During the addition of new, adaptive genes to the genome, if certain sequences are better donors in creating new genes than others, then a kind of selection process results that can improve the ability of the genome to generate new, useful genes. When applied to the evolution of exons, this idea predicts that there should be a predominance, among translated exons, of exons and pairs of exons that are multiples of 3 bases in length. It also predicts that more introns should fall between codons rather than splitting them. Empirical verification of these predictions is described.
“The third and most interesting alternative recognizes that exon shuffling itself imposes a kind of selection, regardless of when or how introns originated. In the assembly of new proteins by shuffling, exons that code for useful pieces of protein structure and function more often give rise to successful new genes (and thus successful new organisms) than exons that do not, even if exon-intron organization were initially random overall. Such exons will be used again and again, and mutations that slightly alter splice sites so as to improve compatibility with other exons will be selected from them. As R. Doolittle put it, "introns that occur between potentially useful domains will have added survival value and will predominate over those that split genes at random locations" (1985, p. 235). For this kind of selection, which L. Altenberg and D. Brutlag (MS) describe as "constructional selection," to be important in giving rise to domain-defining exons, shuffling must be a frequent event in evolution. Although it is impossible to measure the rate of shuffling, the relative stability of gene structure (e.g., for the glycolytic enzymes just discussed) across the broadest possible phylogenetic spectrum suggests that it has been relatively infrequent since the prokaryote-eukaryote divergence. (Constructional selection could, however, have been very important in the kind of precellular genomic evolutionary process described below.)
... Exons that functioned well with other exons would be favored, and thus selection (constructional selection of the sort envisioned by L. Altenberg and D. Brutlag [pers. comm.]) for exons encoding structurally or functionally defined polypeptides would intensify.”
“The other possibility is that exons that encode modules at the level of protein structure and/or function are themselves the product of some prolonged period of selection. L. Altenberg and D. Brutlag (in prep.) call this "constructional selection," arguing that, among exons that code randomly for bits of protein, those that by chance encode shuffle-able modules would be selected by shuffling itself, since they would come to be used again and again in the construction of new genes (and thus, in passing, new organisms or species). This seems intrinsically reasonable, and one then needs only ask when such shuffling started, and how long it has been going on.”
Tomita M,, Shimizu, N., Brutlag, D. L. 1996. Introns and reading frames: correlation between splicing sites and their codon positions. Molecular Biology and Evolution 13(9): 1219-1223.
“The concept of selection of exons based on properties of their ends with respect to codon boundaries was derived from earlier work by Lee Altenberg.”