The possible contribution of transposable elements to the proteome has been discussed in several publications 1234567 and has provoked much debate 8. Many mechanisms are known to increase the protein-domain repertoire, e.g., domain duplication, substitution mutations, insertions, deletions, and domain rearrangements 9. In metazoans, a transposable element may result in transduction, in which a DNA segment downstream of transposable elements is mobilized as part of an aberrant transposition. This may result in gene duplication or exon shuffling, subsequently enriching the protein repertoire 10111213. However, in the process of transduction, the transposable element does not acquire gene fragments as part of its sequence.

In plants, on the other hand, thousands of transposable elements contain duplicated gene fragments, captured in a process termed transduplication. Transduplication is a potentially rich source of novel coding sequences within rice and Arabidopsis thaliana 141516. Recently, transduplications of small nucleolar (sno) RNA by retroposon-like non-LTR transposable elements were found in the C. elegans 17 and platypus genomes 18.

The Harbinger superfamily of "cut-and-paste" DNA transposons was discovered through in silico studies 19. This superfamily is characterized by Harbinger-specific transposases that are distantly related to the transposases encoded by the IS5-like group of bacterial transposons, such as IS5, IS112, and ISL2. Harbinger transposons are not as widespread as the eukaryotic hAT and mariner/Tc1 transposons; they are found in plants and nematodes 1920212223 but not in mammals. Usually, Harbinger transposons are flanked by 3-bp target site duplications and 25- to 50-bp inverted terminal repeats (ITRs).

Turmoil-1 is a 5,024-bp long DNA transposon with 760-bp long ITRs and a Harbinger-specific transposase (Figure 1A). These ITRs are unique to Turmoil-1 and are not found in other members of the Harbinger superfamily of DNA transposons 24. One complete copy of the Turmoil-1 was found on chromosome II of C. elegans; eight Turmoil-1 fragments exist in the genome (for detailed information, see Table 1). An analysis of C. elegans transposable elements 2526 revealed that a 205-bp ITR sequence within Turmoil-1 is highly similar to a region of two exons separated by an intron of the rsp-2 gene (see pairwise alignment using bl2seq 27 Figure 1B). These two exons encode the RNA Recognition Motif (RRM), which is found in many eukaryal and bacterial proteins. Specifically, the type of RRM domain present in the rsp-2 (called RRM1) gene is highly conserved evolutionarily 28. The high similarity between the ITR sequence and the fragment of the rsp-2 gene implies that one originated from the other. The antiquity of this domain and a phylogenetic analysis (Figure 2) indicate that Turmoil-1 has recently acquired a portion of the rsp-2 gene sequence into its ITR. Tree reconstruction was performed with the PhyML program version 2.4.5 29 using among-site rate variation with four discrete rate categories, and the JTT model 30 of sequence evolution.

A comparative analysis of the rsp-2 gene and the 205-bp region of the gene found in the ITR sequence, revealed that the Turmoil-1 sequence has accumulated several point mutations within the 5' splice site that make it non-functional, whereas the 3' splice site of the intron remains intact (the mutations in the 5' splice site region are marked in red in Figure 1B). Since the RRM domain within the Turmoil-1 DNA transposon is not under purifying selection to maintain the reading frame or the functionality of the splice sites, these mutations are not unexpected.

Within the same ITR domain of Turmoil-1, and very close to the site of insertion of the RRM domain of rsp-2 gene, there is evidence of another "DNA kidnapping" event. A 131-bp fragment from the coding region of C. elegans gene cpg-3, which is unique to nematodes, was inserted into the ITR (Figure 1C). No sequences homologous to Turmoil-1 flank the cpg-3 gene. Thus, similar to the rsp-2 case, a gene fragment from the cpg-3 most likely was acquired by Turmoil-1, and not vice-versa. As the gene fragments are present on both sides of the ITR, capture may have occurred through non-homologous recombination.

One of the Turmoil-1 copies (number 1 in Table 1) contains within it an open reading frame (ORF) with the accession number Y48G1BL.4. It contains two putative exons and an intron (Figure 3), which are similar to the RRM domain. The 5' splice site that corresponds to that in the rsp-2 transcript has been mutated. A novel 5' splice site is most likely located nine nucleotides downstream from the original one. At this site, a point mutation changed an AT into a GT dinucleotide (marked in red in Figure 3). Usage of this 5' splice site maintains the ORF equivalent to that of the RRM domain of rsp-2 with the exception of the addition of three amino acids. These additional residues should have a negligible effect on the three-dimensional structure of the RRM domain (Figure 3C). This ORF, however, may not be transcriptionally active as its sequence is only found in the UNIPROT database (accession number Q9N3P9), and there is no EST or cDNA supporting evidence. If this is the case, it would be consistent with reports that indicate that all known transduplicates in rice, in spite of their genomic abundance, are pseudogenes 16.

Our analysis indicates that Turmoil-1 of C. elegans has captured two unrelated coding sequences within its ITRs at proximate locations. The presence of a transduplication "hotspot" in this region may be tentatively inferred. This analysis reveals a transduplication of protein-coding regions in C. elegans and strengthens the hypothesis that protein domains may be mobilized by transposable elements.

RRM: RNA Recognition motif; ITR: inverted terminal repeat; ORF: open reading frame.

The authors declare that they have no competing interests.

NS and AS did the experiments and analysis. WM interpreted the results. GA and TP supervised the study. NS, AS, WM, TP and GA wrote the paper.