The Evolution of Insect Diets

Adaptive evolution occurs by natural selection of genetically variable characteristics. Because most features of most species seem to be genetically variable, we expect them to evolve readily, should environmental conditions change. Yet some features evolve only slowly and slightly, as we can see by comparisons among species. For example, larvae of the 700 species of white butterflies (subfamily Pierinae), which have evolved from a common ancestor over at least 30 million years, feed on plants in the mustard family, almost without exception. Likewise, each of the leaf beetles in the genus Ophraella, which is probably more than 8 million years old, feeds only on one or another plant in the sunflower family (Asteraceae); no species have adapted to other kinds of plants. Furthermore, the most closely related species of Ophraella generally feed on the most closely related species of Asteraceae. These patterns, found also in many other groups of herbivorous insects, suggest that such insects might not possess the genetic variation that would be required to adapt to distantly related plants.

Undergraduates Joseph Walsh, Christine Hoffmann, and Stuart Milstein, several graduate students, and I have been testing this hypothesis with several species of Ophraella by determining whether or not they possess genetic variation in features that would have to evolve if they were to add other plants to their diet. We screen for genetic variation in feeding behavior and larval survival. For example, O. communa in North America is found feeding only on ragweed; it ordinarily will starve to death rather than eat goldenrod (the host plant of O. conferta), but it will feed to some extent on marsh-elder, a close relative of ragweed that is the host of another Ophraella species. We crossed many pairs of O. communa and, for each of several larval offspring in each family, measured how much they ate and how well they survived on these and several other plants. Some families of larvae ate marsh-elder and survived on it much better than others; thus the species is genetically variable in these respects and presumably could adapt to this plant if it had to. We detected no genetic variation in responses to goldenrod: many hundreds of larvae refused to feed, and all died. This suggests some internal "constraint," some perhaps almost insuperable block to feeding, that would make adaptation to goldenrod unlikely.

About 30 years ago, Paul Ehrlich and Peter Raven (names you may recognize for other reasons) pointed out that closely related insects tend to feed on closely related plants. Our studies provide some insight into the reasons for this regularity. The patterns of evolution of insect diets (and of many other features of animals and plants) show that evolution is not random: some paths are more likely than others. We suggest that genetically determined potentialities and constraints contribute to this non-randomness. In the future, we intend to see if, indeed, these insects can rapidly adapt to different, related plant species, by imposing natural selection in the lab: we will be challenging experimental populations to adapt or die.