 , a 2Ј,3Ј-cyclic phosphate (Ͼp) is primarily generated. Mono-or dinucleotide fragments passed through the LC and were not detected. The percent frequencies of modifications were calculated from the ratios of the mass chromatogram peak heights of RNA fragments containing modified bases to those of RNA fragments lacking modified bases.
Nucleoside analysis of tRNAs was performed as described (16,20). Purified tRNA Met (AUR) and yeast phenylalanine tRNA (Roche Applied Science) were digested into nucleosides using nuclease P1 (Wako Pure Chemical Industries) and bacterial alkaline phosphatase (BAP.C75; Takara) for 3 h at 37°C. The digests were then analyzed by a linear ion trap-orbitrap hybrid mass spectrometer (LTQ Orbitrap XL).

RESULTS
Five species of squid mt tRNAs were isolated from 700 A 260 units of total tRNA by RCC, namely, tRNA Met (AUR), tRNA Trp (UGR), tRNA Ser (AGN), tRNA Leu (UUR), and tRNA-Lys (AAR). Each individual tRNA was faintly detectable on the polyacrylamide gel of the total eluted fraction from the RCC column ( Fig. 1). Each band was excised, and tRNA was eluted from the gel pieces for mass spectrometric analysis. Although the final yield of each tRNA could not be precisely measured, we estimated the yield to be within the range of 10 -300 fmol.
Isolated tRNAs were digested by RNase T 1 and subjected to capillary LC/nanoelectrospray ionization MS to analyze the chemical structures of the modifications. RNA fragments from each tRNA were efficiently separated by capillary LC on the basis of their lengths, base compositions, and sequences ( Fig. 2A). Singly and multiply charged anions of RNA fragments were detected with high mass resolution (30,000 units). Modified bases in the fragments were identified on the basis of the precise molecular mass of each RNA fragment, which was calculated by deconvoluting the observed m/z values from the mass spectra (s u p p l e m e n t a l T a b l e 1). Each fragment was further analyzed by an MS/MS experiment using CID to determine the positions of detected modifications within the tRNA sequence (Fig. 2B). The modifications were unambiguously positioned in the sequence by the assignment of y-and c-series product ions (21).
In mt tRNA Met (AUR), we clearly detected the anticodoncontaining fragment (⌿UCAUt 6 ACCCCAAAAAm 5 UGp) bearing N 6 -threonylcarbamoyladenosine (t 6 A) and 5-methyluridine (m 5 U) at positions 37 and 48, respectively (Fig. 2). Although f 5 C has been reported to be partially introduced at the wobble position (14), we could not detect f 5 C at this position (supplemental Fig. 1). This result suggested that the AUA codon is deciphered as Met by the CAU anticodon via non-canonical A-C pairing. We confirmed the presence of 1-methyladenosine (m 1 A) at position 9, N 2 -methylguanos-ine (m 2 G) at position 11, and N 2 ,N 2 -dimethylguanosine (m 2 2 G) at position 26 ( Fig. 3A and supplemental Table 1). In addition, we newly found a methylated U at position 48 (Fig.  2B), which we determined to be m 5 U by co-injection analysis with authentic m 5 U in total nucleosides of yeast tRNA Phe (supplemental Fig. 2).
In mt tRNA Trp (UGR), we detected the anticodon-containing fragment CUUm 5 UCAms 2 i 6 Am 1 GϾp, which bears m 5 U at the wobble position, 2-methylthio-N 6 -isopentenyladenosine (ms 2 i 6 A) at position 37, and m 1 G at position 38 (Fig. 2). This is the first instance of the detection of m 5 U in molluscan mt tRNA. The relative frequency of m 5 U was calculated on the basis of the signal intensity ratio of the modified and unmodified RNA fragments. The wobble base was modified to m 5 U in 89% of mt tRNA Trp (UGR), indicating that UGR codons are mainly deciphered by m 5 UCA anticodons (Fig. 3A), whereas the residual 11% had unmodified U at the wobble position. The hypomodified tRNAs are likely to be the result of their in vivo status rather than different hybridization efficiency with the DNA probes because the length of probes used in this study are long enough to trap all tRNAs with different modification status. In addition, m 1 A and m 2 2 G were found at positions 9 and 26, respectively ( Fig. 3A and supplemental Table 1). Moreover, we detected monomethylated uridines (Uϩme) with unidentified structures at positions 12 and 13 ( Fig. 3A and supplemental Table 1).
L. bleekeri mt tRNA Ser (AGN) contains m 7 G at the wobble position and t 6 A at position 37 (12). The presence of both modifications was confirmed using MS analysis (Figs. 2A and 3A, and supplemental Table 1). m 7 G is distinguishable from the other methylguanosines because RNase T 1 does not cleave the 3Ј end of m 7 G. In addition, because the N-glycosyl bond of m 7 G is unstable, the m 7 G base dissociates easily from the RNA fragment upon CID, generating product ions lacking m 7 G base (Fig.  2B). The mass chromatogram showed that the wobble base was completely modified to m 7 G (Fig. 3A).
Having detected m 5 U in mt tRNA Trp (UGR), we further analyzed mt tRNA Leu (UUR) and mt tRNA Lys (AAR) to identify additional instances of m 5 U and its derivative in squid mt tRNAs. As expected, the wobble bases of both tRNAs were modified to m 5 s 2 U (Fig. 2). The mass chromatograms of the anticodon-containing fragments with different modification statuses (Fig. 3B) showed that 88% of mt tRNA Leu (UUR) and 90% of mt tRNA Lys (AAR) contained m 5 s 2 U, whereas only 7% of both tRNAs contained m 5 U. The remaining molecules had a 2-thiouridine or an unmodified U at the wobble position. Other modifications were found in mt tRNA Leu (UUR) at positions 6 and 10 (m 2 G) and at positions 9 and 37 (m 1 G) (Fig. 3B) and in mt tRNA Lys (AAR) at position 9 (m 1 G), position 10 (m 2 G), and position 37 (t 6 A) (Fig. 3B). Monomethylated uridines (Uϩme) with unidentified structures were also found at positions 16, 48, and 52 ( Fig. 3B and supplemental Table 1). Uϩme at position 48 was presumed to be m 5 U as observed in mt tRNA Met (AUR).

DISCUSSION
We have shown here that L. bleekeri mt tRNA Met (AUR) possesses an unmodified C at the wobble position (Fig. 3A). Our group previously reported that the wobble position was par- tially modified to f 5 C in the same tRNA (14). In the study, a weak f 5 C spot along with a strong C spot was detected on the two-dimensional TLC by the postlabeling method, suggesting a false positive. Otherwise, we might use a different strain of L. bleekeri for this study. There might be some individual differences of squid variably expressing a putative tRNA-modifying   (36). Pseudouridines were determined previously (12,14). Frequency of modifications at the wobble position and position 37 is indicated in parentheses. MARCH 15, 2013 • VOLUME 288 • NUMBER 11 enzyme for f 5 C formation. Judging from the sensitivity of our MS analysis, if there is only 1% of f 5 C in the anticodon-containing fragment, we can surely detect it. Thus, if a partial f 5 C modification does exist, its frequency must be Ͻ1%. In any case, we demonstrated here that a large majority of mt tRNA Met (AUR) is occupied by unmodified C at the wobble position, indicating that AUA codon is deciphered by CAU anticodon via non-canonical A-C pairing in L. bleekeri mitochondria. This may be a remnant of a primitive decoding system 0that has been used early in evolution, because the domain-specific wobble modifications might have been acquired late during evolution, as suggested by the phylogenetic study on tRNA-modifying enzymes (22).

Squid Mitochondrial tRNAs
Some instances of the AUA decoding by a CAU anticodon have been reported in other mitochondrial systems. In D. melanogaster (6), there are two species of mt tRNAs Met ; one has a CAU anticodon with t 6 A at position 37, the other has an f 5 CAU anticodon with an unmodified A at position 37. This indicates that the AUA codon can be read by a CAU anticodon with the assistance of t 6 A37 because t 6 A37 stabilizes tRNA binding to the A site codon (23,24), or by f 5 CAU without the assistance of t 6 A37. In this study, we detected t 6 A37 in mt tRNA Met (AUR) (Fig. 3A). The frequency of the t 6 A modification in mt tRNA Met (AUR) was estimated to be 93% according to the MS data (Fig. 3A). The high modification efficiency indicates the importance of t 6 A37 for the decoding of AUA as Met in L. bleekeri mitochondria. In addition to the anticodon and 3Ј-adjacent modifications, decoding efficiency can be affected by other regions in tRNAs (25,26). In the case of Hirsh suppressor tRNA Trp with G24A and A9C mutations, UGA codon is efficiently deciphered by the CCA anticodon, in which C34 of anticodon pairs with A3 of codon by a non-standard geometry with a single hydrogen bond between N4 imino group of C34 and N1 of A3 (26). These suppressor mutations are considered to enable C34-A3 pairing by facilitating the distortion of tRNA body at A/T state during decoding (26). Thus, there might be some elements in mt tRNA Met (AUR) that enable CAU anticodon to recognize AUA codon efficiently. In Saccharomyces cerevisiae (27,28) and mosquito (Aedes albopictus) (29), mt tRNAs Met are known to have an unmodified CAU anticodon. In S. cerevisiae, the initiator and elongator mt tRNAs Met have CAU anticodons with m 1 G37 and t 6 A37, respectively (27,28). Collectively, the AUA decoding by a CAU anticodon with the assistance of the modified bases at position 37 might be a general rule.
In animal mitochondria, three strategies are used to decipher AUA codon as Met. The first strategy involves the f 5 CAU anticodon in mt tRNA Met from mammals, nematode, and Drosophila. Indeed, AUA decoding by f 5 CAU anticodon has been confirmed biochemically (30,31). The second strategy involves the m 5 UAU anticodon in ascidian (Halocynthia roretzi) mitochondria. In general, the xm 5 (s 2 )U modification prevents misreading of near cognate codons ending in pyrimidines (NNY) (9). m 5 U would enable tRNA to efficiently decode AUR codons as Met. The third strategy involves the unmodified CAU anticodon described in this study. Non-canonical A-C pairing might be involved in AUA decoding by the CAU anticodon, in which t 6 A37 would stabilize the codon-anticodon pairing. Biochemical and structural studies will be required to provide mechanistic insights into this type of decoding. The distribution of different types of AUA decoding in animal mitochondria was assumed to be associated with AUR codon usage. In both human and L. bleekeri mitochondria, AUA codons are used about four times more frequently than AUG codons, indicating that AUA is an abundant codon, whose usage is independent of the type of decoding. Therefore, we were unable to draw any meaningful correlation between AUR codon usage and the types of AUA decoding. The three types of AUA decoding mentioned above rely heavily on the specificities of tRNAmodifying enzymes. In the first type, mt tRNA Met , containing a CAU anticodon, may be recognized by a putative f 5 C-modifying enzyme that has not been identified yet. In the second type, a m 5 U-modifying enzyme, probably GTPBP3/MTO1 (4), recognizes mt tRNA Met with a UAU anticodon. In the third type, although there is no enzyme for f 5 C formation, t 6 A-modifying enzymes (4) recognize mt tRNA Met with a CAU anticodon and introduce t 6 A at position 37. Further studies directed toward determining the evolutionary distribution of these three types of AUA decoding and characterizing tRNA-modifying enzymes will deepen our understanding of AUA decoding in animal mitochondria.
In L. bleekeri, we identified m 5 U in mt tRNA Trp (UGR), and m 5 s 2 U in mt tRNA Leu (UUR) and mt tRNA Lys (AAR). This is the first reported instance of m 5 U and m 5 s 2 U in molluscan mt tRNAs. m 5 (s 2 )U was first identified in mammalian mt tRNAs responsible for Leu(UUR), Trp, Lys, Glu, and Gln (4, 7). Moreover, we also found m 5 (s 2 )U in fish, 4 suggesting that m 5 (s 2 )U may be conserved in vertebrate mt tRNAs. In addition, m 5 (s 2 )U was found in three mt tRNAs responsible for non-universal genetic codes in ascidian (H. roretzi) mitochondria (10). On the other hand, m 5 (s 2 )U was not found in yeast or nematode mt tRNAs, which use 5-carboxymethylaminomethyluridine (cmnm 5 U) instead (32)(33)(34). Due to the chemical similarity between cmnm 5 U and m 5 U, both nucleotides are thought to be synthesized by a similar biosynthetic pathway, which, in the case of m 5 U, results in the replacement of glycine with taurine (7,33). To identify the phylogenetic junction at which m 5 U replaced cmnm 5 U in animal phyla, it will be necessary to analyze additional invertebrate mt tRNAs. Specifically, m 5 U seems to have emerged in cephalopoda mt tRNAs. It is intriguing to note that m 5 s 2 U is present in L. bleekeri mt tRNA Leu (UUR). In general, 2-thiouridine derivatives occur in three tRNA species for Lys, Glu, and Gln. In mammalian mitochondria, m 5 s 2 U is found in mt tRNAs for Lys, Glu, and Gln, but not in mt tRNAs for Leu(UUR) or Trp (4). This is because MnmA, a 2-thiouridylase, strictly discriminates the anticodons of tRNAs (35). In human mitochondria, MTU1, a homolog of MnmA, recognizes mt tRNAs for Lys, Glu, and Gln (33). A MTU1 homolog in L. bleekeri may have evolved to recognize mt tRNA Leu (UUR).
Another intriguing finding is that m 5 U is present at position 48 in the extra loop of squid mt tRNA Met (AUR), and probably in mt tRNA Lys (AAR) as well. In many tRNAs, m 5 U can be found at position 54 in the T-loop, and m 5 U54-methyltransferase is widely distributed across domains of life. Although various modifications occur in the extra loop of many tRNAs, this is the first instance of tRNAs containing m 5 U48. We speculate that a specific methyltransferase for m 5 U48 exists in L. bleekeri and that m 5 U48 is widely distributed in cephalopod and molluscan mt tRNAs. m 5 C is often observed at positions 48 and 49 in the extra loop of tRNAs (36), and m 5 C49 contributes to the structural stabilization of tRNA (37,38). We speculate that m 5 U48 might have a similar function for the stabilization of tRNAs. In addition, m 1 G38 in squid mt tRNA Trp (UGR) is a unique modification that has never been detected in any other tRNAs to date. m 1 G37 is commonly found in many tRNAs, including squid mt tRNA Leu (UUR). Future studies should examine whether mitochondrial m 1 G37-methyltransferase in L. bleekeri has acquired the specificity to introduce m 1 G at position 38 in mt tRNA Trp (UGR). There are still four unidentified methyluridines at positions 12 and 13 in tRNA Trp (UGR) and positions 16 and 52 in tRNA Lys (AAR). These modifications might be required for tRNA stabilization as proposed in general (39).
The decoding system for non-universal genetic codes in L. bleekeri mitochondria is summarized in Fig. 4. The AUA codon is deciphered as Met by the CAU anticodon of mt tRNA Met (AUR), UGA codon is deciphered as Trp by the m 5 UCA anticodon of mt tRNA Trp (UGR), and AGR codons are deciphered as Ser by the m 7 GCU anticodon of mt tRNA Ser (AGN).