A novel wobble rule found in starfish mitochondria. Presence of 7-methylguanosine at the anticodon wobble position expands decoding capability of tRNA.

In the starfish mitochondrial (mt) genome, codons AGA and AGG (in addition to AGU and AGC) have been considered to be translated as serine. There is, however, only a single candidate mt tRNA gene responsible for translating these codons and it has a GCT anticodon sequence, but guanosine at the first position of the anticodon should base pair only with pyrimidines according to the conventional wobble rule. To solve this enigma, the mt tRNA GCUser was purified, and sequence determination in combination with electrospray liquid chromatography/mass spectrometry revealed that 7-methylguanosine is located at the first position of the anticodon. This is the first case in which a tRNA has been found to have 7-methylguanosine at the wobble position. It is suggested that methylation at N-7 of wobbling guanosine endows the tRNA with the capability of forming base pairs with all four nucleotides, A, U, G, and C, and expands the repertoire of codon-anticodon interaction. This finding indicates that a nonuniversal genetic code in starfish has been generated by base modification in the tRNA anticodon.

An unusual mode of wobble base interaction with mRNA codons exists in mitochondria (1,2). Uridine at the wobble position base pairs with all four nucleotides, while modification from uridine to 5-carboxymethylaminomethyluridine (cmnm 5 U) 1 prevents pairing with pyrimidines. Codons with a pyrimidine in the third position are translated by tRNAs having unmodified guanosine at the first anticodon position. This rule was first proposed for fungal mitochondria (3,4) and has been considered to hold for most codon boxes in animal mitochondria. However, a few exceptional codon-anticodon pairings have been also observed in this organelle.
In most animal mitochondria, the methionine codons AUA and AUG are translated by a single methionine tRNA which has 5-formylcytidine (f 5 C) in the first anticodon position, indicating that f 5 C plays a role similar to cmnm 5 U (5). In starfish and sea urchin mitochondria, however, AUA is restored to the isoleucine codon as in the universal genetic code (6 -9), while AAA and AGR (R ϭ A, G) are presumably translated as lysine and serine, respectively (6 -9); these codon assignments are different from those in other known animal mitochondrial systems. For example, AGR codons are utilized for the termination and glycine codon in mammalian (10) and ascidian (11) mitochondria, respectively. In Drosophila, AGA is the serine codon but the other AGR codon, AGG, is absent from the mitochondrial genome (12)(13)(14). In contrast, both AGA and AGG appear in reading frames encoded on starfish and sea urchin mt DNAs and are presumably assigned to serine judging from the sequence alignment (6 -9). In nematodes, AGN codons are translated with a serine tRNA whose gene possesses a TCT anticodon sequence (15). We have shown that unmodified uridine is located at the wobble position of the tRNA (16), indicating that the codon-anticodon interaction definitely follows the mitochondrial wobble rule. In contrast, the mitochondrial gene encoding the tRNA that translates all AGN codons in starfish and sea urchin has the anticodon GCT (6 -9). The question that arises from this is how the GCT anticodon can recognize the AGA and AGG codons, in addition to AGY (Y ϭ U, C), in echinoderms.
We have previously speculated on possible answers to this question (1,2,17). One possibility is that G at the first anticodon position of the starfish mt tRNA GCU Ser might be modified so as to allow it to pair with all four AGN codons. Alternatively, it might be that a region(s) in the tRNA other than the anticodon influences the decoding ability of AGR codons, because most metazoan mt tRNA GCU Ser have unusual secondary structures in which the D arm is lacking or incomplete (2,17). This truncated D arm could be responsible for G⅐R pairing, in addition to G⅐Y pairing, between the anticodon and codons. Another feature is the G⅐C pair at the bottom of the anticodon stem, which is present in the tRNA GCU Ser of mitochondria of invertebrates, but not in those of most vertebrates, in which its place is taken by an A⅐U pair (2,17,18). Since tRNA GCU Ser of all metazoan mitochondria except for that of nematode have identical anticodon loops (15), there could be some structural differences in regions other than the anticodon loop, depending on which codons, AGY (vertebrates) or AGN (N ϭ A, G, C, or U) (invertebrates), are translated as serine.
With the aim of determining which of these speculations is actually the case, the starfish mt tRNA GCU Ser responsible for the noncanonical decoding in the AGN codon box was purified and sequenced, and the modified nucleoside content was determined by electrospray liquid chromatography-mass spectrom-* This work was supported by Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Science, Sports and Culture of Japan (to K. W. and T. U.), by grants from the Human Frontier Science Program Organization (to K. W. and J. A. M), and the National Institutes of Health, NIGMS Grant GM29812 (to J. A. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed: Tel. and Fax: 81-3-5800-6950; E-mail: kw@kwl.t.u-tokyo.ac.jp. 1 The abbreviations used are: cmnm 5 U, 5-carboxymethylaminomethyluridine; mt, mitochondrial; m 7 G, 7-methylguanosine; m 5 C, 5-methylcytidine; m 3 C, 3-methylcytidine; f 5 C, 5-formylcytidine; ms 2 t 6 A, 2-methylthio-N 6 -threonylcarbamoyladenosine; m 6 t 6 A, N 6 -methyl-N 6threonylcarbamoyladenosine; Y, wyosine; ES, electrospray; LC, liquid chromatography; MS, mass spectrometry. etry (ES-LC/MS). The results showed the first of the speculations posited above to be correct; guanosine at the wobble position of starfish mt tRNA GCU Ser has been completely converted to 7-methylguanosine (m 7 G). This strongly suggests that m 7 G is capable of base pairing with all four nucleotides, thus resulting in a genetic code change in echinoderm mitochondria. Based on the primary structure of the mt tRNA GCU Ser , we also present a scenario that could explain evolutionary changes in the genetic code that have occurred in animal mitochondria.

EXPERIMENTAL PROCEDURES
Purification of tRNA from Asterina amurensis-The starfish A. amurensis was harvested at Miyako, Iwate prefecture, Japan. The ovaries were removed, immediately frozen on dry ice, and stored until RNA preparation. Crude ovary RNA was prepared by phenol extraction and precipitated by ethanol. It was purified further by DE52-cellulose column chromatography using an elution solvent containing 0.7 M NaCl, 20 mM Tris-HCl (pH 7.5), and 10 mM magnesium acetate. The RNA was recovered by ethanol precipitation and applied onto a Q-Sepharose column to eliminate residual polysaccharides. RNA was eluted by the same buffer used for the DE52 column chromatography. Serine tRNA was purified by hybridization methods using the 3Ј biotinylated oligonucleotide CGAAACTTCTATGGATTTGAACCACGATT, immobilized on a streptavidin-agarose matrix as described previously (16,19,20). This 30-mer oligonucleotide complementary to the sequence of the 3Ј region of the tRNA was designed based on our unpublished DNA sequence data (DDBJ accession no. is D17543).
Sequence Determination of Starfish mt tRNA GCU Ser -The purified tRNA GCU Ser was subjected to sequencing by the methods of Donis-Keller (21) and Kuchino et al. (22). RNase T1, RNase PhyM, and T4 RNA ligase were purchased from Pharmacia Biotech Inc., RNase T2 and RNase CL3 from Seikagaku Kogyo, and Escherichia coli alkaline phosphatase from Takara Shuzo. T4 polynucleotide kinase and other chemical reagents of analytical grade were from Wako Pure Chemicals.
[ 32 P]pCp and [␥-32 P]ATP were obtained from Amersham, Japan. In Kuchino's method, nucleotides in the tRNA were analyzed by twodimensional TLC utilizing two different development systems. System A consisted of isobutyric acid/concentrated ammonia/water (66:1:33 by volume), in the first dimension, and 2-propanol/HCl/water (75:15:15 by volume) in the second dimension. In system B, the first dimension was the same as that for system A, but 0.1 M sodium phosphate (pH 6.8)/ ammonium sulfate/1-propanol (100 ml:60 g:2 ml) (22, 23) was used for the second dimension. The base numbering conforms to the literature

Directly Combined High Performance Liquid Chromatography/Electrospray Mass Spectrometry (ES-LC/MS) of Nucleosides from Hydrolyzed mt tRNA GCU
Ser -Two micrograms of purified tRNA GCU Ser were digested to nucleosides with nuclease P1 (Life Technologies, Inc.), phosphodiesterase I (type VII; Sigma), and bacterial alkaline phosphatase (Calbiochem) as described previously (24). The hydrolysate was injected directly onto the column without further purification.
The nucleosides were fractionated on a 250 ϫ 2-mm LC-18S column with a matching 20 ϫ 2-mm guard column (Supelco). The solvent system consisted of 5 mM ammonium acetate (pH 5.3; buffer A) and acetonitrile/H 2 O (4:6, v/v; buffer B) at a flow rate of 300 l/min, using a complex multilinear gradient from 100% A to 100% B (25).
The liquid chromatograph (Hewlett-Packard 1090 with a diode array detector) was interfaced directly to a Fisons Quattro II mass spectrometer (Micromass, Inc.) fitted with an electrospray ion source. The entire effluent was conducted directly into the mass spectrometer. The interface temperature was 180°C. Instrument voltages were: needle 3.08 kV and lens 0.23 kV; the cone was ramped from 35 to 5 V over the acquisition mass range from 100 to 500 m/z units. One-second scans (including 0.1-s interscan delay) were acquired continuously throughout the 50-min analysis.
Cleavage of Starfish mt tRNA GCU Ser at m 7 G Position by Alkaline Treatment-To verify the presence and the position of m 7 G, purified tRNA GCU Ser was treated as follows (26). First, purified tRNA GCU Ser was labeled with [ 32 P]pCp at its 3Ј end and purified by denaturing polyacrylamide gel electrophoresis. Labeled tRNA was incubated in 0.1 M Tris-HCl (pH 9.5) at 50°C for 4.5 h. tRNA was recovered twice by ethanol precipitation and dissolved in 0.3 M aniline (pH 3.5). The solution was incubated at 50°C for 4 h and subjected to denaturing gel electrophoresis. Yeast phenylalanine tRNA possessing m 7 G in the variable loop was treated identically as a control experiment.

Purification and Primary Sequence of Starfish mt tRNA GCU
Ser -Mitochondrial tRNA GCU Ser was prepared from a starfish ovary RNA fraction and purified as described under "Experimental Procedures." The tRNA was quite pure judging from the polyacrylamide gel electrophoretic pattern. A portion of the tRNA sample was sequenced (21,22) and the remainder was hydrolyzed to nucleosides and analyzed by electrospray LC/MS.
The purified tRNA was first analyzed using Donis-Keller's (21) method. As shown in Fig. 1, the nucleotide representing the first letter of the anticodon was resistant to RNase T1, suggesting that this guanosine is, therefore, likely to be completely modified. The first nucleotide of the anticodon loop (position 32), which is predicted to be C from the DNA sequence, was resistant to RNase CL3, suggesting that this C is also modified. The nucleotide at position 37, expected to be an A from the gene sequence, is probably modified, since no band was observed in the RNase U2 lane. The remainder of the nucleotide sequence predicted from this analysis corresponds to that of the gene (DDBJ accession no. is D17543), and shows that the tRNA has not undergone editing but has been posttranscriptionally modified.
Sequence Location of Modified Nucleotides in mt tRNA GCU Ser -The nucleotide sequence was also determined by Kuchino et al.'s (22) method using two-dimensional TLC. The nucleotide at the first position of the anticodon loop (position 32) gave two distinct spots consistent with 3-methylcytidine (m 3 C) and 5-methylcytidine (m 5 C) on the TLC plate (Fig. 2a). Although m 5 C is about twice as abundant as m 3 C judging from the intensity of the spots on TLC, it suggests that two different methylases may be involved in the modification at this site. TLC analysis of the nucleotide at the wobble position (position 34) using two different elution systems, in conjunction with ES-LC/MS data (see below), indicated that the most probable candidate is m 7 G (Fig. 2b).
The TLC mobility of the nucleotide adjacent to the 3Ј end of the anticodon (A-37) is quite similar to that of 2-methylthio-N 6 -threonylcarbamoyladenosine (ms 2 t 6 A) in solvent system A and to that of N 6 -methyl-N 6 -threonylcarbamoyladenosine (m 6 t 6 A) in solvent system B (Fig. 2c). The mobilities of ms 2 t 6 A in solvent system B and that of m 6 t 6 A in solvent A have not been reported to date (23). However, ES-LC/MS analysis (see below) showed that the A-37 derivative is m 6 t 6 A. This tRNA also contains dihydrouridine at position 14 in the D loop and two pseudouridines at position 40 in the anticodon stem and at position 55 in the T loop. The uridine at position 48 is partially modified to an unknown form (data not shown).
Electrospray LC/MS Analysis of Hydrolyzed tRNA GCU Ser -Starfish mt tRNA GCU Ser was hydrolyzed completely to nucleosides (24) and analyzed by electrospray LC/MS. The peak eluting at 3.9 min contains dihydrouridine and pseudouridine (Fig.  3a). Based on sequencing by Kuchino's method, these nucleotides are located in the D loop, the anticodon stem and the T loop. The peak at 10.6 min contains two different methylcytidines, m 3 C and m 5 C, which are assigned to position 32 of the anticodon loop. The peak around 12.7 min is assigned as m 7 G, based on its mass spectrum (MH ϩ m/z 298; BH 2 ϩ m/z 166; Fig.  3b) and elution position. The smaller peak height for m 7 G in the high performance liquid chromatography elution profiles compared with those of other modified nucleosides is a likely consequence of depurination during nuclease digestion (24), resulting in the production of 7-methylguanine. Observation of a component of m/z 166 eluting at 12.7 min is consistent with this interpretation. We assigned the peak at 24.3 min as m 6 t 6 A, based on the electrospray mass spectrum (BH 2 ϩ and MH ϩ m/z 327 and 459, respectively as calculated), and on the retention time (25). In the sequence determination using Kuchino's method (22), an unknown partially modified U derivative was observed at position 48. We were unable to identify the peak eluting just after the guanosine peak, which might correspond to this peak. Uridine derivatives are known to yield poor re- Wobble Rule in Mitochondria 3366 sponse to electrospray ionization (27). The peaks with retention times of 14.7, 16.9, and 19.6 min are assigned from ES-LC/MS data as deoxyguanosine, thymidine and deoxyadenosine, respectively. The shoulder observed in the uridine peak (around 7.8 min) is deoxycytidine. These deoxynucleosides are presumably contaminants from the oligodeoxynucleotide probes used to purify the tRNA.
Alkali-induced Cleavage of mt tRNA Ser at the Location of m 7 G-Since m 7 G (identified from sequencing and ES-LC/MS data) has not previously been reported to occur at the wobble position in the anticodon of any tRNA (28), mt tRNA GCU Ser was subjected to aniline treatment to effect characteristic chain cleavage at the m 7 G site (26). As shown in Fig. 4, scission of mt tRNA GCU Ser at position 34 (the wobble position) by alkaline treatment is clearly observed; yeast phenylalanine tRNA was likewise cleaved at the expected positions of nucleosides wyosine (Y) and m 7 G in a control experiment. Thus, this experiment, together with sequence determination and ES-LC/MS analyses, clearly shows that the wobble nucleotide of starfish mt tRNA GCU Ser is m 7 G, leading to the sequence of starfish mt tRNA GCU Ser shown in Fig. 5. DISCUSSION We report here the first known occurrence of m 7 G at the wobble position of a tRNA. Even since we first discovered that in starfish mitochondria AGN codons seem to be translated as serine by the tRNA GCU Ser from the sole corresponding gene located on the mt genome (6), the question as to how tRNA GCU Ser possessing a GCU anticodon decodes all four AGN codons has attracted our interest, especially with regard to the decoding mechanism of mt tRNAs toward nonuniversal codons. This long-pending puzzle has now been solved by the finding that a modified residue (m 7 G) exists at the wobble position of tRNA GCU Ser , which is probably responsible for decoding AGN codons. This solution seems to be the simplest one among the speculations so far postulated (see Introduction) (1, 2, 17).
We thus consider that the occurrence of m 7 G at the wobble position, which has emerged uniquely in echinoderm mitochon-dria, leads to non-canonical base pairing. This, together with the results concerning codon-anticodon relationships obtained so far for various animal mitochondria (3,4,29), enables us to draw up a set of general rules for codon-anticodon interaction in mitochondria. These are summarized in Fig. 6.
In conventional mitochondrial wobble rules so far elucidated (3,4,29), tRNAs with unmodified guanosine in the first anticodon position are considered to translate only codons terminating in a pyrimidine at the third position, while anticodons with modified uridine at the same position translate codons having a purine at the third position (Fig. 6a).
In Drosophila mitochondria, alignment of mt genes suggests that AGA is a serine codon, whereas AGG is an unassigned codon (2), because AGG has never appeared in Drosophila mt genomes. We have isolated Drosophila melanogaster tRNA GCU Ser , which presumably recognize codons AGC, AGU, and AGA (15), and found that unmodified guanosine is the wobble nucleotide (unpublished observation). Based on the mt genomic sequence, it is unlikely that any tRNA competing with tRNA GCU Ser with regard to the translation of codon AGA exists. Thus, unmodified guanosine is probably able to base pair with adenosine at the codon third position, unless a tRNA that efficiently decodes this codon (a competitor tRNA) does exist simultaneously in the translation system (Fig. 6b).
This hypothesis is supported by the following fact. In starfish mitochondria, AUA is an isoleucine codon, which differs from other known animal mitochondria in which AUA is used as a methionine codon (2). We found that isoleucine and methionine tRNAs of starfish have unmodified GAU and CAU anticodons, respectively (data not shown). Our observation that the wobble nucleoside of the sole mammalian mt methionine tRNA is 5-formylcytidine led us to propose the possibility that this modification allows decoding of AUA in addition to AUG (5). These codon-anticodon relationships for methionine and isoleucine in mammalian and starfish mitochondria support the above interpretation on the decoding properties of unmodified guanosine at the wobble position. Since the unmodified CAU anticodon of methionine tRNA has no capability of decoding AUA (1, 2) due to the lack of modification at the wobble position, the tRNA Met does not compete with tRNA Ile in translating the AUA codon, so that tRNA Ile possessing the GAU anticodon can translate not only AUC and AUU, but also AUA. Thus it can be concluded that guanosine at the wobble position decodes C, U, and A.
In ascidian mitochondria, AGR codons are specific for glycine, while AGY codons are translated as serine (11). The corresponding mt tRNAs were isolated and sequenced; tRNA GCU Ser was found to have unmodified G at the wobble position, whereas tRNA Gly specific for AGR has cmnm 5 U at the same position (30). In this case, tRNA Gly may compete with tRNA GCU Ser , preventing AGR codons from being mistranslated as Ser. In mammalian mitochondria, only AGU and AGC codons are translated by tRNA GCU Ser possessing the anticodon GCU, and there are no tRNAs that translate the AGA and AGG codons (1,2). Although it is apparently in contradiction to the abovementioned codon-anticodon rule, AGG and AGA are utilized as termination codons in mammalian mitochondria (10,31). We propose here that a release factor (1, 2) recognizing these codons plays a role similar to that of the competitor tRNA (Fig.  6c). Thus, only AGY codons remain for assignment to serine.
The present finding that m 7 G is the wobble nucleotide in starfish mt tRNA GCU Ser suggests that methylation of G at N7 expands codon-anticodon interaction to permit base pairing with G, in addition to A, C, and U, at the third position of the codon (Fig. 6d). Although the influence of the altered charge distribution (resulting form N7 methylation) on the base pairing of m 7 G with G is unknown, we are now carrying out experiments using an in vitro translation system to confirm the decoding properties of tRNA possessing m 7 G at the wobble position, and also to verify the ability of m 7 G to base pair with all four nucleotides with respect to the structural factors.