We have previously reported 19 the identification of 285 D. melanogaster nuclear genes encoding mitochondrial proteins that represent the counterparts of human peptides annotated in the Swiss-Prot database as mitochondrial 20. On the basis of comparative evidence obtained by BLASTP analysis, 78 of these genes are involved in the OXPHOS system, encoding 66 proteins known to be components of the five large respiratory complexes and 12 proteins involved in oxidative phosphorylation as accessory proteins. To identify the putative counterparts of the D. melanogaster OXPHOS genes in D. pseudoobscura and A. gambiae we performed a TBLASTN search 1321 on the whole genome sequences of these species using the amino-acid sequences of the 78 D. melanogaster peptides as queries. Sequences giving the best reciprocal BLAST hits were tentatively assumed to identify functional counterparts in two species if they could be aligned over at least 60% of the gene length and the BLAST E-score was less than 10^-30. By these criteria, all the 78 D. melanogaster OXPHOS genes investigated have a counterpart both in D. pseudoobscura and in A. gambiae. To better compare the structure of the OXPHOS genes in the three dipteran species, we used the predicted coding sequences as queries for a search of expressed sequence tags (EST) 21, and used the retrieved sequences to annotate the transcribed noncoding sequences of the A. gambiae genes investigated. Although little EST information is available for D. pseudoobscura, it was still possible to predict unambiguously the exon-intron gene structure of the OXPHOS genes in this species, as well as the amino-acid sequence of their full-length products, by exploiting the high level of similarity with D. melanogaster. The results of BLAST analysis, together with the construction of phylogenetic trees that also include other genes that show lesser but still significant sequence similarity to the 78 genes assumed to be one-to-one orthologs in the three species investigated (see below), strongly suggest that the newly identified D. pseudoobscura and A. gambiae genes are the functional counterparts of the 78 D. melanogaster genes used as probes.

Table 1 lists the 78 putative orthologous OXPHOS genes in the three dipteran genomes and their cytological location. For each gene, a record showing the gene map and reporting the annotated genomic sequences as well as the mRNA and protein sequences is available and can be queried at the MitoComp website 22 (see also Additional data files). MitoComp also compares the structure of the D. melanogaster, D. pseudoobscura and A. gambiae putative orthologous genes and their duplications when present (see below), and aligns the orthologous coding sequences (CDS), and also aligns their deduced amino-acid products with the corresponding human protein.

For the products of the OXPHOS genes investigated, the D. melanogaster/D. pseudoobscura average amino-acid sequence identity is 88%, compared to 64% between D. melanogaster and A. gambiae. Figure 1 shows the frequency distribution of sequence identities, and Additional data file 1 lists all pairwise identity values between the products of the 78 OXPHOS genes when orthologous D. melanogaster/D. pseudoobscura, D. melanogaster/A. gambiae and D. melanogaster/human gene products are compared. A multiple alignment of each cluster of homologous proteins is shown at the MitoComp website 22.

It should be kept in mind that identity values reported in Figure 1 and in the table in Additional data file 1 were calculated on the whole sequence of the predicted unprocessed proteins; they are much higher if the putative amino-terminal pre-sequences are excluded, since such sequences, possessed by most mitochondrion-targeted products, show little amino-acid sequence conservation 2324, although they do share specific physicochemical properties 2526. When only the predicted mature protein is considered, the average percentage identity increases to 90% between D. melanogaster and D. pseudoobscura, and to 70% between D. melanogaster and A. gambiae.

A striking example of evolutionary conservation is provided by the genes encoding cytochrome c (an essential and ubiquitous protein found in all organisms) in the three dipteran species: the amino-acid sequences of the gene products are identical in D. melanogaster and D. pseudoobscura, whereas 96% identity is preserved between Drosophila and Anopheles. Coding sequences are also extremely conserved, suggesting that the nucleotide sequence itself is subject to strong evolutionary constraints, maybe due to codon usage bias. Only synonymous substitutions (21 out of 108 codons) were found on comparing D. melanogaster and D. pseudoobscura cytochrome c coding sequences, whereas 28 synonymous substitutions and only four nonsynonymous substitutions were observed between D. melanogaster and A. gambiae (see MitoComp website 22).

It is well known that a given function may be supplied in different species by genes that are not directly derived from a common ancestor, that is, by paralogous, not orthologous, genes. Therefore, we thought it would be interesting to compare the structural organization of the OXPHOS genes in the three species investigated, on the principle that it should be possible to infer derivation from a common ancestor, that is, 'structural orthology', if an identical or very similar overall structure was preserved. As the introns of the putative orthologous OXPHOS genes in the three species are, as expected, too divergent in DNA sequence to be aligned, we used conservation of number of introns, conservation of their location in the coding sequence, and preservation of the reading frame with respect to the flanking exons as our primary criteria.

With the only exception of Dpse\CG5037, putatively encoding protoheme IX farnesyltransferase, whose 5' genomic sequence was impossible to find in the relevant contig assembly, all other investigated D. pseudoobscura genes show a structural organization almost identical to that of their D. melanogaster counterparts. Of the 78 Anopheles genes studied, 39 maintain the structural organization observed in Drosophila, whereas gain or loss of introns occurred in 33, and in six the location of introns is not preserved at all. In agreement with a previous report 4, the intron-exon structure of the gene appears to be conserved in all three dipteran species when splicing of alternative coding exons occurs: the alternative splice forms of both the Drosophila NADH-ubiquinone oxidoreductase acyl carrier protein (mtacp1, CG9160) 27 and the Drosophila ATP synthase epsilon chain (sun, CG9032) 19 have very similar counterparts in Anopheles, as shown by genomic structure comparison, alignment of splice variants and EST mapping (Figure 2).

Genes encoding the acyl carrier protein (mtacp1) in the three species are characterized by the mutually exclusive use of homologous exons that are repeated in tandem (Figure 2a). The duplicate exons occur at the same location in the aligned amino-acid sequences, and are flanked on both sides by a phase 1 intron. When the sequences of the duplicated exons are compared, they show the expected divergence pattern (that is, the similarity between duplicate exons within a gene is less than the similarity of each exon to its equivalent in the orthologous gene). Evidence from genomic and transcribed sequences (GenBank accession numbers BI510891 and BI508135) shows that the duplicated mtacp1exons are also preserved in the more distantly related insect Apis mellifera (honeybee) (Figure 2c,d), indicating a specific adaptive benefit for this gene structure, as also suggested by the evolutionary convergence leading to the occurrence of alternative splicing in members of three different ion-channel gene families from Drosophila to humans 28. However, there is no evidence from ESTs that duplicated mtacp1 exons undergo alternative splicing in vertebrates and nematodes.

Interspecies comparison of the introns of putative orthologous genes indicates that there is little constraint on their nucleotide sequence, which undergoes nucleotide substitutions at a rate comparable to that of pseudogenes 29. However, several observations suggest that intron size is subject to natural selection. For example, in D. melanogaster and several other organisms the distribution of intron length has been shown to be asymmetrical, with a large group of introns falling into a narrow distribution around a 'minimal' length and the remaining showing a much broader length distribution, ranging from hundreds to thousands of base-pairs 303132.

Of the introns that interrupt the coding sequence in the 78 OXPHOS genes investigated in the present study, 88 (64.7%) of 136 in D. melanogaster, 96 (70.5%) of 136 in D. pseudoobscura and 87 (67.9%) of 128 in A. gambiae fall into the short-size class (Figure 3a). However, in A. gambiae the length distribution of these introns appears slightly broader (62-150 bp, compared with 51-100 bp in both Drosophila species). The remaining introns show a broad length distribution, ranging from 151 to 4,702 bp with no clear boundary between classes.

A comparison of the length of introns in corresponding positions in the putative D. melanogaster, D. pseudoobscura and A. gambiae orthologs suggests that changes from the short-size to the long-size (more than 300 bp) intron class, or the converse, have been rare in the evolutionary history of these species: only seven class changes were observed comparing D. melanogaster and D. pseudoobscura introns, and six between D. melanogaster and A. gambiae (Figure 3b). On the whole, our data confirm the highly asymmetrical intron length distribution in D. melanogaster and extend this finding to the introns of the D. pseudoobscura and A. gambiae OXPHOS genes.

It is generally accepted that gene duplication is the basic process that underlies the diversification of genes and the origination of novel gene functions 33; however, many features of this process are still elusive. To obtain more information on the molecular evolution of the genes involved in the OXPHOS system, we searched the genomes of D. melanogaster, D. pseudoobscura and A. gambiae for duplications of the 78 OXPHOS genes whose orthologs we have identified in the three species.

Duplicate gene pairs were tentatively identified within each genome as best reciprocal hits with an E-value of less than 10^-20 in both directions in a TBLASTN search using the default parameters. Deciding whether two proteins may be considered homologous becomes difficult when their sequence identity is within the 20-30% range (the so-called 'twilight zone' 34), and so the following additional criteria were used: first, the two sequences could be aligned over more than 60% of their length; second, the putative processed proteins encoded had to have more than 40% identity; and third, amino-acid percentage similarity had to be larger than percentage identity 35. Even if meeting these criteria and reported as different genes in the ENSEMBL database 36, identical Anopheles nucleotide sequences were excluded from further analysis, as they are likely to reflect annotation artifacts.

Duplications, or in some instances triplications, of 24 OXPHOS genes were found. Overall, we identified 47 genes (20 in D. melanogaster, 19 in D. pseudoobscura and eight in A. gambiae) each of which shows significant similarity with one of the 78 OXPHOS genes reported above. When the structure of a member of a paralogous gene set indicates that it has been produced by retroposition, it seems reasonable to assume that it is derived from a pre-existing 'parent' gene. For duplicates not clearly originating by retroposition, we also assume, on the basis of the much higher level of conservation and expression, that the genes we find to be the structural orthologs in all three species are the parent ones, and in this case also we will henceforth refer to their paralogs as OXPHOS gene duplicates. The amino-acid percentage identity between the products of duplicate gene pairs ranges from 40% to 85%. For each of the OXPHOS gene duplicates, cytological localization, number of exons interrupting the coding sequence, and number of ESTs found in the D. melanogaster and A. gambiae EST databases are reported in Table 2. Neighbor-joining trees derived from distance matrix analysis and showing the inferred evolutionary relationship between members of each gene cluster are available at the MitoComp website 22.