A distinct neural ensemble to integrate contradictory information and form long-term memory in Drosophila

Abstract

Interpreting contradictory information to favor survival is a crucial challenge for the brain. Although the parallel competing mode and memory convergence mode are proposed to resolve how contradictory information is processed in the brain, no detailed regulatory mechanism is proposed, and the formation of a memory trace to support either one is also lacking. The current study demonstrated that competition and integration existed after contradictory training, i.e., aversive/appetitive, followed by appetitive/aversive conditioning. Although information from aversive and appetitive training competes within 3 hours after conditioning, approach behavior gradually prevails, and avoidance behavior decays. The training type–dependent neural network consolidating and storing contradictory memory differs from the circuit for sole aversive and reward memory. A contradictory memory trace formation in the brain, indicating aversive and reward memories, is integrated into one approach memory. Our study reconciles the current two modes and proposes a competition-integration mode for contradictory memory expression.

A competition-integration mode for contradictory memory expression.

INTRODUCTION

In natural habitats, animals frequently encounter the challenge of seeking food or water while avoiding potential threats. An animal’s ability to behave according to the constant change in its environment to favor survival suggests that behavior is adaptive and can be adjusted through learned experiences. The brain generates behavior based on recent experience or adopts the most seemingly beneficial decision from accumulating past experiences. However, evaluating encountered experience is not always straightforward. Sometimes, animals encounter contradictory cues or situations where individual stimulus seems equally beneficial or detrimental. Memory engram undergoes constant neuronal and circuitry changes until specific information is stored for a particular behavior (1–3). How the brain processes contradictory experiences to form circuitry to perform proper behavior is largely unexplored.

In Drosophila melanogaster, extensive studies have attempted to depict which Kenyon cells (KCs) within the mushroom body are involved in specific types of memories. The ability to acquire new information depends on the KCγ, and memory consolidation relies on the KCα’β’. Memory retrieval depends on the KCαβ (4–8). KCs form synapses with mushroom body (MB) output neurons (MBONs), conveying signals to other brain regions (9, 10). The most prevalent modulatory neurons innervating the KCs are dopaminergic neurons (DANs), which are divided into two clusters located in the protocerebral posterior lateral 1, which relay negative reinforcement (11, 12), and the protocerebral anterior medial, which relay positive reinforcement (13, 14). Information flow of an odor from the distinct KCs to the MBONs is assigned to either a positive or negative valence, directing behavioral response opposite to the valence signaled by the corresponding DANs (15–18).

Previous works have suggested that the brain forms parallel memories of two opposite kinds of memory, aversive and appetitive, competing to guide learned behavior of either avoidance or approach, respectively (19), based on the decay rates of each memory (20, 21). Fruit flies exposed to insect repellent (DEET, an aversive stimulus) mixed with sucrose (an appetitive stimulus) during olfactory conditioning exhibited the simultaneous formation of both memories, highlighting their independent nature (22). A recent study from Caenorhabditis elegans also demonstrated that reward and punishment memory compete (23). On the other hand, other studies suggest that aversive and appetitive information is integrated into the brain. Single neurons in the lateral habenula and ventral tegmental area respond to both aversive and reward stimuli in mice and primates, revealing the intricate interplay between aversion and reward processing in the brain and suggesting convergence of contradictory information (24, 25). A similar finding is also observed in orbitofrontal cortex in the monkey (26). In addition, a recent study on the extinction of fruit flies showed that reexposed to CS+ odor, odor paired with electric shock, after training recruits a neuronal circuit to write a competing memory and causes the brain to reconsider the meaning of the first training (27). Although what causes these inconsistent findings remains unclear except for the differences in animal models, most studies use different training protocols and test the aversive and reward circuits separately. The memory traces to represent the integrated contradictory memory have not been demonstrated, and how the brain integrates the inconsistent information is largely unsettled. In addition, the debate between memory competition and convergence is continuous. Therefore, how contradictory information is processed within a neuronal circuit after animals face contradictory environments remains elusive.

In the present study, we developed a contradictory training paradigm in fruit flies by enforcing two types of reinforcement training, aversive conditioning and appetitive conditioning, to assess the contradictory memory formation, based on the counterconditioning protocol established in mice (28). Our results suggested that contradictory information causes interference with initial learning, which is susceptible to disruption for an excessive period, and the underlying neural circuit used to process contradictory information is training type dependent. Our findings are consistent with a traditional model in which KCγ is required for acquisition, but different sets of circuits are recruited for memory expression depending on the conditioning paradigms. Our results showed that while KCαβ is needed to retrieve conventional appetitive memory, all three KCs are required to store the contradictory memory. Furthermore, compared to its phenotypic-like counterpart, conventional reward memory, contradictory memory shows an unequivocal delay in forming protein synthesis–dependent long-term memory (PSD-LTM). Our genetic behavioral studies and live-cell imaging found no traditional aversive and reward 24-hour memory trace; a different contradictory LTM trace was observed after contradictory training, suggesting that two memories are integrated into one memory. The discrepancies in the behavioral outcomes between conventional and contradictory training paradigms indicate that contradictory memory seems to be distinct from any traditional known memory. Our study reconciles the current two modes in describing how contradictory information is processed in the brain; competition and integration strategies work together to form a different contradictory LTM.

RESULTS

Contradictory training interferes with initial learning

Traditionally, flies are trained in either an aversive (punishment) paradigm, resulting in an avoidance behavior, or an appetitive (reward) paradigm, resulting in an approach behavior (29, 30). Such designs present animals with only one type of information to be learned. Therefore, we combined these two traditional paradigms, similar to counterconditioning (28), to generate contradictory information, presenting the flies with contradictory stimuli. In one training cycle, the same odor paired with a negative reinforcement, electric shock, is now also paired with a positive reinforcement of sugar, training consecutively in either order of appetitive first or aversive first. Two different training procedures were performed and compared, aversive conditioning followed by appetitive conditioning (P + R) and appetitive conditioning followed by aversive conditioning (R + P) (Fig. 1A). We intend to investigate (i) the contradictory memory that could be established with different protocols, and (ii) the underlying neural ensemble to process contradictory information is training protocol dependent.

Fig. 1. Contradictory training interferes with initial learning.

(A) Contradictory training paradigm: Wild-type flies were trained consecutively under two different olfactory classical conditioning paradigms, aversive (punishment) and appetitive (reward) associative conditioning. The same odor paired with a negative reinforcement (electric shock) was also paired with a positive reinforcement (sugar). (B) Reward + punish (R + P)–trained flies exhibited avoidance behavior after training (n = 6 for each group, P < 0.0001 and P = 0.04 respectively). (C) Punish + reward (P + R)–trained flies produced approach behavior after training (n = 7, 7, and 9) (P = 0.003 and P < 0.0001). (D) The P + R effect was reduced when there were 0.5- to 1-hour resting intervals between punish and appetitive trainings (n = 6 for each group, P = 0.005, 0.035, 0.265, and 0.198). (E) The effect of R + P was reduced when flies were given a 1-hour break between punishment and appetitive training (n = 9, 6, and 7, P = 0.0043 and 0.528). (F and G) Approach behavior was observed in 3-hour memory in (F) P + R (n = 6 for each group, P = 0.087 and 0.046) and (G) R + P condition (n = 6, P = 0.203 and P < 0.0001). Statistical analysis was performed using one-way analysis of variance (ANOVA) with Sidak’s post hoc test. Asterisks indicate statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. ns, not significant. In all figures, each value represents mean ± SEM. h, hours.

Our results showed that flies tended to remember the latest information they encountered. R + P–trained flies showed avoidance behavior (Fig. 1B), whereas P + R–trained flies exhibited reward approach behavior (Fig. 1C). Although the performance indexes (PIs) in both contradictory paradigms were lower than those observed when a single reinforcement (electric shock or sugar) was used alone, this result suggests that contradictory information generates memory interference, which may alter what is initially learned.

We were curious whether such interference would be less prominent or even abolished if flies were not immediately, sequentially trained but instead given a resting interval between aversive and appetitive conditioning. We chose time intervals of 30 min and 1 hour to experiment. After training wild-type flies with single-cycle aversive/appetitive conditioning, flies were placed back into vials without food and allowed to rest in a light-tight box to avoid external interference. They were then trained with single-cycle appetitive/aversive conditioning and tested for memory immediately. Both in 30 min and 1 hour, flies showed approach behavior in response to their most recent appetitive training experience in the P + R condition, although a subtle difference may indicate a lingering effect of the prior punishment conditioning (Fig. 1D). The same result was also found in the R + P condition (Fig. 1E). Avoidance behavior was similar between R + P flies and flies only trained with punishment alone when given a resting interval. Therefore, information may be readily disrupted within a 30-min window, information after 30 min is no more contradictory, and prediction tends to be based solely on their most recent experience.

Reward memory predominates over time

Since contradictory information generates interference during acquisition, we investigated what memory is processed over time. After single-cycle contradictory conditioning, flies were assayed for 3-hour memory. Our results showed a trend of approach behavior in both contradictory paradigms, although the approach behavior was not exactly the same as the reward training. Trained flies show a tendency to engage in approach behavior in P + R (Fig. 1F) and R + P (Fig. 1G) after 3 hours. These findings suggest that, irrespective of training sequence, the approach behavior gradually emerges, indicating the dominance of appetitive memory. To confirm that the observed contradictory memory generation is due to the contradictory training, same odor pair for aversive and reward conditioning, we trained the flies with 3-octanol (OCT)/4-methylcyclohexanol (MCH) for reward and isoamyl acetate (IA)/ethyl acetate (EA) for aversive training, different odor pairs for aversive and reward conditioning. After 3 hours, not only was an approach behavior found for OCT/MCH testing, but avoidance behavior for IA/EA testing was also shown (fig. S1A), suggesting that (i) the finding we observed is due to information contradictory and (ii) the existence of parallel pathways for reward and punishment, when there is no contradiction. Our results also suggest that reward signals over time still prevail despite interference caused by contradictory information during acquisition, as shown through approach behavior. We also manipulated the electric shock using 30 and 90 V to examine the effect of different strengths of aversive memory on the result of contradictory 3-hour memory in the P + R protocol. Aversive training with 30-V electric shocks generated little to no contradictory memory effect compared to 90-V electric shocks (fig. S1B).

Consolidation of contradictory memory is slow

It has been demonstrated that one cycle of appetitive conditioning is sufficient to form stable memory quickly (29). Since approach behavior was also seen after P + R conditions, we are curious to know how the memory is consolidated in P + R paradigm. We applied cold shock to assess the memory status, as cold shock–induced anesthesia has been demonstrated in flies to damage unconsolidated memory and leave consolidated memory. Although there is a trend of decrease in reward memory after cold shock, the approach behavior was abolished after cold shock in P + R–trained flies (Fig. 2A). Although flies show approach behavior, the pathway to process reward memory in contradictory-training flies is not the same as in appetitive-trained flies.

Fig. 2. Reward memory predominates over time.

(A) P + R contradictory trained flies followed by cold shock did not show stable memory (n = 6, 8, 7, 7, and 9; P = 0.999, 0.197, 0.140, 0.002, and 0.003). (B to E) Cold shock treatment was applied at different time points to examine the anesthesia resistant memory (n = 7 and 9, 6 and 7, 6 and 6, and 8 and 8; P = 0.010, P = 0.014, P = 0.049, and P = 0.204). (F) A progressive increase in consolidated memory formation was observed from 3 to 24 hours. (G) The 24-hour c-LTM was observed in R + P training (n = 8, 7, and 6; P = 0.041 and P < 0.0001). (H) CHX-fed flies exhibited disrupted 24-hour memory (n = 7, 8, and 9; P = 0.030). (I) The 24-hour P + R and R + P memories were compromised in CHX-fed flies compared to the untreated flies (n = 8 and 8; 7 and 9; P = 0.038 and P = 0.014). All figures display values as mean ± SEM. Statistical comparison was carried out by two-tailed unpaired t test for [(B) to (E), (H), and (I)], *P < 0.05. However, for other groups, one-way ANOVA with Sidak’s post hoc test was conducted to analyze statistical significance. Asterisks indicate statistical significance. *P < 0.05, **P < 0.01, and ****P < 0.0001.

To further delineate the consequent disruptions of memory formation from contradictory training, we were determined to observe when contradictory memory is stabilized in the P + R condition. We tested memory performance at extended hours: 7, 11, 15, and 24 hours, all performing cold shock 1 hour before testing. Unexpectedly, flies’ memory for up to 15 hours showed impairment after being cold-shocked (Fig. 2, B to E). The flies’ memory after contradictory training showed prolonged susceptibility to disruption, which is different from conventional reward memory, where previous studies showed reward memory to be resistant to disruption by cold shock anesthesia (31). We see a gradual increase in the ratio of stable memory being formed, starting from 3 to 24 hours (Fig. 2F). The delayed stabilization of P + R contradictory memory suggests that its consolidation process is distinct from that of reward memory, possibly due to the need to integrate and resolve contradictory aversive and reward signals.

Moreover, to eliminate possible contributing factors to the phenomenon seen in our results, we took P + R as an example and imitated aversive conditioning before training the flies in an appetitive manner by not pairing the CS+ with the electric shock. Food-deprived wild-type flies were given 2 min of OCT or MCH odor before a single-cycle appetitive conditioning. They were then tested for memory performance by conducting the cold shock protocol. Flies preexposed to the CS+ odors did not show a difference in memory performance compared to the control flies (fig. S2A). In another set of experiments, food-deprived wild-type flies experienced 1 min of 90-V electric shock before a single cycle of appetitive training. Results showed no difference in memory performance of preshocked flies with flies of single-cycle appetitive conditioning (fig. S2B). These results support the previous finding that consolidated reward memory is quickly formed (31) and eliminate the possibility that the nature of aversive conditioning (odor or electric shock) experienced before appetitive conditioning may play a role in disrupting 3-hour memory formation. Therefore, complete training cycles are required to constitute the effect of disruption of stable reward memory formation.

The formation of 24-hour memory was not limited to P + R training; the approach behaviors of 24-hour memory could also be observed after R + P training (Fig. 2G). The avoidance behavior gradually changed to approach behavior (fig. S3A). Further analysis showed that animals’ self-status regulates the rate of reversed behavior from avoidance to approach. We fed the flies after contradictory training and found that the avoidance behavior lasted for more than 12 hours (fig. S3B). However, reward memory was still observed after 24 hours of feeding (fig. S3C), suggesting that feeding after training only delays but does not block the approach behavior.

To distinguish whether the memory in our results is protein synthesis–dependent contradictory information formed long-term memory (c-LTM), we fed the flies a protein synthesis inhibitor, cycloheximide (CHX) and the cold shock protocol in the P + R condition. The resulting memory declined, indicating the presence of protein synthesis–dependent c-LTM (Fig. 2H). In contrast to a single appetitive training that forms protein-dependent memory within a few hours (31), we further demonstrated that 8 hours after contradictory training, there is no protein-dependent memory formation (fig. S3D), and this result is consistent with our previous finding in Fig. 2B. Consistently, feeding of CHX damaged the 24-hour memory in the R + P condition, suggesting that a PSD-LTM is also formed in the R + P condition (Fig. 2I).

Neurons to retrieve 24-hour contradictory memory are training dependent

A contradictory training–induced behavioral change suggests that the neural circuit that regulates memory formation has also changed. To identify which KC is responsible for 24-hour contradictory memory retrieval, three different mushroom body lobe GAL4 lines, VT30604-GAL4 for KCα’β’, VT49246-GAL4 for KCαβ, and R16A06-GAL4 for KCγ, were used to overexpress shi^ts (32–34). Shi^ts is a reversible temperature-sensitive shibire mutant, which damages vesicle endocytosis at a restrictive temperature. The 24-hour contradictory memory of P + R retrieval was altered when the output activity was inhibited in the KCαβ and KCα’β’ (Fig. 3, A to C). Further study suggested that inhibiting the output activity of MBON-β’2 (MB011B) and β’1 (MB057B) blocked the 24-hour P + R memory retrieval (Fig. 3D and fig. S4A). In addition, no substantial change in behavior was observed when the experiment was performed in the 23°C (fig. S4B), indicating that the observed effect was due to the inhibition of neural activity at a restrictive temperature. Furthermore, the 24-hour reward memory was damaged when the output activity was inhibited in the KCαβ (Fig. 3E). The 24-hour contradictory memory of R + P retrieval (approach) was compromised when the output activity was inhibited in the KCα’β’ and KCγ (Fig. 3, F and G). Inhibiting the output activity of MBONs demonstrated that it is MBON-β’1 and γ1pedc > αβ (MB112C) mediating R + P 24-hour memory expression (Fig. 3H). No behavior change was found when the experiment was performed at 23°C (fig. S4C). On the other hand, increased activity of MBON-β’1 by expressing TrpA1 promotes a stronger preference for CS+ odor during retrieval at 31°C (fig. S4D). Moreover, the inhibition of MBON-β’1 activity did not appear to alter 24-hour appetitive memory (fig. S4E).

Fig. 3. KCα’β’ and KCαβ are necessary for contradictory memory retrieval.

(A to H) Transgenic flies expressing shi^ts were trained in contradictory paradigms and rested for 24 hours at a permissive temperature of 23°C, followed by testing at a restrictive temperature of 31°C. [(A) and (B)] The 24-hour P + R memory was compromised in VT49246-GAL4 > UAS-Shi^ts (KCαβ) and VT30604-GAL4 > UAS-Shi^ts (KCα’β’) flies (n = 6, 6, and 6; P = 0.0001 and P = 0.001; n = 6, 9, and 8; P = 0.005 and P = 0.004). (C) No 24-hour P + R memory deficit was observed in R16A06-GAL4 > UAS-Shi^ts (KCγ) flies (n = 5, 8, and 9; P = 0.829 and P = 0.989). (D) Recall of 24-hour P + R memory was disrupted in MB011B-GAL4 > UAS-Shi^ts (MBON-β’2) and MB057B-GAL4 > UAS-Shi^ts (MBON-β’1) flies (n = 8, 7, and 8; P = 0.036 and P = 0.029). (E) Disruption of KCαβ output activity affected 24-hour reward memory, whereas 24-hour R + P memory was intact (n = 6, 6, and 6; P = 0.01 and P = 0.001; n = 6, 6, and 6; P = 0.997 and 0.749). [(F) and (G)] Inhibition of KCγ and KCα’β’ activity led to impairment of R + P 24-hour memory (n = 6, 6, and 6; P = 0.951 and 0.996; n = 6, 6, and 6; P = 0.999 and 0.971; n = 6, 6, and 6; P = 0.035 and 0.002). [(F) and (G)] Inhibition of KCγ and KCα’β’ activity led to impairment of R + P 24-hour memory (n = 6, 6, and 6; 6, 6, and 6; 6, 6, and 6; P = 0.004 and P = 0.0002; n = 7, 7, and 7; P = 0.004 and P = 0.027). (H) The 24-hour R + P memory was disrupted in MB057B-GAL4 > UAS-Shi^ts and MB112C-G GAL4 > UAS-Shi^ts (MBON- γ1pedc > αβ) flies (n = 9 and 10, P = 0.027; n = 6 and 6, P = 0.004). (I) The 1-hour R + P memory was disrupted in C739-GAL4 > UAS-Shi^ts (KCαβ) flies (n = 7, 6, and 7; P = 0.0002). VT57244-GAL4 labeled KCα’β’. (J) VT30604-GAL4 > UAS-Shi^ts flies trained with R + P contradictory condition showed 24-hour memory damage (n = 6 and 6, P = 0.013). All figures display values as mean ± SEM. Statistical comparisons were carried out by one-way ANOVA with Sidak’s post hoc test. Asterisks indicate statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

To further confirm the effect of contradictory training, we compared the memory performance when animals received consistent and contradictory training. Consistent training means the odor (CS+) in the reward was used as (CS−) in later punishment training. A clear memory performance difference was observed between these two when the output activity was inhibited in KCα’β’ during retrieval, suggesting that contradictory training changes memory storage (Fig. 3J). As 1 hour after R + P training showed avoidance behavior, we further analyzed the 1-hour contradictory memory of R + P, and retrieval (avoidance) was impaired when the output activity was inhibited only in KCαβ (C739) (Fig. 3I). There was no obvious change observed in C739-GAL4 > UAS-shi^ts flies when the experiment was performed in 23°C (fig. S4F). These results suggest that when two similar pieces of training with contradictory information are presented sequentially to the animals, the brain evaluates and processes the information accordingly. The contradictory training changes the course of memory formation differently from the individual training used to process the memory. Different training protocols use different neurons to retrieve 24-hour contradictory memory.

Protein synthesis in KCs ensures c-LTM consolidation

Since our behavioral data demonstrated that both P + R and R + P c-LTM require protein synthesis (Fig. 2I), both paradigms recruit different neuronal circuits for memory expression. We are then curious how protein synthesis forms c-LTM in different conditions by using RICIN^CS to assay the specific neurons’ protein synthesis (see Materials and Methods for details) (35, 36) on the three MB lobe GAL4 lines. After training, flies were stored at a restrictive temperature of 31°C, activating RICIN^CS and blocking protein synthesis during memory consolidation. Our data demonstrated that inhibition of translation in the KCαβ and KCα’β’ failed to alter c-LTM memory in P + R (Fig. 4, A and B). However, RICIN^CS expressed in KCγ appeared to impair 24-hour memory of the contradictory trained flies in P + R (Fig. 4C). We sought to examine whether artificial activation of KCγ would allow us to facilitate the process of long-term contradictory memory. We activated KCγ with TrpA1, a heat-activated cation channel (37, 38). We chose to test the flies’ memory at 8 hours after conditioning as our previous results suggested that contradictory trained flies had not shown to form stable memory after 7 hours (Fig. 2). After training, flies were immediately stored at 31°C to activate TrpA1 during memory consolidation. Results showed that flies fed with CHX showed altered memory at 31°C but not at 23°C, indicating the presence of protein synthesis–dependent memory (fig. S5A). Facilitated protein synthesis–dependent memory formation was not observed in the KCα’β’ with the same experimental condition (fig. S5B). The same experiment procedure was applied to the R + P paradigm to identify neurons for new protein synthesis to support the c-LTM expression. The results showed that new protein synthesis was required in all KCs, KCα’β’, KCαβ, and KCγ (Fig. 4D). These data suggest that contradictory training can facilitate new PSD-LTM and that new protein synthesis is not restricted to specific cells; it is dependent on the training protocol.

Fig. 4. Protein synthesis is crucial for c-LTM expression.

(A to C) R16A06-GAL4 > UAS-RICIN^cs flies but not VT49246 > UAS-RICIN^cs and VT30604 > UAS- RICIN^cs flies exhibited loss of 24-hour P + R memory (n = 8, 7, and 6; P = 0.993 and 0.769; n = 7, 7, and 8; P = 0.965 and 0.173; n = 9, 11, and 11; P = 0.005 and 0.009). (D) Expression of RICIN^cs in KCαβ, KCα’β’, and KCγ showed disrupted memory in R + P (n = 6, 7, 6, and 6; P = 0.004, P = 0.035, and P = 0.023). (E to H) VT30604-GAL4 > UAS-Shi^ts, VT49246-Gal4 > UAS-Shi^ts, and R16A06-Gal4 > UAS-Shi^ts flies were subjected to P + R, R + P, or appetitive training at 31°C. R16A06-GAL4 > UAS-Shi^ts flies exhibited a marked impairment across all training types (n = 5, 6, 6, and 6; P = 0.994, 0.999, 0.592, and 0.999; n = 6, 6, 6, and 6; P = 0.994, 0.948, 0.999, and 0.938; n = 5, 4, and 4; P = 0.012 and 0.022; n = 6, 7, and 6; P = 0.0001 and 0.0001; n = 6, 7, and 6; P = 0.007 and 0.0001). Statistical analysis was performed using one-way ANOVA with Sidak’s post hoc test. Asterisks indicate statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. In all figures, each value represents mean ± SEM.

Next, we sought to find which KC subset was required for the acquisition stage of contradictory memory. Transgenic flies with expressing shi^ts were trained at a restrictive temperature of 31°C to inhibit the neural output activity transiently. They were then moved back to 23°C for 24 hours until testing. Inhibited KCαβ and KCα’β’ output activity during training suggested no observable change in memory performance (Fig. 4, E and F). Emerging evidence has shown that the KCγ is required to acquire aversive and reward information (4, 8, 22). Here, we found that inhibiting the output of KCγ in flies during the training phase caused memory impairment in P + R and R + P conditions (Fig. 4, G and H). Although neurons to retrieve the contradictory memory are training dependent, the KCγ is the gateway for information uptake and the first checkpoint for processing all acquired information, including contradictory information.

Circuit to consolidate contradictory PSD-LTM

Until now, we have shown that different neural circuits are recruited to mediate different LTM formations. As memory consolidation is crucial to stabilize the memory and keep the memory in the brain, we intended to know whether the circuit for the consolidation process is the same or different between different conditions. We used shi^ts to inhibit the neuronal output activity during the consolidation phase. Transgenic flies expressing shi^ts were trained at 23°C and then immediately shifted to 31°C to render shi^ts active throughout the resting period of the flies until at 24 hours when they were returned to 23°C to be tested. We found that expressed shi^ts in KCγ displayed impaired c-LTM in P + R (Fig. 5, A to C), suggesting that contradictory memory consolidation, P + R, occurs within the KCγ. These data are consistent with fig. S5A. As for the R + P paradigm, we found that inhibition of KCαβ and KCγ during the consolidation phase appears to impair the 24-hour memory expression (Fig. 5, D to F).

Fig. 5. KCγ is critical to consolidating contradictory memory.

(A to C) Memory deficit was observed in P + R–conditioned R16A06-Gal4 > UAS-Shi^ts flies but not VT30604-GAL4 > UAS-Shi^ts and VT49246-GAL4 > UAS-Shi^ts flies when the output activity was inhibited during consolidation (n = 6, 7, and 7; P = 0.239 and 0.704; n = 6, 6, and 6; P = 0.866 and 0.602; n = 9, 8, and 8; P = 0.009 and 0.017). (D to F) Memory deficit was observed in appetitive conditioning in R16A06-GAL4 > UAS-Shi^ts and VT30604-GAL4 > UAS-Shi^ts flies but not VT49246-GAL4 > UAS-Shi^ts flies when the output activity was inhibited during consolidation. However, memory deficit was observed in R + P conditioning in R16A06-GAL4 > UAS-Shi^ts and VT49246-GAL4 > UAS-Shi^ts flies but not VT30604-GAL4 > UAS-Shi^ts flies (n = 8, 6, and 6; P = 0.417 and 0.332; n = 6, 6, and 6; P = 0.0003 and 0.0001; n = 6, 6, and 6; P = 0.003 and 0.0001; n = 6, 6, and 6; P = 0.0001 and 0.0001; n = 7, 6, and 6; P = 0.0001 and 0.0001; n = 6, 6, and 6; P = 0.999 and 0.514). Statistical significance was assessed using one-way ANOVA with Sidak’s post hoc test. Asterisks denote the significance level; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Values in all figures are shown as mean ± SEM.

DANs, PPL1 cluster, regulate behavioral transition

We hypothesized that aversive and appetitive information is integrated after contradictory training, engaging distinct MB compartments for consolidation and retrieval. Consistent with the notion, our results indicate that for retrieval, P + R relies on α’/β’ and α/β lobes, while R + P involves α’/β’ and γ lobes. For consolidation, R + P engages α/β and γ lobes, whereas P + R primarily depends on γ lobes. Furthermore, our results also showed that the approach behavior was gradually formed in the first few hours after training. As accumulated studies have demonstrated that dopamine signal, especially for PPL1-γ1pedc neuron, is important in modifying behaviors (39–41), we speculated that PPL1-γ1pedc neuron activity regulates c-LTM formation. Since R + P shows the dynamic change of behavior, from avoidance to approach, the R + P protocol was first examined. Although either inhibition or activation of MB320C-labeled neuron, PPL1-γ1pedc, immediately after training did not affect 3-hour memory (fig. S6A), 20-min interval between training and inhibition or activation of PPL1-γ1pedc neuron promoted approach behavior in 3-hour memory (fig. S6B). The following studies suggested that inhibition and activation of PPL1-γ1pedc neuron induced approach behavior using different mechanisms; inhibition of PPL1-γ1pedc neuron reduced aversive behavior, which allowed approach memory to form, while PPL1-γ1pedc neuron activation facilitated approach behavior. (i) Inhibited PPL1-γ1pedc neuron immediately after training reduced 1-hour memory avoidance behavior, while activated PPL1-γ1pedc neuron did not affect 1-hour memory avoidance behavior (fig. S6C); we also confirmed that there was no behavioral change when the same experiment was conducted in the 23°C (fig. S6D). (ii) Activation of PPL1-γ1pedc neuron produced stronger approach behavior within 2 hours (fig. S6E). (iii) Inhibition of PPL1-γ1pedc neuron for 3 hours produced more approach behavior of 24-hour memory, while activation of PPL1-γ1pedc neuron did not (Fig. 6A). We also examined the effect of PPL1-γ1pedc neuron in the P + R condition. Output inhibition of PPL1-γ1pedc neuron for 3 hours in P + R conditions after training also promoted approach behavior in 24-hour memory, while PPL1-γ1pedc neuron activation did not affect performance (Fig. 6B). The promoted approach behavior was not observed when the experiment was done at 23°C (fig. S6F). Activation or inhibition of PPL1-γ1pedc neuron for 3 hours, at 3 hours after training, did not affect 6-hour memory approach behavior, suggesting that the first 3 hours after training is critical for merging the two information (Fig. 6C). This is consistent with previous data showing that reward memory is dominant after 3 hours. According to these results, we suspected activity shifting during the first 3 hours; the activity of PPL1-γ1pedc neuron was gradually increased. We did not know why inhibition or activation immediately after training produced no behavior in 3-hour memory. Further, we could not exclude the nonspecific effect of overstimulation of dopamine neurons, i.e., more dopamine receptors are overactivated; we suspected that dopamine signals must be balanced right after training to form memory. We also confirmed that 20 min of PPL1-γ1pedc neuron activation 20 min after training did not affect 1-hour aversive memory (fig. S6G). Inhibition of PPL1-γ1pedc neuron for 20 min immediately after training did not affect 1-hour reward memory (fig. S6H).

Fig. 6. Regulation of behavioral transition by DANs.

(A) MB320C-GAL4 > UAS-Shi^ts and MB320C-GAL4 > UAS-TrpA1 flies were trained in R + P training and maintained at 23°C for 20 min and then moved to 31°C for 3 hours. Enhanced 24-hour memory was observed in MB320C-GAL4 > UAS-Shi^ts (n = 9, 7, 6, and 7; P = 0.034, 0.020, 0.885, and 0.999). (B) Flies with P + R training were initially maintained at 23°C for 20 min and then moved to 31°C for 3 hours. Enhanced 24-hour memory was observed in MB320C-GAL4 > UAS-Shi^ts (n = 6, 6, 6, and 6; P = 0.029 and 0.999). (C) Transgenic flies were moved to 31°C for 3 hours after 3 hours of training. No significant difference was observed in either training condition (n = 8, 6, and 9; P = 0.999 and P = 0.999; n = 6, 6, and 9; P = 0.999 and P = 0.999). (D) Transgenic flies with P + R training fed with either CHX or water were moved to 31°C for 2 hours after training and tested for memory. CHX-fed MB504B-GAL4 > UAS-Shi^ts flies exhibited significant disruption (n = 6, 5, and 6; P = 0.590 and 0.207; n = 6, 5, and 6; P = 0.032 and 0.038). Statistical analysis was conducted using a one-way ANOVA followed by Sidak’s post hoc test. Statistical significance is indicated by asterisks: *P < 0.05. All figures show values as mean ± SEM.

An inhibited dopamine signal promotes contradictory memory formation, which was also observed in P + R condition. Flies were divided into two groups, both contradictorily trained, P + R, fed with CHX or not. To see the early effect of PPL1 neurons activation, similar to the R + P condition, PPL1 neurons were only blocked for 2 hours after training. After 2 hours, the flies were moved to 23°C for another 6 hours before testing. Unexpectedly, when the output of PPL1 neurons labeled by MB504B-GAL4 was blocked, the formation of protein synthesis–dependent memory seemed to be facilitated (Fig. 6D). Although there was a subtle decrease in a non-CHX treatment, the memory performance was abolished after CHX treatment. The original cold shock–sensitive memory shown in Fig. 2 now presents protein synthesis–dependent memory. We additionally tested the PPL1-α’3 neuron (MB304B), which is not included in the MB504B cluster (40), and our observation indicated that c-LTM remained largely unaffected (fig. S6I).

Contradictory training recruits a different circuit to store long-term contradictory information

Until now, our behavioral studies suggest that contradictory training uses different neurons to process contradictory information. However, the existence of individual punishment and reward memory trace remains unclear. An accumulation of studies has demonstrated that 24-hour memory of appetitive and aversive is stored in the KCαβ (42, 43). To reveal the formation of memory traces, we used live-cell imaging. As the R + P paradigm showed the change of behavior from avoidance to approach, we used this paradigm to examine the existence of individual punishment and reward memory. Our data showed a difference between CS+ and CS− fluorescent response 24 hours after individually aversive and appetitive trained flies in the α3 region of KCαβ but not after contradictory training (Fig. 7, A to C). Similar results were found in the α2 region of KCαβ (fig. S7, A to C). These data suggest that the 24-hour memory trace of punishment and appetitive alone is not formed in the α2 and α3 regions after contradictory training. Our behavioral study showed that long-term contradictory memory is stored in the β’1 region of KCα’β’ and γ1pdec region of KCγ. Consistently, our imaging results revealed a noticeable difference in fluorescence between CS+ and CS− in the β’1 region of KCα’β’ (Fig. 7, D to F) and γ1pdec region of KCγ for R + P condition 24 hours after training (Fig. 7, G to I), suggesting that R + P memory traces were formed. No memory traces were observed in aversive and appetitive training in the γ1pdec region of KCγ 24 hours after training. Although CS+ and CS− fluorescent differences were observed in the β’1 region of KCα’β’ in aversive conditioning, it was not observed in appetitive conditioning (Fig. 7, D to F). These results confirm our hypothesis that a different contradictory memory circuit is formed. Our behavioral study showed a transition from avoidance to approach within 3 hours after R + P training and was regulated by PPL1-γ1pedc neuron. We were curious how memory trace is formed in KCγ after training. Although a memory trace was observed in the aversive condition but not appetitive condition 2 to 3 hours after training in the γ1pdec region of KCγ, the trace pattern was different; the CS− trace was higher than the CS+ trace in the aversive training, but the CS+ trace was higher than the CS− trace in appetitive condition (fig. S7, D to F). The trace pattern of R + P training resembled appetitive training, with no apparent difference. However, the CS+ trace is higher than the CS− trace, suggesting that the effect of the approach is beginning to form, which is consistent with our behavior result. It is worth mentioning that the R + P 24-hour memory trace also showed that the CS+ trace was higher than the CS− trace, indicating that this was approach behavior. Although a memory trace was observed in the γ2 region of KCγ 2 to 3 hours after the appetitive and aversive training, no memory trace was found in R + P training (fig. S7, G to I). However, again, the pattern of the CS+ trace higher than the CS− trace was shown in R + P training. No CS+ and CS− fluorescent differences were observed in the γ2 region of KCγ in the R + P condition 24 hours after training (fig. S7, J to L), suggesting the specificity of the γ1pdec region of KCγ in the R + P condition. These results indicate that a contradictory memory is formed, and the traditional individual’s aversive and reward memory trace is changed. The neural circuit to mediate appetitive, P + R, and R + P conditioning is illustrated in Fig. 7J.

Fig. 7. Contradictory training uses different neural pathways for long-term contradictory information storage.

(A to C) Top: Experimental setup and protocol. Bottom: GCaMP7f was expressed in KCαβ using VT49246-GAL4 driver line. The α3 region of KCαβ was recorded in appetitive, aversive, or R + P–trained flies. Increased GCaMP7f signal was found in response to CS+ in appetitive and aversive trained flies, but not in R + P group (n = 10, P = 0.035; n = 7, P = 0.040; n = 10, P = 0.107). (D to F) VT30604-GAL4 was used to express GCaMP7f in KCα’β’, and the β’1 region was recorded in appetitive, aversive, or R + P–trained flies. The GCaMP7f signal showed an increase of CS+ signal in flies with R + P and aversive conditioning (n = 12, P = 0.995; n = 12, P = 0.016; n = 24, P = 0.035). (G to I) GCaMP7f was expressed in the KCγ, and γ1pedc region was recorded in flies conditioned with appetitive, aversive, or R + P. Increased GCaMP7f signal was observed in R + P–conditioned flies upon exposure to CS+ (n = 16, P = 0.128; n = 13, P = 0.067; n = 18, P = 0.005). Calcium levels are represented as changes in fluorescence intensity (ΔF/F). (J) The current study proposed competition-integration mode: Contradictory training recruits different circuits to process the contradictory memory, which is also training type dependent. All data are presented as mean ± SEM. Statistical comparisons were carried out by two-tailed paired t test. Asterisks indicate statistical significance; *P < 0.05 and **P < 0.01.

DISCUSSION

The current study on contradictory memory formation offers many substantial findings. We propose a competition-integration mode based on our behavioral and live-cell imagining study (Fig. 7J). We identified a different memory trace used to process contradictory memory, using a circuit distinct from conventional aversive and reward traces. Our study suggests the existence of a unique, training-dependent neural circuit for integrating contradictory information. We hypothesized that different training strengthens plasticity between different neurons to establish distinct neural circuits. Our competition-integration mode suggests that the brain reconciles the contradictory information and then conducts a behavior suitable for survival. Memory integration is also energy saving for a brain to store a unified memory instead of multiple contradictory memories. Within a proper training time interval, two competing memories integrate within shared neural circuits to form a different PSD-LTM. Two different memories compete in the beginning and then integrate into one memory. Although our two training protocols, P + R and R + P, each contain the same aversive and appetitive training, different training sequences recruit similar but not the same neural circuits to process the information. (i) KCγ is for the acquisition in both conditions. This result is consistent with previous findings that KCγ is the primary neuron for acquisition for many different training procedures. (ii) Our results suggested that the dopamine signal, PPL1-γ1pedc neuron, is involved in integrating contradictory memory in both P + R and R + P condition. (iii) KCγ consolidates P + R memory, while KCαβ and KCγ are for R + P memory consolidation. (iv) Although neurons to retrieve the memory are different, KCαβ and KCα’β’ output is needed for 24-hour P + R memory retrieval, KCγ and KCα’β’ output is needed for 24-hour R + P memory retrieval, and MBON-β’1 is needed for both conditions to retrieve the memory. These results suggest that memory formation is highly sensitive to training. Different types of training recruit different circuits to form memory.

In our contradictory training paradigm, we observed two contrasting valences competed only during an early stage. The reward memory gradually dominated animals’ behavior in 3 hours in P + R condition, and the avoidance behavior gradually decayed within 3 hours in the R + P condition. Therefore, we hypothesize that parallel competition only briefly appears in the brain after training. Animals prefer their most recent training experience after complex training when immediately evaluated. This phenomenon has been deemed the recency effect (44), where resulting flies showed memory of the most recently trained odor (30, 45). Our finding that flies tend to act their recent training experience leads us to suggest that the recency effect is dominant but limited to the early stage, and contradictory information will eventually be integrated.

Although animals’ behavior is predominantly approaching behavior after memory integration, we observed that contradictory memory contains some characteristics found in aversive and appetitive conditioning. The finding of both P + R and R + P showing protein synthesis–dependent memory, which can last for at least 24 hours after one training cycle, is similar to appetitive training. Reward memory is distinctive in its PSD-LTM persistence after one training cycle (29, 31, 46). Despite these similarities, we found that c-LTM is unequivocally slower in consolidating stable form memory than conventional reward memory. This slows c-LTM formation process resembles the slow memory formation seen in aversive LTM. These results suggest that memory integration is a mixture, representing part of each memory’s characters.

How contradictory memory is formed? Our data showed that the primary entrance is the KCγ circuit, PPL1-γ1pedc neuron, KCγ, and MBON-γ1pedc > αβ. The KCγ has been associated with early olfactory information processing (4, 8, 22). Evidence then proposed the KCγ to partake in all types of memory, including LTM (47, 48). We showed that the activity of the KCγ circuit after training is critical to regulating c-LTM formation for the following reasons: (i) blocking output of PPL1-γ1pedc neuron during the early stage of consolidation affects c-LTM formation; (ii) overexpression of RICIN^cs in KCγ to prevent protein synthesis decreases c-LTM formation; (iii) increased KCγ circuit activity after training facilitates protein synthesis–dependent memory formation in P + R condition and promoting R + P c-LTM formation. (iv) The 24-hour memory trace was observed in KCγ. Accumulated evidence has demonstrated that the PPL1-γ1pedc neuron is critical in mediating aversive and reward memory (39–41, 49). Activating PPL1-γ1pedc neuron and odor stimulation form an aversive memory that makes animals avoid the odor cue (50). PPL1-γ1pedc neuron mediates reward memory expression in hunger flies (51, 52). Our results lead us to hypothesize that the dopamine signal is critical to integrating two memories; inhibiting the PPL1-γ1pedc neuron reduces avoidance memory to allow the formation of approach behavior, while PPL1-γ1pedc neuron activation facilitates approach behaviors in contradictory training. A dynamic change in PPL1-γ1pedc neuron activity after contradictory training where the activity of PPL1-γ1pedc neuron, in the beginning, is low but gradually increases. This dynamic change of dopamine signal activity in the early consolidation phase promotes memory integration to form c-LTM formation. The dynamic change of dopamine signal to regulate reward and aversive memory expression is consistent with recent findings in mice (20).

Why is approach behavior gradually dominant? We reasoned at least two possibilities: consolidation rate and internal status. It has been shown that consolidation of reward memory is faster than aversive memory, as has been demonstrated in another study (31). The difference in the consolidation rate may bias the memory integration process toward a reward-favored direction. In fruit flies, appetitive conditioning requires starvation to promote reward memory, and feeding after conditioning affects reward memory expression (51). We speculate that starvation is a facilitator in promoting approach behavior formation, as feeding after contradictory training delays the prominence of approach behavior. Although our findings are limited to starved animals, a parallel memory formation mode in fruit flies is also built on the same training protocol. Appetitive conditioning for laboratory animals is based on motivation for food or chemicals, either due to starvation or preference differences; the internal desire drives animals to obtain the reward. Removing the internal desire to build a reward memory in animals is difficult. It is of great interest for future studies to use nonstarved animals or design a protocol to minimize internal desire in animals to study contradictory memory formation to evaluate aversive and reward motivation in one training condition equally.

Our memory integration findings do not seem consistent with the findings of two recent studies. Unlike the study of Kong et al. (20), where the reward experience facilitates aversive memory extinction, our current study demonstrated that contradictory training formed a different memory trace rather than suppressing the aversive memory. This discrepancy could be due to differences in experiment procedures. Kong et al. (20) presented training cues of reward and aversive during the extinction phase, while we used contradictory training and examined the memory expression afterward. A cricket study used a similar experimental paradigm to ours, showing that appetitive and aversive conditioning form parallel memories and compete (19). It has been known in cricket that reward and aversive memory can be selectively damaged by octopamine receptor and dopamine receptor antagonists, respectively (53, 54). However, dopamine is the primary neurotransmitter in flies to reinforce aversive and appetitive learning. Therefore, the discrepancy could be due to the circuit difference. Although a recent study from C. elegans showed competition after conditioning (23), our study also showed competition after training but integration for LTM. Furthermore, our conclusion is not only based on genetic behavioral study but also on live-cell imaging, which demonstrates integrated memory trace formation.

Despite the trait of aversive memory, behaviorally, within 3 hours, our studies showed that the neuronal circuits for the approach behavior of c-LTM are different from either aversive or reward memory. We concluded that contradictory training recruits different neuronal circuits for the c-LTM formation, which are listed below. (i) Protein synthesis is required for c-LTM in KCγ. For aversive training, protein synthesis resides externally from MB, in the dorsal-anterior-lateral neurons (55). As for the reward LTM, although we could not exclude the possibility that protein synthesis is not within the KCs, as is the limitation of RICIN^cs (56), our data did not find LTM impairment after induction of RICIN^cs. (ii) Neurons consolidating c-LTM differ from aversive and reward LTM formation (5, 7, 8). (iii) The involvement of neurons in retrieving the c-LTM also differs from the aversive and reward LTM formation (7, 8, 32). (iv) Different memory trace storage. All these factors urge the consideration of reevaluating current modes of contradictory memory as the embodiment of a parallel system where independent reward and aversive memories form and where the behavioral outcome is merely a race in decay rates (15, 20, 57).

Contradictory memory formation is an intermingling of positive and negative modulating neuronal activity. This entanglement resulted in modifying multiple network configurations, creating a different circuitry unique to traditional known memory pathways. In our condition, c-LTM favors reward memory after a brief competition, which basic animal instincts may explain. We speculated that a brain evaluates the incoming information based on the similarity and the self-status to form a favorable circuit to conduct an appropriate behavior for survival. To our knowledge, our study is the first to reveal a neural circuit integrating contradictory information and localizing the memory trace to confirm a distinct memory trace formation. Together, our data suggest and support the ideology of viewing memory as a multicomplex system interacting dynamically with all acquired information rather than as independent components acting in competition (58).

MATERIALS AND METHODS

Chemicals, fly strains, and software used in this study

Mineral oil (heavy) (330760), OCT (99%) (218405), MCH (98%) (153095), EA (99.8%) (270989), and sucrose (99.5%) (57501) are from Sigma-Aldrich. Isopentyl acetate (99%) (B21618) is from Alfa Aesar. CHX (14126) is from Cayman. W1118, UAS-Shibire^ts is from Y. Zhong. UAS-TrpA1 (26263), UAS-ricin^cs (38623), R16A06-GAL4 (48709), C739-GAL4 (7362), MB057B-GAL4 (68277), MB011B-GAL4 (68294), MB320C-GAL4 (68253), MB112C-GAL4 (68263), MB504B-GAL4 (68329), MB304B-GAL4 (68367), and 20xUAS-jGCaMP7f (80906) are from Bloomington Drosophila Stock Center. VT49246-GAL4, VT30604- GAL4, and VT57244- GAL4 are from the Vienna Drosophila Resource Center. FV31S-SW is from olympus-lifescience.com. Fiji (59) is from https://imagej.net/software/fiji/.

Behavioral analysis

The olfactory avoidance and appetitive paradigms were performed by training 2- to 5-day-old flies in T-maze as described (45). Flies were food-deprived for 16 to 20 hours before conditioning in empty plastic vials containing water-dampened filter paper. The two odorants, OCT and MCH, were used for conditioning. Aversive conditioned flies were exposed to the first odor (CS+; OCT or MCH) for 1 min, paired with 12 × 1.5 s pulses 90-V dc electric shock, followed by 45 s of clean air stream, and then exposed to the second odor (CS−; OCT or MCH) without pairing of electric shock. For appetitive conditioning, flies were placed in the CS− tube (no sucrose) and exposed to the first odor (CS−; OCT or MCH) for 2 min, followed by 30 s of clean air. Then, they were transferred to the CS+ tube (lined with dried 2 M sucrose solution) and exposed to the second odor for 2 min (CS+; OCT or MCH). For testing, flies were given 2 min to choose between the CS+ or CS− odors in a T-maze. The behavior was conducted in a temperature-controlled room with 70% relative humidity under dim red light. For each experiment, the half PI was calculated by subtracting the number of flies choosing the CS+ odor from those choosing the CS− odor and then dividing by the total number of flies. The final PI value is the average of two complementary experiments with each odor.

Cold shock anesthesia for anesthesia-resistant memory (ARM) assays was performed as described previously (60). Flies were transferred to empty prechilled plastic vials 1 hour before memory testing. The vials were submerged into a 4°C ice bucket for 2 min. The immobile flies were transferred back into their previous vials with water-dampened filter paper to allow for recovery before testing.

UAS-TrpA1 and UAS-Shi^ts transgenic flies were raised at 18°C. Experiments using these flies were conducted at 23°C, and they were shifted to 31°C for 15 to 30 min as needed to activate or block the relevant neurons. For transient inhibition of protein synthesis in neuron, a reversible cold-sensitive 28S ribosomal RNA cleavage toxin, RICIN^CS, was expressed in UAS-RICIN^CS transgenic flies (35, 36). UAS-RICIN^CS flies were maintained at 31°C to express RICIN^CS to disrupt protein synthesis. All experiments not involving temperature-sensitive transgenic flies were performed at 27°C.

Drug feeding

Starved flies were fed 35 mM CHX in ddH₂O 16 to 18 hours before and 24 hours after the training. Controlled flies were starved with ddH₂O only for the same duration. For all memory assays, flies were placed back in food vials after training, either containing CHX or ddH₂O only, till they were to be tested.

In vivo calcium imaging

Flies were collected and trained according to the previously mentioned behavioral assay before calcium imaging. Each fly was placed into a trimmed 200-μl tip, securing its head for further preparation. The head was secured using flowable composite (Flow-It ALC, Pentron Clinical, USA) and hardened with ultraviolet light (ELiTEDENT, Rolence Enterprise, Taiwan). Next, the cuticle of the target area was carefully removed with fine tweezers (Dumont#5 500341, World Precision Instruments). Fat bodies and fibers were removed to enhance image acquisition. The prepared fly was then placed in a custom-made chamber on the stage of upright confocal laser scanning microscope (FLUOVIEW, FV3000, Olympus), equipped with 40× water-immersion objective lens (Olympus) and illumination units (Olympus). Next, adult hemolymph-like medium buffer (61) was used as the medium between the fly head and the water immersion objective lens. GCaMP signals were excited using 488-nm light, with varying intensities from 0.2 to 1%. Time-series images were captured using FV31S-SW Software (version 2.6) at a frame rate of 4 per second (62). The air delivery for all time-series images followed this sequence: 20 s of air, 10 s of CS+ odor, 30 s of air, 10 s of CS− odor, and 10 s of air. This process was managed by our custom-made gas control system.

Processing and quantification of calcium imaging data

All images were processed and quantified using Fiji (59). All collected time-series images were registered using TrakEM2, a built-in plugin of Fiji. Regions of interest (ROIs) were manually defined on all images. For normalization, the fluorescence of the ROIs was subtracted from the fluorescence of a background region of the same size and shape near the ROIs at each time point. We calculated ∆F/F₀, which represents the time-dependent change in relative fluorescence intensity. Here, F0 is the average normalized ROI intensity from the frames taken 5 s before the odor was delivered (t = −5 to 0 s). ∆F is obtained by subtracting F0 from the normalized ROI intensity at each time point. The bar graphs show the mean of ΔF/F₀ during the 10-s period of odor delivery.

Statistical analysis

All raw data were tabulated and stored using Microsoft Excel and statistically analyzed using GraphPad Prism version 8.0. One-way or two-way analysis of variance (ANOVA) was used for multiple comparisons followed by Sidak’s post hoc test. For comparison of two experimental groups, P values were calculated by unpaired t test. However, responses of CS+ and CS− were compared by two-tailed paired t test within the same group.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7