Introduction

The phylum Chordata is divided into the three subphyla Urochordates, Cephalochordates, and Vertebrates, with Urochordata, the most basal lineage, made up of the classes Thaliacea, Appendicularia, and Ascidiacea. The subphylum Urochordata groups marine organisms with very different lifestyles and developmental programs. It includes sessil or planktonic, solitary or colonial, sexual or asexual reproducing animals, sharing the presence of a notochord either throughout the entire life cycle or only during the larval stage. Moreover, such organisms are unsegmented and headless. An understanding of the evolution and phylogenetic relationships between these ancient organisms—whose fossil records date back to the Lower Cambrian (550 million years ago) (Shu et al. 2001; Zhang 1987)—is crucial to provide insight into the early stages of chordata evolution, and into the phylogeny of deuterostomes (Cameron et al. 2000; Stach and Turbeville 2002; Swalla et al. 2000; Turbeville et al. 1994). Moreover, the ascidian Ciona intestinalis is widely used as model system for chordate developmental biology and for comparative genomic studies. The tadpole larva of ascidians represents one of the most simple and primitive chordate body plans. Furthermore, C. intestinalis embryos develop rapidly and can be easy manipulated (Satoh 1994). The small nuclear genome of C. intestinalis, about 160 Mbp, is used as a model to investigate the nature of the primitive nonduplicated chordate nuclear genome (Holland et al. 1994; Ohno 1970). The sequencing and annotation of the coding portion of this genome have recently been completed and confirm that the hypothesized large-scale gene duplication event occurred in the vertebrate lineage after its divergence from cephalochordates and urochordates (Dehal et al. 2002). An ongoing cDNA project will complement the genome sequencing project, allowing an easier functional interpretation of the genomic sequences and providing valuable information on chordata gene regulation (Satou et al. 2002).

The animal mitochondrial genome (mtDNA) is a good model system to study evolutionary mechanisms at the genome level. This molecule is quite small (15–20 kb), easy to sequence, extremely compact, and shows a relatively constant gene content (Saccone et al. 1999). Complete mtDNA sequences of a large number of metazoans are available for comparative analysis. However, these are mostly from vertebrate taxa, and until the submission of our data, the complete mtDNA sequence was available for only one urochordate, the ascidian Halocynthia roretzi (order Pleurogona, suborder Stolidobranchiata, family Pyuridae) (Yokobori et al. 1999). The mtDNA sequence of Ciona savignyi (Yokobori, Watanabe, and Oshima, unpublished) was made available shortly after the submission of our sequence. There have long been indications that ascidians belonging to the same suborder as Halocynthia have a different mitochondrial gene organization (Durrheim et al. 1993; Jacobs et al. 1988) and that urochordate mtDNA evolves faster than that of other metazoans (Kakuda 2001; Stach and Turbeville 2002; Yokobori et al. 1999). Even at the nuclear level, a urochordate phylogeny reconstructed using a conserved portion of the 18S rRNA gene shows that the molecular divergence within urochordates is larger than in vertebrates and hemichordates, suggesting that Urochordata should be raised to the phylum status (Cameron et al. 2000; Swalla 2001; Swalla et al. 2000). In the light of these considerations, we decided to determine the complete mtDNA sequence of the model organism C. intestinalis (order Enterogona, suborder Phlebobranchiata, family Cionidae), an ascidian distantly related to Halocynthia, starting from the analysis of the large number of expressed sequence tags (ESTs) available for this organism in the public database (Gissi and Pesole 2003). The availability of the mtDNA sequence of Ciona savignyi (Yokobori, Watanabe, and Oshima, unpublished) has allowed a more detailed analysis of the evolution of ascidian mtDNA. The genus Ciona includes only solitary species, with C. intestinalis a native of northwestern Europe now showing a global distribution and C. savignyi a native of Japan which has recently been found in southern California and the Pacific coast of America (Hoshino and Nishikawa 1985; Lambert 2001). The classification and number of species that constitute the genus Ciona remain quite controversial (Copello et al. 1981; Hoshino and Nishikawa 1985; Lambert et al. 1990).

In this paper, we describe the structural and compositional features of the C. intestinalis mtDNA, stressing the evolutionary trends of ascidian mtDNA compared to those of other metazoans, especially Chordata.

Materials and Methods

The Ciona intestinalis sample was caught in the Tirrenian Sea, Naples Gulf. Total DNA was extracted from the ovary of a single individual using the Puregene Tissue kit (Gentra Systems). The mtDNA was amplified by long-PCR (TaKaRa LA-Taq) in five overlapping fragments using specific primers designed according to the transcribed mitochondrial sequences of Ciona intestinalis inferred from mitochondrial-like ESTs (Gissi and Pesole 2003).

  1. 1

    •Fragment sco (2.7 kb): colF1 5′-ATAGTTTGAGCAATGA GGGG 3′, smR 5′-TAACCGYGRTARCTGGCAC 3′

  2. 2

    •Fragment n2-5 (3.6 kb): nd5F1 5′-GAAGAGTATATTAGGT GAAG-3′, nd2F 5′CTTGGGATATCATTCGAGAG-3′

  3. 3

    •Fragment pg1 (2.6 kb): ap6R 5′-TTTCAACAGTTTTAC GAGG-3′, glR1 5′-AAATGAACTACTGA AAGGGC-3′

  4. 4

    •Fragment cl-2 (2.7 kb); lg1F 5′-TTCAATTAGATTACCTT CAT-3′, lg2R 5′-CAATACCAAATCGCACACTC-3′

  5. 5

    •Fragment cog (6.5 kb): g2F1 5′-ATGATAGTACGTTTTA CCCC-3′, co1R 5′-CTATATGCTCATAAAAGTGG-3′

The amplified fragments were directly cloned in the pCR 2.1-TOPO plasmid using the TOPO-TA Cloning kit (Invitrogen) following the manufacturer’s protocol. Recombinant plasmids were extracted from positive clones using the QIAprep Spin Miniprep kit (Qiagen) and the inserts were completely sequenced by primer walking. Sequencing was performed on an ABI Prism 3100 sequencer (Perkin Elmer). The overlaps between amplified fragments range from 170 bp (overlap between fragment cog and sco) to 1.2 kb (overlap between fragment n2-5 and pg1). In case of sequence uncertainties, the corresponding region of the mtDNA was reamplified from the total DNA and directly sequenced without cloning.

Transfer RNA genes were identified using tRNAscan-SE (version 1.21) with parameters specific for organellar tRNA structures (Lowe and Eddy 1997), by searching for canonical anticodon arm sequences (Kumazawa and Nishida 1993) using the PatSearch program (PesoLe et al. 2000), and by sequence potential to be folded into cloverleaf structures. Small and large subunit ribosomal rRNA genes (rrnS and rrnL) were identified by sequence similarity to homologous animal genes. Ends of rRNA genes were not determined experimentally and were assumed to be adjacent to the ends of bordering tRNA genes. As an exception, the 5′ end of rrnL was tentatively defined by sequence similarity to homologous genes of the ascidians Halocynthia roretzi (AB024528) and Ciona savignyi (AB079784), as described in Results and Discussion. Open reading frames of protein genes were defined using the ascidian mitochondrial translation code (Durrheim et al. 1993; Yokobori et al. 1993) and identified by similarity to the homologous mitochondrial-encoded proteins of the following 16 species representative of deuterostome organisms: Homo sapiens (AF347015), Oryctolagus cuniculus (AJ001588), Gallus gallus (X52392), Alligator mississippiensis (Y13113), Chrysemys picta (AF069423), Dinodon semicarinatus (AB008539), Rana nigromaculata (AB043889), Squalus acanthias (Y18134), Gadus morhua (X99772), Latimeria chalumnae (U82228), Petromyzon marinus (U11880), Myxine glutinosa (AJ404477), Branchiostoma lanceolatum (Y16474), Halocynthia roretzi (AB024528), Ciona savignyi (AB079784), Balanoglossus carnosus (AF051097), and Paracentrotus lividus (J04815). The possibility of unusual initiation and termination codons was considered (Wolstenholme 1992). The gene encoding ATPase subunit 8 (atp8) was identified by comparison against homologous sequences of a large number of species, including representatives of protostomes, cnidarians, porifers, and the yeast Saccharomyces cerevisiae (alignment available on request). Moreover, the presence and location of a transmembrane hydrophobic helix in ATP8 protein were investigated using the DAS software (Cserzo et al. 1997). The definition of gene boundaries was guided by the mapping of mitochondrial-like Ciona ESTs, as described in Gissi and Pesole (2003).

Specific nucleotide patterns, such as the rRNA transcription termination box, were searched using the PatSearch program (PesoLe et al. 2000). Repeated sequences were identified using RepFinder (Benson 1999) and CleanUP (Grillo et al. 1996) programs.

Base compositional analyses were carried out using the Codontree algorithm (Pesole et al. 1996). The GC- and AT-skews, which indicate compositional differences between the two strands, were calculated according to the formulae of Perna and Kocher (1995).

Mitochondrial genome rearrangements within deuterostomes and in all one-strand transcribed mtDNAs were investigated using the minimal breakpoint method, which reconstructs the most parsimonious phylogeny from breakpoint distances (Blanchette et al. 1999).

Blast search analyses were carried out at the NCBI Web site (http://www.ncbi.nlm.nih.gov/BLAST/ ).

The complete mtDNA sequence of Ciona intestinalis is available at the Embl database under accession number AJ517314.

Results and Discussion

Amplification Strategy

The mtDNA of C. intestinalis has been completely amplified from total DNA in five overlapping fragments by taking advantage of the numerous ESTs of this organism available in the Embl database to design PCR primers. As described in Gissi and Pesole (2003), mitochondrial-like ESTs of C. intestinalis have been assembled based on the presence of overlapping regions and on clone information. Eight distinct nonoverlapping mitochondrial transcripts have been inferred, containing typical mitochondrial genes and covering on the whole about 12 kb. Their sequences have been used to design several specific primers that have been used in different combinations to amplify the complete mtDNA. The locations of inferred mitochondrial transcripts and amplified fragments on the mtDNA are reported in Fig. 1. All inferred mitochondrial transcripts have been confirmed at the level of both gene order and sequence (>99% identity).

Figure 1
figure 1

Amplification strategy and genome structure of C. intestinalis mtDNA, compared to those of Ciona savignyi (AB079784) and Halocynthia roretzi (AB024528). Putative C. intestinalis transcripts inferred from EST data (Gissi and Pesole 2003) are indicated by black lines and amplified segments as dark gray boxes. Noncoding regions longer than 9 bp are indicated in black with numbers corresponding to their size (bp). Genes blocks with different positions in C. intestinalis and C. savignyi are indicated as lightgray boxes; gene blocks conserved in C. intestinalis and H. roretzi are indicated by hatched boxes, with reverse hatch for inverted gene order. Feature table of H. roretzi has been modified as reported in Gissi and Pesole (2003). Abbreviations for proteins and rRNA genes are as in the text, except for 4L (nad4L) and 8 (atp8). tRNAgenes are indicated according to the transported amino acid and, in addition: FN, additional tRNA-Phe; G1, Gly(AGR); G2, Gly(GGN); L1, Leu(UUR); L2, Leu(CUN); M1, Met(AUG); M2, Met(AUA); S1, Ser(AGY); S2, Ser(UCN).

Full size image

General mtDNA Features

The Ciona intestinalis mtDNA is 14,790 bp long, very similar in size to other ascidian mtDNAs. Among animals, the smallest sequenced mtDNA belongs to Taenia crassiceps (Nematoda, 13.5 kb) and the largest to Venerupis philippinarum (Mollusca, 22.6 kb) (Embl database, May 2003). The ascidians genomes are the smallest among the completely sequenced mtDNAs of Chordata, with a size closer to that of Echinodermata and Hemichordata (range from 15.6 to 16.2 kb) than to other Chordata, which range from the 15 kb of Cephalochordata to the 18.9 kb of the Agnatha Myxine glutinosa. The absence of a long noncoding region homologous to the mammalian D-loop-containing region, the presence of small intergenic regions, and the relative shortness of some genes (i.e., rRNAs) make the Ciona mtDNA more similar in size to those of invertebrates. Moreover, the genome is highly compact; indeed the coding portion forms 96.9% of the C. intestinalis mtDNA.

We were surprised to observe that the C. intestinalis mtDNA encodes for 39 genes: the standard set of 37 metazoan mitochondrial genes and 2 additional tRNA genes. The 13 genes coding for proteins of the respiratory complexes have all been identified, including a gene coding for a short form of the F0 ATPase subunit 8 not previously annotated in the closely related C. savignyi and H. roretzi mtDNAs (Yokobori et al. 1999). Indeed, we found this gene in both H. roretzi and C. savignyi mtDNAs (see below). In addition to the two genes coding for ribosomal RNAs (rrnS and rrnL), there are 24 tRNA genes: 2 tRNA genes for decoding Leu, Ser, Gly, and Met codons and 1 tRNA for decoding the remaining amino acids. The existence of two tRNA genes for Leu, Ser, and Gly codons is in accordance with the alteration of the mitochondrial genetic code in ascidians compared to the universal one (Durrheim et al. 1993; Yokobori et al. 1999). Furthermore, two distinct tRNA-Met genes can be identified in C. intestinalis, the first possesses the common 5′-CAU-3′ anticodon, trnM(AUG), while the second exhibits the unusual 5′-UAU-3′ anticodon, trnM(AUA). The same situation has been reported for C. savignyi (AB079784). In H. roretzi only the unusual trnM(AUA) has been reported (Yokobori et al. 1999), but an unannotated trnM(AUG) overlapping the end of the nad4l gene can easily be identified (see below).

Genome Organization

All mitochondrial genes of C. intestinalis are encoded on the same strand, as in H. roretzi and C. savignyi. In the deuterostome mtDNAs completely sequenced, most of the genes are transcribed from one strand, whereas a small number of tRNA genes and at least nad6 are encoded by the complementary strand (Boore 1999; Saccone et al. 1999). The transcription of all mitochondrial genes from the same strand has been found only in ascidians and lower animals, that is, in Cnidaria Zoantharia and some protostomes (Nematoda, except for Trichinella spiralis, Platyhelminthes, Brachiopoda, Annelida, and a few representatives of Crustacea and Mollusca) (see Appendix, Table A1). Gene order similarities between these one-strand transcribed mtDNAs have been calculated as breakpoint distances with and without tRNA genes (data not shown), showing that the genome organization is quite conserved only in phylogenetically related organisms with no trace of evolutionary convergence. Thus, no peculiar constraints seem to act on the one-strand transcribed mtDNAs forcing them to adopt a specific gene order.

The gene arrangement of C. intestinalis is markedly different from that reported for H. roretzi, with which it shares only three conserved gene blocks (Fig. 1): cox2–cob, trnR–trnQ, and trnS(UCN)–cox1. The nad1 and rrnL genes are adjacent in both species, however, the gene order is nad1–rrnL in C. intestinalis and rrnL–nad1 in H. roretzi. Excluding tRNA genes, a fourth gene block appears to be conserved between the two mtDNAs, that is, the nad3–nad4 gene pair, with the atp8 located upstream in H. roretzi and downstream in C. intestinalis. Furthermore, gene order is not conserved even within the genus Ciona (Fig. 1). Comparison of C. intestinalis and C. savignyi mtDNA organization reveals the transposition of nad1 and five tRNA genes—trnD, trnM(AUA), trnM(AUG), trnS(AGY), and trnT—with trnD and trnM(AUA) adjacent but inverted in the two species (Fig. 1). There are few cases of mtDNA completely sequenced in different species of the same genus and the only other reported case of intragenus variability in mt gene order has been found in the Platyhelminthes genus Schistosoma, which shows a high rate of gene rearrangement (Le et al. 2000). Hence, at least for invertebrates, gene rearrangements within genus may be more common than expected.

In addition to intraclass genome rearrangements, the ascidian gene order is extremely divergent from that of other Chordata. Indeed, Vertebrata, Cephalochordata, and even the hemichordate Balanoglossus carnosus share an almost-identical basal mt genome organization, with only a few transpositions tolerated, mostly involving tRNA genes and noncoding control region (Saccone et al. 1999). The only gene pair conserved among all Chordata is rrnL–nad1 (not taking into account the tRNA genes) but is missing in Cionidae. Indeed, this gene pair is inverted in C. intestinalis and absent in C. savignyi (Fig. 1). Figure 2 reports the minimal breakpoint tree reconstructed from deuterostome mtDNA gene order data without considering tRNA genes, owing to their high mobility (Blanchette et al. 1999). The breakpoint distances within the ascidian clade are very high compared to those within the Echinodermata and Vertebrata–Cephalochordata–Hemichordata clades. The total breakpoint distance from the common deuterostome ancestor is 29 for ascidians and 5 for the clade including Vertebrata–Cephalochordata–Hemichordata. This implies a fivefold higher gene rearrangement rate in the former with respect to the latter organisms. Such observations confirm that the ascidian mtDNA has a very high gene rearrangement rate, previously supposed to be a feature of only protostome mt genomes (Boore 1999; Saccone et al. 1999). A more surprising feature of the ascidian mtDNAs is that even genes usually adjacent in metazoan mtDNAs, such as rrnS–rrnL and atp8–atp6, are rearranged and distantly located. The rrnS and rrnL are very close and separated by few tRNA genes in most animal mtDNAs, whereas they are 7 and 5 kb distant in Ciona and Halocynthia, respectively. Ascidiacea share this gene organization only with Echinoidea (Echinodermata) among deuterostomes, most Cnidaria, and some protostomes (see Appendix, Table A2). This peculiar gene arrangement is interesting because it prevents a coordinated expression of both rRNAs by the synthesis of a single transcript. Indeed, in mammals the rRNA genes are transcribed at a higher level compared to other mt genes through the usage of a transcription termination signal located downstream from the two adjacent rRNA genes (Christianson and Clayton 1986). Similarly, atp8 immediately precedes and sometimes overlaps atp6 in most metazoans and even in S. cerevisiae (Foury et al. 1998). In mammals the two proteins are translated from the same mature bicistronic transcript, allowing a translational regulation (Fearnley and Walker 1986). In Ascidiacea these genes are 3 or 8.4 kb distant, a situation found in only a few other species (see Appendix, Table A3). It is possible that the different gene order could have important implications for the regulation of gene expression.

Table 1 List of completely sequenced metazoan mtDNAs showing all genes transcribed from one strand
Full size table
Table 2 List of completely sequenced metazoan mtDNAs showing a gene order with the two rRNA genes separated by more than three genes
Full size table
Table 3 List of completely sequenced metazoan mtDNAs showing a gene order with atp6 and atp8 genes distantly located
Full size table
Table 4 Mitochondrial sequences analyzed for the calculation of rrnL and rrnS gene length
Full size table
Figure 2
figure 2

Minimal breakpoint tree reconstructed from mitochondrial arrangement of Deuterostomes, excluding tRNA genes. Numbers indicate breakpoint distance values. Data as follows: Echinoidea—Paracentrotus lividus (J04815), Strongylocentrotus purpuratus (X12631), Arbacia lixula (X80396); Crinoidea—Florometra serratissima (AF049132);Asteroidea—Asterina pectinifera (D16387); Ciona savignyi (AB079784); Halocynthia roretzi (AB024528); Vertebrata/Cephalochordata—general Vertebrata and Cephalochordata mtDNA gene organization (data from Embl, May 2003); Hemichordata—gene organization of hemichordate Balanoglossus carnosus (AF051097) and vertebrate Teleostei Myctophum affine (AP002922) and Conger miriaster (AB038381).

Full size image

Base Composition and Codon Usage

The G + C content of the C. intestinalis mtDNA sense strand is 21.4%. According to their percentage, the relative usage of bases is T > A > G > C, with very low use of C (about 9.5%). This trend is conserved both in highly constrained sites (first and second codon positions, P1 and P2, respectively) and in less constrained sites (third codon positions, P3; third position of quartet codons, P3Q; noncoding sequences, NC) (see Table 1). A similar compositional pattern is present in the C. savignyi and Halocynthia mt genomes. The asymmetric distribution of bases between the two strands (AT- and GC-skew) is very low in C. intestinalis mtDNA (Table 1). Indeed, although C is the lowest represented nucleotide in most sites, G is almost as rare as C in P2, P3, and P3Q. However, an asymmetric base distribution can be observed in C. savignyi and, particularly, in Halocynthia, which show a positive GC-skew and a negative AT-skew (Table 1).

Table 5 List of completely sequenced metazoan mtDNAs showing all genes transcribed from one strand
Full size table

The bias in base composition toward AT and against GC affects both the codon usage pattern and the amino acid composition of proteins. Indeed, codons GGC (Gly) and CGC (Arg) are not used at all in C. intestinalis mtDNA and are among the lowest used codons in other ascidians (data not shown).

Overall, the compositional features of ascidian mtDNA reflect the basal positions of these organisms in the phylum, being closer to those of Cephalochordata than to Vertebrata. The GC% is the lowest among all sequenced mtDNAs of Chordata. The sense strand is C-poor as in Cephalochordata and unlike in Vertebrata, where the sense strand is G-poor (Saccone et al. 1999; Yokobori et al. 1999). AT- and GC-skew are quite variable, being very low in some species and remarkable in others. Moreover, in the case of an asymmetric base distribution between the two strands, the ascidian GC-skew is positive as in Cephalochordata, therefore it is inverted compared to mammals and other vertebrates (negative GC-skew) (Reyes et al. 1998; Saccone et al. 1999).

Protein Genes

The short open reading frame located between trnF and trnC in C. intestinalis mtDNA has been identified as encoding the F0-ATPase subunit 8 (35 amino acids long). A similar ORF is present in the same position in C. savignyi mtDNA (46% amino acid similarity) and has been found in the region between cox1 and nad3 in H. roretzi, which was originally reported as noncoding (33% amino acid similarity) (Yokobori et al. 1999). The ATP8 protein is quite heterogeneous in length between organisms; it is 48 aa long in S. cerevisiae and longer in Drosophila (52 aa), human (68 aa), and cnidarian (72 aa). The orfB gene of plants and protists, which is the atp8 homologue, is even longer (Gray et al. 1998). A peculiarity of this protein is its conservation more at the level of secondary structure and chemical character of amino acids than in the specific amino acid identity (Papakonstantinou et al. 1996b). In yeast, where it has been widely studied, ATP8 is an intrinsic membrane protein composed of three domains: an N-terminal domain located in the intermembrane space and showing a conserved MPQL motif, a central membrane-spanning hydrophobic domain, and a C-terminal positively charged region exposed in the matrix space. The N-terminal motif plays a role in ATPase activity (Devenish et al. 1992), whereas the positively charged amino acids at the C terminal are involved in both assembly and function of the F0 sector (Papakonstantinou et al. 1993, 1996a). The short ORFs identified in the three ascidian mtDNAs all have a MPQL motif at the N terminal, a transmembrane-spanning domain, as predicted by the DAS software, and one (in C. intestinalis) or two (in C. savignyi and H. roretzi) positively charged amino acids in the C-terminal domain. Thus, despite the low global similarity to other ATP8, such ORFs can be unambiguously regarded as an ATP8 form with short peripheral domains, as reported in the alignment in Fig. 4. In yeast, several lines of evidence indicate that ATP8 is an integral component of the ATPase stator stalk, interacting by means of both peripheral domains with remaining F0 sector subunits and performing a primarily structural role (Stephens et al. 2003). The different number and organization of the F0 subunits in the ATPase of different organisms might explain the high divergence of this protein in various species and thus its peculiarity in ascidians.

Figure 3
figure 4

Alignment of ATP8 amino acid sequences from yeast Saccharomyces cerevisiae (AJ011856), mammal Homo sapiens (V00662), Ciona savignyi (AB079784), Halocynthia roretzi (AB024528), hemichordate Balanoglossus carnosus (AF051097), echinoderm Paracentrotus lividus (J04815), insect Drosophila melanogaster(U37541), cnidarian Metridium senile (AF000023), and sponge Tetilla sp. (AF035265). Highly conserved amino acids are indicated in bold face. Positively charged amino acids are shaded gray.

Full size image

ATG is the most common initiation codon, with only nad5 and nad4L genes starting with ATT, nad2 with ATC, and nad4 with ATA. In accordance with the low G% of the genome, codons ending with G might be expected to be avoided. However, this rule appears not to be valid for start codons, suggesting that there is a positive selection for the choice of ATG as initiator codon. Nine protein genes terminate with a TAA stop codon, and only one (nad4L) with a TAG codon. Partial stop codons that are completed by transcript polyadenylation, as reported in mammals, have been hypothesized for three genes: cob, atp8 (TA incomplete stop codon), and nad5 (T incomplete stop codon). The identification of stop codons has been assisted by mapping mt-like ESTs of C. intestinalis onto the mtDNA; indeed EST data allowed the identification of poly(A) stretches representing the poly(A) tails of the primary or mature mitochondrial transcripts (Gissi and Pesole 2003). Based on genomic sequence data, the cob frame stops with a TAG codon inside the downstream trnP gene, however, based on EST inferred transcripts a poly(A) tail starts 9 bp upstream of the TAG codon (Gissi and Pesole 2003). Thus a TAA stop codon should be created from an incomplete TA codon by polyadenylation of the transcript, excluding the existence of an overlap between the cob and trnP genes. A similar situation is found in nad5, where polyadenylation eliminates the possibility of a 5-nt overlap with the downstream trnL(UUR), and in atp8 (avoided overlap of one nucleotide with trnC). In addition, EST mapping suggests that nad4L could use the complete TAG stop codon or an incomplete T/TA codon (Gissi and Pesole 2003). As suggested by Boore (2001), the coupling of incomplete stop codons to complete in-frame stop codons requiring a few-nucleotide overlap to the downstream gene could prevent translational readthrough in case of incorrect transcript maturation.

The size of the identified protein genes is similar to those of their homologues in H. roretzi and C. savignyi, with few exceptions (data not shown). A 25-amino acid extension at the C-terminal of the Halocynthia NAD4L protein has not been found in any of the aligned deuterostome homologues (see Materials and Methods), and a detailed analysis revealed the presence of a tRNA-Met(AUG) secondary structure entirely included in this extra region. Such putative tRNA exhibits a normal cloverleaf secondary structure except for an extremely long TΨC stem (7 bp) (Fig. 3). An overlap of 13 bp is present between these two genes even assuming a nad4L incomplete termination codon. It is unclear whether this tRNA gene is functional or is a pseudogene of recent origin that has become part of a protein reading frame. A similar situation is found in the sea urchin mtDNA, where a trnL(CUN) pseudogene forms the amino end of the NAD5 protein (Cantatore et al. 1987). It is striking that the trnM(AUG) gene is among the genes transposed between Ciona mtDNAs.

Figure 4
figure 3

Secondary structure of the hypothesized tRNA-Met(AUG) of H. roretzi (locations 5838–5913 on AB024528 sequence) and C. savignyi (AB079784). Canonical and G–T base pairs are differently indicated.

Full size image

Gene overlaps were found in all cases of adjacent protein-coding genes with no tRNA genes in between, that is, for the gene pairs cox2–cob and atp6–nad2. The estimated overlaps are 26 and 10 nt, respectively. The overlap between cox2 and cob is observed also in H. roretzi and C. savignyi. The nad2 reading frame overlaps the atp6 gene for 46 nt, but protein similarity data support a shorter overlap of only 10 nt (data not shown). The identification of bicistronic transcripts in the putative mitochondrial transcripts of C. intestinalis inferred from the EST data further support the existence of both overlaps (Gissi and Pesole 2003).

tRNA Genes

Proposed secondary structures for the 24 identified tRNAs are shown in Fig. 5. The genes vary in size from 52 (trnC) to 70 (trnE and trnD) nt. In all tRNAs a T precedes the anticodon and a purine, mostly A, follows it, except in tRNA-Asp, where the anticodon is followed by a T. Each tRNA has been inferred to have an aminoacyl acceptor stem of 7 nt and an anticodon stem of 5 nt, except for tRNA-Ser(AGY), where the anticodon stem is 7 nt long. There are no unpaired nucleotides in the acceptor stems, although one mismatched base pair is present in the anticodon stems of tRNA-Phe and tRNA-Gly(AGR). The DHU and TΨC arms vary greatly in size and sequence. The stems are 3, 4, or 5 nt long in both arms. The TΨC loop size is highly variable between different tRNAs ranging from 2 to 9 nt, whereas the DHU loop is 5 or 6 nt long in most tRNAs except for tRNA-Glu (13 nt long). Only tRNA-Ser(AGY) and tRNA-Cys lack the DHU arm, which is replaced by loops 4 and 5 nt long, respectively. The variable loop is 2, 3, or 4 nt long and the size of spacers between arms is in accordance with the model of animal mitochondrial tRNAs reported by Kumazawa and Nishida (1993). The tRNA secondary structure lacking the DHU arm is typical of metazoan tRNA-Ser(AGY) (Kumazawa and Nishida 1993), whereas in tRNA-Cys this structure has been found only in Cephalochordata (Boore et al. 1999; Spruyt et al. 1998) and in some vertebrates, where it has been correlated with the loss of the immediately adjacent stem–loop structure of the L-strand replication origin (Macey et al. 1999; Seutin et al. 1994; Zardoya and Meyer 1996). Among ascidians, the tRNA-Cys has a normal cloverleaf secondary structure in Halocynthia (Yokobori et al. 1999) and possesses a DHU replacement loop 13 bp long in C. savignyi.

Figure 5
figure 5

Putative secondary structures of the 24 tRNAs encoded by the C. intestinalis mtDNA. Canonical and G–T base pairs are differently indicated.

Full size image

The secondary structures of the two tRNA-Met identified in C. intestinalis are compatible with their functionality (Fig. 5). Most metazoan mtDNAs possess a single trnM gene with a 5′-CAU-3′ anticodon able to decode methionine codons, ATG or ATR, depending on the genetic code, without discriminating initial and internal methionines (Yokobori et al. 2001). Only in the mollusks Mytilus edulis and M. californianus has an additional gene for trnM(AUA) been found and the transcription product identified (Beagley et al. 1999; Hoffmann et al. 1992). In addition, a tRNA-like structure with an anticodon 5′-UAU-3′ has been found in the nematode Trichinella spiralis, but its unusual secondary structure suggests that it may be not functional (Lavrov and Brown 2001). In Mytilus it has been suggested that tRNA-Met(AUG) may be used as initiator, preferentially recognizing the ATG codon, whereas tRNA-Met(AUA) may be used as an elongator, although there is no definitive evidence to support this hypothesis (Beagley et al. 1999). This hypothesis could also be valid for ascidians. Indeed, in C. intestinalis, the initiator codon is ATG in 9 of the 13 mt-protein genes. Moreover, tRNA-Met(AUG) presents two consecutive G–C base pairs close to the bottom of the anticodon. Three consecutive G–C base pairs at the base of the anticodon stem are essential for the function of the E. coli tRNA-Met as initiator (Rajbhandary and Chow 1995) and are conserved in the mitochondrial tRNA-Met(AUG) of most Metazoa (Sprinzl et al. 1998). Considering the low GC% in C. intestinalis mitochondrial tRNAs, the conservation of such feature is even more significant. To shed light on the role performed by the two tRNA-Met, it will be important to define their decoding capacity. The two tRNA-Met could potentially recognize both ATR codons. Indeed, the tRNA-Met(AUA) of H. roretzi has U-derivative at the wobble position, allowing the decoding of both ATR codons (Kondow et al. 1998). In metazoans, tRNA-Met(AUG) decodes both ATR codons when there is a 5-formylcytidine at the wobble position of the anticodon or an A-derivative at the position 3′ adjacent to the anticodon, while it recognizes only the AUG codon when an unmodified nucleotide is present at both positions (Tomita et al. 1999). Also C. savignyi and Halocynthia tRNA-Met (AUG) possess two consecutive G–C base pairs in the anticodon stem (Fig. 3). It would be interesting to verify the existence and functionality of both tRNA-Met in a larger number of ascidian species.

It is striking that among deuterostome mt genetic codes, AUA codes for Met in Chordata but for Ile in Echinodermata and Hemichordata, and at least in the genus Ciona there is an additional tRNA able to decode such codon. To verify the meaning of AUA codons in the ascidian mt genetic code, we looked for sites with 100% conserved Met or Ile in the alignments of the most conserved mitochondrial genes (cox1, cox2, cox3, and cob), which include species as distant as C. elegans and human (20 representative species plus ascidians). There are 10 Met conserved sites in the analyzed alignments, and in all of them at least one ascidian species presents an AUA codon, demonstrating that the AUA codon is used in sites where methionine is conserved over long evolutionary distances. On the contrary, there are no Ile conserved sites with an AUA codon in ascidians. These observations unambiguously demonstrate that AUA codes for Met in ascidians.

In M. edulis, similarities at the 5′ end of the two trnM genes suggest that they arose by gene duplication (Hoffmann et al. 1992). In ascidians, the low similarity between the two trnM genes of the same species seems to exclude the origin of the two genes by recent gene duplication of a common ancestor.

rRNA Genes

The 5′ and 3′ ends of the small subunit rRNA gene (rrnS) have been tentatively defined as immediately adjacent to the ends of the flanking genes trnL(UUR) and trnW (Fig. 1). An alignment of ascidian rrnS shows the remarkable similarity at both ends of the predicted gene, supporting the validity of these boundaries.

The C. intestinalis large subunit rRNA gene (rrnL) is located between the nad1 and trnI genes (Fig. 1). Boundaries have been identified by sequence similarity to other ascidian rrnL and by searching for the TRGCAKAn5G mt rRNA transcription termination signal, which is well conserved near the 3′ end of rrnL in a wide range of organisms (Valverde et al. 1994). In H. roretzi, EST data demonstrate that rrnL transcripts stop with a poly(A) tail added at multiple sites located in an AT-rich region immediately upstream of a typical rRNA termination box, TGGCATAtaaacG (Gissi and Pesole 2003). Searches using the Halocynthia termination box as a probe revealed a degenerate version of the rRNA termination signal (TGTCGGAtaaacG) in the mtDNA of C. intestinalis. This motif is present in the whole genome only in the gene trnI, located immediately downstream of the rrnL. The pattern TGKCRKAtaaayG might be interpreted as an rRNA termination box specific for ascidians. The 5′ end of C. intestinalis rrnL has been tentatively located 100 bp downstream of nad1, that is, after a repeated sequence found in this 100 bp-long region (see below) and starting from the region of high similarity to C. savignyi rrnL.

Both ascidian rRNA genes are the shortest among deuterostomes (Table 2), moreover, they are comparable in length to those of nematodes and platyhelminthes, which possess the shortest rRNAs among metazoa (Hu et al. 2002).

Table 6 List of completely sequenced metazoan mtDNAs showing a gene order with the two rRNA genes separated by more than three genes
Full size table

Both rRNA genes of Ciona intestinalis show very low global similarity to the mitochondrial rRNAs of other metazoan species. Using such genes as queries in a Blastn search against the nonredundant nucleotide database (nr), the only statistically significant matches to other organisms are with very short regions of mt rRNA genes (data not shown). Indeed, the only element which identifies rrnS is a 23-bp-long sequence. In the human rRNA secondary structure (Cannone et al. 2002) this motif is reported as part of a small stem–loop structure, with the loop implicated in tertiary interactions. In C. elegans rrnS, this structure is conserved, although the primary sequence is not identical. Figure 6 reports the alignment of these rRNA regions in ascidians, human, and C. elegans, together with data on secondary structure. As for rrnS, a short region, 35–18 bp long, allows unambiguous identification of the rrnL gene. This region has been mapped on the secondary structure model of human and C. elegans large subunit mt rRNA (Cannone et al. 2002), showing that the ends are included in two distinct stem structures, whereas the internal part is unpaired and highly conserved at the level of the primary sequence (Fig. 6). The high degree of conservation suggests that both regions represent important signatures of mt rRNAs.

Figure 6
figure 6

Alignment of the short conserved sequence pattern of small (rrnS) and large (rrnL) subunit ribosomal RNAs. Positions involved in tertiary interactions are underlined; positions involved in stem formation are shaded gray. Asterisks mark conserved positions. Numbers refer to location on the reported mtDNA entry.

Full size image

The low global similarities to homologous rRNAs suggest that the ascidian rRNAs are subject to a fast evolutionary rate and/or to taxon-specific functional constraints.

Noncoding Regions

There are 447 noncoding bp in the mtDNA of C. intestinalis, representing 3% of the whole genome, a portion comparable to that observed in other ascidian genomes (2.4% in C. savignyi and 3.3% in H. roretzi). Sixteen noncoding regions, ranging in size from 1 to 20 bp, have been considered intergenic spacers. Four additional noncoding regions range from 30 to 50 bp. The two longest noncoding regions are an 85-bp-long region located between cox3 and trnK and a tentatively 100-bp-long noncoding region located between nad1 and rrnS (Fig. 1). These two regions share a sequence 30 bp long, identical except for a single T/A substitution, which is located 11 bp downstream of a further conserved sequence, 13 bp long, that includes the stop codon of the upstream protein-coding gene. Overall, there is an imperfect repeated sequence 54 bp long showing 87% identity and mostly contained in the two longest noncoding regions. As inferred from mitochondrial-like EST analysis (Gissi and Pesole 2003), both such long noncoding regions are transcribed and seem to form a 3′ untranslated region in the cox3 and nad1 transcripts. There are no similar sequences in other regions of C. intestinalis mtDNA or in the mt genome of other ascidians. The functional meaning of this repeated sequence is obscure but it is peculiar that the two copies of this repeat are close to nad1, which is the only protein-gene transposed in both Ciona mtDNAs. In C. savignyi nad1 is flanked by short noncoding regions (18 and 69 bp long upstream and downstream, respectively) with no similarity to those of C. intestinalis. The C. savignyi nad1 has an extra 30-aa-long C-terminal region, absent in homologous deuterostome proteins. This region contains a 9-bp-long motif which is identically repeated in the noncoding region upstream of nad1. Thus, the nad1 C-terminal region could have originated by the extension of the original frame into a secondarily acquired downstream region. Only one other repeated sequence has been found in ascidian mtDNAs: in H. roretzi there are two copies of trnF with a 94% identity and conservation of the tRNA cloverleaf secondary structure, clearly suggesting a recent duplication of the trnF gene (Gissi and Pesole 2003). The imperfect repeats found in C. intestinalis mtDNA could represent traces of the nad1 transposition event—remnants either of the ancestral target sequence or of a tRNA gene transposed together with nad1.

No similarity has been found between the noncoding regions of the three completely sequenced mtDNAs, thus sequence conservation does not help the identification of a putative control region. Only weak speculations can be made as to the existence of a control region in ascidian mtDNAs. Considering that high AT content is a feature of some invertebrate control regions (Saccone and Sbisà 1994), the 37-bp-long noncoding region located between nad6 and trnG(GGN) (97% AT) is a good candidate in C. intestinalis. Using the same criterion, in other ascidian mtDNAs the best candidates should be the 112-bp-long sequence of H. roretzi (79% AT) and the 47-bp-long sequence of C. savignyi (87% AT). However, an alternative hypothesis is that the C. intestinalis 85- and 100-bp-long noncoding regions close to nad1 represent a multipartite control region. Indeed, such regions are characterized by the presence of an imperfect repeat and proximity to a protein gene transposed between two closely related species. The presence of repeats is a peculiar feature of vertebrate and invertebrate control regions and in Chordata the few tolerated gene transpositions mostly involve the control region (Saccone et al. 1999).

No secondary structures similar to the typical stem–loop of vertebrate L-strand replication origins (Hixson et al. 1986) have been found in C. intestinalis or other ascidian mtDNAs.

Conclusions

The C. intestinalis and, more generally, the ascidian mtDNA features are very peculiar and suggest that the evolutionary dynamics of these genomes are very different from those of other Chordata. The mt genomes of deuterostomes are believed to be subject to less frequent gene arrangement than protostomes, with Chordata gene arrangement almost frozen (Boore 1999; Saccone et al. 1999). Ascidians should be considered an exception to this tendency because of their high gene rearrangement rate. The extreme reduction of genome size, due to short noncoding regions and smaller rRNA genes, the low GC%, the strong intraclass gene rearrangement, and the transcription of all genes by the same strand are features shared together only by Ascidiacea, Platyhelmintes, and Nematoda. To date an intragenus rearrangement has been found only in the genera Ciona and Schistosoma (Platyhelminthes). Interestingly, all these taxa show a high evolutionary rate in the nuclear rRNA genes, which leads to long branches in phylogenetic reconstruction (Cameron et al. 2000; Winchell et al. 2002; Aguinaldo et al. 1997; Peterson et al. 2001). Rapid evolution (highly rearranged mt genomes and high substitution rate of nuclear genes) could be the common factor among the above-mentioned taxa. Nevertheless, ascidians still conserve some traits of Chordata mtDNA, particularly of Cephalochordata. Indeed, the absence of a typical vertebrate control region, a C-poor sense strand, and a tRNA-Cys lacking the DHU arm are features common to Cephalochordata and Ascidiacea. Other features, such as the variation of the genetic code with the consequent acquisition of an additional trnG gene (Yokobori et al. 2001) and the presence of two trnM genes whose functionality and decoding ability have to be demonstrated, are almost exclusive of ascidians. The atp8 gene, which was previously undetected in ascidian mtDNAs, is conservatively present in all Chordata mt genomes.

Appendix