Report

Localization of a Short-Term Memory in Drosophila

See allHide authors and affiliations

Science  28 Apr 2000:
Vol. 288, Issue 5466, pp. 672-675
DOI: 10.1126/science.288.5466.672

Abstract

Memories are thought to be due to lasting synaptic modifications in the brain. The search for memory traces has relied predominantly on determining regions that are necessary for the process. However, a more informative approach is to define the smallest sufficient set of brain structures. The rutabaga adenylyl cyclase, an enzyme that is ubiquitously expressed in the Drosophila brain and that mediates synaptic plasticity, is needed exclusively in the Kenyon cells of the mushroom bodies for a component of olfactory short-term memory. This demonstrates that synaptic plasticity in a small brain region can be sufficient for memory formation.

The localization of memory traces has occupied neuroscientists throughout this century (1). Approaches have ranged from surgical ablation to mapping localized necessary gene expression in transgenic animals (2,3). Until recently, attempts to localize a memory trace have relied mainly on determining necessary brain regions (4). However, in a highly integrated network, other components besides the one being studied may also be necessary.

In insects, much attention has been paid to the mushroom bodies as the site for olfactory learning (3, 5–8). InDrosophila, they are made up of about 2500 intrinsic neurons (Kenyon cells), receive multimodal sensory input, preferentially from the antennal lobe to the calyx, and send axon projections to the anterior brain where they bifurcate to form the α/β, α′/β′, and γ lobes (9). Noninvasive intervention techniques can provide mushroom body–less flies. In most respects, these flies show remarkably normal behavior but are deficient in olfactory learning (5). Genes important for olfactory memory have elevated expression levels in the mushroom bodies (6, 8). Additionally, the mushroom bodies are necessary for context generalization in visual learning at the flight simulator and the control of spontaneous walking activity (10, 11).

The rutabaga (rut) gene of Drosophilaencodes a type I Ca2+/calmodulin-dependent adenylyl cyclase (AC). Regulated synthesis of cyclic adenosine 3′,5′-monophosphate by a type I AC (through Ca2+ and heterotrimeric G-protein signaling) is important for learning and synaptic plasticity throughout the animal kingdom (4, 6, 8,12–16). The Drosophila rut mutation affects all learning paradigms tested to date and has abnormal synaptic plasticity at the larval neuromuscular junction (4,6, 8, 12, 14,15).

Our approach is built on the assumption that synaptic plasticity is impaired in general in rut mutants and that it is this cellular defect that causes the various learning deficits. Restoringrut AC in a spatially restricted fashion in a defined set of neurons would furnish synaptic plasticity to only those cells. If in such flies a learning task is rescued, the corresponding memory trace is mapped to the set of neurons expressing the gene, or a subset of these.

Olfactory short-term memory was tested with the apparatus of Tully and Quinn (6). In this assay, flies are sequentially exposed to two odorants, one of which is paired with electric shocks (17). Shortly after training, ∼95% of wild-type flies prefer the odorant not accompanied by punishment. Mutantrut flies show significantly lower memory scores (6, 8) (Fig. 1).

Figure 1

The rut mutant defect in olfactory short-term memory can be rescued with arut + cDNA in several GAL4 enhancer trap lines. Memory was measured about 2 min after classical conditioning (17). Performance indices (PIs) of rutmutant flies (white bar) and rut mutant flies with either a P[UASGAL4-rut +] or GAL4 enhancer trap element (thin diagonal striped bars) were significantly different from wild-type flies (dark gray bar;P's < 0.0005). There was no significant difference between rut mutant flies' PIs rescued with GAL4 enhancer trap elements 247, c772, 30y, 238y, and H24 and the P[UASGAL4-rut +] compared with wild-type flies (dark gray and thick diagonal striped bars, respectively; P's > 0.05). Mutantrut flies' performance was rescued with GAL4 element 201y and the P[UASGAL4-rut +] (P < 0.05) but was also significantly lower than the performance of wild-type flies (P < 0.005). GAL4 enhancer trap lines c232, 189y, and 17d with a P[UASGAL4-rut +] did not rescue the rut mutation (P's > 0.05). Wild-type flies heterozygous for GAL4 enhancer trap elements c232, 189y, and 17d were not significantly different from wild-type flies (dark gray and cross-hatched bars; P > 0.05). Bars represent mean PIs; errors are SEMs; n = 6 for all genotypes.

To test whether the olfactory learning defect of the rutmutant was rescuable, we combined in the rut mutant a P-element expressing a wild-type rut cDNA under the control of a GAL4-sensitive enhancer P[UASGAL4-rut +] with a driver transgene P[elav-GAL4] expressing the yeast transcription factor in all neurons (18, 19). This pan-neuronal expression of rut AC partially restored olfactory learning in the rut mutant (20). The incomplete rescue could be due to insufficient expression levels of the P[UASGAL4-rut +] transgene, a dominant negative effect of the P[elav-GAL4] element, or a negative effect of ectopically expressing this transgene.

Several GAL4 enhancer trap lines were selected for local rescue because of their expression patterns (see below). Mutant rut flies with the enhancer trap GAL4 elements 247, c772, 30y, 238y, and H24 in combination with the P[UASGAL4-rut +] transgene showed memory scores statistically indistinguishable from wild-type flies (Fig. 1). The GAL4 line 201y partially rescued the rutlearning defect. Finally, rut mutant flies with three other GAL4 enhancer trap elements (c232, 189y, and 17d) and the P[UASGAL4-rut +] effector gene hadrut mutant–like short-term memory scores.

Four of the nine enhancer trap lines were previously used to study olfactory learning after locally expressing a constitutively activated G-protein α subunit (Gαs*) (3). In the present experiments, the magnitude of rescue was similar to the suppressive effect of the Gαs* protein in the respective lines. In c232, Gαs* had no effect, in 201y, suppression was about 50%, whereas in c772 [the same expression pattern as c747 in (3)] and 238y, suppression was nearly complete.

Neither the rescuing GAL4 enhancer trap lines without P[UASGAL4-rut +] nor the P[UASGAL4-rut +] line without driver had a dominant rescue effect. Nor did the nonrescuing GAL4 enhancer trap inserts have a negative effect on wild-type flies (Fig. 1). Thus, it is the specific interaction of the GAL4 enhancer trap element with the P[UASGAL4-rut +] effector that can rescue the memory defect in rut mutant flies.

Control experiments were conducted with naı̈ve wild-type,rut mutant, and potentially rescued rut mutant flies (Table 1) to assure that none of the memory scores were due to changes in shock reactivity or perception of the odorants. All genotypes avoided electric shocks at similar levels. Although 189y and 17d had somewhat reduced shock reactivity, it was similar to that of c772, which showed a wild-type–like learning score. Thus, these shock reactivity scores cannot be responsible for the low memory scores. In addition, all genotypes tested avoided the aversive odorants used in the training protocol.

Table 1

The rescue of the rutabaga olfactory phenotype does not alter electroshock or odorant sensitivity. Wild-type, rutabaga (rut) mutant, and resuced flies were tested for responses to both the electroshock and the odorants used in the learning experiments. There were no significant differences between wild-type flies and any mutant or rescuedrut flies in either assay. The genotype of the rescued flies is rut/Y; GAL4 line/+; P[UASGAl4-rut +]/+. Means of six experiments per genotype are shown plus or minus the SEM.

View this table:

To determine what brain structures are minimally sufficient for olfactory short-term memory, we examined the expression patterns of the rescue and nonrescue GAL4 enhancer trap lines (21). Serial sections showed that the common structure labeled in all rescuing GAL4 lines was specifically the mushroom bodies (Fig. 2). Indeed, comparing expression patterns of rescuing and nonrescuing lines indicates that the γ lobes may be especially important. In contrast to the GAL4 lines used in (3), the rescuing line 247 (22) lacks expression in the median bundle. The latter, therefore, is not part of the set of minimally sufficient neurons.

Figure 2

Spatial expression patterns in the brain of rescuing and nonrescuing GAL4 enhancer trap lines with P[UASGAL4-TAU] as a reporter (21). Serial frontal brain sections were examined. Schematic drawings and sections at the level of the γ lobes, α/β and α′/β′ lobes and median bundle (meb), ellipsoid body (eb) and peduncles (ped), and calyces (ca). The first six lines rescued the rut odorant learning defect and showed common expression in the γ lobes of the mushroom body. GAL4 line 247 (22) showed detectable expression only in the mushroom bodies. Lines c772, 30y, 238y, H24, and 201y all showed expression outside the mushroom bodies as well. Lines c772, 30y, 238y, and H24 were expressed in the ellipsoid body and antennal lobes (al). Lines c772 and 30y also showed fan-shaped body expression (20). The bottom three lines did not rescue therut defect and showed little or no expression in the γ lobe of the mushroom body. Line c232 expression was largely restricted to the ellipsoid body. Line 189y showed expression in the ellipsoid body and the mushroom body α/β lobes and faint expression in the mushroom body γ lobe. Line 17d expression was restricted to the α/β lobes of the mushroom bodies and the median bundle. Some GAL4 enhancer trap lines' expression patterns have been previously described (4, 11, 27). Dorsal is up; e, esophagus; an, antennal nerve; Kcb, Kenyon cell bodies. Scale bar, 50 μm. The color outside the mushroom bodies and ellipsoid body in lines 247, c232, and 17d is roughly representative of background staining.

On the basis of the current model of how type I ACs function in synaptic plasticity (16) and on the connectivity of the mushroom bodies (19, 23,24), the short-term memory trace of odors is localized to a single level in the olfactory pathway: the presynaptic sites in the Kenyon cells contacting extrinsic output neurons and possibly other Kenyon cells in the peduncle and lobes. Modulating neurons carrying the reinforcer must project to the peduncle or lobes and contact presynaptic endings of Kenyon cells there. Norut-dependent synaptic plasticity is required in the antennal lobe or calyx for olfactory learning.

Different brain structures are involved in different learning tasks. In the heat box paradigm, the median bundle, antennal lobes, and ventral ganglion are sufficient for rescue of rut-dependent short-term memory (4, 15, 25). To find olfactory and heat box memory at different locations was not unexpected, as mushroom body–less flies do well in heat box learning (26). The task-specific rescue in different GAL4 lines strongly supports the claim that it is the spatial distribution of the rut AC that matters.

Several open questions remain. Our conclusions refer only torut-dependent synaptic plasticity. Although unlikely, the 60% short-term memory remaining in rut mutant flies may reside outside the mushroom bodies. Second, our current understanding of the role of type I ACs in synaptic plasticity and learning is not complete. Third, the P[UASGAL4-TAU] reporter was used to visualize GAL4 expression patterns, and coincidence with P[UASGAL4-rut +] is inferred. Fourth, temporal control of transgene expression in these lines is not yet possible, leaving the faint possibility that the behavioral rescue in some cases might be due to developmental expression. Finally, whether all memory traces of odors reside in the mushroom bodies and how memory traces of odors are organized within the mushroom bodies await further investigation.

The technique of restoring synaptic plasticity in minimally sufficient brain regions has, in two cases, revealed simple, locally confined memory traces. This result is probably due to the simplicity of the learning tasks, requiring the animal to store a single sensory modality for a binary orientation response. It will be of considerable interest to map memory traces of more complex learning paradigms.

  • * These authors contributed equally to this work.

REFERENCE AND NOTES

View Abstract

Navigate This Article