Simple `model' organisms have made major contributions to our molecular understanding of learning and memory; for example, early Drosophilascreens for learning mutants identified flies that displayed problems in olfactory learning. Subsequently, the genes dunce and rutabaga, responsible for the flies' impaired ability, were found to encode a cyclic AMP phosphodiesterase and adenylate cyclase, respectively. These findings implicated cyclic AMP as a major player in memory formation. Since then, elegant genetic mapping techniques have localised memory formation within the Drosophila brain to a pair of three-lobed structures called `mushroom bodies'. Genetic ablation of these structures is known to produce flies that have trouble with certain classes of learning tasks. In a recent Science paper, Guillaume Isabel and co-workers have dissected the role of such ablations in different classes of memory and come up with a class of brain mutation in which, paradoxically, learning tasks actually reduce memory consolidation.

The paper uses two distinct protocols for olfactory learning. In the`short' protocol, odour exposure is accompanied by 12 mild electric shocks;this is thought to produce short-term memory (STM), followed by medium-term memory (MTM), which is eventually stabilised into anaesthesia-resistant memory(ARM). There are mutants for each step; dunce and rutabagainhibit STM, amnesiac inhibits MTM, and radish inhibits ARM. The `long' protocol, by contrast, uses multiple learning sessions to produce long-term memory (LTM). How does LTM correspond to ARM? Traditionally, both derive from MTM and coexist for 24 h after conditioning. The authors show that in alpha lobes absent (ala) flies (which lack one of the three lobes of the mushroom body), memory was initially similar with both protocols. However, memory performance after the `long' protocol was actually worse than for the `short' protocol at the 5 h time point. This implies that,in ala flies (unlike wild-type), the more that flies are trained, the less they seem to remember! Loss of LTM is not surprising, as alaflies lack the necessary neuronal projections, but the loss of ARM requires a re-evaluation of traditional models. The data suggest that, far from ARM and LTM co-existing, LTM-inducing protocols are actually antagonistic to ARM formation.

To further understand the nature of ARM, which had not previously been proved to be associated with mushroom bodies, synaptic transmission was blocked in each of the three lobes of the mushroom body in turn. The results suggested that ARM depended on transmission in the alpha and beta, but not gamma, lobes. Thus, LRM and ARM reside in the same brain regions and could interact antagonistically. Intriguingly, the authors then further overturned the orthodoxy by showing that rutabaga mutants displayed normal ARM,implying that at least a part of ARM is not dependent on cyclic AMP signalling.

This then leaves us with a very different model of olfactory learning in Drosophila. Instead of all such learning moving through the cyclic AMP-dependent STM and MTM pathways, then diverging into LTM and ARM, which coexist in the brain, the authors suggest that the STM/MTM/LTM pathway is different from the ARM pathway but that both converge on, and antagonise each other within, the alpha and beta lobes of the mushroom bodies. It seems that learning and memory, even in this very simple organism, are not as simple as we had supposed.

Isabel, G., Pascual, A. and Preat, T. (
). Exclusive consolidated memory phases in Drosophila.