In vertebrates, most of the protein-coding genes are interrupted by multiple introns that are removed at the donor and acceptor splice sites so that the adjacent exons are spliced. This process is mediated by an elaborate molecular machine, the spliceosome that consists of 5 snRNPs (small nuclear ribonucleoprotein particles) along with numerous less stably associated proteins, and is conserved throughout the eukaryotic world 123. The U2 spliceosome (the major eukaryotic spliceosome) interacts with specific parts of the intron and the flanking exons to ensure accurate and efficient splicing 4. The nucleotides at the intron termini and the adjacent nucleotides in the exons are involved in these interactions and comprise the splicing signal. The (A/C)AG|

GT

The GT dinucleotide in the first two intron positions is the most conserved element of the U2 donor splice signal. However, in a small fraction of donor sites (<1%), GT is replaced by GC; in these cases, the rest of the nucleotides in the donor signal adhere more closely to the consensus sequence, apparently, compensating for the T to C substitution that is unfavorable for splicing 8910. This rare class of donor splice signals has been implicated in alternative splicing 91112. For example, the conserved C at the +2 position of the 10^th intron of the let-2 gene which encodes one of the collagen isoforms is essential for developmentally regulated alternative splicing in the nematode C. elegans. Replacement of the GC donor signal with a moderate or strong GT signal abolishes splicing regulation and leads to excessive usage of exon 10 of let-2 in embryos 11. Generally, a substantial enrichment of GC donor signals in alternatively spliced genes has been observed in human, C. elegans and Arabidopsis 91112.

Pairwise comparisons of GC splicing signals in the nematodes Caenorhabditis elegans and C. briggsae suggested that GC donor signals are not evolutionarily conserved in nematodes: among the 26 C. elegans GC-AG introns, only 5 had a GC-AG counterpart in C. briggsae 11. Frequent switching between GT-AG and GC-AG introns has been reported for 5 vertebrate genomes 10. We were interested in exploring the genome-wide evolutionary dynamics of the donor splice sites and, in particular, sought to determine whether there might be a trend toward depletion or accumulation of GC.

Genomic alignments of 8 vertebrates (chicken, opossum, dog, cow, rhesus macaque, human, mouse and rat) were extracted from the UCSC genome browser (Additional file 1) 1314 and used to map cases of GC > GT and GT > GC conversion to the branches of the mammalian phylogeny (Figure 1). Terminal leaves of the tree (individual genomes) are more vulnerable to the effects of sequencing errors and/or population polymorphism than internal branches 15. Therefore, only those sites were analyzed in which the GC or GT donor signal was shared by at least two sister taxa (e.g., we assign the GC signature to the rodent clade when GC was found in both mouse and rat sequences). It was further required that either GT or GC signal was conserved in all outgroup species for which an alignment was available at the given site (the presence of the dog and cow sequences was unconditionally required). Altogether, there were 122,621 and 253 invariant GT and GC donor signals, respectively, and 656 variable sites of which 338 mapped to internal branches and were employed for the analysis of evolutionary dynamics.

The GC>GT and GT>GC conversion events were reconstructed using maximum parsimony (Figure 1). Unexpectedly, we observed a pronounced excess of GT>GC conversion over GC>GT conversion that is indicative of accumulation of GC donor splice sites in both primate and rodent genomes (Figure 1). The trend is stronger in the rodent lineage than in the primate lineage (Figure 1), an observation that is consistent with the overall fast genome evolution in rodents 16. The apparent accumulation of GC donor signals was further supported by the analysis of the terminal branches of the tree although the excess of GC>GT conversions in macaque compared to human (Figure 1) could be caused by sequencing errors and/or population polymorphism. The observed excess of GT>GC conversions was robust with respect to the composition of the outgroup species set (Additional file 2).

The observed excess of GT>GC conversions hardly can be explained by a nucleotide substitution bias. It has been repeatedly shown that mammalian genomes have a tendency to become more AT-rich 171819. Even if one assumes that, due to unknown reasons, this trend is reversed in the donor sites so that T to C substitutions are twice as frequent as C to T substitutions, such a bias would not account for the observed excess of GT>GC conversion events (P < 10^-10 according to the χ² test for pooled conversion events).

Considering that mutational bias did not seem to be a plausible cause of the observed accumulation of GC donor sites, it seems most likely that this trend has to do with the involvement of GC sites in alternative splicing that is widespread and essential in mammals 91112. As GT>GC conversion can substantially alter the pattern of alternative splicing 11, these changes might become beneficial and eventually would be fixed in the population. Thus, positive selection could be a plausible explanation for the observed accumulation of GC in donor sites. However, an even more plausible scenario would involve evolution of a strong splice site context that would allow neutral fixation of GC sites. The neutrally fixed GC sites, then, could be recruited for alternative splicing and thus would become subject to purifying selection forbidding the reverse GC>GT conversion.

The authors declare that they have no competing interests.

AC performed sequence comparisons, SWH contributed to the interpretation of the results, AC and IBR contributed to the analysis of the results and wrote the initial draft of the manuscript, IBR and EVK incepted the study, contributed to the analysis of the results and wrote the final manuscript, all authors edited and approved the final version.