Olfactory learning in Drosophila requires O-GlcNAcylation of mushroom body ribosomal subunits

O-GlcNAcylation is a dynamic post-translational modification that diversifies the proteome. Its dysregulation is associated with neurological disorders that impair cognitive function, and yet identification of phenotype-relevant candidate substrates in a brain-region specific manner remains unfeasible. By combining an O-GlcNAc binding activity derived from Clostridium perfringens OGA (CpOGA) with TurboID proximity labeling in Drosophila, we developed an O-GlcNAcylation profiling tool that translates O-GlcNAc modification into biotin conjugation for tissue-specific candidate substrates enrichment. We mapped the O-GlcNAc interactome in major brain regions of Drosophila and found that components of the translational machinery, including many ribosomal subunits, were abundantly O-GlcNAcylated in the mushroom body, the computational center of the Drosophila brain. Hypo-O-GlcNAcylation induced by ectopic expression of active CpOGA in the mushroom body decreased local ribosomal activity, leading to olfactory learning deficits that could be rescued by increasing ribosome biogenesis. Our study reveals that O-GlcNAcylation contributes to the links between protein synthesis and cognitive function in the brain learning center, and provides a useful tool for future dissection of tissue-specific functions of O-GlcNAcylation in Drosophila.


Introduction
Protein O-GlcNAcylation is a ubiquitous post-translational modification that occurs on thousands of nuclear and cytoplasmic proteins, conveying various stimuli or stressors such as fluctuating nutrient levels to distinct cellular processes [1][2][3] . It involves reversible attachment of -N-acetylglucosamine (GlcNAc) to the hydroxyl group of serine and threonine residues of protein substrates, catalyzed by a pair of evolutionarily conserved enzymes, O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) 4 . As a monosaccharide modification, the addition and removal of O-GlcNAc moiety are dynamic, with cycling rates ranging from several minutes to the lifetime of a protein 5,6 . By modifying different protein substrates, O-GlcNAcylation exerts critical regulatory functions in a wide range of basic cellular processes, including transcription, translation, and protein homeostasis 1,7 . O-GlcNAcylation is ubiquitously distributed but more abundant in some tissues, such as the brain 8,9 . Given its enrichment in brain tissues and essential biological functions, it is not surprising that O-GlcNAc cycling is required for the development and functions of central nervous system 2,10,11 , and its dysregulation is linked to numerous neurological disorders 7,10,12,13 .
O-GlcNAc homeostasis appears to be required for proper cognitive function, although the molecular connections between the dysregulated O-GlcNAcome and cognitive impairment are not fully understood. Hypomorphic mutations of OGT are implicated in an X-linked intellectual disability syndrome [14][15][16][17][18] , a severe neurodevelopmental disorder now termed OGT-associated Congenital Disorder of Glycosylation (OGT-CDG) 19 . Drosophila models of OGT-CDG that carry the equivalent human disease-related OGT missense mutations manifest deficits in sleep and habituation, an evolutionarily conserved form of non-associative learning 20 . Our recent work has shown that decreased O-GlcNAcylation level in Drosophila, induced through overexpression of a bacterial OGA from Clostridium perfringens (CpOGA), leads to a deficit of associative olfactory learning. More interestingly, ectopic expression of CpOGA during early embryogenesis results in reduced brain size and learning defect in adult flies, likely due to interference of the sog-Dpp signaling required for neuroectoderm specification 21 . These studies reveal that disturbed O-GlcNAc homeostasis can impact cognitive function by compromising neuronal development.
On the other hand, a number of studies have revealed that impaired O-GlcNAcylation is implicated in aging-related neurodegenerative diseases such as Alzheimer's disease (AD) 7,10,12,13,22 . In the cerebrum of AD patients, O-GlcNAcylation levels are significantly lower than that of healthy controls 23 . Increased O-GlcNAcylation levels by limiting OGA activity recovers the impaired cognitive function in AD mice models 24,25 . Interestingly, during normal aging in mice, reduction of O-GlcNAcylation levels also occurs in the hippocampus, and elevation of neuronal O-GlcNAc modification ameliorates associative learning and memory 26 . These results indicate that, in addition to its involvement in neurodevelopoment, O-GlcNAc homeostasis is also required for normal neuronal activity and cognitive function. However, the identity of key O-GlcNAc protein substrates supporting the cognitive abilities in adult brain and their spatial distribution remain largely unknown.
An obstacle for comprehensively identifying the O-GlcNAc conveyors of various cognitive functions is the lack of an effective tissue-specific O-GlcNAc profiling method. Given the structural diversity and relatively low abundance, enrichment of O-GlcNAc modified proteins is required for mass spectrometry (MS)-based profiling of O-GlcNAcylation 27 . The enrichment strategies roughly fall into two categories.
One category involves direct capture of O-GlcNAcylated proteins by antibodies or lectins that recognize the GlcNAc moiety [27][28][29][30][31][32][33] . O-GlcNAc antibodies including RL2 and CTD110.6, as well as O-GlcNAc-binding lectins such as wheat germ agglutinin (WGA), are commonly used for enrichment. In addition, the catalytic-dead mutant of CpOGA that retains the ability to recognize O-GlcNAcylated substrates was successfully repurposed to concentrate many developmental regulators from Drosophila embryo lysates 34 . Another category of enrichment strategies relies on chemoenzymatic or metabolic labeling [27][28][29][30][31][32][33] . Azido-modified intermediates, such as N-azidoacetylglucosamine (GlcNAz) and N-azidoacetylgalactosamine (GalNAz), are used to introduce specific tags (e.g. biotin) to protein substrates via Staudinger ligation or click chemistry, allowing for capture and enrichment of O-GlcNAcylated proteins. A recent study coupled the O-GlcNAc-binding lectin GafD to the proximity labeling TurboID yielding the GlycoID tool 35  Here we generated transgenic Drosophila lines that allow specific expression of CpOGA in different brain regions. Ectopic expression of CpOGA in the major learning center of Drosophila brain, the mushroom body, reduced local O-GlcNAcylation levels and impaired olfactory learning. We further combined a catalytically incompetent CpOGA mutant (CpOGA CD ) with the proximity labeling enzyme TurboID to develop an O-GlcNAcylation profiling tool. By conditional expression of this tool to translate O-GlcNAc modification into biotin conjugation in specific brain structures, we mapped the O-GlcNAc interactomes and generated an O-GlcNAc atlas for different brain regions of Drosophila (tsOGA, http://kyuanlab.com/tsOGA/). Particularly, we detected abundant O-GlcNAc modifications associated with ribosomes in the mushroom body. Lowering the mushroom body O-GlcNAcylation levels reduced the ability of ribosomes to synthesize new proteins, interfering with olfactory learning, which could be reversed by increasing ribosomal biogenesis via overexpression of dMyc. We propose that compromised ribosomal activity in the brain learning center contributes to the cognitive deficits of O-GlcNAcylation insufficiency-associated neurological diseases.
CpOGA DM , which carries two point-mutations (D298N and D401A) that inactivate both the catalytic and binding activities toward O-GlcNAc modification, was used as a control. We dissected brains from the adult flies and validated tissue-specific expression patterns via immunostaining. As expected, Elav-Gal4 induced CpOGA WT expression in the whole brain ( Figure 1B We then evaluated the cognitive ability of these flies using olfactory learning assay as previously reported [37][38][39] . To rule out the possibility that overexpression of CpOGA WT or CpOGA DM differentially disrupted odor preference, we tested their olfactory acuity toward either 4-methylcyclohexanol (MCH) or octanol (OCT) using air as a control.
Tissue-specific expression of CpOGA WT or CpOGA DM in the antennal and optic lobes generated differences in susceptibility, and these flies were not included in subsequent olfactory learning tests ( Figure S1A and S1B). Flies expressing CpOGA WT or CpOGA DM in brain neurons, mushroom body, or ellipsoid body were trained to associate electric shock punishment with an air current containing MCH or OCT, and then tested for the ability to remember the electric shock-associated odor using a T-maze apparatus ( Figure S1C). Compared to CpOGA DM , conditional expression of CpOGA WT in brain neurons or mushroom body compromised the ability to establish the association between odor and electric shock ( Figure 1F), suggesting that decreased O-GlcNAcylation levels in these brain regions resulted in a deficit in olfactory learning. In contrast, flies expressing CpOGA WT or CpOGA DM in the ellipsoid body, as well as the control flies without a Gal4 driver, showed no statistical difference in the learning performance ( Figure 1F). Ectopic expression of CpOGA WT in the mushroom body driven by OK107-Gal4 impacted neuronal development during the larval stages. To directly investigate whether perturbation of O-GlcNAcylation compromised neuronal function in adult flies, we used the temperature-sensitive Gal80 (Gal80 ts ) to restrict CpOGA expression until adulthood ( Figure S1D). This temporally controlled expression of CpOGA WT specifically in the adult mushroom body did not affect the odor acuity but significantly disrupted olfactory learning relative to CpOGA DM control ( Figure 1G, S1A, and S1B). These results suggested that proper O-GlcNAcylation homeostasis is essential for the mushroom body function.

O-GlcNAcylation profiling through CpOGA proximity labeling
The mushroom body is known to be the associative learning center in Drosophila brain 40,41 . Having discovered that O-GlcNAcylation homeostasis in the mushroom body was critical for olfactory learning, we developed an O-GlcNAc profiling method that allows identification of candidate O-GlcNAcylated protein substrates in this brain region. Mutation of the catalytic residue Asp298 to Asn (D298N) of CpOGA Once induced by different tissue-specific drivers, this tool could tag and enrich O-GlcNAc substrates and their interactors in a tissue-specific manner, as endogenous protein biotinylation level is low in most organisms including Drosophila.
As proof of concept, we generated stable HEK293T cells expressing TurboID-CpOGA CD or its reference construct TurboID-CpOGA DM . To characterize labeling activity, treatment with 10 M or 100 M biotin from an aqueous stock was first applied on these cells for 60 min, and the cell lysates were subject to western blot  Figure 2B). We therefore identified 336 O-GlcNAc candidate substrates from HEK293T cells (Table S1). To compare this result with known O-GlcNAc modifications, we compiled two lists of the previously identified O-GlcNAcylated proteins in HEK293T cells via either direct capture 46,47 or chemoenzymatic labeling methods [48][49][50][51][52] (Table S2). Gene ontology (GO) analysis on these three datasets showed that they were enriched in similar biological processes ( Figure S2E). Overlap analysis revealed that 52% (178/336) of the O-GlcNAc candidate substrates identified in our study were also present in previous reports ( Figure 2D). 48 proteins were shared among the three lists (Table S3), encompassing many well-known O-GlcNAcylated proteins such as OGT, NUP153, NUP62, and HCFC1. Protein-protein interaction networks of these 48 proteins highlighted four cellular component clusters: the MLL1 complex, nuclear pores, COPII vesicle coats, and cytoplasmic stress granules ( Figure   2E). Additionally, of the 158 candidate proteins that were unique in our result, 113 were annotated as O-GlcNAcylation substrates in the O-GlcNAc database (www.oglcnac.mcw.edu). These results validated that TurboID-CpOGA CD was able to effectively tag O-GlcNAcylated proteins with biotin for enrichment and identification.

Region-specific O-GlcNAcylation profiling of Drosophila brain
We next generated transgenic flies harboring UAS-TurboID-CpOGA CD or UAS-TurboID-CpOGA DM via C integrase-mediated site-specific recombination.
To test biotinylation efficiency, we used Da-Gal4 to drive ubiquitous expression and raised the flies on biotin-containing food (100 M) from early embryonic stage to adulthood according to previous reports 36,53 ( Figure 3A). Flies were homogenized and equal amounts of lysate were used in immunoprecipitation experiments. Similar to the result with HEK293T cells, TurboID-CpOGA CD catalyzed more biotinylation in the input relative to TurboID-CpOGA DM , and more biotinylated proteins were immunoprecipitated, in which strong O-GlcNAcylation signals were detected ( Figure   3B). To validate whether TurboID-CpOGA CD could achieve brain region-specific labeling of O-GlcNAcome with biotin tag, we selected different Gal4 to drive TurboID-CpOGA CD in distinct brain regions and fed the flies with biotin.
Whole-mount staining of the brains showed that TurboID-CpOGA CD displayed specific expression patterns as expected. More importantly, staining with streptavidin-Cy3 detected strong biotinylation in the brain regions expressing TurboID-CpOGA CD , whereas the rest of the brain showed negligible background signals ( Figure 3C).
Subsequently, we immunoprecipitated biotinylated proteins from these fly brain lysates using streptavidin magnetic beads and performed MS analysis to identify putative O-GlcNAc substrates in different brain regions. Proteins with higher LFQ O-GlcNAcylation modifications in the mushroom body were highly clustered in processes linked to translation, including cytoplasmic translation, translational initiation, ribosome assembly, and ribosome biogenesis. To eliminate possible interference caused by varying abundance of these candidate proteins in different brain regions, we normalized the calculated O-GlcNAc level (log2 FC) of each substrate using its corresponding brain region-specific normalizing factor generated from the single-cell transcriptome atlas of the adult Drosophila brain 3 ( Figure S3E).
For ease of search and use, we created an online database for tissue-specific O-GlcNAcylation Atlas of Drosophila Brain (tsOGA, http://kyuanlab.com/tsOGA/) to host these datasets ( Figure S3F).

O-GlcNAcylation affects cognitive function of Drosophila by regulating ribosomal activity in the mushroom body
The GO analysis revealed that ribosomes were enriched in the mushroom body O-GlcNAc interactome. We calculated the percentage of ribosomal components in all the proteins identified for different brain regions, and found that nearly 10% of the putative O-GlcNAc substrates in the mushroom body were from ribosomes, much higher than that in other brain regions ( Figure S4A). To validate that the observed enrichment was not due to higher expression levels of these ribosomal subunits in the mushroom body, we plotted the normalized O-GlcNAc levels of the putative ribosomal substrates alongside their mRNA abundances in different brain regions.
While the O-GlcNAc levels were highest in the mushroom body, their mRNA abundances were not ( Figure 4A). Moreover, in the mushroom body, the O-GlcNAc levels of these ribosomal proteins showed no correlation with their mRNA abundances ( Figure S4B).
To directly verify whether mushroom body ribosomes were hyper-O-GlcNAcylated, Flag-tagged RPL13A, a core component of the large ribosomal subunit, was expressed in brain neurons or specifically in mushroom body, driven by Elav-Gal4 or OK107-Gal4 respectively. Intact ribosomes were then isolated from these brain regions by anti-Flag immunoprecipitation 54 Figure 1D and 1F). We next investigated whether this cognitive phenotype was due to compromised ribosomal activity. To this end, we selected a panel of representative ribosomal components that were significantly O-GlcNAcylated in the mushroom body, and performed RNA interference (RNAi)-mediated knockdown. The RNAi induced by Da-Gal4 reduced the expression of the targeted ribosomal genes to varying degrees ( Figure S4C). We then crossed the RNAi lines to OK107-Gal4 to drive specific knockdowns in mushroom body, and conducted olfactory learning assay with these flies. Downregulation of RPL11 and RPL24 in the ribosomal large subunit, and RPS3 and RPS6 in the ribosomal small subunit did not alter olfactory acuity ( Figure S4E-S4F), however, they led to compromised olfactory learning ability ( Figure S4D). Consequently, we reasoned that upregulation of ribosomal activity might ameliorate the cognitive defect caused by CpOGA WT -induced hypo-O-GlcNAcylation. To test this, we increased ribosome biogenesis by overexpression of dMyc [56][57][58] , and observed that dMyc expression in mushroom body could restore local protein synthesis and rescue the hypo-O-GlcNAcylation induced olfactory learning deficit ( Figure 4C-4E).

Discussion
Protein O-GlcNAcylation is controlled by a very simple system consisting of only two enzymes, OGT and OGA. Yet it can dynamically modify more than 5000 protein substrates in different tissues to regulate their stability, protein-protein interactions, New protein synthesis is known to be required for formation and consolidation of long-term memories 69,70 . Several ribosomopathies, such as Diamond-Blackfan anemia and distal trisomy 5q, are associated with learning disabilities 71  In addition, recent studies have revealed that stress granules are tightly linked with autism spectrum disorders 74 . The enrichment of stress granule components in the O-GlcNAc substrate list suggests that O-GlcNAcylation dysregulation might be involved in autism as well. We anticipate that this study will galvanize further studies into targeting O-GlcNAcylation insufficiency to ameliorate cognitive defects commonly seen in many neurological diseases. TurboID-CpOGA CD/DM fragments were cloned into pUASz-HS-HA vectors respectively using Gibson assembly (NEB). Constructs with the attB sequence were injected into flies (y1, w67c23; P(CaryP) attP2) to initiate the C31 integrase-mediated site-specific integration (UniHuaii). The resulted adult flies (G0) were crossed to double balancer to get the F1 generations.

Olfactory learning and memory
Behavioral experiments were carried out in an environmental chamber at 25°C and 70% humidity as previously described 37  If the experimental group flies have similar odor avoidance to that of control, they will be used for subsequent olfactory learning test.
After confirming that the flies to be tested have avoidance behavior in response to electric shock, flies were trained to associate an aversive odor (MCH or OCT) used as a conditioned stimulus (CS) with electric shock. The experiment comprised two phases: the flies were trained in the first phase, and the trained flies were tested in the second phase. During training, approximately 100 flies were simultaneously exposed to odor 1 (CS + ) and electric shock (60 V) in a training tube for 1 min. Then, they were exposed to the blank odor (air) for 1 min before receiving odor 2 (CS -) without electric shock for 1 min, followed by the blank odor (air) for 1 min. Immediately after and multiplied by 100%.

PI =[n(CS -)-n(CS + )]/[n(CS + )+n(CS -)]×100%.
In each experiment, we calculated the mean PI from two trials: one in which MCH was the shock-paired odor, and the other in which OCT was the shock-paired odor.
This method removed any potential bias caused by the flies having a stronger preference for one odor over the other. Therefore, each point in the bar graph  Figure S2C and D, cells were cultured in the medium supplemented with 25 μM Thiamet-G (Selleck, s7213) or 25 μM OSMI-1(Sigma, SML1621) for 6 h before lysis. For the experiment in Figure 4D, the gel was stained with Fast Silver Stain Kit (Beyotime, P0017S).

Immunoprecipitation
For the immunoprecipitation experiment in Figure 2C

Measurement of Protein Synthesis
The protein synthesis in fly brains was assessed using the Click-iT Plus OPP Alexa

RT-qPCR
RNA was extracted from flies using TRIzol (Life Technologies, 87804), and 1 μg total RNA was reverse transcribed to generate cDNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, K1621). The cDNA was then used as templates and qPCR was performed using the SYBR Green qPCR Master Mix (Solomon Biotech, QST-100) on the QuantStudio3 Real-Time PCR system (Applied Biosystems). The expression levels for each gene were normalized to Actin. Detailed information about the primers was listed in Table S9.

Protein Identification by LC-MS/MS
The HEK293T cells (2 × 10 7 cells per sample) and fly brains (~200 fly brains that expressed TurboID-CpOGA CD/DM in brain neurons per sample,~800 fly brains that expressed TurboID-CpOGA CD/DM in other brain structures per sample, three biological replicates) were immunoprecipitated with streptavidin magnetic beads as described above. The supernatants were used for SDS-PAGE separation and minimally stained with Coomassie brilliant blue (Solarbio, C8430-10g). The gels were cut into small pieces, and reduced and alkylated in 10 mM DTT and 55 mM IAA (Merck, I6125) respectively. For digestion, 0.5 µg sequencing-grade modified trypsin was added and incubated at 37 ℃ overnight. The peptides were then collected, desalted by StageTip (Thermo Fisher Scientific, 87782) and resolved in 0.1% formic acid before analysis by mass spectrometry. Mass spectrometry analysis was performed using Q Exactive HF-X mass spectrometer (Thermo Fisher Scientific) coupled with Easy-nLC 1200 system. Mobile phases A and B were water and 80% acetonitrile, respectively, with 0.1% formic acid. Protein digests were loaded directly onto an analytical column (75 µm × 15 cm, 1.9 µm C18, 1 µm tip) at a flow rate of 450 nL/min. Data were collected in a data-dependent manner using a top 25 method with a full MS mass range from 400 to 1400 m/z, 60,000 resolutions, and an AGC target of 3 × 10 6 . MS2 scans were triggered when an ion intensity threshold of 4 × 10 5 was reached. A dynamic exclusion time of 30 sec was used. Ions with charge state 6-8 and more than 8 were excluded.

Data analysis
The raw data were imported into the MaxQuant software to identify and quantify the proteins. The following parameters were used: trypsin for enzyme digestion; oxidation of methionine, acetylation of the protein N terminus, biotinylation of lysine and protein N terminus and HexNAc (ST) as variable modifications; carbamidomethyl (C) as fixed modification. We used the canonical human protein database (containing 20379 reviewed protein isoforms) or Drosophila melanogaster protein database (containing 22088 protein isoforms, including reviewed and unreviewed sequences) for database searching separately. The false discovery rate (FDR) was 1% for peptide-spectrum matches (PSM) and protein levels. For the proteomics data of different brain regions of Drosophila, we used label-free quantitation (LFQ) to determine the relative amounts of proteins among 3 replicates.
Perseus software was used to filtered out all contaminates identified by MaxQuant (contaminant proteins, reversed proteins, proteins only identified by site). A pseudo-count of 1 was added to protein intensities in order to avoid taking log of 0.
We generated log2 Fold Change (log2 FC) values for each protein in the TurboID-CpOGA CD group relative to the TurboID-CpOGA DM control. For the proteomics data of HEK293T cell, only proteins identified with at least 2 peptides were considered for further analysis. Proteins were considered as O-GlcNAcylated substrates when differences in log2 FC of TurboID-CpOGA CD group with relative to the TurboID-CpOGA DM control were higher than 1. For the proteomics data from different brain regions of Drosophila, only proteins identified with at least 2 peptides and in at least 2 of the 3 replicates of TurboID-CpOGA CD were included for further analysis. A two-tailed unpaired student's t-test was applied in order to determine the statistical significance of the differences. Proteins were considered as O-GlcNAcylated substrates when differences in log2 FC of TurboID-CpOGA CD group with relative to the TurboID-CpOGA DM control were higher than 1 or statistically significant (p < 0.05).
To adjust the intereference caused by varying abundance of the putative O-GlcNAc substrates in different brain regions, single-cell transcriptomic data of the entire adult Drosophila brain (GEO: GSE107451) 3 was used to generate a normalizing factor for each substrate. Briefly, the annotated cell clusters were categorized into different brain regions. Then, the average mRNA expression level of each gene within a certain brain region was calculated. The normalizing factor was defined as the ratio of the average mRNA expression level of a given gene in neurons from a specific brain region to the average mRNA expression level of the same gene in neurons from the whole brain (Table S10). The normalized O-GlcNAc level was generated as the O-GlcNAc level (log2 FC) of a putative O-GlcNAcylated protein divided by its normalizing factor in a certain brain region (Table S11).

Website
The website was created to browse through the O-GlcNAc database (www.kyuanlab.com/tsOGA), using the database management system Centos and the Uwsgi web framework. Backend servers were developed by Python programming language (version 3.7). GNU/Linux Debian-based systems with gunicorn (Python http) and NginX were used for development and production of the website. The website search function was based on MySQL database.

Quantification and Statistical Analysis
To quantify fluorescent intensities in different Drosophila brain regions, whole brain images were stitched together using the stitching algorithm in ZEN software (Zeiss), and maximum intensity projection was produced. The images were then analyzed using ImageJ software. Mean fluoresecent intensity of the whole brain or ROI was measured, and the relative fluorescent intensity was calculated as a ratio of the mean fluorescent intensity in ROI to that of the whole brain.
GO enrichment analyses of O-GlcNAcome in HEK293T cells and Drosophila were performed using DAVID. Protein-protein interaction (PPI) network of O-GlcNAcome in HEK293T cells was performed using STRING. GraphPad Prism was used for statistical analysis and the student's t-test was used to determine statistical significance. Bubble plots, pie plots and bar graphs were created using Hiplot, venn plots were created using jvenn.

Data availability
The accession numbers for the mass spectrometry data were PXD040547 and PXD040412 on the ProteomeXchange Consortium PRIDE partner repository.

Materials availability
All cells and fly strains generated in this study are available upon request to the lead contact (see above).

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Kai Yuan (yuankai@csu.edu.cn). Table   Table S1. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from HEK293T cells. Table S2. Previously identified O-GlcNAcylated proteins from HEK293T cells ， related to Figure 2D and S2E. Table S3. 48 proteins shared among the three datasets, related to Figure 2D and Figure   2E. Table S4. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from brain neuron of Drosophila. Table S5. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from mushroom body of Drosophila. Table S6. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from antennal lobe of Drosophila. Table S7. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from ellipsoid body of Drosophila. Table S8. O-GlcNAcylated proteins identified by TurboID-CpOGA CD from optic lobe of Drosophila .   Table S9. Sequences of all the primers used in this study. Table S10. Cell clusters in different brain regions generated from single-cell transcriptomic data.

Declaration of interests
The authors declare no competing interests.          Raw data of all western blots from Figure S2. Complete and uncropped membranes of all western blots from Figure S2.   Raw data of all western blots from Figure 3.         unpaired t-test, the stars indicate significant differences (***p < 0.001, **p < 0.01, *p < 0.05, and ns, not significant, p ≥ 0.05). Error bars represent SD. Excel spreadsheet containing source data used to generate Figures S4A-F.