Octopamine signaling from clock neurons plays dual roles in Drosophila long-term memory

In Drosophila, LNds consist of six clock neurons per hemisphere [11]. We previously identified two GAL4 lines—R18H11 and R78G02—in which GAL4 is expressed in subsets of these neurons [9]. R18H11 drives GAL4 expression in two LNds per hemisphere, whereas R78G02 targets two to three [9]. However, in both GAL4 lines, GAL4 also labels brain neurons other than LNds. To achieve more specific labeling, we generated an LNd-specific split-GAL4 line (hereafter referred to as LNd-split-GAL4) using a combination of the R18H11-p65.AD and R78G02-GAL4.DBD lines. In LNd-split-GAL4, split GAL4 is expressed in only two LNds per hemisphere (Fig 1C) [9]. Using this line and UAS-shi^ts1, we previously confirmed that the disruption of neurotransmission in LNd-split-GAL4-positive neurons during the LTM maintenance phase impairs LTM [9]. To investigate whether octopamine or acetylcholine (ACh) in LNds contributes to LTM, we knocked down genes involved in their synthesis (Tdc2, Tbh, and ChAT) using LNd-split-GAL4 in combination with multiple UAS-RNAi lines. The efficacy of Tdc2 RNAi and Tbh RNAi was assessed by qRT-PCR analysis using a pan-neuronal driver, nSyb-GAL4. nSyb-GAL4/UAS-Tdc2 RNAi males did not show a reduction in Tdc2 expression levels (S1A Fig). However, the co-expression of UAS-Dicer2 achieved ~90% knockdown relative to GAL4 control flies (S1B Fig). nSyb-GAL4/UAS-Tbh RNAi #1 and nSyb-GAL4/UAS-Tbh RNAi #2 flies showed about a 70% reduction in Tbh expression levels compared with GAL4 control flies (S1C, S1D Fig). Unlike Tbh and Tdc2 knockdown flies, nSyb-GAL4/UAS-ChAT RNAi flies were lethal, precluding qRT-PCR analysis. This lethality is consistent with the finding that ChAT mutant flies are also lethal (https://flybase.org/reports/FBgn0000303.htm↗), suggesting that ChAT knockdown using UAS-ChAT RNAi reduces cholinergic function.

In LNd-split-GAL4 and UAS control males, the courtship index (CI), an indicator of male courtship activity, of conditioned males was significantly lower than that of naive males on day 5 after 7 h conditioning (Fig 1D). Thus, in these control flies, we detected courtship memory on day 5 after 7 h conditioning (hereafter referred to as ‘5 d memory’). However, in the F₁ hybrid between LNd-split-GAL4 and either UAS-Dicer2; UAS-Tdc2 RNAi, UAS-Tbh RNAi #1, or UAS-Tbh RNAi #2, no significant differences in CI were detected between naive and conditioned males (Fig 1D). To quantify courtship memory, the memory index (MI) was calculated as described in the Methods section. In these F₁ hybrid males, MI was significantly lower than those in LNd-split-GAL4 and UAS control males, indicating that the LNd-specific knockdown of Tdc2 and Tbh impaired LTM (Fig 1D). Tdc2 encodes a tyrosine decarboxylase that converts tyrosine into tyramine, a precursor of octopamine, whereas Tbh catalyzes the conversion of tyramine into octopamine. Thus, these results indicate that reduced octopamine production in two pairs of LNds impairs LTM. However, Tbh knockdown in LNds did not affect courtship STM formation (S2A Fig). Unlike Tdc2 and Tbh knockdown in LNds, ChAT knockdown in LNds had little effect on 5 d memory after 7 h conditioning (Fig 1D).

As shown in Fig 1D, naive CI in LNd-split-GAL4/UAS-Tbh RNAi #1 (mean naive CI = 62%) was significantly lower than that in LNd-split-GAL4 control males (Steel–Dwass test, q = 3.338, P < 0.020), despite the absence of a significant difference between LNd-split-GAL4/+ and UAS-Tbh RNAi #1/ + males (Steel–Dwass test, q = 1.60, P = 0.795). We previously confirmed that the wild-type strains with a mean naive CI of less than 40% exhibit courtship suppression on day 5 after 7 h conditioning [6]. Thus, it is unlikely that the difficulty in detecting courtship memory in LNd-split-GAL4/UAS-Tbh RNAi #1 males is solely due to a reduction in naive CI. Unlike LNd-split-GAL4/UAS-Tbh RNAi #1 males, no significant difference was detected between LNd-split-GAL4 control and LNd-split-GAL4/UAS-Tbh RNAi #2 males (Steel–Dwass test, q = 1.31, P = 0.926). These results suggest that the reduced courtship activity in LNd-split-GAL4/UAS-Tbh RNAi #1 naive males may not be due to the Tbh knockdown itself, but rather by genetic background effects.

In Drosophila, sleep is required for the consolidation of courtship LTM [28,29]. To confirm whether LTM impairments induced by LNd-specific Tbh knockdown result from altered sleep, we measured the sleep amount in light:Dark (LD) cycles using LNd-split-GAL4/UAS-Tbh RNAi #1 males. LNd-split-GAL4/+ and UAS-Tbh RNAi #1/ + males were used as the control. For the total sleep amount during the day and night, no significant differences were detected between LNd-split-GAL4/UAS-Tbh RNAi #1 males and UAS-Tbh RNAi #1/ + control males (S3A-S3C Fig), although significant differences were detected between LNd-split-GAL4/UAS-Tbh RNAi #1 males and LNd-split-GAL4/ + control males (S3A-S3C Fig). We also assessed two additional sleep parameters: sleep bout number and sleep bout duration. The sleep bout number during the day in Tbh knockdown males was higher than that in GAL4 control males and lower than that in UAS control males (S3D Fig). There was no significant difference in sleep bout number during the night between the GAL4 control males and Tbh knockdown males (S3D Fig). The sleep bout duration during the day in Tbh knockdown males was shorter than that in GAL4 control males and longer than that in UAS control males (S3E Fig). The sleep bout duration during the night did not differ significantly between the Tbh knockdown males and the GAL4 control males. However, the waking activity index was elevated in Tbh knockdown males only during the day (S3F Fig). Thus, Tbh knockdown has little effect on the amount and quality of sleep, but it appears to selectively increase spontaneous activity during daytime wakefulness.

We next examined whether LNd-specific Tbh knockdown affects circadian behavioral rhythms under constant darkness. No significant differences were detected between LNd-split-GAL4/UAS-Tbh RNAi #1 and control males in the proportion of rhythmic individuals (% rhythmic) or circadian periods (S4 Fig). Taken together, it is unlikely that the memory impairment observed in Tbh knockdown flies results from alterations in circadian rhythms.

Are Tdc2 and Tbh actually expressed in LNds? Previous chronobiological studies have shown that Tdc2 is present in a subset of LNds, as revealed by immunohistochemical staining with an anti-Tdc2 antibody [30]. To determine whether Tdc2 is expressed in the specific subset labeled by LNd-split-GAL4, we performed immunohistochemical staining with an anti-Tdc2 antibody, which showed Tdc2 signals in two LNd-split-GAL4-positive neurons per hemisphere (Fig 1E). Unlike Tdc2, antibodies for Tbh were unavailable. To examine Tbh expression in LNds, we utilized a fly-TransgeneOme (fTRG) line (VDRC ID: 318242), which expresses a C-terminal superfolder GFP (sGFP)-tagged fusion protein under the control of the endogenous Tbh promoter [31]. In this study, we used a fTRG 318242 line (hereafter referred to as ‘Tbh-GFP’) to visualize Tbh-expressing neurons. Since the circadian clock gene period is expressed in all six LNds [32], we observed all LNds in males homozygous for Tbh-GFP using an anti-Per antibody. The co-expression of Per and Tbh-GFP was detected in all LNds (Fig 1F). Therefore, Tbh is likely expressed in LNd-split-GAL4-positive cells.

To investigate whether octopamine production in LNds is required for the consolidation or maintenance of LTM, we performed temporal Tbh knockdown experiments using tub-GAL80^ts [33]. The tub-GAL80^ts transgene encodes a ubiquitously expressed GAL4 repressor that functions at the permissive temperature (PT, 25 °C) but is inactive at the restrictive temperature (RT, 30 °C). However, since GAL80^ts has no effect on split-GAL4-driven gene expression, we used tub-GAL80^ts/UAS-Tbh RNAi #1; R78G02/ + males in this experiment. tub-GAL80^ts/ + ; R78G02-GAL4/+ and UAS-Tbh RNAi #1/ + males were used as the control. R78G02 expresses GAL4 in three LNds and the 5th s-LNv in each hemisphere [34]. In Drosophila, the ion transport peptide (ITP) is expressed in one LNd and the 5th s-LNv per brain hemisphere [35]. To confirm this expression pattern in R78G02, we performed immunostaining with an anti-ITP antibody. In R78G02/UAS-mCD8::GFP males, the one LNd and the 5th s-LNv were ITP-positive in each hemisphere (Fig 2A). We next examined whether Tdc2 is expressed in R78G02-positive neurons using an anti-Tdc2 antibody. Tdc2 signals were detected in three LNds and the 5th s-LNv in R78G02-positive cells in each hemibrain (Fig 2B and 2C). We further confirmed whether Tbh is expressed in GAL4-expressing cells in R78G02 using Tbh-GFP. Tbh-GFP signals were detected in LNds and the 5th s-LNv in R78G02-positive cells in each hemibrain (S5A and S5B Fig). Furthermore, they were detected in at least two pairs of R78G02-positive non-clock neurons located in the posterior–dorsal region (S5C Fig).

To knock down Tbh in R78G02-positive cells during the memory consolidation, maintenance, or recall phase, we used tub-GAL80^ts/UAS-Tbh RNAi #1; R78G02-GAL4/ + males. tub-GAL80^ts/ + ; R78G02-GAL4/+ and UAS-Tbh RNAi #1/ + males were used as the control. For the temporal knockdown of Tbh, the temperature was increased to RT (30 °C) for 24 h during three different experimental periods: starting at 24 h before the end of conditioning (learning/consolidation phase), 48–72 h after conditioning initiation (maintenance phase), and 24 h before the test initiation (recall phase). In all genotypes, LTM was intact at PT (25 °C) (Fig 2D). On the other hand, the temporal knockdown of Tbh during the learning/consolidation phase or the maintenance phase significantly impaired LTM (Fig 2E and 2F), whereas that during the recall phase did not (Fig 2G). These findings suggest that octopamine signaling from R78G02-positive Tbh-expressing neurons—including a subset of LNds, the 5th s-LNv, and non-clock neurons—plays a critical role in the consolidation and maintenance, but not recall, of courtship LTM.

To determine whether neurotransmitter release from R78G02-positive cells is required for LTM consolidation and maintenance, we experimented with the temporal disruption of neurotransmission using R78G02/UAS-shi^ts1 males. R78G02/UAS-shi^ts1 males showed LTM at PT (25 °C) as was observed in GAL4 and UAS control males (Fig 3A). When the temperature was raised to RT (30 °C) during 7 h conditioning, LTM was impaired in R78G02/UAS-shi^ts1 males (Fig 3B). In addition, when males were kept at 30 °C for 39–63 h after the end of conditioning, LTM was also impaired (Fig 3C). However, the disruption of synaptic transmission from LNds during the test had little effect on LTM (Fig 3D). Consistent with the experimental results of Tbh knockdown, these results indicate that disruption of neurotransmission from R78G02-positive cells impairs LTM consolidation and maintenance. Thus, these results also support our idea that octopamine release from R78G02-positive cells modulates both LTM consolidation and maintenance.

We previously found that the disruption of neurotransmission from LNd-split-GAL4-positive cells impairs only LTM maintenance without affecting LTM consolidation [9]. On the other hand, Tbh knockdown or blocking neurotransmission in R78G02-positive cells impairs both LTM consolidation and maintenance. Notably, LNd-split-GAL4 is expressed in only two LNds in each hemibrain, whereas R78G02-positive LNds include an ITP-positive cell in each hemibrain (Fig 2A). We confirmed that LNd-split-GAL4-positive cells do not express ITP (Fig 4A). Moreover, R78G02-positive cells include a pair of ITP-positive 5th s-LNvs (Fig 2A). The main differences between GAL4-expressing clock neurons in LNd-split-GAL4 and R78G02 can be summarized in Fig 1C. Since ITP-positive LNds and 5th s-LNv are present in R78G02 but absent in LNd-split-GAL4 (Fig 1C), we hypothesized that octopamine in these ITP-positive clock neurons might contribute to LTM consolidation. To examine this possibility, we used a GAL4 line, R54D11. In R54D11, GAL4 is mainly expressed in one LNd and the 5th s-LNv (Fig 1C) [36]. We confirmed that ITP is expressed in R54D11-positive LNds and the 5th s-LNv (Fig 4B). However, the knockdown of either Tbh or Tdc2 in R54D11-positive cells did not affect LTM (Fig 4C and 4D). Thus, it seems unlikely that octopamine release from ITP-positive clock neurons affects LTM.

In summary, octopamine synthesis in the two LNds, excluding ITP-positive LNds, among R78G02-positive cells appears to be necessary for LTM maintenance. On the other hand, octopamine synthesis in R78G02-positive non-clock neurons may be involved in LTM consolidation.

We previously identified that neurotransmission from Pdf neurons is required for LTM consolidation [5]. Tdc2 is also expressed in Pdf neurons [30]. Thus, we investigated whether Tdc2 and Tbh in Pdf neurons are involved in LTM consolidation. First, we confirmed whether Tdc2 is expressed in Pdf-positive l-LNvs and s-LNvs using an anti-Tdc2 antibody (Fig 5A and 5B). In the case of l-LNvs, all 20 brains examined showed Tdc2 signals in four l-LNvs per hemisphere (Fig 5A). In contrast, among the 18 brains analyzed for s-LNvs, although Tdc2 signals were detected in only a single s-LNv in one hemisphere of one brain, they were absent in the s-LNvs of nearly all other brains (Fig 5B). Furthermore, to confirm whether Tbh is also expressed in Pdf neurons, Pdf-positive l-LNvs and s-LNvs in Tbh-GFP males were visualized using an anti-Pdf antibody. In Tbh-GFP males, strong GFP signals were detected in all l-LNvs, whereas weak GFP signals were detected in s-LNvs (Fig 5C and 5D). Taken together, these results suggest that octopamine is synthesized in Pdf neurons, and that synthesis in l-LNvs among Pdf neurons is particularly prominent.

We next investigated the effects of Tdc2 and Tbh knockdown on LTM. In males with knockdown of Tdc2 or Tbh in Pdf neurons, MI was significantly reduced compared with those in GAL4 and UAS control males (Fig 5E–5G). In Fig 5F and 5G, the naïve CI in Pdf-GAL4/UAS-Tbh RNAi #1 and Pdf-GAL4/UAS-Tbh RNAi #2 males was significantly lower than that in GAL or UAS control (Pdf-GAL4/ + vs Pdf-GAL4/UAS-Tbh RNAi #1, Steel–Dwass test, q = 3.90, P < 0.001; UAS-Tbh RNAi #1/ + vs Pdf-GAL4/UAS-Tbh RNAi #1, Steel–Dwass test, q = 4.77, P < 0.001; Pdf-GAL4/ + vs Pdf-GAL4/UAS-Tbh RNAi #2, Steel–Dwass test, q = 2.60, P < 0.05; UAS-Tbh RNAi #2/ + vs Pdf-GAL4/UAS-Tbh RNAi #2, Steel–Dwass test, q = 3.14, P < 0.01). These findings suggest that octopamine in Pdf neurons plays a direct role in regulating male courtship activity. However, Tbh knockdown in Pdf neurons did not affect courtship STM (S2B Fig), suggesting that the reduction in CI by Tbh knockdown does not interfere with courtship STM formation.

To test whether octopamine production in Pdf neurons is required for LTM consolidation, we performed temporal knockdown experiments using Pdf-GAL4/UAS-Tbh RNAi #1; tub-GAL80^ts/ + males. Pdf-GAL4/ + ; tub-GAL80^ts/+ and UAS-Tbh RNAi #1/ + males were used as the control. In the experiments, the PT was 22°C outside the test phase and 25°C during the test phase. In all genotypes, LTM was intact at PT on day 5 after 7 h conditioning (Fig 6A). When males were kept at RT (30 °C) during the period from 17 h before the initiation of conditioning to the end of conditioning, LTM was impaired (Fig 6B). In addition, no significant differences in the CI of naive males were detected between GAL4 control and Tbh knockdown males (Pdf-GAL4/ + ; tub-GAL80^ts/ + vs Pdf-GAL4/UAS-Tbh RNAi #1; tub-GAL80^ts/ + ; Steel–Dwass test, q = 0.878, P = 0.654; UAS-Tbh RNAi #1/ + vs Pdf-GAL4/UAS-Tbh RNAi #1; tub-GAL80^ts/ + , Steel–Dwass test, q = 2.906, P < 0.05). However, raising the temperature to 30 °C during the maintenance phase (39–63 h after conditioning) had little effect on LTM (Fig 6C) and the CI of naive males (Pdf-GAL4/ + ; tub-GAL80^ts/ + vs Pdf-GAL4/UAS-Tbh RNAi #1; tub-GAL80^ts/ + ; Steel–Dwass test, q = 0.175, P = 0.983; UAS-Tbh RNAi #1/ + vs Pdf-GAL4/UAS-Tbh RNAi #1; tub-GAL80^ts/ + , Steel–Dwass test, q = 2.526, P < 0.05). Since we previously confirmed that neurotransmission from Pdf neurons does not affect LTM recall [5], we did not perform recall-phase-specific Tbh knockdown. Taken together with previous findings, these results suggest that octopamine signaling in Pdf neurons mainly contributes to LTM consolidation but not LTM maintenance or recall.

The CI of naive males in Pdf-GAL4/UAS-Tbh RNAi #1 was significantly lower than that in GAL or UAS control males (Fig 5). However, in the temporal knockdown experiments, no significant difference was detected between the CI of the Tbh knockdown males and that of the GAL4 control males. (Fig 6). Thus, the reduced courtship activity observed in Pdf-GAL4/UAS-Tbh RNAi #1 males is likely due to developmental defects caused by the suppression of Tbh expression.

Pdf-expressing l-LNvs are essential for light-dependent arousal in Drosophila [16]. Thus, we examined whether Tbh knockdown in Pdf neurons affects the sleep phenotype using Pdf-GAL4/UAS-Tbh RNAi #1 males. Pdf-GAL4/+ and UAS-Tbh RNAi #1/ + males were used as the control. Regarding the total sleep amount during the day, no significant differences were detected between Pdf-GAL4/UAS-Tbh RNAi #1 males and UAS-Tbh RNAi #1/ + control males (S6A and S6B Fig), whereas significant differences were detected between Pdf-GAL4/UAS-Tbh RNAi #1 males and Pdf-GAL4/ + control males (S6A and S6B Fig). Regarding the total sleep amount during the night, no significant differences were detected among the three genotypes (S6A and S6C Fig). Regarding both sleep bout number and sleep bout duration, no significant differences were detected between Tbh knockdown males and UAS control males (S6D and S6E Fig). Regarding the waking activity index, no significant differences were detected between Tbh knockdown males and UAS control males (S6F Fig). Thus, octopamine signaling in Pdf neurons does not appear to have a major impact on the sleep phenotype or spontaneous activity.

Huang et al. reported that Tdc2 knockdown in Pdf neurons induces arrhythmic locomotor activity [30]. However, it remains unclarified whether Tbh in Pdf neurons is also necessary for rhythmic locomotor activity. To address this possibility, we examined whether Tbh knockdown in Pdf neurons affects circadian behavioral rhythms in constant darkness. Unlike Tdc2 knockdown, Tbh knockdown in Pdf neurons did not affect the percentage of rhythmic flies or circadian periods (Fig 7A–7C). Thus, it is unlikely that octopamine in Pdf neurons affects circadian rhythms. However, about 60% of the 55 Tdc2 knockdown males had arrhythmic locomotor activity (Fig 7D–7G), indicating that Tdc2 in Pdf neurons is required for circadian behavioral rhythms. This finding is consistent with the previous finding [30]. Taken together, tyramine, rather than octopamine, seems to be the key neurotransmitter released from Pdf neurons in regulating circadian rhythms.

All flies were raised on glucose–yeast–cornmeal medium in 12:12 LD cycles at 25.0 ± 0.5°C (45%– 60% relative humidity). Virgin males and females were collected within 8 h after eclosion. The fly stocks used for this study were as follows: wild-type Canton- S (CS), R78G02 [40010, Bloomington Drosophila Stock Center (BDSC)], R18H11-AD (68852, BDSC), R78G02-DBD (69904, BDSC), UAS-mCD8::GFP (5137, BDSC), UAS-shi^ts1(a gift from Dr. Kitamoto), UAS-Tdc2 RNAi (25871, BDSC), UAS-Tbh RNAi #1 (67968, BDSC), UAS-Tbh RNAi #2 (76062, BDSC), UAS-ChAT RNAi (60028, BDSC), tub-GAL80^ts (7017, 7019, BDSC), UAS-Dicer2 (24650, BDSC), R54D11 (41279, BDSC), nSyb-GAL4 (51635, BDSC), UAS-mCherry.NLS (38425, BDSC), UAS-GFP.NLS (24775, BDSC), Pdf-GAL4 (6900, BDSC), and Tbh-GFP [318242, Vienna Drosophila Resource Center (VDRC)]. All lines except for UAS-Tdc2 RNAi, UAS-Tbh RNAi #1, UAS-Tbh RNAi #2, UAS-ChAT RNAi, and R54D11 were outcrossed for at least 6 generations to white¹¹¹⁸ flies with the CS genetic background. R18H11-AD and R78G02-DBD were used to produce an LNd-split-GAL4 line [9].

In this study, only UAS-Tdc2 RNAi was used in combination with UAS-Dicer2. Unlike UAS-Tdc2 RNAi, the TRiP RNAi lines generated using the VALIUM20 vector in this study (UAS-Tbh RNAi #1, UAS-Tbh RNAi #2, and UAS-ChAT RNAi) are expected to be effective even without being combined with UAS-Dicer2 (https://bdsc.indiana.edu/stocks/rnai/rnai_all.html↗). Thus, they were used without being combined with UAS-Dicer2.

The courtship conditioning assay was carried out as previously described [5,9]. For LTM, a virgin male (3–6 d old) was placed with a mated female (4–7 d old) in a conditioning chamber (15 mm diameter × 5 mm depth) containing food for 7 h either with (conditioned) or without (naive) a single premated female (7 h conditioning). After 7 h conditioning, only flies showing courtship behaviors toward the mated female were transferred to a glass tube with food (12 mm diameter × 75 mm depth) and kept in isolation for 5 d until the test. The test was performed using a freeze-killed virgin female in a test chamber (15 mm diameter × 3 mm depth). All procedures in the experiments were performed at 25 ± 1.0°C (45–60% relative humidity) except for the temperature shift experiments. The CI was used to quantify the male courtship behaviors of individual flies and was calculated manually. CI is defined as the percentage of time spent performing courtship behaviors during 10 min. We first measured CI in conditioned and naive males (CI_Conditioned and CI_Naive, respectively), and then the mean CI_Naive and mean CI_Conditioned were calculated. To quantify courtship memory as previously reported [5,8,9,21], MI was calculated using the following formula: MI = (mean CI_Naive - mean CI_Conditioned)/mean CI_Naive.

Neurotransmission was temporally disrupted as previously described [5]. Briefly, control males (GAL4/+ and UAS-shi^ts1/+) and shi^ts1-expressing males were used in the experiments. They were kept at PT (25 °C) until the temperature shift experiments. To disrupt neurotransmission in GAL4-expressing neurons during the memory consolidation, maintenance, or recall phase, the temperature was increased to RT (30 °C) during three experimental periods: 7 h conditioning (memory consolidation phase), 39–63 h after the end of conditioning (memory maintenance phase), and 30 min before the test began and until the test ended (recall phase).

Gene expression using the TARGET system [33] was temporally knocked down as previously described [5,8]. The transgene tub-GAL80^ts used in the TARGET system encodes a ubiquitously expressed, temperature-sensitive GAL4 repressor that is active at PT but not at RT. In Fig 2, the PT was 25°C. In Fig 6, the PT was 22°C outside the test phase and 25°C during the test phase. Using UAS-Tbh RNAi#1 combined with the TARGET system, we knocked down Tbh in GAL4-positive neurons at RT (30°C). In these experiments, we shifted PT to RT and vice versa during three experimental phases: 24 h before the end of conditioning, 39–63 h after conditioning, and 24 h before the end of test.

The effectiveness of Tdc2 RNAi and Tbh RNAi was confirmed by qRT-PCR using a pan-neural GAL4 line, nSyb-GAL4 as previously described [9]. TRizol (Invitrogen, USA) was used to collect total RNA from about 30 male fly heads in each genotype. cDNA was synthesized by the reverse transcription reaction using a PrimeScript RT reagent kit (RR047A, TAKARA, Japan). qRT-PCR was carried out using a LightCycler 96 (Roche, Switzerland) and the THUNDERBIRD SYBR qPCR Mix (QPS-201, TOYOBO, Japan). The primer sequences used for qRT-PCR were as follows: Tdc2-Forward, 5′- GAAGGGTTCCGACAAGCTGAATG- 3′; Tdc2-Reverse, 5′- TAGTCAATGTCCTCAGCGGTCG- 3′; Tbh-Forward, 5′- ACTACTATCCGGCCACCAAACTG- 3′; Tbh-Reverse, 5′- ATGCTCCGGTAATTGGACGACC- 3′; rp49-Forward,

5′- AAGATCGTGAAGAAGCGCAC- 3′; rp49- Reverse, 5′- TGTGCACCAGGAACTTCTTG- 3′. The expression level of each mRNA was normalized to that of rp49 mRNA. The average normalized mRNA expression levels in control flies were calculated using data from five independent assays.

The antibody against ITP was generated in guinea pigs by Scrum Inc. (Tokyo, Japan). A peptide corresponding to the C-terminal part of ITP (EMDKYNEWRDTL) was chemically synthesized and coupled at the N-terminus to keyhole limpet hemocyanin. The synthesized peptide was then used as an antigen to immunize guinea pigs by a conventional method. The specificity of the anti-ITP antibody was validated as previously reported [54].

Immunohistochemistry was performed as previously described with some modifications [9]. For Per staining, brains were stained with a rabbit anti-Per antibody (MBS610604, MyBioSource, 1/1000) followed by Alexa Fluor 568 anti-rabbit IgG (A11011, Thermo Fisher Scientific, USA) as the secondary antibody (1:500). For Tdc2 staining, brains were stained with a rabbit anti-Tdc2 antibody (ab128225, Abcam, 1/250 or pAB0822-P, Covalab, 1/250) followed by Alexa Fluor Plus 555 anti-rabbit IgG (A32732, Thermo Fisher Scientific, USA) as the secondary antibody (1:1000). For Pdf staining, brains were stained with a mouse anti-Pdf antibody (PDF C7-s, Developmental Studies Hybridoma Bank at the University of Iowa, 1:200) followed by Alexa Fluor 568 anti-mouse IgG (A11004, Thermo Fisher Scientific) as the secondary antibody (1:1000). For GFP staining, brains were stained with a chicken anti-GFP antibody (ab13970, Abcam, USA; 1:1000), followed by Alexa Fluor 488 anti-chicken IgG (A11039, Thermo Fisher Scientific, USA; 1:1000) as the secondary antibody. For ITP staining, brains were stained with a guinea pig anti-ITP antibody (1:20000), followed by Alexa Fluor 568 anti-guinea pig IgG (A-11075, Thermo Fisher Scientific, USA; 1:1000) or Alexa Fluor 647 anti-guinea pig IgG (A-21450, Thermo Fisher Scientific, USA; 1:1000) as the secondary antibody.

To investigate whether cells co-localizing with Tbh and ITP exist among R78G02-positive cells, UAS-mCherry.NLS/ + ; R78G02/Tbh-GFP males were used. Thus, we performed triple staining using an anti-DsRed antibody (632496, Clontech, 1/1000), an anti-GFP antibody (ab13970, Abcam, USA; 1:1000), and an anti-ITP antibody (1:20000). The following antibodies were used as secondary antibodies: Alexa Fluor Plus 555 anti-rabbit IgG (A32732, Thermo Fisher Scientific, USA), Alexa Fluor 488 anti-chicken IgG (A11039, Thermo Fisher Scientific, USA; 1:1000), and Alexa Fluor 647 anti-guinea pig IgG (A-21450, Thermo Fisher Scientific, USA; 1:1000). Adult male brains were fixed in PBS containing 4% formaldehyde for about 60 min at room temperature. After three washes in PBST (0.2% Triton X-100 in PBS), the brain samples were incubated for 1 h in 1% normal goat serum in PBST for blocking and then incubated with the three primary antibodies for 2 days at 4 °C. Next, they were incubated with the three secondary antibodies for 2 days at 4 °C after three washes in PBST. Fluorescence signals were observed under a confocal microscope (C2 or AX, Nikon, Japan).

Measurements and analysis of sleep were performed as previously described with some modifications [55]. The DAM system (Trikinetics) was used for sleep measurements. Sleep was measured for 3 days in 12-h L:12-h D (LD) cycles (lights on at 8:30) at 25 °C. We used 5–7-d-old males at the start of sleep measurements. Locomotor activity data were collected at 1-min intervals for 3 days and analyzed with a Microsoft Excel-based program as described previously [56]. Total sleep amount, sleep bout number, sleep bout durations, and waking activity index were analyzed for each 12-h period of LD or DD and averaged over 3 days for each condition.

Measurements and analysis of circadian behavioral rhythms were performed as previously described with some modifications [22]. The flies were entrained to 12:12 LD cycles during their development, and 8–10-d-old single males were placed in glass tubes containing food and monitored for their locomotor activity using the DAM system. Infrared beam crosses in 30 min bins were recorded. The circadian period and rhythmicity were estimated from locomotor activity data collected for 10 days in DD.

All the statistical analyses were performed using IBM SPSS Statistics 22 (IBM Japan, Ltd.) or BellCurve for Excel (Social Survey Research Information Co., Ltd.) except for the comparisons of MI. The circadian period, and rhythmicity were estimated from the locomotor activity data collected for 10 days in DD. All statistical analyses except for the comparisons of MI, circadian period, and rhythmicity were performed as previously described [5,22,55].

In the statistical analysis of CI, when the data were not distributed normally, we carried out the arcsine and square root transformations of the data. When the basic data or transformed data were normally distributed, Student’s t-test was used for comparisons. When the basic and transformed data were not distributed normally, we used the Mann–Whitney U test for comparisons. A randomization test based on the bootstrapping method was used in the statistical analysis of MI with an open R script [57]. The free statistical package R was used for these tests.

In sleep analysis, when the basic or transformed data are distributed normally, one-way ANOVA followed by post-hoc analysis using Scheffe’s test was carried out for multiple pairwise comparisons. For multiple group analysis of onparametric data, we used nonparametric ANOVA (Kruskal–Wallis test) followed by the rank-sum test for multiple pairwise comparisons.

In the statistical analysis of circadian rhythms, significant circadian rhythmicity was defined as the presence of a periodogram power peak that extends above the significance line (P < 0.05) in chi-square analysis. Clocklab software (Actimetrics) was used to analyze the circadian period and rhythmicity in each individual fly. All basic data of the circadian period were not normally distributed. The transformed data using arcsine transformation, log transformation, and square root transformation also did not show normal distribution. Thus, we used the Kruskal–Wallis test for comparisons of the circadian period. We used the G test with William’s correction to assess whether the percentage of rhythmicity (% rhythmic in Figs S4 and 7) differed significantly between the GAL4 control and the UAS control, or between the GAL4 control and the F₁ hybrids carrying both GAL4 and UAS.

In qRT-PCR, the mean (± SEM) ratio was calculated using data from five independent assays. Since the basic or log-transformed data were normally distributed, one-way ANOVA followed by post-hoc analysis using Scheffe’s test was used.