Intellectual disability-associated disruption of O-GlcNAc cycling impairs habituation learning in Drosophila

Abstract

O-GlcNAcylation is a reversible co-/post-translational modification involved in a multitude of cellular processes. The addition and removal of the O-GlcNAc modification is controlled by two conserved enzymes, O-GlcNAc transferase (OGT) and O-GlcNAc hydrolase (OGA). Mutations in OGT have recently been discovered to cause a novel Congenital Disorder of Glycosylation (OGT-CDG) that is characterized by intellectual disability. The mechanisms by which OGT-CDG mutations affect cognition remain unclear. We manipulated O-GlcNAc transferase and O-GlcNAc hydrolase activity in Drosophila and demonstrate an important role of O-GlcNAcylation in habituation learning and synaptic development at the larval neuromuscular junction. Introduction of patient-specific missense mutations into Drosophila O-GlcNAc transferase using CRISPR/Cas9 gene editing leads to deficits in locomotor function and habituation learning. The habituation deficit can be corrected by blocking O-GlcNAc hydrolysis, indicating that OGT-CDG mutations affect cognition-relevant habituation via reduced protein O-GlcNAcylation. This study establishes a critical role for O-GlcNAc cycling and disrupted O-GlcNAc transferase activity in cognitive dysfunction, and suggests that blocking O-GlcNAc hydrolysis is a potential strategy to treat OGT-CDG.

Author summary

Attachment of single N-acetylglucosamine (GlcNAc) sugars to intracellular proteins has recently been linked to neurodevelopment and cognition. This link has been strengthened by discovery of O-GlcNAc transferase (OGT) missense mutations in intellectual disability. Most of these mutations lie outside the catalytic O-GlcNAc transferase domain and it is unclear how they affect cognitive function. Using the fruit fly Drosophila melanogaster as a model organism, we found that a balance in O-GlcNAc cycling is required for learning and neuronal development. Habituation, a fundamental form of learning, is affected in flies that carry patient-specific OGT mutations and increasing O-GlcNAcylation genetically corrects the habituation deficit. Our work establishes a critical role for O-GlcNAc cycling in a cognition-relevant process, identifies defective O-GlcNAc transferase activity as a cause of intellectual disability, and proposes underlying mechanisms that can be further explored as treatment targets.

Introduction

O-GlcNAcylation is an essential and dynamic co-/posttranslational modification that is characterized by the attachment of an N-acetylglucosamine (GlcNAc) molecule to serine or threonine residues of intracellular proteins. O-GlcNAcylation is implicated in a wide range of cellular processes, such as: chromatin remodeling [1–3], transcription [4,5] and translation [6], Ras-MAPK and insulin signaling [7–9], glucose homeostasis [10], mitochondrial trafficking [11], and control of the circadian clock [12]. The addition and removal of the O-GlcNAc modification, termed O-GlcNAc cycling, is controlled by two evolutionarily conserved enzymes, O-GlcNAc transferase (OGT) and O-GlcNAc hydrolase (OGA).

OGT, responsible for the addition of O-GlcNAc, is abundantly expressed in neurons and is enriched in the postsynaptic density (PSD) [13], a protein-dense structure that organizes the postsynaptic signaling machinery. O-GlcNAcylation is altered in brains of patients with Alzheimer’s disease [14] and animal and cellular models of major neurodegenerative diseases [15–19]. In C. elegans models of neurodegeneration, O-GlcNAcylation protects against neurotoxicity [20]. Furthermore, it plays an important role in neuronal regeneration through the synchronization of insulin signaling-dependent regenerative processes [8]. Recent studies also point to important functions in neuronal development, such as neuronal differentiation [21,22], assembly and axonal transport of neurofilaments [23], and synapse maturation [24]. Missense mutations in human OGT gene, located on the X chromosome, are associated with intellectual disability (ID) [25–31], a severe neurodevelopmental disorder that is characterized by impaired cognition. Patients with OGT mutations suffer from a wide array of clinical features, including intrauterine growth retardation, developmental delay, delayed or restricted language skills and severe learning difficulties. The syndrome has been termed OGT-associated Congenital Disorder of Glycosylation (OGT-CDG) [32]. These findings and animal studies suggest that O-GlcNAcylation plays an important function in cognitive processes, such as learning [33,34].

The OGT protein consists of an N-terminal tetratricopeptide (TPR) domain, which contributes to substrate recognition and binding [35] and a C-terminal catalytic domain that is responsible for glycosylation of the TPR-bound proteins. OGT-CDG mutations have been found in both domains. Unlike the mutations in the catalytic domain, the mutations in the TPR domain have not been shown to significantly affect global protein O-GlcNAc levels but they do affect the OGT-TPR domain substrate binding and glycosylation kinetics, as derived from in vitro assays and crystal structure analysis [26,36,37]. However, it remains to be known whether impaired glycosylation is the mechanism that leads to developmental and cognitive defects caused by identified TPR domain mutations. This question is pertinent as the TPR domain has been shown to be essential for cellular functions other than glycosylation [38].

Drosophila as a model organism has contributed to understanding the disease pathogenesis of numerous (ID) syndromes. It offers a combination of well-established gene targeting approaches and a plethora of morphological and functional disease-relevant phenotypes. The Drosophila OGT orthologue is encoded by the polycomb group gene super sex combs (sxc) [39]. It is highly similar to human OGT [40]. Complete loss of sxc results in severe homeotic transformations of adult body structures and pupal lethality [41]. This lethality can be restored by ubiquitous overexpression of human OGT, demonstrating functional conservation [39]. sxc has also been associated with neuronal function in circadian rhythm regulation [42]. Therefore, Drosophila is highly suited to investigate the disrupted mechanisms underlying OGT-CDG.

Here we genetically manipulated both O-GlcNAc transferase and O-GlcNAc hydrolase activity using established sxc and Oga mutants and investigated the effect of O-GlcNAc cycling on cognitive function and neuronal development. We turned to habituation, an evolutionary conserved form of non-associative learning that is characterized by response decrement towards a repeated, non-meaningful stimulus. At the neuronal level, habituation is mediated by adaptive changes in the excitatory activity of the stimulus response pathway, causing attenuated downstream neuronal responses to repeated, familiar stimuli. Habituation thus serves as a filter mechanism that prevents information overload and allows cognitive resources to focus on relevant stimuli [43]. It represents a prerequisite for higher cognitive functions [44–46]. Deficits in habituation have been reported in a number of neurodevelopmental disorders [47] and in more than a hundred Drosophila ID models [48,49]. Synaptic morphology at the Drosophila neuromuscular junction, an established model synapse [50], was assessed as a measure of neuronal development. To independently validate our findings we targeted the endogenous sxc locus by CRISPR/Cas9 editing [51] and generated a strong hypomorph mutation in the O-GlcNAc transferase domain (sxc^H596F). With the same technique we introduced three ID-associated missense mutations found in the conserved TPR domain of OGT: R284P [28], A319T [25] and L254F [29] and generated equivalent sxc^R313P, sxc^A348T and sxc^L283F alleles, respectively. We evaluated their effect on protein O-GlcNAcylation, developmental viability, adult lifespan and locomotor activity, and assessed their effect on cognitive function and neuronal development.

We find that appropriate O-GlcNAc cycling is required for habituation, a fundamental form of learning that is widely disrupted in Drosophila models of ID, and for synaptic development. We show that OGT missense mutations, implicated in ID, outside the catalytic O-GlcNAc transferase domain lead to deficits in habituation learning and that these deficits are caused by disrupted O-GlcNAc transferase activity. We thus unambiguously demonstrate the role of O-GlcNAcylation in brain development and function.

Results

Alteration of O-GlcNAc transferase activity leads to a deficit in habituation

To investigate whether the catalytic activity of Drosophila OGT is required for cognitive function, we tested the effect of an sxc mutation with diminished catalytic activity, sxc^H537A, on habituation. The homozygous hypomorphic flies are viable and except mild wing vein and scutellar bristle phenotypes do not present with any morphological abnormalities [51]. In habituation, an initial strong response towards a repeated but harmless stimulus gradually wanes based on prior experience. To assess this phenotype we used light-off jump habituation, an established non-associative learning assay that meets the strict habituation criteria, including spontaneous recovery and dishabituation with novel stimulus and excluding sensory adaptation and motor fatigue [46,52]. We subjected sxc^H537A homozygous (sxc^H537A/H537A), heterozygous (sxc^H537A/+), and genetic background control flies (+/+) to 100 light-off pulses in the light-off jump habituation assay. While sxc^H537A/+ and control flies exhibited good initial jump responses to the light-off stimuli (61% and 67% initial jumpers out of N = 96 tested flies per genotype; above a required threshold of 50% [49]), sxc^H537A/H537A flies were impaired (36% initial jumpers, N = 64), identifying broader defects that preclude assessment of habituation. Compared to control flies that habituated quickly to the repeated light-off stimulus, sxc^H537A/+ flies displayed significantly slower habituation and needed significantly more light-off pulse trials to suppress their jump response (Trials To no-jump Criterion, TTC). Some mutant flies were not able to suppress their jump response during the entire course of the experiment, as reflected by the high baseline of the average jump response curve (Fig 1A). These results suggest that partial loss of O-GlcNAc transferase activity or altered O-GlcNAcylation kinetics in sxc^H537A/+ mutants impairs habituation.

Fig 1. Catalytic activity of O-GlcNAc transferase in neurons is required for habituation learning.

Jump responses were induced by 100 light-off pulses with 1 s intervals between pulses. The jump response represents the % of jumping flies in each light-off trial. The mean number of trials that flies needed to reach the no-jump criterion (Trials To Criterion, TTC) ± SEM is also shown. (A) Defective habituation of sxc^H537A/+ flies (N = 59, mean TTC ± SD: 12.8 ± 6.7 p = 0.001, in red) compared to their respective genetic background control flies (+/+, mean TTC ± SD: 4.4 ± 0.9, N = 61, in blue). ** p<0.01, based on lm analysis. (B) Habituation defect of sxc^H537A/+ flies (N = 72, mean TTC ± SD: 15.5 ± 7.8 p_adj = 0.045, in red) is corrected by removing one Oga allele in sxc^H537A/+; Oga^KO/+ flies (N = 70, mean TTC ± SD: 10.1 ± 5.9, p_adj = 0.024, in cyan) to the level of control flies (N = 72, mean TTC ± SD: 9.4 ± 3.5, p_adj = 0.677, in blue). Habituation of Oga^KO/+ flies (N = 76, mean TTC ± SD: 17.1 ± 10.4, in purple) is slower but not significantly different from the control flies (p_adj = 0.467). (C) Defective habituation of elav-Gal4>UAS-sxc flies (N = 55, mean TTC ± SD: 12.8 ± 4, p_adj = 3.89x10^-12, in dark blue) compared to control elav-Gal4/+ flies (N = 38, mean TTC ± SD: 2.6 ± 1.3 in light blue). (D) Habituation defect of sxc^H537A/+; elav-Gal4/+ flies (N = 40, mean TTC ± SD: 14.6 ± 6, p_adj = 4.92x10^-12, in red) is corrected by selective expression of UAS-sxc in neurons (sxc^H537A/+; elav-Gal4>UAS-sxc, N = 43, mean TTC ± SD: 4.2 ± 1.3, p_adj = 8.94x10^-7, in green), to the level of the genetic background control flies (+/+, N = 65, mean TTC ± SD: 4.4 ± 0.9, p_adj = 0.68, in blue). * p_adj<0.1, *** p_adj<0.001, n.s. not significant, based on lm analysis with Bonferroni-Holm correction for multiple comparisons. A complete list of p-values and summary statistics is provided in S3 Table.

We validated our conclusion by employing a knockout-out allele of Drosophila Oga [53]. We asked whether partial inhibition of O-GlcNAc hydrolysis, by removing one copy of Oga (Oga^KO/+) could improve habituation of the sxc^H537A/+ flies. Habituation of Oga^KO/+ flies was also slower but not significantly different from the control flies. The transheterozygous sxc^H537A/+; Oga^KO/+ flies showed good initial jump responses, and their habituation was not significantly different to that of control flies, identifying a significant improvement compared to sxc^H537A/+ flies (Fig 1B). The fatigue assay (see Materials and Methods) confirmed that the lower TTC values were not a result of increased fatigue (S2A Fig). These results show that Drosophila is a suitable model to study the role of OGT in cognitive functioning and demonstrate that tight control of protein O-GlcNAcylation is required for proper habituation learning in Drosophila.

O-GlcNAc transferase is required for habituation in neurons

We next sought to determine whether the sxc^H537A/+ habituation deficit originates from reduced OGT function in neurons. We therefore induced neuronal knockdown of sxc by crossing the pan-neuronal elav-Gal4 driver line (see Materials and Methods) to an inducible RNAi line against sxc obtained from Vienna Drosophila Resource Center (#18610, zero predicted off-targets). Progeny from crossing the driver line to the isogenic genetic background of the RNAi line (#60000) were used as controls. While control flies (elav-Gal4/+) showed good initial jump response (56%, N = 64), elav-Gal4>UAS-sxc^RNAi flies—similar to sxc^H537A/H537A flies—exhibited very low initial jump response to light-off stimulus (19% initial jumpers, N = 96). While this detrimental effect prevented assessment of sxc neuron-specific knockdown in habituation learning, it does argue that i) the failed jump response of sxc^H537A/H537A flies is likely due to loss of OGT activity in neurons, and ii) OGT activity is indispensable for basic neuronal function or neuronal development.

We used an alternative strategy to test whether habituation deficits of sxc^H537A/+ flies are of neuronal origin. We asked whether restoration of OGT activity in neurons can correct habituation deficits of sxc^H537A/+ flies, by inducing pan-neuronal overexpression of functional wild-type sxc in the heterozygous sxc^H537A/+ as well as control background. Neuronal overexpression of functional sxc in control flies (elav-Gal4>UAS-sxc) resulted in a habituation deficit (Fig 1C). This is consistent with our previous findings of habituation deficits in Oga^KO/KO flies, which also show increased protein O-GlcNAcylation [53]. In contrast, re-expression of functional sxc in the sxc^H537A/+ flies completely corrected their habituation deficits (Fig 1D). The lower TTCs were not a result of increased fatigue (S2B Fig). Therefore, appropriate levels of OGT activity and O-GlcNAcylation, specifically in neurons, are required for habituation learning.

Neuronal O-GlcNAc transferase activity controls synaptic development

Synapse biology is important for brain development and cognition, and abnormalities in synaptic architecture are characteristic of multiple Drosophila models of neurodevelopmental disorders [54–59]. For these reasons, we asked whether sxc^H537A/+ mutants show any defects in the morphology of the third instar larval neuromuscular junction (NMJ), a well-established model synapse. We labeled NMJs by immunostaining for the postsynaptic membrane marker anti-discs large (Dlg1) to visualize the overall morphology of the NMJ terminal, and for synaptotagmin (Syt), a synaptic vesicle marker that visualizes synaptic boutons. We did not observe a significant change in synaptic length, area, or perimeter (Fig 2A), nor in the number of branches and branching points (S3A Fig) in NMJs of sxc^H537A/+ mutant larvae but we observed a significant increase in bouton number compared to the genetic background control (Fig 2A). Increasing OGT activity by presynaptic overexpression of wild-type sxc (elav-Gal4>UAS-sxc) resulted in an opposite phenotype: a decreased number of boutons compared to both controls, UAS transgene and driver alone. NMJ length was also reduced (Fig 2B). Both parameters were normalized when we neuron-specifically expressed functional sxc in the sxc^H537A/+ mutant background (sxc^H537A/+, elav-Gal4>UAS-sxc) (Fig 2C). These results demonstrate a role of sxc in the synaptic bouton number and NMJ morphology and indicate that tight control of O-GlcNAcylation is important for normal synaptic development.

Fig 2. Both reduced and increased O-GlcNAc transferase activity in neurons cause defects in synaptic morphology.

Data presented as individual data points with mean ± SD. (A) NMJs on muscle 4 of sxc^H537A/+ larvae have a significantly higher number of synaptic boutons (N = 29, in red) as compared to their genetic background control (+/+, N = 24, p = 0.009, in blue,) but not significantly different NMJ length (p = 0.872), area (p = 0.314) and perimeter (p = 0.935). ** p<0.01, based on one-way ANOVA. (B) Elav-Gal4>UAS-sxc larvae have a significantly lower number of boutons (N = 29, in dark blue) compared to their respective background controls elav-Gal4/+ (N = 30, p_adj = 2.1x10^-4, in light blue) and UAS-sxc/+ (N = 26, p_adj = 0.009, in grey), significantly reduced NMJ length (p_{adj/elav-Gal4} = 0.013, p_adj/UAS-sxc = 0.02) and a smaller NMJ perimeter (p_adj/UAS-sxc = 0.008). (C) Neuron-selective expression of sxc in sxc^H537A/+ larvae shows no change in bouton numbers (sxc^H537A/+; elav-Gal4>UAS-sxc (N = 29, in green)), compared to sxc^H537A/+, UAS-sxc/+ larvae (N = 28, p_adj = 0.102, in red) and compared to control (N = 27, p_adj = 0.977, in blue) and no change in NMJ length compared to control (+/+, N = 28, p_adj = 0.974) and sxc^H537A/+, UAS-sxc/+ larvae (p_adj = 0.935). Also NMJ area (p_adj/control = 0.849, p_{adj/sxcH537A; UAS-sxc/+} = 0.691) and NMJ perimeter were not changed (p_adj/control = 0.066), p_{adj/sxcH537A; UAS-sxc/+} = 0.995). (D) Number of synaptic boutons (p_adj = 0.003), NMJ length (p_adj = 0.01), area (p_adj = 0.028) and perimeter (p_adj = 4.5x10^-4) are significantly increased in sxc^H596F/H596F larvae (N = 31, in brown) compared to the genetic background control larvae (+/+, N = 28, in blue) and partially normalized in the sxc^H596F/H596F; Oga^KO/KO larvae (N = 30, in cyan, boutons: p_adj/sxcH596F = 0.011, p_adj/control = 0.923; length: p_adj/sxcH596A = 0.896, p_adj/control = 0.001; area: p_adj/sxcH596A = 0.501, p_adj/control = 0.431; perimeter: p_adj/sxcH596A = 0.08, p_adj/control = 0.26). None of the parameters is significantly affected in the Oga^KO larvae (N = 30, in purple; Oga^KO experiments were performed simultaneously and first published here [53] with significantly increased bouton counts (p <0.05) without multiple testing correction). * p_adj <0.05, ** p_adj <0.01, *** p_adj <0.001, based on one-way ANOVA with Tukey’s multiple comparisons test. A complete list of p-values and summary statistics is provided in S3 Table. (A’- D’) Representative NMJs of wandering third instar larvae labeled with anti-discs large 1 (Dlg, magenta) and anti-synaptotagmin (Syt, green). When appropriate, type 1b synapses are distinguished from other synapses with white arrow. Scale bar, 20μm. The quantitative parameter values of the representative images: (A’) (+/+ | sxc^H537A/+): #Boutons (31 | 39), Length (103.7 | 137.8), Area (374.4 | 369.4), Perimeter (245.5 | 306.7) (B’) (elav-Gal4/+ | UAS-sxc/+ | Elav-Gal4>UAS-sxc): #Boutons (45 | 37 | 28), Length (160.0 | 137.5 | 93.3), Area (480.4 | 478.4 | 363.9), Perimeter (407.6 | 290.8 | 243.9) (C’) (+/+ | sxc^H537A/+, UAS-sxc/+ | sxc^H537A/+; elav-Gal4>UAS-sxc): #Boutons (27 | 34 | 31), Length (107.6 | 144.9 | 114.9), Area (430.6 | 464.7 | 361.7), Perimeter (256.0 | 354.8 | 299.4) (D’) (+/+ | sxc^H596F/H596F | Oga^KO/KO | sxc^H596F/H596F; Oga^KO/KO): #Boutons (23 | 43 | 35 | 30), Length (111.5 | 205.6 | 122.9 | 142.5), Area (351.7 | 691.7 | 417.8 | 377.6), Perimeter (245.6 | 544.0 | 274.1 | 318.9).

The decrease of synaptic length caused by neuronal overexpression of sxc in the control but not sxc^H537A/+ background indicates that strong dysregulation of sxc might be required to uncover its function in synaptic growth. Accordingly, we found a significant increase in NMJ length and perimeter in homozygous sxc^H537A larvae. The increase in the number of synaptic boutons did not reach significance (S3D Fig), potentially due to greater variability. This effect also did not withstand multiple testing correction in the sxc^H537A/+; UAS-sxc/+ larvae that were used as a control in the rescue experiment (Fig 2C). However, there is a quantitatively very consistent increase in bouton number across the three tested H537A mutant conditions (fold changes 1.19, 1.18 and 1.15). We therefore conclude that the bouton phenotype associated with H537A mutation is mild. To validate the synaptic bouton and length/growth phenotypes, we used CRISPR/Cas9 gene editing to generate a stronger catalytic hypomorph, sxc^H596F (S1A and S1B Fig). The in vitro catalytic activity of sxc^H596F was reported to be 3% relative to wildtype OGT activity, less than the reported catalytic activity of sxc^H537A (5.6% activity relative to wildtype) [40]. Indeed, we found total O-GlcNAc levels in sxc^H596F homozygous embryos to be reduced (S4A Fig). sxc^H596F homozygous flies are viable, confirming minimal requirement of endogenous OGT activity for completion of development in Drosophila (S4B Fig). We found that NMJs of sxc^H596F larvae display a significant increase in synaptic bouton number, NMJ length, area and perimeter (Fig 2D), reflecting a more severe NMJ phenotype and indicating the effect of sxc^H537A on synaptic morphology is mild and/or not fully penetrant. We also subjected the sxc^H596F flies to light-off jump habituation assay but homozygous as well as heterozygous sxc^H596F flies showed impaired jump response (38% and 42% initial jumpers), similar to homozygous sxc^H537A and pan-neuronal sxc knockdown flies. This precluded the assessment of habituation.

A knockout of Oga normalized bouton number and partially also the area and perimeter of the sxc^H596F NMJs (Fig 2D). These data show that O-GlcNAcylation controls bouton number and partially also NMJ size.

Characterization of development and locomotor function of sxc mutations associated with Intellectual Disability

Recent studies have reported three hemizygous missense mutations (R248P, A319T, L254F) in human OGT in male individuals with ID. The de novo R248P mutation was identified by trio whole exome sequencing in an affected individual with ID and developmental delay [28]. A319T and L254F mutations were identified by X chromosome exome sequencing. The A319T mutation, present in three individuals with severe ID, was inherited from the mother but segregated with an uncharacterized missense mutation in MED12, a gene already implicated in ID [25]. The L254F mutation was present in three related individuals with moderate to mild ID [27,29]. These mutations reside in the conserved TPR domain, outside of the catalytic O-GlcNAc transferase domain [25,27–29]. To investigate the functional consequences of these mutations, we introduced the equivalent missense mutations (R313P, A348T, L283F) into the sxc gene using CRISPR/Cas9 editing (S1 Fig) and generated three novel sxc ID alleles sxc^R313P, sxc^A348T, and sxc^L283F.

We first characterized the development of the patient-related mutant sxc alleles. We transferred embryos at stage 11–16 to vials with fresh food and counted the number of resulting pupae and adult flies. Homozygous sxc^R313P, sxc^A348T, and sxc^L283F embryos developed normally to adulthood without apparent delay and the percentage of pupae and adults did not statistically differ from the genetic background controls (S5A Fig).

We next investigated locomotor phenotypes in adult sxc^R313P, sxc^A348T, and sxc^L283F flies using the island and negative geotaxis assays. In the island assay, flies were thrown onto a white platform surrounded by water, and the number of individuals remaining on the platform was quantified over time. Homozygous sxc^R313P, sxc^A348T and sxc^L283F flies escaped from the platform with similar efficiency as the genetic background control (S5B Fig), indicating that their startle response is not affected.

In the negative geotaxis assay, climbing performance of homozygous and heterozygous sxc^R313P flies was significantly slower while homozygous sxc^A348T and sxc^L283F flies showed an average climbing speed similar to the control (Fig 3A). We also tested Drosophila lines with homozygous H537A (sxc^H537A) or H596F (sxc^H596F) catalytic hypomorph mutations to investigate whether impaired O-GlcNAc transferase activity affects climbing speed in the negative geotaxis assay. The catalytic hypomorphs exhibited similar climbing speed as the control group (Fig 3A). This suggests that the deficit in coordinated locomotor behavior in the negative geotaxis assay of sxc^R313P flies is independent of O-GlcNAc transferase activity.

Fig 3. Locomotor and biochemical characterization of sxc^R313P, sxc^A348T and sxc^L283F flies.

(A) Climbing locomotor behaviour was assessed based on the climbing speed (mm/s) in an automated negative geotaxis assay. The sxc^R313P/+ and sxc^R313P (N = 9) flies showed reduced climbing speed compared to background control (N = 9) indicating locomotor dysfunction. sxc^A348T, sxc^L283F, sxc^H537A and sxc^H596F flies (N = 9 for all genotypes) did not show significantly reduced climbing speed. Data presented as mean ± SD. * p_adj < 0.05 based on one-way ANOVA with Tukey’s multiple comparisons of mean climbing speed. A complete list of p-values and summary statistics is provided in S3 Table. (B) Western blot on head samples from 1–4 days old male adult Drosophila indicate no significant alteration in the level of protein O-GlcNAcylation in sxc^R313P, sxc^A348T and sxc^L283F samples compared to the genetic background controls, while the homozygous Oga^KO allele caused an increase of O-GlcNAcylation. Western blot was probed with a monoclonal anti-O-GlcNAc antibody (RL2). (C) Quantification of O-GlcNAcylated proteins revealed that protein O-GlcNAcylation in sxc^R313P, sxc^A348T and sxc^L283F flies remain at a similar level as in the control samples. Data presented as mean ± SD. * p_adj = 0.035, based on one-way ANOVA with Tukey’s multiple comparisons test, n = 3 for all lines. A complete list of p-values and summary statistics is provided in S3 Table.

Taken together, similar to catalytic hypomorphs ([51] and S2B Fig) the sxc^R313P, sxc^A348T, and sxc^L283F mutants are fully viable and develop normally to adulthood. Neither the reduction of protein O-GlcNAcylation induced by catalytic hypomorph alleles nor the sxc^A348T and sxc^L283F ID alleles cause severe locomotor defects in adult flies. Only the sxc^R313P allele negatively affects climbing performance. This effect appears to be independent of O-GlcNAc transferase activity. However, a contribution of a potential second site mutation affecting another gene that was not eliminated by six generations of backcrossing cannot be formally excluded.

Patient-related sxc mutant alleles do not affect global protein O-GlcNAcylation

We investigated whether the ID-associated alleles in sxc affect protein O-GlcNAcylation by subjecting lysates from adult heads of sxc^R313P, sxc^A348T, and sxc^L283F flies to Western blotting. Labeling with anti-O-GlcNAc antibody (RL2) that is able to capture O-GlcNAcylation changes in the catalytic hypomorphs (S4B Fig and [40]) indicated that the levels of protein O-GlcNAcylation are not significantly altered in each of the three mutants (Fig 3B and 3C). This is in line with normal levels of O-GlcNAcylation in patient-derived fibroblasts and human embryonic stem cell models of the R313P and L283F equivalent mutations [26,28]. O-GlcNAcylation of the human A348T equivalent has not been investigated. The cell models of OGT-CDG mutations show downregulation of Oga, which may compensate for decreased O-GlcNAcylation. Although existence of such regulatory mechanism has not been shown in Drosophila, we analyzed the O-GlcNAcylation levels in Oga knockout background. Blocking O-GlcNAc hydrolysis in patient-related sxc mutant alleles with Oga^KO increased O-GlcNAc levels to the same degree as in Oga^KO samples (Fig 3B and 3C). We thus conclude that flies carrying ID-associated sxc mutations do not have a grossly affected protein O-GlcNAcylation.

sxc^R313P and sxc^A348T display defective habituation learning

Because of their role in ID in humans, we also investigated the effect of the novel sxc^R313P, sxc^A348T, and sxc^L283F alleles on habituation learning. We first subjected the sxc^L283F flies to 100 light-off pulses in the habituation assay. Despite sufficient locomotor abilities to perform in the island test and negative geotaxis assays, the initial jump response of the sxc^L283F homozygous and heterozygous flies was below the required threshold of 50% therefore deemed non-performers (sxc^L283F/L283F: 36% initial jumpers, N = 96; sxc^L283F/+: 47% initial jumpers, N = 96). Insufficient performance at the beginning of the assay thus precluded the assessment of habituation in these flies. We observed the same phenotype also for the sxc^R313P homozygous flies (49% initial jumpers, N = 64). The initial response in sxc^R313P heterozygous flies was sufficient (67%) and they were not able to suppress their jump response to the repeated light-off stimuli as efficiently as the genetic background control flies (Fig 4A), revealing a learning deficit. Flies heterozygous for the sxc^A348T allele showed a good initial jump response and habituated similar to the control, while sxc^A348T homozygous flies showed a habituation deficit (Fig 4B). In summary, deficits in habituation learning were observed for the R313P (heterozygous) and A348T (homozygous) mutations, while evaluation of the L254F homo- and heterozygous as well as R313P homozygous conditions was precluded by a poor initial jump response.

Fig 4. Assessment of sxc^R313P and sxc^A348T flies in habituation learning.

Jump responses were induced by 100 light-off pulses with 1 s interval between pulses. The jump response represents the % of jumping flies in each light-off trial. The mean number of trials that flies needed to reach the no-jump criterion (Trials To Criterion, TTC) ± SEM is also shown. (A) Deficient habituation of sxc^R313P/+ flies (N = 73, p_adj = 6.18x10^-6, in red) compared to their respective genetic background controls (control, N = 65, in blue). (B) Deficient habituation of sxc^A348T/A348T flies (N = 76, p_adj = 2.1x10^-14, in brown) and no significant habituation deficit of sxc^A348T/+ flies (N = 72, p_adj = 0.095, in red) compared to the genetic background control (control, N = 65, in blue). (C) Deficient habituation of sxc^R313P/+ flies (N = 81, p_adj = 2.63x10^-6, in red) is restored in sxc^R313P/+; Oga^KO/+ flies (N = 53, p_adj = 4.89x10^-5, in cyan). (D) Deficient habituation of sxc^A348T/A348T flies (N = 79, p_adj = 8.84x10^-10, in brown) is restored in sxc^A348T/A348T; Oga^KO/KO flies (N = 62, p_adj = 1.07x10^-4, in cyan). (E) Deficient habituation of sxc^R313P/+ flies (N = 81, p_adj = 2.63x10^-6, in red) is restored in sxc^R313P/+; Oga^D133N/+ flies (N = 64, p_adj = 9.09x10^-6, in cyan). (F) Deficient habituation of sxc^A348T/A348T flies (N = 79, p_adj = 8.84x10^-10, in brown) is restored in sxc^A348T/A348T; Oga^D133N/D133N flies (N = 56, p_adj = 3.74x10^-7, in cyan). *** p_adj<0.001, n.s. not significant, based on lm analysis with Bonferroni-Holm correction for multiple comparisons. A complete list of p-values and summary statistics is provided in S3 Table.

Blocking O-GlcNAc hydrolysis corrects habituation deficits of sxc^R313P and sxc^A348T

The apparently unaltered levels of O-GlcNAcylation in ID-associated sxc mutants (Fig 3B and 3C) may suggest that O-GlcNAc-independent mechanisms underlie their cognitive phenotypes. To test this experimentally, we performed the habituation assay in flies carrying sxc^R313P allele and either the Oga^KO allele or an Oga mutation that specifically blocks its O-GlcNAc hydrolase activity (Oga^D133N). Notably, heterozygous Oga^KO and Oga^D133N flies habituated to a similar degree as the genetic background control flies. When introduced into an sxc^R313P/+ background, Oga^KO and Oga^D133N alleles fully rescued defects seen in the sxc mutants alone (Fig 4C and 4D). Similarly, we attempted a rescue of habituation deficient homozygous sxc^A348T with homozygous Oga^KO and Oga^D133N alleles. We have previously shown that these homozygous Oga mutants also exhibit habituation deficits [53]. Strikingly, blocking O-GlcNAc hydrolysis by Oga^KO/KO or Oga^D133N/D133N in sxc^A348T/A348T flies was sufficient to completely rescue habituation deficits of either single mutant condition (Fig 4E and 4F). All tested flies show a good initial jump response and lower TTCs in the rescue experiments were not caused by fatigue (S4 Fig and S2 Table). In summary, despite seemingly normal gross O-GlcNAc levels in the sxc^R313P and sxc^A348T ID alleles, these genetic experiments provide evidence that their deficits in habituation learning depend on defective OGT enzymatic activities.

sxc^R313P and sxc^A348T and sxc^L283F show deficits in synaptic morphology

To determine whether synaptic phenotypes seen in larvae with hypomorphic catalytic domain mutations are recapitulated in larvae with patient mutations we assessed synaptic morphology at the NMJs of homozygous sxc^R313P and sxc^A348T and sxc^L283F larvae. We found that the NMJs of larvae carrying any of the three patient-related sxc mutant alleles display a significant increase in synaptic bouton number and sxc^R313P larvae also display significantly increased length and perimeter of the NMJ. NMJ length is also increased in the sxc^A348T and sxc^L283F larvae albeit not significantly (Fig 5A). In addition, sxc^R313P and sxc^L283F larvae show an increased number of NMJ branches (S5C Fig). Overall, the NMJ morphology phenotypes of the patient-related sxc mutant alleles resemble those of the catalytic hypomorphs sxc^H537A/+ (Fig 2A) and sxc^H596F/H596F (Fig 2D). This is in line with the observation that the patient-related sxc^R313P and sxc^A348T alleles affect O-GlcNAc transferase activity, which is indispensable for habituation (Fig 4C and 4F). The shared NMJ phenotype signature between the catalytic and patient-related mutants is the increase of synaptic bouton number. Because R313P and L283F mutations also significantly affect other NMJ parameters, we conclude that they are stronger/more detrimental than A348T mutation. This is in line with the observed effect on the jump performance in the light-off jump habituation assay (sxc^L283F/L283F, sxc^L283F/+ and sxc^R313P/R313P non-performers).

Fig 5. Synaptic morphology of sxc^R313P, sxc^A348T and sxc^L283F.

Data presented as individual data points with mean ± SD. (A) NMJs on muscle 4 of sxc^R313P, sxc^A348T and sxc^L283F larvae have a significantly higher number of synaptic boutons (sxc^R313P: N = 21, p = 6.8x10^-5, in red; sxc^A348T: N = 21, p = 0.011, in blue; sxc^L283F: N = 20, p = 0.0225) compared to the control (+/+, N = 42, in grey). sxc^R313P larvae have also significantly higher NMJ length (p = 0.0125) and perimeter (p = 0.001).). * p_adj <0.05, ** p_adj <0.01, *** p_adj <0.001, based on one-way ANOVA with Tukey’s multiple comparisons test. A complete list of p-values and summary statistics is provided in S3 Table. (A’) Representative NMJs of wandering third instar larvae labeled with anti-discs large 1 (Dlg, magenta) and anti-synaptotagmin (Syt, green). When appropriate, type 1b synapses are distinguished from other synapses with white arrow. Scale bar, 20μm. The quantitative parameter values of the representative images (+/+ | sxc^R313P | sxc^A348T | sxc^L283F): #Boutons (20 | 36 | 29 | 37), Length (84.5 | 113.8 | 107.4 | 128), Area (399 | 447.9 | 360.9 | 349.9), Perimeter (254.2 | 329.5 | 313.8 | 302.4).

The Drosophila O-GlcNAc proteome is enriched in genes with function in neuronal development, learning & memory, and human ID gene orthologs

ID-associated mutations in sxc do not globally reduce protein O-GlcNAcylation yet blocking O-GlcNAc hydrolysis can completely restore the learning deficits in light-off jump habituation. We hypothesized that altered O-GlcNAcylation of specific sxc substrates is responsible for the habituation deficits. We therefore attempted to predict candidate substrates by exploration of the Drosophila O-GlcNAc proteome, as previously determined through enrichment with a catalytically inactive bacterial O-GlcNAcase [60]. We first performed an enrichment analysis of neuronal and cognitive phenotypes (as annotated in Flybase, see Materials and Methods) among the Drosophila O-GlcNAc substrates (encoded by in total 2293 genes) and found that they are significantly enriched in phenotype categories learning defective (Enrichment = 1.5, p_adj = 0.032), memory defective (Enrichment = 1.5, p_adj = 0.016), neurophysiology defective (Enrichment, 1.5, p_adj = 3.2x10^-5) and neuroanatomy defective (Enrichment = 1.9, p = 5.27x10^-35) (genes listed in S4 Table), supporting the importance of O-GlcNAcylation for neuronal development and cognitive function. We also found that human orthologs of 269 genes from the O-GlcNAc proteome are proven or candidate monogenic causes of ID (Enrichment = 1.7, p_adj = 1.01x10^-16). When restricting this analysis to proteins with high-confidence mapped O-GlcNAc sites (in total 43) [60], we found orthologs of nine O-GlcNAcylated proteins to be implicated in Intellectual Disability, again representing a significant enrichment (Enrichment = 3.4, p_adj = 0.01). These orthologs are: Atpalpha (human ATP1A2), Gug (human ATN1), Hcf (human HCF1), LanA (human LAMA2), mop (human PTPN23), NAChRalpha6 (human CHRNA7), Ndg (human NID1), Nup62 (human NUP62) and Sas-4 (human CENPJ). They represent potential downstream effectors of sxc that may control habituation learning in the wild-type condition and may contribute to habituation deficits in catalytic and ID-associated sxc mutant conditions. Further analysis will be required to answer the question whether impaired O-GlcNAc transferase activity towards one or more of these targets is responsible for habituation deficits that are associated with OGT-CDG mutations.

Discussion

O-GlcNAcylation is important for habituation and for neuronal development in Drosophila

Habituation, the brain’s response to repetition, is a core element of higher cognitive functions [44–46]. Filtering out irrelevant familiar stimuli as a result of habituation allows to focus the cognitive resources on relevant sensory input. Abnormal habituation was observed in a number of neurodevelopmental disorders, including ID and Autism [47] and characterizes > 100 Drosophila models of ID [48,49]. To address the role of OGT and its O-GlcNAc transferase activity in this cognition-relevant process, we investigated heterozygous sxc^H537A/+ flies [51] in light-off jump habituation. We found that they were not able to suppress their escape behavior as a result of deficient habituation learning (Fig 1A). This finding is in line with the recently published habituation deficit of the complete knock-out of OGT ortholog in C. elegans [61] and shows that altering O-GlcNAc transferase activity is sufficient to induce this deficit. We thus demonstrate the importance of O-GlcNAcylation in habituation learning.

Proper development and maintenance of synapses is an important aspect of neuronal function and cognition. The synaptic connection between motor neurons and muscle cells, termed the neuromuscular junction (NMJ), represents an excellent model system to study the molecular mechanisms of synaptic development in Drosophila [62]. Because NMJ defects were found in several Drosophila disease models with defective habituation [48,54–56], we investigated the synaptic architecture of the sxc^H537A/+ larvae. We found that NMJs of the sxc^H537A/+ larvae are characterized by an increased number of synaptic boutons, recognizable structures that contain the synaptic vesicles (Fig 2A). Larvae with a stronger homozygous catalytic mutation, sxc^H596F/H596F, also show an increase in NMJ length, area, and perimeter. We conclude that O-GlcNAcylation is important for control of synaptic size and synaptic bouton number.

Appropriate O-GlcNAc cycling is required for habituation learning and maintenance of the synaptic size

We recently showed that increased protein O-GlcNAcylation in homozygous Oga knockout flies causes a habituation deficit [53]. Here we show that heterozygous Oga knockout can restore the habituation deficit of sxc^H537A/+ flies (Fig 1B). This indicates that habituation learning depends on O-GlcNAc cycling. Because the loss of one Oga allele does not significantly affect total O-GlcNAc levels [53], we presume that subtle changes in O-GlcNAcylation dynamics rather than gross loss of O-GlcNAc transferase activity inhibits habituation learning.

It is known that postsynaptic expression of OGT in excitatory synapses is important for synapse maturity in mammals [24]. Here we show that presynaptic O-GlcNAc transferase also has role in synapse growth. At the NMJ, the synapses of larvae with neuronal overexpression of sxc are shorter, and the number of synaptic boutons is decreased (Fig 2B). Both length and bouton number are normalized when sxc is overexpressed in neurons of the sxc^H537A/+ larvae (Fig 2C). This phenotype was not observed in Oga knockout larvae with increased O-GlcNAcylation. Knockout of Oga can correct the increased bouton number in larvae with sxc^H596F mutation but not the NMJ size (Fig 2D).

Our data suggest that the NMJ defects associated with decreased O-GlcNAc transferase function are of neuronal origin and that O-GlcNAcylation controls the number of synaptic boutons and partially also synaptic size. Absence of synaptic size defects in Oga knockout larvae and failure of Oga^KO to rescue the NMJ size defects caused by decreased O-GlcNAcylation indicates that other, non-catalytic O-GlcNAc transferase functions may be involved in the control of synaptic size. Levine et al. recently demonstrated that non-catalytic activities of OGT are necessary for its function in some cellular processes, such as proliferation [38].

Drosophila NMJ as a model for the O-GlcNAc-related synaptopathy

The sxc catalytic hypomorph mutations (sxc^H537A, sxc^H596F) as well as the OGT-CDG-patient equivalent mutations (R284P, A319T, L254F) that we introduced with the CRISPR/Cas-9 gene-editing technology in the Drosophila sxc gene (sxc^R313P, sxc^A348T, sxc^L283F), lead to an increase in the number of synaptic boutons, and in some cases also to an increase in synaptic size. NMJ size and the number of synaptic boutons in our model is determined by the level of sxc activity. Dependence of these parameters on gene activity/dosage was previously established in Fmr1 (the Drosophila model of Fragile X Syndrome) [59] and other Drosophila models of neurodevelopmental or neurological disorders, including Prosap/SHANK mutants (modelling Phelan-McDermid Syndrome caused by mutations in SHANK3, characterized by ID and ASD), Neuroligin 4 (ID and ASD caused by mutations in NLGN4), VAP33 (model of Amyotrophic Lateral Sclerosis caused by mutations in VAP-33A) and highwire (potential therapeutical target in traumatic brain injury) [63–66]. The synaptic phenotypes associated with impaired sxc catalytic activity may be linked to increased microtubule polymerization, since it has been shown that O-GlcNAcylation of tubulin negatively regulates microtubule polymerization and neurite outgrowth in mammalian cell lines [67] and Fmr1 and VAP-33A control synaptic growth and bouton expansion through presynaptic organization of microtubules [59,66].

Increased number of synaptic boutons has been also associated with increased excitability at the NMJ [68–70] although not consistently [70–72]. An interesting future direction could involve electrophysiological assessment of NMJ activity to determine whether O-GlcNAc cycling and the patient-related sxc mutations go beyond determining synapse development and affect synapse excitability and/or plasticity. However, these investigations would need to test various aspects of physiology and would still leave the impact of O-GlcNAc on cognition undetermined. For this reason, we assessed habituation as a highly cognition-relevant parameter.

Mutations implicated in OGT-CDG affect habituation via modulation of O-GlcNAc transferase activity

We assessed the effect of OGT-CDG missense mutations on habituation. We found that sxc^R313P and sxc^A348T inhibit habituation in the light-off jump habituation assay (Fig 4A and 4B). sxc^L283F could not be investigated as these mutants displayed a non-performer phenotype in the light-off jump response. While the full spectrum of ID-related phenotypes in an individual with R284P mutation has been attributed to OGT, the A319T mutation segregates with an uncharacterized missense mutation in another gene implicated in ID, MED12 (G1974H) [25]. It was not known which of the mutations is responsible for ID in the affected individuals. We provide evidence that the Drosophila equivalent of the A319T mutation in the TPR domain of OGT causes a cognitive deficit and support a causal role of A319T in OGT-CDG.

Consistent with no detectable O-GlcNAc changes in patient samples and cellular models of the non-catalytic OGT mutations [26,28], no appreciable reduction in protein O-GlcNAcylation was observed in sxc^R313P and sxc^A348T flies. However, habituation learning was restored by increasing O-GlcNAcylation through blocking Oga activity (Fig 4C–4F). This argues that the mechanism by which sxc^R313P and sxc^A348T inhibit habituation is defective O-GlcNAc transferase activity, paralleling impaired O-GlcNAc transferase activity and significant reduction of protein O-GlcNAcylation demonstrated in the catalytic OGT-CDG mutations [30,31]. It is worth noticing that we have previously shown that mutations in Oga also cause habituation deficits [53]. Our finding that genetic combination of loss of OGA with loss of OGT activity rescues the cognitive readout argues that OGA inhibition using available inhibitors may represent a viable treatment strategy. The R284P and A319T reside in the TPR domain (S1 Fig), which is responsible for recognition and binding of OGT substrates [38,73,74]. All OGT-CDG mutations investigated in this study were shown to impair the substrate interaction properties and the glycosyltransferase kinetics [27,36]. The observed habituation deficits may thus be caused by impaired O-GlcNAcylation dynamics towards a specific set of substrates that cannot be captured by standard O-GlcNAc detection assays. Identification of these substrates may pinpoint the underlying defective mechanisms and additional treatment targets.

Potential downstream effectors of O-GlcNAcylation in cognition and cognition-relevant processes

Our explorative analysis found that of 43 established O-GlcNAcylated proteins [60], nine are orthologs of human proteins implicated in ID: ATP1A2, ATN1, HCF1, LAMA2, PTPN23, CHRNA7, NID1, NUP62 and CENPJ. These proteins represent potential downstream effectors and can be investigated in future studies. Particularly the transcriptional co-regulator HCF1 (Host Cell Factor 1) emerges as a top candidate. In mammals, OGT mediates glycosylation and subsequent cleavage of HCF1, which is essential for its maturation [75]. Recombinant OGT with an R284P amino acid substitution is defective in HCF1 glycosylation [28] and HCF1 processing was shown to be completely abrogated by a catalytic OGT-CDG mutation [30].

Drosophila sxc is a member of the polycomb group (PcG), a conserved set of chromatin and transcriptional modifiers that initially have been identified by phenotypic similarity of their mutant phenotypes: homeotic transformations. They are required for maintenance of transcriptional repression (of non-lineage genes) during embryonic development and cell proliferation [76,77]. Missing O-GlcNAcylation of PcG component Polyhomeotic (Ph) is responsible for misexpression of HOX genes and homeotic transformations in sxc null mutants [78]. A recent study has shown that chromatin redistribution induced by interaction between sxc and PcG member Polycomb like (Pcl) controls plasticity of sensory taste neurons [79]. It is not known whether the regulation of PcG activity by sxc/OGT is important for cognitive function, but it can be noted that a series of PcG genes are associated with ID [80,81], and some of them are subject to regulation by OGT in the context of development or cancer. These include PHC1 –human ortholog of Drosophila Ph [82], RING1B (Drosophila Sce) [83,84], EZH2 (Drosophila E(z)) [85–87], YY1 (Drosophila pho) [88,89] and ASXL1 (Drosophila Asx) [90,91]. In addition, OGT regulates expression of PcG genes by O-GlcNAcylation of PcG transcriptional regulators, for example ATN1 (Drosophila Gug), which was identified in the embryonic O-GlcNAc proteome [60]. The encoded proteins represent interesting candidate targets that may link cognitive deficits of OGT-CDG mutations to PcG function.

We propose that in depth clinical phenotyping of patients with mutations in OGT and the above listed genes may give additional hints to the most crucial downstream targets of OGT-mediated O-GlcNAcylation.

In summary, we show that OGT-CDG mutations in the TPR domain negatively affect habituation learning in Drosophila via reduced protein O-GlcNAcylation. The data support a causal role of A319T in OGT-CDG and demonstrate that Drosophila habituation can be used to analyze the contribution of OGT mutations to cognitive deficits. This important aspect of ID has to date not been addressed for any of the OGT-CDG mutations. Moreover, our genetic approach points to a key role of O-GlcNAc transferase activity in ID-associated cognitive deficits and identifies blocking O-GlcNAc hydrolysis as a treatment strategy that can ameliorate cognitive deficits in OGT-CDG patients. Thanks to its high-throughput compatibility, the light-off jump habituation assay can be used with high efficiency for future identification of the downstream effectors and novel therapeutic targets for OGT-CDG.

Materials and methods

Cloning of the guide RNA and repair template DNA vectors for Drosophila CRISPR/Cas9 editing

Novel mutant Drosophila lines, sxc^H596F, sxc^R313P, sxc^A348T, and sxc^L283F, were generated via CRISPR/Cas9 gene editing, following a previously described protocol [51]. Briefly, guide RNA sites were selected using an online tool (crispr.mit.edu) and the annealing primer pairs with appropriate overhangs for BpiI restriction digestion were cloned into pCFD3-dU63gRNA plasmid [92]. Vectors coding for repair template DNA of roughly 2 kb were generated from Drosophila Schneider 2 cell genomic DNA by PCR using GoTaq G2 Polymerase (Promega) and primer pairs appropriate for the desired region (S1 Table). The PCR products were digested with BpiI and inserted into the pGEX6P1 plasmid. The intended mutation, as well as silent mutations required to remove the gRNA sequence (S1 Fig), were incorporated by either site-directed mutagenesis (H596F) using the QuikChange kit (Stratagene) or restriction-free cloning (R313P, A348T and L283F) [93]. The four sets of mutations–H596F, L283F, R313P, and A348T removed restriction sites for HinfI, BfmI, MnlI and BsqI, respectively. DNA products of cloning and mutagenesis were confirmed by sequencing. All primer sequences are listed in S1 Table.

Generation of sxc^H596F, sxc^R313P, sxc^A348T, and sxc^L283F Drosophila lines

Vasa::Cas9 Drosophila embryos (strain #51323 from Bloomington Drosophila Stock Center; bdsc.indiana.edu) were injected with a mixture of CRISPR/Cas9 reagents, 100 ng/μl guide RNA plasmid and 300 ng/μl repair template DNA vector (University of Cambridge fly facility). Injected male flies were crossed with an in-house Sp/CyO balancer stock for two generations, allowing for the elimination of the vasa::Cas9 carrying X chromosome. Candidate F1 males were genotyped exploiting restriction fragment length polymorphism. All lines were validated by sequencing the region approximately 250 base pairs upstream and downstream of the mutations and sequencing the areas outside the repair templates. In addition, all of the predicted off-target sites were PCR-amplified and checked for the presence of any lesions compared with the genomic DNA from the BL51323 line. None of the predicted off-target sites were found to have mutations. To eliminate any other potential off-target mutations introduced during CRISPR, all lines were backcrossed into the w¹¹¹⁸ control genetic background for six generations.

Restriction fragment length polymorphism assay

To assess and confirm the presence of the H596F, L283F, R313P, and A348T mutations in the sxc gene, DNA of candidate individual adult flies was extracted using 10–50 μl of DNA extraction buffer containing 10 mM Tris-HCl pH 8, 1 mM EDTA, 25 mM NaCl and 200 μg/ml freshly added Proteinase K (Roche). The solution was subsequently incubated at 37°C for 30 min, followed by inactivation of Proteinase K at 95°C for 3 min, and centrifuged briefly. 1 μl of the crude DNA extract was used per 25 μl PCR reaction with the relevant diagnostic primers, using a 2x GoTaq G2 Green premix (Promega). 5 μl of the PCR products were used for restriction fragment length polymorphism assay with the appropriate enzymes, followed by agarose gel electrophoresis of the digested products. Reactions which showed the presence of an undigested full-length PCR product resistant to the expected restriction enzyme cleavage indicated CRISPR/Cas9 gene editing event and were sequenced. Precise incorporation of the repair template into the right position of the genome was confirmed by sequencing a second round of PCR products obtained from potential homozygous CRISPR mutants with mixed diagnostic and line-check primer pairs. Primer sequences are listed in S1 Table.

Fly stocks and maintenance

Drosophila stocks and experimental crosses were reared on a standard Drosophila diet (sugar/cornmeal/yeast). An RNAi strain to knockdown sxc (#18610) and a genetic background control strain (#60000) were obtained from the Vienna Drosophila Resource Center (VDRC; www.vdrc.at). In-house sxc^H537A [51] and UAS-sxc [40] strains, and the generated sxc^H596F, sxc^R313P, sxc^A348T, and sxc^L283F strains, were crossed into the VDRC w¹¹¹⁸ control genetic background (#60000) for six generations. The sxc^H537A/CyO; UAS-sxc strain was assembled using the isogenic strains. Oga^KO and Oga^D133N lines were also crossed to this background as described earlier [53]. #60000 was used as isogenic control for the mutant alleles. In the neuromuscular junction analysis of sxc^R313P, sxc^A348T and sxc^L283F alleles, the control flies were derived by crossing the flies from the stock used for microinjection (Bloomington Stock: BL51323) and same crossing scheme as that used to derive the sxc^R313P, sxc^A348T and sxc^L283F homozygotes and eliminate the Cas9 transgene were used. To induce neuronal knockdown and overexpression, a w¹¹¹⁸; 2xGMR-wIR; elav-Gal4, UAS-Dicer-2 driver strain was used. This strain contains a double insertion of an RNAi construct targeting the gene white specifically in the Drosophila eye (2xGMR-wIR) to suppress pigmentation, as required for an efficient light-off jump response [54,55]. Progeny of the crosses between the driver, RNAi/UAS-sxc and #60000 strain were used as controls for knockdown and overexpression experiments. All crosses were raised at 25°C, 70% humidity, and a 12:12h light-dark cycle.

Western blotting

Protein lysates for Western blotting were prepared from adult male (1–4 days old) fly head samples or 0–16 h embryo collection and snap frozen in liquid nitrogen. Samples were homogenized in lysis buffer containing 2x NuPAGE LDS Sample Buffer, 50 mM Tris- HCl (pH 8.0), 150 mM NaCl, 4 mM sodium pyrophosphate, 1 mM EDTA, 1 mM benzamidine, 0.2 mM PMSF, 5 μM leupeptin, and 1% 2-mercaptoethanol. Crude lysates were then incubated for 5 min at 95°C, centrifuged at 13000 rpm for 10 min, and supernatants were collected. Pierce 660 nm protein assay supplemented with Ionic Detergent Compatibility Reagent (Thermo Scientific) was used to determine protein concentration. 20–30 μg of protein samples were separated on RunBlue 4–12% gradient gels (Expedeon) using MOPS running buffer, before being transferred onto nitrocellulose membranes. Western blot analysis was carried out with anti-O-GlcNAc (RL2, Abcam, 1:1000) and anti-actin (Sigma, 1:5000) antibodies. Membranes were incubated overnight with selected primary antibodies in 5% BSA at 4°C. Blots were visualized via Li-Cor infrared imaging with Li-Cor secondary antibodies (1:10000) Signal intensities were quantified using ImageStudioLite software. Significance was calculated using one-way ANOVA with Tukey’s multiple comparisons test (p_adj).

Developmental survival

Stage 11–16 embryos (25 embryos per vial, 100 per genotype per experiment, n = 3) were cultured at 25°C and assessed for lethality by counting the number of pupae and adults derived. Significance was calculated using Student’s t-test with Holm-Sidak’s correction for multiple comparisons when appropriate.

Light-off jump habituation

The light-off jump reflex habituation assay was performed as previously described [49,94]. Briefly, 3- to 7-day-old individual male flies were subjected to the light-off jump reflex habituation paradigm in two independent 16-chamber light-off jump habituation systems. Male progeny of the appropriate control genetic background was tested simultaneously on all experimental days. Flies were transferred to the testing chambers without anesthesia. After 5 min adaptation, a total of 32 flies (16 flies/system) were simultaneously exposed to a series of 100 short (15 ms) light-off pulses with 1 s interval. The noise amplitude of wing vibration following every jump response was recorded for 500 ms after the start of each light-off pulse. A carefully chosen automatic threshold was applied to filter out background noise and distinguish it from jump responses. Data were collected by a custom-made Labview Software (National Instruments). Initial jump responses to light-off pulse decreased with the increasing number of trials and flies were considered habituated when they failed to jump in five consecutive trials (no-jump criterion). Habituation was quantified as the number of trials required to reach the no-jump criterion (Trials To Criterion (TTC)). All experiments were done in triplicates (N = 96 flies). Main effects of genotype on log-transformed TTC values were tested using a linear model regression analysis (lm) in the R statistical software (R version 3.0.0 (2013-04-03)) [95] and corrected for the effects of testing day and system. Bonferroni-Holm correction for multiple testing [96] was used to calculate adjusted p-values (p_adj).

Fatigue assay

Each genotype that was tested in light-off jump habituation was subsequently subjected to fatigue assay. The fatigue assay was used to evaluate whether the lower TTCs in the rescue experiments were not a result of increased fatigue rather than improved habituation/non-associative learning. The assay was performed as previously described [49]. The interval between light-off pulses was increased to 5 seconds, an intertrial interval that is sufficiently long to prevent habituation. The light-off pulse was repeated 50 times. Fatigue was concluded when log-transformed TTC values of the rescue were significantly smaller than log-transformed TTC values of the control (based on lm analysis and Bonferroni-Holm correction; p_adj < 0.05).

Analysis of Drosophila neuromuscular junction

Wandering male L3 larvae were dissected with an open book preparation [97], and fixed in 3.7% paraformaldehyde for 30 minutes. Larvae were stained overnight at 4°C with the primary antibodies against synaptic markers Discs large (anti-dlg1, mouse, 1:25, Developmental Studies Hybridoma Bank) and synaptotagmin (anti-Syt, rabbit, 1:2000, kindly provided by H. Bellen). Secondary antibodies anti-mouse Alexa 488 and anti-rabbit Alexa 568 (Invitrogen) were applied for 2 hours at room temperature (1:500). Projections of type 1b neuromuscular junctions (NMJs) at muscle 4 from abdominal segments A2-A4 were assessed. Individual synapses were imaged with a Zeiss Axio Imager Z2 microscope with Apotome and quantified using in-house developed Fiji-compatible macros [98,99]. Anti-dlg1 (4F3 anti-discs large, DSHB, 1:25) labeling was used to analyze NMJ area, length, number of branches and branching points. Anti-Syt (kind gift of Hugo Bellen, 1:2000) labeling was used to analyze the number of synaptic boutons. Secondary antibodies goat anti-mouse Alexa Fluor 488 (1:200) and goat anti-rabbit Alexa Fluor 568 from Life Technologies were used for visualization. Parameters with a normal distribution (area, length, number of boutons) were compared between the mutants and controls with one-way ANOVA (p) and Tukey’s test for multiple comparisons (p_adj). Parameters without normal distribution (number of branches and branching points) were compared with non-parametric Wilcoxon test (p, single comparisons) and Kruskal-Wallis test with Wilcoxon pairwise test for multiple comparisons (p_adj) in the R statistical software (R version 3.0.0 (2013-04-03)) [95].

Island assay

Locomotor behaviour of 3–6 days old male flies was assessed with the island assay as described previously [100,101]. Each trial was performed using 15 flies. 3–4 repeats were carried out on each test day, and data was collected on 3 consecutive days. In total, data from 11–16 trials were collected per genotype. The percentage of flies on the island platform over time was plotted and area under curve (AUC) was determined for each run. Groups were compared using one-way ANOVA with Holm-Sidak’s multiple comparisons (p_adj) of means for AUC.

Negative geotaxis test

The negative geotaxis assay was performed as described previously [102]. The climbing ability of 3–6 days old male flies was evaluated on groups of 10 animals. Prior to the measurement, flies were transferred into 150 x 16 mm transparent plastic test tubes without anesthesia. Test tubes were secured into a frame that allowed for monitoring of climbing behavior of up to 10 vials at once. Upon release, the frame is dropped from a fixed height onto a mouse pad, thereby tapping the flies to the bottom of the tubes. The climbing assay was repeated 4 times for each loaded frame providing data from 4 runs. The experiment was video-recorded with a Nikon D3100 DSLR camera. ImageJ/FIJI software was used to analyse the resulting recordings. First, images were converted to an 8-bit grey scale TIFF image sequence (10 frames per second) file format. Background-subtraction and filtering were then applied, and the image pixel values were made binary. The MTrack3 plug-in was used for tracking of flies. Mean climbing speed (mm/s) was quantified for each genotype in 2nd, 3rd and 4th runs, between 17–89 data points were collected per run. Groups were compared using one-way ANOVA with Tukey’s multiple comparisons (p_adj) of means on mean climbing speed values calculated for each run.

Enrichment analysis

The O-GlcNAc proteome data was extracted from Selvan et al. (Supplementary dataset 3) [60]. Phenotype annotations of Drosophila gene alleles were extracted from Flybase (Flybase.org, downloaded in April 2016). Human genes implicated in Intellectual disability (ID + ID candidate genes) were extracted from sysid database (https://sysid.cmbi.umcn.nl/, downloaded in April 2016). Enrichment was calculated as follows: (a/b)/((c-a)/(d-b)), whereby a = genes in O-GlcNAc proteome and associated with the phenotype term/human ID gene orthologs, b = genes in O-GlcNAc proteome, c = genes associated with the phenotype term/human ID gene orthologs, d = background/all Drosophila genes. Significance was determined using two-sided Fisher’s exact test in R [95]. p-values were adjusted for multiple testing (p_adj) with Bonferroni-Holm correction [96].

Supporting information

S1 Fig. Generation and characterization of sxc^H596F, sxc^L283F, sxc^R313P and sxc^A348T alleles.

(A) Schematic representation of Drosophila sxc protein showing the location of H596F, L283F, R313P and A348T mutations; purple tetratricopeptide repeat (TPR) domain, green glycosyl transferase (GT) domain. (B)–(E) Sequences of genomic DNA of wild type, sxcH596F, sxcL283F, sxcR313P and sxcA348T Drosophila alleles. The missense mutation and additional silent mutations are highlighted. The restriction digestion sites used for genotyping are shown.

(TIF)

S2 Fig. Jump responses in the fatigue assay.

In the fatigue assay, jump responses were induced with 50 light-off pulses with 5 s interval between pulses that prevents habituation. The jump response is presented as % of jumping flies in each light-off trial. The mean number of trials that flies needed to reach the no-jump criterion (Trials To Criterion, TTC) ± SEM is presented. (A) Jump response of the sxc^H537A/+; Oga^KO/+ flies (N = 85, mean TTC ± SD: 27.4 ± 8, in cyan) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 85, mean TTC ± SD: 34.5 ± 5, p_adj = 0.14, in blue) demonstrating that restored habituation in sxc^H537A/+; Oga^KO/+ flies (Fig 1B) is not confounded by fatigue. (B) Jump response of the sxc^H537A/+; elav-Gal4>UAS-sxc flies (N = 52, mean TTC ± SD: 25.1 ± 7.9, in green) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 55, mean TTC ± SD: 28.3 ± 2.3, p_adj = 0.128, in blue) demonstrating that restored habituation in sxc^H537A/+; elav-Gal4>UAS-sxc flies (Fig 1D) is not confounded by fatigue. (C) Jump response of the sxc^R313P/+; Oga^KO/+ flies (N = 73, mean TTC ± SD: 28.9 ± 7.6, in cyan) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 84, mean TTC ± SD: 26.1 ± 5.1, p_adj = 1, in blue) demonstrating that restored habituation in sxc^R313P/+; Oga^KO/+ flies (Fig 4C) is not confounded by fatigue. (D) Jump response of the sxc^A348T/A348T; Oga^KO/KO flies (N = 78, mean TTC ± SD: 26.9 ± 5.1, in cyan) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 85, mean TTC ± SD: 34.5 ± 5, p_adj = 0.31, in blue) demonstrating that restored habituation in sxc^A348T/A348T; Oga^KO/KO flies (Fig 4D) is not confounded by fatigue. (E) Jump response of the sxc^R313P/+; Oga^D133N/+ flies (N = 83, mean TTC ± SD: 25.5 ± 9.2, in cyan) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 84, mean TTC ± SD: 26.1 ± 5.1, p_adj = 1, in blue) demonstrating that restored habituation in sxc^R313P/+; Oga^{D133N /+} flies (Fig 4E) is not confounded by fatigue. (F) Jump response of the sxc^A348T/A348T; Oga^D133N/D133N flies (N = 63, mean TTC ± SD: 24.4 ± 5.1, in cyan) remains high throughout the entire course of the experiment, similar to control flies (+/+, N = 85, mean TTC ± SD: 34.5 ± 5, p_adj = 0.15, in blue) demonstrating that restored habituation in sxc^A348T/A348T; Oga^D133N/D133N flies (Fig 4F) is not confounded by fatigue. * p_adj<0.1, based on lm analysis with Bonferroni-Holm correction for multiple comparisons. Complete list of p-values and summary statistics is provided in S3 Table.

(TIF)

S3 Fig. NMJ of sxc^H537A/H537A mutant and NMJ branching morphology.

(A) Number of synaptic branches and branching points in sxc^H537A/+ larvae (N = 29, in red) is not significantly different from the genetic background control larvae (+/+, N = 24, branches: p = 0.1439, branching points: p = 0.05648, in blue). P-values are based non-parametric Wilcoxon test analysis. (B) Number of branches and branching points is not affected in elav-Gal4>UAS-sxc larvae (N = 29, in dark blue) compared the elav-Gal4/+ larvae (N = 30, branches: p_adj = 0.8702, branching points: p_adj = 0.8488, in light blue) and to UAS-sxc/+ larvae (N = 26, branches: p_adj = 0.8294, branching points: p_adj = 0.2689, in grey). (C) Branches and branching points are not affected in sxc^H537A/+; UAS-sxc/+ larvae (N = 28, in red) compared to the control larvae (+/+, N = 28, branches: p_adj = 0.7121, branching points: p_adj = 0.2979, in blue). sxc^H537A/+; elav-Gal4>UAS-sxc larvae (N = 29, in green) do not show any changes in number of branches and branching points compared to the sxc^H537A/+; UAS-sxc/+ larvae (branches: p_adj = 0.4097, branching points: p_adj = 0.5928) and control larvae (+/+; branches: p_adj = 0.4301, branching points: p_adj = 0.6927). (D) sxc^H537A/H537A larvae have significantly increased NMJ length (N = 25, p =) and perimeter (N = 21, p =, in red) compared to their genetic background control (+/+, N = 26, in blue) but not significantly different number of boutons (p = 0.085), NMJ area (p = 0.618), number of branches (p = 0.691) and branching points (p = 0.371). * p<0.05, *** p<0.001. P-values for boutons, length, area and perimeter are based on one-way ANOVA. P-values for branches and branching points are based on non-parametric Wilcoxon test analysis. (E) Branches and branching points are not affected in sxc^H956F larvae (N = 31, branches: p_adj = 1, branching points: p_adj = 0.5, in brown), Oga^KO larvae (N = 30, branches: p_adj = 0.75, branching points: p_adj = 0.51, in purple), and sxc^H596F; Oga^KO larvae (N = 30, branches: p_adj = 0.75, branching points: p_adj = 0.5, in cyan) compared to the genetic background control larvae (+/+, N = 28, in blue). sxc^H596F; Oga^KO larvae do not show a significant change in number of branches and branching points compared to the sxc^H956F larvae (branches: p_adj = 0.75, branching points: p_adj = 1). Data presented as individual data points with mean ± SD. P-values are based on Kruskal-Wallis test with Wilcoxon pairwise test for multiple comparisons. Complete list of p-values and summary statistics is provided in S3 Table. (D’) Representative NMJs of genetic background control (+/+) and sxc^{H537A/sxcH537A} wandering third instar larvae labeled with anti-discs large 1 (Dlg, magenta) and anti-synaptotagmin (Syt, green). Scale bar, 20μm. The quantitative parameter values of the representative images (+/+ | sxc^{H537A/sxcH537A}): #Boutons (26 | 30), Length (96.2 | 169.8), Area (415.9 | 464.3), Perimeter (258.4 | 445).

(TIF)

S4 Fig. Western Blot and developmental survival of sxc^H596F flies.

(A) Embryos from either wildtype, sxc^H537A, sxc^H596F homozygotes were assessed for levels of global O-GlcNAc using a pan-O-GlcNAc antibody RL2. The blot was normalized to actin. This blot is a representative of three experiments. (B) Reduced total O-GlcNAc levels in sxc^H596F and sxc^H537A homozygotes are not associated with developmental lethality. Data presented as percentage of pupae and adults derived from stage 11–16 embryos (100 per genotype per experiment, n = 3). Based on Student’s t-test with Holm-Sidak’s correction for multiple testing. Complete list of p-values and summary statistics is provided in S3 Table.

(TIF)

S5 Fig. Developmental and locomotor characterization and NMJ branching morphology of sxc^R313P, sxc^A348T and sxc^L283F.

(A) Control, sxc^R313P, sxc^A348T and sxc^L283F embryos (stage 11–16, 80–100 per experiment) were transferred to fresh food at 25°C, and the numbers of pupae formed and adults eclosed were counted. Development from embryo to pupae or from pupae to adulthood was not significantly affected in sxc^R313P (pupae: N = 4 repeats, p = 0.9, adults: N = 3, p = 0.29, in red), sxc^A348T (pupae: N = 3, p = 0.152, adults: N = 3, p = 0.345, in blue) and sxc^L283F mutants (pupae: N = 4, p = 0.108, adults: N = 3, p = 0.727, in green). Data presented as individual data points with mean ± SD. P-values are based on Student’s t-test. (B) Flight escape performance was assessed in the island assay. 15 flies per measurement were thrown on a white platform surrounded with water. Data was collected over 3 days of measurement (Control: N = 23, sxc^R313P: N = 16, sxc^A348T: N = 15 and sxc^L283F: N = 14 repeats). Floating bars depict mean ± SD area under curve (AUC), a parameter that is derived from data plotted as % flies on the platform over time. One-way ANOVA with Tukey’s multiple comparisons was used to compare the mean AUC between genotypes. Flight escape performance of sxc^R313P, sxc^A348T and sxc^L283F flies revealed no defects in locomotion or fitness. (C) Number of synaptic branches is increased in sxc^R313P (N = 21, p = 0.049, in red) and sxc^L283F larvae (N = 20, in green) compared to the genetic background control (+/+, N = 41, in grey). Number of branching points is not significantly different. Data presented as individual data points with mean ± SD. * p< 0.05. P-values are based on Kruskal-Wallis test with Wilcoxon pairwise test for multiple comparisons. Complete list of p-values and summary statistics is provided in S3 Table.

(TIF)

S1 Table. Primer sequences.

(XLSX)

S2 Table. Light-off jump habituation parameters summary.

(XLSX)

S3 Table. Summary statistics.

(XLSX)

S4 Table. Enrichment in Drosophila phenotypes and orthologs of human genes implicated in intellectual disability.

(XLSX)

Data Availability

All relevant data are within the manuscript and its Supporting Information files.

Funding Statement

This work was supported by a ZonMW Vici grant from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (nwo.nl, grant no. 09150181910022, M.F. received salary from this grant) and the Australian National Health & Medical Research Council Centre for Research Excellence Scheme (www.nhmrc.gov.au, grant no. APP1117394, L.E.R.B received salary from this grant) to A.S., by an h2020 European Research Council (erc.europa.eu) consolidator grant (grant no. ERC-2017-COG 770244, M.C. and E.S. received salary from the grant) to E.S., by a mobility grant from Ministersvo Školství, Mládeže a Tělovýchovy (www.msmt.cz, grant no. CZ.02.2.69/0.0/0.0/20_079 /0017633, M.F. received salary from this grant) to M.F. and by a Wellcome Trust Investigator Award (wellcome.org, grant no. 110061) and the National Centre for the Replacement, Refinement and Reduction of Animals in Research (https://www.nc3rs.org.uk/; grant no. T001682) to D.M.F.A. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.