The transformation of butterflies

What actually happens inside the pupa — and why biologists still argue about it

A Few biological transformations are described so often and understood so poorly as the one that converts a caterpillar into a butterfly. The popular account, repeated in textbooks and nature documentaries alike, holds that the larva dissolves inside its pupal case into an undifferentiated liquid from which the adult is somehow reassembled. The image is memorable and it is substantially wrong. Considerable dissolution does occur, and a pupa punctured at the right moment will indeed release something disconcertingly fluid, but the material is not raw. Distributed through it are structures that were present in the caterpillar from the moment it hatched, and it is these, rather than any process of spontaneous reconstruction, that determine what emerges.

B These structures are the imaginal discs: small pockets of cells, set aside during embryonic development, that remain dormant while the larva feeds and grows. Each corresponds to a component of the adult — a wing, an antenna, a leg, an eye — and each has been mapped in sufficient detail that a researcher can predict which disc will produce which appendage before the pupa is formed. What the dissolution supplies is not a template but a resource. Larval tissues that have no adult equivalent are broken down, and their constituents are consumed by the discs as they proliferate. The animal, in effect, digests the parts of itself it will no longer require in order to finance the parts it has not yet built.

C Understood in this way, metamorphosis is less a demolition than a scheduled handover, and the scheduling is chemical. Two hormones govern the sequence: ecdysone, which triggers each moult, and juvenile hormone, whose presence instructs the insect to remain immature. As long as the second circulates in quantity, a moult produces another and larger larva. When its concentration falls below a threshold, the same trigger produces a pupa instead, and with it the reorganisation that has been deferred since the egg. The elegance of the arrangement lies in its economy: one signal decides timing, the other decides outcome, and no new machinery is required to move an animal from one body plan to an entirely different one.

D That this account is now well established has not prevented the appearance of far more dramatic explanations. In 2009 a paper in a leading American journal proposed that the larval and adult stages had never belonged to a single lineage at all: that at some remote point an ancestral insect had hybridised with a velvet worm, acquiring a second genome, and that the caterpillar was the expression of one genome and the butterfly of the other. The hypothesis had the considerable virtue of explaining why the two stages appear so unrelated. It also generated an immediate and unusually forceful response, and the episode is worth examining, because the objections raised against it illustrate what a biological explanation is expected to supply.

E The criticisms were of two kinds. The first was procedural: the paper had reached publication through a route that allowed a senior member of the academy to sponsor a submission with limited external scrutiny, and a number of researchers argued that the mechanism had been used to bypass the judgement of specialists who would have rejected it. The second, and more damaging, was evidential. A fusion of two distant genomes should leave conspicuous traces — in genome size, in gene sequences, in the relationships that emerge when many species are compared — and the traces were absent. Insect genomes had by then been sequenced in quantity, and they contained nothing that could plausibly be assigned to a velvet worm. An explanation that accounts for a puzzle while predicting evidence nobody can find is not, in the end, an explanation.

F Meanwhile a quieter experiment had complicated the orthodox picture in a more productive direction. Caterpillars of the tobacco hornworm were exposed to a particular odour and simultaneously given a mild electric shock, a procedure that reliably teaches them to avoid the smell; in the assessment that followed, roughly three-quarters of the trained larvae chose the untreated route. The animals were then permitted to pupate, and the surviving adults were tested a month later, after their nervous systems had been comprehensively remodelled. The moths avoided the odour too. Something learned by one body had been retained by another that had been substantially rebuilt around it — a result that says less about memory than about continuity, and that sits awkwardly with any description of the pupa as a fresh start.

G The question that remains is not what happens inside the case but why the arrangement should have evolved at all, and here the disagreement is genuine and unresolved. The dominant proposal treats the larva as an elaboration of an ancestral stage that once passed quickly inside the egg and was gradually extended into an independent feeding phase. Its attraction is that it explains the arrangement's persistence: a caterpillar and a butterfly compete for neither the same food nor the same space, so a single species can occupy two ecological positions at once, which is a considerable advantage. Its weakness is that the fossil record of soft-bodied larvae is close to non-existent, and a hypothesis about origins that cannot be tested against origins remains, however plausible, a hypothesis. What can be said with confidence is that the animal in the pupa is neither dissolved nor duplicated. It is being rebuilt, from materials it has saved, according to instructions it has carried since the egg.