To characterize Alk transcriptional targets in the Drosophila CNS, we employed TaDa. Briefly, transgenic Dam fused with RNA-Pol II (hereafter referred as Dam-Pol II) (Southall et al., 2013; Figure 1a and b) was driven using the pan neuronal C155-Gal4 driver. To inhibit Alk signaling we employed a dominant-negative Alk transgene, which encodes the Alk extracellular and transmembrane domain (hereafter referred as UAS-Alk^DN) (Bazigou et al., 2007; Figure 1a). Flies expressing Dam-Pol II alone in a wild-type background were used as control. Expression of Dam-Pol II was confirmed by expression of mCherry, which is encoded by the primary ORF of the TaDa construct (Southall et al., 2013; Figure 1b, Figure 1—figure supplement 1a and b’). CNS from third instar wandering larvae were dissected and genomic DNA was extracted, fragmented at GA^meTC marked sites using methylation-specific DpnI restriction endonuclease. The resulting GATC fragments were subsequently amplified for library preparation and NGS sequencing (Figure 1—figure supplement 1c). Bioinformatic data analysis was performed based on a previously described pipeline (Marshall and Brand, 2017; Mendoza-Garcia et al., 2021). Initial quality control analysis indicated comparable numbers of quality reads between samples and replicates, identifying >20 million raw reads per sample that aligned to the Drosophila genome (Figure 1—figure supplement 1d–d’). No significant inter-replicate variability was observed (Figure 1—figure supplement 1e). Meta-analysis of reads associated with GATC borders showed a tendency to accumulate close to transcription start sites (TSS) indicating the ability of TaDa to detect transcriptionally active regions (Figure 1—figure supplement 1f). A closer look at the Pol II occupancy profile of Alk shows a clear increase in Pol II occupancy from exon 1 to exon 7 (encoding the extracellular and transmembrane domain) in Alk^DN samples reflecting the expression of the dominant-negative Alk transgene (Figure 1—figure supplement 1g).

To detect differential Pol II occupancy between Dam-Pol II control (C155-Gal4>UAS-LT3-Dam::Pol ll) and UAS-Alk^DN (C155-Gal4>UAS-LT3-Dam::Pol II; UAS-Alk^DN) samples, neighboring GATC-associated reads, maximum 350 bp apart (median GATC fragment distance in the Drosophila genome) were clustered in peaks (Tosti et al., 2018). More than 10 million reads in both control and Alk^DN samples were identified as GATC-associated reads (Figure 1—figure supplement 1d’), and those loci displaying differential Pol II occupancy were defined by logFC and FDR (as detailed in Materials and methods). Greater than 50% of aligned reads were in promoter regions, with 33.55% within a 1 kb range (Figure 1c, Supplementary file 1).

To further analyze transcriptional targets of Alk signaling, we focused on loci exhibiting decreased Pol II occupancy when compared with controls, identifying 2502 loci with logFC≥2, FDR≤0.05 (Figure 1d, Supplementary file 1). Genes previously known to be associated with Alk signaling, such as kirre, RhoGAP15B, C3G, mib2, and mamo, were identified among downregulated loci (Figure 1d). We compared CNS TaDa Alk targets with our previously published embryonic visceral mesoderm TaDa datasets that were derived under similar experimental conditions (Mendoza-Garcia et al., 2021) and found 775 common genes (Figure 1e, Supplementary file 1). Gene Ontology (GO) analysis identified GO terms in agreement with previously reported Alk functions in the CNS (Bai and Sehgal, 2015; Bazigou et al., 2007; Cheng et al., 2011; Gouzi et al., 2011; Orthofer et al., 2020; Pfeifer et al., 2022; Rohrbough and Broadie, 2010; Rohrbough et al., 2013; Woodling et al., 2020) such as axon guidance, determination of adult lifespan, nervous system development, regulation of gene expression, mushroom body development, behavioral response to ethanol and locomotor rhythm (Figure 1f). Many of the differentially regulated identified loci have not previously been associated with Alk signaling and represent candidates for future characterization.

To further characterize Alk-regulated TaDa loci, we set out to examine their expression in scRNA-seq data from wild-type third instar larval CNS (Pfeifer et al., 2022). Enrichment of TaDa loci were identified by using AUCell, an area-under-the-curve-based enrichment score method, employing the top 500 TaDa hits (Aibar et al., 2017; Figure 2a and b, Supplementary file 1). This analysis identified 786 cells (out of 3598), mainly located in a distinct cluster of mature neurons that robustly express both Alk and jeb (Figure 2b, red circle; Figure 2c). This cluster was defined as neuroendocrine cells based on canonical markers, such as the neuropeptides Lk (Leucokinin), Nplp1 (Neuropeptide-like precursor 1), Dh44 (Diuretic hormone 44), Dh31 (Diuretic hormone 31), sNPF (short neuropeptide F), AstA (Allatostatin A), and the enzyme Pal2 (Peptidyl-α-hydroxyglycine-α-amidating lyase 2) as well as Eip74EF (Ecdysone-induced protein 74EF), and Rdl (resistance to dieldrin) (Guo et al., 2019; Hückesfeld et al., 2021; Takeda and Suzuki, 2022; Torii, 2009; Figure 2d–f). Overall, the TaDa-scRNA-seq data integration analysis suggests a role of Alk signaling in regulation of gene expression in neuroendocrine cells.

To further explore the observed enrichment of Alk-regulated TaDa loci in neuroendocrine cells, we used a Dimm transcription factor reporter (Dimm-Gal4>UAS-GFPcaax), as a neuroendocrine marker (Park et al., 2008), to confirm Alk protein expression in a subset of neuroendocrine cells in the larval central brain, ventral nerve cord, and neuroendocrine corpora cardiaca cells (Figure 2g, Figure 2—figure supplement 1). This could not be confirmed at the RNA level, due to low expression of dimm in both our and publicly available single-cell RNA-seq datasets (Brunet Avalos et al., 2019; Michki et al., 2021; Pfeifer et al., 2022).

Loci potentially subject to Alk-dependent transcriptional regulation were further refined by integration of the Alk-regulated TaDa dataset with previously collected RNA-seq datasets (Figure 3a). Specifically, w¹¹¹⁸ (control), Alk^Y1355S (Alk gain-of-function), and Alk^ΔRA (Alk loss-of-function) RNA-seq datasets (Pfeifer et al., 2022) were compared to identify genes that exhibited both significantly increased expression in Alk gain-of-function conditions (w¹¹¹⁸ vs Alk^Y1355S) and significantly decreased expression in Alk loss-of-function conditions (w¹¹¹⁸ vs Alk^ΔRA and control vs C155-Gal4-driven expression of UAS-Alk^DN). Finally, we positively selected for candidates expressed in Alk-positive cells in our scRNA-seq dataset. Notably, the only candidate which met these stringent criteria was CG4577, which encodes an uncharacterized putative neuropeptide precursor (Figure 3b). CG4577 exhibited decreased Pol II occupancy in Alk^DN samples (Figure 3c), and CG4577 transcripts were upregulated in Alk^Y1355S gain-of-function conditions and downregulated in Alk^ΔRA loss-of-function conditions (Figure 3d, Supplementary file 1). In agreement with a potential role as a neuropeptide precursor, expression of CG4577 was almost exclusively restricted to neuroendocrine cell clusters in our scRNA-seq dataset (Figure 3e). Examination of additional publicly available first instar larval and adult CNS scRNA-seq datasets (Brunet Avalos et al., 2019; Davie et al., 2018) confirmed the expression of CG4577 in Alk-expressing cells (Figure 3—figure supplement 1a and b). CG4577-RA encodes a 445 amino acid prepropeptide with a 27 amino acid N-terminal signal peptide sequence as predicted by SignalP-5.0 (Figure 3f; Almagro Armenteros et al., 2019). Analysis of CG4577-PA at the amino acid level identified a high percentage of glutamine residues (43 of 445; 9%), including six tandem glutamine repeats (amino acids 48–56, 59–62, 64–71, 116–118, 120–122, and 148–150) of unknown function as well as a lack of cysteine residues. The preproprotein has an acidic pI of 5.1 and carries a net negative charge of 6. Several poly- and dibasic proprotein convertase (PC) (proprotein convertase) cleavage sites were also predicted (KR, KK, RR, RK) (Pauls et al., 2014; Southey et al., 2006; Veenstra, 2000; Figure 3f). Since the propeptide does not contain cysteine residues, it is unable to form intracellular or dimeric disulfide bridges. A second transcript, CG4577-RB, encodes a 446 amino acid protein with only two amino acid changes (Figure 3—figure supplement 1c). Phylogenetic analysis of CG4577 relative to known Drosophila neuropeptide precursors failed to identify strong homology in keeping with the known low sequence conservation of neuropeptide prepropeptides outside their bioactive peptide stretches. However, we were also unable to find sequence homologies with other known invertebrate or vertebrate peptides. Next, we searched for CG4577 orthologs across Metazoa. We obtained orthologs across the Drosophilids, Brachyceran flies, and Dipterans. No orthologs were found at higher taxonomic levels, suggesting that CG4577 either originated in Dipterans or has a high sequence variability at higher taxonomic levels. To identify conserved peptide stretches indicating putative bioactive peptide sequences, we aligned the predicted amino acid sequences of the Dipteran CG4577 orthologs. This revealed several conserved peptide stretches (Figure 3—figure supplement 2) framed by canonical PC cleavage sites that might represent bioactive peptide sequences. BLAST searches against these conserved sequences did not yield hits outside of the Diptera.

To further characterize CG4577 we generated polyclonal antibodies that are predicted to recognize both CG4577-PA and CG4577-PB and investigated protein expression. CG4577 protein was expressed in a ‘sparkly’ pattern in neurons of the third instar central brain as well as in distinct cell bodies and neuronal processes in the ventral nerve cord, prompting us to name CG4577 as Sparkly (Spar) (Figure 4a and b). Co-labeling of Spar and Alk confirmed the expression of Spar in a subset of Alk-expressing cells, in agreement with our transcriptomics analyses (Figure 4a, Figure 4—figure supplement 1). In addition, we also observed expression of Spar in neuronal processes which emerge from the ventral nerve cord and appear to innervate larval body wall muscle number 8, that may be either Leukokinin (Lk) or cystine-knot glycoprotein hormone GPB5 expressing neurons (Figure 4b; Cantera and Nässel, 1992; Sellami et al., 2011). Spar antibody specificity was confirmed in both C155-Gal4>UAS-Spar-RNAi larvae, where RNAi-mediated knockdown of Spar resulted in loss of detectable signal (Figure 4c–c’), and in C155-Gal4>UAS Spar larvae, exhibiting ectopic Spar expression in the larval CNS and photoreceptors of the eye disc (Figure 4d–d’). To further address Spar expression in the neuroendocrine system, we co-labeled with antibodies against Dimm to identify peptidergic neuronal somata (Allan et al., 2005) in a Dimm-Gal4>UAS-GFPcaax background. This further confirmed the expression of Spar in Dimm-positive peptidergic neuroendocrine cells in the larval CNS (Figure 4e–e’’, Figure 4—videos 1 and 2). Moreover, co-staining of Spar and Dimm in the adult CNS showed similar results (Figure 4—figure supplement 2).

Our initial integrated analysis predicted Spar as a locus responsive to Alk signaling. To test this hypothesis, we examined Spar protein expression in w¹¹¹⁸, Alk^Y1355S_, and Alk^ΔRA genetic backgrounds, in which Alk signaling output is either upregulated (Alk^Y1355S) or downregulated (Alk^ΔRA) (Pfeifer et al., 2022). We observed a significant increase in Spar protein in Alk^Y1355S CNS, while levels of Spar in Alk^ΔRA CNS were not significantly altered (Figure 4f–h, quantified in i). In agreement, overexpression of Jeb (C155-Gal4>jeb) significantly increased Spar levels when compared with controls (C155-Gal4>UAS-GFPcaax) (Figure 4j–l, quantified in m). Again, overexpression of dominant-negative Alk (C155-Gal4>UAS-Alk^DN) did not result in significantly decreased Spar levels (Figure 4l, quantified in m). Thus activation of Alk signaling increases Spar protein levels. However, while our bulk RNA-seq and TaDa datasets show a reduction in Spar transcript levels in Alk loss-of-function conditions, this reduction is not reflected at the protein level. This observation may reflect additional uncharacterized pathways that regulate Spar mRNA levels as well as translation and protein stability, since and notably Spar transcript levels are decreased but not absent in Alk^ΔRA (Figure 3d). Taken together, these observations confirm that Spar expression is responsive to Alk signaling in CNS, although Alk is not critically required to maintain Spar protein levels.

To provide biochemical evidence for the expression of Spar, we re-analyzed data from a previous LC-MS peptidomic analysis of brain extracts from 5-day-old male control flies and flies deficient for carboxypeptidase D (dCPD, SILVER) (Pauls et al., 2019), an enzyme that removes the basic C-terminal amino acid of peptides originating from PC cleavage of the proprotein. This analysis identified several peptides derived from the Spar propeptide by mass matching in non-digested extracts from genetic control brains (Figure 5). These included peptides that are framed by dibasic prohormone cleavage sequences in the propeptide, one of which (SEEASAVPTAD) was also obtained by de novo sequencing (Figure 5). This result demonstrates that the Spar precursor is expressed and is processed into multiple peptides by PCs and possibly also other proteases. Analysis of the brain of svr mutant flies yielded similar results, but further revealed peptides C-terminally extended by the dibasic cleavage sequence (SEEASAVPTADKK, FNDMRLKR) (Figure 5), thereby confirming canonical PC processing of the Spar propeptide. Of note, the phylogenetically most conserved peptide sequence of the Spar precursor (DTQLNPADMLALVALVEAGERA, Figure 3—figure supplement 2) framed by dibasic cleavage sites was among the identified peptides yet occurred only in control but not svr mutant brains (Figure 5).

Additionally, we performed co-labeling with known Drosophila neuropeptides, pigment-dispersing factor (PDF), Dh44, insulin-like peptide 2 (Ilp2), AstA, and Lk, observing Spar expression in subsets of all these populations (Figure 6). These included the PDF-positive ventral lateral neuron (LNv) clock neurons (Figure 6a–b’’), Dh44-positive neurons (Figure 6c–d’’), a subset of Ilp2 neurons in the central brain (Figure 6e–f’’) and several AstA-positive neurons in the central brain and ventral nerve cord (Figure 6g–h’’). We also noted co-expression in some Lk-positive neurons in the central brain and ventral nerve cord, that include the neuronal processes converging on body wall muscle 8 (Figure 6i–l’’; Cantera and Nässel, 1992). Similar Spar co-expression with PDF, Dh44, Ilp2, and AstA was observed in adult CNS (Figure 6—figure supplement 1).

Since previous reports have shown that Jeb overexpression in the larval CNS results in a small pupal size (Gouzi et al., 2011), we measured pupal size on ectopic expression of Spar (C155-Gal4>Spar) and Spar RNAi (C155-Gal4>Spar RNAi), noting no significant difference compared to controls (C155-Gal4>+ and C155-Gal4>jeb) (Figure 7—figure supplement 1). These results suggest that Spar may be involved in an additional Alk-dependent function in the CNS. Further, experiments overexpressing Spar did not reveal any obvious phenotypes. To further investigate the function of Spar, we generated a Spar loss-of-function allele by CRISPR/Cas9-mediated non-homologous end-joining, resulting in the deletion of a 716 bp region including the Spar TSS and exon 1 (hereafter referred as Spar^ΔExon1) (Figure 7a). Immunoblotting analysis indicated a 35 kDa protein present in the wild-type (w¹¹¹⁸) controls that was absent in Spar^ΔExon1 mutant CNS lysates (Figure 7b; Figure 7—source data 1). The Spar^ΔExon1 mutant allele was further characterized using immunohistochemistry (Figure 7c–d’). Spar^ΔExon1 shows a complete abrogation of larval and adult Spar expression, consistent with the reduction observed when Spar RNAi was employed (Figure 7c–d’). Spar^ΔExon1 flies were viable, and no gross morphological phenotypes were observed, similar to loss of function mutants in several previously characterized neuropeptides such as PDF, drosulfakinin, and neuropeptide F (Liu et al., 2019; Renn et al., 1999; Wu et al., 2020).

A previous report noted expression of Spar in the LNv, dorsal lateral neuron (LNd), and dorsal neuron 1 (DN1) populations of adult Drosophila circadian clock neurons (Abruzzi et al., 2017; Figure 7e, Supplementary file 1). A meta-analysis of the publicly available single-cell transcriptomics of circadian clock neurons indicated that almost all adult clusters of clock neurons express Spar (Ma et al., 2021; Figure 7f). Additionally, we noted that the expression of Spar peaks around zeitgeber time 10 (ZT10) (coinciding with the evening peak of locomotor activity) (Figure 7g and h), although the differences in expression level around the clock with LD or DD cycle were not dramatic (Figure 7—figure supplement 2a–c). To confirm the expression of Spar in circadian neurons at the protein level, we co-stained Spar with a clock neuron reporter (Clk856-Gal4>UAS GFP). A subset of Spar-positive larval CNS neurons appeared to be Clk856-Gal4>UAS GFP positive (Figure 7i–j’’). Similarly, a subset of Spar-positive neurons in adults were GFP-positive (Figure 7k–l’’), confirming the expression of Spar protein in LNv clock neurons. Taken together, these findings suggest a potential function of the Alk-regulated TaDa-identified target Spar in the maintenance of circadian activity in Drosophila.

Given the expression of Spar in circadian neurons of the larval CNS, and the previous observations of a role of Alk mutations in sleep dysregulation in flies (Bai and Sehgal, 2015), we hypothesized that Spar^ΔExon1 mutants may exhibit activity/circadian rhythm-related phenotypes. To test this, we first investigated the effects of loss of Spar (employing Spar^ΔExon1) and loss of Alk (employing a CNS-specific loss-of-function allele of Alk, Alk^ΔRA, Pfeifer et al., 2022) on adult lifespan and sleep/activity behavior using the DAM (Drosophila activity monitor) system (Trikinetics Inc). Both Alk^ΔRA and Spar^ΔExon1 mutant flies displayed a significantly reduced lifespan when compared to w¹¹¹⁸ controls, with the Spar^ΔExon1 group exhibiting a significant reduction in survival at 25 days (Figure 8a). Activity analysis in Alk^ΔRA and Spar^ΔExon1 flies under 12 hr light:12 hr dark (LD) conditions indicated that both Alk^ΔRA and Spar^ΔExon1 flies exhibited two major activity peaks, the first centered around ZT0, the beginning of the light phase, the so-called morning peak, and the second around ZT12, the beginning of the dark phase that is called the evening peak (Figure 8b, black arrows). Overall activity and sleep profiles per 24 hr showed increased activity in Spar^ΔExon1 flies (Figure 8b–d, Figure 8—figure supplement 1), that was more prominent during the light phase, with an increase in the anticipatory activity preceding both the night-day and the day-night transition in comparison to Alk^ΔRA and w¹¹¹⁸ (Figure 8b, empty arrows). Actogram analysis over 30 days showed an increased number of activity peaks in the mutant groups, indicating a hyperactivity phenotype, in comparison to wild-type (Figure 8d). Furthermore, mean activity and sleep were also affected; the two mutant groups (Alk^ΔRA and Spar^ΔExon1) displayed significant variations in activity means (Figure 8e and h–h’; Figure 8—figure supplement 2). Analysis of anticipatory activity by quantifying the ratio of activity in the 3 hr period preceding light transition relative to activity in the 6 hr period preceding light transition as previously described (Harrisingh et al., 2007) failed to identify conclusive effects on anticipatory activity in Spar^ΔExon1 flies (Figure 8f and g). Furthermore, both Alk^ΔRA and Spar^ΔExon1 exhibited significant decrease in average sleep during the day per 12 hr at young ages (days 5–7) (Figure 8h, Figure 8—figure supplement 2a). In contrast, older flies (days 20–22) did not show any significant differences in sleep patterns during the day and per 12 hr (Figure 8h’, Figure 8—figure supplement 2a’). The decrease in average sleep in both Alk^ΔRA and Spar^ΔExon1 was accompanied by an increase in number of sleep bouts per 12 hr at young age (days 5–7) (Figure 8—figure supplement 2b) with no difference in number of sleep bouts at older age (Figure 8—figure supplement 2b’). Rhythmicity analysis showed that Alk^ΔRA and w¹¹¹⁸ are more rhythmic in LD compared to Spar^ΔExon1 flies (Figure 8—figure supplement 3a), however when comparing percentage of rhythmic flies among all groups the differences were not significant (Figure 8—figure supplement 3a’). Moreover, free-running period calculation by chi-square periodograms showed that both w¹¹¹⁸ and Spar^ΔExon1 flies exhibit a longer circadian period (higher than 1440 min), with 13% of the latter group having a shorter period (Figure 8—figure supplement 4a–a’). These results demonstrate that Spar is important for normal fly activity and that loss of Spar affects adult sleep/wake activity.

Since Spar^ΔExon1 flies exhibited a hyperactive phenotype during both day and night hours, we sought to investigate a potential role of Spar in regulating the endogenous fly clock by assessing fly activity after shift to dark conditions. While control flies adapted to the LD-DD shift without any effect on mean activity and sleep, Spar^ΔExon1 flies exhibited striking defects in circadian clock regulation (Figure 9a–b’, Figure 8—figure supplement 1a–d’). Comparison of average activity and sleep during 5 days of LD versus 5 days of DD (dark-dark) cycles identified a reduction in mean activity under DD conditions in Spar^ΔExon1 flies (Figure 9b–b’). Actogram profiling showed that Spar^ΔExon1 flies exhibit a hyperactive profile consistent with our previous data in LD conditions and maintain this hyperactivity when shifted into DD conditions (Figure 9c, Figure 8—figure supplement 1). Further, anticipatory peaks were largely absent on transition to DD cycle in Spar^ΔExon1 mutants with no activity peaks observed at either circadian time 0 (CT0) or at CT12 (Figure 9b, empty arrows), consistent with a significant decrease in the a.m. and p.m. anticipatory activity (Figure 8g) and altered activity and sleep bouts in these mutants (Figure 9d, Figure 8—figure supplement 2). To confirm that the circadian clock activity defects observed here were specific to loss of Spar, we conducted a targeted knockdown of Spar in clock neurons, employing Clk856-Gal4. Clk856-Gal4>Spar RNAi flies exhibited a significant disruption in both activity and sleep during the DD transition period, consistent with a hyperactivity phenotype (Figure 9e–g’, Figure 9—figure supplement 1). Further comparison of Clk856-Gal4>Spar RNAi flies relative to controls identified a consistent increase in activity in both LD and DD conditions upon Spar knockdown, with a decrease in sleep observed in DD conditions (Figure 9—figure supplement 2). These findings agree with the expression pattern of Spar in clock neurons (Figure 7), indicating a role for Spar in circadian clock regulation. Rhythmicity analysis comparing LD and DD cycles in Spar^ΔExon1 did not show a significant change indicating that Spar^Δexon1 flies are mostly rhythmic in LD and DD conditions, whereas as expected, control w¹¹¹⁸ flies were less rhythmic in DD conditions (Figure 9—figure supplement 3a–a’). This was also consistent when percentages of rhythmicity were determined, both w¹¹¹⁸ and Spar^Δexon1 flies were rhythmic (Figure 9—figure supplement 3b–b’). In terms of circadian period, the majority of w¹¹¹⁸ and Spar^Δexon1 flies exhibited a longer free running period in DD (Figure 9—figure supplement 3c–d).

Standard Drosophila husbandry procedures were followed. Flies were fed on Nutri-Fly Bloomington Formulation food (Genesee Scientific, Inc) cooked according to the manufacturer’s instruction. Crosses were reared at 25°C. The following stocks were obtained from Bloomington Drosophila Stock Center (BDSC): w¹¹¹⁸ (BL3605), Dimm-Gal4 (also known as C929-Gal4) (BL25373), Clk856-Gal4 (BL93198), and C155-Gal4 (BL458). The UAS-Spar RNAi (v37830) line was obtained from Vienna Drosophila Resource Center. Additional stocks used in this study are the following: UAS-LT3-NDam-Pol II (Southall et al., 2013), UAS-Alk^DN (P{UAS-Alk.EC.MYC}) (Bazigou et al., 2007), UAS-Jeb (Varshney and Palmer, 2006), UAS-GFPcaax (Finley et al., 1998), Alk^Y1335S (Pfeifer et al., 2022), Alk^ΔRA (Pfeifer et al., 2022), Spar^ΔExon1 (this study), UAS-Spar (this study).

Pan neuronal C155-Gal4 expressing animals were crossed with either UAS-LT3-Dam::Pol II (Control) or UAS-LT3-Dam::Pol II; UAS-Alk^EC (Alk dominant-negative sample) and crosses were reared at 25°C. Approximately 100–150 third instar larval brains were dissected in cold PBS for each technical replicate. Genomic DNA was extracted using a QIAGEN blood and tissue DNA extraction kit and methylated DNA was processed and amplified as previously described (Choksi et al., 2006; Sun et al., 2003) with the following modifications; after genomic DNA extraction, non-sheared gDNA was verified on 1.5% agarose gel, and an overnight DpnI digestion reaction set up in a 50 µl reaction volume. The digestion product was subsequently purified using QIAGEN MinElute PCR Purification Kit and eluted in 50 µl MQ water. 50 µl of DpnI digested and purified DNA was further used for adaptor ligation. Adaptor ligated DNA was amplified using the adaptor-specific primer to generate the TaDa-seq library. Amplified DNA from all experimental conditions was repurified (QIAGEN MinElute PCR Purification Kit) into 20 µl of MQ water and 200 ng aliquots were run on 1% agarose gel to verify amplification of TaDa library (DNA fragments ranging from 500 bp to 3 kb). The TaDa library was used for PCR-free library preparation followed by paired-end sequencing on an Illumina HiSeq 10× platform (BGI Tech Solutions, Hong Kong).

TaDa FASTQ paired-end reads of the control sample with three biological replicates and dominant-negative samples with two biological replicates (with two technical replicates for both control and dominant-negative samples) were obtained for a total of 10 samples and used for subsequent analysis. After base quality assessment, reads were mapped to the Dm6 reference genome of D. melanogaster using Bowtie2 (--very-sensitive-local) (Langmead and Salzberg, 2012) and post-alignment processes were performed with sam tools and BED tools (Barnett et al., 2011; Quinlan, 2014). The D. melanogaster reference sequence (FASTA) and gene annotation files were downloaded from Flybase and all GATC coordinates were extracted using fuzznuc (Rice et al., 2000) in BED format. Replicates were merged using Sambamba (merge) (Tarasov et al., 2015), and fold changes between control and dominant-negative samples, obtained by deeptools bamCompare (--centerReads --scaleFactorsMethod readCount --effectiveGenomeSize 142573017 --smoothLength 5 -bs 1) (Ramírez et al., 2014) for BIGWIG (BW) file generation. Counts of reads mapped to GATC border fragments were generated using a perl script (GATC_mapper.pl) from DamID-Seq pipeline (Maksimov et al., 2016). GATC level counts were converted to gene level counts using Bedtools (intersectBed) (Quinlan, 2014). GATC sites were merged into peaks based on a previous study (Tosti et al., 2018). Log2FC for individual GATC sites were generated using Limma for dominant-negative vs control (p<1e-5) and GATC sites were merged into peaks based on median GATC fragment size in the Drosophila genome assembly using mergeWindows (tol = 195, max.width=5000) and combineTests function from the csaw package (Lun and Smyth, 2016). Peaks were assigned to overlapping genes and filtered for FDR <0.05 and mean log2FC≥2. All peak calling and statistical analysis was performed using the R programming environment. TaDa data can also be visualized using a custom UCSC (University of California, Santa Cruz) Genome Browser session (https://genome-euro.ucsc.edu/s/vimalajeno/dm6). WebGestaltR (Liao et al., 2019) was used for GO for significantly downregulated TaDa candidates.

Previously published wild-type third instar larval brain scRNA-seq data (GSE198850) was employed (Pfeifer et al., 2022). Cellular heterogeneity was determined with eight different types of cells, including immature neurons, mature neurons, early neuroblast, NB-enriched cells, NB proliferating cells, optic lobe epithelium, Repo-positive cells and Wrapper-positive cells. The mature neuron population was divided into two groups for the current study: mature neurons and neuroendocrine cells. The neuroendocrine cell cluster was determined based on canonical markers (Guo et al., 2019; Hückesfeld et al., 2021; Nässel, 2018; Takeda and Suzuki, 2022; Torii, 2009). Subsequent analysis, including dimensionality reduction/projection or cluster visualization, and marker identification was performed using R (Seurat) (Stuart et al., 2019) and Python (Scanpy) (Wolf et al., 2018) packages. Marker genes for each cluster were identified by FindAllMarkers function (Seurat) (Stuart et al., 2019). Clusters were visualized using two-dimensional Uniform Manifold Approximation and Projection (UMAP). The top 500 significantly downregulated genes from TaDa data (FDR<0.05 and mean logFC≥2) were analyzed in the third instar larval brain scRNA-seq data. These 500 candidates were used as gene signatures, and signature enrichment analysis carried out using AUCell to determine whether a subset of the input gene set was enriched for each cell (with an enrichment threshold set at >0.196), and the clusters projected in UMAP based on the signature score (AUC score) (Aibar et al., 2017). Violin plots, dot plots, feature plots, heatmaps, and matrix plots were used to visualize gene expression in the scRNA-seq data. Functional enrichment analysis for the common significantly downregulated genes from the TaDa analysis was compared to neuroendocrine cell markers using WebGestaltR (Liao et al., 2019).

Publicly available circadian neuron scRNA-seq data (10×) from the GEO database (GSE157504) was employed to investigate expression of CG4577 in circadian neurons (Ma et al., 2021). The dataset includes two conditions: LD and DD, as well as six time points: 2 hr, 6 hr, 10 hr, 14 hr, and 22 hr. After preprocessing, 3172 and 4269 cells remained for the LD and DD samples respectively, with a total of 15,743 and 15,461 RNA features. Subsequent analysis, including integration, dimensionality reduction/projection, and cluster visualization, was performed using R (Seurat) (Stuart et al., 2019). Based on clustering, 17 clusters were defined and visualized using two-dimensional UMAP. Violin plots, dot plots, and feature plots were employed to visualize gene expression.

Relevant tissue (larval CNS or body wall muscle preparation) was dissected in cold PBS and tissues fixed in 4% formaldehyde at 4°C for 1 hr. Samples were washed three times with 0.1% PBS Triton X-100, followed by overnight incubation in 4% goat serum, 0.1% PBS Triton X-100. The following primary antibodies were used: guinea pig anti-Alk (1:1000, Lorén et al., 2003), rabbit anti-Alk (1:1000, Lorén et al., 2003), and rabbit anti-Dimm (1:1000, Allan et al., 2005), chicken anti-GFP (1:1000, Abcam #ab13970), mouse mAb anti-PDF (1:1000, DSHB: C7), rabbit anti-Ilp2 (1:1000, Veenstra et al., 2008), anti-Dh44 (1:1000, Cabrero et al., 2002), rabbit anti-AstA (1:3000, Stay et al., 1992; Vitzthum et al., 1996, Jena Bioscience GmbH), rabbit anti-Lk (1:1000, Cantera and Nässel, 1992), guinea pig anti-Spar (1:2000, this study), and Alexa Fluor-conjugated secondary antibodies were from Jackson ImmunoResearch.

Spar fluorescence intensity (Figure 4i and m) was quantified for the minimum complete confocal z-series of each third instar larval brain using Fiji (Schindelin et al., 2012). Confocal images from the 488 nm wavelength channel were analyzed as a Z project. Using a selection tool, Spar-positive areas were demarcated, and measurements recorded. Corrected total cell fluorescence (CTCF), in arbitrary units, was measured for each third instar brain as follows: CTCF  = integrated density – (area of selected cell × mean fluorescence of background readings) (Bora et al., 2021; McCloy et al., 2014). Calculated CTCFs were represented in the form of boxplots (n=12 each for w¹¹¹⁸, Alk^Y1255S, Alk^RA. n=5 each for C155-Gal4>UAS-GFPcaax and C155-Gal4>UAS-Alk^DN, n=7 for C155-Gal4>UAS Jeb).

Third instar larval brains were dissected and lysed in cell lysis buffer (50 mM Tris-Cl, pH 7.4, 250 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% Triton X-100, complete protease inhibitor cocktail, and PhosSTOP phosphatase inhibitor cocktail) on ice for 20 min prior to clarification by centrifugation at 14,000 rpm at 4°C for 15 min. Protein samples were then subjected to SDS-PAGE and immunoblotting analysis. Primary antibodies used were: guinea pig anti-Spar (1:1000) (this study) and anti-tubulin (Cell Signaling #2125, 1:20,000). Secondary antibodies used were: Peroxidase Affinipure Donkey Anti-Guinea Pig IgG (Jackson ImmunoResearch #706-035-148) and goat anti-rabbit IgG (Thermo Fisher Scientific # 32260, 1:5000).

Polyclonal antibodies against Spar (CG4577) were custom generated in guinea pigs by Eurogentec. Two Spar peptides corresponding to epitopes LQEIDDYVPERRVSS (amino acids 212–226) and PVAERGSGYNGEKYF (amino acids 432–446) of Spar-PA were injected simultaneously.

Peptidomic data from our previous study on the role of Drosophila carboxypeptidase D (SILVER) in neuropeptide processing (Pauls et al., 2019) was re-examined for the occurrence of Spar. Peptides were extracted from brains from 5-day-old male flies and analyzed on an Orbitrap Fusion mass spectrometer (Thermo Scientific) equipped with a PicoView ion source (New Objective) and coupled to an EASY-nLC 1000 system (Thermo Scientific). Three (controls) and two (mutants) biological samples (pooled brain extracts from 30 flies) were measured in technical duplicates. The raw data is freely available at Dryad (https://doi.org/10.5061/dryad.82pr5td, for details see Pauls et al., 2019). Database search was performed against the UniProt D. melanogaster database (UP000000803; 22070 protein entries) with PEAKS XPro 10.6 software (Bioinformatics Solutions) with the following parameters: peptide mass tolerance: 8 ppm, MS/MS mass tolerance: 0.02 Da, enzyme: ‘none’; variable modifications: oxidation (M), carbamidomethylation (C), pyro-glu from Q, amidation (peptide C-term). Results were filtered to 1% PSM-FDR.

To identify Spar precursor sequences in other insects and arthropods, tblastn searches with the PAM30 matrix and a low expectation threshold against the whole Drosophila Spar precursor or partial peptides flanked by canonical cleavage sites were performed against the NCBI databank (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The obtained sequences were aligned by the MUSCLE algorithm and plotted using JalView 2 (Waterhouse et al., 2009).

The Spar^ΔExon1 mutant was generated using CRISPR/Cas9 genome editing. Design and evaluation of CRISPR target sites was performed using the flyCRISPR Optimal Target Finder tool (Gratz et al., 2015). Single guide RNA targeting sequences (sequences available in Supplementary file 1) were cloned into the pU6-BbsI-chiRNA vector (Addgene, Cat. No. 45946) and injected into vasa-Cas9 (BDSC, #51323) embryos (BestGene Inc). Injected flies were crossed to second chromosome balancer flies (BDSC, #9120) and their progeny were PCR-screened for a deletion event. Mutant candidates were confirmed by Sanger sequencing (Eurofins Genomics).

UAS-Spar was generated by cloning (GeneScript) the coding sequence of CG4577-RA into EcoRI/XbaI-cut pUASTattB vector followed by injection into fly embryos (BestGene Inc) using attP1 (second chromosome, BDSC#8621) and attP2 (third chromosome, BDSC#8622) docking sites for phiC31 integrase-mediated transformation. Injected flies were crossed to second or third chromosome balancer flies, and transgenic progeny identified based on the presence of mini-white marker.

Late pupae of the indicated genotype were collected and placed on glass slides with double-sided tape. Puparium were imaged with a Zeiss Axio Zoom.V16 stereo zoom microscope with a light-emitting diode ring light and measured using Zen Blue edition software. Both female and male pupae, picked randomly, were used for measurements.

Up to 32 newly eclosed male flies were transferred into individual glass tubes containing food media (1% agar and 5% sucrose), which were each placed into a DAM2 Drosophila activity monitor (Trikinetics Inc). Monitors were then placed in a 25°C incubator running a 12:12 hr LD cycle, at a constant 60% humidity. Activity was detected by an infrared light beam emitted by the monitor across the center of each glass tube. The experiment was carried out for 1 month, and the raw binary data was acquired by the DAMSystem310 software (Trikinetics Inc). The LD/DD experiment was performed according to previously published work (Chiu et al., 2010); adult flies were first entrained for 5 days in normal LD cycle and on the last day (LD5), the light parameters were switched off and flies were then conditioned in complete DD settings for 7 days. Raw data analysis was carried out using a Microsoft Excel macro (Berlandi et al., 2017) taking into consideration 5 min of inactivity as sleep and more than 24 hr of immobility as a death event. The activity and sleep parameters are calculated for each day of the experiment as average from data of all living animals at this time and are displayed over the duration of the experiment. Calculation of anticipatory activity was performed accordingly to previously published work (Harrisingh et al., 2007) by quantifying the ratio of activity in the 3 hr preceding light transition to activity in the 6 hr preceding light transition, defined as a.m and p.m anticipation for the 6 hr period before lights on and 6 hr period before lights off, respectively. Actogram activity profile charts were generated using ActogramJ 1.0 (https://bene51.github.io/ActogramJ/index.html) and ImageJ software (https://imagej.nih.gov/ij/). ActogramJ was further used to generate the chi-square periodogram for each single fly in order to calculate the power value of rhythmicity and the percentage of rhythmic flies. All statistical analysis were performed using GraphPad Prism 8.4.2.

Schematics were generated at Biorender.com and Bioicons.com. The pipeline icon by Simon Dürr (https://twitter.com/simonduerr) is licensed under CC0 (https://creativecommons.org/publicdomain/zero/1.0/). Boxplots in Figure 3d and Figure 7—figure supplement 1 were generated using BoxplotR (http://shiny.chemgrid.org/boxplotr/). Boxplots in Figure 4 were generated using GraphPad Prism 9.