The arthropod muscle myosin heavy chain genes were identified by TBLASTN searches against the corresponding genome data of the different species using the Drosophila melanogaster protein as query (Figure 1, see Additional file 1). The species analysed were the mosquitos Aedes aegypti, Culex pipiens quinquefasciatus and Anopheles gambiae, the silkworm Bombyx mori, the honeybee Apis mellifera, the jewel wasp Nasonia vitripennis, the waterflea Daphnia pulex, the rust-red flour beetle Tribolium castaneum, the body louse Pediculus humanus corporis, and thirteen Drosophila species (Table 1). According to the general nomenclature for myosin sequences 12 the alternatively spliced muscle myosin heavy chain genes are named Mhc1, and the non-muscle myosin heavy chain genes are denoted Mhc2. The sequences were assigned by manual inspection of the genomic DNA sequences. Exons have been confirmed by the identification of flanking consensus intron-exon splice junction donor and acceptor sequences (Figure 1) 30. Because of the five to nine clusters of mutually exclusive exons and the included or excluded penultimate exon, automatic identification of all exons failed. The genomic sequences of Apis mellifera and Bombyx mori contain several gaps that at least in one case must have contained missing exons. The expression of the myosin genes including the transcription of some of the mutually exclusive exons has been confirmed by analysis of corresponding EST data.

The untranslated first exons of the genes have been assigned by analysing EST data, if possible. Because untranslated 5' exons were found for all those species for which EST data covering the amino-termini of the genes is available, it is expected that the other arthropod myosin genes also contain untranslated first exons. Accordingly, the unambiguously identified exons have been numbered starting with exon two. Duplicated exons were named in alphabetical order according to the direction of transcription, the exception being the alternatively spliced exon 11 of the Drosophila Mhc1 of which the first of the mutually exclusive exons was named 11e for historical reasons 9. The differentially included penultimate exons of the Drosophila species have been predicted based on their similarity at the DNA level. Although this exon mainly consists of untranslated bases and its identity between the Drosophila species is almost as low as that found in intron regions, the exon borders are conserved enough to be recognised. The carboxy-terminal exons of the other arthropod Mhc1 genes have been confirmed by analysing EST data, if possible. For TicMhc1 and DapMhc1 only one carboxy-terminal exon could be confirmed by EST data. However, given the exon conservation between all arthropod Mhc1 genes it is expected that both genes contain another carboxy-terminal exon. For Nasonia, EST data is not available. The carboxy-terminal exon of the NavMhc1 gene was identified based on its homology to the other Mhc1 exons. An exon corresponding to the penultimate exon of the other genes could not be identified.

The Drosophila sp. Mhc1 genes, the AeaMhc1 and the CpqMhc1 gene contain consensus polyadenylation signals AATAAA, while the Mhc1 genes of Ang, Am, Dap, Nav, Pdc, and Tic contain polyadenylation signals of type AAAAAA. For the DmMhc1 gene it has been shown that the use of either polyadenylation site is not regulated 3132 and the same might be true for the two or multiple polyadenylation sites of the other arthropod genes.

Surprisingly, a second muscle myosin heavy chain gene has been identified in Aedes aegypti (Figure 1) and named Mhc3. The Mhc3 gene contains the same exon organisation as Mhc1 except that it does not have any cluster of alternatively spliced exons and misses the two carboxy-terminal exons (Figure 1). Many EST clones provide supporting evidence for the deduced carboxy-terminus, the amino-terminal untranslated exon1, and other parts of the gene. The exons related to the alternatively spliced exons of Mhc1 are either identical ("exon3b") or very similar to one of the Mhc1 exons. The protein sequence of Mhc3 has an overall sequence identity of 91.4% to Mhc1. Besides the different carboxy-termini, the largest differences are in loop-1, which is three residues shorter in Mhc3, and in loop-2, which has only six instead of ten glycines and might therefore be structurally more restricted. The Culex pipiens quinquefasciatus genome encodes another two muscle myosin heavy chain genes that are very similar to each other and have been named Mhc3 and Mhc4 (Figure 1). Both have the same exon organisation as the CpqMhc1 gene except that they do not have any cluster of alternatively spliced exons and miss the two carboxy-terminal exons. Another difference is that alternative exons 8 are fused to the following constitutive exons in the Mhc3 and Mhc4 genes. The protein sequence identity between CpqMhc3 and CpqMhc4 is 92.0%, the identity to CpqMhc1 is 84.4% and 90.4%, respectively. Surprisingly, AeaMhc3, CpqMhc3 and CpqMhc4 retained identical variants of the alternatively spliced exons of the corresponding Mhc1 genes.

The analysis of the BmMhc1, TicMhc1, PdcMhc1, and DapMhc1 genes revealed further clusters of alternatively spliced exons compared to the DmMhc1 gene. All further sets of alternative exons encode for sequence that is part of the motor domain. The additional alternative exon of Bm, Pdc and Tic is conserved between these three organisms, and is also encoded within the Dap Mhc1 gene. It is located between the alternatively spliced exons 11 and 17 (Bm), alternative exon 13 and constitutive exon 19 (Pdc), and alternative exons 12 and 16 (Tic), respectively, and separated from the neighbouring alternatively spliced exons by constitutively expressed exons (Figure 1). In contrast to the other alternatively spliced exons, these alternatively spliced exons are different in length and amino acid conservation (see Additional file 2, figure S6A). The first part of the exon encodes part of loop-2 (see below), that is a very flexible loop involved in actin-binding. In the arthropod genes it mainly consists of glycines, arginines, and lysines. Thus, the alternatively spliced exons of Bm, Tic, Pdc, and Dap encode different numbers and compositions of these residues. The second part of the alternatively spliced exon is part of the following alpha-helix and hence completely conserved in length and strongly conserved in composition. In addition to this cluster of alternatively spliced exons, the DapMhc1 gene contains three further sets of alternatively spliced exons extending its number of clusters of alternatively spliced exons to nine (compared to five in Drosophila). Alternative exon 6 encodes an alternative P-loop to loop-1 sequence, alternative exon 11 directly follows the alternative exon encoding a structural part near the ATP-binding site, and alternative exon 18 encodes an alternative version of the sequence after loop-2 (Figure 1).

The Pediculus humanus corporis Mhc1 gene contains the most reduced set of alternative exons (Figure 1). It has four sets of alternative exons each comprising two variants. However, the sequence encoding part of the converter domain, which is encoded by sets of three to five alternative exons in the other arthropod genes, has been fused to the following exon forming one constitutive exon in the PdcMhc1 gene (exon 19, Figure 1). Also, the part in the tail domain encoded by a set of two alternative exons in all other arthropod genes is represented by only one exon in the PdcMhc1 gene (exon 25). Altogether, the alternative exons encode for 16 different versions of the motor domain and 32 different mRNAs of the PdcMhc1 gene, compared to potentially 120 different combinations of alternative exons for only the motor domain of the Drosophila Mhc1 gene.

The number of variants differs between the arthropod species for many of the alternatively spliced exons (Figures 1 and 2). For the first set of alternatively spliced exons two variants have been found in all Mhc1 genes. Both differ by two absolutely conserved residues, namely the amino acids alanine and aspartate at positions 25 and 26 in the "a" variants of the exon that are substituted by serine and asparagine in the "b" variants (Figure 3). A slightly less conserved marker for the "b" variants is a cysteine at position 21. Variant 3a of the DapMhc1 is an exception as it has an additional residue at the N-terminus compared to the other Mhc1 variant "a" exons. The DapMhc1 gene encodes three clusters of alternatively spliced exons not found in the other arthropod Mhc1 genes. For all three clusters exons variant "b" is more homologous to the corresponding amino acid sequences of the other Mhc1 proteins than variant "a" (see Additional file 2, figures S2, S4, and S6B). The alternatively spliced exons of BmMhc1, DapMhc1, PdcMhc1 and TicMhc1 covering loop-2 are different in length and starting position (see Additional file 2, figure S6A). However, the "a" variants are more similar to each other than to the "b" variants and the corresponding amino acid sequences of the other Mhc1 proteins. Thus, the common ancestor of Bm, Dap, and Tic had in all probability already contained an "a" and a "b" variant. Completely conserved residues characterizing the "a" variant are a serine at the end of loop-2, a glutamate at position 3, and a leucine at position 8 of the following helix ([G/K/R 8-9]S [G/A]F [Q/M]TVS [S/A]LYR). Except for PdcMhc1, all arthropod Mhc1 genes have two variants of the mutually exclusively spliced exon in the tail (Figure 2; see also Additional file 2, figure S8). The most conserved differences between the two variants are an aspartate at position 14 in variant "b" (either an asparagine or a glutamine in variant "a") and an asparagine at position 24 (an arginine in variant "a"). In addition, at position 15 the "b" variants have a large hydrophobic residue (leucine, methionine, or phenylalanine) while the "a" variants have a small polar residue (serine or threonine). In contrast to the other Mhc1 genes, the "a" variant of DapMhc1 is closer related to the "b" variants than to the other "a" variants.

The situation is more complex for the remaining clusters of mutually exclusive exons that contain three to six variants. The exon encoding a loop-helix motif adjacent to the ATP-binding site (blue color in Figure 1) is not as conserved as the other alternatively spliced exons (Figure 2; see also Additional file 2, figure S3). Therefore, it is difficult to identify characteristic residues/motifs for the respective variants. Except for the PdcMhc1 and TicMhc1 genes all genes contain four variants. The variant with the most characteristic residues is variant "c". It is characterized by a positively charged residue at position 8 (arginine or histidine), a conserved arginine at position 21, and a conserved asparagine at position 26. None of these residues appear in any of the other variants at the respective positions. The TicMhc1, PdcMhc1, and DapMhc1 genes have lost this variant. The only strong characteristic of variant "d" is a conserved isoleucine or valine at position 20 that is found in all Mhc1 genes. Variants "a" and "b" do not contain any distinguishing residues.

The alternatively spliced exons spanning the relay helix and the relay loop are the longest and most conserved of the mutually exclusive exons (see Additional file 2, figure S5). The variability ranges from two variants in the Pediculus Mhc1 gene to six variants in the Nasonia gene (Figures 1 and 2). The least conserved part of the exon is the relay loop that is not embedded in the motor domain. In this region, characteristic residues for certain variants are found. Variant "c" is characterized by a conserved glutamine at position 49 and either a glutamine or an asparagine at position 50. A copy of this variant is present in all Mhc1 genes except that of Tic. Another conserved variant is variant "d" characterized by a glutamine at position 49 followed by a proline at position 50. This variant appears in the Mhc1 genes of Aea, Ang, Cpq, Tic, and Bm. Similar to the situation for the alternatively spliced exon at the ATP-binding site, the other variants are not conserved enough to define characteristic residues. It is thus not clear which were present in the ancient arthropod gene and which arose through exon duplication in the individual genes. Again, the DapMhc1 is the exception because its first two variants, characterized by two conserved methionines at positions 42 and 55, differ from all other variants.

The variants of the cluster of alternative exons encoding part of the converter domain also show a high degree of variability (Figure 2; see also Additional file 2, figure S7). Two of the variants have characteristic features. Variant "a" is the most conserved of the variants at the protein level having a conserved methionine at position 9 and a conserved cysteine at position 26. These residues do not appear in any of the other variants of this cluster. Variant "a" of this cluster is conserved in the Mhc1 genes of all species and therefore must have been present in their common ancestor. The last of the variants has a characteristic feature at the DNA level. The intron following the last variant always has a GT 5' splice site. This is in contrast to all other variants of this exon whose following introns have a GC 5' splice site. At the amino acid level this variant is characterized by a lysine at position 2, a cysteine at position 5 and a glutamate at position 20.

Wherever EST and/or cDNA data was available a differentially excluded penultimate exon could be identified. These exons are very short (one to thirteen residues) and not conserved (see Additional file 2, figure S9), and therefore similar exons have not been predicted for the species for which EST data is not available. For Ang three carboxy-termini have been identified. Based on EST data the AngMhc1 transcript may also end with a short extension to the antepenultimate exon. This C-terminus is similar to that found for AeaMhc3 and CpqMhc4 and might be used in a similar combination of the other alternatively spliced exons.

A phylogenetic tree of all arthropod Mhc1 protein sequences, always incorporating the first variant of the clusters of alternatively spliced exons and excluding the differentially included penultimate exon, has been generated (Figure 4). In general, the tree reflects the phylogenetic relationship between the species. The AeaMhc3 sequence is most closely related to the CpqMhc3 and the CpqMhc4 sequence implicating that the last common ancestor of Aedes and Culex already had one of these genes. The phylogeny of the Drosophila species slightly differs compared to other analyses 23. Thus, the DaMhc1 sequence would have been expected to separate after the divergence of the DpMhc1 sequence. Similarly, the DseMhc1 gene would have been expected to be the closest relative of the DssMhc1 sequence. Overall, the sequence identity is very high. Between DapMhc1 and the other sequences the identity is 70.6 – 77.9%, while it is between 77.0% and 99.7% between the other species.

Whenever intron positions are shared between the genes, the corresponding type of splice site is conserved, with the exception of the shared exon 9 (AmMhc1), exon 10 (TicMhc1), exon 9 (BmMhc1), and the alternatively spliced exon 11 of DapMhc1 (Figure 5). All introns have consensus dinucleotide borders except those downstream of the last variant of the cluster of alternative exons encoding part of the motor domain (homologs of exon 11 in DmMhc1), which have a GC dinucleotide at the 5' donor site instead of the consensus GT. The 3' exons of these alternatively spliced exons again have a consensus GT site. Exon '10a' of AeaMhc3 is almost identical to exon 10a of AeaMhc1 and the following intron also has a GC dinucleotide at the 5' donor site. In contrast to the introns following the exons 9 of AmMhc1, NavMhc1, and BmMhc1, and the intron following exon 10 of PdcMhc1 that have a consensus GT site, exon 10 of TicMhc1 has a GC 5' donor site. The intron following exon 11a of DapMhc1 starts with a consensus GT site, while the intron following exon 11b starts with the absolutely rare GA dinucleotide. Also, all split codons are shared between the genes.

In the part encoding the motor and the neck domain, all intron positions are shared by at least two genes (Figure 5). In the coiled-coil tail domain, all genes have lost several introns so that the exons are considerably longer and the intron positions in many cases are not identical. Assuming, that introns have in most cases been lost and were not gained during evolution 33, an ancient arthropod Mhc1 gene can be reconstructed (Figure 5). The ancient Mhc1 gene is expected to contain all intron positions that appear in at least one of the analysed Mhc1 genes. In the motor domain, the proposed ancient Mhc1 gene structure completely resembles the DapMhc1 gene. The exon lengths are between 30 and 210 bp. The exons in the tail domain are considerably longer (up to 480 bp).

The locations of the alternatively spliced exons of DmMhc1 in the motor domain have been discussed in detail elsewhere 34. The position of the additional alternatively spliced exons of the BmMhc1, TicMhc1, PdcMhc1, and DapMhc1 genes in the structure of the motor domain are shown in Figure 6. The alternative exons of DapMhc1 encoding the structural part from the P-loop to loop-1 have identical P-loop sequences. The loop-1 sequences are identical in length but differ significantly in composition. Studies have shown that the flexibility of this loop affects the rate of ADP and phosphate release, with greater flexibility leading to an enhancement in the rate of product release 35. Although the amino acid composition is different between the alternative variants, both contain two glycines and a similar overall charge. The alternative exons of DapMhc1 including loop-4 are similar in length and composition. This region of the motor domain has not been investigated so far and therefore functional consequences of differences in the two variants cannot be drawn. Loop-4 has been postulated to be important for the proper localization of class-I myosins that contain elongated loops that sterically interact with actin-binding proteins 36 but the loop-4 sequences are almost identical between the two DapMhc1 variants and the two variants must therefore modulate a different property of the motor domain. The loop-2 sequence is modulated by alternative exons in the BmMhc1, DapMhc1, PdcMhc1, and TicMhc1 genes. By studies of the Dictyostelium class-2 myosin with its loop-2 replaced with the analogous loop from four other myosins with different enzymatic activities, loop-2 was shown to be involved in the weak and the strong binding interactions with actin 37. It also plays an important role in the rate-limiting step of P_i release 3839. The exon variants of the BmMhc1, DapMhc1, PdcMhc1, and TicMhc1 genes encoding the loop-2 sequence have identical numbers of lysine and arginine residues. The "a" variants are always one residue shorter and have only four instead of five glycines. These differences are, however, very subtle and their influence on actin binding is expected to be very small. The variants of the alternative exon in DapMhc1 following loop-2 are very similar. This part of the motor domain has also not been investigated so far.

The genes for Aea, Ang, Am, Bm, Cpq, Dm, Drp, Dp, Dse, Dss, Dy, Dw, Pdc, and Tic Mhc1 and Mhc3 have been obtained by TBLASTN searches against the insects section of the NCBI wgs database (Table 1)44. The genes for the Da, Der, Dg, Dmo, and Dv Mhc1 have been obtained using the BLAT alignment tool 45 against the UCSC Genome Browser database 4647. The DhMhc1 sequence was derived from the NCBI nonredundant database. The DapMhc1 sequence has been obtained by a TBLASTN search against the 9× assembly of the Daphnia pulex genome provided by the DOE Joint Genome Institute 48 and the Daphnia Genomics Consortium 49. The NavMhc1 gene was derived from version 1.0 of the Nasonia vitripennis assembly provided by the Human Genome Sequencing Center at Baylor College of Medicine 50. The exons of the genes were predicted by manual inspection of the nucleotide sequences. For the correct prediction of the transcriptional start and the 3' terminal exons, the analysis of cDNA and EST data, that has been obtained from the EST section of NCBI's nucleotide database, was necessary. In particular, the following data has been obtained: For TicMhc1, only a small amount of EST data is available, confirming the prediction of exon2. There is not enough data to exclude a further untranslated 5' exon, as well as further C-terminal exons. For AngMhc1, several EST and cDNA clones support exon1 and the different C-termini. The C-termini of AeaMhc1 are also supported by several EST clones (e.g. GenBank ID DV384821). Exon1 of AeaMhc3 is supported by EST data. Exon1 of AeaMhc3 has been used for the identification of exon1 of AeaMhc1, as there is no direct evidence by EST data. Surprisingly, it is found 26,432 bp before the translation start codon ATG. For AmMhc1, the N-terminus is not supported by EST or cDNA data. Therefore it is not clear whether there might be an additional 5' untranslated exon. The C-termini are supported by several EST and cDNA clones (e.g. GenBank ID CK629939). The C-terminus of DapMhc1 is supported by EST data (e.g. GenBank ID BJ927473), while there is no EST data for the N-terminus. For BmMhc1, exon2 is supported by EST data. However, the corresponding EST clones are not long enough to exclude a further 5' untranslated exon. Both C-termini of BmMhc1 are supported by EST clones (e.g. GenBank ID BP179837). The genomic DNA of the BmMhc1 gene contains a gap in the coiled-coil tail region. The missing amino acid sequence has been derived from EST data. However, the exon/intron structure in the corresponding region remains unresolved.

All alternatively spliced exons have been aligned manually. Some kind of relationship is already obvious from these sequence alignments. To get a more quantitative description, sequence identity matrices have been calculated for each set of aligned exons. Subsequently, sets of homologous exons from all Mhc1 genes have been clustered by sequence similarity. We have visualized the results in graphs that have to be read in columns. The highest identity between an exon listed on top and any variant of a certain Mhc1 protein listed on the left side has been set to 1 (red colour) while the differences between the values of the lower identity exons and the value of the highest identity have been plotted for the other combinations of exons. Thus, in every column the highest identity of the named exon to one of the variants of the other Mhc1 proteins is visualized.

The phylogenetic tree was generated using neighbour joining and the Bootstrap (1,000 replicates) method as implemented in ClustalW (standard settings) 51 and drawn by using TreeView 52. The sequence of DapMhc1 has been used as outgroup.