Main

The mechanisms by which splice sites are selected and the evolutionary patterns are different in ‘small’ (up to 80 base pairs) and ‘large’ (more than 80 base pairs) introns in Drosophila3,8,9,10. Size-class transitions in homologous introns are frequent (about 20%) between D. melanogaster and D. virilis or D. pseudoobscura but rare among most closely related species3,11. We investigated the relation between intron size and recombination rate by analysing each intron type separately (Fig. 1a), and found that recombination occurs in small introns on average at 2.45±0.04 centimorgans per million base pairs (cM Mb−1), whereas the value for large introns is 2.20±0.05 cM Mb−1, a difference that is significant at the 0.001 level (by analysis of variance).

Figure 1: Association between intron size and recombination rate.
figure 1

a, Relation between local recombination rate and intron size over 1,817 introns (from 590 GenBank accessions) in expressed genes of Drosophila melanogaster (linear regression on log-transformed size: b=0.026, P=0.001, r2=0.006). Recombination rates were estimated as described5. b, Average rates of recombination (±one standard error) for introns of less than 60, 60–80, and more than 80 base pairs (bp); n is 370, 665 and 782, respectively.

Full size image

Within large introns, there is no significant association between recombination rate and intron size (b=−0.012, P>0.2), so whatever causes this relation must act equally against all large introns, irrespective of their absolute size. In small introns, there is a significant, positive association between recombination rate and intron size (b=+0.002, P=0.02), indicating that very small introns are also deleterious: they tend to occur in regions of low recombination, where natural selection is less efficient.

The average rate of recombination for introns of less than 60 base pairs is 2.26±0.07 cM Mb−1, whereas the rate for introns of 60 to 80 base pairs is 2.56±0.06 cM Mb−1, a difference significant at the 0.001 level (Fig. 1b). These findings are not surprising, as it is known that very small introns do not splice well8.

Another explanation for these results is that intron size is a neutral trait, with recombination alone generating the association between recombination rate and intron size by its effect on DNA insertion and deletion processes. However, very small and large introns are both associated with regions of low recombination, excluding not only this possibility, but also any other that does not involve the direct action of natural selection on intron size. Fewer than 2% of the large introns in our sample show any sequence similarity to transposable elements, so our results cannot be explained by the accumulation of these elements in regions of low recombination3,12.

Intron size usually changes through small deletions and insertions of DNA2,3,13. Our results suggest that natural selection in Drosophila acts on the variants generated by these processes, selecting against large introns and very short introns, as both tend to occur in regions of low recombination where selection is less effective.