Evolution of trichome patterns through cis-regulatory evolution of shavenbaby

Does morphological evolution occur primarily by changes in protein sequences or through changes in cis-regulatory regions? In recent years, this rather simple question has attracted considerable attention and surprisingly robust debate. On one hand, most of the evidence for the molecular causes of phenotypic evolution comes from studies of protein-coding regions. On the other hand, developmental biologists have argued for several decades that changes in the cis-regulatory regions of genes seem a more probable route to developmental evolution.

Nailing down the role of cis-regulatory regions in developmental evolution has been much more difficult than gathering the evidence that protein-coding changes contribute to phenotypic evolution. This is largely because the cis-regulatory "code" is not as well understood as the genetic code for proteins and because we possess more limited tools for studying cis-regulatory function than for studying protein function. Nonetheless, in recent years many groups have succeeded in demonstrating an important role for cis-regulatory evolution.

Several outstanding questions remain. How precisely do cis-regulatory regions evolve?; Is cis-regulatory evolution more common for genes in particular parts of regulatory networks?; And can we detect when and how natural selection has acted on cis-regulatory regions?

We did not set out to investigate cis-regulatory evolution, but we stumbled into this problem by performing an unbiased experiment to identify the genetic causes of a simple morphological difference between species, the loss of trichomes in one species of Drosophila.

The first-instar larvae of Drosophila sechellia display a different pattern of dorsal trichomes than the closely related species D. melanogaster, D. simulans and D. mauritiana. D. sechellia larvae are missing a broad swath of dorsal trichomes and a patch of lateral trichomes on most segments along the body axis.


We first determined, through standard genetic crosses, that this entire difference in trichome patterning resulted from evolution of the expression pattern of a gene called shavenbaby-ovo (svb) (Sucena & Stern, 2000). We subsequently showed that the changed expression pattern resulted from changes in at least three separate cis-regulatory modules (McGregor et al., 2007). Each regulatory module generates part of the complete trichome pattern and multiple mutations of relatively small effect must have been fixed in the D. sechellia lineage to generate their naked cuticle.


We have continued to work on this problem by expanding our search for additional enhancers of the svb gene and by performing detailed dissections of the known enhancers. This work is in progress.

Our data indicate that, in this case, morphology has evolved through an abundance of cis-regulatory changes and no protein coding changes. This case study has already provided preliminary answers to the first two questions posed above. For this trait, cis-regulatory regions evolved through multiple, very particular mutations and all of the mutations are clustered in a single gene.

These results help to explain a confusing discrepancy between many observations in evolutionary genetics -- which tend to indicate that evolution often occurs by the accumulation of mutations of small effect -- and the results from evolutionary developmental biology -- which tend to suggest that morphological evolution often occurs by changes at key regulatory genes. In fact, both may be true. Long-term evolutionary changes may have resulted from the accumulation of small-effect mutations at a few, special loci. We call these hot-spot genes and we suspect that the structure of developmental networks helps to explain why mutations at these hot-spot genes are preferred over mutations elsewhere in the network (Stern & Orgogozo, 2008,2009).


If some genes really are hot spots for evolutionary change, then we would expect to see these genes involved in similar evolutionary changes in other lineages. We therefore examined species of the D. virilis clade of flies, which also display a diversity of larval trichome patterns. In this group, we found that the pattern of svb expression is precisely correlated with larval trichome patterns, suggesting that svb is indeed involved in the evolution of trichome patterns in this group of species as well (Sucena et al., 2003).


We have continued to work on svb evolution in the D. virilis group and we will soon have the results of a definitive test of whether evolution of svb has caused trichome evolution in this group of species. Stay tuned!