A Cost of Long-Term Memory in Drosophila

See allHide authors and affiliations

Science  20 May 2005:
Vol. 308, Issue 5725, pp. 1148
DOI: 10.1126/science.1111331


Two distinct forms of consolidated associative memory are known in Drosophila: long-term memory and so-called anesthesia-resistant memory. Long-term memory is more stable, but unlike anesthesia-resistant memory, its formation requires protein synthesis. We show that flies induced to form long-term memory become more susceptible to extreme stress (such as desiccation). In contrast, induction of anesthesia-resistant memory had no detectable effect on desiccation resistance. This finding may help to explain why evolution has maintained anesthesia-resistant memory as another form of consolidated memory, distinct from long-term memory.

Two distinct forms of consolidated associative memory are known in Drosophila: long-term memory (LTM) and so-called anesthesia-resistant memory (ARM) (1, 2). In the context of Pavlovian aversive olfactory learning, LTM will only form after repeated conditioning events separated by rest intervals (a spaced protocol), whereas ARM will also form if consecutive conditioning events immediately follow one another (a massed protocol) (2). LTM is more stable, but unlike ARM, its formation requires protein synthesis (2). LTM is thus likely to be energetically costly, but it is not clear if the additional energy demand has any effect on an animal's fitness.

We studied how induction of consolidated memory affects the resistance of adult flies to extreme stress imposed by an absence of food and water. We used an outbred wild-type Drosophila melanogaster line that has been artificially selected for improved memory and shows particularly good LTM (3). We trained the flies to associate an odor with an aversive mechanical shock (associative conditioning) (4). Five consecutive training sessions were either separated by 20-min intervals (the spaced protocol) or followed one another immediately (the massed protocol). Both protocols induced avoidance of the odor previously associated with the shock if flies were assayed 24 hours after conditioning (Fig. 1A). However, the response after the spaced protocol was dependent on protein synthesis (i.e., involved LTM), whereas the response after the massed protocol was not (Fig. 1A). In control experiments, other flies were either exposed to mechanical shocks without any odors (shock only) or exposed to both shocks and odors but not concurrently (nonassociative conditioning), which did not lead to formation of even short-term associative memory (Fig. 1A, red bars) (4).

Fig. 1.

(A) Olfactory memory (mean ± standard error; n = 8 replicate memory scores per treatment) induced by spaced and massed conditioning protocols. In both protocols, associative conditioning induced 24-hour memory in normal flies (CXM–), but treatment with the protein synthesis inhibitor cycloheximide (CXM+) erased the response after the spaced protocol only, showing that only this response is based on LTM. No associative memory was detectable 20 min after nonassociative conditioning (red bars). (B) Time until death in the absence of food and water (mean ± standard error) of flies subject to different conditioning treatments (n = 30 flies per treatment and sex).

We subjected each fly to one of seven conditions: associative massed conditioning (inducing ARM), associative spaced conditioning (inducing LTM), nonassociative conditioning (massed or spaced), shock only (massed or spaced), and untreated control. Directly afterwards, we assayed their individual desiccation and starvation resistance, measured as time until death in the absence of food and water (4). Despite exposure to the same shock and odors, flies conditioned in the associative spaced protocol died on average 4 hours (19%) earlier than flies subjected to nonassociative spaced conditioning (F1,116 = 28.1, P < 10–3) (Fig. 1B). This effect was the same for males and females (interaction F1,116 = 0.1, P = 0.74). This difference might have been due to greater activity induced by associative conditioning, but video tracking of individual flies indicated no difference in locomotor activity (fig. S1). No difference in time to death under stress was observed for flies subject to associative versus nonassociative conditioning in the massed protocol, in which LTM is not formed (F1,116 = 0.7, P = 0.42) (Fig. 1B).

The earlier death of flies subjected to associative spaced conditioning, relative to those subjected to other conditioning treatments, is thus likely to reflect additional strain due to the formation and maintenance of LTM. Given the importance of resistance to desiccation and starvation in natural Drosophila populations (5, 6), this result suggests that long-term memory has not only benefits but also ecologically relevant costs. Whether, and to what extent, natural selection favors improvements of memory will depend on the balance of the costs and benefits (7). Such costs may also help to explain why evolution has maintained ARM as another form of consolidated memory, distinct from LTM: Although LTM is more stable than ARM, it is also apparently more expensive.

Supporting Online Material

Materials and Methods

Fig. S1

References and Notes

View Abstract

Navigate This Article