The beautiful patterns that decorate the wings of butterflies have fascinated lepidopterists and evolutionary biologists alike for generations, both of them captivated by their incredible range of colours and shapes. For evolutionary biologists, however, the source of this variation — more precisely, the genetic variants on which natural selection acts to generate such variation — has often been quite elusive. In a new study, Patrícia Beldade and colleagues have used a candidate gene approach, together with artificial selection experiments, to get to the genetic basis of variation in wing eyespot size in the tropical butterfly Bicyclus anynana. Their work strongly implicates the Distal-less (Dll) gene in generating variation in eyespot size and highlights the strength of evo–devo for studying important adaptive features.

The authors chose to study Dll, which encodes a transcription factor, because of its previously reported organizing activity during eyespot development (see Highlights December 2001). But knowing that Dll is involved in determining the position of future eyespots and showing that variation at Dll is responsible for intra-individual variation in eyespot size are two different things. Rather than trying to match Dll variants to eyespot size in a natural population, the authors artificially selected butterflies for nine generations, and obtained two lines – one with large and one with small eyespots (high and low line, respectively) — in the expectation that Dll polymorphisms would segregate with one line or the other. That the selection experiment worked indicates that eyespot size is a highly heritable trait. Reassuringly for their hypothesis, Beldade et al. found not only that the domains of Dll expression correlated with eyespot size, but also that the phenotype of backcross progeny between a high–low hybrid and either a high or low parent segregated with a line-specific Dll polymorphism. Of course, as the authors point out, the linkage of Dll to this trait is not formal proof that Dll itself is involved, as their study does not rule out the involvement of linked loci. The case for the contribution of Dll to the difference in backcross phenotypes was further strengthened when it was found that the LOD score for a putative QTL peaks, or was very high, at Dll, compared with neighbouring regions.

Identifying which, if any, Dll alleles underlie variation in eyespot size is still some way away and, as the experiments revealed, more factors than Dll alone must be invoked to explain the phenotypic difference between the artificially selected lines. Even so, this work is a nice example of how a known developmental pathway can be linked to the standing quantitative varation in a population, made more significant in this case by involving a trait of known adaptive significance.