Optogenetic induction of appetitive and aversive taste memories in Drosophila

  1. Meghan Jelen
  2. Pierre-Yves Musso
  3. Pierre Junca
  4. Michael D Gordon  Is a corresponding author
  1. Department of Zoology and Life Sciences Institute, University of British Columbia, Canada

Abstract

Tastes typically evoke innate behavioral responses that can be broadly categorized as acceptance or rejection. However, research in Drosophila melanogaster indicates that taste responses also exhibit plasticity through experience-dependent changes in mushroom body circuits. In this study, we develop a novel taste learning paradigm using closed-loop optogenetics. We find that appetitive and aversive taste memories can be formed by pairing gustatory stimuli with optogenetic activation of sensory neurons or dopaminergic neurons encoding reward or punishment. As with olfactory memories, distinct dopaminergic subpopulations drive the parallel formation of short- and long-term appetitive memories. Long-term memories are protein synthesis-dependent and have energetic requirements that are satisfied by a variety of caloric food sources or by direct stimulation of MB-MP1 dopaminergic neurons. Our paradigm affords new opportunities to probe plasticity mechanisms within the taste system and understand the extent to which taste responses depend on experience.

Editor's evaluation

Through a new operant learning assay and fly genetics, this important work convincingly shows that taste memory formation requires the same circuit substrates and mechanisms as olfactory memory formation. While the exact mechanisms remain to be elucidated, the convincing data and approach represent a valuable foundation for the study of molecular and circuit mechanism underpinning taste memory formation and the role of brain energy therein. This study will be of particular interest to the large community of scientists studying the mechanisms and circuits of memory formation in the fly and possibly beyond.

https://doi.org/10.7554/eLife.81535.sa0

Introduction

Food selection is influenced by a complex set of factors including external sensory input, interoceptive circuits signaling internal state, and plasticity driven by past feeding experiences. The gustatory system plays a critical role in evaluating the nutritional qualities of foods, and is generally thought to evoke innate appetitive or aversive behavioral responses. However, the degree to which taste processing can be modified by learning is unclear.

In flies, taste detection is mediated by gustatory receptor neurons (GRNs) located on the proboscis, pharynx, legs, wing margins, and ovipositor (Stocker, 1994). GRNs express a range of chemosensory receptors for detecting sugars, bitters, salts, and other contact chemical cues (Chen and Dahanukar, 2020). GRNs project to the subesophageal zone (SEZ) of the fly brain, where taste information is segregated based on modality, valence, and organ of detection (Marella et al., 2006; Thorne et al., 2004; Wang et al., 2004).

Although the valence of a specific taste is generally set, the intensity of the response can vary substantially according to internal state. Starvation increases a fly’s sensitivity to sweet tastes and blunts bitter responses through direct modulation of GRN activity (Inagaki et al., 2014; Inagaki et al., 2012; LeDue et al., 2016; Marella et al., 2012). Moreover, flies lacking essential nutrients such as amino acids and salts exhibit increased nutrient-specific preference toward foods containing those substances (Corrales-Carvajal et al., 2016; Jaeger et al., 2018; Steck et al., 2018).

In addition to internal state-dependent changes in nutrient drive, fly taste responses can be altered by experience. Most notably, short-term taste-specific suppression of appetitive responses can be achieved through pairing the appetitive taste with either bitter stimulation or noxious heat (Keene and Masek, 2012; Kirkhart and Scott, 2015; Masek et al., 2015; Tauber et al., 2017). This plasticity requires an integrative memory association area called the mushroom body (MB), which is known to represent different sensory modalities, including olfaction and taste (Cohn et al., 2015; Davis, 2005; Keene and Masek, 2012; Keene and Waddell, 2007; Kirkhart and Scott, 2015; Masek et al., 2015; Schwaerzel et al., 2003). Thus, while taste responses are executed by innate circuits, they also exhibit experience-dependent changes driven by the adaptable networks of the MBs (Colomb et al., 2009; Kirkhart and Scott, 2015; Krashes et al., 2009).

The MBs are composed of approximately ~4000 intrinsic Kenyon cells (KCs), whose dendrites receive inputs from different sensory systems (Kirkhart and Scott, 2015; Schwaerzel et al., 2003; Tanaka et al., 2008; Vogt et al., 2014). KCs form en passant synapses with mushroom body output neurons (MBONs), and MBONs send projections to neuropils outside of the MBs to modulate behavior (Crittenden et al., 1998; Tanaka et al., 2008). Each MBON receives KC input in a specific region of the MB called a ‘compartment’, and activation of an MBON typically evokes either a positive (approach) or negative (avoidance) valence (Perisse et al., 2013). Integration across many MBON responses is thought to produce the final behavioral valence and intensity (Aso et al., 2014).

Much of what we know about the mechanisms of associative memory formation in the MB comes from studies pairing olfactory stimuli with either sugar (reward) or electric shock (punishment). In these paradigms, an odor serves as the conditioned stimulus (CS) and produces sparse activation of a unique combination of KCs (Beck et al., 2000; Tempel et al., 1983; Tully, 1984; Tully and Quinn, 1985). Meanwhile, sugar or electric shock serves as the unconditioned stimulus (US) by evoking activity in distinct populations of dopaminergic neurons (DANs) – protocerebral anterior medial (PAM) DANs are activated by sugar, while protocerebral posterior lateral 1 (PPL1) DANs are activated by shock (Burke and Waddell, 2011; Gervasi et al., 2010; Mao and Davis, 2009; Tomchik and Davis, 2009). DANs target specific MB compartments, where dopamine functions to depress the synaptic connections between active KCs and the compartment’s MBONs (Aso et al., 2014; Cohn et al., 2015; Hige et al., 2015; Perisse et al., 2013). Strikingly, rewarding PAM DANs generally target compartments with MBONs carrying negative valence, while PPL1 DANs target compartments with MBONs carrying positive valence. Thus, the resulting change in synaptic weights following concurrent activation of KCs with either PAM or PPL1 skews behavior toward either approach or avoidance (Aso et al., 2010; Perisse et al., 2013).

Consistent with this model, direct activation of DANs in the absence of any rewarding or punishing stimulus can function as a US in some fly associative learning paradigms (Aso et al., 2012; Claridge-Chang et al., 2009; Colomb et al., 2009; Liu et al., 2012). Optogenetic or thermogenetic activation of PAM DANs following, or in coincidence with, an odor results in the formation of an appetitive memory. Meanwhile, activation of punishing PPL1 DANs leads to the formation of an aversive memory (Cohn et al., 2015; Yamagata et al., 2015). A similar phenomenon has also been demonstrated in mice, where phasic optogenetic activation of specific dopaminergic subsets can lead to the formation of conditioned behaviors, even in the absence of a physical reward (Saunders et al., 2018).

DAN populations are also segregated by the type of memory formed, as appetitive short-term memories (STM) and long-term memories (LTM) are formed by independent PAM subpopulations (Burke and Waddell, 2011; Colomb et al., 2009; Musso et al., 2015; Yamagata et al., 2015). Moreover, whereas STM may be formed with a sweet tasting reward on its own, the formation of LTM requires a sweet and nutritious US (Burke and Waddell, 2011; Musso et al., 2015). Caloric sugars are thought to gate memory consolidation by promoting sustained rhythmic activity of MB-MP1 DANs (Musso et al., 2015; Plaçais et al., 2017; Plaçais et al., 2012). Interestingly, this signaling may occur up to 5 hr post ingestion, suggesting that there is a critical time window for the formation of LTM (Musso et al., 2015; Pavlowsky et al., 2018).

Although flies are known to exhibit aversive short-term taste memories, the full extent to which taste behaviors are modifiable by learning is unknown. Can taste responses be enhanced by appetitive conditioning? Can flies form LTM about taste? These are difficult questions to answer using traditional methods for several reasons. First, appetitive association paradigms generally rely on food as the US, which interferes with the representation of a taste CS and can also modify future taste behaviors through changes in satiety state. Second, taste is an active sense, and animals typically have behavioral control over exposure to the stimulus. Thus, repeated temporal pairing of a taste CS with a US is difficult to achieve in flies without immobilization, making LTM difficult to test. Moreover, the self-control over taste exposure under more natural conditions could support operant learning with neural and molecular mechanisms that are distinct from classic olfactory conditioning (Brembs, 2009).

To probe the potential of taste learning, we developed an optogenetic learning paradigm that couples a taste (the CS) with optogenetic GRN or DAN stimulation (the US). Using this novel paradigm, we show that flies can form both appetitive and aversive short- and long-term taste memories. As in olfaction, appetitive taste memories are driven by discrete PAM populations, and activation of a single PAM subpopulation is sufficient to induce appetitive LTM. The formation of appetitive LTM requires de novo protein synthesis and is contingent on caloric intake. Moreover, sugar, certain amino acids, and lactic acid can provide the energy required to support LTM formation, and this requirement is also satisfied by thermogenetic activation of MB-MP1 neurons.

Results

Pairing GRN activation with a food source leads to taste memory formation

We previously developed a system called the sip-triggered optogenetic behavioral enclosure (STROBE), in which individual flies are placed in an arena with free access to two odorless food sources (Musso et al., 2019). Interactions (mostly sips) with one food source triggers nearly instantaneous activation of a red LED, which can be used for optogenetic stimulation of neurons expressing CsChrimson. We reasoned that sipping on a tastant (the CS+) that triggers activation of neurons providing either positive or negative reinforcement may produce a change in the number of interactions a fly initiates upon subsequent exposure to the same CS+ (Figure 1A).

Figure 1 with 1 supplement see all
Gustatory receptor neurons (GRNs) produce punishment and reward signals capable of facilitating taste memory formation.

(A) Diagram outlining sip-triggered optogenetic behavioral enclosure (STROBE) memory paradigm. Training: 24 hr starved flies freely interact with a LED-activating tastant (CS+) and a non-LED-activating tastant (CS-) for 40 min. LED activation stimulates CsChrimson-expressing neurons. Testing: associative memory is measured by assessing a fly’s preference for the CS+ tastant compared to the CS- for 1 hr. In the short-term memory (STM) assay testing occurs 10 min after training. In the long-term memory (LTM) assay testing occurs 24 hr after training. (B) Aversive STM measured after pairing 25 mM sucrose (CS+) with bitter neuron optogenetic activation. Preference indices (left) and tastant interactions (right) for Gr66a>CsChrimson flies compared to controls during training and testing. The interaction numbers for individual flies are connected by lines. (C) Cumulative average preference indices over the course of training and testing in (B), (n=16–30). (D) Appetitive STM measured after pairing 75 mM NaCl (CS+) with sweet neuron optogenetic activation. Preference indices (left) and interactions (right) for Gr43a>CsChrimson flies compared to controls in the short-term memory assay. (E) Preference index of flies in (D) over time during training and testing (n=12–23). (F) Appetitive LTM measured after pairing of 75 mM NaCl (CS+) with sweet neuron optogenetic activation. Preference indices (left) and interactions (right) for Gr43a>CsChrimson flies compared to controls in the LTM assay. (G) Average preference index as a function of time for the training and testing in the LTM assay (n=14–30). All flies were starved for 24 hr prior to training. Preference index is mean ± SEM, Kruskal-Wallis with Dunn’s multiple comparison test: **p < 0.01, ****p < 0.0001.

We began by testing the efficacy of the STROBE in inducing aversive and appetitive memories through optogenetic activation of bitter and sweet GRNs, respectively. Bitter GRN stimulation is known to activate PPL1 DANs, while sweet GRNs activate PAMs (Keene and Masek, 2012; Kirkhart and Scott, 2015; Liu et al., 2012; Masek et al., 2015). Moreover, bitter or sweet GRN activation with Gr66a- or Gr43a-Gal4 is sufficient for STM induction in taste and olfactory associative learning paradigms (Keene and Masek, 2012; Yamagata et al., 2015). Therefore, we tested whether pairing GRN activation with feeding on a single taste modality could create an associative taste memory that altered subsequent behavior to the taste.

In the aversive taste memory paradigm, interactions with 25 mM sucrose (CS+) during training triggered LED activation of Gr66a bitter neurons expressing CsChrimson (Figure 1—figure supplement 1A). This led to CS+ avoidance relative to plain agar (CS-) during training (Figure 1B). During testing, we disabled the STROBE lights and measured preference toward 25 mM sucrose (CS+) relative to agar (CS-) to see if flies have formed aversive taste memories. Indeed, 10 min after training, flies that experienced bitter GRN activation during training showed a lower sugar preference than control flies lacking the obligate CsChrimson cofactor all-trans-retinal or not expressing CsChrimson. Like most of the experiments that will follow, there was high variance in the behavior of individual flies during both training and testing, undoubtedly reflecting a combination of individual variation in internal state, past experiences, response to training, as well as stochastic effects during the measurement time period. Examining the preference indices over time revealed that the difference in preference emerged after about 30 min of testing, which could either reflect a progressive divergence in behavior between the groups or, more likely, increased reliability of the preference measurement as sips accumulate over time (Figure 1C). A similar aversive memory was also produced by the activation of PPK23glut ‘high salt’ GRNs, which carry a negative valence in salt-satiated flies (Figure 1—figure supplement 1A, B). Importantly, these effects are not due to heightened satiety in trained flies, because training in this paradigm is associated with fewer food interactions than controls (Figure 1C and Figure 1—figure supplement 1B).

For appetitive training, we chose 75 mM NaCl as the CS+, since flies show neither strong attraction nor aversion to this concentration of salt (Zhang et al., 2013). Interactions with the CS+ in this paradigm triggered optogenetic activation of sweet neurons, either with Gr43a-Gal4, which labels a subset of leg and pharyngeal sweet neurons in addition to fructose-sensitive neurons in the protocerebrum, or Gr64f-Gal4, which labels most peripheral sweet GRNs (Figure 1—figure supplement 1A). In both cases, sweet GRN activation produced an increased preference for the salt CS+ during training and testing 10 min later (Figure 1D and Figure 1—figure supplement 1C). The increased preference is evident early during testing and maintained throughout the testing phase (Figure 1). Like the aversive memory paradigm, the effects of appetitive conditioning cannot easily be explained through changes in internal state, since trained flies interacted more with the food during training and therefore should have a lower salt drive during testing. Interestingly, refeeding flies with standard medium directly after training in the appetitive paradigm led to a long-term preference for the CS+, revealed by testing 24 hr later (Figure 1F and G and Figure 1—figure supplement 1D). This stands in contrast to the aversive paradigm, where reduced preference for sugar following bitter GRN activation was absent 24 hr later (Figure 1—figure supplement 1A, E).

DAN activation is sufficient for the induction of short- and long-term taste memories

We next asked whether direct activation of DANs during feeding could drive the formation of taste memories. Aversive short-term taste memory depends on multiple PPL1 DANs, including PPL1-α’2 α2 and PPL1-α3 (Masek et al., 2015), while appetitive short-term taste memories have not been previously reported. We first tested whether activating PPL1 DANs coincident with tastant interactions would lead to STM formation in the STROBE. Stimulation of PPL1 neurons reduced sucrose preference during training, and a reduced preference was also observed during STM testing 10 min later (Figure 2A). This decreased preference was sustained throughout the entire period of testing (Figure 2B). Interestingly, unlike activation of bitter sensory neurons, PPL1 activation also produced a long-term aversive memory that was expressed 24 hr after training and remained stable through the duration of testing (Figure 2C and D).

Figure 2 with 1 supplement see all
PPL1 and protocerebral anterior medial (PAM) neural activation is sufficient for the induction of short- and long-term taste memories.

(A) Aversive short-term memories (STM) measured following PPL1 neuron optogenetic activation paired with 25 mM sucrose (CS+) vs agar (CS-). Flies lacking retinal or one genetic element for expression of CsChrimson serve as controls (n=19–31). (B) Preference indices over time for the experiment shown in (A). (C) Aversive long-term memories (LTM) measured following PPL1 optogenetic activation paired with 25 mM sucrose (CS+) vs agar (CS-) (n=20–33). (D) Preference indices over time for the experiment shown in (C). (E) Appetitive STM measured following PAM neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=25–38). (F) Preference indices over time for the experiment shown in (E). (G, H) Appetitive LTM measured following PAM neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=17–35). Flies were refed with standard food for 1 hr directly after training unless otherwise indicated as delayed refeeding (8 hr after training) or no refeeding. (H) Preference indices over time for the experiment shown in (G). All flies were food deprived for 24 hr prior to the start of experimentation. Preference indices are mean ± SEM, Kruskal-Wallis with Dunn’s multiple comparison test: *p < 0.05, **p < 0.01, ***p < 0.001.

To test the effect of appetitive DAN activation, we used flies expressing CsChrimson in PAM neurons under control of the broad PAM driver R58E02-Gal4. Intriguingly, although optogenetic activation of PAM neurons signals reward to the MB, it did not affect immediate preference toward light-paired 75 mM NaCl (CS+) during training. Nonetheless, this pairing resulted in appetitive memory expression during testing 10 min and 24 hr after training (Figure 2E and G). These taste memories were stable throughout the entire duration of testing (Figure 2F and H). Thus, optogenetic activation of PAM neurons in the STROBE was able to write both short- and long-term appetitive taste memories in the absence of acute effects on feeding.

Given that the taste memories we observe are created in a novel and uncharacterized paradigm, we did additional experiments with PAM activation to establish that the memories were specific to the CS+ tastant. First, flies trained with NaCl as the CS+ and agar as the CS- showed no preference between two identical agar options during testing, ruling out the possibility that the increased CS+ preference observed in prior experiments was driven by a spatial memory or other non-CS+ local cues such as deposited pheromones (Figure 2—figure supplement 1A). This remains true when using sweet sensory neuron activation as the US, which, unlike PAM stimulation, drives elevated preference for the salt option during training (Figure 2—figure supplement 1B). Next, we found that a second tastant, monopotassium glutamate (MPG), could replace NaCl as the CS+. MPG is approximately equally appetitive to NaCl (Figure 2—figure supplement 1C), and pairing of MPG with PAM activation resulted in a robust appetitive memory to MPG (Figure 2—figure supplement 1D). Moreover, training with NaCl as the CS+ and MPG as the CS- produced an appetitive memory for NaCl (Figure 2—figure supplement 1E). Finally, flies trained with NaCl as the CS+ and agar as the CS- did not show elevated preference for MPG introduced as a novel tastant during testing (Figure 2—figure supplement 1F). All these observations support the conclusion that memories formed during STROBE training are taste memories specific to the trained CS+.

We also sought to establish the energy requirements for appetitive LTM formed through PAM activation. Based on the critical role of energy in long-term olfactory memory formation (Musso et al., 2015; Plaçais et al., 2017; Plaçais et al., 2012), we designed our LTM paradigm to include a brief 1 hr exposure to food after training. To confirm the necessity of this feeding, we tested flies that were not fed after training or were fed 7 hr post training, after the memory consolidation time period defined in olfactory memory (Figure 2G). Neither of these groups expressed taste memories during testing. Thus, the contingencies governing the formation and expression of taste memories in Drosophila appear similar to those previously discovered for olfaction.

The MBs are required for short- and long-term taste memory formation

The intrinsic neurons of the MB are required for aversive taste memory formation (Masek et al., 2015). To demonstrate that the MBs are also required for appetitive taste memory formation, we silenced KCs throughout both our STM and LTM assays using tetanus toxin expressed under control of the pan-KC driver R13F02-LexA. KC silencing eliminated both short-term and long-term appetitive memories formed by activation of Gr43a sensory neurons (Figure 3A and B) or PAM neurons labeled by the driver R58E02-Gal4 (Figure 3C and D). These findings indicate that MB intrinsic neurons play a pivotal role in the formation of appetitive taste memories.

The mushroom body (MB) is required for the formation of short- and long-term taste memories.

(A, B) Appetitive short-term memories (STM) (A) and long-term memories (LTM) (B) measured following sweet taste neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) with Kenyon cells (KCs) silenced by expression of tetanus toxin (n=16–34 for STM and n=13–27 for LTM). Controls are missing one genetic element for KC silencing and therefore exhibit memory. (C, D) Appetitive STM (C) and LTM (D) measured following protocerebral anterior medial (PAM) neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) with KCs silenced by expression of tetanus toxin (n=24–28 for STM and n=17–23 for LTM). Controls are missing one genetic element for KC silencing and therefore exhibit memory. Memory assays when the MB is silenced, compared to controls. (E) Appetitive LTM measured following PAM neuron optogenetic activation with 75 mM NaCl (CS+) vs agar (CS-). Flies were either fed retinal or retinal plus cycloheximide (n=17–22). (F) Model of appetitive taste memory formation via gustatory receptor neuron (GRN)/PAM activation. All flies were starved for 24 hr prior to training. Preference indices are mean ± SEM, Kruskal-Wallis with Dunn’s multiple comparison test (A–D) or Mann-Whitney test (E): *p < 0.05, **p < 0.01.

Olfactory LTM requires de novo protein synthesis during memory consolidation (Colomb et al., 2009). To test whether the same is true for taste memories, we fed flies the protein synthesis inhibitor cycloheximide (CXM). As expected, flies fed CXM prior to training were unable to form long-term taste memories, in contrast to vehicle controls (Figure 3E). These results confirm that the taste memories being formed are protein synthesis dependent, consistent with the classic characteristics of LTM (Figure 3F).

Distinct PAM subpopulations induce appetitive short- and long-term taste memories

Distinct subpopulations of PAM neurons – those targeting β’2, γ4, and γ5 compartments labeled by R48B04-Gal4 and those targeting α1, β’1, β2, and γ5 compartments labeled by R15A04-Gal4 – mediate the formation of appetitive short- and long-term olfactory memories, respectively (Yamagata et al., 2015). Moreover, it has been hypothesized that two differential reinforcing effects of sugar reward – sweet taste and nutrition – are encoded by these segregated STM and LTM neural populations (Yamagata et al., 2015). We tested both populations in our appetitive STROBE memory assays to determine if the activation of these separate PAM clusters would support the formation of parallel short- and long-term taste memories. Indeed, activation of the β’2, γ4, and γ5 regions drove appetitive short-term but not long-term taste memories, as shown by the higher salt preference of flies expressing active CsChrimson during STM testing but not LTM testing (Figure 4A and B). Conversely, activation of the α1, β’1, β2, and γ5 compartments produced LTM but not STM (Figure 4C and D). These results indicate that, much like appetitive olfactory memory, short- and long-term taste memories are formed by distinct PAM subpopulations.

Figure 4 with 1 supplement see all
Distinct protocerebral anterior medial (PAM) subpopulations induce appetitive short- and long-term taste memories.

(A, B) Appetitive short-term memories (STM) (A) and long-term memories (LTM) (B) measured following β’2, γ4, and γ5 PAM neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=21–28 for STM and n=15–17 for LTM). (C, D) Appetitive STM (C) and LTM (D) measured following α1, β’1, β2, and γ5 PAM neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=11–15 for STM and n=20–27 for LTM). (E, F) Appetitive STM (E) and LTM (F) measured following PAM-α1 neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=11–14 for STM and n=19–22 for LTM). (G, H) Appetitive STM (G) and LTM (H) measured following PAM-β2β′2a neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=20–27 for STM and n=10–15 for LTM). All flies were starved for 24 hr prior to training. Preference indices are mean ± SEM, Mann-Whitney test: **p < 0.01.

Next, we wondered whether activation of a single PAM cell subtype, PAM-α1, would be sufficient to induce taste memories. PAM-α1 neurons project to an MB compartment innervated by MBON-α1, which in turn feeds back onto PAM-α1 to form a recurrent reward loop necessary for the formation of appetitive olfactory LTM (Aso and Rubin, 2016; Ichinose et al., 2015). Consistent with its role in olfactory memory, activation of this PAM cell type in the STROBE with drivers MB043B-Gal4 or MB299B-Gal4 was sufficient to drive appetitive long-term, but not short-term, taste memory formation (Figure 4E and F and Figure 4—figure supplement 1A, B).

Interestingly, activation of the PAM-β2β′2a subset labeled by MB301B-Gal4 produced a higher preference for the salt CS during training, yet no sustained changes in taste preference during STM or LTM testing were observed (Figure 4G and H). This demonstrates that the reward signaling associated with PAM cell activation occurs on multiple timescales to produce acute, short-, or long-term changes in behavior, consistent with past results demonstrating the context-dependent effects of DAN activation (Rohrsen et al., 2021) Notably, the trend toward lower salt preference during testing in this experiment may reflect a reduced salt drive due to increased salt consumption during training.

Caloric food sources are required for the formation of associative long-term taste memories

Because refeeding with standard fly medium shortly after training is permissive for the consolidation of appetitive long-term taste memories, we next asked what types of nutrients support memory formation. As expected, refeeding with L-glucose, a non-caloric sugar, did not lead to formation of associative long-term taste memories (Figure 5A and B). However, along with sucrose, refeeding with lactic acid, yeast extract, and L-alanine promoted LTM, while L-aspartic acid did not. These results indicate that, in addition to sucrose, other caloric nutrients can provide sufficient energy for long-term taste memory formation. Moreover, 7 hr delayed refeeding of each nutrient failed to support memory formation (Figure 5B). Thus, similar to olfactory LTM, the formation of appetitive taste LTM is dependent on an energy source being readily available during the memory consolidation window (Fujita and Tanimura, 2011; Musso et al., 2015).

Caloric food sources are required for the formation of associative long-term taste memories.

(A) Schematic of the conditions and mushroom body (MB) compartments innervated by broad protocerebral anterior medial (PAM) driver R58E02-Gal4. (B, C) Training (B) and testing (C) of appetitive long-term memories (LTM) measured following PAM neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=13–28). Flies were fed the indicated compounds for 1 hr immediately after training or after an 8 hr delay where indicated. All flies were starved for 24 hr prior to training. Preference indices are mean ± SEM, Kruskal-Wallis with Dunn’s multiple comparison test: *p < 0.05, **p < 0.01.

Our findings concerning the formation and expression of appetitive taste LTM bear striking similarities to those of olfactory LTM in terms of MB circuitry, dependence on protein synthesis, and energetic requirements. This led us to wonder if MB-MP1 neurons, which signal onto the MB and promote energy flux in MB neurons during LTM, perform a similar function in taste memory (Musso et al., 2015; Plaçais et al., 2017; Plaçais et al., 2012). To test this hypothesis, we activated MB-MP1 neurons directly after training using UAS-TRPA1 and delayed refeeding to outside the memory consolidation window. Compared to genetic controls, flies in which MB-MP1 neurons were activated post training showed significantly elevated memory scores during testing (Figure 6A and B). This confirms that MB-MP1 activation is sufficient to drive memory consolidation during long-term appetitive taste memory formation (Figure 6C).

Mushroom body (MB)-MP1 neuron activation post training replaces energy signal required for the formation of long-term memories (LTM).

(A) Appetitive LTM measured following protocerebral anterior medial (PAM) neuron optogenetic activation paired with 75 mM NaCl (CS+) vs agar (CS-) (n=18 = 29). Flies were transferred to 29°C for 1 hr immediately after training to thermogenetically activate MB-MP1 using R30E11>TRPA1. Controls are lacking one genetic element for MB-MP1 thermogenetic activation. Graphic of timeline followed for the LTM taste assay with thermogenetic activation of MB-MP1 neurons. All flies were starved for 24 hr prior to training. Preference indices are mean ± SEM, Kruskal-Wallis with Dunn’s multiple comparison test: **p < 0.01.

Discussion

Gustation plays a vital role in determining the suitability of foods for ingestion. Yet, little is known about how experience influences higher-order taste representations and contributes to the continuous refinement of food selection. In fact, a memory system for the recollection of appetitive taste memories has not been described in flies. In this study, we use the STROBE to establish a novel learning paradigm and further investigate the formation and expression of taste memories. We demonstrate that flies can form short- and long-term appetitive and aversive taste memories toward two key nutrients – salt and sugar. Much like olfactory memory, associative taste memory formation occurs within the MB and follows many of the same circuit and energetic principles.

It is perhaps not surprising that olfactory and taste memories share common principles; however, important distinctions exist between olfactory and taste learning paradigms that justify the possibility that this may not have been the case. Most notably, using taste as a CS in a free feeding situation where reinforcement is temporally coupled to food contact creates the potential for operant, rather than classical, conditioning. These two types of learning can employ distinct neural circuits in rodents (Ostlund and Balleine, 2007) and are separable by their synaptic properties and molecular mechanisms in invertebrates (Brembs, 2009; Brembs and Plendl, 2008; Hawkins and Byrne, 2015). Nevertheless, when flies are tethered in a flight arena and punishment is predicted by a mixture of classical (color) and operant (self-motion) cues, the classical conditioning system overrides the operant conditioning system (Brembs and Plendl, 2008). In the STROBE, our data suggests that the fly learns the tastant as a classical cue, despite the operant component of the reinforcement contingencies. Thus, the conservation between olfactory and taste learning mechanisms is consistent with past studies.

Although aversive taste memories have been established, prior evidence for appetitive taste memories has been sparse. Rats’ hedonic response to bitter compounds can be made more positive through pairing with sugar, and human studies suggest that children’s taste palates are malleable based on positive experiences with bitter vegetables (Breslin et al., 1990; Figueroa et al., 2020; Forestell and LoLordo, 2000; Wadhera et al., 2015). Therefore, despite the difficulties of measuring taste memories in the lab, appetitive taste plasticity is very likely an ethologically important process.

We observed enhanced salt feeding following pairing of salt taste with sweet sensory neuron stimulation. This may be surprising, given that NaCl on its own activates sweet GRNs (Jaeger et al., 2018; Marella et al., 2006). However, 75 mM NaCl moderately activates only about one third of sweet GRNs (Dweck et al., 2022), and thus appetitive memory formation may be driven by strong activation of the broader sweet neuron population. Nevertheless, using direct stimulation of DANs as the US afforded us the ability to reduce this complication and also interrogate the roles of specific DAN populations. Taking a hypothesis-driven approach, we confirmed that PAM neural subpopulations reinforce taste percepts much like olfactory inputs, and that STM and LTM are processed by distinct subpopulations. For example, activating β’2, γ4, and γ5 compartments with R48B04-Gal4 produces STM in both olfactory and taste paradigms, while activation of α1, β’1, β2, and γ5 with R15A04-Gal4 produces LTM in both. These results confirm that appetitive short- and long-term taste memories are processed in parallel in the MB (Trannoy et al., 2011; Yamagata et al., 2015). Given that tastes, like odors, activate the KC calyces (Kirkhart and Scott, 2015), we speculate that optogenetic stimulation of PAM neurons during feeding modulates the strength of KC-MBON synaptic connections. Notably, activation of single PAM cell types produced different forms of memory in the STROBE. For example, stimulating PAM-α1 neurons during feeding drives appetitive taste LTM, while activation of PAM-β’1 was immediately rewarding.

The activation of bitter GRNs paired with sucrose led to the formation of STM, which agrees with previous research demonstrating that thermogenetic stimulation of bitter GRNs can negatively reinforce short-term taste learning (Keene and Masek, 2012). However, unlike sweet neuron activation, bitter neuron activation was not sufficient for the formation of LTM in our assay. One possible explanation is that the strong feeding inhibition evoked by bitter GRN activation leads to an insufficient number of CS-US pairings to induce LTM. Consistent with this idea, PPL1 activation, which induced LTM, is less aversive than bitter neuron activation during training, and therefore allows more associations.

A unique aspect of our long-term taste learning paradigm is that we uncoupled the US from a caloric food source. By doing this we were able to probe the energetic constraints gating LTM formation. It has long been reported that LTM formation in Drosophila requires the intake of caloric sugar. Here, we demonstrate that the caloric requirements of LTM formation can be fulfilled by food sources other than sucrose, including lactic acid and yeast extract. Moreover, it seems that at least one amino acid, L-alanine, is able to provide adequate energy, while others like L-aspartic acid cannot. We theorize that these foods may provide flies with readily accessible energy, as neurons are able to metabolize both lactic acid and L-alanine into pyruvate to fuel the production of ATP via oxidative phosphorylation (de Tredern et al., 2021).

Energy gating in the MB is thought to be regulated by the MB-MP1-DANs. MB-MP1 neuron oscillations activate increased mitochondrial energy flux within the KCs, which is both necessary and sufficient to support LTM (Plaçais et al., 2017). To demonstrate sufficiency in our assay, we activated MB-MP1 neurons with TRPA1 directly after fly training. This effectively substitutes for a caloric food source and allows LTM formation (Figure 6C). These results suggest that MB-MP1 neurons integrate energy signals during the formation of multiple types of LTM, and may be influenced by a variety of caloric foods.

Despite the advantages of replacing natural stimuli with optogenetic stimulation, there are also limitations. Most notably, optogenetic activation may not closely replicate temporal dynamics, intensity, or population features of natural stimulus encoding. Contact with food in the STROBE activates LED illumination with a relatively low latency of about 37±17 ms, but would not be expected to precisely mimic the onset of activation from natural taste stimuli (Musso et al., 2019). Moreover, bitter and acidic stimuli are known to evoke OFF responses that would not be replicated in the STROBE (Devineni et al., 2021; Stanley et al., 2021). Given the proposed importance of GRN temporal dynamics to higher-order neuronal plasticity, the critical role of timing between KC and DAN activation for MB plasticity, and the broad importance of timing to various synaptic plasticity mechanisms, it is easy to imagine that temporal differences between optogenetics and natural stimuli could differentially affect learning (Cohn et al., 2015; Devineni et al., 2021; Handler et al., 2019). Optogenetics and natural stimuli also undoubtedly activate different neuron populations. For example, Gr64f-Gal4 labels most sweet sensory neurons, but the distribution of these neurons on different taste organs makes coincident activation of all these populations unlikely under natural conditions (Fujii et al., 2015). Conversely, direct DAN stimulation affects only a small subset of the neurons activated upon sugar taste detection and consumption, and likely therefore does not capture all of the effects that sugar has on appetitive conditioning (Wang et al., 2004). Nevertheless, the features of optogenetic activation are clearly sufficient to drive plasticity and learning.

Overall, our results suggest that lasting changes in the value of specific tastes can occur in response to temporal association with appetitive or aversive stimuli, raising the possibility that such plasticity plays an important role in animals’ ongoing taste responses. It is interesting to speculate on what could serve as the US under more natural conditions. One obvious possibility is that pairing of different tastes (e.g. sugar and salt) in complex foods allows one taste to serve as the US and modifies future responses to the other. Intriguingly, tastes may also have the ability to self-reinforce over time, as shown for some odors (Kato et al., 2022). Another possibility is that natural association of tastes with non-taste reinforcers such as pain or mating could modify subsequent behavior. Future experiments using the STROBE paradigm could further probe the molecular and circuit mechanisms underlying taste memories and advance our understanding of how taste preferences may be shaped by experience over an animal’s lifetime.

Materials and methods

Fly strains

Request a detailed protocol

Fly stocks were raised on a standard cornmeal diet at 25°C, 70% relative humidity. For neuronal activation, 20XUAS-IVS-CsChrimson.mVenus (BDCS, stock number: 55135) was used. Dopaminergic PAM expression was targeted using previously described lines: R58E02-GAL4 (Musso et al., 2015); R58E02-LexA, R48B04-GAL4, R15A04-GAL4, R13F02-LexA, and R30E11-LexA obtained from Bloomington (BDCS, stock numbers: 52740, 50347, 48671, 52460, 54209); and MB split-GAL4 lines MB043B-GAL4, MB504B-GAL4, MB299B-GAL4, MB301B-GAL4 from Janelia Research Campus (Aso et al., 2014). GRN expression was driven using Gr43a-GAL4, Gr64f-GAL4 (Dahanukar et al., 2007), Gr66a-GAL4 (Wang et al., 2004), and PPK23glut-GAL4, PPK23-GAL4, Gr66a-LexA::VP16, LexAop-Gal80 (Jaeger et al., 2018). LexAop-tnt was previously described (Liu et al., 2016). For temperature activation experiments, LexAop-TrpA1 was used (Liu et al., 2012).

STROBE experiments

Request a detailed protocol

Mated female Drosophila were collected 2–3 days post eclosion and transferred into vials containing 1 ml of standard cornmeal medium supplemented with 1 mM all-trans-retinal (Sigma #R2500) or an ethanol vehicle control. Flies were maintained on this diet for 2 days in a dark environment. 24 hr prior to experimentation, flies were starved at 25°C, 70% relative humidity, on 1% agar supplemented with 1 mM all-trans-retinal or ethanol vehicle control.

STROBE training protocol

Request a detailed protocol

During the training phase for the STM experiments the STROBE was loaded with 4 μl of tastant (salt: Sigma #S7653 or sucrose: Sigma #S7903) on channel 1 and 4 μl 1% agar on channel 2. The red LED was triggered only when a fly interacted with the tastant in channel 1. The duration of the training period was 40 min. For the STM training protocol, flies were then transferred to clean empty vials for 10 min while the experimental apparatus was cleaned. The training and testing phases of LTM experiments were performed as described for the STM experiments with the following exception: after the 40 min training period flies were transferred individually into vials containing standard cornmeal diet or nutrient of interest in 1% agar (500 mM sucrose: Sigma #S7903, 500 mM L-glucose: Sigma #G5500, 250 mM lactic acid: Sigma #69785, 10% yeast extract: Sigma #Y1625, 250 mM L-alanine: Sigma #05129, 250 mM L-aspartic acid: Sigma #11230) and allowed to feed for 1 hr. They were then transferred into 1% agar starvation vials and kept at 18°C until the testing component of the experiment. For MB-MB1 activation experiments, after training flies were placed at 29°C, 70% relative humidity for 1 hr on 1% agar starvation vials. They were then transferred to 18°C and refed 8 hr later, outside of the memory consolidation. After 1 hr of feeding they were once again transferred into 1% agar starvation vials and kept at 18°C until the retrieval component of the experiment. The preference index for each individual fly was calculated as: (sips from channel 1 – sips from channel 2)/(sips from channel 1+sips from channel 2). All experiments were performed with a light intensity of 11.2 mW/cm2 at 25°C, 70% relative humidity.

STROBE testing protocol

Request a detailed protocol

During testing, 4 μl of the same tastant (salt: Sigma #S7653, sucrose: Sigma #S7903, MPG: Sigma #G1501) was reloaded into channel 1 and 4 μl of 1% agar on channel 2. The optogenetic component of the system was deactivated such that the red LED would no longer trigger if a fly interacted with the tastant. Flies were reloaded individually into the same arenas. The duration of the testing phase was 1 hr. The preference index for each individual fly was calculated as: (sips from channel 1 – sips from channel 2)/(sips from channel 1+sips from channel 2).

Immunofluorescence microscopy

Request a detailed protocol

Brain staining protocols were performed as previously described (Chu et al., 2014). Briefly, brains were fixed for 1 hr in 4% paraformaldehyde and dissected in PBS + 0.1% Triton-X. After dissection brains were blocked in 5% NGS diluted with PBST for 1 hr. Brains were probed overnight at 4°C using the following primary antibody dilutions: rabbit anti-GFP (1:1000, Invitrogen #A11122, RRID: AB_221569) and mouse anti-brp (1:50, DSHB #nc82, RRID: AB_2392664). After a 1 hr wash period, secondary antibodies – goat anti-rabbit Alexa-488 (1:200, Invitrogen #A11008, RRID: AB_143165) and goat anti-mouse Alexa-568 (1:200, Invitrogen #A11030, RRID: AB_2534072) – were applied and incubated for 1 hr at room temperature to detect primary antibody binding. Slowfade gold was used as an antifade mounting medium.

Slides were imaged under a 25× water immersion objective using a Leica SP5 II Confocal microscope. All images were taken sequentially with a z-stack step size at 1 µm, a line average of 2, speed of 200 Hz, and a resolution of 1024×1024 pixels. ImageJ was used to compile slices into a maximum intensity projection (Jaeger et al., 2018).

Statistical analysis

Request a detailed protocol

All statistical analyses were executed using GraphPad Prism 6 software. Sample size and statistical tests performed are provided in the Figure legends. Non-parametric tests were used because data did not always adhere to a normal distribution. For Dunn’s multiple comparison tests, the experimental group was compared to all controls and the highest p-value reported over the experimental bar. Replicates are biological replicates, using different individual flies from two or more crosses. Sample sizes were based on previous experiments in which effect size was determined. Data was excluded on the basis of STROBE technical malfunctions for individual flies and criteria for data exclusion are as follows: (i) if the light system was not working during training for individual arenas, (ii) if during training or testing a fly did not meet a standard minimum # of interactions for that genotype, (iii) if during training or testing the STROBE recorded an abnormally large # of interactions for that genotype, (iv) technical malfunctions due to high channel capacitance baseline activity, and (v) if a fly was dead in an arena.

Code availability

Request a detailed protocol

All STROBE software is available for download from GitHub:

FPGA code: https://github.com/rcwchan/STROBE-fpga ( Chan, 2018a).

All other code: https://github.com/rcwchan/STROBE_software/ (Chan, 2018b).

Data availability

All data generated or analyzed during this study are included in the manuscript; spreadsheets of raw numerical data are provided as source data files attached to each figure.

References

    1. Tully T
    2. Quinn WG
    (1985) Classical conditioning and retention in normal and mutant Drosophila melanogaster
    Journal of Comparative Physiology. A, Sensory, Neural, and Behavioral Physiology 157:263–277.
    https://doi.org/10.1007/BF01350033

Decision letter

  1. Ilona C Grunwald Kadow
    Reviewing Editor; University of Bonn, Germany
  2. Claude Desplan
    Senior Editor; New York University, United States
  3. Thomas Dieter Riemensperger
    Reviewer; University of Cologne, Germany

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Optogenetic induction of appetitive and aversive taste memories in Drosophila" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Claude Desplan as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Thomas Riemensperger (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

All three reviewers expressed a high interest and, provided you can address the following issues, are supportive of publication of your manuscript in eLife. The reviewers were particularly impressed by the assay you have developed and its potential in resolving more dynamic aspects of perception and learning. In a revised version of the manuscript, please address especially the following points:

1. The reviewers recommend taking better advantage of your operant assay. Please show the data as original traces and not just box plots for the learning and recall phase. In addition, please explain why this is not 'simply' associative learning with one pairing as usually used in standard olfactory learning assays.

2. Explain the putative perceptual differences between optogenetic activation and real stimuli, for instance for sugar reward vs. exogenous activation of sweet taste neurons. Along the same lines, please show the data during learning as traces to help evaluate how actual stimulus perception during training impacts on learning.

3. Add reciprocal controls for at least one of the conditions.

4. We recommend making the manuscript more accessible to scientists not familiar with Drosophila memory paradigms by providing more explanation and interpretation of your data.

Reviewer #1 (Recommendations for the authors):

A few suggestions to improve the manuscript further:

Introduction:

(Line numbering would have been helpful…)

You write at the end of the first page, that DANs are activated by reward and punishment. They are also activated/modulated, for instance, by metabolic state and movement. Both of these aspects might be highly relevant in your operant assay.

End of p 3: 'while maintaining similar satiation states': this statement contrasts somewhat with some other statements you make regarding interaction with the food source and even the shown bar graphs during training. This needs to be taken into consideration and explained a little better.

Beginning of page 8 and Figures 5 and 6: please correct the references to and/or labeling of the panels. In addition, I did not find the data for the 7-hour delayed refeeding. Figure 6C does not exist.

In the schemes, the light dot is always shown only on one side of the brain. Is this correct that you are activating only unilaterally? I guess not and I found it somewhat confusing…

There were several typos or small mistakes that must be corrected.

Reviewer #2 (Recommendations for the authors):

Figures:

The authors describe the learning as a preference index, which is calculated from the number of interactions (sips) with the CSs. Does also the length of sip bouts change? Maybe it would be worth looking also at overall feeding time. This could be especially interesting as flies that have been trained with GR66a activation, still prefer the sugar CS over several minutes at the beginning of the test. Maybe these are just very short bouts. Is there an explanation for this phenotype? Are the flies maybe more food deprived, because they receive less sugar in the training?

It is unclear what L/R means in Figure 1C/E/G, as this is also not explained in the legend. I assume it is the left/right side of CSs. Could this be renamed as CS+/Cs-, or red light/ no red light?

The LLM/LTM can only be induced by feeding caloric food to the flies directly after training, this is a very important piece of information that does not show up in the figures. The authors should explain this fact a bit better in the text and maybe add this below the relevant graphs too (refeeding, yes/no), this would be especially helpful in Figure 6 where they do not receive food at all after training? For people not familiar with LTM this is hard to follow.

In Figure 3 some of the dots below the graphs do not seem to be in the right place. Please check 3A/B, they are different from 3C/D, but should be the same?

Discussion:

The paradigm that the authors present uses operant training, where the fly can decide how much they want to eat from each tastant presented. It would be great to have a paragraph in the discussion where the similarities and differences to previously studied classical conditioning are elucidated and results are compared. The authors could also cite work from the Brembs lab. In terms of reinforcement vs. US function of dopamine neurons, there was a recent biorxiv paper that could be cited if that is possible.: Pain is so close to pleasure: the same dopamine neurons can mediate approach and avoidance in Drosophila | bioRxiv Rohrsen et al., 2021.

Reviewer #3 (Recommendations for the authors):

The manuscript Optogenetic induction of appetitive and aversive taste memories in Drosophila by Jelen and colleagues is interesting, well written and the experiments elaborated, and the data provided by the authors justifies the title of the manuscript. The technique provided by the authors is very interesting and will attract a broad readership mainly of the Drosophila community as gustatory learning paradigms are yet tricky and not feasible for mass assays. The novel technique provided by Jelen and colleagues will pave the way to investigate gustatory learning now in more detail. In addition, the fact that the animals can freely move greatly improves the possibilities to investigate gustatory learning. Despite the high potential of the manuscript to reach out to a broad readership I would like to address some points of concern that the authors may want to address before publication.

In the first set of experiments the authors pair the artificial activation of bitter gustatory neurons with a low concentrated sugar solution as (CS+), whereas pure agar is presented as (CS-). To control for the learnability of a conditioned stimulus the authors have to ensure that the (CS-) and (CS+) are both equally learnable. Therefore, the authors should on the one hand provide first a naïve preference index between the two stimuli to demonstrate that both are equally perceived by the fly. On the other hand, the authors should equilibrate the conditioned stimuli in such a way that once agar is reinforced and once 25% sugar and display a learning index instead of the performance index. This of course should be done similarly for the NaCl conditioned stimulus.

It further appears that the (CS+) source is always situated on the left side of the experimental setup and the (CS-) always on the right side, if the indication of the interaction numbers are well understood. In order to avoid any place component that may interfere with the actual learning pathways the authors investigate the authors should also equilibrate the place of (CS+) presentation in relation to the (CS-).

According to many learning theories the learnability of a (CS+) is strongly increased when (CS+) and (US) are presented overlapping but with a time delay. Seen that the artificial activation of gustatory neurons together with the presentation of low-concentrated sugar may not only affect learning circuits but directly the perception of the (CS+) itself such a time delay would be even more important. Therefore, I suggest that the authors may introduce a time delay between the sipping and the opto-genetic activation of neurons in accordance with the published olfactory learning paradigms.

The authors need to explain the figures much more in detail. For example the given figure for figure 1C/D lets assume that the box blots in figure 1C reflect the data of the first ten minutes of the cumulative preference index. Indeed, this is apparently not the case, but the cumulative preference index is depicted over the entire 60 min but 10 min after the training. The figure description is misleading and would need some amendments. Further the authors should explain more in detail what the interaction values are and what they reflect. The information given by the authors is cryptic and does not allow a straight-forward understanding of the figure.

As stated above the cumulative preference index indicates a strong delay between the two groups in their memory retrieval (Figure 1D). The authors do not really discuss this effect in detail that per se is very interesting as it is very different in its dynamics compared with the other learning experiments provided by the authors.

For the statistical analysis the authors use an ANOVA with a Dunnet's post hoc correction throughout. In this regard it is unclear which of the data groups serve as a reference for the test. Normally a Dunnet's correction is used for multiple test groups that are compared to one single control group e.g., Placebo against different medical treatments. Here the Placebo group would serve as reference. In the case of the data provided by the authors, the situation is drastically different, as we have one test group and three control groups. As such the Dunnet's correction may not be the most adequate way for a multiple comparison of data and the authors may want to think about employing a more standard correction such as Bonferroni or Tukey.

The authors use two terms when referring to forms of memory exceeding short-term memory, long-term memory (LTM) and long lasting memory (LLM). However, they miss to explain when and why they employ the two different terms.

Further, the authors should help the reader and indicate more rigorously the compartments that are innervated by the individual lines. Descriptions like R48B04>CsChrimson or the "activation of R15A04-Gal4 neuron" are difficult to follow for readers that not directly related to the field.

Lastly, I would like to encourage the authors to employ their intuitive technique to expand the field of gustatory learning instead of asking questions that were already answered for olfactory conditioning now for gustatory conditioning. Of course, it is interesting to see the parallels between gustation and olfaction but the cellular mechanisms and energy availability would rather be a surprise if they would differ in their mode of action between the two forms of learning. However, the technique described by Jelen and colleagues would allow much more detailed circuit-oriented and temporal analysis of gustatory learning.

https://doi.org/10.7554/eLife.81535.sa1

Author response

Essential revisions:

All three reviewers expressed a high interest and, provided you can address the following issues, are supportive of publication of your manuscript in eLife. The reviewers were particularly impressed by the assay you have developed and its potential in resolving more dynamic aspects of perception and learning. In a revised version of the manuscript, please address especially the following points:

1. The reviewers recommend taking better advantage of your operant assay. Please show the data as original traces and not just box plots for the learning and recall phase. In addition, please explain why this is not 'simply' associative learning with one pairing as usually used in standard olfactory learning assays.

All graphs now include raw data points in addition to the summary statistics. In the text we now also note and comment on the individual variability seen in the data.

We have added a section to the introduction in which we note the operant nature of our paradigm and how it differs from classical conditioning (lines 125-130). We have also added a paragraph to the Discussion where we comment on our results in the context of the differences between operant and classical conditioning, as well as past studies exploring relationships between these two types of learning in flies (lines 340-352).

2. Explain the putative perceptual differences between optogenetic activation and real stimuli, for instance for sugar reward vs. exogenous activation of sweet taste neurons. Along the same lines, please show the data during learning as traces to help evaluate how actual stimulus perception during training impacts on learning.

We have added a paragraph to the Discussion in which we describe some limitations of our approach, with particular focus on the ways that optogenetic stimulation may not replicate all the features of natural stimuli (lines 409-427).

We have also added time curves showing the evolution of preference for the two stimuli during both training and testing of all the experiments in Figures 1 and 2. Since the rest of the paper explores manipulations that derive from the basics laid out in Figures 1 and 2, we decided that it is sufficient to show the temporal dynamics for only those earlier experiments.

3. Add reciprocal controls for at least one of the conditions.

We have now expanded Figure 2—figure supplement 1 and its associated text in order to present and describe several controls to demonstrate the specificity of learning the CS+ in our paradigm. In order to avoid redundancy, we will discuss these additions in more detail response to the more specific reviewer comments below.

4. We recommend making the manuscript more accessible to scientists not familiar with Drosophila memory paradigms by providing more explanation and interpretation of your data.

We appreciate this suggestion and have extensively rewritten the manuscript to add clarity and accessibility to both the background material and also our results. We have also reworked the figures and figure legends to increase clarity.

Reviewer #1 (Recommendations for the authors):

A few suggestions to improve the manuscript further:

Introduction:

(Line numbering would have been helpful…)

You write at the end of the first page, that DANs are activated by reward and punishment. They are also activated/modulated, for instance, by metabolic state and movement. Both of these aspects might be highly relevant in your operant assay.

We appreciate you mentioning this and agree that this is relevant for operant learning. However, in rewriting this section for clarity, we have now explicitly introduced the DANs in the context of olfactory classical conditioning where their activity is usually evoked by sugar or shock. However, we have added a passage near the end of the Discussion (lines 431-437) where we briefly discuss what could function to activate DANs during natural taste learning.

End of p 3: 'while maintaining similar satiation states': this statement contrasts somewhat with some other statements you make regarding interaction with the food source and even the shown bar graphs during training. This needs to be taken into consideration and explained a little better.

This is an excellent point. Our intention was to point out that changes in satiety could influence behavior in a way that appears to be learning, whereas in our experiments we point out that any differences in feeding during training are actually in the opposite direction to what could explain the change in behavior (e.g. flies eating more salt during training and then still showing increased salt feeding during testing). Nonetheless, during revision of the text for clarity we have removed the statement altogether.

Beginning of page 8 and Figures 5 and 6: please correct the references to and/or labeling of the panels. In addition, I did not find the data for the 7-hour delayed refeeding. Figure 6C does not exist.

The delayed refeeding is indicated for each nutrient in the figure (gold bars). We have corrected the other errors.

In the schemes, the light dot is always shown only on one side of the brain. Is this correct that you are activating only unilaterally? I guess not and I found it somewhat confusing…

We have tried, where possible, to expand the red light to suggest global illumination (which is the case).

There were several typos or small mistakes that must be corrected.

We have extensively rewritten and edited the manuscript, so we hope to have caught everything.

Reviewer #2 (Recommendations for the authors):

Figures:

The authors describe the learning as a preference index, which is calculated from the number of interactions (sips) with the CSs. Does also the length of sip bouts change? Maybe it would be worth looking also at overall feeding time. This could be especially interesting as flies that have been trained with GR66a activation, still prefer the sugar CS over several minutes at the beginning of the test. Maybe these are just very short bouts. Is there an explanation for this phenotype? Are the flies maybe more food deprived, because they receive less sugar in the training?

These are interesting questions but unfortunately the code we use to run the STROBE does not collect data on sip length, bout duration, or any of the other more detailed metrics that are possible with the original FlyPad code.

It is certainly likely that flies trained with Gr66a activation are more food deprived than controls. However, we also caution (as discussed above) against interpreting preference indices early in the assay. At these early time points the flies have made few sips and the overall preference measurement is likely to not be very reliable.

It is unclear what L/R means in Figure 1C/E/G, as this is also not explained in the legend. I assume it is the left/right side of CSs. Could this be renamed as CS+/Cs-, or red light/ no red light?

We have changed these to “agar” and “salt” or “suc” to indicate the food option in these figures.

The LLM/LTM can only be induced by feeding caloric food to the flies directly after training, this is a very important piece of information that does not show up in the figures. The authors should explain this fact a bit better in the text and maybe add this below the relevant graphs too (refeeding, yes/no), this would be especially helpful in Figure 6 where they do not receive food at all after training? For people not familiar with LTM this is hard to follow.

We now have indicated the refeeding status for each LTM figure in the schematic associated with the figure.

In Figure 3 some of the dots below the graphs do not seem to be in the right place. Please check 3A/B, they are different from 3C/D, but should be the same?

Thank you for this very important catch. We have corrected the error.

Discussion:

The paradigm that the authors present uses operant training, where the fly can decide how much they want to eat from each tastant presented. It would be great to have a paragraph in the discussion where the similarities and differences to previously studied classical conditioning are elucidated and results are compared. The authors could also cite work from the Brembs lab. In terms of reinforcement vs. US function of dopamine neurons, there was a recent biorxiv paper that could be cited if that is possible.: Pain is so close to pleasure: the same dopamine neurons can mediate approach and avoidance in Drosophila | bioRxiv Rohrsen et al., 2021.

Thank you for these very insightful suggestions. We have added a paragraph where we discuss the interpretation of our data in light of the possible contributions from operant and classical conditioning and what is known about these two forms of learning from important earlier work from Brembs. We have also added the suggested reference in the results when noting the differential effects of different PAM subpopulations.

Reviewer #3 (Recommendations for the authors):

The manuscript Optogenetic induction of appetitive and aversive taste memories in Drosophila by Jelen and colleagues is interesting, well written and the experiments elaborated, and the data provided by the authors justifies the title of the manuscript. The technique provided by the authors is very interesting and will attract a broad readership mainly of the Drosophila community as gustatory learning paradigms are yet tricky and not feasible for mass assays. The novel technique provided by Jelen and colleagues will pave the way to investigate gustatory learning now in more detail. In addition, the fact that the animals can freely move greatly improves the possibilities to investigate gustatory learning. Despite the high potential of the manuscript to reach out to a broad readership I would like to address some points of concern that the authors may want to address before publication.

In the first set of experiments the authors pair the artificial activation of bitter gustatory neurons with a low concentrated sugar solution as (CS+), whereas pure agar is presented as (CS-). To control for the learnability of a conditioned stimulus the authors have to ensure that the (CS-) and (CS+) are both equally learnable. Therefore, the authors should on the one hand provide first a naïve preference index between the two stimuli to demonstrate that both are equally perceived by the fly. On the other hand, the authors should equilibrate the conditioned stimuli in such a way that once agar is reinforced and once 25% sugar and display a learning index instead of the performance index. This of course should be done similarly for the NaCl conditioned stimulus.

We have now expanded Figure 2—figure supplement 1 to include a number of controls to address the issues raised. We chose to focus on using NaCl and MPG as the conditioned stimuli, with PAM activation serving to drive appetitive memory formation. We show first that the naïve preference for NaCl vs MPG is roughly equal (panel C). Next, we show that, like NaCl, flies are able to learn to prefer MPG over agar when MPG is trained as the CS+ (panel D). We also show that an appetitive memory for NaCl can be formed when NaCl is the CS+ and MPG is the CS-. Finally, we show that the memory formed when NaCl is the CS+ is not generalized to MPG, as training with NaCl as the CS+ does not result in an elevated preference for MPG vs agar (panel E). Although we acknowledge that these are not the exact experiments requested, we were very limited in the number of additional experiments we were able to perform, and feel that these demonstrate the same qualities that the reviewer is looking for in their comment.

It further appears that the (CS+) source is always situated on the left side of the experimental setup and the (CS-) always on the right side, if the indication of the interaction numbers are well understood. In order to avoid any place component that may interfere with the actual learning pathways the authors investigate the authors should also equilibrate the place of (CS+) presentation in relation to the (CS-).

In panels A and B of Figure 2—figure supplement 1, we now demonstrate that flies trained with NaCl as the CS+ do not exhibit a positional preference when agar is presented on both sides during testing. This is true whether the training is done with PAM activation (panel A) or sweet neuron activation (panel B). We performed both these experiments since sweet neuron activation drives a positive preference during training, while PAM activation does not. Along with the other experiments described above, these results demonstrate the specificity of training to the CS+ in the STROBE paradigm.

According to many learning theories the learnability of a (CS+) is strongly increased when (CS+) and (US) are presented overlapping but with a time delay. Seen that the artificial activation of gustatory neurons together with the presentation of low-concentrated sugar may not only affect learning circuits but directly the perception of the (CS+) itself such a time delay would be even more important. Therefore, I suggest that the authors may introduce a time delay between the sipping and the opto-genetic activation of neurons in accordance with the published olfactory learning paradigms.

This is a great idea and would be an excellent way to leverage the STROBE to manipulate different temporal dynamics during training. However, we do not currently have the expertise in the lab to modify the sip detection and light triggering algorithm to introduce a systematic delay. Nonetheless, we should point out that, while the STROBE LED activates with low latency, there is still a delay between sips and LED illumination. We previously measured this delay to be about 37 ms.

The authors need to explain the figures much more in detail. For example the given figure for figure 1C/D lets assume that the box blots in figure 1C reflect the data of the first ten minutes of the cumulative preference index. Indeed, this is apparently not the case, but the cumulative preference index is depicted over the entire 60 min but 10 min after the training. The figure description is misleading and would need some amendments. Further the authors should explain more in detail what the interaction values are and what they reflect. The information given by the authors is cryptic and does not allow a straight-forward understanding of the figure.

We have now modified all the figures and figure legends with the goal of increasing clarity. We paid particular attention to the schematics and the presentation of timing. We have removed the timeline schematic for each figure, as they were redundant and, upon reflection, confusing for several reasons. We now rely on Figure 1A to clearly demonstrate the paradigm and the timing for both STM and LTM assays. The schematics associated with individual panels now illustrate only the neuron population being optogenetically activated and the identity of the CS+ and CS-, as these are the parameters that change from experiment to experiment.

As stated above the cumulative preference index indicates a strong delay between the two groups in their memory retrieval (Figure 1D). The authors do not really discuss this effect in detail that per se is very interesting as it is very different in its dynamics compared with the other learning experiments provided by the authors.

As described in response to another reviewer’s comment above, we believe that the calculated preference indices are relatively unreliable in the early portion of testing, due to the small number of interactions during that period and the stochastic effects of flies encountering one choice vs the other. Thus, we caution against reading too much into changes in preference during these early times. Nonetheless, it is possible that the control group develops a stronger preference for sugar over the course of the assay as some hunger builds up after their feeding during training, and this combines with higher reliability as sips build up to drive a higher preference. We now mention our general interpretation of the testing dynamics in the text.

For the statistical analysis the authors use an ANOVA with a Dunnet's post hoc correction throughout. In this regard it is unclear which of the data groups serve as a reference for the test. Normally a Dunnet's correction is used for multiple test groups that are compared to one single control group e.g., Placebo against different medical treatments. Here the Placebo group would serve as reference. In the case of the data provided by the authors, the situation is drastically different, as we have one test group and three control groups. As such the Dunnet's correction may not be the most adequate way for a multiple comparison of data and the authors may want to think about employing a more standard correction such as Bonferroni or Tukey.

From a statistical standpoint, we struggle to see the distinction between comparing one control to several test groups and comparing one test group to several controls. The similarity between these situations is exactly why we chose the Dunnet’s test, and then report the lowest level of significance (highest p-value) across the comparisons with the different controls. Nonetheless, we tested several experiments with using a Bonferroni test instead of the Dunnet’s test and there was no difference in the categorical result (i.e. the number of stars). In the end we elected to continue with the Dunnet’s test.

The authors use two terms when referring to forms of memory exceeding short-term memory, long-term memory (LTM) and long lasting memory (LLM). However, they miss to explain when and why they employ the two different terms.

After further consideration, we have decided to simply use LTM throughout. Originally we used LLM in the portion of the manuscript preceding the demonstration that the memory is protein synthesis dependent. However, we agree this is unnecessarily confusing and have elected to simplify by only using LTM.

Further, the authors should help the reader and indicate more rigorously the compartments that are innervated by the individual lines. Descriptions like R48B04>CsChrimson or the "activation of R15A04-Gal4 neuron" are difficult to follow for readers that not directly related to the field.

We have now added this information to the text as we describe each experiment.

Lastly, I would like to encourage the authors to employ their intuitive technique to expand the field of gustatory learning instead of asking questions that were already answered for olfactory conditioning now for gustatory conditioning. Of course, it is interesting to see the parallels between gustation and olfaction but the cellular mechanisms and energy availability would rather be a surprise if they would differ in their mode of action between the two forms of learning. However, the technique described by Jelen and colleagues would allow much more detailed circuit-oriented and temporal analysis of gustatory learning.

We appreciate this suggestion and the sentiment behind it. As we performed all of the experiments presented in the manuscript, we were always pulled between the desire to demonstrate the robustness and utility of the assay versus the desire to tread new ground and explore major differences in learning mechanisms. For us, each result that confirmed a similarity with olfactory learning was additional evidence that the assay was really measuring what we thought it was measuring, and this contributed to the appeal of covering all the similarities with olfactory learning. As with many projects, circumstances (in particular the pandemic in this case) also contributed to how much we could achieve. Nonetheless, we look forward to probing more taste learning mechanisms in the future.

https://doi.org/10.7554/eLife.81535.sa2

Article and author information

Author details

  1. Meghan Jelen

    Department of Zoology and Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
  2. Pierre-Yves Musso

    Department of Zoology and Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  3. Pierre Junca

    Department of Zoology and Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Michael D Gordon

    Department of Zoology and Life Sciences Institute, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Supervision, Funding acquisition, Visualization, Writing – original draft, Writing – review and editing
    For correspondence
    michael.gordon@ubc.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5440-986X

Funding

Natural Sciences and Engineering Research Council of Canada (RGPIN-2016-03857)

  • Michael D Gordon

Natural Sciences and Engineering Research Council of Canada (RGPAS 492846-16)

  • Michael D Gordon

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Celia Lau for the original SEZ diagram models, and members of the Gordon lab for comments on the manuscript. This work was funded by Natural Sciences and Engineering Research Council (NSERC) grants RGPIN-2016-03857 and RGPAS 492846-16.

Senior Editor

  1. Claude Desplan, New York University, United States

Reviewing Editor

  1. Ilona C Grunwald Kadow, University of Bonn, Germany

Reviewer

  1. Thomas Dieter Riemensperger, University of Cologne, Germany

Version history

  1. Preprint posted: November 13, 2021 (view preprint)
  2. Received: July 4, 2022
  3. Accepted: September 22, 2023
  4. Accepted Manuscript published: September 26, 2023 (version 1)
  5. Version of Record published: October 9, 2023 (version 2)

Copyright

© 2023, Jelen et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 576
    Page views
  • 129
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Meghan Jelen
  2. Pierre-Yves Musso
  3. Pierre Junca
  4. Michael D Gordon
(2023)
Optogenetic induction of appetitive and aversive taste memories in Drosophila
eLife 12:e81535.
https://doi.org/10.7554/eLife.81535

Further reading

    1. Neuroscience
    Connon I Thomas, Melissa A Ryan ... Benjamin Scholl
    Research Article

    Postsynaptic mitochondria are critical for the development, plasticity, and maintenance of synaptic inputs. However, their relationship to synaptic structure and functional activity is unknown. We examined a correlative dataset from ferret visual cortex with in vivo two-photon calcium imaging of dendritic spines during visual stimulation and electron microscopy reconstructions of spine ultrastructure, investigating mitochondrial abundance near functionally and structurally characterized spines. Surprisingly, we found no correlation to structural measures of synaptic strength. Instead, we found that mitochondria are positioned near spines with orientation preferences that are dissimilar to the somatic preference. Additionally, we found that mitochondria are positioned near groups of spines with heterogeneous orientation preferences. For a subset of spines with a mitochondrion in the head or neck, synapses were larger and exhibited greater selectivity to visual stimuli than those without a mitochondrion. Our data suggest mitochondria are not necessarily positioned to support the energy needs of strong spines, but rather support the structurally and functionally diverse inputs innervating the basal dendrites of cortical neurons.

    1. Neuroscience
    Weiwei Qui, Chelsea R Hutch ... Darleen Sandoval
    Research Article

    Several discrete groups of feeding-regulated neurons in the nucleus of the solitary tract (nucleus tractus solitarius; NTS) suppress food intake, including avoidance-promoting neurons that express Cck (NTSCck cells) and distinct Lepr- and Calcr-expressing neurons (NTSLepr and NTSCalcr cells, respectively) that suppress food intake without promoting avoidance. To test potential synergies among these cell groups we manipulated multiple NTS cell populations simultaneously. We found that activating multiple sets of NTS neurons (e.g., NTSLepr plus NTSCalcr (NTSLC), or NTSLC plus NTSCck (NTSLCK)) suppressed feeding more robustly than activating single populations. While activating groups of cells that include NTSCck neurons promoted conditioned taste avoidance (CTA), NTSLC activation produced no CTA despite abrogating feeding. Thus, the ability to promote CTA formation represents a dominant effect but activating multiple non-aversive populations augments the suppression of food intake without provoking avoidance. Furthermore, silencing multiple NTS neuron groups augmented food intake and body weight to a greater extent than silencing single populations, consistent with the notion that each of these NTS neuron populations plays crucial and cumulative roles in the control of energy balance. We found that silencing NTSLCK neurons failed to blunt the weight-loss response to vertical sleeve gastrectomy (VSG) and that feeding activated many non-NTSLCK neurons, however, suggesting that as-yet undefined NTS cell types must make additional contributions to the restraint of feeding.