mod(mdg4) is a complex locus encoding >25 different mRNAs with protein products that are believed to be involved in the regulation of higher-order chromatin structure (Dorn et al. 1993; Gerasimova et al. 1995; Buchner et al. 2000). All mod(mdg4) mRNAs share the first four exons, which encode a BTB domain, and differ in the fifth and sixth exons encoding the variable C terminus of the protein (Gerasimova et al. 1995; Buchner et al. 2000; Dorn and Krauss 2003). The first indication of a requirement for trans-splicing in the generation of Mod(mdg4) proteins came after the realization that the two DNA strands of the gene have coding capabilities and contain coding sequences present in mature mRNAs that are translated into functional proteins (Labrador et al. 2001). Further analysis of the encoded products of the mod(mdg4) gene revealed that as many as seven out of 27 mRNAs are encoded by the complementary DNA strand (Dorn et al. 2001; Dorn and Krauss 2003). The finding that single molecules of mRNA could originate from independent mod(mdg4) transgenes located in different chromosomal positions or from two trans-heterozygous mutant alleles (Dorn et al. 2001; Mongelard et al. 2002) was further evidence supporting the involvement of trans-splicing in the maturation of mod(mdg4) mRNAs and discarded alternative hypothesis such as the existence of somatic DNA rearrangements of the locus. Because potentially all eukaryotic cells have the capability of performing trans-splicing, it is surprising that so far only one well-characterized example of trans-spliced mRNAs has been found. One can argue that the absence of additional examples of trans-splicing, even after the sequencing of multiple eukaryotic genomes, suggests that the mechanism is not biologically relevant. However, the annotation of mod(mdg4) by the Drosophila genome project failed to detect the involvement of trans-splicing in the maturation of the mod(mdg4) encoded mRNAs, even though a wealth of information was already known about the gene and its transcripts. Therefore, it is still possible that trans-splicing is not uncommon in eukaryotic cells, and only after a thorough genomic and proteomic analysis, will we have a full picture of the relevance of this process in the generation of protein diversity. Alternatively, it is also possible that trans-splicing occurs only rarely and has thus remained elusive to experimental detection. In either case, gaining further insights into the mechanisms of trans-splicing and understanding how it can be experimentally induced to obtain a specific mRNA encoding a predicted combination of exons may be of particular interest for the correct interpretation of genomic data, for the development of in vivo molecular tools, or for the improvement of gene therapy technology.

The genomic approach used to compare the structure of the mod(mdg4) locus between phylogenetically related species has provided valuable information on the evolution and the origin of the trans-splicing process associated with the maturation of several mod(mdg4) mRNAs. In addition, the results show that the comparative analysis of genome sequences could be efficiently used to identify new potential examples of trans-splicing. This and previous reports have shown that a large number of mod(mdg4) variable exons are found in both DNA strands of the D. melanogaster gene and that this distribution requires alternative trans-splicing to explain the presence of hybrid mRNAs and the encoded proteins in the cell (Dorn et al. 2001; Labrador et al. 2001). Similar trans-splicing events have been described elsewhere (Eul et al. 1996; Caudevilla et al. 2001a,b). What makes trans-splicing of the Drosophila mod(mdg4) gene unique compared with other described examples is that the protein encoded by the hybrid mod(mdg4)2.2 mRNA is a major protein with a functional role as a component of the gypsy insulator (Gerasimova et al. 1995; Gerasimova and Corces 1998). The significance of the structure of the gene for the function of the encoded proteins is evident from the conservation of the same structure, likely by natural selection, for >25 million years in two independent lineages leading to the D. pseudoobscura and D. melanogaster species.

It is not clear from our results, however, whether trans-splicing is important for the regulation of expression of the proteins encoded by the mod(mdg4) gene in A. gambiae. In this species, all variable exons are found in the same DNA strand as the constant exons, implying that all mRNAs can be generated by cis-splicing, by trans-splicing, or by both mechanisms. Whether the contribution of trans-splicing to the pool of mRNAs encoded by the mod(mdg4) gene in mosquito is significant can only be explored experimentally. However, when similar exon arrangements are found in other complex genes, such as the Drosophila Dscam or the protocadherin genes in the mouse, cis-splicing apparently accounts for the presence in the cell of all functional alternative variants (Schmucker et al. 2000; Tasic et al. 2002). Interestingly, like in mod(mdg4), the mouse protocadherin α, β, and γ genes encode multiple proteins consisting of common and variable regions, the latter encoded by a number of variable exons. Experimental evidence suggests that transcription of the gene systematically produces trans-spliced mRNAs involving premature RNAs transcribed from different promoters located at the 5′ of the variable exons. However, the level of these trans-spliced RNAs is so low that it is difficult to picture a biological role for the encoded proteins (Tasic et al. 2002). It is possible that the same is true for the mosquito mod(mdg4) gene, in which trans-splicing, similarly to the protocadherin α, β, and γ genes, may be occurring apparently without any biological significance, hence explaining the orderly arrangement of all exons in a single DNA strand. The presence of putative promoters along the sequence of the Drosophila gene suggests that transcripts may be produced at different transcription start sites in the 5′ of sequences encoding the variable region of the protein. These transcripts will later trans-splice with the mRNAs encoding the common region. Increasing evidence suggests that small RNAs transcribed by the complementary strand of genes may have an important regulatory role in gene transcription (Allshire 2002). Interestingly, putative small RNAs originally involved in the regulation of transcription of the gene may also be the source for the origins of transcription in the opposite strand, necessary to explain trans-splicing in D. melanogater. The same presence of such RNAs transcribed from the opposite strand raises the question of how mod(mdg4) escapes or benefits from the silencing presumably induced by RNAi.

According to this scenario, the ancestral mod(mdg4) organization would be similar to that of A. gambiae, and the derived organization will correspond to the one observed in Drosophila, with the bulk of rearrangements occurring only in the Drosophila lineage prior to the split between D. pseudoobscura and D. melanogaster. The significance of trans-splicing in the ancestral organization would be similar to that observed in the protocadherin α, β, and γ genes and will only gain biological relevance in the Drosophila lineage, concomitantly with the emergence of DNA rearrangements. Unfortunately, there are no data available at the moment to test this hypothesis by determining the organization of the locus in an ancestor common to both lineages. Although biologically possible, the alternative hypothesis, that is, the last common ancestor between Drosophila and Anopheles shared the same kind of gene organization as Drosophila, is less parsimonious because it requires that most rearrangements in the mosquito lineage arose toward the perfect alignment of the exons in a single DNA strand. The high number of sequence rearrangements and the subsequent structural stability observed within the D. pseudoobscura and D. melanogaster lineages suggest that the reorganization of the locus may have occurred under the control of positive selection that may have reinforced the role of trans-splicing in the maturation of mod(mdg4) mRNAs, perhaps adding a new level of regulation to the expression of the gene.

One of the most remarkable findings of this work is the large number of breakpoints within mod(mdg4) required to explain the evolution of the gene. A rough estimate of the rate of chromosome breakage and fixation of ~1.2 sequence disruptions per million years per Mb can be obtained if we consider that at least 11 breakpoints were produced in a DNA sequence encompassing ~40 kb during a time lapse of 250 million years. This number is surprisingly large considering that the rearrangements took place within the transcribed region of a gene and that Drosophila has the highest rate of chromosomal evolution reported so far, with an estimated number of sequence disruptions per Mb per million years of only 0.066 to 0.05 (Ranz et al. 2001). We have tested the possibility that transposable elements could be involved in the generation of these rearrangements and concluded that no evidence or traces of such repetitive sequences can be found in the current sequence of the locus in the three species studied (data not shown). We cannot rule out, however, that these sequences may have been eliminated during evolution due to the rapid turnover at which some transposable elements are subject to in Drosophila (Petrov et al. 2000). An alternative possibility is that the number of chromosome breaks described in the literature has traditionally been obtained based on in situ hybridization analysis and without genome sequence data. The numbers thus derived might be an underrepresentation of the total breakage rate, because small inversions may be undetectable by the low resolution of this technique. Interestingly, this is the case in Saccharomyces cerevisiae, in which frequent small inversions found in the genome will go undetected by alternative large-scale detection methods. With a size of 14 Mb, a total of 1100 small single gene inversions are necessary to explain the differences in gene arrangement observed between the S. cerevisiae and Candida albicans genomes. Considering that the divergence between the two species is ~140 million years, an estimated rate of 1.2 sequence disruptions per Mb and million years is necessary to account for such reorganization (Seoighe et al. 2000). A similar magnitude of rearrangement rates of ~0.4 to 1.0 chromosomal breakages per Mb per million years has been found when partial regions of the genomes of Caenorhabditis elegans and Caenorhabditis briggsae were compared (Coghlan and Wolfe 2002). This rate is at least four times that of the previously reported rate for Drosophila and is comparable to what we have observed within the mod(mdg4) locus. An analysis of small inversions inducing differences in gene orientation between genomes of closely related species of yeast suggests that such inversions are in fact small gene duplications followed by differential sequence degeneration (Fischer et al. 2001). Considering that our data show that exon duplications are frequent in mod(mdg4), a similar mechanism could at least partially explain some of the rearrangements that took place during the evolution of this gene.