Introduction

Although exceptions have been reported (cf. Hoffmann et al. 1992; Okimoto et al. 1992; Beagley et al. 1998), metazoan mitochondrial (mt) genomes usually comprise the following: 13 protein genes [cytochrome oxidase subunits I-III (cox1–3), apocytochrome b (cob), NADH dehydrogenase subunits 1–6 and 4L (nad1–6 and 4L), and ATP synthetase subunits 6 and 8 (atp6 and 8)], 2 rRNA genes [large subunit ribosomal RNA (rrnL) and small subunit ribosomal RNA (rrnS)], and 22 tRNA (trnA, etc.) genes (cf. Boore 1999).

A good number of vertebrate mt genomes have been completely sequenced (see Boore 2001). Most share the same gene arrangement, and a few (i.e., the birds and marsupials) are known to have derived gene arrangements. The mt genomes of the cephalochordates Branchiostoma lanceolatum and B. floridae have been reported to possess highly similar arrangements to those of vertebrates, although with some differences (Spruyt et al. 1998; Boore et al. 1999). The hemichordate Balanoglossus carnosus mt genome has also been reported to possess a similar gene arrangement to vertebrate mt genomes, although it also bears some similarities to echinoderm mt genomes (Castresana et al. 1998).

Yokobori et al. (1999) determined the complete mtDNA sequence of the ascidian Halocynthia roretzi (Urochordata, Ascidiacea, Pleurogona); they found that the Halocynthia mt genome does not encode atp8 but does encode an additional tRNAGly gene anticodon UCU. The gene arrangement of the Halocynthia mt genome is radically different from any other metazoans sequenced to date, including other known deuterostomes.

The metazoan mitochondrion uses a derived genetic code: with the exception of the diploblasts, most metazoan mt genetic systems analyzed so far use UGA as a Trp codon, AUA as a Met codon, and AGRs as a Ser codon (reviewed by Yokobori et al. 2001). The deuterostome mt genetic code varies among (sub)phyla. AUA specifies Met in chordate mitochondria but specifies Ile in echinoderm and hemichordate mitochondria. AAA specifies Asn in echinoderm mitochondria and Lys in chordate mitochondria but does not appear in hemichordate mitochondria. AGRs specify Ser in echinoderm mitochondria, Gly in urochordate (Halocynthia) mitochondria, and terminate peptide chain synthesis in vertebrate mitochondria. AGA specifies Ser in hemichordate and cephalochordate mitochondria, but AGG does not appear in either the hemichordate or the cephalochordate mitochondria. The most radical change to the genetic code is the reassignment of AGRs to Gly in urochordate mitochondria, because it requires the creation of a tRNAGly that is capable of recognizing AGRs.

In the class Ascidiacea, there are two orders, Pleurogona and Enterogona. Halocynthia and Pyura stolonifera, of which fragments of mt genome have been sequenced (i.e., Durrheim et al. 1993), belong to the family Pyuridae in the Pleurogona order. There are no data for the mtDNA sequence or mt genetic code for species belonging to the Enterogona. To determine when and how the unique mt genetic code found in Halocynthia and Pyura arose, it is necessary to study the mtDNA sequence data for other groups within the Urochordata.

In this paper, we report the complete nucleotide sequence of the other ascidian belonging to the order Enterogona, Ciona savignyi. We found rather different gene arrangements between the Ciona and the Halocynthia mt genomes, including an additional tRNA gene in the former. We discuss gene rearrangement in tunicate mt genomes and the evolution of the metazoan mt genetic code.

Materials and Methods

Preparation of DNA

Adults of Ciona savignyi were caught at Mutsu Bay and obtained through the Asamushi Marine Laboratories of Tohoku University, Aomori, Japan. Total DNA was prepared from Ciona muscle by the conventional phenol-extraction method (Sambrook et al. 1989).

Amplification, Cloning, and Sequencing of Ciona mtDNA

The following primers specific for cox1, cox3, and cob genes, synthesized by the oligonucleotide synthesis service of Amersham Pharmacia Biotech, were used for PCR amplification: COI-5, TTDGTDGATGCDWGDCCDTGRCC; COI-3, GCHGCHG CATAATAHGTATCATG; COIII-5/1, TTDGTDGATGCDW GDCCDTGRCC (for first PCR); COIII-3/1, ACHACATCHAC AAAATGYCAATA (for first PCR); COIII-5/2, TGRTGRCGD GATGTDGTDCGDGA (for second PCR); COIII-3/2, AAA TAHAAAAAAATYCAHACHAC (for second PCR); Cytb-5, WSDTTTTGRRGDGCDACDCTDAT; and Cytb-3, GCAAA HAAAAAATAYCAMACHGG. PCR amplification was carried out with EX-Taq DNA polymerase (Takara Shuzo, Kyoto, Japan). For the amplification of cox3, nested PCR was performed. PCR conditions were as follows: 94°C for 30 s, 40°C for 30 s, and 72°C for 2 min, for 30 cycles. The PCR fragments were cloned using the TOPO-TA cloning system (Invitrogen). Purification of plasmids was performed by the alkali-lysis method, either by hand or with the PI-50 plasmid purification system (Kurabo). Sequencing was performed using the PRISM 377 and PRISM 377-18 DNA sequencers with the BigDye Terminator Sequencing kit (Applied Biosystems).

Long-PCR (Barnes 1994) was performed to amplify the DNA fragments between cox1 and cox3, between cox1 and cob, and between cox3 and cob. From the partial sequences of cox1, cox3, and cob, the following oligonucleotides were synthesized as long-PCR primers: LCOI-5, AGATGTAAAATAAGCTCGAGAA TCTACATC; LCOI-3, TGATACTCAAGAAAATACCTTAATT CTAGT; LCOIII-5, AATGAGAACAAGTAACTGTAATCCC TGAAC; LCOIII-3, TGACTTTTAGTTACTATTGGTTTGG GAGTC; LCytb-5, GATTTTATACTTTTCATTTTCTATTCC CAT; and LCytb-3, CAATGAAATTGGATTTTTGACCT TAC TATG. Long PCR was done using LA-Taq DNA polymerase (Takara Shuzo), according to the manufacturer’s protocol. The long-PCR DNA fragments were cloned using the TOPO-TA cloning kit/TOPO-TA cloning XL kit (Invitrogen). The recombinant plasmids were purified by the alkali-lysis method. Three fragments were successfully amplified and then sequenced; one contained the 3′-end of cox1 and 5′-end of cob at each end, the second fragment contained the 3′-end of cob and 5′-end of cox3 at each end, and the third fragment contained the 3′-end of cox3 and 5′-end of cox1 at each end.

One of the PCR clones covering the sequence from the 3′-end of cox1 and 5′-end of cob was subjected to shotgun sequencing. The GPS-1 kit (New England Biolabs) was used to make the template set. One to three additional clones were sequenced with synthetic oligonucleotides, the sequences of which were based on the sequence of the clone read with the shotgun method. For shorter Long PCR, fragments were sequenced using the primer walking method. In all cases, each clone was sequenced for both strands.

After completing sequence determination of the long-PCR fragments described above, eight PCR fragments covering the whole mt genome of Ciona were amplified, cloned, and sequenced again to confirm the sequence of Ciona mt genome. The nucleotide sequences of the primers used were based on the sequences of the long-PCR fragments.

Data Analyses

The analyses of sequences were performed with GENETYX (Software Development Co.). The identification of each protein gene and each rRNA gene was carried out by comparing the Ciona mtDNA sequence with the Halocynthia mtDNA sequence (Yokobori et al. 1999) as well as those of B. floridae (Boore et al. 1999), Balanoglossus (Castresana et al. 1998), and Homo (Anderson et al. 1981). tRNA genes were searched by reconstructing their cloverleaf structures. The DDBJ accession number of complete nucleotide sequence of the Ciona mtDNA is AB079784.

Phylogenetic Analyses

The following mtDNAs were used for phylogenetic analyses: Homo sapiens (Genbank/EMBL/DDBJ accession number J01415), Mustelus manazo (AB015962), Petromyzon marinus (U11880), Myxine glutinosa (AJ404477), Branchiostoma floridae (AF098298), Halocynthia roretzi (AB024528), Balanoglossus carnosus (AF051097), Paracentrotus lividus (J04815), Asterina pectinifera (D16387), Florometra serratissima (AF049132), Drosophila yakuba (X03240), Daphnia pulex (AF117817), Limulus polyphemus (AF216203), Lumbricus terrestris (U24570), Platynereis dumerilii (AF178678), Terbratulina retusa (AJ245743), Laques rubellus (AB035869), Katharina tunicata (U09810), Loligo bleekeri (AB029616), Metridium senile (AF000023), Allomyces macrogynus (U41288), Podospora anserina (X55026), and Pichia canadensis (D31785). Data from the above are subsequently referred to as belonging to the 24-species set. Data from a reduced set of species (i.e., Metridum, Homo, Branchiostoma, Halocynthia, Ciona, Balanoglossus, Asterina, Limulus, and Katharina) are hereafter referred to as the 9-species set.

With the exception of atp8, all the mt protein genes encoded by metazoan mt genomes were used for the phylogenetic analyses. Amino acid sequences of each protein gene were aligned with CLUSTAL X (Thompson et al. 1997) using the default conditions. Small manual adjustments were subsequently made to the alignments. The regions where alignments were not clear were removed by using GBLOCKS (Castresana 2000). The combined data were used for further analyses. The alignments used for this study can be located at The web site of this journal as supplemental data.

24-Species Set

A total of 2129 sites was included in the analyses. For the maximum likelihood (ML) analysis, MOLPHY 2.3b (Adachi and Hasegawa 1996) was used. The ML distance matrix was constructed using the ML distance estimation option in PROTML. Then the neighbor-joining (NJ) tree was constructed using the NJDIST in MOLPHY. The ML tree was estimated using the local rearrangement search option in PROTML. The NJ tree topology was used as the initial tree topology for the search. The mtREV-f model (Adachi and Hasegawa 1996) was used for these analyses. For the quartet puzzling (QP) analysis, TREE-PUZZLE 5.0 (Schmidt et al. 2002) was used with the following options: slow estimation of parameters, gamma distribution model for among-site rate variation, and use of the mtREV model

9-Species Set

A total of 2269 sites was used for the analyses. Detailed analyses of the 9-species set were carried out during the ML analyses. In these analyses, Halocynthia and Ciona, and Limulus and Katharina, were treated as monophyletic groups, respectively. Metridium was used as an outgroup. Under these conditions, 945 unrooted trees are possible. The CODEML of PAML 3.0c (Yang 1997), under the mtREV24-f model, was used for estimating the log-likelihood (lnL) of each tree and several other values. The 925 possible topologies were used for the ML analyses with various gamma parameter values (0–0.9, with 0.1 intervals). In addition, QP analysis was performed as described above.

Results and Discussion

General Features of Ciona mtDNA

By reviewing the sequences of the PCR fragments, Ciona mtDNA was concluded to comprise 14,737-bp-long circular DNA. This length is as short as that of the Halocynthia mt genome [14,771 bp (Yokobori et al. 1999)] and approx. 2 kp shorter than typical vertebrate mt genomes, such as the human mt genome [l6,569 bp (Anderson et al. 1981)].

The Ciona mt genome contains a slightly different number and type of genes than do other metazoan mt genomes (Fig. 1). The atp8 gene is absent, as is the case in the mt genomes of Halocynthia (Yokobori et al. 1999), several nematode species [cf. Caenorhabditis and Ascaris (Okimoto et al. 1992)], and Mytilus edulis (Hoffmann et al. 1992). The Ciona mt genome encodes trnG(ucu) as is the case in the Halocynthia mt genome (Yokobori et al. 1999). On the other hand, the Ciona mt genome also contains trnM(cau), which has been identified in all metazoan mt genomes studied so far, with the exception of Halocynthia (Kondow et al. 1998; Yokobori et al. 1999). However, as in Halocynthia, the Ciona mt genome also encodes an additional tRNAMet gene with the anticodon UAU (discussed below) (Yokobori et al. 1999). We therefore conclude that the Ciona mt genome encodes 12 protein, 2 rRNA, and 24 tRNA genes (Fig. 1).

Figure 1
figure 1

Structure of the Ciona mt genome compared with the Halocynthia mt genome. The Halocynthia mt gene organization was taken from Yokobori et al. (1999). The circular mt genomes are presented here in linear form. All the genes are encoded on a single strand. tRNA genes are indicated by a one-letter code for their corresponding amino acids: G1, G2, L1, L2, S1, and S2 represent trnG(ucc), trnG(ucu), trnL(uaa), trnL(uag), trnS(uga), and trnS(gcu), respectively.

Full size image

All of the Ciona mt genes are encoded by the same strand, as is also the case in Halocynthia (Yokobori et al. 1999), Mytilus (Hoffmann et al. 1992), brachiopod (Stechmann and Schlegel 1999; Noguchi et al. 2000; Helfenbein et al. 2001), and Caenorhabditis (Okimoto et al. 1992) mt genes. However, the gene arrangement of the Ciona mt genome differs greatly from that of other metazoan mt genomes, including the Halocynthia mt genome (discussed below).

The longest noncoding region is found between trnF and trnC. The length is 115 bp. We were unable to find any significant similarity to the equivalent noncoding region within the Halocynthia mt genome.

In the coding strand of the Ciona mtDNA, T occurs more frequently than A, and G occurs more frequently than C, as is the case in both Halocynthia and Branchiostoma mtDNAs, but not in Asterina, Balanoglossus, and Homo mtDNAs (Table 1). Among the six mtDNAs compared in Table 1, the Ciona mtDNA is extremely AT-rich. The frequency of G in the coding strand of the Ciona mtDNA is 9.1% lower than in the coding strand of the Halocynthia mtDNA. The lack of G in the coding strand of the Ciona mtDNA is compensated by an excess of A, which is found more abundantly than in the Halocynthia mtDNA.

Table 1 Nucleotide composition (%) of selected deuterostome mt genomes
Full size table

Gene Arrangement

Although the mt gene arrangements of the Vertebrata (e.g., Anderson et al. 1981), Cephalochordata [Branchiostoma spp. (Spruyt et al. 1998; Boore, Daehler and Brown 1999)], and Hemichordata [Balanoglossus (Castresana et al. 1998)] are well conserved, the gene organization of the Halocynthia mt genome has been reported to differ from that of other metazoan mt genomes (Yokobori et al. 1999). The gene organization of the Ciona mt genome is also very different from that of other metazoans, including Halocynthia. As shown in Fig. 1, there is a large difference between the Ciona and the Halocynthia mt gene arrangements. Between these two ascidian mt genomes, there are only three orders of identical gene pairs — cox2–cob, trnS(uga)–cox1, and trnR–trnQ. Excluding tRNA genes from the comparison of gene arrangements, the only common orders are nad3–nad4 and cox2–cob. Two partial orders, cox2–cob and trnS(uga)–cox1, are unique to these two ascidian mt genomes; they are found in no other known metazoan mt genomes.

The order trnR–trnQ, common to both Ciona and Halocynthia mt genomes, is also found in nematode (Caenorhabditis, Ascaris, and Onchocerca) mt genomes (Okimoto et al. 1992; Keddie et al. 1998). However, this could be a result of parallel evolution, because the mt gene arrangements of Caenorhabditis, Ascaris, and Onchocerca (Okimoto et al. 1992; Keddie et al. 1998) are, like Ciona and Halocynthia, believed to be highly derived. In addition, because urochordate and nematode mt genomes encode all their genes on a single strand, the number of possible orders of gene pairs is much reduced compared to the situation in which both strands of the DNA duplex are coding strands. This might have caused the observed coincidence of the shared gene pair in the urochordate and nematode mt genomes.

The Halocynthia mt genome shares only one partial gene order trnH–trnS(gcu) with most vertebrate, cephalochordate, hemichordate (Balanoglossus), and echinoderm mt genomes but none with the Ciona mt genome (Fig. 2). The Ciona mt genome does not share any partial gene orders with vertebrate, cephalochordate, hemichordate, or echinoderm mt genomes (Fig. 2). However, it does share the order trnI–nad3 with the Katharina (Mollusca) (Boore and Brown 1994) and Terebratulina (Brachiopoda) (Stechmann and Schlegel 1999) mt genomes, the order cox3–trnK with the Katharina (Mollusca) (Boore and Brown 1994) and Onchocerca (Nematoda) (Keddie et al. 1998) mt genomes, and the order nad1–trnP with the Loligo (Mollusca) (Tomita et al. 2002) mt genome. In addition, the Halocynthia mt genome shares the order nad1–atp6 with the Caenorhabditis and Ascaris (Nematoda) mt genomes (Okimoto et al. 1992) and the order nad3–trnA with most arthropod mt genomes (e.g., Clary and Wolstenholme 1985). These coincidences of gene order between urochordate mt genomes and nondeuterostome mt genomes might also be due to the much diversified mt gene arrangements of the urochordates.

Figure 2
figure 2

Comparison of mitochondrial gene arrangement among Ciona, Halocynthia (Yokobori et al. 1999), Balanoglossus (Castresana et al. 1998), and Branchiostoma (B. floridae) (Boore et al. 1999). For abbreviations of gene names, see the legend to Fig. 1.

Full size image

As shown above, the Ciona and Halocynthia mt gene organizations do not show any particular similarity with the mt gene organizations of other metazoans studied so far. Ciona and Halocynthia differ as much in their mt gene organizations as Ciona/Halocynthia do from other metazoan mt gene organizations. The use of the Ciona/Halocynthia mt gene organization for inferring their phylogenetic positions seems at present to be difficult; further mt gene organization data for other urochordate species are required to allow this. In particular, to analyze the relationship between urochordates and other metazoans, the sequence of a urochordate mt genome in which the gene organization is known to have evolved slowly would be desirable.

The high similarity among the hemichordate, cephalochordate, and vertebrate mt gene organizations suggests that the urochordates also originated from a similar mt gene organization. The model—gene duplication followed by random loss of duplicated genes—can explain how present-day urochordate mt gene organizations were created. However, urochordate mt gene organizations are so different from those of other metazoans that we cannot easily trace how they have evolved in the absence of further data.

Genetic Code

To predict the genetic code used in the Ciona mt genome, the well-conserved cox1 sequence was compared between Ciona and several other metazoans: Halocynthia (Yokobori et al. 1999), B. floridae (Boore et al. 1999), Petromyzon (Lee and Kocher 1995), Balanoglossus (Castresana et al. 1998), Asterina (Asakawa et al. 1995), D. yakuba (Clary and Wolstenholme 1985), Katharina (Boore and Brown 1994), and Metridium (Beagley et al. 1998). AGA appears 34 times in Ciona cox1. At the corresponding positions of cox1 in the eight comparison species, Gly codons appear 24–26 times, but Ser codons appear only 1–3 times for each species. AGG appears five times in Ciona cox1. All the corresponding positions are conserved as Gly in the eight species compared. Therefore, we conclude that AGRs specify Gly in Ciona mitochondria. Twenty-six AUA codons are found in Ciona cox1. At the corresponding positions of cox1 of the eight comparison species, Met codons appear 14–17 times, but Ile codons appear only 1–4 times. Thus, we conclude that AUA specifies Met in Ciona mitochondria. At the corresponding positions in the eight species for the 13 positions of UGA in Ciona cox1, Trp codons appear 9–13 times. Therefore, we conclude that UGA specifies Trp in Ciona mitochondria. AAA appears 11 times in Ciona cox1, and in the corresponding positions of the eight comparison species, Lys codons appear in only 2–6 positions. However, Asn codons appear only once in Halocynthia cox1 and twice in Asterina cox1 at the corresponding positions where AAA appears in Ciona cox1, and no Asn codons are found in these positions in Petromyzon, Branchiostoma, Balanoglossus, Drosophila, Katharina, and Metridium cox1. Therefore, we conclude that AAA in Ciona mitochondria specifies Lys rather than Asn.

In summary, the Ciona mt genome uses the same codon table as the Halocynthia mt genome: AUA specifies Met, UGA specifies Trp, and AGA/AGG specify Gly.

Codon Usage

In Ciona mt protein genes, codons terminating with C are less frequently used in all four codon families and in all NNY families (Table 2). Indeed, two codons—TGC for Cys and CGC for Arg—are not found at all (Table 2). However, these might not be unassigned codons, because the sequences of trnC (anticodon GCA) and trnR (anticodon UCG) suggest that they can be recognized. They could instead simply reflect the nucleotide composition of the Ciona mtDNA, which is rich in A and T rather than G and C (see above).

Table 2 Comparison of codon usage between Ciona and Halocynthia
Full size table

When codon usage is compared between the Ciona and Halocynthia mt protein genes, some differences are observed. First, codons ending with G are more abundantly used in Halocynthia than in Ciona. Second, A is more frequently found at the first codon position than G in Ciona compared to Halocynthia mt protein genes. For example, 65% of Gly codons in Ciona mt protein genes are encoded by the AGA codon, whereas GGN codons (62%) are more abundant in Halocynthia mt protein genes than AGR codons. In addition, Ile codons (ATY) and Met codons (ATR) are used more frequently in Ciona than in Halocynthia mt protein genes, whereas Val codons (GTN) are used more frequently in Halocynthia than in Ciona mt protein genes.

As shown in Table 3, amino acid composition differs among the Ciona, Halocynthia (Yokobori et al. 1999), Balanoglossus (Castresana et al. 1998), and Branchiostoma (Spruyt et al. 1998) mt protein genes (In the cases of Balanoglossus and Branchiostoma, nad6 is excluded from the analysis, because it is encoded by the opposite strand to that of the other mt protein genes). The choice of amino acid might be affected by the nucleotide composition and/or directional mutation pressure of the protein gene’s encoding strand. For example, in Halocynthia, which is rich in G and T rather than C and A, respectively, Val encoded by GTN codons is used twice as often as in Balanoglossus, which is rich in C and A, rather than G and T, respectively. Because the Ciona mt genome is richer in A and less rich in G than the Halocynthia mt genome, only three quarters of the amount of Val in Halocynthia mt protein genes is used in Ciona mt protein genes, and two thirds of the amounts of Ile and Met, which are encoded by ATY and ATR codons, respectively, in Ciona mt protein genes are used in Halocynthia mt protein genes. However, the total number of neutral amino acids is very similar among these four species, suggesting little change in protein properties.

Table 3 Comparison of amino acid (aa) usage among Ciona, Halocynthia, Balanoglossus, and Branchiostoma mt-protein genes
Full size table

tRNA Genes

Within the Ciona mt genome, we found 24 cloverleaf structures bearing a resemblance to tRNAs, as mentioned above (Fig. 3A). Ciona mtDNA putatively contains two additional tRNA genes, trnM(uau) and trnG(ucu), to the standard set of metazoan mt tRNA genes.

Figure 3
figure 3

A Cloverleaf structures of Ciona mitochondrial tRNA genes. B Comparison of Ciona mt tRNAGly and tRNAMet with Halocynthia (Yokobori et al. 1999). In the Ciona mt tRNA MetCAU different nucleotides from Ciona mt tRNA MetUAU are shown with lowercase letters. In the Halocynthia mt tRNAMet (anticodon UAU), different nucleotides from Ciona mt tRNA MetUAU are shown with outlined letters. In Ciona and Halocynthia mt tRNA MetUCU , different nucleotides from Ciona and Halocynthia mt tRNA MetUCU , respectively, are shown with lowercase letters. In Halocynthia mt tRNA GlyUCC and tRNA GlyUCU , different nucleotides from Ciona mt tRNA MetUCC and tRNA MetUCC , respectively, are shown with outlined letters. Overlapping nucleotides in Ciona mt tRNA GlyUCU with the downstream tRNA gene at the gene level are shown with italic letters.

Full size image

trnC

The trnC has a truncated D arm, as do most metazoan mt trnS(gcu). Other than mt tRNA SerGCU , few metazoan mt tRNA species are known to have truncated D arms—mt tRNA SerUGA from several species [including nematodes (e.g., Caenorhabditis elegans and Ascaris suum) (Okimoto and Wolstenholme 1990; Okimoto et al. 1992) and a cephalopod, Loligo bleekeri (Tomita et al. 2002)], and the tRNACys gene for a reptile mitochondrion (Seutin et al. 1994), for example. In general, tRNA secondary structures should be important for processing the 5′- and 3′-ends (Rossmanith et al. 1995; Rossmanith 1997). Exactly how such D-arm-truncated tRNA is processed is an important issue for the understanding of the mt transcription process.

trnM(cau) and trnM(uau)

As presented in Fig. 3B, the Ciona mt tRNA gene with anticodon TAT shows a high similarity to Halocynthia mt trnM(uau), which has been shown to be expressed (Kondow et al. 1998) and which is the only candidate tRNAMet gene for the Halocynthia mt genome (Yokobori et al. 1999). However, the Ciona mt genome encodes the trnM(cau) in addition to trnM(uau) (Fig. 3). The similarity between the two Ciona mt tRNAMet genes is much less than that between the Ciona and the Halocynthia mt tRNA MetUAU genes (Fig. 3B). The greatest similarities between Ciona and Halocynthia mt trnM(uau)’s are found in the D and anticodon arms, rather than the T arm and acceptor stem. Conversely, the acceptor stem of the two Ciona mt tRNAMet genes is more similar than the other regions.

In eubacterial and eukaryotic systems, some characteristics of initiator tRNA are known (RajBhandary and Chow 1995). One is that the initiator tRNA of both eubacterial and eukaryotic has a run of three G–C pairs at the bottom of the acceptor stem. Metazoan mt tRNAMet, such as those of Homo and Asterina, also have a similar run of three G–C pairs (see Sprinzl et al. 1998). Ciona mt trnM(cau) has similar runs of G’s and C’s, although the bottom base pair is not G–C but an A–C mismatch. The G–C pair run is not found in either Ciona or Halocynthia mt trnM(uau)’s, although they have a G–C pair at the bottom of the acceptor stem.

trnG(ucc) and trnG(ucu)

We previously reported that the product of the trnG(ucu) of the Halocynthia mt genome is expressed (Kondow et al. 1999). Therefore, the product of the tRNA gene with anticodon TCT on the Ciona genome is concluded to be the tRNAGly for AGR codons.

Adenine at the discriminator position is well conserved both in cytoplasmic tRNAsGly and in the mt tRNAsGly of eukaryotes (Nameki et al. 1997; Sprinzl et al. 1998). This discriminator base A is one of the major identity elements for recognition by Saccharomyces glycyl tRNA synthetase (GlyRS) (Nameki et al. 1997). However, Ciona mt trnG(ucu) has a T in the same position. This tRNA gene overlaps for two nucleotides (5′-AT-3′) with the downstream trnY in the same direction (Fig. 3). As has been found in other metazoan mitochondria (Yokobori and Pääbo 1995a, b, 1997; Reichert et al. 1998; Reichert and Mörl 2000), alternation of the discriminator from U to A might occur as a result of tRNA editing.

Evolutionary Considerations on mt tRNA and the mt Genetic Code in the Metazoa

The mt genomes of two urochordates, Halocynthia and Ciona, share unique features encoding trnM(uau) and trnG(uau) and use the same codon table—AUA specifying Met, UGA specifying Trp, and AGR specifying Gly—suggesting that it would have been established in the common ancestor of the Ascidaceans. As we have discussed previously (Yokobori et al. 1993, 1999, 2001; Kondow et al. 1999), the creation of tRNA GlyU*CU is essential for the establishment of the ascidian mt genetic code table.

Yokobori et al. (2001) proposed that the common ancestor of chordate mitochondria lost a methyl group at the anticodon first position of tRNA SerGCU (7-methylguanosine to G). The modification should be key for decoding the AGG codon by tRNA SerGCU . In vertebrate mitochondria, the appearance of a release factor might have caused the AGR codon to have become a termination codon. On the other hand, the appearance of the tRNA GlyU*CU gene might have allowed ascidian mitochondria to use AGR as the glycine codon. During the period when the assignment of AGRs changed from Ser to Gly, competition for AGA (and AGG) between tRNA SerGCU and tRNA GlyU*CU might have occurred.

Cytoplasmic and mt GlyRS originate from a single gene in Bombyx mori (Nada et al. 1993) and humans (Mudge et al. 1998). In Drosophila melanogaster and Caenorhabditis elegans, only a single putative GlyRS gene is found throughout the entire genome (The C. elegans Sequencing Consortium 1998; The FlyBase Consortium 2002). Therefore, products of a single GlyRS gene can be used in both the cytoplasm and mitochondria of Ciona. An important question is how tRNA GlyUCU can be recognized by GlyRS, because a tRNA with a UCU anticodon might be tRNAArg in the cytoplasm and therefore must not be recognized by GlyRS.

As discussed in the preceding section, the Ciona mt genome encodes trnM(cau), which has been lost in the Halocynthia mt genome. The Ciona trnM(cau) shows more similar characteristics to other metazoan mt trnM(cau)s, rather than to the trnM(uau) of Ciona and Halocynthia. Ciona and Halocynthia mt trnM(uau) might have originated from trnM(cau), because in almost all metazoan mt genomes, including the cnidarians, only a single tRNAMet gene is encoded (Boore 2001).

In nonmetazoan genetic systems, there are two types of tRNAMet. One is the initiator tRNAMet (or initiator tRNAfMet); the other is the elongator tRNAMet. The methionine is formylated (fMet) after charging of the initiator tRNA in eubacterial and mitochondrial systems (RajBhandary and Chow 1995). Mammalian mt methionyl tRNA transformylase has recently been characterized (Takeuchi et al. 2001). In most metazoan mt genomes, only one tRNAMet gene is encoded, and there is no experimental evidence in metazoans for the import of tRNAMet from the cytoplasm to the mitochondria, although tRNALys has recently been shown to be imported from the cytoplasm to the mitochondria in marsupials (Dörner et al. 2001). Therefore, it has been thought that a single tRNAMet molecule has two roles, one for initiation and one for elongation.

The existence of two methionine tRNA genes with anticodons CAT and TAT have been reported for the bivalve mollusk Mytilus edulis mtDNA (Hoffmann et al. 1992). Interestingly, in Mytilus mt tRNAMet genes, too, the trnM(cau) carries a G–C run at the bottom of the anticodon stem, but the trnM(uau) does not.

The C at the anticodon first position is a strong identifying element for tRNAMet. E. coli minor tRNAIle has the anticodon LAU, where L (lysidine) is a modified nucleotide of C. Muramatsu et al. (1988) demonstrated that the E. coli minor tRNAIle can be charged with methionine, if it has an unmodified C at the anticodon first position. On the other hand, in the eukaryotic cytoplasmic system, tRNA IleUAU exists (Sprinzl et al. 1998) and must not be methionylated. Thus, how the tRNA MetUAU in urochordate mitochondria is recognized as tRNAMet is an important issue for the evolution of the recognition mechanism of MetRS.

The appearance of trnM(uau) and trnG(ucu) might be related to the scrambling of the mt gene arrangement in the ancestor of Halocynthia and Ciona, as predicted from the present-day mt gene arrangements of Halocynthia and Ciona. As mentioned earlier, Halocynthia and Ciona mt gene arrangements are quite different from those of other metazoans. This means that many rearrangement events have taken place in uorchordate mt genomes. As stated before, the partial duplication of the mt genome followed by the random loss of one of the duplicates is the most widely accepted model of the evolution of mt genome rearrangement. In this process, there is a period when the duplicated genes must coexist on the genome. During this period, one of the duplicates might be free from functional constraints and through the action of mutation might therefore adopt a different function. This possibility seems to be likely not only for urochordate mt genomes but also for other metazoan mt genomes. The Mytilus mt genome carries an extra trnM [trnM(uau)] and has a highly rearranged gene arrangement (Hoffmann et al. 1992). This situation might also be the same in urochordate mt genomes. In addition, pseudogenes of tRNA genes that were created by gene rearrangement have been reported from several metazoan mt genomes, such as the gekko Heteronotia binoei (Zevering et al. 1991) and squid Loligo bleekeri (Sasuga et al. 1999). Information regarding the presence or absence of mt trnM(uau) and mt trnG(ucu) could be a useful marker for the reconstruction of the urochordate phylogeny, if the mt genomes of other urochordate groups are also analyzed.

Phylogenetic Analyses

The ML tree for 24 species (see Materials and Methods) with mt protein gene sequences, reconstructed by PROTML (Adachi and Hasegawa 1996), is shown in Fig. 4A. In this tree, the urochordate branch is basal to all other triploblast animal groups. However, as Yokobori et al. (1999) pointed out, the Halocynthia mt genes evolved much more rapidly than other deuterostome mt genes. As shown in Fig. 4A, the same is also likely for the Ciona mt genes. Figure 4B shows the QP tree (Schmidt et al. 2002) that considers among-site variation in substitution rates. In this tree, a deuterostome group is recognized. However, although the grouping of hemichordates and echinoderms in this tree matches recent molecular phylogenetic studies (e.g., Castresana et al. 1998), the relationship among subphyla of the chordates and the hemichordate + echinoderm group is not yet resolved.

Figure 4
figure 4

Phylogenetic trees of 21 metazoans and 3 fungi based on the amino acid sequences of 12 combined mitochondrial protein genes. A The ML tree constructed by PROTML (local rearrangement search) in MOLPHY (Adachi and Hasegawa 1996). The lnL (log-likelihood) with standard deviation for the tree is 58,383.82 886.11. The local bootstrap probability of each node (50%) is shown. Species included in this tree are listed in Phylogenetic Analysis under Materials and Methods. B The QP tree constructed by TREE-PUZZLE (Schmidt et al. 2002). The estimated of the distribution of rate variation among sites with standard deviation is 0.62 0.02, and the lnL for the tree is 54,762.72. The bootstrap probability of each node (50%) is shown.

Full size image

By reducing the number of species to nine (Ciona, Halocynthia, Homo, Branchiostoma, Balanoglossus, Asterina, Limulus, Katharina, and Metridium), more detailed ML analyses were possible using PAML (Yang 1997). In this analysis, Ciona and Halocynthia are constrained to be monophyletic, as are Limulus and Katharina.

We varied the parameter α of the Γ distribution for among-site heterogeneity of substitution rates (0.1–0.9). When the log-likelihoods (lnL) of the ML trees are compared among different α, the lnL of the ML tree when α is 0.6 is best (data not shown). In Table 4, the trees with the top 10 lnL including the ML tree are shown. In the ML tree, the urochordates appear as the sister group of a group—protostomes + deuterostomes except urochordates—as in the case of the 24 species tree reconstructed by protml (Table 4). Among the 945 possible trees, 43 trees could not be rejected because differences between their lnL and that of the ML tree were less than two standard deviations. The third tree for lnL (indicated by boldface in Table 4) has a topology in which the urochordates are the sister taxa for the Branchiostoma and Homo clade.

Table 4 Top 10 maximum likelihood (ML) trees and the quartet-puzzling (QP) tree of nine species
Full size table

The QP tree was also reconstructed for the topology of the same data set. The topology of the QP tree (α = 0.57; Table 4) is identical to that of the second tree in the ML analysis (Table 4).

As shown in Fig. 4 and Table 4, the lengths of the branches for Ciona and Halocynthia are much longer than those for the other species. Ciona and Halocynthia mt protein genes might be under relaxed selection pressure compared with their counterparts used for the reconstruction of their phylogeny. Metazoan mt genomes with a much different gene organization from that of other metazoans tend also to show greatly differing primary sequences in their protein genes (accelerated substitution rate at the protein level) (e.g., Yamazaki et al. 1997; Yokobori et al. 1999). Analyses of the relationships between various aspects of the stability of mt genomes (e.g., substitution rate and gene rearrangement rate) might be needed to understand how the mt genetic system (replication, transcription, and translation) has evolved.