The bursicon gene CG13419 in Drosophila melanogaster, named burs by Dewey et al. (2004), encodes one-half of the functional heterodimeric insect neurohormone bursicon, which regulates cuticle sclerotization and wing inflation after ecdysis in Drosophila and other insects (Luo et al. 2005; Mendive et al. 2005). This protein, which is also known as the α-chain of bursicon, is conserved throughout insects and related arthropods (Van Loy et al. 2006). In Drosophila, Burs is a 173-amino-acid protein encoded by three exons. Here we show by comparison with all available insect and related arthropod genomes that the ancestral insect gene had four coding exons and in all except mosquitoes it is a conventionally spliced locus. In mosquitoes bursicon is translated from a mRNA derived from four exons (one 5′ UTR and three coding); however, the middle coding exon is on a different chromosome arm (2R) from the first, second, and fourth exons, which are together on chromosome arm 2L. To our knowledge this is the only example of trans-splicing known to involve internal exons of genes on nonhomologous chromosome arms, although this has been demonstrated experimentally for the complicated trans-spliced mod(mdg4) locus in Drosophila (Gabler et al. 2005).

The unusual Anopheles gambiae bursicon gene structure was initially inferred bioinformatically using the Drosophila protein as query in TBLASTN searches of the genome sequences and by comparison of two 5′ ESTs (NAP1-P41-B-11-5 and NAP1-P111-F-12-5) and the available genome sequence (Holt et al. 2002). We obtained the cDNA clones from which these ESTs were generated and completed their sequences (GenBank nos. AY735442 and AY735443). These unambiguously confirm that the first, second, and fourth exons are present contiguously on the same scaffold on chromosome arm 2L (on the reverse strand at ±10.5 Mb on this ±49-Mb chromosome arm), while the third exon is present alone on chromosome arm 2R (on the reverse strand at ±22 Mb on this ±67-Mb chromosome arm) (Figure 1). These regions of the genome are both well assembled for at least 10 kb on each side of the bursicon exons as judged by comparison with the raw traces available in the Trace Archive at NCBI, indicating that this unusual gene structure is not the result of misassembly of the genome sequences. We amplified the locus with exons 1, 2, and 4 from the upstream promoter region to the 3′ UTR, and the single clean PCR product size (1430 bp) and DNA sequence agree fully with the assembled genome sequence.

Furthermore, this trans-spliced gene structure encoding bursicon is also present in the Aedes aegypti and Culex pipiens genomes (Figure 1). Although the chromosome locations of the contigs containing the 1, 2, 4, and 3 exons are not known in Ae. aegypti, microsynteny on either side with orthologs of the same genes in An. gambiae and Ae. aegypti implies that they are again on separate chromosome arms (microsynteny cannot be examined easily for C. pipiens until the genome assembly is available, but given its relatively close relationship to Ae. aegypti it is likely to be the same). This unusual gene structure has therefore existed for at least the ±150 million years (MY) of mosquito evolution since the split of the culicid and anophelid family lineages, but is younger than the ±250 MY split of the dipteran suborders Nematocera and Brachycera (containing the Culicidea and Drosophilidae, respectively) (Krzywinski et al. 2006). We propose to call these two complementary mosquito bursicon-encoding loci burs124 and burs3.

This unusual trans-splicing of two separate loci encoding complementary parts of bursicon is restricted to these three mosquito genomes. All other available insect and related arthropod genome sequences contain a single locus encoding bursicon from 3 or 4 coding-exon genes (Figure 2). It appears that the ancestral bursicon locus in insects, at least, consisted of 4 coding exons separated by three introns (Figure 2) (the presence of introns in the 5′ UTRs cannot be conclusively demonstrated or excluded in these species in the absence of cDNA evidence). The drosophilids, Tribolium, and Ixodes lost the last phase 0 intron, the mosquitoes lost the middle phase 0 intron, and the water flea Daphnia pulex and deer tick Ixodes scapularis lost the first phase 1 intron (or this intron is unique to insects). The position of the first phase 1 intron within or near the signal peptide is rather variable, but this kind of intron movement within highly variable coding sequences such as signal sequences is not unusual.

We attempted to monitor the trans-splicing of the An. gambiae bursicon mRNA by RT–PCR from the second to fourth exons using primers designed to amplify the entire coding region from start to stop codons (Figure 1). If the bursicon mRNA trans-splicing is a slow inefficient process, then in addition to the expected final product of ~490 bp we would anticipate obtaining either a relatively short product of ~230 bp, reflecting the splicing of the second and fourth exons prior to or instead of trans-splicing, or a long product of ~1300 bp, reflecting the prespliced transcript containing the second and fourth exons connected by the out-of-frame phase1/0 “intron” between them. Instead, the only product we obtain is the final fully trans-spliced product of ~490 bp (data not shown). Thus this unusual trans-splicing of an internal exon into the middle of a separate transcript is achieved efficiently.

How might this unusual example of trans-splicing have evolved? We hypothesize that it resulted from duplication of the bursicon gene followed by essentially simultaneous inactivation of the third exon in one copy and at least the second or fourth in the other. Association of the two pseudogenic transcripts might have allowed trans-splicing leading to an intact transcript that became indispensable. Subsequent decay and loss of the third exon from the 2L gene copy and the second and fourth exons from the 2R copy would yield the present-day situation. This would be an unusual series of events, accounting perhaps for the rareness of this kind of trans-splicing of an internal exon. If such a duplication involved more than just the bursicon locus we might find hints of genomic similarity on either side of the burs124 and burs3 loci; however, the genes on either side of these loci—in An. gambiae, at least—are unrelated to each other. A duplication might have involved only this locus, or genome flux over the past >150 MY might have removed any remnants of a larger duplication.