In 1988, John Cairns published a study claiming that Escherichia coli cells could respond to selection by directing advantageous mutations to specific genes. This process — adaptive mutation — presented a fundamental challenge to established views about the way evolution works. Not surprisingly, many geneticists responded by investigating whether this observation could be explained by more conventional mechanisms. Heather Hendrickson and colleagues now present just such a model for adaptive mutation, which might also be relevant to genetic changes that occur in cancer.

The experiment that started off the debate involved an E. coli strain that carried an F′ plasmid with a mutant lacZ gene. When the strain was incubated for several days on lactose-containing medium — lactose cannot be utilized by the lac− mutants — lacZ revertants slowly accumulated at many times the expected frequency. It looked like mutations were being directed to occur in the very gene required to get the cells growing again. Subsequently, the revertant colonies were shown to carry mutations elsewhere in the genome. This suggested that, in this system, selection induces a mutable state that increases the occurrence of useful mutations.

Hendrickson et al. propose a rather different model. The lacZ mutation is known to be 'leaky', so the authors suggest that, if the mutant gene is duplicated, there would be enough activity to support slow growth. Further amplification of the lacZ gene during growth of individual clones would confer an additional selective advantage. Recombination products derived from the repeated copies would then cause the induction of the SOS DNA repair system, leading to increased mutability and to occasional reversion of the lacZ mutation. Finally, loss of mutant copies of the gene will turn off the SOS system, return the level of mutability to normal and leave cells with a stable lac+ phenotype. So, adaptive mutation reflects a sequence of events, each of which confers a selective advantage.

The authors tested several predictions of the model, all of which were upheld: for example, when amplification was inhibited, reversion did not occur; the lac+ phenotype of colonies with an amplified mutant lacZ gene was more unstable, relative to colonies that carry a reversion mutation; and mutations that abrogated the SOS response reduced, but did not abolish, the recovery of revertant mutations.

This model for adaptive mutation provides a neat explanation for the phenomenology of the lac/F′ system, although this is certainly not the end of the story. However, the authors also discuss how this model could account for the accumulation of multiple mutations in tumour cells. A key point is that an initial duplication event provides sufficient selective advantage to initiate a clone within which secondary mutations can arise. So, although the concept of 'adaptive' mutation might gradually be eroded, the provocative claims associated with the initial observations have served to stimulate some exciting new ideas.