IP₃-mediated Ca²⁺ signals regulate larval to pupal transition under nutrient stress through the H3K36 methyltransferase Set2

Integration of nutrient uptake with developmental processes is essential for organismal growth and survival (Britton et al., 2002; Wang and Lei, 2018; Ward and Thompson, 2012; Koyama et al., 2020). In holometabolous insects such as Drosophila, growth occurs primarily in the larval stages and is heavily dependent on access to nutrients until the onset of pupariation, which is a nutrient-independent period of development (Nijhout, 2003). Larvae that have reached the ‘critical weight’ checkpoint can overcome protein deprivation and pupariate (Boulan et al., 2015; Mirth et al., 2005). Pupariation on a protein-deficient diet (PDD) requires acetylcholine-stimulated Ca²⁺-release from the inositol 1,4,5-trisphosphate receptor (IP₃R) in a subset of interneurons (Jayakumar et al., 2016). Rhythmic IP₃-mediated Ca²⁺ (IP₃/Ca²⁺) signals in the interneuron subset stimulate neuropeptide release for upregulation of ecdysone synthesis genes and, interestingly, also help maintain neuronal gene expression (Jayakumar et al., 2016, 2018).

Gene expression changes in neurons are primarily thought to occur by activity-driven transcription of immediate-early genes and their downstream signaling mechanisms (Chen et al., 2016; Dolmetsch, 2003). Little notice has been given to how IP₃/Ca²⁺ changes, downstream of myriad signaling pathways initiated by metabotropic receptors, regulate neuronal gene expression. Histone modifications, in response to starvation, are known to occur in Drosophila (An et al., 2017; Slade and Staveley, 2016). We report here an essential role for the Histone 3 Lys 36 methyltransferase (H3K36me) Set2 (Fig. 1A) downstream of neuronal IP₃/Ca²⁺ signals, for pupariation on a PDD. In Drosophila, as well as in other organisms (McDaniel et al., 2017), H3K36 trimethylation by Set2 (or its orthologues; Fig. S1A,B) is required for efficient transcriptional elongation (Schaft et al., 2003) and serves to upregulate gene expression (Bannister et al., 2005).

Regulation of Set2 expression by intracellular Ca²⁺ signaling was first suggested in a transcriptomic screen from the Drosophila pupal central nervous system with pan-neuronal knockdown of the endoplasmic reticulum (ER) Ca²⁺ sensor Stim (Fig. 1B; Richhariya et al., 2017). To test whether Set2 expression might be altered by Ca²⁺ release from the ER, we measured levels of Set2 mRNA in IP₃R mutant Drosophila larvae (itpr^ka1091/ug3 or itpr^ku; Joshi et al., 2004), which are known to exhibit reduced IP₃-mediated Ca²⁺ release upon stimulation of the muscarinic acetylcholine receptor (mAChR; Jayakumar et al., 2016). Larval itpr^ku brains exhibit greater than 50% reduction in levels of Set2 mRNA (Fig. 1C). A similar decrease in Set2 levels was observed in larval brains with pan-neuronal knockdown of the IP₃R (elav^C155GAL4>itpr^IR; Fig. 1C), suggesting that Set2 expression in neurons is indeed sensitive to IP₃-mediated Ca²⁺ release through the IP₃R.

The Drosophila genome encodes a single H3K36 trimethylation (H3K36me3) protein: Set2 (Fig. S1A,B; Stabell et al., 2007). Hence, the extent of H3K36me3 is a direct measure of Set2 activity. Quantitative measurement of H3K36me3 in lysates from itpr^ku mutant larval brains revealed a significant reduction in trimethylated H3K36 (Fig. 1D), indicating a reduction in Set2 protein.

From previous studies, we know that IP₃-mediated Ca²⁺ release enables larval-pupal transitions on PDD. The focus of this deficit lies in a subset of glutamatergic interneurons, vGlut^VGN6341, in the ventral ganglion (Jayakumar et al., 2016, 2018). Peripheral cholinergic inputs activate mAChR, IP₃ and Ca²⁺ signaling in vGlut^VGN6341 neurons (Fig. 1E, green) and stimulate ecdysone synthesis from the prothoracic gland through an intermediate circuit of peptidergic neurons (Fig. 1E, pink; Jayakumar et al., 2016). This signaling mechanism appears redundant on a normal diet (ND) but essential on a PDD. To assess the physiological consequences of deficient H3K36 trimethylation, we measured pupariation on a ND and a PDD, by Set2 knockdown with two RNAi lines: Set2^IR-1 and Set2^IR-2. Pan-neuronal knockdowns exhibit significant Set2 downregulation in the brain, whereas overexpression of Set2 (Set2^OE) achieved a 1.8-fold increase in expression (Fig. S1C). All wild-type (CS) larvae on a ND (Fig. 1F) and a PDD (Fig. 1G), when transferred as mid-third instars, develop normally to form pupae. Specific depletion of Set2 in vGlut^VGN6341 neurons with Set2^IR-2 has no effect on pupariation rates on a ND. In agreement with earlier findings, itpr^ku mutant larvae show a minor decrease in pupariation on a ND (Fig. 1F). However, when mid-third instar larvae with Set2 knockdown in vGlut^VGN6341 neurons were switched from a ND to a PDD, there was a significant reduction of pupariation (Fig. 1G), as observed previously for itpr^ku larvae (Jayakumar et al., 2016). Remarkably, reduced pupariation on a PDD in itpr^ku larvae could be rescued from <20% to >70% by overexpressing Set2 in vGlut^VGN6341 neurons (Fig. 1G). Thus, a systemic deficit due to loss of intracellular Ca²⁺ signaling in IP₃R mutant larvae can be compensated for by overexpression of Set2 in a subset of glutamatergic interneurons.

Set2 levels in the brains of wild-type larvae, either on a PDD or a ND, were similar (Fig. 1H). itpr^ku mutant brains demonstrate reduced Set2 mRNA levels in both dietary regimes (Fig. 1H), indicating that attenuated ER Ca²⁺ release through the IP₃R, affects the pan-neuronal expression of Set2, independently of diet.

Expression of Set2 in larval neurons, including the vGlut^VGN6341 neurons, is supported by robust H3K36me3 signals in the ventral ganglion (Fig. 2A). The cellular consequence of reduced Set2 in vGlut^VGN6341 neurons was examined next in ex vivo larval brains (Fig. 2B; Fig. S2). Calcium responses were recorded from vGlut^VGN6341 interneurons upon stimulation with carbachol (CCh), a mAChR agonist (Fig. 2C). vGlut^VGN6341 neurons from wild-type larvae exhibit a robust Ca²⁺ transient upon CCh stimulation (Fig. 2C-E; Jayakumar et al., 2016). Knockdown of either Set2 (Fig. 2C,D) or the IP₃R (Fig. 2E) in vGlut^VGN6341 neurons, significantly attenuated the Ca²⁺ response to CCh, as evident from reduced peak values and areas under the curve (AUC), in comparison with controls (Fig. 2F,G). Loss of Set2 in vGlut^VGN6341 neurons thus impedes Ca²⁺ release through the IP₃R in response to CCh stimulation.

Next, we tested Ca²⁺ responses upon overexpression of Set2 in vGlut^VGN6341 neurons. The attenuated Ca²⁺ response seen in vGlut^VGN6341>itpr^IR, was rescued back to wild-type levels by overexpression of Set2 (Fig. 2E), as evident from the peak values and AUC (Fig. 2F,G; Fig. S2). Overexpression of Set2 in wild-type vGlut^VGN6341 neurons resulted in significant increase in the evoked response to CCh when compared with wild type (Fig. 2E-G). Acute amino acid withdrawal evokes a series of slow calcium transients in vGlut^VGN6341 that are lost in itpr^ku (Fig. 2H). Initiation of these Ca²⁺ transients requires cholinergic inputs, whereas long-term propagation depends upon neuropeptide receptors (Jayakumar et al., 2018). Knockdown of Set2 (Set2^IR-2) abrogated the Ca²⁺ transients, whereas overexpression of Set2 in vGlut^VGN6341 neurons of itpr^ku restored the initial, but not the longer, response (Fig. 2H-K). These data demonstrate that Set2 rescues the CCh-mediated response in vGlut^VGN6341 neurons. The neuropeptide-mediated longer response is very likely dependent on alternate signaling mechanisms.

Expression of Set2 is regulated by IP₃/Ca²⁺ signaling (summarized in Fig. 3A). However, loss of CCh-stimulated IP₃/Ca²⁺ signals upon knockdown of Set2 followed by their rescue upon overexpression of Set2 suggests feedback regulation by Set2 of key components of IP₃/Ca²⁺ signaling in neurons. This idea was investigated next by measuring transcript levels of mAChR and itpr, two critical components of the pathway. Brains with pan-neuronal knockdown of Set2 did not exhibit a detectable change in the expression of either mAChR or itpr (Fig. 3B). Pan-neuronal knockdown of itpr resulted in a significant decrease in Set2 levels, but not mAChR, in the whole brain (Fig. 3C).

Because both the systemic (Fig. 1) and cellular phenotypes of IP₃R and Set2 (Fig. 2) derived specifically from vGlut^VGN6341 neurons, we hypothesized that Set2 might regulate mAChR and itpr expression in that specific neuronal subset. Gene expression in vGlut^VGN6341 neurons was measured by enriching them through fluorescence-activated cell sorting (FACS; Fig. 3D). A greater than 50-fold enrichment of GFP-tagged vGlut^VGN6341 neurons was achieved from all required genotypes (Fig. 3E).

Peak Ca²⁺ response of vGlut^VGN6341 neurons reduces in wild-type larvae on a PDD (Jayakumar et al., 2016). Expression of mAChR, itpr and Set2 also reduced significantly on a PDD, whereas expression of a starvation response gene, Thor (Teleman et al., 2005), increased in vGlut^VGN6341 neurons (Fig. 3G). Although attenuated, Ca²⁺ signals in vGlut^VGN6341 neurons on a PDD remain essential for pupariation, because their abrogation by knockdown of itpr prevents the larval-pupal transition (Fig. 3I; Jayakumar et al., 2016). These data support a central role for vGlut^VGN6341 neurons in development on a PDD (Jayakumar et al., 2018).

Next, we tested the reciprocal effects of Set2 knockdown and reduced Ca²⁺ signaling on expression of mAChR, itpr and Set2 in vGlut^VGN6341 neurons. Expression of both mAchR and itpr was downregulated by 0.25 and 0.3-fold, respectively, when compared with wild type in sorted vGlut^VGN6341>Set2^IR-1 neurons (Fig. 3H, purple), in agreement with the attenuated Ca²⁺ responses seen earlier (Fig. 2D). itpr knockdown in vGlut^VGN6341 neurons also reduced expression of mAChR to ∼0.5-fold (Fig. 3H, maroon). The reciprocal effects of Set2 and the IP₃R on expression of their respective genes and on mAchR, suggest a positive-feedback regulatory loop between them in vGlut^VGN6341 neurons (Fig. 3I). This idea is further supported by Set2 overexpression in vGlut^VGN6341 neurons in the presence of itpr RNAi, as seen by the significantly heightened expression of both mAchR and itpr (Fig. 3H; green). These results indicate that Set2 and H3K36 trimethylation regulate gene expression in vGlut^VGN6341 neurons, and maintain a level of responsiveness to acetylcholine through mAChR and the IP₃R that is essential for pupariation on a PDD (Fig. 3I). Reciprocal regulation of genes by Set2/H3K36me3 and IP₃/Ca²⁺ signaling may be relevant in other developmental and physiological contexts that need investigation.

To understand the breadth of neuronal gene expression changes mediated by IP₃/Ca²⁺ and Set2, we performed RNA-seq using third instar larval central nervous systems (CNSs) expressing pan-neuronal Set2 RNAi (Fig. 4A), followed by comparison with a published RNA-seq for pan-neuronal itpr RNAi from an equivalent stage (Fig. 4B) (Jayakumar et al., 2018). Set2^IR and itpr^IR conditions exhibit a greater proportion of downregulated rather than upregulated genes (Fig. 4A,B, Fig. S3A). Higher fold-change values were observed in a greater number of downregulated when compared with upregulated genes (Fig. S3A). Consequently, further analysis was performed primarily on downregulated genes. Dysregulated genes with greater fold changes in either Set2 RNAi or itpr RNAi conditions exhibited higher baseline expression as measured from their respective control genotypes (RNAi/+; Fig. S3B).

Next, we compared downregulated genes from Set2^IR and itpr^IR pan-neuronal knockdown datasets and identified 277 genes in common (Fig. 4C; P<0.05). At a lower stringency (P<0.1), the number of common genes increased to 413. Developmental expression of the 277 downregulated genes (P<0.05) in wild-type Drosophila (Celniker et al., 2009) identified four clusters with differential peaks of expression (Fig. 4D). Among these, 233 out of 277 (86%; clusters I and II) are also expressed in larvae and pupae. Set2 levels increase from mid third instar to pupariation, with consistently higher expression in the CNS (Contrino et al., 2012; Larkin et al., 2021; Fig. S3C). These analyses suggest that IP₃/Ca²⁺ signaling and Set2-mediated H3K36me3 modification work together for appropriate regulation of developmental gene expression in the larval brain, in a broader set of neurons than only vGlut^VGN6341 neurons. In agreement with this, larval brains express mAChR and Set2 in multiple clusters that include cholinergic, GABAergic and glutamatergic cells, as determined by single cell sequencing analysis (Fig. S3D-H; Avalos et al., 2019), albeit with a greater overlap (60%) between VGlut-(glutamatergic) and Set2-expressing cells (Fig. S3I).

Gene ontology and pathway analysis of the 277 common downregulated genes revealed a significant enrichment in GO categories that are likely to aid the larval-to-pupal transition (Fig. 4E), such as ecdysone synthesis (steroid hormone metabolism; Cyp9b1, Cyp4s3, Cyp6a21 and Cyp4d1), breakdown of the larval cuticle (peptidases; Nepl5, Nepl8 and Nepl18) and synthesis of the pupal/adult cuticle (chitin metabolism; Sgs4, mmy and Muc96D). Misexpression of the neprilysin family of peptidases has been implicated during impaired food intake (Hallier et al., 2016). The GO categories of DNA damage-dependent checkpoint activation (CHK kinases, CHKoV2) and transcription factors (sens and EAChm) possibly relate to the extensive changes in gene expression profiles that occur between larvae and pupae. Similar GO categories, in addition to a few others, were enriched among downregulated genes in Set2^IR-2 (Fig. S4A) and itpr^IR (Fig. S4B) datasets. Taken together, these analyses of RNA-seq data support the idea that IP₃/Ca²⁺ signals and Set2 expression allow for timely expression of multiple neuronal genes during the larval to pupal transition.

Based on existing knowledge that Set2 activity catalyzes H3K36 tri-methylation on the gene body and is frequently associated with transcriptional activation and elongation (Kizer et al., 2005; Schaft et al., 2003), we hypothesized that gene bodies in common between Set2^IR and the itpr^IR datasets should normally appear enriched in H3K36me3 marks. This was assessed using a published dataset from ChIP-on-chip performed in Drosophila ML-BG3 cells, which are derived from the 3rd instar larval CNS (Celniker et al., 2009). For this analysis, we selected commonly downregulated genes (413; P<0.1) identified from Set2 and itpr knockdown datasets (Fig. 4C). An appreciable enrichment of H3K36me3 was observed across the gene body of these genes and, as expected (Kharchenko et al., 2011; Kouzarides, 2007), we observed a drop in the flanking regions that constitute the transcriptional start and stop sites (Fig. 4F).

Next, we asked whether the severity of transcription loss upon Set2 knockdown is directly proportional to the enrichment of H3K36me3 on these genes. Interestingly, such correlation yielded two categories of genes. Approximately half the genes belong to Group 1, where higher extents of downregulation in itpr^IR and Set2^IR-2 conditions associate with enhanced enrichment of H3K36me3 along the gene body. These genes are likely to be direct targets of the catalytic activity of Set2, downstream of IP₃/Ca²⁺ signals. Group 2 represents genes with smaller fold changes and concomitant lower enrichment of H3K36me3, indicating an indirect effect of the IP₃/Ca²⁺- and Set2/H3K36me3-driven transcriptional program (Fig. 4G). In accordance with this analysis, H3K36me3 levels were enriched at the itpr and Set2 loci, as well as some candidate genes belonging to the enriched GO terms (Fig. 4E), in ML-BG3 cells and third instar larvae (Fig. S4C-H).

The IP₃R/Ca²⁺-H3K36me3 axis identified here suggests a possible mechanism by which neurons establish a brain-wide transcriptional profile, which in Drosophila larvae culminates in endocrine outputs such as ecdysone biosynthesis essential for pupariation. Several other studies suggest a link between intracellular Ca²⁺ signaling and histone modifications that regulate transcription (Sharma et al., 2014; Whitlock et al., 1983). Dysfunction of the IP₃R has been implicated in neurodegenerative conditions such as Alzheimer's, Huntington's disease and spino-cerebellar ataxias (Egorova and Bezprozvanny, 2018; Hasan and Sharma, 2020). Rodent models for Alzheimer's disease and accelerated senescence also show a decrease in brain-wide levels of H3K36 methylation (Wang et al., 2010). Post-mortem brains of individuals with Huntington's disease show association of SETD2, the mammalian ortholog of Set2, with morphological deposits linked to pathology (Passani et al., 2000), and human mutations in SETD2 have been associated with intellectual disabilities and autism spectrum disorders (Iossifov et al., 2014; Lumish et al., 2015). Future studies that investigate the IP₃/Ca²⁺- H3K36me3 link identified here in mammalian neurons could lead to a better understanding of neurodegenerative and neurological disease pathologies, and help to identify potential diagnostic markers and treatment.

Drosophila strains were grown on standard cornmeal medium consisting of 80 g corn flour, 20 g glucose, 40 g sugar, 15 g yeast extract, 4 ml propionic acid, 5 ml p-hydroxybenzoic acid methyl ester in ethanol and 5 ml ortho-butyric acid in a total volume of 1 l (ND) at 25°C under a light-dark cycle of 12 h and 12 h. The protein-deprived diet (PDD) consisted of 100 mM sucrose with 1% agar. Canton S was used as wild type throughout. Mixed sex populations were used for all experiments. Several fly stocks used in this study were sourced using FlyBase (https://flybase.org), which is supported by a grant from the National Human Genome Research Institute at the US National Institutes of Health (#U41 HG000739), the British Medical Research Council (#MR/N030117/1) and FlyBase users from across the world. The stocks were obtained from the Bloomington Drosophila Stock Centre (BDSC) supported by NIH P40OD018537. All fly stocks used and their sources are listed in Table S1.

A full-length cDNA Gold clone, LD27386, encoding Set2 (Transcript B) was obtained from the Drosophila Genomic Resource Center (DGRC). The cDNA sequence was PCR amplified using Phusion enzyme (NEB, M0530S) and appropriate primers with overhangs for NotI and KpnI restriction enzymes (sequences given in Table S2). The PCR product and pUAST-attB vector were digested with NotI and KpnI restriction enzymes (NEB,) and ligated to obtain the desired clones. The complete cDNA sequence was verified by Sanger sequencing. Set2 flies were obtained by injection of embryos employing standard methods for Drosophila transgenics by the Fly facility at NCBS, Bengaluru.

Larvae at 84±4 h post egg laying were transferred to a PDD or a ND in batches of 25 and were scored for pupariation. Data for each genotype on each media were obtained from at least six independent batches of larvae. These are reported as percentage pupariation.

Central nervous systems (CNSs) from larvae of the appropriate genotype and age were dissected in 1× phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄ and 1.8 mM KH₂PO₄) prepared in double distilled water treated with DEPC. Each sample consisted of five CNS homogenized in 500 µl of TRIzol (Ambion, ThermoFisher Scientific) per sample. At least three biological replicate samples were made for each genotype. After homogenization the sample was kept on ice and either processed further within 30 min or stored at −80°C for up to 4 weeks before processing. RNA was isolated following the manufacturer's protocol. Purity of the isolated RNA was estimated by NanoDrop spectrophotometer (ThermoFisher Scientific) and integrity was checked by running it on a 1% Tris-EDTA agarose gel.

Approximately 100 ng of total RNA was used per sample for cDNA synthesis. DNAse treatment and first-strand synthesis were performed as described previously (Pathak et al., 2015). Quantitative real time PCRs (qPCRs) were performed in a total volume of 10 µl with Kapa SYBR Fast qPCR kit (KAPA Biosystems) on an ABI 7500 fast machine operated with ABI 7500 software (Applied Biosystems). Technical duplicates were performed for each qPCR reaction. The fold change of gene expression in any experimental condition relative to wild type was calculated as 2^−ΔΔCt, where ^ΔΔCt=[Ct (target gene) –Ct (rp49)] Expt. – [Ct (target gene) – Ct (rp49)]. Primers specific for rp49 and ac5c were used as internal controls. Sequences of all primers used are provided in Table S2.

Larval CNSs of appropriate genotypes were dissected in ice-cold PBS. Between 5 and 10 brains were homogenized in 50 μl of NETN buffer [100 mM NaCl, 20 mM Tris-Cl (pH 8.0), 0.5 mM EDTA, 0.5% Triton-X-100, 1× Protease inhibitor cocktail (Roche)]. The homogenate (10-15 μl) was run on an 8% SDS-polyacrylamide gel. The protein was transferred to a PVDF membrane by standard semi-dry transfer protocols (10 V for 10 min). The membrane was incubated in the primary antibody overnight at 4°C. Primary antibodies were used at the following dilutions: rabbit anti-H3K36me3 (1:5000, Abcam, ab9050) and rabbit anti-H3 (1:5000, Abcam, ab12079). Secondary antibodies conjugated with horseradish peroxidase were used at dilution of 1:3000 (anti-rabbit HRP; 32260, Thermo Scientific). Protein was detected by a chemiluminescent reaction (WesternBright ECL, Advansta K12045-20). Blots were first probed for H3K36me3, stripped with a solution of 3% glacial acetic acid for 10 min, followed by re-probing with the anti-H3 antibody.

Third instar larval brains were dissected in insect hemolymph-like saline (HL3) [70 mM NaCl, 5 mM KCl, 20 mM MgCl₂, 10 mM NaHCO₃, 5 mM trehalose, 115 mM sucrose, 5 mM HEPES an 1.5 mM Ca²⁺ (pH 7.2)] after embedded in a drop of 0.1% low-melting agarose (Invitrogen). Embedded brains were bathed in HL3. GCaMP6m (Chen et al., 2013) was used as the genetically encoded calcium sensor. Images were taken as a time series on a xy plane at an interval of 1 s using a 20× objective with an NA of 0.7 on an Olympus FV1000 inverted confocal microscope. A 488 nm laser line was used to record GCaMP6 m Ca²⁺ measurements. All live-imaging experiments were performed with at least 10 independent brain preparations. For creating the effect of an acute loss of amino acid levels, we incubated the ex vivo preparations in 0.5×Essential Amino Acids (EAA) obtained from a 50×EAA mixture lacking glutamine (ThermoFisher Scientific) dissolved in HL3. At the point of withdrawal, the amino acid levels were diluted 10-fold using more HL3, thus creating the effect of amino acid withdrawal. For the dilution experiments, images were taken as a time series on a xy plane across six z planes at an interval of 4 s using a 10× objective with an NA of 0.4 and an optical zoom of 4 on a SP5 inverted confocal microscope mounted with a resonant scanner (Leica Microsystems). The z projection across time was then obtained as a time series.

The raw images were extracted using FIJI (based on ImageJ version 2.1.0/1.53c; Schindelin et al., 2012). ΔF/F was calculated from selected regions of interest (ROIs) using the formula ΔF/F=(F_t-F₀)/F₀, where F_t is the fluorescence at time t and F₀ is baseline fluorescence corresponding to the average fluorescence over the first 40 time frames. Mean ΔF/F time-lapses were plotted using MATLAB (R2019b; License number-1122786). A shaded error bar around the mean indicates the 95% confidence interval for CCh (50 µM) responses and the s.e.m. for amino acid withdrawal responses. Area under the curve was calculated from the point of stimulation, which was considered as 0th second for stimulation up to 300 s, using Microsoft Excel and plotted using BoxPlotR (Spitzer et al., 2014).

CNS were dissected in ice-cold PBS, fixed with 4% PFA, washed with 0.2% PTX, blocked and incubated overnight in primary chick anti-GFP antibody (1:10,000; Abcam, 13970) and rabbit anti-H3K36me3 (1:5000; Abcam, 9050). They were then washed and incubated with anti-chick AlexaFluor-488 (1:5000; #A1108, Invitrogen, RRID:AB_143165), anti-rabbit AlexaFluor-568 (1:5000; #A1108, Invitrogen, RRID:AB_143165) and DAPI, and mounted in 70% glycerol. Confocal images were obtained on the Confocal FV3000 microscope (Olympus) with a 40×, 0.7 NA objective. Images were visualized on FIJI (based on ImageJ version 2.1.0/1.53c).

Fluorescence-activated cell sorting (FACS) was used to enrich cells from larval CNS, where neurons of interest were genetically labeled with GFP using the GAL4/UAS system (Brand and Perrimon, 1993). The following genotypes were used for sorting: wild type (vGlut^VGN6341GAL4>UAS-eGFP); itpr^IR (vGlut^VGN6341GAL4>itpr RNAi); Set2^IR-1 (vGlut^VGN6341GAL4>Set2^IR-1) and itpr^IR; Set2^OE (vGlut^VGN6341GAL4>itpr RNAi; UAS-Set2). Approximately 50 third instar larval CNSs per sample were washed in 1×PBS and 70% ethanol. Larval CNSs were dissected in Schneider's medium (ThermoFisher Scientific) supplemented with 10% fetal bovine serum, 2% PenStrep, 0.02 mM insulin, 20 mM glutamine and 0.04 mg/ml glutathione. Post dissection, the larval CNS were treated with an enzyme solution [0.75 g/l collagenase and 0.4 g/l dispase in Rinaldini's solution (8 mg/ml NaCl, 0.2 mg/ml KCl, 0.05 mg/ml NaH₂PO₄, 1 mg/ml NaHCO₃ and 0.1 mg/ml glucose)] at room temperature for 30 min. They were then washed and resuspended in ice-cold Schneider's medium and gently triturated several times using a pipette tip to obtain a single-cell suspension. This suspension was then passed through a 40 mm mesh filter to remove clumps and kept on ice until sorting (less than 1 h). Flow cytometry was performed on a FACS Aria Fusion cell sorter (BD Biosciences) with a 100 mm nozzle at 60 psi. The threshold for GFP-positive cells was set using dissociated neurons from a non GFP-expressing wild-type strain (Canton S). The same gating parameters were used to sort other genotypes in the experiment. GFP-positive cells were collected directly in Trizol and then frozen immediately in dry ice until further processing.

For RNA-seq from larval CNSs, RNA isolation was performed from 15 larval CNSs of 84 h AEL larvae of both control (UAS-Set2 IR/+) and Set2 KD (vGlut^VGN6341GAL4>UAS-Set2^IR) genotypes using Trizol (ThermoFisher Scientific) following the manufacturer's protocol. Libraries with 500 ng total RNA per sample were prepared as described previously (Richhariya et al., 2017). Libraries were run on a Hiseq2500 platform. Biological triplicates were used for control and Set2 KD. itpr KD RNA-seq data (GEO Accession Number GSE109637) were obtained from previous work (Jayakumar et al., 2018) and analyzed in the same manner. 50-70 million unpaired sequencing reads per sample were aligned to the dm3 release of the Drosophila genome using HISAT2 (Kim et al., 2015, 2019) and an overall alignment rate of 95.2-96.8% was obtained for all samples. Featurecounts (Liao et al., 2014) was used to assign the mapped sequence reads to the genome and obtain read counts. Differential expression analysis was performed using two independent methods: DESeq2 (Love et al., 2014) and edgeR (Robinson et al., 2010). A fold change cutoff of a minimum twofold change was used. Significance cutoff was set at an FDR-corrected P value of 0.05 for DESeq2 and edgeR.

Volcanoplots were generated using VolcaNoseR (https://huygens.science.uva.nl/). Comparison of gene lists and generation of Venn diagrams was performed using Whitehead BaRC public tools (http://jura.wi.mit.edu/bioc/tools/). Gene ontology analysis for molecular function was performed using DAVID (Huang et al., 2009a,b). Two-hundred and seventy-seven downregulated genes common to itpr KD and Set2 KD were used as the target set and all genes in Drosophila were used as background. Developmental gene expression levels were measured for downregulated genes using FlyBase (Larkin et al., 2021) and DGET (Hu et al., 2017) (www.flyrnai.org/tools/dget/web/), and were plotted as a heatmap using ClustVis (Metsalu and Vilo, 2015) (https://biit.cs.ut.ee/clustvis/).

H3K36me3 enrichment data was obtained from a ChIP-chip dataset (ID_301) generated in Drosophila ML-DmBG3-c2 cells submitted to modEncode (Celniker et al., 2009). Enrichment scores for genomic regions were calculated using ‘computematrix’ and plotted as a tag density plot using ‘plotHeatmap’ from deeptools2 (Ramírez et al., 2016). All genes were scaled to 2 kb with a flanking region of 250 bp on either end. A 50 bp length of non-overlapping bins was used for averaging the score over each region length. Genes were sorted based on mean enrichment scores and displayed on the heatmap in descending order. For each downregulated gene, fold changes upon itpr KD, Set2 KD and relative enrichment H3K36me3 were compared and clustered using ClustVis (Metsalu and Vilo, 2015) (https://biit.cs.ut.ee/clustvis/). For each row of genes, row centering and unit variance scaling were applied prior to plotting. Genes were clustered based on Pearson's correlation and ordered based on highest mean value. Two primary clusters emerged: high H3K36me3, with greater extent of downregulation; and low H3K36me3, with a lesser extent of downregulation.

A single-cell RNA-seq dataset from 1st instar larval brains generated using 10× genomics technology by Avalos et al. (2019) (GEO Accession Number GSE134722) was analyzed. The sequenced files were processed using Cell Ranger (version 2.2.0) count and aggr, followed by subsequent analysis using Seurat version 3.0 (Butler et al., 2018; Satija et al., 2015; Stuart et al., 2019) on R version 4.0.4. The matrices produced by cell ranger were processed to produce a Seurat object. A non-linear dimensional reduction was performed to visualize the results as UMAP plots. The DotHeatMap function was used to generate expression heatmaps for genes of interest.

Amino acid sequences for different H3K36 methyltransferases belonging to yeast, human and Drosophila species were obtained from UniprotKB (https://www.uniprot.org/). Multiple sequence alignment for the sequences was performed using the alignment tool ClustalW. A phylogenetic tree was constructed based on the neighbor joining method using ClustalW2 Phylogeny tool.

Datasets were evaluated for normality using the Kolmogorov–Smirnov test. Normally distributed datasets were compared using ANOVA followed with post-hoc Tukey test. Non-parametrically distributed datasets were evaluated using the Kruskall–Wallis test, followed with the Mann–Whitney U-test for pairwise comparisons. Pairwise comparisons for parametrically distributed datasets were performed using a two-tailed Student's t-test. Exact P values are listed in Table S3.

We thank the Central Imaging and Flow Cytometry Facility (CIFF, NCBS) for maintenance and use of microscopes, and the Drosophila facility (Flyfacility, NCBS) for stock maintenance and development of transgenics. We also thank members of the Notani Lab for discussions relating to ChIP-Seq analysis.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199018