Neuropeptides control flexible mating behaviors based on context in fruit flies

We have described two distinct male Drosophila melanogaster behaviors as a model for investigating the neural circuit principles that determine interval timing. Male fruit flies exhibit the LMD behavior, which is characterized by a prolonged duration of mating after exposure to competing males. This behavior is believed to be an adaptation to male mating competition (S1A Fig) [34,35,37,52–59]. SMD is a behavior in which sexually experienced males manifest a shorter duration of mating. This behavior is believed to be an adaptation that allows male flies to conserve energy by mating quickly then continuing to other activities (S1B Fig) [36].

The SIFaR, the receptor for SIFa, has been linked to circadian rhythm, feeding, courtship, sleep, and memory extinction; however, its functional properties remain poorly understood [40–44,47,60]. When we used a pan-neuronal elav^c155 driver to knock down SIFaR in a neuronal population, both LMD and SMD were disrupted (Fig 1A and 1B) compared to controls (Fig 1C–1F). However, RNA interference-mediated knockdown of SIFaR using a pan-glial repo-GAL4 driver did not significantly alter either LMD or SMD behaviors (S1C and S1D Fig), suggesting that glial expression of SIFaR is not essential for the regulation of interval timing behaviors. The inclusion of elav-GAL80, which suppresses GAL4 activity in a pan-neuronal context, was found to restore both LMD and SMD behaviors when SIFaR was knocked down by a pan-neuronal elav^c155 driver (Fig 1G and 1H) compared to controls (Fig 1I and 1J). The observed reduction in SIFaR expression, driven by elav^c155, is deemed sufficient to induce significant disruptions in LMD and SMD behaviors.

To ensure that RNAi did not have an off-target effect, we tested three independent RNAi strains and found that all three RNAi successfully disrupted LMD/SMD when expressed in neuronal populations. (S1E–S1J Fig). We chose to use the HMS00299 line as SIFaR-RNAi for all our experiments because it efficiently disrupts LMD/SMD without UAS-dicer expression. Employment of broad drivers, including the tub-GAL4 and the strong neuronal driver nSyb-GAL4, with HMS00299 line consistently results in 100% embryonic lethality (S1O–S1R Fig). This phenotype mirrors the homozygous lethality observed in the SIFaR^B322 mutant. The efficiency of HMS00299 SIFaR-RNAi lines was also validated through quantitative PCR analysis (Fig 1K). The silencing of SIFaR mRNA was achieved at approximately 50% using the HMS00299 knockdown line in combination with tub-GAL80^ts, with RNAi induction lasting for three days (bottom diagram in Fig 1K). Notably, the same tub-GAL4 driver, when used without the tub-GAL80^ts combination, resulted in embryonic lethality while still reducing SIFaR mRNA levels by 50% after three days of RNAi induction. This finding suggests that SIFaR knockdown using the HMS00299 line with GAL4 drivers is likely sufficient to elicit the observed LMD and SMD behaviors. This rationale underscores the effectiveness of our experimental approach and its potential implications for understanding the role of SIFaR in mating behaviors. When we used tub-GAL80^ts to knock down SIFaR in all neuronal populations in an adult-specific manner, both LMD and SMD were also impaired compared to temperature control (Compare S1K and S1L Fig with S1M and S1N Fig).

We used the newly developed chemoconnectome (CCT) knock-in SIFaR-GAL4 (SIFaR^2A-AC.GAL4, SIFaR^2A hereafter) strain [61] to examine SIFaR expression in the nervous system, while the expression pattern of SIFaR in the brain has been previously characterized [61], our findings reveal abundant SIFaR expression throughout the VNC highly overlapping with neuropil marker nc82 (Fig 1L). Neuropil is a synaptic dense region in the nervous system composed of mostly unmyelinated axons, dendrites, and glial cell processes [62]. The monoclonal antibody nc82, which identifies Drosophila Bruchpilot (Brp), a constituent of the CAST1/ERC family and a component of the active zone, has been employed to examine the localization of certain neuronal populations. The significant overlap between the membrane marker of these neurons and the neuropil marker nc82 indicates a possible presence at the active zone [63]; nonetheless, this should not be construed as direct evidence of functionality. The studies demonstrate that SIFaR is extensively expressed in the CNS. Although the expression pattern is indicative, it is crucial to acknowledge that additional tests are necessary to confirm a direct correlation between SIFaR expression in these neurons and the initiation of interval timing behaviors.

To identify the minimal region labeling GAL4 drivers for testing the functionality of SIFaR-mediated interval timing behaviors, we examined four custom GAL4 drivers [64] targeting the SIFaR regulatory region (Fig 2A). LMD and SMD disappeared when SIFaR-RNAi was co-expressed with GAL4^57F10 (S2E and S2F Fig) and GAL4^24F06 (Fig 2B and 2C) compared to controls (Fig 2D and 2E) but not GAL4^23G06 and GAL4^24A12 (S2A–S2D Fig), indicating that SIFaR expression in these GAL4 driver-expressing cells is required for interval timing behaviors. Targeted reduction of SIFaR expression specifically in the GAL4^24F06 neuronal subset resulted in altered mating behavior. Both singly reared and sexually experienced flies exhibited an extended MD relative to naïve flies, contrary to the expected reduction. This observation indicates a deficit in the neural mechanism responsible for modulating MD, suggesting a disinhibition-like effect within the neural circuitry governing mating behavior. We have also previously observed a similar phenotype when sNPF peptidergic signaling is inhibited in specific neuronal circuits [65]. The prolonged MD observed after SIFaR knockdown suggests a disinhibition-like effect within the mating circuitry. This phenotype is similarly evident in Drosophila light preference and ethanol sensitization behaviors, indicating a broader regulatory role of SIFaR in inhibitory control [66,67].

To determine the cell populations in which SIFaR expression is sufficient to maintain interval timing behaviors, we conducted genetic rescue experiments with the SIFaR^B322 homozygous lethal mutant strain. As shown in Table 1, the expression of UAS-SIFaR with GAL4^57F10 and GAL4^23G06 drivers in a SIFaR^B322 mutant background cannot rescue the lethality of the mutant, whereas GAL4^24A12 and GAL4^24F06 can rescue the lethality of the SIFaR^B322 mutant by expressing UAS-SIFaR. Among these two GAL4 drivers, only GAL4^24F06 can rescue both lethality and interval timing behaviors in a SIFaR^B322 mutant background by expressing UAS-SIFaR (Fig 2F and 2G) compared to controls (Fig 2H–2K), indicating that the cells labeled by GAL4^24F06 driver contain neurons that are critical for both lethality and interval timing behaviors via SIFa-SIFaR neuropeptidergic signaling.

The expression patterns (Fig 2L) and the number of GAL4^24F06 expressing cells (Fig 2M) are similar between male and female brains, these results indicate that GAL4^24F06 labeled cells are highly expressed in the brain and VNC of both sexes. To compare the expression patterns of GAL4^24F06-positive neurons with those of all other SIFaR-expressing cells, we utilized genetic intersectional methods with a recently developed SIFaR knock-in line, SIFaR^2A driver (S2G Fig) [61]. The images indicate that the newly identified GAL4^24F06 neurons are pertinent SIFaR-positive cells, with particularly high expression in optic lobe (OL), sub-esophageal ganglion (SOG), and abdominal ganglion (AG) neurons. To better characterize the coexpression of GAL4^24F06 and SIFaR^2A, we employed an FLP-dependent fluorescence reporter system. These results (S2H Fig) confirm and extend our previous findings, demonstrating consistent overlap between these genetic markers in OL, SOG, and AG.

To further characterize the molecular identity of GAL4^24F06 neuronal population, we first examined potential sexual dimorphism using intersectional approaches with dsx^FLP and fru^FLP-based fluorescent reporters. Surprisingly, no GAL4^24F06 neurons expressed dsx or fru (S1A and S1B Supplemental information), indicating these neurons lack canonical sexual identity markers. We next mapped neuropeptide coexpression patterns using lexA^24F06 and GAL4 lines for peptidergic pathways known to modulate LMD/SMD behaviors (S1 and S2 Tables) [68]. Strikingly, in the SOG and AG, lexA^24F06 neurons showed extensive colocalization with six neuropeptides: AstA, Capa, FMRFa, Lk, Ms, and Proc (S1C–S1H Supplemental information). These findings suggest that: (1) the SOG and AG serve as critical integration hubs where GAL4^24F06 neurons receive SIFa inputs from the PI region, and (2) these neurons may coordinate interval timing decisions across social-sexual contexts through regulation of these neuropeptidergic pathways.

To determine the neuronal architectures between SIFa neurons and the identified essential SIFaR-expressing neurons, we labeled both SIFa and SIFaR neurons using genetic intersectional methods [69]. As previously reported, SIFa neurons arborize extensively throughout the CNS, but the neuronal processes of GAL4^24F06-positive neurons are enriched in the OL, SOG, and AG (GFP signal in Fig 2L). Neuronal processes that are positive for SIFa and SIFaR strongly overlap in the prow (PRW), prothoracic and metathoracic neuromere (ProNm and MesoNm), and AG regions (yellow signals in S3A Fig). We quantified these overlapping neuronal processes between SIFa- and SIFaR-positive neurons and found that approximately 18% of SIFa neurons and 52% of GAL4^24F06-positive neurons overlap in brain (S3B and S3C Fig), whereas approximately 48% of SIFa and 54% of GAL4^24F06-positive neurons overlap in VNC (S3D and S3E Fig). These findings suggest that SIFa neurons and GAL4^24F06-positive neurons form more neuronal processes in the VNC than in the brain. These results imply that GAL4^24F06-positive neurons are interconnected in specific regions of the brain and the VNC (S3H Fig). While GAL4^24F06 labels diverse neurons, the partial overlap with SIFaR highlights specific subregions (e.g., SOG and AG) where SIFa-SIFaR interactions are anatomically plausible, guiding future cell-type-specific investigations.

The observed synaptic alterations in GAL4^24F06 neurons may arise from the integrated activity of both SIFaR-expressing and non-expressing subpopulations. Nevertheless, given the functional significance of GAL4^24F06 neurons in rescuing lethality (Table 1) and behavioral deficits (Fig 2F and 2G) in genetic complementation assays, our data support a model wherein SIFa-SIFaR signaling modulates interval timing behaviors through dual mechanisms: direct synaptic modulation and indirect network-level integration.

We utilized synaptobrevin-GRASP (GFP Reconstitution Across Synaptic Partners) [70], a potent activity-dependent marker for synapses in vivo for assessing neuronal connectivity, to examine the synaptic changes between SIFa and SIFaR within brain neurons in various social contexts [71]. Therefore, we employed the SIFa^2A-lexA driver, which specifically labels ventral-posterior SIFa neurons (SIFa^VP) that only project to the brain [48]. Using GRASP experiments, we determined that SIFa^VP-SIFaR neurons form synapses in the gall (GA) of the brain (Fig 3A). GA are made up of a small group of neurons with numerous synapses and relatively few glia around them. They stretch out from the top of the lateral accessory lobe (LAL), adjacent to the ventrolateral protocerebrum and below the spur of the mushroom body (MB) [72]. Since GA are linked to the EB and protocerebral bridge (PB), they may facilitate learning and memory (S4A Fig) [73–75]. We utilized the “Virtual Fly Brain (VFB)” platform, an interactive tool designed for exploring neuronal connectivity, to gain insights into the connectivity of SIFa neurons with four other neurons, specifically GA, FB, and AL (Figs 3B and S4B) [76]. While VFB provides valuable information, it does not offer precise locations of synapses originating from SIFa neurons. To address this limitation, we incorporated data from the FlyEM hemibrain connectome (v1.2.1), which allowed us to confirm that SIFa projections indeed form actual synapses with GA, AL, FB, and SMP (S3F and S3G Fig) [77]. This multi-faceted approach enhances the robustness of our findings by integrating different data sources to validate neuronal connections.

Males reared in groups (we named naïve condition, hereafter) had stronger connections between SIFa and SIFaR neurons in CB, PI, GA, and fan-shaped body (FB) than males reared in isolation (compare naïve and single in Fig 3C; 3E; 3G; 3I). In addition, the strength of the connections in CB, PI, GA, and FB was reduced in males with sexual experience compared to males without sexual experience (compare naïve and exp. in Fig 3D; 3F; 3H; 3J). Besides, the intensity of synapses decreased when males were socially isolated (S4F and S4G Fig) and sexually experienced (S4I and S4J Fig). However, the size of synapses in different conditions remained constant (S4H and S4K Fig), indicating that the changes of SIFa-SIFaR synapses by sociosexual experiences are authentic. In contrast to synapses between SIFa and SIFaR formed in PI, GA, and FB, synapses between SIFa and SIFaR neurons in AL remained constant or did not vary significantly across sociosexual conditions (Figs 3E–3J; S4C–S4E). In VNC, no synapses were found between SIFa^2A-lexA and GAL4^24F06 (S4L Fig). All of these findings indicate that sociosexual experience may have a substantial effect on the connectivity between SIFa and SIFaR neurons (Fig 3K).

Given the strong expression of SIFaR^2A cells in the AG of VNC, we employed GAL4^SIFa.PT, a GAL4 driver line that project to both AG and the brain area [78], to examine the synaptic connections between SIFa and SIFaR neurons in VNC across various social contexts (Raw data provided in S2 Supplemental information). Social isolation and sexual experiences led to a decrease in the overall synaptic connections between SIFa and SIFaR in the brain and VNC. This reduction was particularly observed in the decline of OL when socially isolated and AG when both socially isolated and sexually experienced (Figs 4A–4G; S5A–S5E). These findings indicate that social isolation leads to a decrease in the SIFa-SIFaR synapses in the visual pathways (OL), as well as a significant decrease in the SIFa-SIFaR synapses in the AG of VNC when males experience social isolation or sexual interactions (Fig 4H). The subtle differences in GRASP signals observed in Fig 3A may stem from the distinct expression patterns of the SIFa^2A-lexA and GAL4^SIFa.PT drivers. We would like to emphasize that the SIFa^2A driver labels only a subset of SIFa neurons in other regions [48].

Next, we investigated the calcium response properties of GAL4^24F06 neurons in different social experiences of fly. We analyzed neuronal activity using a CaLexA-based transcriptional reporter system (Raw data provided in S3 Supplemental information) in which sustained neural activity drives GFP expression [79]. We found that the levels of CaLexA signals in both the brain and VNC were significantly elevated in both socially isolated and sexually experienced males when compared to naïve males (S5F–S5J Fig). CaLexA signals originating from GAL4^24F06-positive SOG neurons were found to be at higher levels in males reared singly and with sexual experience than in naïve males (Fig 4I–4K). In comparison to the SOG region, the calcium signals observed in SIFaR^24F06-positive OL neurons exhibited an increase solely in the singly reared condition, while no significant difference was observed in the sexually experienced condition when compared to the naïve condition (Fig 4L and 4M).

The CaLexA signals originating from the AG region in the VNC showed a decrease when males were subjected to social isolation (Fig 4N). Conversely, these signals exhibited an increase when males had prior sexual experience (Fig 4O). In contrast to the calcium signal changes of AG, the CaLexA signals originating from the VAC region exhibited an increase when males were socially isolated, but no significant differences were observed when sexually experienced (S5K and S5L Fig). These findings suggest a strong link between the synaptic plasticity of SIFa-SIFaR and calcium activity in GAL4^24F06-positive neurons in males, in response to different social contexts (Fig 4P).

Additionally, SIFa neurons express SIFaR, which enables them to produce SMD behavior [48]. In order to assess the influence of significant alterations in calcium activity that occur precede SIFa-SIFaR synaptic changes, we examined the SIFaR-to-SIFa synaptic changes using tGRASP. No synapses along the SIFaR-to-SIFa direction were observed in the VNC (S6A–S6C Fig). Significant synaptic plasticity has been observed between SIFaR and SIFa neurons in the PI, PRW, and OL regions of the brain. Additionally, synapses around the PI region also showed an increase in experienced males compared to group-reared males, while a decrease was observed in singly reared males compared to group-reared males (S6D and S6E Fig). The synapses between SIFaR and SIFa in PRW exhibited a decrease both in singly reared males and sexually experienced males (S6F and S6G Fig). The synapses between SIFaR and SIFa in OL exhibited a decrease solely in males who were socially isolated, rather than in those who had sexual experiences (S6H and S6I Fig). Therefore, it can be demonstrated that the synaptic modifications of SIFa-to-SIFaR take place when there are changes in the calcium activity of SIFaR neurons, resulting in changes in SIFaR-to-SIFa synapses (S6J Fig). The evidence indicates that neuronal circuits involving SIFa-to-SIFaR-to-SIFa utilize feed-forward augmentation to modify the internal states of the CNS in response to different social contexts (S6K and S6L Fig).

To investigate alterations in the synaptic terminals of the GAL4^24F06-labeled neuron across various social-sexual contexts, we utilized UAS-sty.eGFP (Fig 5A). Our findings revealed a significant increase in the synaptic terminals within the OL in the single condition, whereas no differences were observed between the naïve and experienced conditions (Fig 5B and 5C). In the SOG, the synaptic terminal intensity under naïve conditions was lower compared to both the single and experienced conditions (Fig 5D and 5E). In the AG region, the synaptic terminal intensity was higher in the experienced condition than in the naïve condition, with no significant difference observed between the naïve and single conditions (Fig 5F and 5G). Similar modifications in synaptic terminal dynamics have also been noted in the brain and the VNC (S7A–S7D Fig). These results suggest that GAL4^24F06-labeled neuron in these regions undergoes significant alterations in synaptic terminal dynamics, which are likely involved in modulating interval timing behaviors.

To investigate the response of SIFaR neurons in the AG region following the activation of SIFa neurons in the PI region, we conducted live calcium (Ca²⁺) imaging in the AG region of the VNC, where SIFaR^24F06 neurons are located (Fig 5H). We tracked Ca²⁺ fluctuations in these neurons using two-photon microscopy in conjunction with a genetically encoded calcium indicator, GCaMP7s, which was expressed under the control of the GAL4^24F06 driver. To selectively stimulate SIFa neurons, we utilized an optogenetic strategy that involved the red-light-sensitive channelrhodopsin (CsChrimson), expressed using the lexA^SIFa.PT driver (Fig 5I). Upon optogenetic activation of SIFa neurons, we noted a significant rise in the activity of SIFaR^24F06 neurons in the AG region (Fig 5J–5L), as indicated by a prolonged elevation in intracellular Ca²⁺ concentrations that remained at a high level before slowly returning to baseline. This increase in activity was in stark contrast to the control group, which did not receive all-trans retinal (ATR). The results suggest that SIFaR neurons in the AG region respond to the activation of SIFa neurons, which is likely involved in the modulation of interval timing behaviors.

The neuropeptide Corazonin (Crz) has been identified as a participant in the regulation of male Drosophila MD within the AG of the VNC. The male AG’s four neurons that express Crz also express fruitless (fru), and their silencing results in reduced sperm transfer and extended MD [80]. Furthermore, the Crz peptide and Crz-expressing neurons have been characterized as pivotal relay signals in the SIFa-to-SIFaR pathway, which is essential for modulating interval timing behaviors.

Neurons expressing Crz were observed in close proximity to SIFaR^24F06-expressing neurons within the PRW-SOG of the brain (panels of SOG in Fig 6A). The modulation of interval timing in LMD is primarily facilitated through the visual pathway [34,35]. Given the established absence of Crz peptide in the OL [81], SIFaR^24F06-expressing neurons within this region may represent a critical neural substrate for LMD behavior that operates independently of Crz signaling. In comparison to the brain, a smaller proportion of SIFaR^24F06 neuron processes exhibited overlap with Crz-GAL4 in VNC (S7E Fig). However, the overlapping increased in the AG region. The signals of SIFaR^24F06 located outside of the AG region in VNC exhibited minimal overlap with Crz signals (S7E Fig).

By employing the genetic intersectional method, we have determined that approximately 100 neurons in the OL and 2 neurons in Super Intermediate Protocerebrum (SIP) express Crz in conjunction with SIFaR^24F06 neurons (Fig 6B and 6C) [82]. However, the Crz-positive AG neurons in the VNC are not associated with the SIFaR^24F06 driver (S7F Fig), indicating that only Crz neurons in the brain are positive for SIFaR^24F06.

We have previously determined that Crz expression in SIFaR^24F06 cells is required for both LMD and SMD by a systematic screening of GAL4^24F06-mediated neuropeptide knockdown (Fig 6D and 6E) compared to controls (Fig 6F and 6G), whereas SIFaR expression in Crz neurons is only required for SMD and not LMD (Fig 6H and 6I) compared to controls (Fig 6J and 6K) [68]. It has been reported that KCNJ2, which functions by inwardly rectifying the K⁺ channel to impede Crz neuronal activity, can substantially extend the duration of mating [80]. Previous research has documented a significant augmentation in the MD exhibited by these males. However, this genetic manipulation only disrupted LMD and not SMD (Fig 6L and 6M) compared to controls (Fig 6N and 6O).

To elucidate the direct response of Crz neurons to the activity of SIFa neurons, we conducted live calcium (Ca²⁺) imaging in the SIP, OL, and AG region of the VNC, where Crz neurons are situated (Figs 7A, 7B, and 7E; S7G). Upon optogenetic stimulation of SIFa neurons, we observed a tendency to maintain the activity of Crz neurons in OL and AG regions (Figs 7B–7D, S7G–S7I), evidenced by a sustained activity in intracellular Ca²⁺ levels that persisted in a high level compared to control ATR⁻ condition which shows gradual declining to baseline levels (Fig 7B–7D). In contrast to the OL and AG regions, the cells in the upper region of the SIP consistently show a decrease in Ca²⁺ levels following stimulation of the SIFa neurons (Fig 7E–7G). These calcium level changes were in contrast to the control group (without ATR) (Figs 7B–7G; S7G–S7I). These findings confirm that Crz neurons in OL and AG are activated in response to SIFa neuronal activity, but the activity of Crz neurons in SIP is inhibited by the activation of SIFa neuron, reinforcing their role as postsynaptic effectors in the neural circuitry governed by SIFa neurons. Moreover, these results support a model where long-range neuropeptide signaling - specifically SIFa-SIFaR in the brain coupled with Crz-CrzR in the AG—contributes significantly to the neural mechanisms governing interval timing (Fig 7H).

The majority of CrzR expression can be found in oenocytes (S8A Fig), and adult oenocytes (S8B Fig) have an enrichment of CrzR, according to the analysis of the fly Scope scRNA sequencing dataset [83]. However, the suppression of CrzR in oenocytes did not have any impact on both LMD and SMD behaviors in adults (S8C and S8D Fig) compared to controls (S8J and S8K Fig), indicating that oenocyte-expressing CrzR is not necessary for interval timing behaviors. Remarkably, the pan-neuronal suppression of CrzR exhibits no discernible impact on either LMD or SMD (S8E and S8F Fig) compared to controls (S8L and S8M Fig). Interestingly, our findings indicate that the downregulation of CrzR in the glial population resulted in the disruption of LMD rather than SMD (S8G and S8H Fig) compared to controls (S8N and S8O Fig). The efficiency of the CrzR-RNAi line has been confirmed by quantitative PCR (S8I Fig). All these data indicate that the relay of SIFa-SIFaR/Crz-CrzR neuropeptide signaling necessitates the presence of non-neuronal population in order to regulate interval timing behaviors.

Drosophila melanogaster were raised on cornmeal-yeast medium at similar densities to yield adults with similar body sizes. Flies were kept in 12 h light: 12 h dark cycles (LD) at 25 °C (ZT 0 is the beginning of the light phase, ZT12 beginning of the dark phase) except for some experimental manipulation. For temperature-controlled experiments, including those utilizing the temperature-sensitive tub-GAL80^ts driver, the flies were initially crossed and maintained at a constant temperature of 22 °C within an incubator. The temperature shift was initiated post-eclosion. Once the flies had emerged, they were transferred to an incubator set at an elevated temperature of 29 °C for a defined period, after which the experimental protocols were carried out. Wild-type flies were Canton-S (CS).

Following lines used in this study, Canton-S (#64349), Df(1)Exel6234 (#7708), Crz-GAL4 (#51977), elav^c155; UAS-Dicer (#25750), lexAop-CD8GFP; UAS-mLexA-VP16-NFAT, lexAop-rCD2-GFP (#66542), lexAop-nSyb-spGFP1-10, UAS-CD4-spGFP11 (#64315), UAS-post-t-GRASP, lexAop2-pre-t-GRASP (#79039), UAS-pre-t-GRASP, lexAop2-post-t-GRASP (#79040), PromE-GAL4 (#65405), repo-GAL4 (#7415), SIFaR^B322 (#16202), GAL4^23G06 (#49041), GAL4^24A12 (#49061), GAL4^24F06 (#49087), GAL4^57F10 (#46391), lexA^24F06 (#52695), SIFaR-RNAi^JF01849 (#25831), SIFaR-RNAi^HMS00299 (#34947), tub-GAL80^ts (#7108), UAS-CD4tdGFP (#35839), UAS-RedStinger (#8546), UAS-FRT-stop-FRT-myrGFP (#55811), fru-FLP (#66870), CrzR-RNAi^JF02042 (#26017), lexAop-flp (#90892), mCherry-RNAi (#35785), UAS-Denmark, UAS-syt.eGFP (#33065), UAS-KCNJ2 (#6596), UAS-mCD8RFP, lexAop-mCD8GFP (#32229), UAS-GCaMP7s (#80905), AstA-GAL4^T2A (#84593), FMRFa-GAL4^T2A (#84633), Proc-GAL4 (#51972), and lexAop-CsChrimson (#55138) were obtained from the Bloomington Drosophila Stock Center at Indiana University. The following lines, SIFaR-RNAi^v1783(#1783), and Crz-RNAi^v30670 (#30670) were obtained from the Vienna Drosophila Resource Center. The following lines, SIFa-lexA^T2A (#FBA00116), SIFa-GAL4^T2A (#FBA00103), SIFaR-GAL4^T2A (#FBF00102), Capa-GAL4^T2A (#FB100010↗), Ms-GAL4^T2A (#FBA00065), Lk-GAL4^T2A (#FBA00235), and SIFaR-lexA^T2A (#FBF00086) were obtained from Qidong Fungene Biotechnology in China. The GAL4^SIFa.PT was a gift from Jan A. Veenstra. The UAS-SIFaR was a gift from Young-Joon Kim in GIST. The fly genotypes used in each figure of this study are listed in Table 2.

The CS background was selected as the experimental background due to its well-characterized and consistent LMD and SMD behaviors. To ensure that genetic variation did not confound our results, all GAL4, UAS, and RNAi lines employed in our assays were rigorously backcrossed into the CS strain, often exceeding ten generations of backcrossing. This approach was undertaken to isolate the effects of our genetic manipulations from those of genetic background. We assert that the extensive backcrossing to the CS background, in concert with the internal control in LMD and SMD, provides a stable platform for the accurate interpretation of the LMD and SMD phenotypes observed in our experiments. To reduce the variation from genetic background, all flies were backcrossed for at least 10 generations to CS strain. For the generation of outcrosses, all GAL4, UAS, and RNAi lines employed as the virgin female stock were backcrossed to the CS genetic background for a minimum of ten generations. Notably, the majority of these lines, which were utilized for LMD assays, have been maintained in a CS backcrossed state for long-term generations subsequent to the initial outcrossing process, exceeding ten backcrosses. Based on our experimental observations, the genetic background of primary significance is that of the X chromosome inherited from the female parent. Consequently, we consistently utilized these fully outcrossed females as virgins for the execution of experiments pertaining to LMD and SMD behaviors. Contrary to the influence on LMD behaviors, we have previously demonstrated that the genetic background exerts negligible influence on SMD behaviors, as reported in our prior publication [36]. The mutants and transgenic lines utilized in this study have been previously characterized, with the exception of the novel transgenic strains that we generated and describe herein.

The MD assay in this study has been reported [34–36]. To enhance the efficiency of the MD assay, we utilized the Df(1)Exel6234 (DF hereafter) genetic modified fly line in this study, which harbors a deletion of a specific genomic region that includes the sex peptide receptor (SPR) [95,96]. Previous studies have demonstrated that virgin females of this line exhibit increased receptivity to males [96]. We conducted a comparative analysis between the virgin females of this line and the CS virgin females and found that both groups induced SMD. Consequently, we have elected to employ virgin females from this modified line in all subsequent studies. For group-reared (naïve) males, 40 males from the same strain were placed into a vial with food for 5 days. For single-reared males, males of the same strain were collected individually and placed into vials with food for 5 days. For experienced males, 40 males from the same strain were placed into a vial with food for 4 days then 80 DF virgin females were introduced into vials for last 1 day before assay. Forty DF virgin females were collected from bottles and placed into a vial for 5 days. These females provide both sexually experienced partners and mating partners for MD assays. At the fifth day after eclosion, males of the appropriate strain and DF virgin females were mildly anaesthetized by CO₂. After placing a single female in to the mating chamber, we inserted a transparent film then placed a single male to the other side of the film in each chamber. After allowing for 1 h of recovery in the mating chamber in 25 °C incubators, we removed the transparent film and recorded the mating activities. Only those males that succeeded to mate within 1 h were included for analyses. Initiation and completion of copulation were recorded with an accuracy of 10 s, and total MD was calculated for each couple. Part of genetic controls with GAL4/+ or UAS/+ lines were omitted from supplementary figures, as prior data confirm their consistent exhibition of normal LMD and SMD behaviors [34–37,59]. Hence, genetic controls for LMD and SMD behaviors were incorporated exclusively when assessing novel fly strains that had not previously been examined. In essence, internal controls were predominantly employed in the experiments, as LMD and SMD behaviors exhibit enhanced statistical significance when internally controlled. Within the LMD assay, both group and single conditions function reciprocally as internal controls. A significant distinction between the naïve and single conditions implies that the experimental manipulation does not affect LMD. Conversely, the lack of a significant discrepancy suggests that the manipulation does influence LMD. In the context of SMD experiments, the naïve condition (equivalent to the group condition in the LMD assay) and sexually experienced males act as mutual internal controls for one another. A statistically significant divergence between naïve and experienced males indicates that the experimental procedure does not alter SMD. Conversely, the absence of a statistically significant difference suggests that the manipulation does impact SMD. Hence, we incorporated supplementary genetic control experiments solely if they deemed indispensable for testing. All assays were performed from noon to 4 PM. We conducted blinded studies for every test.

To generate the SIFa^PT-lexA driver, the putative promotor sequence of the gene was amplified by PCR using wild-type genomic DNA as a template with the following primers GCCAATTGGCTGAATCTCCTGACCCTCA and GCAGATCTCTTGCAGTTTTCGGTGAGC as mentioned before [78]. The amplified DNA fragment (1482 base pairs located immediately upstream of the SIFa coding sequence) was inserted into the E2 Enhancer-lexA vector. This vector, supplied by Qidong Fungene Biotechnology Co., Ltd. (http://www.fungene.tech/↗), is a derivative of the pBPLexA::p65Uw vector (available at https://www.addgene.org/26231↗). The insertion was achieved by digesting the fragment and the vector with EcoRI and XbaI restriction enzymes to create compatible sticky ends. The genetic construct was inserted into the attp2 site on chromosome III to generate transgenic flies using established techniques, a service conducted by Qidong Fungene Biotechnology Co., Ltd.

The dissection and immunostaining protocols for the experiments are described elsewhere [34]. After 5 days of eclosion, the Drosophila brain was taken from adult flies and fixed in 4% formaldehyde at room temperature for 30 min. The sample was washed three times (5 min each) in 1% PBT and then blocked in 5% normal goat serum for 30 min. The sample next be incubated overnight at 4 °C with primary antibodies in 1% PBT, followed by the addition of fluorophore-conjugated secondary antibodies for one hour at room temperature. The brain was mounted on plates with an antifade mounting solution (Solarbio) for imaging purposes.

Samples were imaged with Zeiss LSM880. Antibodies were used at the following dilutions: Chicken anti-GFP (1:500, Invitrogen), mouse anti-nc82 (1:50, DSHB), rabbit anti-DsRed (1:500, Rockland Immunochemicals), Alexa-488 donkey anti-chicken (1:200, Jackson ImmunoResearch), Alexa-555 goat anti-rabbit (1:200, Invitrogen), Alexa-647 goat anti-mouse (1:200, Jackson ImmunoResearch).

To ascertain calcium levels and synaptic intensity from microscopic images, we dissected and imaged five-day-old flies of various social conditions and genotypes under uniform conditions. For group-reared (naïve) flies, the flies were reared in group condition and dissect right after 5 days of rearing without any further action. For single-reared flies, the flies were reared in single condition and dissect at the same time as group-reared flies right after 5 days of rearing without any further action. For sexual experienced flies, the flies were reared in group condition after 4 days of rearing and will be given virgins to give them sexual experience for 1 day, those flies will also be dissected at the same time as group and single-reared flies after one day. The GFP signal in the brains and VNCs was amplified through immunostaining with chicken anti-GFP primary antibody. Image analysis was conducted using ImageJ software. For the quantification of fluorescence intensities, an investigator, blinded to the fly’s genotype, thresholded the sum of all pixel intensities within a sub-stack to optimize the signal-to-noise ratio, following established methods [70]. The total fluorescent area or region of interest (ROI) was then quantified using ImageJ, as previously reported. For CaLexA signal quantification, we adhered to protocols detailed by Kayser and colleagues [97], which involve measuring the ROI’s GFP-labeled area by summing pixel values across the image stack. This method assumes that changes in the GFP-labeled area are indicative of alterations in the CaLexA signal, reflecting synaptic activity. ROI intensities were background-corrected by measuring and subtracting the fluorescent intensity from a non-specific adjacent area, as per Kayser and colleagues [97]. For the analysis of GRASP or tGRASP signals, a sub-stack encompassing all synaptic puncta was thresholded by a genotype-blinded investigator to achieve the optimal signal-to-noise ratio. The fluorescence area or ROI for each region was quantified using ImageJ, employing a similar approach to that used for CaLexA quantification [70].

To prepare representative images for figures, grayscale channels were used for single-fluorophore display to improve clarity and avoid color-based misinterpretation. For overlays involving multiple reporters (e.g., GFP, RFP, and nc82), distinct non-overlapping pseudocolors were applied, with channel identities labeled below each panel. Thresholding for all displayed images was performed conservatively and consistently across genotypes and conditions, with the primary goal of preserving visible biological signals while avoiding oversaturation. For quantified data, identical threshold settings were applied across conditions. All brightness and contrast adjustments were applied uniformly to the entire image and were restricted to linear modifications for visual clarity only. No nonlinear enhancements or selective adjustments were applied. These standardized procedures ensured both accurate visualization and reproducible quantification. Raw image data are available upon request.

Before the colocalization analysis, an investigator, blinded to the fly’s genotype, thresholded the sum of all pixel intensities within a sub-stack to optimize the signal-to-noise ratio, following established methods [70]. To perform colocalization analysis of multi-color fluorescence microscopy images in this study, we employed ImageJ software [98]. In brief, we merged image channels to form a composite with accurate color representation and applied a threshold to isolate yellow pixels, signifying colocalization. The measured “area” values represented the colocalization zones between fluorophores. To determine the colocalization percentage relative to the total area of interest (e.g., GFP or RFP), we adjusted thresholds to capture the full fluorophore areas and remeasured to obtain total areas. The quantification of the overlap was performed using confocal images with projection by standard deviation function provided by ImageJ to ensure precise measurements and avoid pixel saturation artifacts. The colocalization efficiency was calculated by dividing the colocalized area by the total fluorophore area. All samples were imaged uniformly.

Before the particle analysis, an investigator, blinded to the fly’s genotype, thresholded the sum of all pixel intensities within a sub-stack to optimize the signal-to-noise ratio, following established methods [70]. To quantitatively measure particle intensity of cell number and synaptic puncta in microscopic images, we applied ImageJ software. Initially, the image is converted to grayscale to reduce complexity and enhance contrast. Subsequently, thresholding techniques are employed to binarize the image, distinguishing particles from the background. This binarization can be achieved through automated thresholding algorithms or manual adjustment to optimize the segmentation. The results of these measurements are then available for review by conducting “Analyze Particles” function of ImageJ. All specimens were imaged under identical conditions.

The expression levels of SIFaR and CrzR after knockdown by SIFaR-RNAi lines and CrzR-RNAi were analyzed by quantitative real-time RT-PCR with SYBR Green qPCR MasterMix kit (Selleckchem). The primers of SIFaR are F:5′-AAGCAGGAGAGCGAGTTCAG-3′; R: 5′-TTCGCCTTGTTTTGTCACAG-3′ [99]. The primers of CrzR are F: 5′-TGGTGCCGCTCCTGAAATCGT-3′; R: 5′-TTCCGCCATCGGTGGTGCTTC-3′. qPCR reactions were performed in triplicate, and the specificity of each reaction was evaluated by dissociation curve analysis. Each experiment was replicated three times. PCR results were recorded as threshold cycle numbers (Ct). The fold change in the target gene expression, normalized to the expression of internal control gene (GAPDH) and relative to the expression at time point 0, was calculated using the 2^−ΔΔCT method as previously described [100]. The results are presented as the mean ± SD of three independent experiments.

All trans-retinal powder (Sigma) was dissolved in EtOH as a 100 mM stock solution. 200 μL of this stock solution was diluted in 50 mL of melted normal food to prepare 400 μM of all trans-retinal (ATR) food. For CsChrimson experiments, male flies of proper genotype were selected and separated into two groups (with and without ATR) after 5 days of eclosion for at least 3 days prior to any optogenetic experiments.

To measure the neuron activity in AG region, the flies were carefully sedated with ice and then placed in an inverted posture on a plastic plate coated with UV glue, which securely held flies in the middle. Whole fly was exposed to AHL for red light activation. The plate was filled with Drosophila Adult Hemolymph-Like Saline (AHLS) buffer [108 mM NaCl, 5 mM KCl, 4 mM NaHCO₃, 1 mM NaH₂PO₄, 15 mM ribose, 5 mM Hepes (pH 7.5), 300 mosM, CaCl₂ (2 mM), and MgCl₂ (8.2 mM)] to ensure the flies’ neurons remained in an active state throughout the experiment [101,102]. Subsequently, the flies were dissected to specifically expose the VNC for detailed examination. Calcium imaging was performed using Zeiss LSM880 microscope with a 20× water immersion objective. Images were acquired at 2 frames per second at a resolution of 512 × 512 pixels. GCaMP7 slow (GCaMP7s) signals were recorded with an 880 nm laser and optogenetic stimulation was achieved with a 590–595 nm 20 W red light pulses at 50–60 Hz (0.16 mW/mm²), LED stimulation lasts 2 s. ROIs were manually selected from the cell body in AG area with ImageJ.

ΔF/F₀ = (F_t − F₀)/F₀ × 100% and Peak ΔF/F₀ = (Peak F_t − F₀)/F₀ × 100% were used to determine the fluorescence change, where Ft is the fluorescence at time point n and F₀ is the fluorescence from the average intensity of 10 frames fluorescence before optogenetic stimulation. To quantify neuronal activity in the brain region, the brains of male flies with the appropriate genotype were dissected and secured in a plastic dish. All subsequent steps were conducted as previously described.

The snRNAseq dataset analyzed in this paper is published in [83] and available at the Nextflow pipelines (VSN, https://github.com/vib-singlecell-nf↗), the availability of raw and processed datasets for users to explore, and the development of a crowd-annotation platform with voting, comments, and references through SCope (https://flycellatlas.org/scope↗), linked to an online analysis platform in ASAP (https://asap.epfl.ch/fca↗). For the generation of the tSNE plots, we utilized the Fly SCope website (https://scope.aertslab.org/#/FlyCellAtlas/*/welcome↗). Within the session interface, we selected the appropriate tissues and configured the parameters as follows: ‘Log transform’ enabled, ‘CPM normalize’ enabled, ‘Expression-based plotting’ enabled, ‘Show labels’ enabled, ‘Dissociate viewers’ enabled, and both ‘Point size’ and ‘Point alpha level’ set to maximum. For all tissues, we referred to the individual tissue sessions within the ‘10X Cross-tissue’ RNAseq dataset. Each tSNE visualization depicts the coexpression patterns of genes, with each color corresponding to the genes listed on the left, right, and bottom of the plot. The tissue name, as referenced on the Fly SCope website is indicated in the upper left corner of the tSNE plot. Dashed lines denote the significant overlap of cell populations annotated by the respective genes. Coexpression between genes or annotated tissues is visually represented by differentially colored cell populations. For instance, yellow cells indicate the coexpression of a gene (or annotated tissue) with red color and another gene (or annotated tissue) with green color. Cyan cells signify coexpression between green and blue, purple cells for red and blue, and white cells for the coexpression of all three colors (red, green, and blue). Consistency in the tSNE plot visualization is preserved across all figures.

Single-cell RNA-sequencing (scRNA-seq) data from the Drosophila melanogaster were obtained from the Fly Cell Atlas (FCA) website (https://doi.org/10.1126/science.abk2432↗). Oenocytes gene expression analysis employed UMI (Unique Molecular Identifier) data extracted from the 10× VSN oenocyte (Stringent) loom and h5ad file, encompassing a total of 506,660 cells. The Seurat (v4.2.2) package (https://doi.org/10.1016/j.cell.2021.04.048↗) was utilized for data analysis. Violin plots were generated using the “Vlnplot” function, the cell types are split by FCA.

Whole brain connectomics data were obtained from VFB (https://v2.virtualflybrain.org↗) [76]. The right SIFa (VFB ID: VFB_jrchk542 and VFB_jrchk541) dataset was used to gather information on the synaptic connections between the presynaptic and the postsynaptic neurons of interest. The connectivity was visualized with Sankey diagram and doughnut diagram by the NetworkD3 R package (https://github.com/christophergandrud/networkD3↗).

Statistical analysis of MD assay was described previously [34–36]. More than 50 males (naïve, experienced, and single) were used for MD assay. Our experience suggests that the relative MD differences between naïve and experienced condition and singly reared are always consistent; however, both absolute values and the magnitude of the difference in each strain can vary. So, we always include internal controls for each treatment as suggested by previous studies [103]. Therefore, statistical comparisons were made between groups that were naïvely reared, sexually experienced, and singly reared within each experiment. As MD of males showed normal distribution (Kolmogorov–Smirnov tests, p > 0.05), we used two-sided Student’s t tests. The mean ± standard error (s.e.m) (**** = p < 0.0001, *** = p < 0.001, ** = p < 0.01, * = p < 0.05). All analysis was done in GraphPad (Prism). Individual tests and significance are detailed in figure legends.

Besides traditional t test for statistical analysis, we added estimation statistics for all MD assays and two group comparing graphs. In short, ‘estimation statistics’ is a simple framework that—while avoiding the pitfalls of significance testing—uses familiar statistical concepts: means, mean differences, and error bars. More importantly, it focuses on the effect size of one’s experiment/intervention, as opposed to significance testing [104]. In comparison to typical NHST plots, estimation graphics have the following five significant advantages such as (1) avoid false dichotomy, (2) display all observed values (3) visualize estimate precision (4) show mean difference distribution. And most importantly (5) by focusing attention on an effect size, the difference diagram encourages quantitative reasoning about the system under study [105]. Thus, we conducted a reanalysis of all our two group data sets using both standard t tests and estimate statistics. In 2019, the Society for Neuroscience journal eNeuro instituted a policy recommending the use of estimation graphics as the preferred method for data presentation [106].