Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Neuron. Author manuscript; available in PMC 2007 December 7.
Published in final edited form as:
PMCID: PMC1779901

Drosophila α/β Mushroom Body Neurons Form a Branch-Specific, Long-Term Cellular Memory Trace After Spaced Olfactory Conditioning


Functional optical imaging showed that odor or electric shock stimuli presented to the fly causes transient calcium influx into the two major axon branches of α/β mushroom body (MB) neurons. One pairing of odor and electric shock stimuli or multiple pairings in a massed configuration did not alter odor-evoked calcium influx. In contrast, animals that received multiple pairings in a spaced configuration exhibited a robust increase in calcium influx into the MB axons when tested at 9 or 24 hr after training but not at 3 hr post-training. This modification occurred only in the α/β branch of the neurons and was blocked by mutation at the amnesiac gene, inhibition of protein synthesis, or the expression of a protein blocker of the transcription factor Creb. Thus, behavioral long-term olfactory memory appears to be encoded as a branch-specific modification of calcium influx into the α/β MB neurons that occurs after spaced training in a protein synthesis-, Creb-, and amnesiac-dependent way.


Memories are formed from the modifications that occur within neurons due to learning. These modifications, or cellular memory traces, can be molecular, biophysical, or cellular in nature. They may involve changes in ion channel properties and distribution which govern the excitability of neurons, growth or retraction of neuronal processes to facilitate the establishment or abolishment of neural connectivity, modifications of downstream cell signaling molecules that alter a neuron’s response to sensory information, and morphological and functional synaptic changes that affect a neuron’s ability to stimulate its synaptic partners. A major goal in neuroscience is to understand the nature of cellular memory traces, the neurons in which they develop, the mechanisms by which they form, their duration, and how the complete set of cellular memory traces within different areas of the nervous system guide behavior after learning.

In Drosophila, olfactory classical conditioning is a robust and well-studied type of learning in which olfactory cues are usually paired with electric shock, such that conditioning leads to learned avoidance behavior of the odor used for conditioning (Tully and Quinn, 1985). Behavioral studies have revealed that different conditioning protocols generate different forms of memory. For instance, a conditioning protocol consisting of a 1 min presentation of an odor (conditioned stimulus, CS+) along with 12 electric shock pulses (unconditioned stimulus, US), followed by a 1 min presentation of a counter odor (conditioned stimulus, CS-) without shock, generates robust initial memory that decays over the period of about 1 day (Tully et al., 1994). A protocol consisting of 5-10 such conditioning cycles with no rest between each cycle (massed conditioning) generates robust initial memory that decays over a few days. A protocol consisting of 5-10 conditioning cycles with a rest between cycles (spaced conditioning) generates a longer-lasting memory of 4-7 days that is dependent upon protein synthesis at the time of conditioning (Pascual and Preat, 2001; Isabel et al., 2004; Perazzona et al., 2004). Since different conditioning protocols generate different forms of memory, they must also generate distinct cellular memory traces that underlie the conditioned behavior.

Some of the cellular memory traces that potentially underlie conditioned behavior have recently been discovered. A single cycle of olfactory classical conditioning generates an immediate but short-lived (7 min) memory trace in the antennal lobe (AL) that is registered as the recruitment of additional sets of projection neurons into the representation of the CS+ odor (Yu et al., 2004). A second memory trace forms after conditioning in the dorsal paired medial (DPM) neurons (Yu et al., 2005). This trace is displayed as increased calcium influx and synaptic transmission in response to the CS+ odor in one branch of the DPM neurons that innervates the vertical lobes of the mushroom bodies (MB). It forms with a delay, appearing 30 min after training and persisting for at least 1 hr, and requires the function of the amnesiac (amn) gene. Dopaminergic neurons that innervate the axons of the MB neurons have been reported to form a calcium-based memory trace immediately after conditioning (Riemensperger et al., 2005) and there is an increase in synaptic protein synthesis in regions of the AL one day after spaced olfactory conditioning (Ashraf et al., 2006). Although it remains unclear whether these memory traces form independently of one another and to what extent that they guide post-conditioning behavior, such studies of olfactory conditioning using cellular functional optical imaging have begun to reveal the nature, duration, and location of candidate cellular memory traces that may influence post-conditioning, olfactory behavior.

Conspicuously missing from the above list of neurons reported to form memory traces after olfactory conditioning are the MB neurons themselves. This is surprising, since a great deal of evidence has accumulated suggesting that these neurons are a major site for the formation and storage of olfactory memory (McGuire et al., 2001; Dubnau et al., 2001; Davis, 2004; Davis, 2005). Here, we present data that defines the elusive MB memory trace. The cellular memory trace is expressed as an increased calcium influx into one class of MB neurons, the α/β neurons, and is generated only by a spaced conditioning protocol, not by single cycle or massed conditioning. The trace is delayed, forming between 3 and 9 hr after conditioning, and intriguingly is axon branch specific, forming only in the αaxon branch of the α/β MB neurons and not in the β branch. In addition, the data suggest that the memory trace is dependent on protein synthesis at the time of conditioning, the activity of the transcription factor, Creb, and the activity of the amnesiac (amn) gene. Thus, the data reveal a delayed, branch specific memory trace that is protein synthesis-, amn-, and Creb-dependent, which is formed by a conditioning protocol that also generates protein synthesis-, amn-, and Creb-dependent long-term behavioral memory.


The α/β MB neurons respond to odor and shock stimuli with calcium influx

To search for a cellular memory trace in the MB neurons, we first tested whether MB axons respond with calcium influx to odors delivered to the antennae and electric shock pulses delivered to the abdomen of Drosophila. Flies carrying the calcium reporter transgene Uas-G-CaMP expressed by c739-Gal4 were prepared for in vivo imaging of brain by mounting them stably under a laser-scanning confocal microscope to detect basal fluorescence and the resulting change in fluorescence following the application of odor or electric shock. The c739-Gal4 transgene is an enhancer detector element that drives expression from Uas-transgenes specifically in the α/β MB neurons using anonymous enhancers in the vicinity of the hr39 gene (unpublished data). This Gal4 driver was chosen because it impairs memory retrieval at 3 and 24 hr after conditioning when used in combination with the conditional blocker of synaptic transmission, Uas-Shits (McGuire et al., 2001; Isabel et al., 2004). The cell bodies of the α/β MB neurons reside in the dorsal and posterior cellular cortex. Each neuron extends a single axon that splits into two branches - the αand β branches - near the anterior edge of the brain (Figure 1A). We collected imaging data across time from thick optical sections at two different depths in the brain in order to visualize calcium influx separately into the α or β branches of the α/β MB neurons (Figure 1A).

Figure 1
The α/β mushroom body (MB) neurons respond with calcium influx into their αand β axon branches when odors or electric shock stimuli are delivered to Drosophila. (A) Cartoon illustrating the Drosophila α/β ...

Stimulation of the flies with the odors 3-octanol (Oct) or benzaldehyde (Ben) elicited significant and reproducible changes in calcium content in both the αand β branches of the α/β MB neurons. Figure 1B illustrates the time course of the percent change in fluorescence (%ΔF/Fo) during odor stimulation. An increase in the %ΔF/Fo was observed shortly after odor onset. The response peaked within one sec and persisted over the course of odor presentation. A representative false color image of the percent change in fluorescence that occurs with odor stimulation for these odors at the peak of the response in both branches of the neurons is illustrated in Figure 1C, with group data for the averages of the peak responses shown in Figure 1D. The group data show a measurable and reproducible change in fluorescence of about 4% in both the αand β branches. Similarly, application to the abdomen of the fly of electric shock pulses of the same intensity, duration, and frequency as those used for behavioral conditioning (Roman and Davis, 2001) produced significant calcium responses in both the αand β branches (Figures 1E and F). These data show that calcium responses to odor and electric shock can readily be detected in the αand β branches of the α/β MB neurons and that the magnitude of the responses between axon branches is similar. It is important to note that the response of brain neurons to electric shock delivered to the abdomen has specificity. Neither olfactory receptor neurons nor GABAergic neurons in the antennal lobe respond to electric shock pulses to the abdomen (Yu et al., 2004), whereas projection neurons in the antennal lobe (Yu et al., 2004), DPM neurons (Yu et al., 2005), and MB neurons all respond to this stimulation. These data therefore show that the α/β MB neurons are activated by odor or electric shock stimulation to the fly and that their responsiveness is registered by calcium influx into the αand β branches using G-CaMP as a reporter.

The α/β MB neurons form a branch-specific, long-term memory trace

The accumulated evidence supporting a role for the MB neurons in olfactory conditioning and the calcium responses of the α/β MB neurons to both odor and electric shock led us to examine the possibility that a cellular memory trace might form in the axon branches of the α/β MB neurons after olfactory classical conditioning. Flies were trained to associate an odor with electric shock in standard behavioral training tubes using three experimental protocols (Figure 2A): (a) a single cycle conditioning protocol (1x forward), (b) a massed conditioning protocol of 5 consecutive single conditioning cycles (5x massed forward), and (c) a spaced conditioning protocol consisting of 5 single conditioning cycles with a 15 min intertrial interval (5x spaced forward). Each conditioning protocol was discriminative using both a CS+ odor paired with electric shock and a CS-odor without electric shock. In addition, we performed the conditioning in a backward fashion, wherein the CS+ odor was presented after the onset of the electric shock stimuli. Backward conditioning protocols generally fail to produce the robust excitatory behavioral conditioning that occurs with forward conditioning (Rescorla, 1988). For each group of flies that received olfactory conditioning, we tested a group of naïve flies that were passed through the same manipulations except that they did not receive the associative conditioning of odor with electric shock stimuli. The performance of each naïve group was subtracted from the performance of the conditioned group with which it was paired in order to obtain a measure (ΔPerformance Index, ΔPI, see Experimental Procedures) of the performance gains due to associative conditioning with the CS+ odor.

Figure 2
Conditioning protocols: Spaced conditioning produces robust 24 hr performance. (A) Diagram illustrating the seven conditioning protocols that were used for these experiments. Flies carried one copy of c739-Gal4 and one copy of Uas-G-CaMP. Naïve ...

Flies conditioned using the spaced forward conditioning protocol exhibited a significantly higher level of associative memory when tested at 24 hr after conditioning than flies conditioned using the 1x forward or 5x massed forward conditioning protocols (Figure 2B). This was observed using either Oct or Ben as the CS+ odor. In addition, flies conditioned using the 1x forward conditioning protocol exhibited associative memory indistinguishable from flies conditioned using the 5x massed forward conditioning protocol. In each case, the corresponding backward conditioned group failed to show any significant performance gains. These data are consistent with prior studies that established these conditioning protocols and indicate that spaced forward conditioning induces more robust 24 hr memory than either 5x massed forward or 1x forward conditioning (Tully et al., 1994).

Some of the flies conditioned using the various conditioning protocols (Figure 2A) using Oct as the CS+ were removed immediately prior to behavioral testing at 24 hr and mounted for imaging the calcium responses to odor. All of the groups, save one, exhibited CS+ odor-responses in the αbranch that quantitatively are typical of naïve flies (Figure 1D) and these responses were indistinguishable from each other for the CS+ odor of Oct (Figures 3A and D). These included the 1x (forward and backward), 5x massed (forward and backward), and 5x spaced backward groups. The only exception was the 5x spaced forward group, which exhibited an odor-elicited calcium response nearly twice as large as any other group. This modification in odor response properties resulted from an increased response magnitude during the odor stimulation without altering the length of the response, as illustrated by the response time course to odor for the representative 5x spaced forward and 5x spaced backward groups (Figure 3B). This modification in odor response properties was specific for the αbranch of the α/β MB neurons and to the CS+ of Oct, since it was not detectable in the β branch with the CS+ odor, or with the CS-of Ben in either the αor β branch. Similar experiments using Ben as the CS+ and Oct as the CS-confirmed that spaced forward conditioning significantly modifies the CS+-elicited calcium response at 24 hr after conditioning only in the α branch of these neurons (Figures 3C and D).

Figure 3
A branch-specific, long-term memory trace forms in the αbranches of the α/β MB neurons after spaced forward conditioning of Drosophila. (A) Response to the CS+ and CS-odors 24 hr after conditioning. A significant increase in %ΔF/F ...

Five considerations contribute to the conclusion that the axon branch specificity observed for the memory trace is not an artifact of the imaging procedures. First, for optical imaging we utilized a wide open pinhole, producing an optical slice calculated to be 32 micrometers with the numerical aperature of the objective lens and at the excitation wavelength used for these experiments. This eliminates the possibility that we inadvertently and consistently sampled a specific optical plane from the β lobe that failed to form the calcium trace, since the depth of the β lobe along the dorsal/ventral axis is approximately 25 micrometers. Second, the basal fluorescence from the β branches was of the same magnitude as that for the αbranches using the same gain and laser power for excitation (Figure 1C and E). Thus, distribution of the G-CaMP reporter between branches was evenly distributed between branches so that responses in both branches could be observed (Figure 1C-F). Third, the value of interest in evaluating the formation of a memory trace is the %ΔF/Fo, the percent change in fluorescence. If basal fluorescence is observed, then significant changes from this baseline may be detectable. Fourth, the quantitative change in %ΔF/Fo after spaced forward conditioning in the αbranch is nearly twice that observed after other conditioning protocols. In contrast, the quantitative change in %ΔF/Fo after spaced forward conditioning in the β branch fails to exhibit even a trend of an increase relative to other conditioning protocols (Figure 3A). Indeed, there is a trend of an increase with spaced backwards conditioning over spaced forward conditioning in the β branch (Figure 3A and C). Fifth, recent experiments have detected a memory trace in the γlobe of the MBs that forms under certain conditions (not shown), indicating that memory traces can be detected in structures residing at the same imaging depth and oriented in the same direction as the β branches. These considerations together strongly argue for branch specificity in memory trace formation, although we cannot eliminate the unlikely possibility that differences in branch orientation between the two lobes relative to the optical axis may impose an observational bias.

The branch-specific cellular memory trace forms between 3 and 9 hr after conditioning

We then assayed flies at various times after spaced forward conditioning to delimit the time during which the cellular memory trace forms. Behavioral memory decays after spaced forward conditioning by 9 hr to the level observed at 24 hr after conditioning (Figure 4A). The cellular memory trace formed in response to the CS+ exhibits the opposite trend (Figure 4B), such that at 3 hr after training the response in the αbranch of the α/β MB neurons approximates that observed in naïve flies (Figure 1D). The enhanced calcium response in the αbranch was observed at 9 hr after conditioning to be indistinguishable from the response observed at 24 hr (Figure 4B). The calcium responses in the β branch remain constant across the measured time points (Figure 4C).

Figure 4
The branch-specific calcium trace forms between 3 and 9 hr after spaced forward conditioning and requires the function of the memory gene, amnesiac. (A) Memory measured by odor-avoidance response after spaced forward conditioning decreases with time. ...

The branch-specific and long-term MB memory trace requires the function of the amn gene product

Our previous studies have revealed that DPM neurons form a medium-term memory trace in one branch that extensively ramifies within the αlobe of the MBs, but not in another branch that ramifies within the β lobe (Yu et al., 2005). This memory trace is first detected at 30 min after training and persists for at least one hr. In addition, its formation is dependent on the normal expression of the amn+ gene product. Because the DPM neuron memory trace forms prior to the long-term trace detected in the MBs, and because the DPM neuron trace forms in processes that innervate the αlobe of the MBs and probably synapse with the αbranches of MB neurons, we asked whether the MB long-term memory trace requires the prior formation of the DPM neuron memory trace.

Flies carrying the Uas-G-CaMP and c739-Gal4 transgenes along with the amnchpd mutation were tested for behavioral memory at 9 and 24 hr after 5x spaced forward conditioning. No behavioral memory was observed at 24 hr and only a small amount was present at 9 hr (Figure 4A). Prior studies have revealed that amn mutants acquire (learn) odor:shock associations at a rate that is indistinguishable from wild-type flies, but are impaired in the memory of these learned associations (Yu et al., 2005). Consistent with these measurements showing no behavioral memory at 24 hr after 5x spaced forward conditioning, there was no significant increase in odor-evoked calcium influx into the αbranch of the MB neurons in amn mutants at either 9 or 24 hr after conditioning (Figure 4B). Therefore, the formation of long-term behavioral memory and the long-term memory trace in the αbranch of MB neurons requires the function of the amn gene product, probably through its requirement in the DPM neurons for the formation of the medium-term memory trace. Answering where in the brain that amn+ function is required for long-term memory trace formation and pursuing other possible experiments such a synaptic silencing with UAS-Shits to probe the interrelationships of the various memory traces will require the development of a new series of promoter or driver elements that are independent of the Gal4:Uas system, so that rescuing or inactivating transgenes can be expressed independently of the reporters required for imaging.

A protein-synthesis inhibitor administered during training and expression of a dCREB repressor molecule blocks the branch-specific cellular memory trace that forms in the α/β MB neurons

Because spaced forward conditioning induces protein-synthesis dependent long-term behavioral memory, we tested whether the branch-specific memory trace induced by this training was also dependent on protein synthesis. Flies were fed on glucose solution with or without the protein synthesis inhibitor cycloheximide (35 mM) prior to spaced forward conditioning and tested 24 hr after conditioning. We observed a significant reduction in 24 hr behavioral performance after spaced forward conditioning of flies fed on cycloheximide compared to flies fed on the vehicle alone (Figure 5A), but this pharmacological treatment had no effect on immediate memory tested after 1x forward conditioning (Figure 5C). We then studied whether cycloheximide treatment alters the formation of the branch-specific memory trace. Functional optical imaging revealed that re-exposure to either Oct or Ben when used as the CS+ odor following spaced forward olfactory conditioning produces a significantly higher calcium response compared to re-exposure to the CS-odor (Figure 5B) confirming earlier results (Figures 3AC-D). This effect was completely blocked by feeding flies on cycloheximide prior to spaced forward conditioning (Figure 5B). Thus, the cellular memory trace that forms in the αbranch of the α/β MB neurons, like long-lasting behavioral memory, is dependent upon active protein synthesis.

Figure 5
Long-term memory and the αbranch-specific calcium trace are both disrupted by inhibiting protein synthesis. (A) Flies were fed on a 5% glucose solution with or without 35 mM cycloheximide for 12-15 hr prior to 5x spaced forward conditioning. After ...

An accumulation of evidence also indicates that long-term memory formation requires the functional participation of the transcription factor Creb, the cAMP response element binding protein (Yin et al., 1994; Kandel, 2001). Expression of a repressing isoform named dCreb2-b in wild-type flies under the control of a heat-shock promoter was reported to produce a heat-shock dependent impairment of long-term memory formation (Yin et al., 1994). Other investigators have observed that the hs-Creb2-b transgene may also impair long-term memory in the absence of heat shock due to basal expression (Perazzona et al., 2004).

We constructed a Uas-dCreb2-b transgene to determine whether expression of the dCreb2-b repressor molecule in the α/β MB neurons blocks the formation of long-term behavioral memory. This transgene was combined with the Uas-G-CaMP transgene to obtain control flies, and with Uas-G-CaMP and c739-Gal4 to obtain flies that express both the reporter and the dCreb-2 repressor in the α/β MB neurons. Behavioral experiments revealed that control flies carrying Uas-elements without the Gal4 driver have 24 hr memory after spaced forward conditioning comparable to wild-type flies (PI of 0.4-0.6), and that flies additionally carrying the c739-Gal4 driver element are impaired in long-term memory (Figure 6A). Expression of the dCreb2-b repressor molecule in the α/β MB neurons had no effect on 3 min memory after 1x forward conditioning (Figure 6B), confirming that expression of the repressor impairs long-but not short-term memory. These results demonstrate that long-term memory is impaired by the expression of a dCreb2-b repressor in the α/β MB neurons, mapping a requirement for Creb activity in long-term memory to these neurons. Flies expressing the dCreb2-b repressor in the α/β MB neurons were also functionally imaged 24 hr after spaced forward conditioning. This expression also abolished the branch specific long-term memory trace that is normally observed with either Oct or Ben as the CS+ odor, but had no effect on responses to the CS-odor (Figure 6D). We conclude that both long-term behavioral memory and the branch-specific long-term memory trace that normally forms following spaced forward conditioning are impaired by the expression of a dCreb2-b repressor in the α/β MB neurons.

Figure 6
Long-term memory and the branch-specific calcium trace are both disrupted by expression of a Creb repressor. (A) The 24 hr performance scores after 5x spaced forward conditioning were severely reduced by the expression of a dCreb2-b repressor in the MB ...

We extended these results by testing the effect of the previously generated hs-Creb2-b transgene on both long-term behavioral memory and the long-term memory trace. We conditioned flies containing the hs-dCreb2-b repressor transgene and the transgenic elements used for functional optical imaging (c739-Gal4; Uas-G-CaMP). Flies carrying hs-Creb2-b were impaired in 24 hr memory compared to control flies in the presence or absence of heat shock (Figure 6C), confirming behavioral studies conducted in another laboratory showing that the hs-Creb2-b transgene impairs long-term memory without heat shock (Perazzona et al., 2004). Moreover, expression of the hs-Creb2-b transgene also impaired the formation of the long-term memory trace with or without heat shock (Figure 6D).


These data indicate that increases in calcium influx in response to presentation of a conditioned odor occur specifically in the αbranch of the α/β MB neurons only after spaced forward conditioning. Neither 1x forward, 5x massed forward conditioning, nor any variation of backward conditioning is sufficient to promote the post-conditioning increase in calcium influx in this axon branch in response to the CS+. The memory trace was detected as a significant increase in G-CaMP fluorescence in response to the CS+ odor from the three-dimensional region of interest of the brain imaged in flies that received spaced forward conditioning compared to the response from the same area in naïve flies or in flies receiving other types of conditioning that were ineffective at producing long-term behavioral memory. Because the comparison was made for a specific region of interest between flies conditioned by various protocols, we believe that the failure to observe a memory trace in the β branch indicates that the physiological changes producing the memory trace in the αbranch do not occur in the β branch. Nevertheless, we cannot rule out the possibility that some unknown technical constraint, possibly related to the orientation of fibers relative to the angle of light detection, obscured the detection of an authentic trace in the β branch. Furthermore, we find that the formation of the cellular memory trace using functional optical cellular imaging is disrupted by feeding flies an inhibitor of protein synthesis, by the absence of the amn gene product, or by expressing a repressor of the transcription factor, Creb, in the α/β MB neurons. Thus, the evidence together indicates that the calcium-based memory trace discovered from these studies is localized in the α/β MB neurons, formed only in one axon branch of these neurons, formed only after spaced forward conditioning, and is dependent on amn+ activity, normal protein synthesis and Creb activity during training.

The strong correlation between the conditions necessary for the formation of the branch-specific cellular memory trace described here and those that underlie long-term behavioral memory provide an extremely strong argument that the newly discovered memory trace may guide behavior long after the conditioning event. Long-term behavioral memory in Drosophila, like the observed cellular memory trace, is induced by spaced forward conditioning, is disrupted by inhibiting protein synthesis, and requires the activity of dCreb and as shown here, the function of the amn gene. Moreover, behavioral studies of mutant animals with brain structural defects have hinted at a possible tie between long-term behavioral memory and the vertical lobes of the MB neurons. Pascual and Preat (2001) reported that long-term memory induced with spaced forward conditioning is abolished in ala (alpha-lobes absent) flies that are missing the α/α ’ branches of the MB neurons but not the β/β ’ branches, although the low penetrance of the ala mutant weakened this conclusion. Nevertheless, this observation is consistent with the possibility that long-term memory induced by spaced forward training either forms in the α/α ’ branches of the MB neurons or that these branches are required for the circuit-based retrieval of long-term memory formed after spaced forward conditioning. The observation of a long-term cellular memory trace in the αbranch of the α/β MB neurons supports that idea that the memories are formed and stored in this branch of this specific class of MB neuron.

Why is it that a memory trace representing early memories was not discovered during these studies, given that the MBs have been widely implicated in all temporal phases of memory? For instance, ablation of the MBs during development by hydroxyurea poisoning impairs the memory of adult flies immediately after training (de Belle and Heisenberg, 1994), suggesting the importance of the MBs for short-term memory expression. The expression of a wild-type rutabaga transgene specifically in the MB of adult flies rescues the memory impairment of rutabaga mutants that is observed immediately after olfactory conditioning (McGuire et al., 2003; Mao et al., 2004), indicating that normal rutabaga function in the MBs is required for the expression of early memory. Expression of Uas-Shits specifically in MB neurons impairs olfactory memory retrieval at 0.5, 3, and 24 hr after conditioning (McGuire et al., 2001; Dubnau et al., 2001; Isabel et al., 2004), strongly indicating that early memories either form in the MBs or are expressed through these neurons.

There exist at least two explanations for why we detected a trace that is specific to long-term memories. First, it is possible that memory traces that underlie early memories are formed in neurons other than the α/β MB neurons. For instance, early memories may be subserved by the short-term memory trace detected in the antennal lobes (Yu et al., 2004) and/or by memory traces that form in other types of MB neurons such as the α’/β’ or γMB neurons. Second, memory traces produced by cellular mechanisms other than increased calcium influx likely exist and these may subserve early memory. In either case, the expression of both short- and long-term memory appears to be via the output synapses of the MB neurons (McGuire et al., 2001; Dubnau et al., 2001; Isabel et al., 2004), irrespective of where various memory traces may form upstream of these synapses within the olfactory nervous system.

Two other discovered olfactory memory traces - the immediate and short-lived antennal lobe PN memory trace and the medium-term memory trace formed in DPM neurons - occur prior to the long-term memory trace that forms in the MB neurons. This raises the general question of whether the various memory traces are dependent upon one another in a time series for their formation. We have performed an initial query into this question by determining whether the long-term MB trace forms in amn mutants, which fail to form the medium-term memory trace in the DPM neurons (Yu et al., 2005). As shown by the data, amn mutants fail to form long-term behavioral memory measured at 24 hr and they fail to form the long-term memory trace in the αbranch as detected in control flies. These results are therefore consistent with the model that the formation of the medium-term trace in DPM processes that appear to innervate the αbranch of the MB neurons is required for formation of the long-term memory trace in these axon branches. Perhaps the increased calcium influx and synaptic release that occurs in response to an odor CS+ after training from the DPM processes innervating the αlobe are responsible for guiding or permitting the formation of the long-term memory trace in the putative DPM targets - the αbranches of the MB neurons. Further tests of issues regarding the interdependence of the various cellular memory traces require the development of new gene promoter systems that are independent of the Gal4:Uas system to permit expressing the neuronal reporters and transgenes that block synaptic transmission, for instance, in different sets of neurons. Although the mechanisms underlying the branch-specific modifications that result in calcium influx in response to the CS+ after spaced forward conditioning remain to be elucidated, the discovery of multiple memory traces that form at different times after training for different durations in various parts of the olfactory nervous system leads to the general proposition that behavioral memory is guided from multiple and discrete memory trace elements rather than from a single, continuously decaying memory trace.

Experimental Procedures

Transgenic animals and fly culture

Flies were cultured on standard medium at room temperature and transferred overnight to a 25°C incubator before training. Flies carrying Uas transgenes containing Drosophila the Creb2-b repressor coding region (Uas-dCreb2-b, see below) or G-CaMP (Uas-G-CaMP, Wang et al., 2003; Wang et al., 2004) were employed along with the c739-Gal4 driver (c739-Gal4 flies were from K. Kaiser, Division of Molecular Genetics, University of Glasgow, Scotland). For heat shock experiments we used the hs-dCreb2-b repressor line 17-2 (Yin et al., 1994) and crossed these flies with flies containing the c739-Gal4 driver and Uas-G-CaMP. The w(CS10) flies (CS flies carrying the w1118 mutation) served as a wild-type control in certain experiments.

The Uas-dCreb2-b repressor transgene was constructed by Dr. B. Perazzona using a procedure previously described for the generation of a Uas-dCreb2-a transgene (Perazzona et al., 2004). Briefly, genomic DNA from the transgenic fly lines carrying a hs-Creb2 transgene (Yin et al., 1994) was isolated using the Purogene DNA isolation kit (Gentra Systems, Minneapolis, MN). PCR was performed using primers spanning the 5’ and 3’ untranslated and coding regions of the hs-dCreb2 transgene and the PCR products were subcloned into topoisomerase (TOPO) vector pCR 2.1 and sequenced. The dCreb2-b coding cassette was isolated and engineered into the expression vector pPBretU (Roman et al., 1999) followed by germline transformation using standard procedures. For our experiments, we used the Uas-dCreb2-b repressor line T25.2 crossed with c739-Gal4 flies.

Behavioral assays

Drosophila olfactory learning is frequently assayed using olfactory classical conditioning (Tully and Quinn, 1985). Usually, flies are exposed to two odors in succession, one odor (the conditioned stimulus, CS+) paired with electric shock pulses (unconditioned stimulus; US) followed by a second odor (the CS-) without electric shock. The flies are then presented with the two odors in a T-maze and their avoidance of the CS+ is computed as one-half of the performance index (P.I.). To eliminate naïve odor bias, experiments are performed in a counterbalanced design and averaged, with one group of flies used in the calculation of the performance index being trained to the first odor and a second group to the second odor. Prior test-retest experiments (Tully et al., 1994; Beck et al., 2000) have revealed that the numerical index of P.I. represents the probability that trained flies will make the correct choice of the CS+ odor in the T-maze at testing, rather than representing a probability that an odor:shock association has formed in individual flies. Most or all trained flies probably develop an association during training and the strength of this association at the time of testing affects their choice behavior. Therefore, it is expected that most or all individual flies will develop a cellular memory trace after training, but only a fraction of these will display the correct choice behavior as reflected in the P.I. at testing.

We modified this protocol by adding a corresponding naïve control for each trained group. These flies underwent all of the manipulations as the trained flies in except they were not administered odor or electric shock. After training, the flies were incubated at 25°C for the indicated times (3 min, 3, 9 or 24 hr) before testing in a T-maze. Some conditioned flies were removed prior to behavioral testing and these were analyzed for cellular memory by functional imaging at the indicated times. For each group of flies trained with odor, a naïve group was tested simultaneously. The Performance Indices were then calculated for both the naïve and trained group and the ΔP.I. was obtained by subtracting the naïve score from the score of the corresponding trained group. In all cases, only experiments where the naïve flies exhibited naïve performance scores that were not significantly different from zero (t-test) were used (Figure S1). This modified protocol allowed us to make meaningful comparisons among groups since performance indices were normalized to naïve odor biases rather than counterbalancing with a CS-odor. In addition, the assay allowed us to obtain an index of the performance gains due to conditioning with each specific odor as the CS+, so that these gains could be compared to the results obtained after optically imaging individual flies.

Cycloheximide (CXM) feeding and heat-shock protocol

Flies were collected and placed into feeding tubes (17×100 mm polystyrene test tube Falcon 2017) containing one 1.0 × 2.5 cm Whatmann 3MM filter paper strip soaked with 125 μl of 5 % glucose in 3% ethanol solution or 35 mM CXM in the same solution. Following a feeding period of 12-15 hr at 25°C, the flies fed either CXM or vehicle alone were transferred to fresh food vials and allowed 30 min to recover before training. Previous experiments have shown that 12-14 hr CXM-feeding prior to associative conditioning is sufficient to reduce 24-hr memory following spaced training (Yin et al., 1994).

Heat-shock induction was performed according to previously published protocols (Yin et al., 1994; Perazzona et al., 2004). Briefly, flies were collected and incubated at 25°C overnight before heat-shock treatment. The next day, flies were transferred to empty vials and submerged for 30 min in a 37°C water bath. After heat-shock, the flies were transferred to fresh food vials and allowed a 3 hr recovery period prior to training.

Functional cellular imaging

Functional imaging procedures were similar to those already described (Yu et al., 2004; Yu et al., 2005). Flies containing both a Gal4 driver and Uas-G-CaMP were aspirated from a culture bottle without anesthesia, mounted in pipette tips, and their exposed heads secured to the tip opening with silicon cement. A small area of cuticle was removed from the top of the head capsule and the opening covered with a piece of plastic wrap. The flies were then mounted beneath a 20X objective (NA = 0.7) of a Leica TCS confocal microscope and imaged with a 488 nm excitation line using a completely open pinhole. Under these conditions, the thickness of the optical slice collected was calculated to be 32 micrometers. The emitted light was collected from 520±15 nm. Odors were delivered with pressurized air flowing at a rate of 100 mls per min. Odorants were spread on a small piece of filter paper inside of a syringe barrel and the syringe barrel placed in line with the pressurized air. Concentrated odorants were diluted in mineral oil. The delivery of odorants was accomplished with a 3-way teflon valve under the control of a programmable timer, such that fresh air could be delivered to the animals for a determined period with an instantaneous switch to odor-laced air without altering the overall flow rate. Electric shock pulses were applied to the fly’s abdomen. A total of 12 pulses of electric shock at 90v were delivered with each shock lasting 1.25 sec. To detect the long-term memory trace optically, some conditioned flies were collected after training and tested at 3, 9, and 24 hours after training and tested for calcium influx into the MB axons when the CS+ and CS-odors were delivered with 5-minute intervals.

Data analysis

Images were acquired at 5 frames/per second at a resolution of 256×256 pixels. The image data were analyzed as already described (Yu et al., 2004; Yu et al., 2005). In general, the raw fluorescence images were first smoothed with a 7x7 gaussian convolution filter and then registered. Regions of interest were circumscribed and a pseudo-color image of the %[increment]F/Fo ratio produced. The value Fo was calculated for each pixel within the region of interest as the fluorescence prior to odor application or the first electric shock as averaged over 5 successive frames. The value was calculated for each pixel within the region of interest as the difference between the maximum average intensity during the 3 sec odor application for 5 successive frames and Fo. The value [increment]F for electric shock application was calculated as the average difference between the pixel intensities within each individual time frame and those within Fo.

Supplementary Material



Supported by grants NS19904 and NS52352 from the NINDS and the Mathers Charitable Trust. R.L.D. is the R. P. Doherty-Welch Chair in Science at the Baylor College of Science.


  • Ashraf SI, McLoon AL, Sclarsic SM, Kunes S. Synaptic protein synthesis associated with memory is regulated by the RISC pathway in Drosophila. Cell. 2006;124:191–205. [PubMed]
  • Beck CDO, Schroeder B, Davis RL. Learning performance of normal and mutant Drosophila after repeated conditioning trials with discrete stimuli. J. Neurosci. 2000;20:2944–2953. [PubMed]
  • Davis RL. Olfactory learning. Neuron. 2004;44:31–48. [PubMed]
  • Davis RL. Olfactory memory formation in Drosophila: From molecular to systems neuroscience. Ann. Rev Neurosci. 2005;28:275–302. [PubMed]
  • de Belle SJ, Heisenberg M. Associative odor learning in Drosophila abolished by chemical ablation of mushroom bodies. Science. 1994;263:692–695. [PubMed]
  • Dubnau J, Grady L, Kitamoto T, Tully T. Disruption of neurotransmission in Drosophila mushroom body blocks retrieval but not acquisition of memory. Nature. 2001;411:476–480. [PubMed]
  • Isabel G, Pascual A, Preat T. Exclusive consolidated memory phases in Drosophila. Science. 2004;304:1024–1027. [PubMed]
  • Kandel ER. The molecular biology of memory storage: a dialogue between genes and synapses. Science. 2001;294:1030–1038. [PubMed]
  • Mao Z, Roman G, Zong L, Davis RL. Pharmacogenetic rescue in time and space of the rutabaga memory impairment using Gene-Switch. Proc. Natl. Acad. Sci. USA. 2004;101:198–203. [PubMed]
  • McGuire SE, Le PT, Davis RL. The role of Drosophila mushroom body signaling in olfactory memory. Science. 2001;293:1330–1333. [PubMed]
  • McGuire SE, Le PT, Osborn AJ, Matsumoto K, Davis RL. Spatio-temporal Rescue of Memory Dysfunction in Drosophila. Science. 2003;302:1765–1768. [PubMed]
  • Pascual A, Preat T. Localization of long-term memory within the Drosophila mushroom body. Science. 2001;294:1115–1117. [PubMed]
  • Perazzona B, Isabel G, Preat T, Davis RL. The role of cAMP response element-binding protein in Drosophila long-term memory. J. Neurosci. 2004;24:8823–828. [PubMed]
  • Rescorla RA. Behavioral studies of Pavlovian conditioning. Annu. Rev. Neurosci. 1988;11:329–352. [PubMed]
  • Riemensperger T, Voller T, Stock P, Buchner E, Fiala A. Punishment prediction by dopaminergic neurons in Drosophila. Curr. Biol. 2005;15:1953–1960. [PubMed]
  • Roman G, Davis RL. Molecular biology and anatomy of Drosophila olfactory associative learning. BioEssays. 2001;23:571–581. [PubMed]
  • Roman G, He J, Davis RL. New series of Drosophila expression vectors suitable for behavioral rescue. BioTechiques. 1999;27:54–56. [PubMed]
  • Tully T, Quinn WG. Classical conditioning and retention in normal and mutant Drosophila melanogaster. J. Comp. Physiol. 1985;157:263–277. [PubMed]
  • Tully T, Preat T, Boynton SC, Del Vecchio M. Genetic dissection of consolidated memory in Drosophila. Cell. 1994;79:35–47. [PubMed]
  • Wang JW, Wong AM, Flores J, Vosshall LB, Axel R. Two-photon calcium imaging reveals an odor-evoked map of activity in the fly brain. Cell. 2003;112:271–282. [PubMed]
  • Wang Y, Guo HF, Pologruto TA, Hannan F, Hakker I, Svoboda K, Zhong Y. Stereotyped odor-evoked activity in the mushroom body of Drosophila revealed by green fluorescent protein-based Ca2+ imaging. J. Neurosci. 2004;24:6507–6514. [PubMed]
  • Yin JC, Wallach JS, Del Vecchio M, Wilder EL, Zhou H, Quinn WG, Tully T. Induction of a dominant negative CREB transgene specifically blocks long-term memory in Drosophila. Cell. 1994;79:49–58. [PubMed]
  • Yu D, Ponomarev A, Davis RL. Altered representation of the spatial code for odors after olfactory classical conditioning. Neuron. 2004;42:437–449. [PubMed]
  • Yu D, Keene AC, Srivatsan A, Waddell S, Davis RL. Drosophila DPM neurons form a delayed and branch-specific memory trace after olfactory classical conditioning. Cell. 2005;123:945–957. [PubMed]