Localization of Long-Term Memory Within the Drosophila Mushroom Body

See allHide authors and affiliations

Science  02 Nov 2001:
Vol. 294, Issue 5544, pp. 1115-1117
DOI: 10.1126/science.1064200


The mushroom bodies, substructures of theDrosophila brain, are involved in olfactory learning and short-term memory, but their role in long-term memory is unknown. Here we show that the alpha-lobes-absent (ala) mutant lacks either the two vertical lobes of the mushroom body or two of the three median lobes which contain branches of vertical lobe neurons. This unique phenotype allows analysis of mushroom body function. Long-term memory required the presence of the vertical lobes but not the median lobes. Short-term memory was normal in flies without either vertical lobes or the two median lobes studied.

The organization of theDrosophila brain, which shows highly organized and specialized structures despite its small size, in combination with its sophisticated behavioral repertoires and powerful molecular genetic tools render this organism a model of choice for the study of integrative brain functions, such as associative learning and memory. The mushroom bodies (MBs) constitute a prominent bilateral structure of the insect brain. The MBs are formed and rearranged sequentially during embryonic and postembryonic development (1–3). In adult Drosophila, they are composed of about 5000 neurons, which receive, through the calyx, olfactory inputs from the antennal lobes. The MBs are essential for associative learning and memory (4–6). Several proteins involved in learning and short-term memory are detected at high levels in the MBs (7), and chemical ablation ofDrosophila MBs abolishes olfactory learning (6). Synaptic transmission from the MBs is required for retrieval of short-term memories but not for acquisition or storage (8, 9). With intensive and spaced training,Drosophila can also display long-term memory (LTM), which depends on protein synthesis after the conditioning paradigm (10). We have now tested whether Drosophila MBs are involved in LTM formation.

Three categories of MB intrinsic neurons (Kenyon cells), associated with five sets of lobes, have been described (Fig. 1A) (1, 11). Two types of neurons branch to give rise to a vertical and a median lobe (α/β lobes and α′/β′ lobes, respectively). The third type composes the median γ lobe. Uniquely identifiable efferent neurons originate from specific parts of the medial and vertical lobes, and send their axons to characteristic regions of the forebrain (12). Afferents from the forebrain also invade specific parts of the lobes (12). The implication of this architecture, which also characterizes other insect MBs (13), suggests that the lobes are not identical and may support quite distinct functions.

Figure 1

ala mutants show specific MB defects. (A) Composite confocal images of wild-type CS adult mushroom body (25). Expression of the UAS-mCD8-GFPtransgene driven by the P insertion GAL4-OK107 allows visualization of the whole mushroom body. Three sets of neurons generate five axonal lobes. The γ lobe is outlined in white, the α and β lobes are outlined in red, and the α′ and β′ lobes are outlined in yellow. The color code is conserved in (B) and (C). The median bundle is also revealed with GAL4-OK107. Scale bar, 40 μm. (B) alaE13 MBs without β β′ lobes. (C) alaE13 MBs without α α′ lobes. In these flies, the β and β′ lobes cross the midline.

Genetic characterization of the alpha-lobes-absent(ala) mutant, which shows abnormal MBs, was previously reported (14). The original mutation corresponds to the insertion of a P-element and is recessive. Phenotypic revertants as well as new mutant alleles were produced by excision of the P-element (14). We reassessed the ala brain defect using the GAL4-OK107 enhancer-trap line (15), which labels all five lobes (Fig. 1), and the antibody to FasciclinII (FasII), which strongly labels α/β lobes and weakly labels γ lobes (11, 16). Three batches of brain analyses showed that 10.5 ± 2.8% of ala individuals possessed all five lobes in both hemispheres, 36 ± 2.4% lacked the β and β′ lobes in both hemispheres, and 4.5 ± 1.1% lacked α and α′ vertical lobes. The remaining flies showed different combinations of phenotypes in the left and right hemisphere.ala flies without vertical lobes also showed a fusion of the left and right β and β′ lobes (Fig. 1C). This fusion phenotype is also observed in a fraction of ala flies with all lobes present (3.6 ± 1.4% of total) (16). A similar phenotype distribution was observed in ala/Df(1)4b18individuals (17). γ lobes appeared to be normal inala mutants. This observation was reinforced by the fact that in second instar larvae ala mutants possessed normal vertical and median γ projections (18).

The ala mutant was trained to associate an odor with electric shocks by using three different experimental paradigms (19, 20): (i) a single training cycle protocol to induce short-term memory (21); (ii) an intensive spaced conditioning protocol, consisting of 10 individual training sessions with a 15-min rest interval between each session, to induce LTM (10); and (iii) a massed conditioning protocol, consisting of 10 consecutive training sessions without rest, to induce 24-hour memory but not protein-synthesis dependent LTM (10).

In spite of the severe brain phenotype, 1-hour, 3-hour, 24-hour spaced, and 24-hour massed memory performances were normal in alamutant (Fig. 2A). We next tested whether the absence of specific MB lobes results in specific memory defects. Several thousand ala flies were trained and tested in the three conditioning protocols. These protocols are discriminative because during the test flies tend to avoid the previously shock-associated odor and move toward the second odor (not previously associated with shock). Flies that made the correct choice were collected separately from those that made the wrong choice, and their brains were analyzed by staining with antibody to FasII (16). This allowed us to correlate a memory performance index (PI) with each category of ala brain defect (Fig. 2B). There are three advantages of such an experiment. First, it allowed a regional analysis of the MB function, which could not be performed with traditional genetic tools. Second, all flies came from the same siblings and were treated equally under conditioning and testing conditions. Third, the experiment was performed blindly with respect to brain defects. We also took advantage of the fact that flies with different memory abilities are known to perform independently of each other when mixed for conditioning and testing (21,22).

Figure 2

LTM is abolished in absence of MB vertical lobes. (A) ala mutant stock has normal olfactory memory. PIs were measured 1 hour and 3 hours after one conditioning cycle, and 24 hours after 10 cycles of training with (spaced) or without (massed) 15-min intervals between each training cycle (19, 20). Memory scores of ala flies (gray bars) were not significantly different from wild-type flies (black bars) at all times tested (t test). Bars represent means of PIs; errors are SEMs (standard errors of the mean);n = 8 to 12 groups. (B) Memory scores of three categories of ala flies: all lobes present (black bars), β β′ lobes absent (gray bars), and α α′ lobes absent (striped bars). LTM was significantly decreased in flies without α α′ lobes in comparison with flies with all lobes present (χ2 P = 0.0012), shown by asterisks (26). On the contrary, flies with β β′ lobes absent were not affected (P > 0.5). The 24-hour memory increased after massed training in flies with α α′ lobes absent was not significant (P > 0.25) (27). 1530 flies were tested at 3 hours (n = 8 groups), with a distribution of 96 flies with all lobes present, 599 flies with β β′ lobes absent, and 99 flies with α α′ lobes absent. After spaced training, 1273 flies were tested (n = 10 groups), with a distribution of 111 flies with all lobes present, 480 flies with β β′ lobes absent, and 53 flies with α α′ lobes absent. After massed conditioning, 1607 flies were analyzed (n = 11 groups), with a distribution of 204 flies with all lobes present, 503 flies with β β′ lobes absent, and 44 flies with α α′ lobes absent.

In agreement with our first analysis, 3-hour memory was not reduced in flies lacking either α α′ or β β′ lobes, confirming that these structures are not required for proper odor or shock perception (6). LTM was normal in the absence of β β′ lobes but fully abolished in the absence of α α′ lobes, despite the fact that these two sets of lobes originate from the same neurons during development. In contrast, 24-hour memory generated after massed training was not reduced in flies lacking vertical lobes. The LTM deficit associated with the absence of vertical lobes was not detected immediately (Fig. 2A) because these flies represent less than 5% of the ala population. The reduction in LTM cannot be attributed to the fused β β′ defect because ala flies with all lobes showed normal scores when their β β′ lobes were fused (23).

In Drosophila, the dendrites of efferent neurons leading from the MBs reside in the lobes or at the junction between the lobes and the pedunculus, but not posteriorly in the pedunculus itself (12). The fact that short-term memory was not affected by the lack of α α′ or β β′ lobes has two possible explanations. First, functional redundancy of neural connections may have prevented us from observing short-term memory defects in the absence of two of these four lobes. This view is supported by the fact that memory loss was observed after transient inactivation of neurons that form α/β lobes (9). Alternatively, γ lobes, which are normal inala flies, could represent the main neuronal substrate for short-term memory. This interpretation is supported by the fact that the rutabaga (rut) learning defect can be rescued by tissue-specific rut + expression in γ lobes (24).

This study shows that the MBs play an essential role in LTM formation. Furthermore, 24-hour memory after spaced training was fully abolished in the absence of vertical lobes, suggesting that in normal flies a unique molecular and cellular pathway underlies this memory phase. In a previous study, this phase was not completely eliminated after protein synthesis inhibition (10). At the time, this observation was interpreted to mean that LTM represented only half of the memory displayed at 24 hours after spaced training. The present results, however, suggest that residual ribosomal activity was the cause of this partial inhibition. In the larva, vertical axon branches of neurons associated with the medial γ lobe might fulfill a similar role to adult vertical projections in long-term information processing. Loss of these projections during metamorphosis could erase some long-term information specific to the larval stage. In adultDrosophila, few efferent neurons from the γ lobe extend to around the α α′ lobes (12), which could represent a pathway that converts information from short-term to long-term memory. Alternatively LTM may form independently of short-term memory. Further analyses must be performed to resolve that issue and to determine the role of α and α′ lobes in LTM formation.

  • * To whom correspondence should be addressed. E-mail: preat{at}


View Abstract

Navigate This Article