Introduction

Learning defines a behavioral change that results from acquiring information about the environment, and memory refers to the process by which the information is encoded, stored, and later retrieved. Learning and memory forms the basis for higher brain functions, including cognition and decision making, which shapes our individuality¹. On the cellular and physiological level, learning and memory is achieved by neuroplastic changes in circuits, including neuronal excitability and synaptic plasticity. Usually distinct types of neurotransmitters, such as dopamine, modulate these changes.

Dopamine (DA) plays an important role in many mammalian brain functions, including motor functions, motivation, reinforcement, addiction, and learning and memory^2–4. Dopaminergic neurons (DAN) are mainly located in the mesencephalon: DANs in the substantia nigra (SN) are responsible for motor functions, while those in the ventral tegmental area (VTA) are important in reward, addiction, learning and memory^3,5. Dopamine achieves its functions via two families of G protein-coupled receptors (GPCR): excitatory D1-like and inhibitory D2-like receptors⁴. All D1-like receptors are located post-synaptically; in contrast, D2-like receptors both function post-synaptically and play an important presynaptic role, regulating dopamine release through negative feedback⁴. All these receptors are important in mammalian associative learning^6,7. D1-like receptors elevate intracellular cAMP by activating adenylyl cyclase (AC) via Gα_s, while D2-like receptors repress cAMP by inhibiting AC via Gα_i/o. cAMP activates protein kinase A (PKA), leading to the phosphorylation of DARPP-32 (dopamine and cyclic AMP-regulated phosphoprotein, 32kDa), ion channels, and CREB (cAMP response element-binding protein). In addition, dopamine receptors also activate the PLC-PKC, MAPK, and CaMKII pathways^2–4,8–10.

Although mammalian studies reveal mechanisms more relevant to human beings, the complexity of the nervous system impedes our understanding about the basic or more universal principles of learning and memory applicable generally to all nervous systems¹¹. With a simple central nervous system (CNS) and powerful genetic tools, the fruit fly Drosophila melanogaster has become a popular model organism in learning and memory research^12,13. With conserved genes in dopamine metabolism and signaling¹⁴, as well as fundamental similarities in the olfactory circuitry compared to mammals^15,16, Drosophila can perform olfactory associative learning in both larvae^17–19 and adults^20–24. Olfactory associative learning is a type of classical conditioning in which flies are trained under positive or negative reinforcement paired with an odorant. Different from the naïve reaction to the odorant, flies approach the odorant after being trained with rewards (e.g., sucrose; appetitive)²¹, but avoid the odorant when trained with punishments (e.g., electric shock, bitter taste chemicals; aversive)²⁰. Several genes related to the cAMP-PKA signaling pathway, including dunce (dnc)²⁵ and rutabaga (rut)²¹, are expressed in the mushroom body (the center for Drosophila learning and memory)^26,27. Mutations of these genes lead to learning deficiencies, indicating their roles in olfactory learning^21,25.

Drosophila larvae offer several advantages for studying olfactory learning compared to adults. Notably, compared to neural circuits underlying olfactory learning, their simpler neural circuitry²⁸, characterized by fewer olfactory receptor neurons (ORNs)^29,30, projection neurons (PNs) ³¹, mushroom body neurons (MBNs)³², and dopaminergic neurons (DANs)^33,34, facilitates the elucidation of underlying mechanisms. Additionally, larvae exhibit simpler behavioral patterns, facilitating experimental manipulations and observations. Furthermore, their translucent cuticles enable convenient application of techniques such as optogenetics³⁵, further enhancing the experimental versatility of larval studies.

Like in mammalian brains, dopamine achieves its functions via four dopamine receptors in flies, two D1-like receptors dDA1³⁶ (or Dop1R1) and DAMB³⁷ (or Dop1R2), one D2-like receptor D2R³⁸ (or Dop2R), and one non-canonical receptor DopEcR^14,39. dDA1 is mainly found in the mushroom body^40,41 and is necessary for appetitive and aversive olfactory learning in larvae and adults^41,42. D2R in the mushroom body is necessary for anesthesia-resistant memory⁴³. In addition, D2R in GABAergic anterior paired lateral (APL) neurons is known to secure aversive conditioning in adult flies⁴⁴. Although D2R expression has been reported in the ventral nerve cord⁴⁵, neither its expression in larval brains, nor its functions in larval olfactory learning have been investigated.

By using a GFP-tagged D2R strain, we detected the expression pattern of D2R in the third- instar larval brain, specifically in dopaminergic and mushroom body neurons. Knockdown of D2Rs in DAN-c1 impaired aversive learning, while knockdown of D2R in mushroom body neurons led to deficits in both aversive and appetitive learning. These results revealed that D2Rs in distinct brain structures mediate different learning tasks. The newly discovered roles of D2R in the larval brain provides new insights into the mechanisms underlying larval associative learning.

Results

Distinct dopaminergic neurons innervate different compartments of the mushroom body

The connectome of larval learning has been investigated in both first- and third-instar larvae^28,46. The mushroom body works as the learning center in Drosophila, which is composed by αβ, α’β’, and γ neurons in adult brains⁴⁷. In larvae, axons from γ neurons bifurcate and form the vertical and medial lobes, as αβ and α’β’ neurons are not mature^33,48,49. These lobes are divided into 11 compartments (refer to Figure 1) based on the coverage of neurites from both mushroom body extrinsic neurons (MBEN) and mushroom body output neurons (MBON)²⁸. Around 21 dopaminergic neurons are found in each brain hemisphere, and categorized into 4 clusters: DM1 (dorsomedial), pPAM (primary protocerebral anterior medial, or DM2)⁵⁰, DL1 (dorsolateral), and DL2³⁴. DL1 neurons project to the vertical lobe⁴¹, while pPAM neurons innervate the medial lobe⁵⁰ (refer to Figure 1a-c).

Figure 1.

In this study, we wanted to functionally identify individual DANs that mediate larval olfactory learning. The first step was to search for DAN-specific driver strains that mark single dopaminergic neurons, which subsequently can be used to target genetic manipulations of corresponding neurons. A total of 56 driver strains identifying dopaminergic neurons were screened in this study (Table 1). These strains were chosen based on previous studies, either identifying a single dopaminergic neuron in larvae, or identifying only several in adult brains and indicating the potential of identifying single DAN in larvae. TH-GAL4, a traditional dopaminergic neuronal driver⁵¹, identifies all dopaminergic neurons except those in pPAM (Figure 1d, Figure S1a). Split-GFP reconstitution across synaptic partners (GRASP) technique was used to investigate the “direct” synaptic connections from DANs to the mushroom body (MB), in which portions of GFP were specifically expressed in corresponding neurons (Figure S2d). GRASP results showed neurons under TH-GAL4 formed synapses in the vertical lobe and lower peduncle (white dash lines in Figure 1d’), consistent with previous electron microscopy data⁴⁶. We found three DAN driver strains that identify single dopaminergic neurons in the third-instar larval brain hemisphere. DAN driver R76F02-AD;R55C10-DBD identifies a dopaminergic neuron innervating the lower peduncle (LP), which would be DAN-c1 based on previous published nomenclature²⁸ (Figure 1f). MB296B driver identifies the dopaminergic neuron (DAN-d1) projecting to the lateral appendix (LA) (Figure 1g), as well as many non-dopaminergic neurons. SS1716 driver identifies one dopaminergic neuron forming synapses in the lower vertical lobe (LVL), indicating it is DAN- g1 (Figure 1h).

Table 1.

In pPAM, R58E02 driver identifies DAN-h1, i1, and j1, innervating the shaft (SHA), intermediate toe (IT) and upper toe (UT) (Figure 1i), and R30G08 driver identifies DAN-h1 and i1 (Figure 1j). As described in a previous report²⁸, SS864 driver identifies DAN-i1, innervating the upper toe (Figure 1l), and SS1757 driver identifies DAN-k1 which innervates the lower toe (LT) (Figure 1m). In contrast, SS1696 driver identifies not only DAN-h1, but also i1 and one DL1 neuron not innervating the mushroom body (Figure 1K).

In summary, our results show that five DL1 and four pPAM DANs innervate nine distinct mushroom body compartments in a one-to-one pattern (Figure 1b and c). DL1 neurons innervate the vertical lobe and peduncle, while pPAM neurons project to the medial lobe. The single neuronal driver strains screened can be used to investigate the roles of individually identified DANs in larval olfactory learning.

Dopamine release from DAN-c1 is both necessary and sufficient for larval aversive learning

Dopamine plays an important role during olfactory associative learning in both adults and larvae^52,53. In adults, dopaminergic neurons in PPL1 regulate aversive learning^54–57, while those in PAM mediate reward signals in appetitive learning^58–60. In larvae, DL1 neurons innervating the vertical lobe and the peduncle are required for aversive learning^18,41, while those in pPAM projecting to the medial lobe are involved in appetitive learning⁵⁰.

In Figure 1 and Table 1, three driver strains identifying single dopaminergic neurons in DL1 were discovered, which could be candidates to investigate their roles in larval aversive learning. The R76F02-AD;R55C10-DBD strain identifies MB-MP1 in the adult brain⁶¹, which is a dopaminergic neuron involved in adult aversive learning⁶². This driver strain identifies one neuron per hemisphere in the third-instar larval brain (Figure 2a-c, Figure S1c). Using a UAS- DenMark;UAS-sytGFP strain, its dendrites were labeled with RFP and axonal terminals were marked by GFP. Its dendrites were localized in the dorsomedial protocerebrum (dml), and its axonal terminals located in the lower peduncle of the mushroom body (Figure 2a and d), with GRASP results supporting the existence of synapses in this compartment (Figure 2E). All these characteristics are consistent with the previously published nomenclature²⁸, indicating that this neuron is DAN-c1 (Figure 2f), thus, this strain will now be referred to simply as DAN-c1.

Figure 2.

To reveal the role of DAN-c1 in larval olfactory learning, a single odor learning paradigm and thermogenetic tools were applied^17,18,63. Compared to those trained with distilled water (DW), control larvae exhibited repulsion to the odorant pentyl acetate (PA) after being trained with quinine (QUI) paired with PA, reflecting aversive learning. In contrast, larvae were attracted to PA after being trained with sucrose (SUC) paired with PA, reflecting appetitive learning. The extent of repulsion or attraction was represented with a response index (R.I.) that is compared to the DW group (Figure 2g).

To examine the role of DAN-c1 in aversive learning, we used a Shibire^ts1 strain, which encodes a thermosensitive mutant of dynamin blocking neurotransmitter release when the ambient temperature is higher than 30°C by repressing endocytosis and vesicle recycling functions¹⁷. When trained at 34°C, the complete inactivation of dopamine release from DAN-c1 with Shibire^ts1 impaired aversive learning (Figure 2h), while Shibire^ts1 did not affect learning when trained at room temperature (22°C). These data indicated that dopamine release from DAN-c1 is necessary for larval aversive learning to occur.

In the next experiments, a fly carrying temperature-sensitive cation channel, dTRPA1, was used to excite the DAN-c1 neuron because it can be activated at temperatures higher than 30°C⁵⁸. Activation of DAN-c1 with dTRPA1 at 34°C during training induced repulsion to PA in the distilled water group (Figure 2k). These data suggested that DAN-c1 excitation and presumably increased dopamine release is sufficient for larval aversive learning in the absence of gustatory pairing.

Combining the blockade results with Shibire^ts1, these data revealed that dopamine released from DAN-c1 activation is both necessary and sufficient in larval aversive learning. However, when paired with a gustatory stimulus (QUI or SUC), activation of DAN-c1 during training impairs both aversive and appetitive learning (Figure 2k). We suggest that these data indicate a critical role for the amount of DAN-c1 dopamine release in larval associative learning, as dTRPA1 stimulation may result in excessive dopamine release.

The expression pattern of D2R in the third-instar larval brain

Although dopamine D1-like receptors have been proven important for learning⁴², the role of D2-like receptors has not been fully investigated. In addition, the expression pattern of D2R in fly brains was not reported. A fly strain expressing GFP-tagged D2R (BDSC#60276) was used to reveal the expression pattern of D2R in the third-instar larval brain (Figure 3a). D2Rs were found in dopaminergic neurons and the mushroom body. In dopaminergic neurons (Figure 3b-g), D2Rs were found in DM1, pPAM, DL2b, and some DL1 neurons. In the mushroom body, D2Rs were expressed in the soma and lobes, but not in the calyx (Figure 3h and i). Even though D2Rs were widely found in vertical lobes, medial lobes, and peduncles, they were not expressed in every mushroom body neuron. A transection of the peduncle region showed the absence of D2Rs in the core area (Figure S3i), which is composed of densely packed newly created fibers and lacks Fasciclin II (FAS II)⁴⁹.

Figure 3.

To inspect whether the pattern of GFP signals indeed reflected the expression of D2R, three D2R enhancer driver strains (R72C04, R72C08, and R72D03-GAL4) were crossed with the GFP-tagged D2R strain. R72C08-GAL4 covered three DM1 dopaminergic neurons (Figure S3c), and R72C04-GAL4 labeled one DM1 and two DL2b dopaminergic neurons (Figure S3d and e). R72D03-GAL4 identified parts of mushroom body neurons, whose axons spread on the surface of the mushroom body lobes (Figure S3f); R72C08-GAL4 also identified a subset of neurons from four MB neuroblasts, with soma in four clusters and a converged bundle of axons (Figure S3g and h). These results revealed the expression of D2R in the mushroom body and dopaminergic neurons in the third-instar larval brain.

D2R in DAN-c1 influences larval aversive learning

Our previous work reported that D2R knockdown (UAS-RNAi) in dopaminergic neurons driven by TH-GAL4 impaired larval aversive learning⁶³. Using a microRNA strain (UAS-D2R-miR)⁶¹, a similar deficit was observed (Figure S5c). To further understand the roles of D2R in aversive learning, its expression in distinct DANs, as well as corresponding learning assays were both investigated. Crossing the GFP-tagged D2R strain with a DAN-c1-mCherry strain demonstrated the expression of D2R in DAN-c1 (Figure 4a).

Figure 4.

To reduce the expression of D2R in dopaminergic neurons, a microRNA strain UAS-D2R-miR was used when crossing with distinct driver strains. The efficiency of D2R knockdown was confirmed by crossing the GFP-tagged D2R strain with TH-GAL4;UAS-D2R-miR. In these larval brains, GFP signals in DM1 were not detected, while those in pPAM were still intact (Figure 4b and c). Quantification showed a significant decrease of GFP signals in the knockdown group compared to the control (Figure 4d), indicating reduced transcripts of D2R linked GFP by D2R-microRNA (Figure S4f).

To investigate the roles of D2R in distinct dopaminergic neurons during larval associative learning, UAS-D2R-miR strain was crossed with distinct single dopaminergic neuronal drivers. Among them, the knockdown of D2R in DAN-c1 impaired aversive learning with the odorant pentyl acetate, while appetitive learning was unaffected (Figure 4e). In contrast, although D2R was also found in DAN-d1 and DAN-g1, neither D2R knockdown in DAN-d1 nor in DAN-g1 affected larval olfactory learning (Figure S4c-e). As the naïve sensory and motor functions were not affected, this deficiency was caused by impairment in learning abilities (Figure 4f-i). Similar learning deficits were observed in the same strain trained with another odorant, propionic acid (Figure S5a), as well as in larvae with D2R knockdown using UAS-RNAi (Figure S5b). These results demonstrated that D2Rs are expressed in DAN-c1, and they are necessary for larval aversive learning. Presumably the knockdown of presynaptic inhibitory D2R autoreceptors on DAN-c1 will result in increased and excessive dopamine release, which leads to aversive learning deficiency. These results are consistent with the activation studies with dTRPA1 above, in which increased dopamine release during training results in impaired aversive learning (Figure 2k).

Over-excitation of DAN-c1 during learning impairs larval aversive learning

To exclude possible chronic effects of D2R knockdown during development, optogenetics was applied at distinct stages of the learning protocol. Channelrhodopsin2 (ChR2) is a blue light activated cation channel from algae, which can be used to activate target neurons^64,65. Over-excitation of DANs under a TH-GAL4 driver with ChR2 during training impaired aversive learning but left appetitive learning intact (Figure S7), which is consistent with D2R knockdown results. To investigate the mechanisms with a better temporospatial resolution, ChR2 was expressed in DAN-c1, and blue light was applied at distinct stages of the learning protocol (Figure 5a). Optogenetic activation of DAN-c1 during training impaired aversive learning, not appetitive learning (Figure 5b-d). This result is consistent with the effect of D2R knockdown in DAN-c1, indicating that increased, excessive dopamine release during training leads to impaired aversive learning.

Figure 5.

D2R in mushroom body mediates larval learning through inhibition

We have shown that D2R in DAN-c1 plays a critical role in larval aversive learning. Since D2Rs are also expressed in soma and axons in most mushroom body neurons (Figure 3h and i), we examined the role of D2R in MB neurons, the center for learning in Drosophila. Knockdown of these D2Rs by D2R-miR impaired both appetitive and aversive learning (Figure 6a). Similarly, optogenetic activation of mushroom body neurons during training led to deficiencies in both appetitive and aversive learning (Figure 6b and d). These deficiencies were not observed in larvae with activation during the resting stage. As D2Rs are inhibitory receptors, and optogenetic activation leads to greater neuronal excitation like what may occur with knockdown of D2Rs, these data show that the inhibitory effect of D2Rs in mushroom body neurons is necessary for larval olfactory associative learning to occur.

Figure 6.

Discussion

The dopaminergic system plays an important role in Drosophila olfactory associative learning, but the roles of D2R in this process have not been fully explored. In this study, we systematically investigated the expression pattern of D2R in the third-instar larval brain as well as its role in larval aversive and appetitive learning. One driver strain singly identifying the DAN-c1 neuron was discovered and learning assays with thermogenetic tools (Shibire^ts1, dTRPA1) determined that the blockade of dopamine release from DAN-c1 impeded aversive learning, while its activation during training led to repulsion toward the odor in the absence of unconditioned stimulus (i.e., QUI). These results revealed that DAN-c1 activation (i.e., presumably leading to the release of synaptic dopamine) is both necessary and sufficient for larval aversive learning to occur. Subsequently, the expression pattern of D2R was explored by using a GFP-tagged D2R strain, including distinct dopaminergic and mushroom body neurons.

D2Rs were expressed in DAN-c1, and the knockdown of these receptors resulted in aversive learning deficiency. These data suggest that presynaptic D2Rs in a single dopaminergic neuron, DAN-c1, regulate dopamine release during excitation whereas knockdown of these same receptors leads to excessive dopamine release, causing deficits in aversive learning to occur. Furthermore, the activation of DAN-c1 with optogenetic tools during training, resulting in excessive dopamine release, impaired aversive learning, as well. Finally, it was demonstrated that either the knockdown of postsynaptic D2R or activation of mushroom body neurons led to learning deficits. These data demonstrate that D2Rs in distinct brain locations are critically involved in associative learning.

Insights into the neuronal circuits underlying larval olfactory associative learning

Mushroom body and dopaminergic neurons play important roles in Drosophila associative learning^66–69. In larvae, olfactory information (odors, Conditioned Stimulus, CS) is received by olfactory sensory neurons (OSN), then transmitted to the mushroom body via projection neurons (PN)⁷⁰. The mushroom body (MB) is the learning center of Drosophila and composed of Kenyon cells (KC, or mushroom body neurons, MBN)^71–73. Their dendrites form the calyx, receiving olfactory information from projection neurons. The axons converge into peduncles, then branch into the vertical and medial lobes. Distinct gustatory cues (taste, Unconditioned Stimuli, US) are sensed by gustatory sensory neurons (GSN) and transferred to dopaminergic neurons in different clusters (Figure 7A). DAN-c1 in the DL1 cluster mediates aversive cues, projects to the lower peduncle (LP) in the mushroom body. The plasticity of synapses from MBNs to MB output neurons (MBON) is negatively modulated by dopamine^67,74,75. The MBN-MBON synapses in the vertical lobe and peduncle are responsible for attraction, while those in the medial lobe are for repulsion^67,75.

Figure 7.

When only the odorant appears, the subset of Kenyon cells representing this odor are depolarized, inducing calcium influx⁷³. As a balance exists between compartments across the mushroom body lobes, the response to the odor depends on the naïve olfactory circuits from projection neurons to the lateral horn (LH). In aversive learning, in addition to olfaction induced calcium influx, gustatory stimuli also lead to dopamine release from DAN-c1 and subsequently, activation of Gα_s in LP⁵⁴. The co-existence of calcium and Gα_s activates a Ca²⁺-dependent adenylate cyclase (AC), rutabaga (rut)^76–79. Rutabaga converges information from both olfaction and gustation, working as the coincidence detector of associative learning. Its downstream signaling inhibits attractive MBN-MBON synapses in the LP, breaking the balance between distinct compartments. As a result, after learning, the larvae will exhibit repulsion when exposed to the odor again. In contrast, dopaminergic neurons in the pPAM convey appetitive cues, which project to compartments in the medial lobe. The co-existence of olfactory and appetitive gustatory stimuli leads to inhibition of these repulsive MBN-MBON synapses, inducing attraction.

The conserved role of DAN-c1 in aversive learning throughout Drosophila development

Adult Drosophila share similar neuronal circuits of learning with larvae^54,66–69. In adult brains, dopaminergic neurons are classified into 13 clusters, named as PAM (protocerebral anterior medial), PAL (protocerebral anterior lateral), PPM (protocerebral posterior medial), PPL (protocerebral lateral), and PPD (protocerebral posterior dorsal) clusters³². DANs in the PAM cluster innervate the medial lobe, while those in PPL1 project to the vertical lobe⁶⁹.

R76F02-AD;R55C10-DBD identifies two dopaminergic neurons in adult brains, MB-MP1 in PPL1 and ALT-PLPC in PPL2ab ⁶¹. MB-MP1 is also named PPL1-γ1pedc, innervating both γ1 and the peduncle of the β lobe⁶⁹. Activation of this neuron induced aversive learning⁶², and activation of its corresponding MBON-γ1pedc>α/β led to approach⁸⁰. During metamorphosis, dopaminergic neurons in DL1 develop into the PPL1 cluster, DL2a neurons develop into PPL2ab, and those in pPAM will develop into the PAM cluster⁸¹. This driver strain only identifies DAN-c1 from DL1 in larvae, which innervates the lower peduncle of the mushroom body. Interestingly, previous reports revealed that memory can be transferred from larvae to adults⁸², indicating the maintenance of neuronal circuitry architecture during metamorphosis. This evidence supports that DAN-c1 is the corresponding neuron of PPL1-γ1pedc in larvae, which performs similar functions in aversive learning.

Pre- and post-synaptic D2Rs regulate cAMP in the mushroom body during aversive learning

The molecular mechanisms underlying Drosophila learning have not been fully determined. In a traditional view, gustatory cues elevate dopamine release, which binds to D1 receptors and then activates Gα_s. The co-existence of Gα_s and calcium elicited by olfactory cues activates rutabaga in axons of MBNs. Rutabaga transforms ATP into cAMP, activating PKA signaling pathway. Mutant flies with either insufficient (rutabaga) or excessive cAMP (dunce) showed aversive learning deficiency, indicating that the level of cAMP should be kept in an optimal range to achieve aversive learning^21,25 (Figure 7B).

Our results have shown that D2Rs in dopaminergic neurons and the mushroom body are important for larval aversive learning, suggesting a “dual brake” role in regulating cAMP levels in the MBNs through both pre- and postsynaptic components. On the presynaptic side, D2R in DAN-c1 decreases the release of dopamine under gustatory stimuli, reducing the probability of postsynaptic D1R activation in MBNs. On the postsynaptic side, D2R in MBNs inhibits the coincidence detector rutabaga (AC) both via activation of Gα_i/o and inhibition of voltage-gated calcium channels², indicating postsynaptic D2R functioning as a “brake of the coincidence detector”. Combining these, “dual brake” D2R ultimately regulates the mushroom body cAMP level within a physiologically optimal range during aversive learning. In addition, D2R fine tunes the functional concentration spectrum of dopamine with higher resolutions, as its dopamine affinity is 10- to 100-fold greater than D1 receptors. Overall, D2Rs work in a “dual brake” system both expanding the representation of a dynamic range of gustatory signal intensity with high signal to noise ratio and preventing postsynaptic overexcitation, which increases the reliability of DAN-c1 and MBN circuits for the larval aversive learning (Figure 7C & D).

Recent studies showed that the approach/repulsion in learning is achieved via inhibition of the repulsive/attractive representing compartments^67,74,75, which indicates dopamine inhibits acetylcholine release from MBN to MBON⁸³. However, the PKA signaling pathway usually elevates neuronal excitability and increases neurotransmitter release^2,3, which is contradictory to the recent findings. In addition to the “dual brake” role of D2R, our results suggest a third role of D2R in aversive learning. D1 and D2 receptors can form heteromeric receptors, whose downstream Gα_q activates PKC and CaMKII signaling pathways. The activation of these signaling pathways may reduce acetylcholine release⁸⁴ (Figure 7C & D).

Explaining the results of thermogenetic and optogenetic experiments

Activation of DAN-c1 with dTRPA1 induced aversive learning, while the repulsion disappeared when DAN-c1 was activated in the quinine group (Figure 2K). Our explanation is that quinine stimulation and temperature activation led to over-excitation of DAN-c1, which impaired aversive learning. This is consistent with the learning deficiency in larvae with D2R knockdown in DAN-c1 (Figure 4E). Results from optogenetic activation of DAN-c1 during aversive learning also support this (Figure 5B). However, in contrast to results with thermo-activation, larvae with optogenetic activation of DAN-c1 showed neither repulsion after being trained with distilled water (Figure 5C), nor did they show reduced attraction in the sucrose group (Figure 5D). One possible explanation is that the thermo-activation is relatively mild compared to the optogenetic activation. Based on this, thermo-activation of DAN-c1 is still in the physiological range of cAMP under the upper cutoff (Figure 7B), resulting in repulsion in the distilled water (DW) group, neutralized attraction in SUC group, and impaired repulsion in QUI group. In contrast, optogenetic activation of DAN-c1 overwhelmed the physiological conditions, leading to failure of repulsion in DW group (Figure 5C). This repulsive failure did not affect appetitive learning (Figure 5D), and a stronger over-excitation in QUI group induced similar failure (Figure 5B).

Distinct dopaminergic neurons may have different roles in larval aversive learning

Previous work reported that aversive olfactory learning was induced through the optogenetic activation of DAN-d1, f1, or g1, but not DAN-c1⁸⁵. This discrepancy can be explained as the optogenetic overexcitation of DAN-c1, similar to our optogenetic or D2R knockdown results. Our learning assay results from larvae with D2R knockdown in DAN-d1 or g1 also supported this: aversive learning was not affected by D2R knockdown (Figure S4E). These data indicate D2Rs in DAN-d1 or g1 are not involved in larval aversive learning. Additionally, the DAN-c1 strain used in the previous work (SS02160-split-GAL4) not only labels DAN-c1, but also marks other non-dopaminergic neurons, which may affect the results. Besides, live calcium imaging showed that DAN-d1, f1, and g1 responded to the activation of mechanosensory and nociceptive neurons⁸⁵, indicating a functional differentiation from gustatory activated DAN-c1^62,86.

In future studies, the molecular signaling downstream of D2R needs to be explored, as well as the comprehensive neuronal circuit architectures of larval learning. The neuronal circuits underlying learning and memory are complex networks, sharing similarities with the regulatory networks of gene expression. Studies of the mechanisms of learning and memory help us understand the essential principles of the non-linear dynamic characteristics in these complex systems. On one hand, the progress in larval learning provides useful information for helping our understanding about more complex systems, from brains in adult flies to those in mammals. On the other hand, the architecture of larval learning circuits could improve either hardware design or algorithm structures in artificial intelligence, which may bring more powerful tools, such as navigation systems regulating multiple auto-drive vehicles in complex 3-dimentional environments.

In conclusion, we explored the expression pattern of D2R in the third-instar larval brain and investigated their roles during larval olfactory learning. D2Rs were found in DAN-c1, and their knockdown induced deficiency in aversive learning. D2Rs were also expressed in MBNs, knockdown of which impaired both aversive and appetitive learning. This research revealed the important roles of D2Rs in Drosophila larval olfactory learning and enriched our understanding regarding the mechanisms underlying the learning process.

Materials and Methods

Fly stocks

All fly strains used in this study are listed in Table 1 and Table S1. Flies were maintained on standard medium, which consists of cornmeal, yeast, dextrose, sucrose and agar in water. Flies were kept in a 12/12-hour light/dark cycle at 25°C. Canton-S genotype (WT) and yw¹¹¹⁸ were used as wild-type. Strains carrying more than one transgene were constructed by standard genetic crosses with the w¹¹¹⁸; CyO/Sco; TM2/TM6 multiple balancer chromosome strain. Strain UAS-Syb::spGFP1-10, LexAop-CD4::GFP11, LexAop-rCD2::RFP /CyO; MB247-LexA::Up16/TM6B was made by chromosome swapping between UAS-Syb::spGFP1-10, LexAop-CD4::GFP11/CyO (II, second chromosome) and LexAop-rCD2::RFP (II).

The D2R gene is located on the X chromosome, with 6 different alternative splicing products. The GFP-tagged D2R strain is inserted with a GFP gene in the second intron, generating D2R molecules tagged with GFP^87,88. The D2R-miR strain produces microRNA recognizing the sequence across the third and fourth exons, which is not affected by the GFP insertion (Figure S4F).

GRASP

Split-GFP reconstitution across synaptic partners (GRASP) was used to investigate whether neurons formed synapses⁸⁹ (Figure S2d). Half of split GFP tethered to the presynaptic synaptobrevin (UAS-syb::spGFP1-10) was expressed in one type of neuron using UAS/Gal4 binary system, and the complementary split GFP linked to a membrane protein (LexAop-CD4::spGFP11) was expressed in another category of neuron with LexA/LexAop. If synapses between these neurons exist, the split GFPs would form a complete one and be recognized by a mouse antibody (Figure S2a-c).

Immunohistochemistry

All staining processes were performed in 1.5 ml Eppendorf tubes. Late third-instar (96-100h after egg laying) larval brains were dissected in dissection solution (300 mOsmol/L). Brains were fixed in 4% paraformaldehyde (PFA, Electron Microcopy Sciences, Cat. No. 15713) for 1 hour on ice. After three washes (0.1% bovine serum albumin in 10mM PBS, Sigma Life Sciences A9647), brains were incubated in the blocking and permeabilization solution (0.2% Triton X-100 in 10mM PBS with 5% normal goat serum; Triton-X 100, Sigma, T8532; NGS, Sigma-Aldrich, G9023) for 2 hours at room temperature. Incubation with primary antibodies was done overnight at 4°C. After three washes, brains were incubated in the secondary antibodies for two hours. Both the primary and secondary antibodies were diluted in the blocking and permeabilization solution. GFP antibody (Rabbit, Thermo Fisher Scientific, Cat. No. A6455, 1:1000), TH antibody (Mouse, Immunostar, Cat. No. 22941, 1:1000), goat anti-rabbit with green fluorescence (Invitrogen, Alexa Fluor 488 conjugate, Cat. No. A-11035, 1:1000), and goat anti-mouse IgG with far-red fluorescence (Alexa Fluor 633 conjugate, Cat. No. 21052, 1:500) secondary antibodies were used. After three washes, brains were transferred and mounted in the Fluoro-Gel with Tris Buffer (Electron Microcopy Sciences, Cat. No. 17985-10) on a piece of micro cover glasses (Electron Microcopy Sciences, Cat. No. 72200-41). Finally, the samples were covered with another piece of micro cover glasses.

A seven-day staining protocol was used for staining of GFP-tagged D2R or GRASP, in which brains were fixed in 1% paraformaldehyde in Schneider’s insect medium (Sigma Life Sciences, Cat.NO. S0146) overnight at 4°C. In the second day, the brains were rinsed and washed twice with PAT3 solution (0.5% Triton X-100 in 10mM PBS with 0.5% BSA), each for 1 hour. Then brains were incubated in the blocking and permeabilization solution (3% NGS in PAT3) for 2 hours at room temperature. Incubation of primary antibodies was done overnight at 4°C. On the third day, brains were rinsed and washed twice with PBT solution, then incubated in the secondary antibodies for five days. Both the primary and secondary antibodies were diluted in the PBTN solution. In the staining of GFP-tagged D2R, GFP antibody (Rabbit, ThermoFisher Scientific, Cat. No. A6455, 1: 1000), TH antibody (Mouse, Immunostar, Cat. No. 22941, 1:1000), and mCherry antibody (Rat, ThermoFisher Scientific, Cat. No. M11217, 1:1000) were used. For staining of GRASP, GFP antibody (Mouse, ThermoFisher Scientific, Cat. No. G6539, 1:100) was used. Goat anti-rabbit with green fluorescence (Invitrogen, Alexa Fluor 488 conjugate, Cat. No. A-11035, 1:1000), goat anti-mouse with green fluorescence (Alexa Fluor 488 conjugate, Cat. No. A11029, 1:1000), goat anti-mouse with far-red fluorescence (Alexa Fluor 633 conjugate, Cat. No. A21052, 1:500), and goat anti-rat with red fluorescence (Alexa Fluor 546 conjugate, Cat. No. A-11081, 1:1000) secondary antibodies were used. In the seventh day, brains were rinsed and washed twice with PBT solution, and then transferred and mounted in the Fluoro-Gel with Tris Buffer on a piece of micro cover glasses. Finally, the samples were covered with another piece of micro cover glasses.

Confocal imaging

All images were obtained with a Zeiss Laser Scanning Microscope 510 (LSM510, Carl Zeiss, Inc., USA). Under 40× and 100× objective magnifications, images were collapsed from confocal stacks of 1.0 μm optical slices. Under 25× objective magnifications, images were collapsed from confocal stacks of 2.0 μm optical slices. Under 10× objective magnifications, images were collapsed from confocal stacks of 12.5 μm optical slices. ImageJ software was used to remove other signals outside of the mushroom bodies, as the background noise in GRASP is strong.

Larval olfactory learning assays

A single odor learning paradigm was slightly modified from previous publication^17,18. In brief, 25 to 50 third-instar larvae (92–96 h after egg laying) were trained on a 2.5% agar plate (100mm petri dish) covered with 2 mL of 1 M sucrose solution (SUC, Sigma, Cat. No. S1888) or 0.1% quinine hemisulfate solution (QUI, Sigma, Cat. No. 22640). Distilled water (DW) was used as a control. An odorant pentyl acetate (PA, 10μL, Sigma-Aldrich, CAT. No. 109584) was placed on a small piece of filter inside the lid. After 30 minutes, larvae were rinsed and transferred to the middle line of a new 2.5% agar plate. A small piece of filter paper with 2.5μL pentyl acetate was placed on one side of the plate, while distilled water on the other side. After 5 min, the numbers of larvae in the two semicircular areas were counted and the response index (R.I.) was calculated with the following equation (Figure 2g):

Naïve Olfactory Test

Larvae were transferred into the midline of test plates. 2.5μL of odorant were added on one side and distilled water on the other side. The number of larvae in two semicircular areas were counted and the R.I. was calculated after 5 min.

Naïve Gustatory Test

A petri dish with a median separator was used. Both sides were filled with 1% agar, with 2 mL of distilled water on the control side, and with 1 M sucrose (SUC) solution, or 0.1% quinine hemisulfate (QUI) solution on the test side. Twenty larvae were put on each side near the midline and allowed to move for 5 min. Gustatory R.I. was calculated using the larvae numbers on two sides⁹⁰.

Larval locomotion assay

Individual larvae were placed on the surface of a plate of 2.5% agar mixed with 1mL India ink. They were allowed to acclimate for 1 min, and then a video was recorded for 30 seconds using a Moticam3 digital camera (Motic) and Motic Images Plus 2.0 software. The video was analyzed by the MTrack2 plug-in (from http://valelab.ucsf.edu/∼nico/IJplugins/MTrack2.html) in ImageJ. The path was recorded; scores were quantified as the length traveled per minute as previously described⁹¹. As the locomotion speed of DAN-c1 homozygous was slow, DAN-c1 × WT was used as the control group.

Learning assays with thermogenetics

In learning assays with thermogenetics, 25 to 50 third-instar larvae (92–96 h after egg laying) were trained on a 2.5% agar plate (100mm petri dish) covered with 2 mL of 1 M SUC or 0.1% QUI. DW was used as a control. An odorant PA was placed on a small piece of filter inside the lid. Training plates were put in a water bath either under 22°C or 34°C. After 30 minutes, larvae were rinsed and transferred to the testing plate. After 5 min, the response index (R.I.) was calculated.

Learning assays with optogenetics

In learning assays with optogenetics, egg laying plates with 1mM ATR (all-trans retinal, Sigma, Cat. No. R2500) were used. All-trans retinal (ATR) is a necessary light-isomerizable chromophore for ChR2, which is not synthesized by Drosophila.^92,93 Around 50 third-instar larvae were trained in a 35mm petri dish with 2 mL of either 1 M SUC or 0.1% QH solutions. DW was used as a control. During training, an odorant was placed on a small piece of filter inside the lid. To activate channelrhodopsin2, a LED (Luxeon Rebel Color LEDs, 07040 PB000-D, wavelength 470 nm) with a power supply (GW Instek, Laboratory DC power supply Model GPS-1830D) was used. The intensity of the blue light was 25 mW, measured by a laser power meter (Sanwa, LP1). After being trained for 30 minutes, larvae were rinsed and transferred to the middle line of a 2.5% agar plate in 100 mm test plate. A small piece of filter paper with pentyl acetate was placed on one side of the plate, while distilled water on the other side. Then the number of larvae in the two semicircular areas were counted and the R.I. was calculated after 5 min.

Quantification of D2R knockdown

Quantification of the fluorescent intensity of D2R knockdown was performed as follows. TH signals were used to define dopaminergic neurons, and the mean fluorescent intensity of GFP in each neuron was calculated with subtraction of the background. The mean intensity of DM1 was divided by that of pPAM, and the value in the knockdown group was subsequently normalized with the control group.

Statistical Analysis

Information for statistical analysis is provided in figure legends.

Author contributions

Qi. C.: Conceptualization, Methodology, Formal analysis, Investigation, Writing – Original Draft, Visualization, Qian. C.: Writing – Review & Editing, Steijvers. E.: Writing – Review & Editing Colvin. R.: Writing – Review & Editing Lee. D.: Conceptualization, Writing – Review & Editing, Supervision, Funding acquisition

Acknowledgements

This work was partially supported by an NIH grant (AG065925) and an International Collaboration Grant from Korea Institute of Science & Technology (Brain Science Institute), Seoul, Korea. CQ was a recipient of the SEA and GSR awards from Ohio University. We thank Dr. J Hirsh (University of Virginia), Dr. M. Wu (Johns Hopkins University), Dr. M. Zlatic and Dr. C. Eschbach (HHMI Janelia Research Campus), Dr. M. Gallio (Northwestern University), Dr. A. Kopin (Tufts-New England Medical Center), Dr. S. Tanda (Ohio University), Dr. B. Condron (University of Virginia), Dr. T. Kitamoto (University of Iowa) for their kind gift strains.

Declaration of Interest

The authors declare no competing interests.

Supplemental information

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.

Figure S7.

Supplementary Tables

Table S1.

Table S2.

Table S3.

Table S4.