It has been shown that neurotransmitter output from α′β′ neurons is critical for wLTM consolidation but not acquisition or retrieval (Shyu et al., 2017), which led us to ask whether there are time-dependent requirements of neurotransmission in α′β′-related neurons during wLTM consolidation. To address this, we performed acute manipulation of neurotransmitter output by expressing temperature-sensitive shibire (shi ^ts ) allele in specific neurons that inhibits dynamin-mediated endocytosis for recycling of neurotransmitters at restrictive temperatures (Dubnau et al., 2001; Kitamoto, 2001; McGuire et al., 2001). Results showed that blocking neurotransmitter output from α′β′ neurons (via VT30604-GAL4) during the first 4-h period right after training disrupts wLTM (Figures 1A,B). However, blocking neurotransmitter output in α′β′ neurons at– 4–12 or 12–20 h after training did not affect wLTM (Figure 1B).

Since neurotransmission from α′β′ neurons is specifically required during the early consolidation period (0–4 h after training) of wLTM (Figures 1A,B), we asked whether the α′β′-MBONs are also involved in wLTM consolidation. There are several different types of MBONs, whose dendrites are distributed in different α′β′ lobe subdomains (Aso et al., 2014b). Therefore, we individually tested the involvement of these α′β′-MBONs in wLTM consolidation. MBON-β′1 was labeled using MB057B-GAL4, MBON-γ3β′1 was labeled using MB083C-GAL4, MBON-γ5β′2 was labeled using MB011B-GAL4, MBON-α′3 was labeled using MB027B-GAL4, MBON-α′1α′3 was labeled using MB543B-GAL4, MBON-β2β′2a was labeled using VT0765-GAL4, MBON-γ2α′1 was labeled using MB051B-GAL4, MBON-α′2 was labeled using MB091C-GAL4, and MBON-β′2 was labeled using VT41043-GAL4 (Figures 1C–K). Using shi ^ts to block the neurotransmitter output from these α′β′-MBONs individually, we found that blocking neurotransmission only in MBON-γ3β′1, MBON-α′2, or MBON-α′1α′3 during the 4-h period right after training disrupts wLTM (Figure 1L). We further asked whether neurotransmitter output from MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 is required during training (acquisition) or testing (retrieval). By genetically expressing shi ^ts , we found that neurotransmission in MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 is required only for wLTM consolidation and not for acquisition or retrieval (Figure 2).

To dissect the role of MBONs during the initial 4-h period after training, we genetically expressed the shi ^ts transgene in MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 and performed behavioral assays. All flies were raised and trained at 23°C, maintained at 32°C for 2 h immediately after the training, and shifted back to 23°C for another 22 h and tested. We observed that blocking neurotransmitter output in MBON-γ3β′1 and MBON-α′1α′3 during the 0–2 h period after training disrupted wLTM (Figure 3A). We further investigated the 2–4 h period post-training. The flies were trained at 23°C, maintained at 23°C for 2 h immediately after the training, shifted back to 32°C for 2 h, and again shifted back to 23°C for another 20 h and tested. We observed that blocking neurotransmitter output in MBON-α′2 and MBON-α′1α′3 during the 2–4 h period after training disrupted wLTM (Figure 3B).

MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 receive inputs not only from α′β′ neurons but may also receive inputs from MB modulatory neurons. It has been shown that MB modulatory neurons, the DPM neurons that project to and innervate the MB lobes, play a role in mediating the synaptic activity of MBONs. Previous studies have suggested that DPM neurons are critical for olfactory associative memories, including shock-punishment and sugar-reward conditioning (Waddell et al., 2000; Keene et al., 2004; Yu et al., 2005; Keene et al., 2006; Wu et al., 2011). It has also been reported that DPM neurons release the neurotransmitter 5HT (Lee et al., 2011), which led us to ask whether 5HT synthesis in DPM neurons affects wLTM. In the 5HT biosynthesis pathway, L-tryptophan is first hydroxylated by the enzyme tryptophan hydroxylase (Trh) to 5-hydroxytryptophan, which is then decarboxylated by the enzyme dopa decarboxylase (Ddc) to 5HT (Blenau and Thamm, 2011). Our results showed that RNAi-mediated silencing of Trh and Ddc using the GAL4/UAS binary system inhibits 5HT biosynthesis in DPM neurons (via VT64246-GAL4) and disrupts wLTM (Figures 4A,B). However, inhibiting 5HT biosynthesis in MB neurons did not affect wLTM (Supplementary Figure S1). We further analyzed the knockdown efficiency of UAS-Trh ^RNAi ; UAS-Ddc ^RNAi in VT64246-GAL4 line. We observed that VT64246-GAL4 > UAS-Ddc ^RNAi ;;UAS-Trh ^RNAi flies showed over 50% reduction of anti-5HT antibody immunoreactive signals in both soma and fibers of DPM neurons as compared to the control groups (Figures 4C–E). To exclude the developmental effects of inhibiting 5HT biosynthesis in DPM neurons, we next tested the flies with inducible knockdown of Ddc and Trh genes using tub-GAL80 ^ts in the VT64246-GAL4 line. Our results showed that adult-stage-specific knockdown of Ddc and Trh in DPM neurons also impaired wLTM (Figure 4F).

Since wLTM formation requires 5HT biosynthesis in DPM neurons, we asked that neurotransmission in DPM neurons is involved in which memory phases. We genetically expressed shi ^ts transgene in DPM neurons and all transgenic flies were maintained at 23°C and shifted to 32°C to inhibit neurotransmitter output during training (acquisition), trial interval (consolidation), or testing (retrieval). We found that blocking neurotransmission in DPM neurons during consolidation impaired wLTM (Figure 5A). However, blocking neurotransmission in DPM neurons during acquisition or retrieval did not affect wLTM (Figure 5A). Since MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 showed different time-dependent requirements for wLTM (Figure 3), we asked that in which period during memory consolidation DPM neurons are involved. The flies were trained at 23°C and then maintained at 32°C for 0–2 h or 2–4 h right after the training, and shifted back to 23°C for the resting interval and tested. We found that blocking neurotransmitter output in DPM neurons during the 0–2 h or 2–4 h period disrupted wLTM (Figure 5B). Taken these data together suggest that 5HT neurotransmitter output from DPM neurons is critical for wLTM consolidation (Figures 4, 5).

Considering the gap junctions between DPM and anterior paired lateral (APL) neurons are critical for aversive anesthesia-sensitive memory in Drosophila (Wu et al., 2011), which led us to ask whether inhibiting neurotransmission during memory consolidation in APL neurons affects wLTM. Blocking neurotransmission in APL neurons during the 0–4 h period after training did not affect the wLTM, implying that neurotransmission in APL neurons is not critical for wLTM consolidation (Supplementary Figure S2). It is possible that APL neurons are involved in other memory phases rather than wLTM consolidation.

The dendritic regions of MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 are located in the horizontal and vertical lobes of α′β′ neurons. Considering that axons of DPM neurons are highly innervated into the horizontal and vertical lobes of α′β′ neurons, we investigated the connectivity of DPM neurons with these three MBONs. We utilized the GRASP tool by expressing two split-GFP partners, spGFP_1-10 and spGFP₁₁, to check whether these three types of MBONs are anatomically connected to DPM neurons. Our previous study showed that L0111-LexA could label the DPM neurons (Wu et al., 2011) therefore, we chose L0111-LexA to drive the expression of spGFP ₁₁ and Brp::mCherry (labels the axons of DPM neurons). In flies carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB083C-GAL4 > UAS-spGFP _1-10 , we observed GRASP signals in β′1 and γ3 subregions of MB horizontal lobes (Figures 6A–C). In flies carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB091C-GAL4 > UAS-spGFP _1-10 , we observed GRASP signals specifically in the α′2 region of the MB vertical lobes (Figures 6D–F). Finally, in flies carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB543B-GAL4 > UAS-spGFP _1-10 , we observed GRASP signals mainly in α′1 and α′3, along with some signals in α′2 subregions of the MB vertical lobes (Figures 6G–I).

The axons of DPM neurons highly innervate the MB lobe (especially α′β′ lobes), whereas the dendrites of MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 are located in α′β′ lobes. Our GRASP results showed the connectivity between DPM neurons and the three MBONs. DPM neurons showed overlapping time period requirements with MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3. We asked whether 5HT release from DPM neurons affects MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3 activity during wLTM consolidation. To address this, we suppressed 5HT receptor expression in these three MBONs and assessed wLTM formation. Several 5HT receptor genes have been identified in the Drosophila genome including 5HT1A, 5HT1B, 5HT2A, 5HT2B, and 5HT7 (Majeed et al., 2014). Using the GAL4/UAS binary system, we expressed RNAi against each 5HT receptor gene (Supplementary Figure S3) in MBON-γ3β′1, MBON-α′2, and MBON-α′1α′3, and performed behavioral assays. We found that silencing of 5HT1A, 5HT2A, or 5HT7 in MBON-γ3β′1 (via MB083C-GAL4) impaired wLTM (Figure 7A). Moreover, silencing of 5HT1A, 5HT1B, 5HT2A, or 5HT2B in MBON-α′1α′3 (via MB543B-GAL4) also impaired wLTM (Figure 7B). Finally, silencing of 5HT1B or 5HT2A in MBON-α′2 (via MB091C-GAL4) impaired wLTM (Figure 7C). All the memory-impaired flies showed normal odor avoidance and water preference in the thirsty state, suggesting that memory impairment was not caused by the deficiency of sensory inputs (Supplementary Figure S4).

All flies were reared on standard cornmeal food at 25°C and 60% relative humidity on a 12 h:12 h light-dark cycle. R13F02-GAL4, VT30604-GAL4, VT0765-GAL4, VT41043-GAL4, VT64246-GAL4, APL-GAL4, L0111-LexA, tub-GAL80 ^ts , UAS-shi ^ts , and UAS-mCD8:GFP;UAS-mCD8:GFP fly strains have been used in our previous studies (Wu et al., 2013; Shih et al., 2015; Wu et al., 2015; Yang et al., 2016; Shyu et al., 2017; Wu et al., 2018; Shyu et al., 2019; Lee et al., 2020). The lexAop-spGFP ₁₁ , UAS-spGFP _1-10 , lexAop-Brp::mCherry fly strains were obtained from Tzu-Yang Lin. The UAS-DDC ^RNAi and UAS-Trh ^RNAi fly strains were obtained from Ann-Shyn Chiang (Lee et al., 2011). The UAS-GAD ^RNAi (v32344) fly strain was obtained from Vienna Drosophila RNAi Center. MB057B-GAL4, MB083C-GAL4, MB011B-GAL4, MB027B-GAL4, MB543B-GAL4, MB051B-GAL4, MB091C-GAL4, UAS-5HT1A ^RNAi (TRiP.JF01852), UAS-5HT1B ^RNAi (TRiP.JF01851), UAS-5HT2A ^RNAi (TRiP.JF02157), UAS-5HT2B ^RNAi (TRiP.HMJ22882), UAS-5HT7 ^RNAi (TRiP.JF02576), UAS-ChAT ^RNAi (TRiP.JF01877), and UAS-VAChT ^RNAi (TRiP.JF02764) fly strains were obtained from the Bloomington Drosophila Stock Center.

The fly brains were dissected in phosphate-buffered saline (PBS) solution and immediately transferred to 4% paraformaldehyde (PFA) for fixation for 20 min. Fixed brain samples were incubated in penetration and blocking buffer (PBS containing 2% Triton X-100 and 10% normal goat serum) for 2 h. During the penetration and blocking period, the brain samples were subjected to a degassing procedure. Brains were then incubated in the dilution buffer (PBS containing 0.25% Triton X-100, 1% normal goat serum) containing mouse 4F3 anti-discs large (DLG) monoclonal antibody (1:10, AB 528203, Developmental Studies Hybridoma Bank, University of Iowa) at 25°C for 24 h. After incubation, the tissues were washed in PBS containing 1% Triton X-100 (PBS-T) three times (each time 15 min), and incubated with biotinylated goat anti-mouse IgG (1:200, 31,800, Thermo Fisher Scientific) at 25°C for 24 h. Next, the brain samples were washed and incubated with Alexa Fluor 635 streptavidin (1:500, S32364, Thermo Fisher Scientific) at 25°C for 24 h. After intensive washing in PBS-T for three times, the brain samples were cleared and mounted in FocusClear (FC-101, CelExplorer).

Fly brains were immunostained with rabbit anti-5HT polyclonal antibody (1:200, S5545, Sigma-Aldrich) at 25°C for 24 h. After three washes in PBS-T, the samples were incubated in biotinylated goat anti-rabbit IgG (1:200, 31,822, Thermo Fisher Scientific) at 25°C for 1 day. The brain samples were then washed and incubated in Alexa Fluor 635 streptavidin (1:500, S32364, Thermo Fisher Scientific) at 25°C for 1 day. After three-time extensive washing, the brain samples were cleared and mounted in FocusClear (FC-101, CelExplorer). Brain images were obtained using a Zeiss LSM 700 confocal microscope under the same confocal settings for each brain sample, and the images were further analyzed using ImageJ. Single optical sections were used to calculate the average intensity values per voxel of the 5HT immunopositive signals in the soma and fibers of DPM neurons, and the signals were normalized to the GFP intensity.

The fly brain samples were imaged under the Zeiss LSM 700 confocal microscope with a ×63 Plan-Apochromat oil-immersion objective lens (N.A. value, 1.4; working distance, 170 μm) or a ×40 C-Apochromat water-immersion objective lens (N.A. value, 1.2; working distance, 220 μm). The confocal pinhole (optical section) was set at 1.5 μm when using ×63 objective lens or 2 μm when imaging thorough ×40 objective lens. The brain image stacks were processed using the ZEN software.

Adult fly brains carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB083C-GAL4 > UAS-spGFP _1-10 were used for DPM and MBON-γ3β′1 connectivity assay. Adult fly brains carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB091C-GAL4 > UAS-spGFP _1-10 were used for DPM and MBON-α′2 connectivity assay. Adult fly brains carrying L0111-LexA > lexAop-spGFP ₁₁ , lexAop-Brp::mCherry, and MB543B-GAL4 > UAS-spGFP _1-10 were used for DPM and MBON-α′1α′3 connectivity assay. All the brain sample were dissected in PBS solution and immediately transferred to 4% PFA for fixation and degassing procedure. The fixed brain samples were than cleared and mounted in FocusClear and imaged under Zeiss LSM 700 confocal microscope by using 488 and 555 nm lasers for GFP and mCherry excitation respectively.

Flies were deprived of water by keeping them in a glass milk bottle containing a 6 cm × 3 cm piece of dry sucrose-soaked filter paper (the filter paper was soaked in saturated sucrose solution and allowed to dry before use) at 23°C and 20–30% humidity for 16 h before water conditioning. Approximately 50 water-deprived flies were transferred to the plastic training tube of the T-maze (TM-101, CelExplorer) and provided with a stream of relatively odorless room air for 1 min. The flies were exposed to the first odor for 2 min [unconditioned stimulus, CS–: 3-octanol (OCT) or 4-methylcyclohexanol (MCH)] in a plastic tube lined with dry filter paper, followed by 1 min of room air. The flies were then transferred to another plastic tube containing a water-soaked filter paper and exposed to a second odor for 2 min (conditioned stimulus, CS+: MCH or OCT). Finally, the flies were transferred to a clean training tube and exposed to fresh room air for 1 min. The trained flies were kept in a plastic vial containing a 1.5 cm × 3 cm piece of dried sucrose-soaked filter paper during the 24 h interval between training and testing. In the testing phase, flies were presented with a choice between CS+ and CS–odors in a T-maze for 2 min. The performance index (PI) was calculated as the number of flies running toward the CS + odor minus the number of flies running toward the CS–odor, divided by the total number of trained flies and multiplied by 100. For calculating individual PI, naive flies were first trained by pairing water with OCT (CS+), and the index (PI_O) was calculated. Next, another group of naive flies was trained by pairing water with MCH (CS+), and the index (PI_M) was calculated. The individual PI was calculated as the average of the single PI_O and PI_M values.

Groups of approximately 50 naive flies were given a choice between either MCH or OCT versus ‘fresh’ room air for 2 min in the T-maze for the odor avoidance assay. The odor avoidance index was calculated as the number of flies in the fresh room air tube minus the number of flies in the MCH or OCT tube, divided by the total number of flies, and multiplied by 100. Groups of approximately 50 naive water-deprived flies were given 2 min to choose between tubes containing a water-soaked filter paper or a dry filter paper in T-maze in the water preference assay. The water preference index was calculated as the number of flies in the tube containing the water-soaked filter paper minus the number of flies in the tube containing the dry filter paper, divided by the total number of flies, and multiplied by 100.

The efficiency of gene silencing in each 5HT receptor RNAi line from collections was verified with qPCR. Flies for qPCR experiments were generated by crossing virgin elav-GAL4 flies to either wild-type males or the various male UAS-5HT receptor ^RNAi flies. The RNA from isolated heads of adult flies was extracted by the TRIzol Reagent (T9424, Sigma-Aldrich). The extracted RNA was used to synthesize first strand of cDNA with SuperScript^TMIII First-Strand Synthesis SuperMix Kit (18080400, Thermo Fisher Scientific). The expression levels of mRNA were quantified with SYBR Green PCR Master Mix on a StepOnePlus System.

All the raw data were analyzed parametrically using the Prism 5.0 software (GraphPad). Comparison between more than two groups was performed using one-way analysis of variance (ANOVA) followed by the Tukey’s multiple comparison test. Comparison between two groups was performed using the paired t-test. Statistical significance was set at p < 0.05. Data are presented as mean ± standard error of the mean (SEM).