Abstract
Olfactory associative learning in Drosophila is mediated by synaptic plasticity between the Kenyon cells of the mushroom body and their output neurons. Both Kenyon cells and their inputs are cholinergic, yet little is known about the physiological function of muscarinic acetylcholine receptors in learning in adult flies. Here we show that aversive olfactory learning in adult flies requires type A muscarinic acetylcholine receptors (mAChR-A) specifically in the gamma subtype of Kenyon cells. Surprisingly, mAChR-A inhibits odor responses in both Kenyon cell dendrites and axons. Moreover, mAChR-A knockdown impairs the learning-associated depression of odor responses in a mushroom body output neuron. Our results suggest that mAChR-A is required at Kenyon cell presynaptic terminals to depress the synapses between Kenyon cells and their output neurons, and may suggest a role for the recently discovered axo-axonal synapses between Kenyon cells.
Introduction
Animals learn to modify their behavior based on past experience by changing connection strengths between neurons, and this synaptic plasticity often occurs through metabotropic receptors. In particular, neurons commonly express both ionotropic and metabotropic receptors for the same neurotransmitter, where the two may mediate different functions (e.g., direct excitation/inhibition vs. synaptic plasticity). In mammals, where glutamate is the principal excitatory neurotransmitter, metabotropic glutamate receptors (mGluRs) have been widely implicated in synaptic plasticity and memory (Jörntell and Hansel, 2006; Lüscher and Huber, 2010). Given the complexity of linking behavior to artificially induced plasticity in brain slices (Schonewille et al., 2011; Yamaguchi et al., 2016), it would be useful to study the role of metabotropic receptors in learning in a simpler genetic model system with a clearer behavioral readout of synaptic plasticity. One such system is Drosophila, where powerful genetic tools and well-defined anatomy have yielded a detailed understanding of the circuit and molecular mechanisms underlying associative memory (Busto et al., 2010; Cognigni et al., 2017; Hige, 2017). The principal excitatory neurotransmitter in Drosophila is acetylcholine, but, surprisingly, little is known about the function of metabotropic acetylcholine signaling in synaptic plasticity or neuromodulation in Drosophila. Here we address this question using olfactory associative memory.
Flies can learn to associate an odor (conditioned stimulus, CS) with a positive (sugar) or a negative (electric shock) unconditioned stimulus (US), so that they later approach ‘rewarded’ odors and avoid ‘punished’ odors. This association is thought to be formed in the presynaptic terminals of the ~2,000 Kenyon cells (KCs) that make up the mushroom body (MB), the fly’s olfactory memory center (Busto et al., 2010; Cognigni et al., 2017; Hige, 2017). These KCs are activated by odors via second-order olfactory neurons called projection neurons (PNs). Each odor elicits responses in a sparse subset of KCs (Campbell et al., 2013; Lin et al., 2014) so that odor identity is encoded in which KCs respond to each odor. When an odor (CS) is paired with reward/punishment (US), an odor-specific set of KCs is activated at the same time that dopaminergic neurons (DANs) release dopamine onto KC presynaptic terminals. The coincident activation causes long-term depression (LTD) of synapses from the odor-activated KCs onto mushroom body output neurons (MBONs) that lead to approach or avoidance behavior (Aso and Rubin, 2016; Aso et al., 2014b; Cohn et al., 2015; Hige et al., 2015b; Owald et al., 2015; Perisse et al., 2016; Séjourné et al., 2011). In particular, training specifically depresses KC-MBON synapses of the ‘wrong’ valence (e.g., odor-punishment pairing depresses odor responses of MBONs that lead to approach behavior), because different pairs of ‘matching’ DANs/MBONs (e.g. punishment/approach, reward/avoidance) innervate distinct regions along KC axons (Aso et al., 2014a). Yet recent studies show that the mushroom body contains not just KC->MBON and DAN->KC synapses, but also KC->DAN, DAN->MBON and even KC->KC synapses (Cervantes-Sandoval et al., 2017; Eichler et al., 2017; Takemura et al., 2017), suggesting that this model may be incomplete.
Both MB input (PNs) and output (KCs) are cholinergic (Barnstedt et al., 2016; Yasuyama and Salvaterra, 1999), and KCs express both ionotropic (nicotinic) and metabotropic (muscarinic) acetylcholine receptors (Croset et al., 2018; Davie et al., 2018). The nicotinic receptors mediate fast excitatory synaptic currents (Su and O’Dowd, 2003), while the physiological function of the muscarinic receptors is unknown. Muscarinic acetylcholine receptors (mAChRs) are G-protein coupled receptors; out of the three mAChRs in Drosophila (mAChR-A, mAChR-B and mAChR-C), mAChR-A (also called Dm1, mAcR-60C or mAChR) is the most closely homologous to mammalian mAChRs (Collin et al., 2013). Mammalian mAChRs are typically divided between ‘M1-type’ (M1/M3/M5), which signal via Gq and are generally excitatory, and ‘M2-type’ (M2/M4), which signal via Gi/0 and are generally inhibitory (Caulfield and Birdsall, 1998). Drosophila mAChR-A seems to use ‘M1-type’ signaling: when heterologously expressed in Chinese hamster ovary (CHO) cells, it signals via Gq protein (Collin et al., 2013; Ren et al., 2015) to activate phospholipase C, which produces inositol trisphosphate to release Ca2+ from internal stores.
Recent work indicates that mAChR-A is required for aversive olfactory learning in Drosophila larvae, as knocking down mAChR-A expression in KCs impairs learning (Silva et al., 2015). However, it is unclear whether mAChR-A is involved in olfactory learning in adult Drosophila, given that mAChR-A is thought to signal through Gq, and in adult flies Gq signaling downstream of the dopamine receptor Damb promotes forgetting, not learning (Berry et al., 2012; Himmelreich et al., 2017). Moreover, it is unknown how mAChR-A affects the activity or physiology of KCs, where it acts (at KC axons or dendrites or both), and how these effects contribute to olfactory learning.
Here we show that mAChR-A is required in KCs for aversive olfactory learning in adult Drosophila. Surprisingly, genetic and pharmacological manipulations of mAChR-A suggest that mAChR-A is inhibitory and acts in part on KC axons. Moreover, mAChR-A knockdown impairs the learning-associated depression of odor responses in a key MB output neuron, MB-MVP2. We suggest that mAChR-A is required to depress synapses between KCs and their outputs.
Results
mAChR-A expression in KCs is required for aversive olfactory learning in adult flies
Drosophila larvae with reduced mAChR-A expression in KCs show impaired aversive olfactory learning (Silva et al., 2015), but it remains unknown whether mAChRA in KCs also functions in learning in adult flies. We addressed this question by knocking down mAChR-A expression in KCs using two UAS-RNAi lines, “RNAi 1” and “RNAi 2” (see Methods). Only RNAi 2 requires co-expression of Dicer-2 (Dcr-2) for optimal knockdown. To test the efficiency of these RNAi constructs, we expressed them pan-neuronally using elav-GAL4 and measured their effects on mAChR-A expression levels using quantitative real time polymerase chain reaction (qRT-PCR). Both RNAi lines strongly reduce mAChR-A levels (RNAi 1: 36±4% of elav-GAL4 control, i.e., 64±4% below normal; RNAi 2: 34±2% of normal; mean±s.e.m.; see Figure 1A). We then examined whether knocking down mAChR-A in KCs using the pan-KC driver OK107-GAL4 affects short term aversive conditioning in adult flies. We used the standard training protocol and odors used in the field (i.e. 3-octanol, OCT, and 4-methylcyclohexanol, MCH; see Methods). Under these conditions both UAS-RNAi transgenes significantly reduced aversive conditioning (Figure 1B,C and Figure S1A). Knocking down mAChR-A had no effect on naive avoidance of MCH and OCT (Figure 1D; see Methods) or flies’ reaction to electric shock (Figure S1B), showing that the defect was specific to learning, rather than reflecting a failure to detect odors or shock.
Given that mAChR-A is expressed in the larval MB and indeed contributes to aversive conditioning in larvae, it is possible that developmental effects underlie the reduced learning observed in mAChR-A KD flies. To test this, we used tub-GAL80ts to suppress RNAi 1 expression during development. Flies were grown at 23°C until 3 days after eclosion and were then transferred to 31 °C for 7 days. Adult-only knockdown of mAChR-A in KCs reduced learning (Figure 1E), just as constitutive knockdown did, indicating that mAChR-A plays a physiological, not purely developmental, role in aversive conditioning. To further verify that GAL80ts efficiently blocks RNAi expression (i.e., that GAL80ts is not leaky), flies were grown at 23°C without transferring them to 31°C, thus blocking RNAi expression also in adults. When tested for learning at 10 days old, these flies showed normal learning (Figure 1E).
mAChR-A is required for olfactory learning in γ KCs, not αβ or α′β′ KCs
Kenyon cells are subdivided into three main classes: γ neurons project to the horizontal lobes only, while the axons of αβ and α′β′ neurons bifurcate to form the α and α′ portions of the vertical lobes and the β and β′ portions of the horizontal lobes. These different classes play different roles in olfactory conditioning (Guven-Ozkan and Davis, 2014; Krashes et al., 2007). To unravel in which class(es) mAChR-A functions, we used a Minos-mediated integration cassette (MiMIC) line to investigate where mAChR-A is expressed (Venken et al., 2011). The MiMIC insertion in mAChR-A lies in the first 5’ non-coding intron, creating a gene trap where GFP in the MiMIC cassette should be expressed in whichever cells endogenously express mAChR-A. Because the GFP in the original mAChR-A MiMIC cassette produced very little fluorescent signal (data not shown), we used recombinase-mediated cassette exchange (RMCE) to replace the original MiMIC cassette with a MiMIC cassette containing GAL4 (Venken et al., 2011). These new mAChR-A-MiMIC-GAL4 flies should express GAL4 wherever mAChR-A is endogenously expressed. To reveal the expression pattern of mAChR-A, we crossed mAChR-A-MiMIC-GAL4 and 20xUAS-eGFP flies. mAChR-A-MiMIC-GAL4 drove GFP expression throughout the brain, consistent with previous reports (Blake et al., 1993; Croset et al., 2018; Davie et al., 2018; Hannan and Hall, 1996) and with the fact that the Drosophila brain is mostly cholinergic. In the mushroom bodies, GFP was expressed in the αβ and γ lobes, but not the α′β′ lobes (Figure 2A). No GFP signal was observed with an inverted insertion where GAL4 is inserted in the MiMIC locus in the wrong direction (data not shown). Consistent with these MiMIC results, two recently reported databases of single-cell transcriptomic analysis of the Drosophila brain (Croset et al., 2018; Davie et al., 2018) confirm that mAChR-A is more highly expressed in αβ and γ KCs than in α′β′ KCs (Figure S2).
The higher expression of mAChR-A in αβ and γ KCs compared to α′β′ KCs suggests that learning would be impaired by mAChR-A knockdown in αβ or γ, but not α′β′, KCs. To test this, we expressed mAChR-A RNAi in different KC classes. As expected, aversive olfactory learning was reduced by knocking down mAChR-A in αβ and γ KCs together using MB247-GAL4, but not by knockdown in α′β′ KCs using c305a-GAL4. To examine if αβ and γ KCs both participate in the reduced learning observed in mAChR-A knockdown flies, we sought to limit mAChR-A RNAi expression to either αβ or γ neurons. While strong driver lines exist for αβ neurons, the γ GAL4 drivers we tested were fairly weak (H24-GAL4, MB131B, R45H04-GAL4, data not shown), perhaps too weak to drive mAChR-A-RNAi enough to knock down mAChR-A efficiently. Therefore, we used MB247-GAL4, which was strong enough to affect behavior, and blocked GAL4 activity in either αβ or γ KCs by expressing the GAL80 repressor under the control of R44E04-LexA (αβ KCs) or R45H04-LexA (γ KCs) (Bräcker et al., 2013). These combinations drove strong, specific expression in αβ or γ KCs (Figure S3). Learning was reduced by mAChR-A RNAi expression in γ, but not αβ, KCs (Figure 2B). These results suggest that mAChR-A is specifically required in γ KCs for aversive olfactory learning and short-term memory.
mAChR-A suppresses odor responses in γ KCs
We next asked what effect mAChR-A knockdown has on the physiology of KCs, by expressing GCaMP6f and mAChR-A RNAi 2 together in KCs using OK107-GAL4. Knocking down mAChR-A in KCs increased odor-evoked Ca2+ influx in the mushroom body calyx, where KC dendrites reside (Figure 3). This result is somewhat surprising because mAChR-A is a Gq coupled receptor whose activation leads to Ca2+ release from internal stores (Ren et al., 2015), which predicts that mAChR-A knockdown should decrease, not increase, odor-evoked Ca2+ influx in KCs. However, some examples have been reported of inhibitory signaling through Gq by M1-type mAChRs (see Discussion), and Drosophila mAChR-A may join these as another example of an inhibitory mAChR signaling through Gq.
Because mAChR-A is required for aversive learning in γ KCs, not αβ or α′β′ KCs (Figure 2), we next asked if odor responses in αβ, α′β′ and γ KCs are differentially affected by mAChR-A knockdown. αβ, α′β′ and γ KC dendrites are not clearly segregated in the calyx, so we examined odor responses in the axonal lobes. Indeed, although odor responses in all lobes were increased by mAChR-A knockdown, only in the γ lobe was the effect statistically significant for both MCH and OCT (Figure 3). This result is consistent with the behavioral requirement for mAChR-A only in γ KCs. However, we do not rule out the possibility that mAChR-A knockdown affects αβ and α′β′ odor responses in a more subtle way that does not affect short-term memory, especially as αβ and α′β′ odor responses were somewhat, though non-significantly, increased.
Do increased odor responses in γ KCs prevent learning by increasing the overlap between the γ KC population representations of the two odors used in our task (Lin et al., 2014)? When GCaMP6f and mAChR-A-RNAi 2 were expressed in all KCs, mAChRA knockdown did not affect the sparseness or inter-odor correlation of KC population odor responses (Figure 4A-C) even though it increased overall calyx responses. To focus specifically on γ KCs, we expressed GCaMP6f and mAChR-A-RNAi 1 only in γ KCs, using mb247-Gal4, R44E04-LexA and lexAop-GAL80, as in Figure 2B. GCaMP6f was visible mainly in the γ lobe (Figure 4D). γ-only expression of mAChR-A-RNAi 1 increased odor responses in the calyx (here, dendrites of γ KCs only) and, in the case of OCT, in the γ lobe (Figure 4E,F). Note that γ KC odor responses are increased by both RNAi 1 (Figure 3A,B) and RNAi 2 (Figure 4E,F). As with pan-KC expression, γ-only expression of mAChR-A-RNAi 1 did not affect the sparseness or inter-odor correlation of γ KCs (Figure 4G-I). Thus, mAChR-A knockdown does not impair learning through increased overlap in KC population odor representations.
KC odor responses are decreased by an mAChR agonist
RNAi-based knockdown of mAChR-A might induce homeostatic compensation that obscures or even reverses the primary effect of reduced mAChR-A expression. To test the acute role of mAChR-A in regulating KC activity, we took the complementary approach of pharmacologically activating mAChR-A. Initially we bath-applied 10 μM muscarine, an mAChR-A agonist (Drosophila mAChR-B is 1000-fold less sensitive to muscarine than mAChR-A is (Collin et al., 2013), and mAChR-C is not expressed in the brain (Davie et al., 2018)). Muscarine strongly decreased odor responses in all subtypes of KCs (Figure 5A,B). However, muscarine did not significantly affect odor responses in PNs (Figure 5C), suggesting that the effect of muscarine on KCs arose in KCs, not earlier in the olfactory pathway. KCs can be silenced by an inhibitory GABAergic neuron called the anterior paired lateral (APL) neuron (Lin et al., 2014; Masuda-Nakagawa et al., 2014; Papadopoulou et al., 2011), so we asked whether muscarine reduces KC odor responses indirectly by activating APL, rather than directly inhibiting KCs. We applied muscarine to flies with APL-specific expression of tetanus toxin (TNT), which blocks inhibition from APL and thereby greatly increases KC odor responses. In these flies, APL is labeled stochastically, so hemispheres where APL was unlabeled served as controls (Lin et al., 2014) (see Methods). Muscarine decreased KC odor responses both in control hemispheres and hemispheres where APL synaptic output was blocked by tetanus toxin (Figure 5D), and the effect of muscarine was not significantly different between the two cases (Figure 5E). This result indicates that muscarine does not act solely by activating APL or by enhancing inhibition on KCs (e.g., increasing membrane localization of GABAA receptors).
To further clarify whether muscarine directly affects KC axons, we locally applied muscarine to the MB horizontal lobe by pressure ejection (Figure 6). Red dye included in the ejected solution confirmed that the muscarine did not spread to the calyx (Figure 6B). In the absence of odor stimuli, locally applied muscarine on the horizontal lobe decreased GCaMP6f signal in the β and γ lobes during the period 0.5 – 1.5 s after application, suggesting that mAChR-A reduces baseline Ca2+ levels in some KCs (Figure 6A,C). Interestingly, in α′β′ KCs, local muscarine application caused a slower rise in GCaMP6f signal at 3 – 4 s after application (Figure 6A,D), suggesting that mAChR-A has different effects in different KCs or on different timescales. To test whether locally applied muscarine also modulates odor-evoked Ca2+ influx, we applied muscarine 1 s before odor onset. Locally applied muscarine decreased KC odor responses for both odors in the γ and β lobes, and for OCT in all lobes (Figure 6F,G). The more obvious effect with OCT might arise from OCT activating more KCs than MCH does (i.e., MCH responses were sparser: Figure 4B, Table S1), consistent with previous reports (Lin et al., 2014; Perisse et al., 2013). Although the red dye suggests that muscarine applied to the horizontal lobe did not spread substantially to the vertical (α and α′) lobes (Figure 6B), these lobes may have been affected in the case of OCT because muscarine reached KC axons at the branch point. Notably, with only one exception, locally applied muscarine had no effect on either GCaMP6f signal or odor responses on the opposite hemisphere from the site of application (Figure 6G); nor did muscarine applied on the horizontal lobe affect GCaMP6f signal in the calyx (Figure 6A). In the case of OCT, during muscarine application to the horizontal lobe, odor responses in the calyx were lower on both the same side and opposite side, but given that the opposite side was affected in no other case, this exception may reflect sensory adaptation between the ‘before’ and ‘+muscarine’ recordings, rather than an effect of muscarine per se. Together, these results suggest that mAChR-A acts on KC axons to inhibit local Ca2+ influx.
To test what effect mAChR-A exerts on the calyx, we locally applied muscarine to the calyx. Surprisingly, applying muscarine to the calyx in the absence of odor stimuli increased GCaMP signal in the calyx and α lobe (Figure 6A,E). It also decreased GCaMP signal in the α′ and β′ lobes around 1-2 s after application (Figure 6A), although this effect was not statistically significant. The increased Ca2+ in the calyx most likely did not reflect increased excitability, as applying muscarine to the calyx did not increase the calyx odor response (Figure 6F,G). If anything, it likely decreased the calyx odor response, because the Ca2+ increase induced by muscarine alone (no odor) lasted ~6-7 s and thus would have continued into the odor pulse in the muscarine + odor condition. If the odor response was unaffected by muscarine, the muscarine-evoked and odor-evoked increases in GCaMP6f signal should have summed. Instead, the peak calyx ΔF/F during the odor pulse was the same before and after locally applying muscarine, suggesting that the specifically odor-evoked increase in GCaMP6f was decreased by muscarine. Indeed, applying muscarine to the calyx suppressed odor responses in KC axons (Figure 6F,G). Given that calyx muscarine suppresses α′β′ axonal odor responses, the decrease in α′β′ KC GCaMP signal in the absence of odor likely reflects suppression of spontaneous action potentials (Figure 6A,E), as α′β′ KCs have the highest spontaneous spike rate out of the three subtypes (Groschner et al., 2018; Turner et al., 2008). The increase in calyx Ca2+ induced by muscarine alone (without odor) might reflect Ca2+ release from internal stores triggered by Gq signaling, which then inhibits KC excitability (thus smaller odor responses). Note that muscarine on the calyx is unlikely to reduce KC odor responses via presynaptic inhibition of PNs, because bath muscarine does not affect PN odor responses in the calyx (Figure 5C), although we cannot rule out Ca2+-independent inhibition. Interestingly, KC axons in many cases show opposite responses to local muscarine application in the horizontal lobe vs. the calyx (e.g., α′β′ and γ KCs in Figure 6A), suggesting that mAChR-A may act via different signaling pathways in KC dendrites and axons.
mAChR-A knockdown prevents training-induced depression of MBON odor responses
The finding that muscarine can locally inhibit Ca2+ influx in KC axons suggests that mAChR-A may play a role in the weakening of KC->MBON synapses that underlies olfactory learning. For example, mAChR-A might be required in KC presynaptic terminals to integrate the US (shock) and CS (odor) to induce synaptic plasticity. If so, increasing the strength of the US might overcome the aversive learning defect in mAChR-A knockdown flies. In contrast, if mAChR-A functions only in dendrites and modulates how KCs process the CS, not the US or CS-US integration, changing the US strength should not affect the learning defect in mAChR-A knockdown flies. We tested this by increasing the electric shock strength from 50 V (as in Figure 1A) to 90 V (Figure 7A). Indeed, increased electric shock strength improved learning (compare scores in Figure 7A with Figure 1A). In particular, mAChR-A knockdown flies no longer show any learning defect. As only the US input to KC presynaptic terminals was modified while the CS (i.e. odor input to the calyx) was unchanged, this result is consistent with the hypothesis that mAChR-A exerts its effect on learning and memory at KC presynaptic terminals. This result also backs up the imaging results in Figure 4 showing that mAChR-A knockdown does not affect KC odor coding (i.e., the sparseness or inter-odor correlation of KC population responses).
We sought to further test whether mAChR-A contributes to the synaptic plasticity underlying aversive olfactory learning. In Drosophila, olfactory associative memories are stored by weakening the synapses between KCs and output neurons that lead to the “wrong” behavior. For example, aversive memory requires an output neuron downstream of γ KCs, called MBON-γ1pedc>α/β or MB-MVP2. MB-MVP2 leads to approach behavior, and aversive conditioning reduces MB-MVP2’s responses to the aversively-trained odor (Hige et al., 2015b; Perisse et al., 2016). We tested whether knocking down mAChR-A would prevent this depression. We knocked down mAChR-A in KCs using OK107-GAL4 and UAS-mAChR-A-RNAi 1, and expressed GCaMP6f in MB-MVP2 using R12G04-LexA and lexAop-GCaMP6f (Figure 7B). We trained flies in the behavior apparatus and then imaged MB-MVP2 odor responses (3 h after training to avoid cold-shock-sensitive memory). Because overall response amplitudes were variable across flies, for each fly we measured the ratio of the response to MCH (the trained odor) over the response to OCT (the untrained odor). Consistent with previous published results (Hige et al., 2015b; Perisse et al., 2016), in control flies not expressing mAChR-A RNAi, the MCH/OCT ratio was substantially reduced in trained flies relative to mock-trained flies (Figure 7C). This was not because the OCT response increased, because there was no difference between trained and mock-trained flies in the ratio of the response to OCT over the response to isoamyl acetate, a ‘reference’ odor that was absent in the training protocol. In contrast, in flies expressing mAChR-A RNAi in KCs, the MCH/OCT ratio was the same between trained and mock-trained flies (Figure 7C), indicating that the mAChR-A knockdown impaired the learning-related depression of the KC to MB-MVP2 synapse. These results support the notion that mAChR-A plays a role at the presynaptic terminal of KCs where LTD occurs.
Discussion
Here we show that mAChR-A is required in γ KCs for aversive olfactory learning and short-term memory in adult Drosophila. Knocking down mAChR-A increases KC odor responses, while the mAChR-A agonist muscarine suppresses KC activity at both KC dendrites and axons. Knocking down mAChR-A prevents aversive learning from reducing responses of the MB output neuron MB-MVP2 to the conditioned odor, suggesting that mAChR-A is required for the learning-related depression of KC->MBON synapses.
Why is mAChR-A only required for learning in γ KCs, not αβ or α′β′ KCs? Although our mAChR-A MiMIC gene trap agrees with single-cell transcriptome analysis that α′β′ KCs express less mAChR-A than do γ and αβ KCs (Croset et al., 2018; Davie et al., 2018), transcriptome analysis indicates that α′β′ KCs do express some mAChR-A (Figure S2). Moreover, γ and αβ KCs express similar levels of mAChR-A. It may be that the RNAi knockdown is less efficient at affecting the physiology of αβ and α′β′ KCs than γ KCs, whether because the knockdown is less efficient at reducing protein levels, or because αβ and α′β′ KCs have different intrinsic properties or a different function of mAChR-A such that 30% of normal mAChR-A levels is sufficient in αβ and α′β′ KCs but not γ KCs. This interpretation is supported by our finding that mAChR-A RNAi knockdown significantly increases odor responses only in the γ lobe, not the αβ or α′β′ lobes. Alternatively, γ, αβ and α′β′ KCs are thought to be important mainly for short-term memory, long-term memory, and memory consolidation, respectively (Guven-Ozkan and Davis, 2014; Krashes et al., 2007); as we only tested short-term memory, mAChRA may carry out the same function in all KCs, but only its role in γ KCs is required for short-term (as opposed to long-term) memory. Indeed, the key plasticity gene DopR1 is required in γ, not αβ or α′β′ KCs, for short-term memory (Qin et al., 2012).
mAChR-A seems to inhibit KC odor responses, because knocking down mAChRA increases odor responses in the calyx and γ lobe, while activating mAChR-A with bath or local application of muscarine decreases KC odor responses. Some details differ between the genetic and pharmacological results. In particular, while mAChR-A knockdown mainly affects γ KCs, with other subtypes inconsistently affected, bath application of muscarine reduces responses in all KC subtypes, and local axonal application of muscarine reduces responses in the γ and β lobe (for MCH) or all lobes (for OCT). What explains these differences? mAChR-A might be weakly activated in physiological conditions, in which case gain of function would cause a stronger effect than loss of function. Similarly, pharmacological activation of mAChR-A is likely a more drastic manipulation than a two-thirds reduction of mAChR-A mRNA levels. Although we cannot entirely rule out network effects from muscarine application, the effect of muscarine does not stem from PNs or APL (Figure 5C,D) and locally applied muscarine would have little effect on neurons outside the mushroom body. A previous report found that local application of ACh at the horizontal lobe did not affect GCaMP5 signal in KC axons, but this experiment may have missed the small dip in GCaMP6f signal (~ -5-10% ΔF/F in the γ lobe), and could not have measured modulation of odor responses as it was done ex vivo (Barnstedt et al., 2016).
How does mAChR-A inhibit odor-evoked Ca2+ influx in KCs? Given that mAChRA signals through Gq when expressed in CHO cells (Ren et al., 2015), that muscarinic Gq signaling normally increases excitability in mammals (Caulfield and Birdsall, 1998), and that pan-neuronal artificial activation of Gq signaling in Drosophila larvae increases overall excitability (Becnel et al., 2013), it may be surprising that mAChR-A inhibits KCs. However, Gq signaling may exert different effects on different neurons in the fly brain, and some examples exist of inhibitory Gq signaling by mammalian mAChRs. M1/M3/M5 receptors acting via Gq can inhibit voltage-dependent Ca2+ channels (Gamper et al., 2004; Kammermeier et al., 2000; Keum et al., 2014; Suh et al., 2010), reduce voltage-gated Na+ currents (Cantrell et al., 1996), or trigger surface transport of KCNQ channels (Jiang et al., 2015), thus increasing inhibitory K+ currents. Drosophila mAChRA may inhibit KCs through similar mechanisms.
What function does mAChR-A serve in learning and memory? Our results indicate that mAChR-A knockdown prevents the learning-associated weakening of KC-MBON synapses, in particular for MBON-γ1pedc>α/β, aka MB-MVP2 (Figure 7). One potential explanation is that the increased odor-evoked Ca2+ influx observed in knockdown flies increases synaptic release, which overrides the learning-associated synaptic depression. However, increased odor-evoked Ca2+ influx per se is unlikely on its own to straightforwardly explain a learning defect, because other genetic manipulations that increase odor-evoked Ca2+ influx in KCs either have no effect on, or even improve, olfactory learning. For example, knocking down GABA synthesis in the inhibitory APL neuron increases odor-evoked Ca2+ influx in KCs (Lei et al., 2013; Lin et al., 2014) and improves olfactory learning (Liu and Davis, 2008).
Alternatively, mAChR-A may be involved directly in learning-associated synaptic depression. In mammals, where the principal excitatory neurotransmitter is glutamate (vs. ACh in Drosophila), metabotropic glutamate receptors (mGluRs) underlie many forms of LTD (Lüscher and Huber, 2010), most notably at the synapse between parallel fibers and Purkinje cells in the cerebellum, where LTD is thought to underlie motor learning (Jörntell and Hansel, 2006). The mushroom body has many architectural similarities to the cerebellum, and the KC-MBON synapse appears to fulfill an analogous role to the parallel fiber–Purkinje cell synapse (Farris, 2011). Using a metabotropic receptor for the main excitatory neurotransmitter to mediate synaptic depression may be yet another conserved feature of this ‘cerebellum-like’ circuit. Although LTD occurs postsynaptically at the parallel–fiber-Purkinje cell synapse and presynaptically at the KC-MBON synapse, mGluRs can also mediate presynaptic LTD (Pinheiro and Mulle, 2008). In addition, mAChRs are also involved in synaptic plasticity and memory in mammals. M1 mAChR knockout mice have impaired synaptic plasticity and altered memory (Anagnostaras et al., 2003), and M1 mAChRs together with metabotropic glutamate receptors mediate LTD in the hippocampus (Jo et al., 2010; Kamsler et al., 2010; Volk et al., 2007). It may be that mAChR-A similarly co-operates with dopamine receptors to depress KC-MBON synapses. Note that the impaired synaptic depression in mAChR-A knockdown flies could explain the increased odor responses in KC axons. KC-MBON synapses are modulated by individual flies’ experiences before their physiology can be measured experimentally, a phenomenon that requires the plasticity gene rutabaga (Hige et al., 2015a). If such plasticity involves depression of KC-MBON synapses and is impaired in mAChR-A knockdown flies, that could explain why mAChR-A knockdown flies have increased odor-evoked Ca2+ influx in KC axons. Regardless, mAChR-A likely also inhibits KCs directly, given that muscarine inhibits Ca2+ influx in KCs on a time scale of ~1 s.
What role might mAChR-A play in synaptic plasticity? An intriguing possibility is suggested by an apparent paradox: both mAChR-A and the dopamine receptor Damb signal through Gq (Himmelreich et al., 2017), but mAChR-A promotes learning while Damb promotes forgetting (Berry et al., 2012). How can Gq mediate apparently opposite effects? Perhaps Gq signaling aids both learning and forgetting by generally rendering the synapse more labile. Indeed, although damb mutants retain memories for longer than wildtype, their initial learning is slightly impaired (Berry et al., 2012); damb mutant larvae are also impaired in aversive olfactory learning (Selcho et al., 2009). Although one study reports that knocking down Gq in KCs did not impair initial memory (Himmelreich et al., 2017), the Gq knockdown may not have been strong enough; also, that study shocked flies with 90 V shocks, which also gives normal learning in mAChRA knockdown flies (Figure 7A).
For mAChR-A to contribute to KC-MBON synaptic depression, which is thought to occur in KC presynaptic terminals, mAChR-A must act at least in part at KC presynaptic terminals. An axonal site of action is supported by our finding that local muscarine application on the horizontal lobe decreases GCaMP6f signal and odor responses in KC axons, including those of γ KCs, the subtype of KCs where mAChR-A is required for learning. Mammalian mAChRs often mediate presynaptic inhibition, most commonly by M2-type mAChRs (Allen and Brown, 1993; Bellingham and Berger, 1996; Slutsky et al., 1999) but also in some cases by Mi-type mAChRs (de Vin et al., 2015; Kamsler et al., 2010; Sheridan and Sutor, 1990). Drosophila mAChR-A may combine the signaling pathway of M1-type mAChRs (Gq rather than Gi/o) with inhibitory action more typical of M2-type mAChRs.
What is the source of ACh which activates mAChR-A and modulates odor responses? In the calyx, cholinergic PNs are certainly a major source of ACh. However, KCs themselves are cholinergic (Barnstedt et al., 2016) and release neurotransmitter in both the calyx and lobes (Christiansen et al., 2011). In the lobes, KCs are the only known source of ACh, as cholinergic MBONs do not have presynaptic specializations in the MB and non-KC intrinsic MB neurons like APL and DPM are not thought to be cholinergic (Haynes et al., 2015; Wu et al., 2013). Thus, mAChRs could function as autoreceptors to prevent excess release of an excitatory neurotransmitter, as in mammals, where metabotropic glutamate receptors mediate presynaptic inhibition on glutamatergic neurons to prevent excess glutamate release (Scanziani et al., 1997). However, mAChRs may also mediate lateral interactions between KCs. Numerous KC-KC synapses have been seen by electron microscopy both in Drosophila (Takemura et al., 2017) and other insects (Leitch and Laurent, 1996; Schürmann, 2016; Strausfeld and Li, 1999). Thus, mAChR-A may mediate lateral inhibition, in conjunction with the lateral inhibition provided by the GABAergic APL neuron (Lin et al., 2014), or KC-KC signaling may enhance memory by aiding LTD via mAChR-A.
Methods
Fly Strains
Fly strains (see below) were raised on cornmeal agar under a 12 h light/12 h dark cycle and studied 1–10 days post-eclosion. Strains were cultivated at 25 °C unless they expressed temperature-sensitive gene products (GAL80ts); in these cases the experimental animals and all relevant controls were grown at 23 °C. To de-repress the expression of RNAi with GAL80ts, experimental and control animals were incubated at 31 °C for 7 days. Subsequent behavioral experiments were performed at 25 °C.
Experimental animals carried transgenes over Canton-S chromosomes where possible to minimize genetic differences between strains. The following transgenes were used: UAS-GCaMP6f, lexAop-GCaMP6f (Barnstedt et al., 2016; Chen et al., 2013), UAS-mAChR-A RNAi 1 (TRiP.JF02725, Bloomington #27571), UAS-mAChR-A RNAi 2 (VDRC ID 101407), UAS-Dcr-2 (Bloomington #24651), lexAop-GAL80 (Bloomington #32216), tub-GAL80ts (McGuire et al., 2003), mb247-dsRed (Riemensperger et al., 2005), GH146-GAL4 (Stocker et al., 1997), OK107-GAL4 (Connolly et al., 1996), c305a-GAL4 (Krashes et al., 2007), mb247-GAL4 (Zars, 2000), R44E04-LexA (Bloomington #52736), R45H04-LexA (Bräcker et al., 2013), R12G04-LexA (Bloomington #52448) (Jenett et al., 2012), elav-GAL4 (Lin and Goodman, 1994), NP2631-GAL4, GH146-FLP, tub-FRT-GAL80-FRT, UAS-TNT, UAS-mCherry, mb247-LexA (Lin et al., 2014), 20xUAS-6xGFP (Shearin et al., 2014), UAS-mCD8-GFP.
Behavioral Analysis
Behavioral experiments were performed in a custom-built, fully automated apparatus (Claridge-Chang et al., 2009; Lin et al., 2014; Parnas et al., 2013). Single flies were housed in clear polycarbonate chambers (length 50 mm, width 5 mm, height 1.3 mm) with printed circuit boards (PCBs) at both floors and ceilings. Solid-state relays (Panasonic AQV253) connected the PCBs to a 50 V source.
Air flow was controlled with mass flow controllers (CMOSens PerformanceLine, Sensirion). A carrier flow (2.7 l/min) was combined with an odor stream (0.3 l/min) obtained by circulating the air flow through vials filled with a liquid odorant. Odors were prepared at 10 fold dilution in mineral oil. Therefore, liquid dilution and mixing carrier and odor stimulus stream resulted in a final 100 fold dilution of odors. Fresh odors were prepared daily.
The 3 liter/min total flow (carrier and odor stimulus) was split between 20 chambers resulting in a flow rate of 0.15 l/min per half chamber. Two identical odor delivery systems delivered odors independently to each half of the chamber. Air or odor streams from the two halves of the chamber converged at a central choice zone. The 20 chambers were stacked in two columns each containing 10 chambers and were backlit by 940 nm LEDs (Vishay TSAL6400). Images were obtained by a MAKO CMOS camera (Allied Vision Technologies) equipped with a Computar M0814-MP2 lens. The apparatus was operated in a temperature-controlled incubator (Panasonic MIR-154) maintained at 25 °C.
A virtual instrument written in LabVIEW 7.1 (National Instruments) extracted fly position data from video images and controlled the delivery of odors and electric shocks. Data were analyzed in MATLAB 2015b (The MathWorks) and Prism 6 (GraphPad).
A fly’s preference was calculated as the percentage of time that it spent on one side of the chamber. Training and odor avoidance protocols were as depicted in Figure 1. The naïve avoidance index was calculated as (preference for left side when it contains air) – (preference for left side when it contains odor). During training, MCH was paired with 12 equally spaced 1.25 s electric shocks at 50 V (Tully and Quinn, 1985). The learning index was calculated as (preference for MCH before training) – (preference for MCH after training). Flies were excluded from analysis if they entered the choice zone fewer than 4 times during odor presentation.
Functional Imaging
Brains were imaged by two-photon laser-scanning microscopy (Ng et al., 2002; Wang et al., 2003). Cuticle and trachea in a window overlying the required area were removed, and the exposed brain was superfused with carbogenated solution (95% O2, 5% CO2) containing 103 mM NaCl, 3 mM KCl, 5 mM trehalose, 10 mM glucose, 26 mM NaHCO3, 1 mM NaH2PO4, 3 mM CaCl2, 4 mM MgCl2, 5 mM N-Tris (TES), pH 7.3.
Odors at 10-1 dilution were delivered by switching mass-flow controlled carrier and stimulus streams (Sensirion) via software controlled solenoid valves (The Lee Company). Flow rates at the exit port of the odor tube were 0.5 or 0.8 l/min.
Fluorescence was excited by a Ti-Sapphire laser centered at 910 nm, attenuated by a Pockels cell (Conoptics) and coupled to a galvo-resonant scanner. Excitation light was focussed by a 20X, 1.0 NA objective (Olympus XLUMPLFLN20XW), and emitted photons were detected by GaAsP photomultiplier tubes (Hamamatsu Photonics, H10770PA-40SEL), whose currents were amplified and transferred to the imaging computer. Two imaging systems were used, #1 for Figures 3–6 except 5C, and #2 for Figure 5C and Figure 7, which differed in the following components: laser (1: Mai Tai eHP DS, 70 fs pulses; 2: Mai Tai HP DS, 100 fs pulses; both from Spectra-Physics); microscope (1: Movable Objective Microscope; 2: DF-Scope installed on an Olympus BX51WI microscope; both from Sutter); amplifier for PMT currents (1: Thorlabs TIA-60; 2: Hamamatsu HC-130-INV); software (1: ScanImage 5; 2: MScan 2.3.01). Volume imaging on System 1 was performed using a piezo objective stage (nPFocus400, nPoint). Muscarine was applied locally by pressure ejection from patch pipettes (resistance ~10 MOhm; capillary inner diameter 0.86 mm, outer diameter 1.5 mm; concentration in pipette 20 mM; pressure 12.5 psi) using a Picospritzer III (Parker). A red dye was added to the pipette to visualize the ejected fluid (SeTau-647, SETA BioMedicals) (Podgorski et al., 2012).
Movies were motion-corrected in X-Y using the moco ImageJ plugin (Dubbs et al., 2016), with pre-processing to collapse volume movies in Z and to smooth the image with a Gaussian filter (standard deviation = 4 pixels; the displacements generated from the smoothed movie were then applied to the original, unsmoothed movie), and motion-corrected in Z by maximizing the pixel-by-pixel correlation between each volume and the average volume across time points. ΔF/F, activity maps, sparseness and inter-odor correlation were calculated as in (Lin et al., 2014). We excluded non-responsive flies and flies whose motion could not be corrected.
Structural Imaging
Brain dissections, fixation, and immunostaining were performed as described (Pitman et al., 2011; Wu and Luo, 2006). To visualize native GFP fluorescence, dissected brains were fixed in 4% (w/v) paraformaldehyde in PBS (1.86 mM NaH2PO4, 8.41 mM Na2HPO4, 175 mM NaCl) and fixed for 20 min at room temperature. Samples were washed for 3×20 min in PBS containing 0.3% (v/v) Triton-X-100 (PBT). The neuropil was counterstained with nc82 (DSHB) and goat anti-mouse Alexa 647 or Alexa 564. Primary antisera were applied for 1-2 days and secondary antisera for 1-2 days in PBT at 4 °C, followed by embedding in Vectashield. Images were collected on a Leica TCS SP5, SP8, or Nikon A1 confocal microscope and processed in ImageJ.
APL expression of tetanus toxin was scored by widefield imaging of mCherry. mCherry expression in APL was distinguished from 3XP3-driven dsRed from the GH146-FLP transgene by using separate filter cubes for dsRed (49004, Chroma: 545/25 excitation; 565 dichroic; 605/70 emission) and mCherry (LED-mCherry-A-000, Semrock: 578/21 excitation; 596 dichroic; 641/75 emission).
Statistics
Statistical analyses were carried out in GraphPad Prism as described in figure legends and Table S1. In general, no statistical methods were used to predetermine sample sizes, but where conclusions were drawn from the absence of a statistically significant difference, a power analysis was carried out in G*Power to confirm that the sample size provided sufficient power to detect an effect of the expected size. The experimenter was blind to which hemispheres had APL neurons expressing tetanus toxin before post-experiment dissection (Figure 5) but not otherwise.
Author contributions
NB: Methodology, investigation, formal analysis, writing–review & editing, visualization. HA: Methodology, investigation, formal analysis, writing-original draft, writing—review & editing, visualization. AAA: Methodology, investigation, formal analysis, writing–review & editing, visualization. ER: Methodology, investigation, formal analysis, software, writing–review & editing, visualization. HL: Investigation, formal analysis, writing-review & editing. WH: Investigation, writing-review & editing, visualization. ACL: Conceptualization, methodology, investigation, formal analysis, software, writing–original draft, writing–review & editing, visualization, supervision, funding acquisition. MP: Initiated the project, conceptualization, methodology, investigation, formal analysis, software, writing–original draft, writing–review & editing, visualization, supervision, funding acquisition.
Acknowledgments
We thank Vincent Croset, Christoph Treiber and Scott Waddell for sharing mAChR-A expression data before publication. We thank Oren Schuldiner, Andreas Thum, Scott Waddell, the Bloomington Stock Center, the Vienna Drosophila RNAi Center, and the Kyoto Drosophila Genetic Resource center for fly strains. We thank Anton Nikolaev for comments on the manuscript. This work was supported by the European Research Council (676844, MP; 639489, AL).