Abstract
Long-term memory (LTM) requires learning-induced synthesis of new proteins allocated to specific neurons and synapses in a neural circuit. Not all learned information, however, becomes permanent memory. How the brain gates relevant information into LTM remains unclear. In Drosophila adults, a single training session in an olfactory aversive task is not sufficient to induce protein synthesis-dependent LTM. Instead, multiple spaced training sessions are required. Here, we report that initial learning induces neural activity in the early α/β subset of Kenyon cells of the mushroom body (MB), and output from these neurons inhibits LTM formation. Specifically in response to spaced training, Schnurri activates CREBB expression which then appears to suppress the inhibitory output from MB. One training session can enhance LTM formation when this inhibitory effect is relieved. We propose that learning-induced protein synthesis and spaced training-induced CREBB act antagonistically to modulate output from early α/β MB neurons during LTM formation.
Introduction
Drosophila continues to demonstrate its utility as a model system to study memory, more than four decades after the first mutant was described (Dudai et al., 1976). Genetic dissection of olfactory aversive memory formation using various single-gene mutants has revealed at the behavioral level several distinct temporal phases, including short-term memory (STM), middle-term memory (MTM), anesthesia-resistant memory (ARM) and long-term memory (LTM) (Tully et al., 1994; Tully et al., 1990; Tully, 1996; Quinn and Dudai, 1976). The initial learning event (acquisition) after a single training session (1x) appears to induce STM, MTM and ARM, while spaced training (10 training sessions with 15 min rest intervals between each, 10xS) appears uniquely required to induce LTM consolidation. Manipulations of several of these “memory genes” also have established cases where memory formation is either impaired or enhanced, revealing bi-directional biochemical modulation of memory formation (Yin et al., 1994; Yin et al., 1995a; Ge et al., 2004; Presente et al., 2004; Wu et al., 2007; Pavlopoulos et al., 2008; Huang et al., 2012; Tubon et al., 2013; Fropf et al., 2014; Lee et al., 2018; Scheunemann et al., 2018).
As the neural substrates of olfactory memory formation are elucidated in flies, a remarkable “memory circuit” is emerging. Olfactory information delivered from the antennal lobe (AL) by projection neurons (PN) and foot shock reinforcement delivered by dopaminergic neurons (DAN) both converge on mushroom body (MB) neurons in the central brain where their coincidence triggers cascading cellular events that underlie learning (Dubnau and Chiang 2013; Perisse et al., 2013; Davis, 2015; Cognigni et al., 2018). MBs play a predominant role in subsequent memory formation, together with several groups of extrinsic MB neurons. Sequential genetically-defined memory phases map onto distinct subpopulations of these neurons. STM involves γ, α′/β′ and α/β neurons and two classes of MB output neurons (MBON: MB-M4, MB-M6) (Blum et al., 2009; Scheunemann et al., 2012; Bouzaiane et al., 2015). MTM involves neural activity in γ, α/β and MB-V2 neurons (Blum et al., 2009; Scheunemann et al., 2012; Bouzaiane et al., 2015). ARM requires neural activity in MB γ, α′/β′, α/β neurons, dorsal paired medial (DPM) neurons, anterior paired lateral (APL) neurons, DAN and four different MB output neurons (MB-M4, MB-M6, MB-V2, MBON-β2β′2a) (Lee et al., 2011; Knapek et al., 2011; Placais et al., 2012; Wu et al., 2013; Bouzaiane et al., 2015; Yang et al., 2016; Scholz-Kornehl and Schwärzel, 2016; Kotoula et al., 2017; Shyu et al., 2019). LTM involves neural activity in late MB α/β neurons with output from DPM, serotonergic projection neurons (SPN) and three classes of MBONs (MB-V3, MB-M4, MBON-γ3,γ3β′1). Cyclic AMP response element binding protein (CREB)-dependent consolidation of LTM also requires activity in dorsal anterior lateral (DAL) neurons (Chen et al., 2012; Pai et al., 2013; Tonoki and Davis 2015; Bouzaiane et al., 2015; Wu et al., 2017; Scheunemann et al., 2018). Finally, memory retrieval depends on neural activity in DAL, pioneer α/β neurons and four classes of MBONs (MB-V2, MB-V3, MB-M4, MBON-γ3,γ3β′1) (Séjourné et al., 2011; Chen et al., 2012; Pai et al., 2013; Bouzaiane et al., 2015; Wu et al., 2017).
Here, we describe another enlightening property of olfactory memory in Drosophila: inhibition of LTM formation at the circuit level. Output from the early α/β subpopulation of MB neurons appears initially to inhibit LTM formation, but with spaced training, transcription of crebB (dcreb2, repressor) is induced therein, apparently reducing neural output therefrom and thereby enabling LTM formation. Thus, persistent olfactory memory formation appears modulated or “gated” at the level of neural activity in early α/β MB neurons. These observations presage the need for a more general deconvolution of biochemical mechanisms into distinct neuronal subtypes within a memory circuit.
Results
Learning inhibits LTM
Sixty of the single-gene mutants mentioned above were generated using transposon mutagenesis and were screened for impairments of memory one day after 10xS training (Dubnau et al., 2003). Twenty-two of these lines carried P-Gal4 enhancer traps, which enabled us to drive targeted inducible expression of a temperature-sensitive RicinCS transgene and then block protein synthesis after 10xS (Chen et al., 2012; Pai et al., 2013; Wu et al., 2017). With protein synthesis inhibited in this manner, we found impairments of 1-day memory in nine of these lines (figure supplement 1A-C; table supplement 1) (Tully et al., 1994; Yin et al., 1994). Remarkably, GFP was expressed in DAL neurons in all nine cases (figure supplement 1B), an observation that contributed to our characterization of DAL neurons extrinsic to the MB as bona fide “LTM neurons” (Chen et al., 2012). Two of the enhancer-trap memory mutants that we screened were particularly informative. In umnitza flies, GFP was expressed in DAL neurons but very weakly in MB, and 1-day memory after 10xS was impaired. Conversely in norka flies, GFP was expressed in MB but not in DAL neurons and 1-day memory after 10xS was normal.
What then might be going on in the seven enhancer-trap mutants with normal memory and with GFP expression in both MB & DAL neurons (Figure 1A; figure supplement 1A)? First, we confirmed that active RicinCS inhibition of protein synthesis in different subsets of MB neurons did not impair 1-day memory after 10xS (figure supplement 2A-B and 3). In contrast, 1-day memory after 10xS was impaired by active RicinCS in DAL or MB-V3 neurons (Chen et al., 2012; Pai et al., 2013; Wu et al., 2017) but was normal after massed training (10x training sessions with no rest intervals; 10xM) or in control flies with inactive RicinCS (18 °C) (figure supplement 3).
We next tested the hypothesis that inhibition of protein synthesis in MB might enhance LTM formation, thereby off-setting the impairment produced by blocking protein synthesis in DAL neurons. Using cry-Gal80 or MB-Gal80, we blocked transgenic RicinCS expression outside (i.e. DAL neurons) or inside of MB, respectively, in these seven enhancer-trap memory lines (Figure 1B-C) and then subjected them to suboptimal 3xS training. Surprisingly, LTM formation was enhanced in all seven lines by cry-Gal80 subtraction (Figure 1D-F).
The MB is composed of approximately 2,500 intrinsic neurons (Kenyon cells; KCs) developmentally derived from four neuroblasts and distinguished by their projections that form the γ, α′/β′ and α/β lobes (Ito et al., 1997; Zhu et al., 2003; Lin et al., 2007). We looked among these neuronal subpopulations to identify where LTM enhancement might reside (Figure 2; figure supplement 2A). Active Ricin was expressed in all KCs or in γ, α′/β′, or α/β neurons separately. Enhanced LTM was observed after 3xS only in α/β neurons, which were previously shown to have a role in LTM formation (Blum et al., 2009; Yu et al., 2006) (Figure 2A).
The α/β neurons are subdivided further into three types: pioneer α/β, early α/β and late α/β neurons based on their birth sequences (Zhu et al., 2003; Lin et al., 2007; Tanaka et al., 2008; Aso et al., 2014). When active Ricin was expressed in these three subpopulations separately, we observed enhanced LTM after 3xS only when protein synthesis was blocked in early α/β neurons (Figure 2A, left). Enhanced LTM lasted for at least 4 days (Figure 2A, right) and was not observed in control flies (inactive RicinCS) after 3xS training or in flies with active RicinCS after 3xM (figure supplement 2C). Blocking protein synthesis in early α/β neurons enhanced 1- and 4-day memories even after only 1x training (Figure 2B). Notably, LTM enhancement after 1x training required two copies of transgenic RicinCS. Together, these results indicate that inhibition of protein synthesis in early α/β neurons yields a bona fide enhancement of LTM formation.
We next inquired about the most effective time after training when inhibition of protein synthesis would enhance LTM. RicinCS in early α/β neurons was activated for 12 h in a series of time windows staggered by 2 h during the first 24 h after 1x training (Chen et al., 2012; Wu et al., 2017). LTM was enhanced when protein synthesis was blocked beginning from 0- to 4-h but not from 6- to 12-h after training (Figure 3A). Shortening the inhibition period to 3 h, we resolved the window of protein-synthesis-dependent LTM inhibition to the first 6 h after training (Figure 3B).
Output from early α/β neurons inhibits LTM
To address whether the inhibitory effect on LTM in early α/β lobes depends on their neural output, we blocked synaptic transmission from pioneer α/β, late α/β or early α/β neurons using UAS-shits (Dubnau et al., 2001; McGuire et al., 2001). In these experiments we found that 1- and 4-day memory after 3xS training were enhanced by this manipulation in early α/β (Figure 4A-B), but not in pioneer or in late α/β neurons (figure supplement 4A). Interestingly, we also established that LTM after 3xS was enhanced when synaptic transmission was blocked from early α/β during the first 8 h period after training but not 9-24 h after training or during 1-day memory retrieval (Figure 4A). Thus, this temporal requirement for synaptic transmission from early α/β neurons corresponds to the requirement for protein synthesis (Figure 3). We confirmed these results using two additional Gal4 drivers expressing specifically in early α/β neurons (Figure 4C; figure supplement 2A) and observed normal 1-day memory after 3xM (Figure 4B, middle) as in control flies carrying UAS-shits alone, Gal4 alone (Figure 4B, right) or at the permissive temperature for shits (Figure 4A, left and 4C, right). Finally, we established that LTM after 10xS training was not impaired when neural output from early α/β neurons was blocked (figure supplement 4B). Together, these results indicate that, in the absence of spaced training, neural output from early α/β neurons inhibits LTM formation. To examine the influence of early α/β neuron membrane excitability on memory, we used ectopic expression of either transgenic hyperexcitors (UAS-ShawDN, a dominant-negative Shaw potassium channel and UAS-NaChBac, a sodium channel) or hypoexcitors (UAS-Shaw, a Shaw potassium channel and UAS-Kir2.1::GFP, an inward-rectifying potassium channel Venken et al., 2011). Temporal control of these transgenes was enabled using a tub-Gal80ts transgene (conditional expression of Gal80 suppresses Gal4 expression at 18 °C but not at 30°C McGuire et al., 2003). We found that increasing membrane excitability of early α/β neurons impaired 1-day memory after 10xS training (Figure 5A, left) without affecting (1) memory after 1x training (fig. S5A), (2) 1-day memory after 10xM training (figure supplement 5B) or (3) 1-day memory after 10xS training when flies were kept at permissive temperature (18°C) (figure supplement 5C). This inhibitory effect appears to be complete, because inhibition of protein synthesis by feeding flies cycloheximide (CXM) did not further reduce 1-day LTM after 10xS training (Figure 5A, right). Decreasing membrane excitability of early α/β neurons with ectopic expression of hypoexciter transgenes, on the other hand, enhanced both 1- and 4-day memory after 1x training (Figure 5B, left and middle), whereas enhancement of 1-day memory was not observed in control transgenic flies kept at the permissive temperature (18°C) (Figure 5B, right). We also found normal 1-day memory in these transgenic flies after 10xS training (Figure 5C). Thus, sufficient spaced training appeared to occlude the enhancing effects on LTM formation of decreased membrane excitability in early α/β neurons. These data further support the notion that neural activity from early α/β neurons inhibits LTM formation downstream (Dubnau and Chiang 2013; Pai, et al., 2013; Wu et al., 2017).
cAMP signaling in early α/β neurons enhances LTM
Neural excitability is suggested to be modulated by cAMP signaling (Davis et al., 1998; Baines, 2003), which in MB is also involved in LTM formation (Blum et al., 2009). Accordingly, we inducibly overexpressed rutabaga+ (rut+) adenylyl cyclase (AC) or constitutively active cAMP-dependent protein kinase (Pkaact1) transgenes in early α/β neurons and found that 1-day memory after 1x training was enhanced to levels normally seen after 10xS in both cases (Figure 6A and 7A). Moreover, inducible RNAi knockdowns of these genes impaired 1-day memory after 10xS (Figure 6b and 7b; figure supplement 6). Together, these results suggest that LTM formation is also modulated by cAMP in early α/β neurons.
CREBB in early α/β neurons enhances LTM
Protein synthesis-dependent LTM formation also depends on CREBB transcription factors, and expression of CREBB protein is thought to be dependent on the expression level of protein kinases involved in cAMP signaling (Lee et al., 2018). Consistent with our Rutabaga and PKA knockdown results, we induced RNAi knockdown of CREBB in early α/β neurons and observed impairment of 1-day memory after 10xS training. Further impairment was not seen in combination with systemic protein synthesis inhibition by feeding CXM (Figure 8C). Because inhibition of protein synthesis in early α/β neurons instead produced an enhancing effect on LTM formation, we inducibly expressed crebB repressor transgenes (Zhang et al., 2019) in these neurons only, with the expectation that the manipulations would lead to enhanced memory. Indeed, expressing crebB enhanced 1-day memory after 1x training to levels normally seen after 10xS training (Figure 8A-B; figure supplement 6).
Spaced training induces crebB transcription
We next generated a crebB promoter-driven Gal4 transgene containing an 11-kb 5′ genomic sequence just upstream of CREBB (see Methods) (Yin et. al., 1995b). This crebB-Gal4 drives GFP expression in most glia cells and brain neurons, including most MB neurons, though higher levels of expression can be seen in α/β compared to α′/β′ or γ neurons (Figure 9A). By photo converting pre-existing green KAEDE to red prior to training (Chen et. al., 2012), we measured significantly more newly synthesized crebB-Gal4 green KAEDE in the MB α-lobe during 24-h intervals after 5xS or 10xS training, but not after 1x training or 10xM training in comparison with naïve control flies (Figure 9B-C). This training-induced increase in crebB KAEDE appeared specific to the MB neurons because spaced training did not significantly change the levels of new crebB KAEDE in ellipsoid body (EB) or glia (Figure 9B, right). These results demonstrate that multiple sessions of spaced training increases CREBB expression in early α/β neurons. Our finding that inhibition of protein synthesis in early α/β neurons enhanced LTM formation (Figure 1-3) suggests that crebB gene products function to repress protein synthesis.
Schnurri (Shn) regulates CREBB-dependent LTM formation
How does spaced training induce CREBB expression? To address this question, we sought to identify positive regulators of crebB transcription during LTM formation. Yeast two-hybrid and chromatin immunoprecipitation experiments previously revealed several such candidates that bind in the crebB promoter region (data not shown). Prominent among these were (1) CREB family protein CREBA, a leucine-zipper transcription factor (Smolik et al., 1992) and (2) Shn, a zinc finger C2H2 transcription factor encoded by the shn gene (Marty et al., 2000) which also was identified in a transposon mutagenesis screen for impairment of 1-day memory after 10xS as the umnitza mutant (Dubnau et al., 2003) (described above, see figure supplement 1).
Together, these findings implicated CREBA and Shn as candidate regulators of LTM formation through transcriptional activation of crebB. Interestingly, we inducibly overexpressed each in early α/β neurons and found enhanced 1-day memory after 1x or 3xS training with this manipulation of the shn+ transgene but not crebA+ (Figure 10A; figure supplement 7A-B). We also observed strongly elevated CREBB protein expression in transgenic shn+ flies, but not in transgenic rut+ or Pkaact1 flies (figure supplement 8). Moreover, inducible RNAi knockdowns of shn but not crebA impaired 1-day memory after 10xS (Figure 10B; figure supplement 7C), and further impairment was not seen with systemic protein synthesis inhibition after CXM feeding (Figure 10B, right). Memory after 10xS was fully rescued in shn knockdown flies by crebB co-expression (Figure 10C) and was enhanced relative to controls after 1x (Figure 10D). Taken together, these results show that CREBB expression in early α/β neurons in response to spaced training is Shn-dependent.
Discussion
Our data suggest that MB neurons provide a compelling cellular gating mechanism for LTM formation. A single training session is sufficient to increase early α/β neuronal excitability, the output from which produces a downstream inhibitory effect on LTM formation. After spaced training, cAMP signaling regulates neural excitability and/or Shn increases CREBB expression, the net effects of which we suggest then represses further protein synthesis, thereby reducing early α/β output and relieving the inhibitory effect on LTM formation. Remarkably, our observations emerged from a screen of enhancer trap memory mutants using RicinCS protein synthesis inhibition (figure supplement 1). 1-day memory after 10xS was impaired in nine lines, eight of which showed expression in both MB & DAL neurons. Curiously, another seven lines were not impaired in 1-day memory after 10xS training – but they, too, showed enhancer expression patterns in MB & DAL neurons. We hypothesized that blocking protein synthesis in DAL neurons impaired LTM but doing so in (some) MB neurons might actually enhance LTM, negating the inhibitory effects in DAL neurons.
We tested this idea by maintaining RicinCS expression in MB while blocking RicinCS expression outside of MB using cry-Gal80 (Figure 1). Surprisingly, LTM now was enhanced in all seven of these enhancer trap lines (Figure 1). We then identified early α/β as the subset of MB neurons responsible for this enhancing effect (Figure 2). Inhibition protein synthesis in early α/β neurons during the first 6 h, or blocking synaptic transmission from early α/β neurons during the first 8 h after training was sufficient to enhance LTM (Figure 3 and 4). Increasing excitability of early α/β neurons impaired LTM, but decreasing excitability again enhanced LTM (Figure 5). We next asked whether these neural excitability-dependent effects were also cAMP dependent. RNAi mediated knockdown of Rutabaga or PKA in early α/β impaired LTM, while overexpression of a rut+ or Pkaact1 transgene enhanced LTM (Figure 6 and 7). CREBB expression is suggest to be synergistically and post-transcriptionally regulated by protein kinases responding to cAMP signaling (15) and accordingly, our RNAi mediated knockdown of CREBB in early α/β impaired LTM, while overexpression of a crebB transgene enhanced LTM (Figure 8). Finally, using a crebB promoter driven Gal4 transgene, we show that CREBB transcription increases after 5xS or 10xS spaced training but not after 1x training (Figure 9). Thus, spaced training-dependent expression of CREBB repressor proteins in early α/β neurons blocks this inhibitory output from early α/β neurons, thereby allowing LTM formation (downstream) to proceed.
An enhancing role associated with Shn-induced expression of CREBB repressor is a novel aspect of this LTM gating mechanism (Figure 10 and figure supplement 8). Previous reports have claimed that chronic expression of a CREBB repressor or RNAi transgenes in all α/β neurons impaired 1-day memory after spaced training (Yu et al., 2006; Lee et al., 2018). Chen et al., (2012) documented, however, that these chronic disruptions of CREBB produced developmental abnormalities in MB structure. In contrast, acute induced expression of active RicinCS or CREBB repressor only in adult α/β neurons did not impair 1-day memory after spaced training (and did not produce structural defects). Using a different inducible system (MB247-Switch) to acutely expresses CREBB in γ and α/β neurons, Hirano et al., (2016) showed a mild impairment of 1-day memory after spaced training. More interestingly, they used various molecular genetic tools to show that interactions among CREBB, CREB Binding Protein (CBP) and CREB Regulated Transcription Coactivator (CRTC) in MB clearly were involved in LTM formation or maintenance, respectively. Using the same inducible gene switch tool, Miyashita et al., (2018) showed a fascinating positive regulatory loop between Fos and CREBB in MB during LTM formation – but they did not show behavioral data pertaining to manipulation of CREBB per se- and they did not restrict their experiments to early α/β neurons.
A recent study that features cyclic AMP-response element (CRE)-driven transgenes is pertinent to this report. Zhang et al., (2015) expressed a CRE-luciferase transgene in different subpopulations of MB neurons and then monitored luciferase activity in live flies at various times after spaced training. Immediately after spaced training, they showed in some cases luciferase expression decreased (OK107 expressing in all MB neurons; c739 expressing in all α/β neurons; 1471 expressing in γ neurons), in others expression increased (c747 and c772 expressing variably in all MB neurons) or in some no changes were detected (c320 expressing variably in γ α′/β′ and α/β subpopulation, 17d expressing primarily in late α/β and in early α/β neurons). Indeed, these authors point out that, because CRE-luciferase was expressed in more than one subpopulation of MB neurons, only net effects of CREB function could be quantified. Obviously, such a conclusion must be drawn from any behavioral data collected after CREBB manipulations in multiple subpopulations of MB neurons. Our study provides a dramatic example of this point. By restricting our manipulation only to the early α/β neurons and only in adult stage animals, we show that acute overexpression or knockdown of CREBB enhances or impairs LTM formation, respectively (Figure and that spaced training serves to increase the expression of CREBB in these neurons (Figure 9).
Of particular relevance to our future studies is the curious discovery that output from early α/β neurons specifically inhibits LTM formation. We find no evidence of inhibitory transmitter (i.e., GABA) synthesis or signaling in early α/β neurons, however, and others have suggested that memory-relevant MB output synapses are cholinergic (Barnstedt et al.,2016). Thus, we presume that inhibition of LTM lies somewhere downstream in the memory circuit. Furthermore, we note that ARM appears to involve α/β neurons (Lee et al.,2011; Knapek et al.,2011; Scholz-Kornehl and Schwärzel, 2016; Kotoula et al., 2017; Shyu et al., 2019) and to inhibit LTM formation (Isabel et al., 2004; Placais et al., 2012). Thus, a molecular link between ARM and LTM may reside in early α/β neurons.
More generally, our results underscore the need to study behavior-genetic relations in each of the seven MB neuronal subpopulations (Aso et al., 2014) separately before drawing firm conclusions about a role for MB in specific memory phases or in the dynamics of a larger memory circuit involving neurons intrinsic and extrinsic to MB. With the more complex circuitries in vertebrate animal models, such deconstruction of memory formation into specific neuronal subtypes will be even more critical and enlightening.
Materials and Methods
A collection of Drosophila P-Gal4 transposon insertions were previously selected in an enhancer trap mutagenesis screen for long-term memory phenotypes (Dubnau et al., 2003). The resultant Gal4 expression patterns in seven of these mutants were leveraged to drive and temporally control cold-sensitive RicinCS activity to block protein synthesis in the identified neuron subsets. In addition, we spatially restricted RicinCS activity by inhibiting Gal4 with MB or DAL neuron-specific expression of Gal80. We used an automated olfactory aversive learning task (Tully et al., 1994) and assessed LTM after blocking protein synthesis, inhibiting consolidation in these temporally and spatially restricted domains to identify the subsets of neurons critical for this task. Blocking transmission from these neurons with Gal4-targetted temperature-sensitive Dynamints after training was used to test the implicated roles of these neurons in LTM consolidation (Dubnau et al., 2001; McGuire et al., 2001). Spatial and temporal regulation of K+ and Na+ channel activity with transgene overexpression and RNAi knockdown within these neurons was used to assess the downstream impacts of signaling valence on LTM. Similarly, restricted expression of transgenes was used to examine the training-responsive effects on LTM. We evaluated training-responsive CREBB expression with confocal microscopy using a Gal4-targeted UV-sensitive KAEDE reporter system (Chen et al., 2012). In various experiments, flies were fed CXM to provide a systemic level of protein synthesis inhibition. Detailed procedures for all methods are described in the supplementary materials.
Flies
Fly stocks were maintained on standard corn meal/yeast/agar medium at 25 ± 1 °C or 18 ± 1 °C and 70% relative humidity on a 12:12-h light:dark cycle. All genotypes and sources are listed in table supplement 2.
Behaviour
Olfactory associative learning was evaluated by training 6- to 7-day-old flies in a T-maze apparatus with a Pavlovian olfactory conditioning procedure (Tully and Quinn, 1985) as described previously (Chen et al., 2012; Pai et al., 2013; Wu et al., 2017). All experiments were conducted in the dark in an environment-controlled room at the required temperatures and 70% relative humidity. The odours used were 3-octanol (OCT) and 4-methylcyclohexanol (MCH). Each experiment consisted of two groups of approximately 100 flies, each of which was conditioned with one of the two odours. Flies were exposed sequentially to two odours that were carried through the training chamber in a current of air (odours were bubbled at 750 ml/min). In a single training session, flies first were exposed for 60 s to the conditioned stimulus (CS +), during which time they received the unconditioned stimulus (US), which consisted of 12 1.5-s pulses of 60 V dc electric shock presented at 5-s interpulse intervals. After the presentation of the CS+ condition, the chamber was flushed with fresh air for 45 s. Then flies were exposed for 60 s to the unpaired CS−. To evaluate memory retention immediately after single-session training (acquisition), flies were gently tapped into an elevator-like compartment immediately after training. After 90 s, the flies were transported to the choice point of a T-maze, in which they were exposed to two converging currents of air (one carrying OCT, the other MCH) from opposite arms of the maze. Flies were free to choose between and walk toward the CS+ and CS− for 120 s, at which time they were trapped inside the respective arms of the T-maze (by sliding the elevator out of the register), anesthetised, and counted. Flies that chose to avoid the CS+ran into the T-maze arm containing the CS−, whereas flies that chose to avoid the CS− ran into the arm containing the CS+. For each experiment, a performance index (PI1,2) = (NCS− – NCS+)/(NCS− + NCS+) was calculated and averaged over these two complementary experiments, with the final PI = (PI1 + PI2)/2. Averaging of the two reciprocal scores eliminated any potential biases originating from the machine, naïve odour preferences, or non-associative changes in olfaction. For 24-h memory experiments, flies were subjected to single-session training, training massed together without rest, or training spaced out with 15-min rest intervals. For these training protocols, robotic trainers were used. All genotypes were trained and tested in parallel and rotated among all of the robotic trainers to ensure a balanced experiment. The genetic backgrounds of all fly strains were equilibrated to the “Canton” wild-type background by five or more generations of backcrossing. In tub-Gal80ts experiments, flies raised at 18 °C were transferred to 30 °C for at least five days before the experiments.
Pharmacological treatment
To block protein synthesis, flies were fed 35 mM cycloheximide (Sigma) in 5% glucose 1 day before training until immediately before the test (Tully et al., 1994).
crebB promoter construct
To engineer the crebB promoter construct, polymerase chain reaction (PCR) was performed using genomic DNA from the wild-type Canton-S w1118 (iso1CJ) fly line as the template together with the forward primer 5′GAAAAGTGCCACCTGCTGCATGTCTACCAACAGTTCGAG 3′ and the reverse primer 5′CCGGATCTGCTAGCGGTTCCAGCTGCTGTCTGTATGAC 3′. A 11.6-kb PCR product was generated and inserted into the pBPGAL4.2Uw-2 vector, was digested with AatII and KpnI using In-Fusion® cloning system (Clontech). The promoter construct was injected into attP40-containing fly strains to obtain the transgenic fly lines.
KAEDE measurement
KAEDE is a photoconvertible green fluorescent protein, irreversibly changing its structure to a red fluorescent protein upon ultraviolet irradiation (Ando et al., 2002). Taking advantage of circadian transcription and protein synthesis in the lateral clock neurons, we previously validated de novo KAEDE synthesis in per-Gal4>UAS-kaede flies, in which it faithfully reports the cyclic transcriptions of the period gene. Feeding cycloheximide also suppressed green KAEDE synthesis, while not affecting the already-converted red KAEDE (Chen et al., 2012). To measure the amount of newly synthesised KAEDE in MB neurons, we used procedures adapted from a previous study (Chen et al., 2012). Briefly, pre-existing KAEDE proteins were photoconverted into red fluorescent proteins by 365–395 nm UV irradiation generated from a 120-W mercury lamp. For behavioural testing, approximately 15–20 flies kept in a clear plastic syringe were directly exposed to UV light at a distance of 5 cm for 1 h. Individual neurons expressing KAEDE were directly visualised through an open window in the fly’s head capsule. Living samples were used because the signal- to-noise ratio of green to red KAEDE is greatly reduced after chemical fixation. KAEDE neurons were located in less than 5 s by a fast pre-scanning of red KAEDE excited by a 561-nm laser, to avoid unnecessary fluorescence quenching of green KAEDE during repeated scanning. A single optical slice through the MB α-lobe tip was imaged at a resolution of 1024×1024 pixels under a confocal microscope with a 40× C-Apochromat water-immersion objective lens (N.A. value 1.2, working distance 220 μm). All brain samples in the experiment were imaged with the same optical settings maximised for green and red KAEDE immediately before and after photoconversion, respectively. In all cases, both green KAEDE (excited by a 488-nm laser) and red KAEDE (excited by a 561-nm laser) were measured. By using the amount of red KAEDE as an internal standard to calibrate individual variation, we calculated the rate of increase in green KAEDE synthesis after photoconversion with the formula (ΔF) = %(Ft1 –average Ft0)/average Ft0, where Ft1 and Ft0 are the ratios of the averaged intensities of green (G) to red (R) KAEDE (Gt0/Rt0) immediately after photoconversion (t0) and at a later specific time point (t1), respectively.
Spatiotemporal inhibition of protein synthesis
RicinCS, a mutated Ricin A chain, inactivates eukaryotic ribosomes by hydrolytically cleaving the N-glycosidic bond (A4324) of the 28S ribosomal RNA subunit at high temperatures (30°C), but not at low temperatures (18°C) (Endo et al., 1987; Endo and Tsurugi, 1987; Moffat et al., 1992; Allen et al., 2002). We previously validated the spatiotemporal effect of RicinCS inhibition in the Drosophila brain using lateral clock neurons. We found that RicinCS can effectively inhibit ∼80% of protein synthesis at a permissive temperature (30°C), which is quickly reversed to normal levels after shifting to a restrictive temperature (18°C) (Chen et al., 2012). This suggests a quick restoration of ribosomal synthesis once RicinCS becomes inactive. While active RicinCS is a potent cytotoxin for inhibiting protein synthesis, it tends not to be lethal, as RicinCS eventually inhibits its own synthesis (Chen et al., 2012, Moffat et al., 1992; Allen et al., 2002). In the current experiments, two copies of RicinCS was used to block protein synthesis. All flies were raised at 18°C to keep RicinCS inactive. Before or after training at 18°C, the Gal4> UAS-ricinCS;UAS-ricinCS flies were transferred to 30°C for 24 h to activate RicinCS, and then shifted back to 18°C for 1 h to inactivate RicinCS before the experiments. Temporal control of RicinCS activation is indicated in the figures for the relevant experiments.
Immunohistochemistry
Brains were dissected in phosphate-buffered saline (PBS), fixed with a commercial microwave oven (2,450 MHz, 1100 Watts) in 4% paraformaldehyde on ice for 60 s three times, and then immersed in 4% paraformaldehyde with 0.25% Triton X-100 for 60 s three times. After being washed in PBS for 10 min at room temperature, brain samples were incubated in PBS containing 2% Triton X-100 (PBS-T) and 10% normal goat serum, and then degassed in a vacuum chamber to expel tracheal air for four cycles (depressurizing to –70 mmHg and then holding for 10 min). Next, brain samples were blocked and penetrated in PBS-T at 4 °C overnight, and then incubated in PBS-T containing (1) 1:40 mouse 4F3 anti-DLG antibody (Developmental Studies Hybridoma Bank, University of Iowa) to label Disc large proteins, and (2) 1:500 mouse anti-CREBB α657 antibody (from Jerry Yin (Tubon et al., 2013)) at 4 °C for 1 day. Samples were subsequently washed in PBS-T three times and incubated in PBS-T containing 1:200 biotinylated goat anti-mouse IgG (Molecular Probes) as the secondary antibody at 25 °C for 1 day. Brain samples were then washed and incubated with 1:500 Alexa Fluor 635 streptavidin (Molecular Probes) at 25 °C for 1 day. Finally, after extensive washing, immunolabeled brain samples were directly cleared for 5 min in FocusClear, an aqueous solution that renders biological tissue transparent (Chiang et al., 2001), and mounted between two cover slips separated by a spacer ring with a thickness of ∼200 μm. Sample brains were imaged under a Zeiss LSM 780 or 880 confocal microscope with a 40× C-Apochromat water-immersion objective lens (N.A. value 1.2, working distance 220 μm).
Statistics
All raw data were analysed parametrically with SigmaPlot 10.0 and SigmaStat 3.5 statistical software. All the data including the behaviour Performance Index (PI) or KAEDE image (ΔF) were evaluated via unpaired t-test (two groups) or one-way analysis of variance (ANOVA) (> two groups). Data were evaluated with the Mann-Whitney Rank Sum Test in cases of unequal variances. Data in all figures are presented as the mean ± SE. Experiments were replicated using multiple Gal4 drivers with equivalent expression patterns, and multiple effector genes and reagents that impact shared cellular functions.
Funding
This work was financially supported by
The Brain Research Center under the Higher Education Sprout Project co-funded by the Ministry of Education and the Ministry of Science and Technology in Taiwan
Yushan Scholar Program from the Ministry of Education in Taiwan
Dart NeuroScience LLC in U.S.A.
Author contributions
Conceived the project, analysed the data, and wrote the manuscript: C.C.C., J.S.D., T.T. and A.S.C.
Imaging experiments: H.W.L.
Behavioural experiments: C.C.C. and F.K.L.
Generated creb2-Gal4 transgenic flies: R.Y.J. and L.C.
Competing interests
All other authors declare they have no competing interests.
Data and materials availability
All data are available in the main text or the supplementary materials.
Acknowledgments
We thank the Bloomington Drosophila stock center, Vienna Drosophila RNAi Center (VDRC) and Kyoto Drosophila Genomics Resource Centers (DGRC) for fly stocks. We also thank the Developmental Studies Hybridoma Bank for the antibodies.