Introduction

Complete mitochondrial (mt) DNA sequences have been determined for a number of vertebrates (269 species as of 2 June 2003; NCBI Web site: http://www.ncbi.nlm.nih.gov/). The vertebrate mtDNAs characterized to date are double-stranded, circular molecules approximately 16 kb in size and encode genes for 13 proteins, 2 rRNAs, and 22 tRNAs, as well as having a major noncoding region (control region) that contains initiation sites for transcription and replication (Clayton 1992). Most vertebrate mtDNAs contain a short intergenic spacer that contains the origin of light-strand replication, which is found in mammals, cartilaginous and teleost fish, and frogs but is lacking in jawless fish, blind snakes, crocodilians, and birds (Macey et al. 1997). The arrangement of these coding and noncoding regions was once considered to be fairly conservative, but it has been recognized that deviations from the typical gene order are not rare (Boore 1999) and that tRNA genes are involved in most cases of such gene rearrangement (Macey et al. 1997). The gene rearrangements have been proposed as occurring by tandem duplication of gene regions as a result of slipped-strand mispairing, followed by deletions of redundant genes (Levinson and Gutman 1987; Moritz and Brown 1987; Boore 2000; Lavrov et al. 2002).

Genomic size of animal mitochondrial DNA is usually minimized over time (Rand 1993). Therefore, when duplications of a partial region do occur, they are followed by a rapid elimination of redundant material, which may or may not result in a rearrangement of gene order (Macey et al. 1997). In contrast to this general view, some stabilized duplications that have resulted in large stretch of apparently noncoding regions were reported, such as a control region (CR)-like duplication in snakes being evolutionary stable for at least 70 million years (Myr) (Kumazawa et al. 1996) and a partial CR-Iike duplication in salamanders existing for at least 20 Myr (McKnight and Schaffer 1997). Bakke et al. (1999), however, reported a tRNAThr–tRNAPro intergenic spacer shared by two groups of cod fishes (Gadidae and Phycidae), the common ancestor of which originated from at least the beginning of the Oligocene (approximately 40 Myr ago) (Nolf and Steurbaut 1989). The reason why these noncoding regions have been maintained has been suggested to be positive selection due to the retention of function as an alternative site of initiation of replication or transcription (Kumazawa et al. 1996) or of heavy-strand transcript termination (Bakke et al. 1999). There was, however, another explanation which regarded stable stem-and-loop structures formed by the spacers as the possible factor against deletion (McKnight and Schaffer 1997).

In this paper, we report the presence of stable tRNA pseudogenes, defined as derivatives of tRNA genes that have lost their function as tRNA genes, found in the mitogenome of the teleost fish family Scaridae. During the course of a molecular phylogenetic study of the fish suborder Labroidei (cichlids, surfperches, damselfish, wrasses, odacids, and parrotfish), we determined a complete mitochondrial nucleotide sequence for a scarid, Chlorurus sordidus, and found a gene rearrangement, including a tRNA pseudogene, in an IQM tRNA gene-cluster region. In order to explore evolutionary origins of this rearrangement, we sequenced the corresponding partial region for an additional 10 scarid species and 20 possible labroid sister species. Observed phylogenetic distribution of the derived gene order indicated that the gene rearrangement occurred in a common ancestor of the parrotfishes. Phylogenetic analysis of the tRNAMet genes and pseudogenes demonstrated that the pseudogenes also had a single origin and survived at least 14 Myr. Based on their potential secondary structures (retaining “top halves” of the clover-leaf structures), locations within the mitogenomes (flanking the 5′ ends of the ND2 genes), and stabilities over time (survived at least 14 Myr), they were presumed to function as punctuation marks for mitochondrial mRNA processing.

Materials and Methods

Specimens

Mitochondrial DNA sequences were obtained from 11 species of Scaridae, representing seven of the 10 currently recognized genera (Calotomus, Cetoscarus, Chlorurus, Leptoscarus, Nicholsina, Scarus, and Sparisoma). Complete mitochondrial nucleotide sequence was determined for 1 of the 11 species, Chlorurus sordidus. To assess the origin of the novel gene order found in Scaridae, partial sequences were also obtained from 20 species of Labridae, representing 19 of the 68 recognized genera (Parenti and Randall 2000), and a single species of Odacidae (including 4 genera with 12 species [Gomon and Paxton 1985]). Although the interrelationships among scarids and the two additional families remain unclear, they are currently placed within the family Labridae (e.g., Kaufman and Liem 1982; Bellwood 1994). According to the classification of “labroid” genera used in Gomon (1997), the species examined in the present study represent all 10 tribes of “Labridae,” which includes Scaridae and Odacidae as tribes (Table 1). One of the examined labrid species, Pseudodax moluccanus, has been inferred to be an immediate sister group of the Scaridae based on osteological data (Gomon 1979; Bellwood 1994). Tropheus duboisi (Cichlidae) and Abudefduf vaigiensis (Pomacentridae) were used as outgroups, on the basis of traditionary accepted classification (Nelson 1994). We added two other teleost fish, Aulostomus chinensis (Gasterosteiformes, Aulostomidae) and Kurtus gulliveri (Perciformes, Kurtoidei, Kurtidae), to the analysis of gene orders, both of which were indicated to exhibit gene rearrangements in IQM region. Species used in this study are listed in Table 1 with accession numbers of sequences.

Table 1 List of species used in this study with DDBJ/EMBL/GenBank accession numbers
Full size table

Laboratory Protocols

Genomic DNAs were extracted from muscle tissue preserved in 99.5% ethanol, using the Quiagen DNEasy tissue kit.

The entire mitogenome of Chlorurus sordidus was amplified using a long-PCR technique (Cheng et al. 1994). Two sets of fish-versatile primers (L2508-16S + H1065-12S and L12321-Leu + S-LA-16S-H; for sequences, see Kawaguchi et al. [2001]) were used to amplify the entire mitogenome in two reactions (Fig. 1). The long-PCR products were diluted with TE buffer (1:20) for subsequent use as short-PCR templates. A total of 47 fish-versatile primers (Miya and Nishida 1999, 2000; Inoue et al. 2000, 2001ab) was used to amplify contiguous, overlapping segments of the entire mitogenome. A total of six species-specific primers was designed in cases where no appropriate fish-versatile primers were available. Most partial sequences, including IQM-related regions from the other fish, were amplified using the following primer pair: L4160-ND1 (Kumazawa and Nishida 1993) + H4866-ND2 (Miya and Nishida 1999). Alternative two primer pairs, S-LA-16S-L (Miya and Nishida 2000) + H4866-ND2 and L4160-ND1 + H5334-ND2 (Miya and Nishida 1999), were used for Cetoscarus bicolor and Pseudodax moluccanus, respectively. A list of the PCR primers used in this study is available in the supplementary table (***).

Figure 1
figure 1

Linearized representation of the mitochondrial gene arrangement for Chlorurus sordidus. All protein-coding genes are encoded by the H-strand with the exception of the ND6 gene, which is coded by the L-strand. Transfer RNA (tRNA) genes are designated by single-letter amino acid codes, those encoded by the H-strand and L-strand being shown above and below the gene map, respectively. Two pairs of long-PCR primers (L2508-16S + H1065-12S and L12321-Leu + S-LA-16S-H; for primer sequences, see Kawaguchi et al. [2001]) amplify two segments that cover the entire mitochondrial genome. 12S and 16S indicate 12S and 16S ribosomal RNA genes; ND1–6 and 4L, NADH dehydrogenase subunits 1–6 and 4L genes; ATPase 6 and 8, ATPase subunits 6 and 8 genes; cyt b, cytochrome b gene; CR, control region; OL, origin of light-strand replication; ψM tRNAMet pseudogene.

Full size image

Long PCR and subsequent short PCRs were carried out as described by Miya and Nishida (1999). Double-stranded PCR products, purified using a Pre-Sequencing Kit (USB), were subsequently used for direct cycle sequencing with dye-labeled terminators (Applied Biosystems). Primers used were the same as those for PCR. All sequencing reactions were performed according to the manufacturer’s instructions. Labeled fragments were analyzed on a Model 3100 DNA sequencer (Applied Biosystems).

Data Analysis

DNA sequences were analyzed using the computer software package program DNASIS version 3.2 (Hitachi Software Engineering). Locations of the 13 protein-coding genes in the mitogenome of Chlorurus sordidus were determined by comparisons of DNA or amino acid sequences of teleost fish mitogenomes. The two rRNA genes in the mitogenome of the species were identified by sequence homology and secondary structure (Gutell et al. 1993). The tRNA genes in the mitogenome of C. sordidus and those in the partial sequences from other fishes were identified by their clover-leaf secondary structures (Kumazawa and Nishida 1993) and anticodon sequences. Sequence data are available from DDBJ/EMBL/GenBank with accession numbers as shown in Table 1.

In order to examine the origin of the scarid tRNA pseudogenes, neighbor-joining (NJ) (Saitou and Nei 1987) and maximum parsimony (MP) analyses were performed using PAUP* 4.0b8a (Swofford 1998). The sequences of the pseudogenes and genes from 11 scarid species were aligned together with those from 21 labrids, a single odacid, and two outgroup species, using the computer program ClustalX (Thompson et al. 1997). The resultant aligned sequences were adjusted by eye and subjected to NJ and MP analyses. In NJ analysis, evolutionary distances were calculated using the Jukes and Cantor (1969) model. To evaluate the robustness of the internal branches of the NJ tree, 500 bootstrap replications (Felsenstein 1985) were calculated using PAUP* 4.0b8a. Unweighted MP analysis was performed using heuristic search option, 100 random addition sequence replicates, and tree–bisection–reconnection algorithm in PAUP* 4.0b8a.

Results and Discussion

Organization of the Mitogenome of Chlorurus sordidus

The total length of the Chlorurus sordidus mitogenome is 16,681 bp, including 2 rRNA, 22 tRNA, and 13 protein-coding genes as found in other vertebrates (Fig. 1). Also, as in other vertebrates, most genes were encoded on the H-strand, except for the ND6 and eight tRNA genes (Fig. 1), and all genes were similar in length to those in other teleost fish. The order of these genes was largely identical to that of typical vertebrate mitogenomes, except for the relative positions of the three tRNA genes, tRNAIle (I), tRNAGln (Q), and tRNAMet (M), clustering between the ND1 and the ND2 genes (Fig. 1). In vertebrate mitogenomes, the three tRNA genes are usually arranged in this order (IQM); in C. sordidus, however, the tRNAMet gene was inserted between the tRNAIle and the tRNAGln genes (IMQ) (Figs. 1 and 2).

Figure 2
figure 2

Partial sequence of mitochondrial DNA from Chlorurus sordidus including an IQM tRNA gene-cluster region. This segment is amplified by a primer set, L4160-ND1 (Kumazawa and Nishida 1993) + H4866-ND2 (Miya and Nishida 1999), which is indicated by serial arrows. The most common arrangement for this region in vertebrates consists of the genes encoding ND1, tRNAIle (I), tRNAGln (Q), tRNAMet (M), and ND2. In C. sordidus, the tRNAMet gene is inserted between the tRNAIle and the RNAGln genes (IMQ), and the tRNAGln gene is followed by the tRNAMet pseudogene. Sequences are presented as L-strand sequence from 5′ to 3′ ends. Direction of transcription for each gene indicated by arrows, and beginning and end of each gene indicated by vertical bars. Transfer RNA genes are boxed, and corresponding anticodons are indicated in black boxes. In protein-coding genes, amino acids are positioned under the first codon position, and stop codon is indicated by asterisk. Sequence data are available from DDBJ/EMBL/GenBank under accession number AB006567.

Full size image

The mitogenome of C. sordidus contained three noncoding regions longer than 40 bp (Fig. 1). The longest one (920 bp), located between the tRNAPro and the tRNAPhe genes, appears to correspond to the control region, because it had termination-associated sequences (TAS; Doda et al. 1981) and conserved sequence blocks (CSB; Walberg and Clayton 1981) characteristic of this region. The shortest noncoding region (43 bp), located in a cluster of the five tRNA genes (WANCY), appeared to correspond to the origin of light strand replication (OL), because it had the potential to fold into a stable stem-loop secondary structure, and the conserved motif 5′-GCCGG-3′ was found at the base of the stem within the flanking tRNACys gene (C). Both of the functional noncoding regions were located at the identical positions to those in the mitogenomes of other teleost fish. The remaining noncoding region of 66 bp, however, was unique to the parrotfish and appeared to be a tRNAMet pseudogene on the basis of the following two reasons. First, the reverse complement sequence of this region has potential for forming a clover-leaf secondary structure. Second, it exhibited some sequence similarity (70%) to the tRNAMet gene, while similarities to other 21 tRNA genes were either ≤49% or unalienable. It was successfully aligned with the tRNAMet gene (Fig. 3) and could form a clover-leaf secondary structure (Fig. 4B), whereas anticodon sequence was degenerated from CAU into UAC (Figs. 4A and B), indicating loss of function as a tRNA gene.

Figure 3
figure 3

Aligned sequences of the 10 putative tRNAMet pseudogenes (ψM) of 11 species of Scaridae and a tRNAMet gene (M) from Chlorurus sordidus. The sequence of tRNAMet pseudogene from Scarus guacamaia (*) is identical to that from S. ghobban. Dots indicate sequence identity with the first sequence (tRNAMet gene of C. sordidus), and dashes indicate alignment gaps. Sequence in black box indicates anticodon for the tRNAMet gene, and underlined sequences indicate its stem regions. Note that the corresponding anticodon sequences in the pseudogenes are degenerated in most species.

Full size image
Figure 4
figure 4

Potential secondary structures of the tRNAMet from Chlorurus sordidus (A) and transcripts of the putative tRNAMet pseudogenes encoded in the mitochondrial genomes of seven parrotfish (B, Chlorurus sordidus; C, Scarus guacamaia; D, Cetoscarus bicolor; E, Leptoscarus vaigiensis; F, Calotomus japonicus; G, Nicholsina usta; H, Sparisoma chrysopterum), a lanternfish, Diaphus splendidus (I), a pearlfish, Carapus bermudensis (J), and a codlet, Bregmaceros nectabanus (K). Dashes represent Watson–Crick bonds, and pluses G–U pairs. Note that the transcripts of pseudogenes well sustain stem structures corresponding to the “top half” of the tRNA secondary structure.

Full size image

Phylogenetic Distribution of the Derived tRNA Gene Order and Pseudogene

Sequencing of the IQM-related regions for other 10 scarid species demonstrated that all the examined species shared the same gene order and insertion pattern as in Chlorurus sordidus. On the other hand, all possible sister groups examined had the typical gene order of vertebrate mitogenomes.

Streelman et al. (2002) proposed a phylogenetic hypothesis among scarid genera based on ≍2 kb of nuclear and mitochondrial DNA sequences. Although 3 of the 10 recognized scarid genera (Bolbometopon, Cryptotomus, and Hipposcarus) were not examined in the present study, mapping of the known gene order and insertion patterns on the phylogenetic hypothesis indicated a single origin of the derived pattern (Fig. 5): the distribution of the derived pattern was parsimoniously explained by a single gene-rearrangement event in a common ancestor of the family or in an ancestor older than that species. On the other hand, the scarid gene order and insertion pattern were not found in any of the possible sister groups examined. These results lead to the conclusion that the derived gene order and insertion pattern shown in scarids probably originated in a common ancestor of the family.

Figure 5
figure 5

Scarid mitochondrial gene and pseudogene orders mapped on their proposed phylogenetic hypothesis (Streelman et al. 2002). I, M, Q, and ψM represent the tRNAIle, tRNAMet, and tRNAGln genes and tRNAMet pseudogene, respectively. Note that the derived gene order and pseudogene are found throughout scarids, indicating their single origin in a common ancestor of the family. Time scales for Calotomus and two major lineages are from Bellwood and Schultz (1991) and Streelman et al. (2002), respectively.

Full size image

It should be noted that the novel scarid gene order was also found in the mitogenome of another teleost fish, Diaphus splendidus (order Myctophiformes) (Fig. 6A) (Miya et al. 2001). Because the fish is clearly distantly related to the parrotfish (order Perciformes), the shared derived gene order should have two independent origins. Interestingly, another teleost fish, Carapus bermudensis (order Ophidiiformes), with another type of novel gene order possesses the tRNAMet pseudogene between the tRNAGln (Q) and the ND2 genes like the above fish (Fig. 6B) (Miya et al. 2003). As discussed below, some constraints seem to exist on the presence and location of the tRNA pseudogene.

Figure 6
figure 6

Vertebrate mitochondrial gene rearrangements shown within the IQM region. A, Diaphus splendidus (data from Miya et al. 2001); B, Carapus bermudensis (Miya et al. 2003); C, Bregmaceros nectabanus (Miya et al. 2003); D, Kurtus gulliveri (this study); E, Aulostomus chinensis (this study); F, Eurypharynx pelecanoides (Inoue et al. 2003); G, Saccopharynx sp. (Inoue et al. 2003); H, Agamid lizard (Macey et al. 2000); I, Akamata (Kumazawa et al. 1996); J, Texas blind snake (Kumazawa and Nishida 1995). Genes are depicted and labeled as in Fig. 1. Note that most of the vertebrate mitochondrial gene rearrangements including the “IQM” region have held the tRNAMet gene just before the ND2 gene (D–J), and that, even in a few exceptional cases (A–C), the tRNA pseudogene is located in that position.

Full size image

General Features and Origin of the tRNAMet Pseudogene

The putative tRNAMet pseudogenes found in scarid fishes were aligned as in Fig. 3. They varied in sequence lengths (53–69 bp) among species. Regression of their pairwise sequence divergence estimates (Jukes–Cantor distance) on those of the tRNAMet gene demonstrated that the pseudogenes have evolved about two times faster than the tRNAMet genes (Fig. 7). Most of the pseudogenes retained clover-leaf secondary structures (Figs. 4B–H), yet anticodon sequences were mostly degenerated (Figs. 3 and 4), indicating their losses of tRNA function.

Figure 7
figure 7

Regression of pairwise sequence divergence estimates of the tRNAMet pseudogene on those of the tRNAMet gene for the 11 scarid species. Sequence divergences were calculated using the Jukes–Cantor (1969) model. A single plot corresponding to pairwise comparison between Calotomus japonicus and Leptoscarus vaigiensis was eliminated here, because their divergence estimate for the pseudogene was more than 1.0. Note that almost all plots are located above the diagonal line of Y = X, indicating that the tRNA pseudogenes have evolved more rapidly than the tRNA genes.

Full size image

NJ analysis of these pseudogenes and tRNAMet genes from various scarid and allied fishes resulted in the tree shown in Fig. 8. Although bootstrap values were mostly low owing to short sequence lengths, the resultant NJ tree demonstrated that the scarid tRNAMet genes and pseudogenes were reciprocally monophyletic and the two clades formed a sister group. Such relationships among the tRNA genes and pseudogenes were reproduced in all of the 5487 equally most-parsimonious trees (tree length = 233). Both of these results indicated that the duplication of tRNAMet gene occurred before diversification of the scarid species.

Figure 8
figure 8

Neighbor-joining tree for the tRNAMet genes (M) and pseudogenes (ψM) from the 11 scarid, 20 labrid, single odacid, and two outgroup species. Evolutionary distances are calculated using the Jukes–Cantor (1969) model. Values along branches are the bootstrap probabilities from 500 replications (only values ≥50% are shown). The sequence of the tRNAMet gene from Chlorurus sordidus is identical to those from Scarus forsteni, S. ghobban, and S. niger, and the sequence of the tRNAMet pseudogene from S. guacamaia is identical to that from S. ghobban. Note that the scarid tRNAMet pseudogenes and genes are reciprocally monophyletic, and the two clades form a sister group.

Full size image

In both of the NJ and MP trees, topologies within clades of the scarid tRNAMet genes and pseudogenes were not identical to each other, probably because of insufficient phylogenetic information owing to short sequence length and conservative nature of tRNA genes. It is, however, notable that, in the NJ tree, pseudogenes replicated the identical intergeneric relationships to that proposed by Streelman et al. (2002), except for the position of Leptoscarus. It should be noted that, in both of the NJ and MP trees, Pseudocheilinus octotaenia was an immediate sister taxon of parrotfish rather than Pseudodax moluccanus, which have been inferred to be an immediate sister group to parrotfish based on osteological data (Gomon 1979).

Possible Mechanisms far the Scarid Gene Rearrangement

Because the scarid gene rearrangement in the IQM region was not found in the labroid fish other than scarids, it is reasonable to assume that the ancestral scarid gene order is identical to the typical vertebrate gene order. Gene rearrangements have been proposed as occurring by tandem duplication of gene regions as a result of slipped-strand mispairing, followed by deletions of redundant genes (Levinson and Gutman 1987; Moritz and Brown 1987). The present gene order and associated pseudogenes can be parsimoniously explained by such process as follows (Fig. 9): tandem duplication occurred in the IQM region, followed by deletions of the redundant first copy of Q and second copy of I. If this is the case, the intergenic spacer located between Q and the ND2 genes corresponds to the second copy of M. Actually, as described above, the sequence of the intergenic spacer was most similar to that of the tRNAMet gene among the other 21 tRNA genes constituting the mitogenome of Chlorurus sordidus. In addition, all the examined scarid species had short vestigial sequences (5–15 bp) between I and M and between M and Q genes (e.g., Fig. 2), while other comparative labroid species had no (few, if any) inserted nucleotide(s) between these tRNA genes. These facts together support the above explanation based on the tandem duplications of gene regions followed by deletion of redundant genes.

Figure 9
figure 9

Proposed mechanisms of tRNA gene rearrangements in scarid species under a model of tandem duplication of gene regions and subsequent gene deletions. Genes are depicted and labeled as in Fig. 1, and genes encoded in the L-strand are underlined. After tandem duplication in the IQM region is produced, the redundant first copy of Q and second copy of I are deleted. Note that the region between Q and ND2 for the gene-rearranged sequence (ψM) corresponds to the degenerated tRNAMet gene (M).

Full size image

Probable Function of the tRNAMet Pseudogene

Using a molecular clock, Streelman et al. (2002) estimated that two major lineages within the scarids (reef and seagrass clades) diverged approximately 42 Myr ago (Fig 5). On the other hand, the first fossil scarids (genus Calotomus) were recorded from Middle Miocene (approximately 14 Myr ago) of Austria (Bellwood and Schultz 1991) (Fig. 5). The origin of scarids, thus, goes back to at least 14 Myr ago. As discussed above, the gene duplication of tRNAMet gene is assumed to have occurred in the common ancestor of the family, which leads to the conclusion that the tRNAMet pseudogene has been a stable component of the scarid mitogenomes during at least 14 Myr. It is, however, surprising that the tRNA pseudogenes lacking tRNA function have been sustained during such a long period, because animal mitochondrial genomic size is usually minimized over time (Rand 1993). We suggest that the scarid tRNA pseudogenes have not been eliminated because they retain some function(s) within the mitogenome.

All of the scarid tRNA pseudogenes flanked the 5′ ends of the ND2 genes. As mentioned above, tRNA pseudogenes found in Diaphus splendidus and Carapus bermudensis are also located at just upstream of the ND2 gene (Figs. 6A and B). These facts seem to indicate that their locations are associated with their function. In a sea urchin, a tRNA gene lost its tRNA function and became part of a flanking protein-coding gene (Cantatore et al. 1987). But this is not the case here, because the present pseudogenes do not have a start codon at the 5′ end and sometimes have a stop codon(s) inside the pseudogene.

In the vertebrate mitogenome, genes for tRNAs flank nearly all protein genes and the two rRNA genes (e.g., Fig. 1). Such unique genetic organization has led to the proposal that the secondary structures of the tRNA sequences provide the punctuation marks for proper processing of the mitochondrial primary transcript (Ojala et al. 1981). This proposal has been supported by identifications of key enzymatic activities (i.e., precise cleavage at the 5′ and 3′ ends of tRNAs) in the human mitochondrial system (Rossmanith et al. 1995). Furthermore, maternally transmitted human diseases have been observed which were caused by mutations in mitochondrial tRNAs, due to end processing defects (King et al. 1992; Bindoff et al. 1993; Rossmanith and Karwan 1998; Levinger et al. 2001, 2003).

Interestingly, most of the vertebrate mitochondrial gene rearrangements, including the IQM region, have held the tRNAMet gene just upstream of the ND2 gene (Figs. 6D–J), and even in a few exceptional cases, including the present ones, the tRNA pseudogenes are located in that position (Figs. 6A–C). This fact indicates that the ND2 gene favors or requires the tRNA gene or its substitute at the 5′ end. On the other hand, most of the scarid tRNA pseudogenes flanking the 5′ end of the ND2 gene retained clover-leaf secondary structures (Figs. 4B–H), inspite of the loss of function as a tRNA and relatively rapid sequence evolution (Fig. 7). Considering these features together, these tRNA pseudogenes are likely to function as punctuation marks for precise processing of the 5′ end of the ND2 mRNA. RNA end processing reactions require only the “top half” of the clover-leaf secondary structure (see Fig. 4A) (Maizels and Weiner 1999). Therefore, if these pseudogenes function only as punctuation marks, the “bottom half” of the secondary structure (see Fig. 4A) should be less important in its function. Interestingly, the tRNA pseudogenes from the two scarid fish, Nicholsina usta and Sparisoma chrysopterum, retained the top halves of the secondary structures, while the bottom halves were largely degenerated (Figs. 4G, H). Such tendencies on secondary structures are more conspicuous in the putative tRNAMet pseudogenes from D. splendidus, C. bermudensis, and Bregmoceros nectabanus (Figs. 4I–K). These facts agree with the above reasoning on secondary structures and support the idea that the scarid tRNA pseudogenes have a function as punctuation marks. To confirm these ideas, an experimental approach which investigates whether these pseudogenes are actually recognized by proper endonucleases will be effective.

Provided that the tRNAMet pseudogenes function as punctuation marks, one problem arises as follows. The pseudogene flanks a noncoding strand of the tRNAGln at the 5′ end (Figs. 1 and 2), and the noncoding strand can form clover-leaf-like secondary structures. If the noncoding strand could serve as proper punctuation marks, processing of the 5′ end of the ND2 mRNA would be successfully completed without the tRNAMet pseudogene. In such a situation, why is the tRNAMet pseudogene not deleted from this location? Two plausible explanations of this problem can be proposed. (1) The noncoding strand of tRNAGln may not serve as proper punctuation marks in the first place. Macey et al. (1998) presented the possibility that a noncoding strand of tRNAPro gene could not signal precise processing of the primary transcript, where tandem duplication of the tRNAThr–tRNAPro gene cluster occurred and the first copy of tRNAThr has been sustained between an incomplete stop codon of the cytochrome b gene and the noncoding strand of the tRNAPro gene. The first copy of the tRNAThr gene was thought to be maintained there, because the noncoding strand of the tRNAPro gene cannot signal the precise processing of the 3′ end of the primary transcript of the cytochrome b gene, and the successful processing of it cannot be done without the first copy of the tRNAThr gene. (2) The tRNAMet pseudogene may not be eliminated because it cannot pass through the halfway situation to complete deletion: When the pseudogene becomes nonfunctional and fails to be processed out of the primary transcript, the ND2 mRNA would have to have many extra nucleotides before the start codon, which might prevent the mRNA from being properly translated. In other words, the tRNAMet pseudogene is sustained simply because it is difficult to eliminate without translation error. In a codlet, Bregmaceros nectabanus, a probable tRNAMet pseudogene is inserted between the tRNAIle and the ND2 genes (Fig. 6C), and can fold into the secondary structure corresponding to the top half of the clover-leaf structures (Fig. 4K). Because the flanking tRNAIle gene is encoded by the same strand as the ND2 gene, the inserted pseudogene should not be indispensable for precise processing of the 5′ end of the ND2 mRNA. This case seems to support the second explanation.