Results and discussion

- The complete mitochondrial genome of a basal teleost, the Asian arowana (Scleropages formosus, Osteoglossidae)

The complete mitochondrial genome of Asian arowana was sequenced with shotgun sequencing method (min. 6X, average 9X coverage). Its total size was found to be ca. 16,651 bp [GenBank:DQ023143]. Except the mitochondrial control region the size of Asian arowana mitochondrial genome was found to be similar to that of silver arowana, butterfly fish and goldeneye [9] [see Additional file 1 for the exact sizes]. The GC content of Asian arowana mitochondrial genome was 46.1%, the highest among mitochondrial genomes of all Osteoglossiformes available in Genbank (silver arowana – 43%, butterfly fish – 39% and goldeneye – 42%).

On the whole, the structure of the Asian arowana mitochondrial genome is very similar to that of silver arowana, butterfly fish and bichir [see Additional file 2]. The number and order of genes in the Asian arowana mitogenome [see Additional file 3] were found to be the same as common vertebrate form [1]. It contains 24 RNA and 13 protein-coding genes: 7 subunits of the NADH ubiquinone oxidoreductase complex (nad1-6 and nad4L), 3 subunits of the cytochrome c oxidase complex (cox1-3), a single subunit of the ubiquinol cytochrome c oxidoreductase complex (cob), 2 subunits of ATPase (atp6 and atp6), 2 ribosomal RNA (rrnL and rrnS) and 22 transfer RNA (trn) genes. The non-coding control regions situated between the trnP and trnF genes contain the heavy strand origin of replication (O_H). A smaller control region containing the putative light strand origin of replication (O_L) was found between trnW and trnY genes.

Eleven potential overlaps between genes have been observed in the Asian arowana mitogenome. The longest one (10 bp) involving the two ATPase genes appears to be common in most vertebrate mitochondrial genome, and its size in fish (7–10 bp [16]) is smaller than that in mammals (40–46 bp; [2]). The second largest overlap is 7 bp long, (between nad4 and nad4L genes), whereas the remaining nine were in the size range of 1–5 bp.

The Asian arowana mtDNA's heavy strand control region, also known as D-loop, contains O_H and is ca 980 bp long. Similar to typical vertebrate mitogenomes, this non-coding region can be divided into three different domains [17,18] (Figure 1A). Domain I which is 400 bp long, consists of a termination associated sequence (TAS: TACATAAATTG) [19] and several copies of a previously described conserved palindromic motif without any known function [20]. A 37 bp tandem repeat array, suggested to be involved in the regulation of mitochondrial genome replication by forming a thermostable "hairpin" [21], was also found in this domain (see next section for details). Domain II – commonly known as the central conserved block – covering the 401–641 bp stretch in the control region, showed high similarity to domain II of rainbow trout [22] and sturgeon [21] (data not shown). In domain III, a TA-dinucleotide microsatellite repeat was present in all the six individuals from which the control region was sequenced. Two conserved sequence blocks (CSB1; 724–742 bp and CSB3; 813–839 bp) found in this domain showed high similarity to CSBs detected earlier in other species [23] whereas CSB2 described earlier in teleosts [24] was not found.

A smaller control region (34 bp) for O_L exhibited high sequence similarity to the corresponding region in silver arowana, bichir and butterfly fish (data not shown). The AT content of Asian arowana O_L which was 35.3%, is similar to that of butterfly fish but higher than in silver arowana (31.4%) and much lower than in bichir (44%). The secondary structure of O_L was suggested earlier to regulate light strand replication [25]. In Asian arowana this secondary structure consists of a perfect 9 bp stem (CCTCCCGCC/GGAGGGCGG) and loop structure. Despite the fact that the control region is the most variable region in animal mtDNAs, most part of the stem (TCCCGCC and AGGCGGA) was found to be conserved in the mitogenomic O_L of several fish species (including the Asian arowana) and even mammalian ones [26].

The mtDNA of all six Asian arowana individuals tested possess a heteroplasmic tandem repeat array in domain I (Figure 1B). The tandem repeat arrays in the six individual fishes sequenced contained 3 to 6 repeat units, resulting in variable length of the heavy strand control region (976 to 1094 bp long). A partial repeat unit could also be found at the beginning and at the end of the array indicating that it might have been formed by replication slippage [21,27].

The tandem repeat units were highly similar with only a few base substitutions (Figure 1B). Each repeat unit in the array was 37 bp long (TACATATTATGCATAATCATGCATATATATGTACTAG). The conserved motif TACAT (previously described only in mammals [28] and lungfish [29]) and its complement ATGTA, were both located in the stem region providing the theoretical ability of forming a stable hairpin loop (Figure 2A). Our investigation of the other three members of Osteoglossiformes with fully sequenced mitogenome (i.e. silver arowana, butterfly fish and goldeneye) has shown that this conserved motif could also be found in a similar arrangement in their heavy strand control region (Figure 2A). Further investigation revealed that the two motifs could also be found in the heavy strand control region of several eel species (Anguilliformes) (Figure 2B). The conservation of this motif across various vertebrate taxa suggests that it serves an important role in the mitochondrial heavy strand control region. An extensive search of the literature database showed that since it was reported more than a decade ago, no extensive study was published to investigate its function. Based on the position of the motif, we speculate that it might be required for the formation of a thermostable hairpin involved in replication of the tandem repeat array. We also cannot rule out the possibility of the motif being binding sites for proteins involved in replication.

Another type of repeat – a TA-type dinucleotide microsatellite – was present at the opposite end of the heavy strand control region in domain III (Figure 1B). The number of TA core units ranged from 8 to 10 in the six individuals sequenced. Although both tandem repeats alone (e.g [30-33]) or microsatellites in the tandem repeat array [34] has been reported earlier in the heavy strand control regions of some species, to our knowledge no one has reported the existence of both types of repeats on the same heavy strand control region.

The start codon usage in the Asian arowana mitogenome was found to be the same as that of zebrafish [16]. All but one of the 13 protein coding genes began with the orthodox ATG start codon, only cox1 used GTG start codon [see Additional file 2]. Ten genes ended in a complete termination codon, either TAA, TAG or AGA. The remaining three genes (cox2, nad4 and cob) did not possess a complete stop codon, but did show a terminal T. This condition is known to be common to vertebrate mitochondrial genomes whereby post-transcriptional polyadenylation provides the two adenosine residues required for generating the TAA stop codon [35].

The total nucleotide length for the 13 coding genes was found to be 11,403 bp, shorter than that of silver arowana and butterfly fish, but longer than that of bichir [see Additional file 2 for the exact sizes]. The coding sequences in Asian arowana consisted of 28.0% A, 25.1% T, 14.8% G and 32.1% C bases. The corresponding ranges for silver arowana, butterfly fish and bichir were 29.2–30.4% (A), 27.9–31.3% (T), 13.4–14.2% (G), and 24.3–28.8% (C). These data further support the observations: i) the GC content of Asian arowana mitogenome is higher than that of other teleost, including other known Osteoglossoidei species; and ii) the frequency of G is the lowest among the four bases in fish mitochondrial genomes [2].

Comparison of amino acid sequences for the 13 proteins among Asian arowana, silver arowana, butterfly fish and bichir confirmed the closer taxonomic relatedness of Asian arowana to silver arowana, than to butterfly fish or bichir (Table 1). In agreement with others' data [2,36], cox1 was the most conserved gene and atp8 was the most variable.

The pattern of codon usage in Asian arowana mtDNA was also studied. The most frequently used amino acids were leucine (16.9%), followed by threonine (8.9%), alanine (8.4%) and glycine (7.8%) [see Additional file 4]. At the third codon position, the order of nucleotide usage frequency was C > A > T > G (Figure 3), the same order was described earlier for the mitochondrial genome of Japanese fugu [37]. The order was somewhat different in the silver arowana, butterfly fish and bichir, where A became the most frequently used base in the third codon position, albeit the frequency of G remained the lowest (Figure 3).

For amino acids with fourfold degenerate third codon position, codons ending with A were the most frequent (42.7%), followed by C (36.5%), T (14.1%) and G (6.2%). Genes located on the heavy strand showed a typical native GC and positive AT skew (Figure 4), whereas the nad6 gene located on the light strand displayed an opposite pattern. With regard to the absolute value, the GC skew was always higher than the AT skew: the former ranged from 0.60 to 1, whereas the latter ranged from 0.33 to 0.72 (Figure 4). Similar patterns were also seen in silver arowana, butterfly fish and bichir (data not shown). The GC and AT skews in Asian arowana were not correlated (R = 0.094, P > 0.05).

The twenty-two tRNA genes typical of vertebrate mitochondrial genomes have all been detected in the Asian arowana mitogenome. All tRNA genes possessed anticodons that match the vertebrate mitochondrial genetic code. The length of tRNA genes ranged from 67 bp to 74 bp [see Additional file 3] with a total length of 1,550 bp, similar to silver arowana and bichir, but shorter than that in butterfly fish [see Additional file 2]. The inferred secondary structure of the 22 tRNA genes had several uniform features: 7 bp in the aminoacyl stem, 5 bp in the TC and anticodon stem, 4 bp in the DHU stem and 7 bp in the anticodon loop. A "U" residue before the anticodon was found in 19 of the 22 tRNA, whereas a purine was detected in the position immediately 3' to the anticodon. In the stem regions, there were several non-complementary pairings, mainly A-C type. A similar structure has been found in the silver arowana, whereas different kinds of non-standard base pairings have also been described in other fish species [2]. The original sequences and the secondary structure of the tRNA genes were quite different in genetically distant related species.

 

Like the mitochondrial genome of other fishes, the Asian arowana mitogenome was found to possess two ribosomal RNA (rRNA) genes, a small rRNA gene (rrnS) and a large rRNA gene (rrnL), the two being separated by trnV [see Additional file 3]. The length of rrnS and rrnL are 956 and 1,698 bp, respectively [see Additional file 3]. These sizes are similar to those in the other three species used for comparison [see Additional file 2]. Substitution rates of the two rRNAs among Asian arowana, butterfly fish and bichir were lower than those of protein coding genes. Secondary structures found in the four species seemed to be conserved across large evolutionary distances, as described earlier for teleosts [2].

 

Several studies have been published recently on the phylogeny of Osteoglossoidei suborder using morphological data [11], partial mitochondrial sequences [38], a few nuclear genes [39] or the combination of the latter two [40]. On the other hand, there is only a single study that analysed the phylogenetic relationship of the osteoglossids based on all genes present in the mitochondrial genomes [41] but the Asian arowana was not included as its complete mitogenomic sequence was not available.

To determine whether the addition of the complete Asian arowana mitogenome causes any difference in the evolutionary position of the Osteoglossomorpha from the cladograms produced earlier [11,38-40], we used the complete Asian arowana mitogenome sequence obtained in this study and other osteoglossids' complete mitogenome sequences to carry out phylogenetic analysis. Beside the osteoglossid species our analysis also included mtDNA from three fish species of ancestral lineages: four members of the Chondrostei taxon and representatives for both the Elopomorpha and Clupeocephala taxa (for complete list of species used, refer to Additional file 1). Using both nucleotide and amino acid sequences of different kinds of mitochondrial genes (see Materials and Methods for details) the systematic arrangements were reconstructed as monophyletic which is in agreement with the relationship tree of basal Actinopterygians produced by Inoue and colleagues [41].

Phylogenetic trees constructed with the various data sets using three different methods (i.e. MP, BI and ML) showed little variations within the data set, indicating that variation mainly originated from the type of data and not the methods used (data not shown). On all trees the Asian arowana was clustered into one group with the silver arowana, butterfly fish and goldeneye (all three from the Osteoglossomorpha superorder) with a high bootstrap support value. However, within Teleostei taxon the position of Osteoglossomorpha clade varied in the trees generated using concatenated protein-coding cum tRNA nucleotide sequences and concatenated protein-coding cum tRNA cum rRNA nucleotide sequences data sets. On the other hand, trees generated using concatenated protein-coding nucleotide sequences and concatenated amino acid sequences data sets consistently placed the Osteoglossomorpha clade at the basal level (Figure 5). This is in agreement with trees constructed earlier by others using various molecular [9,38-41] and morphological data [11]. In addition the proximity of Osteoglossomorpha clade to that of basal teleost clades in our study further supports the position of osteoglossids among the early branches of living teleosts' stem lineages (see e.g. [41])

The placement of goldeneye and butterfly fish was different in osteoglossid cladograms produced earlier on the basis of morphological data [41-46]. While most publications predicted that during the evolution of osteoglossids the ancestor of goldeneye split off earlier from the arowanas, than from the butterfly fish [42-45]. One study proposed exactly the opposite [46]. Our cladogram based on full mtDNA sequences similarly to the data from [41] from four osteoglossids supports the former situation (Figure 5).

Since goldeneye is the only complete mtDNA sequence reported for Notopteroidei suborder, additional full mitogenomic sequences from this taxonomic group will have to be obtained for a more detailed analysis.