Evolutionary biology

Déjà vu

Article metrics

A long-term study of fruitflies adds to the evidence that evolution can run backwards. To what extent the genetic underpinnings revert to the original is unclear.

We tend to think of evolutionary change as resulting in something different or new. But evolution going backwards, in which an organism reverts to a previous state, has its own possibilities and puzzles. Teotónio and Rose1 (page 463 of this issue) show that, in captive populations of the fruitfly Drosophila melanogaster, the effects of 100–200 generations of adaptation to a particular environment can be undone in as little as 20 generations when the flies are returned to their original environment. A few traits did not revert, however, even when the flies had enough genetic variation to go all the way back. These cases shed light on why evolution does not necessarily reverse.

Evolution undoing previous evolution has often been thought to be improbable. But one must distinguish between molecular genetic and phenotypic reversibility. In the first instance, the question is whether the original genetic changes have been undone, recreating the ancestral genotype. In the second, the question is whether the organism's morphological, physiological or behavioural traits have reverted to an ancestral state (for example, the re-evolution of wings in a flightless insect), without regard to the genetic basis of that reversion. As in most earlier work on reversibility, Teotónio and Rose look at the evolution of phenotypic traits.

The common view that phenotypic evolution is largely irreversible is codified in Dollo's law, dating from the late 1800s. A broad interpretation of this law is simply that evolution cannot go back to recreate an organism in exact detail. More narrowly, and sometimes recognized as Abel's law or Meyrick's law, it has been interpreted to mean that lost structures (wings, say) are not regained2,3. The narrow sense of Dollo's law certainly has its exceptions. Darwin recognized that domesticated organisms sometimes regain ancestral features when returned to the wild. Many lab experiments also show that when artificial selection on a trait is relaxed, that trait often reverts fully or partially to the ancestral condition4.

A striking case of reversibility is seen in the live, attenuated virus of the Sabin poliovirus vaccine. This virus often reverts to a more virulent form within the vaccinated person; such reversion does not harm the person concerned but can cause polio when transmitted to others who have not been vaccinated or previously infected5.

Teotónio and Rose conducted a large experiment using caged populations of Drosophila that were originally maintained for other purposes. For 25 years, one population has been kept under standard rearing conditions (the 'standard' line). Twenty years ago, a subset of the standard flies was used to start another population that was then selected for late-life fecundity (this line was originally used to test evolutionary theories of senescence). Eleven years ago, flies from the standard line were used to found two additional populations, one being selected for resistance to starvation, the other for intermediate generation time. Finally, eight years ago, a fourth population was started from the standard line, selecting for flies that developed into adults especially quickly.

About three years ago, flies from these four lines were returned to the original standard conditions. Since then, eight different phenotypic traits have been monitored across five replicates of each line: male and female development time; male and female resistance to starvation; early fecundity at high and low population densities; female dry body weight; and female lipid content. Teotónio and Rose observed three different patterns of reversibility: complete, partial and none. Most traits reversed completely, although the time taken for this varied from 20 to 50 generations. Partial reversals were not merely slower versions of complete reversals, because their initial rate of reversal was rapid but then reached a plateau.

Why did some traits — such as fecundity at high population densities — fail to reverse at all or to completely attain the ancestral condition? The limitation could be genetic — that is, genetic variation was absent or exhausted by selection. Or epistasis could have blocked the return. Epistasis is an interaction in the effects of different mutations, such that the consequence of one mutation changes depending on whether another mutation is present or absent.

An example is seen in bacteria that have become resistant to antibiotics. Although mutations that give rise to drug resistance are, by themselves, generally deleterious in the absence of antibiotic, the ending of antibiotic treatment does not ensure the evolutionary reversal to antibiotic sensitivity. This is because drug-resistance mutations often impair essential cellular functions and so are followed by other, compensatory mutations that restore those functions. In the absence of antibiotics, the antibiotic-resistance gene and the compensatory mutations are each singly more harmful than they are in combination. So reverse evolution is blocked because neither gene can undergo reversal on its own6.

To explore similar possibilities in the fruitfly, Teotónio and Rose created hybrid populations between the selected and standard lines. These hybrid populations were genetically variable, and so should have reverted more easily because they lacked many types of genetic barriers to reversion. Correcting for the fact that the hybrid populations started their return from intermediate values, the replicates behaved essentially as the evolved populations did, suggesting that genetic limitations did not explain the lack of reversion. Teotónio and Rose speculate that the form of selection was instead altered by the original evolution — that is, for some traits the nature of selection depended in subtle ways on the genetic background of the population, which changed over the course of the experiment.

But is phenotypic reversal accompanied by genetic reversal? Genetic reversal seems less likely than phenotypic reversal, because there are often several genetic routes to an adaptation, and because mutational biases and epistasis may effectively preclude a reversal back down the original pathway. Genetic reversions have been observed in microbes7,8, but resolution of this question in Drosophila awaits further work. Nonetheless, Teotónio and Rose's study adds to a growing body of body of work on adaptation, in which natural or semi-natural replicates tend to reveal that natural selection has a powerful influence on phenotype. Their work suggests that there is essentially no intrinsic barrier opposing the return to formerly adaptive states, and that reversals of many types of phenotypic traits can occur.


  1. 1

    Teotónio, H. & Rose, M. R. Nature 408, 463–466 (2000).

  2. 2

    Gould, S. J. J. Hist. Biol. 3, 189–212 (1970).

  3. 3

    Minkoff, E. C. Evolutionary Biology (Addison Wesley, Reading, MA, 1983).

  4. 4

    Hill, W. G. & Caballero, A. Annu. Rev. Ecol. Syst. 23, 287–310 (1992).

  5. 5

    Guillot, S. et al. Vaccine 12, 503–507 (1994).

  6. 6

    Schrag, S. J. et al. Proc. R. Soc. Lond. B 264, 1287– 1291 (1997).

  7. 7

    Crill, W. D. et al. Genetics 154, 27–37 (2000).

  8. 8

    Levin, B. R. et al. Genetics 154, 985– 997 (2000).

Download references

Author information

Correspondence to J. J. Bull.

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.