Neurodevelopmental defects in a mouse model of O-GlcNAc transferase intellectual disability

Intellectual disability (ID) represents a heterogeneous group of neurodevelopmental disorders that are predicted to affect 1–3% of the worldwide population (Maulik et al., 2011). ID appears during childhood and is characterized by impaired cognitive function (with IQ below 70) and adaptive behavior (Shaffer, 2005). Frequent comorbidities are autism spectrum disorders (ASDs), attention deficit and hyperactivity disorder (ADHD), and epilepsy (Posada de la Paz et al., 2017).

The addition of O-linked β-N-acetylglucosamine (O-GlcNAc) to serine and threonine residues of proteins (referred to as O-GlcNAcylation) is a dynamic and highly conserved posttranslational modification that regulates several cellular processes, including transcription (Lamarre-Vincent and Hsieh-Wilson, 2003; Constable et al., 2017), signaling (Ong et al., 2018) and metabolism (Whelan et al., 2008). Only two enzymes regulate O-GlcNAcylation: O-GlcNAc transferase (OGT) is responsible for the addition of a single O-GlcNAc moiety on substrates (Kreppel and Hart, 1999) and O-GlcNAcase (OGA) for its removal (Gao et al., 2001). O-GlcNAcylation has been shown to be essential in mammals. Ogt deletion leads to impaired embryogenesis and early development (Shafi et al., 2000; O'Donnell et al., 2004), whereas Oga is critical for perinatal survival (Keembiyehetty et al., 2015; Stichelen et al., 2017; Muha et al., 2021). Dysregulation of O-GlcNAcylation homeostasis has also been associated with pathological conditions such as diabetes, neurodegeneration and cancers (Bond and Hanover, 2013; Balana and Pratt, 2021; Ciraku et al., 2022).

O-GlcNAcylation and the two cycling enzymes are abundant in the mammalian brain (Gao et al., 2001; Okuyama and Marshall, 2003; Akimoto et al., 2003), in particular at pre- and post-synaptic compartments, where many proteins important for neuronal structure and synaptic function have been found to be O-GlcNAcylated (Vosseller et al., 2006; Khidekel et al., 2004; Lagerlof et al., 2017; Cole and Hart, 2001). Previous work has demonstrated that modulation of O-GlcNAcylation affects neuronal processes important for brain function, including synaptic maturation, synaptic function and axon morphology, and learning and memory (Lagerlof et al., 2017; Francisco et al., 2009; Taylor et al., 2014; Yang et al., 2017; Rexach et al., 2012; Wheatley et al., 2019). Several OGT substrates are also ID-related proteins whose functions have been shown to be regulated by O-GlcNAcylation. O-GlcNAc represses CREB (cyclic AMP response element-binding protein)-dependent transcription that is associated with changes in neurite growth and long-term memory (Rexach et al., 2012). O-GlcNAcylation at T306 also blocks interactions between SynGAP (also known as SynGAP1) and PSD-95 (also known as DLG4) leading to inhibition of liquid-liquid phase separation that could affect postsynaptic density condensate formation (Lv et al., 2022). Although these observations collectively suggest an important role for neuronal O-GlcNAcylation in cognition, our understanding of how O-GlcNAcylation and the cycling enzymes regulate brain function is still limited.

Recently, missense variants of OGT have been identified in individuals affected by ID, giving rise to a syndrome named OGT-linked congenital disorder of glycosylation (OGT-CDG) (Pravata et al., 2020). As OGT is located on the X chromosome, almost all individuals with OGT-CDG are male, with the exception of a report of monozygotic female twins both harbouring a de novo missense variant (Pravata et al., 2019a). OGT-CDG is a clinically heterogenous disorder in which those affected present with ID, developmental delay and very restricted language skills. Individuals with OGT-CDG also commonly present with dysmorphic features including craniofacial characteristics with broad and high forehead, hypertelorism, broad nasal root, full or long philtrum, and clinodactyly. Brain and eye abnormalities are also observed in most cases (Pravata et al., 2020).

OGT is composed of a catalytic domain and an N-terminal tetratricopeptide repeat (TPR) domain consisting of 13.5 TPRs that is responsible for substrate binding and protein-protein interactions (Hart et al., 2011). In addition to installing O-GlcNAc on proteins, OGT is involved in the proteolytic processing and activation of host cell factor 1 (HCF1) (Capotosti et al., 2011), which is encoded by a known ID gene (Castro and Quintana, 2020), and possesses non-catalytic functions implicated in cellular proliferation (Levinea et al., 2021). To date, 17 OGT-CDG missense variants have been reported in OGT, with missense and exon-skipping variants in the TPR and catalytic domains giving rise to similar clinical features. This suggests that there are common mechanisms affected in these OGT-CDG variants; these mechanisms remain unknown as no vertebrate models are currently available to dissect them. Investigating the function of OGT in neurodevelopment and the brain has been hampered by mouse lethality caused by global loss of Ogt (Shafi et al., 2000; O'Donnell et al., 2004). Although several brain-cell-specific Ogt knockout (KO) mouse models also lead to postnatal lethality, these have highlighted the role of OGT in neurodevelopment, neuronal survival and structure (O'Donnell et al., 2004; Wheatley et al., 2019; Shao et al., 2022; Cheng et al., 2020; Wang et al., 2016). Modeling OGT-CDG in vivo will help to more precisely understand how OGT regulates processes that are essential for neurodevelopment and brain function.

Here, we report the use of a CRISPR/Cas9 genome editing approach to generate a mouse model carrying a catalytically impaired OGT-CDG variant, C921Y. Unlike previous Ogt KO models, mice carrying the C921Y variant of OGT (OGT^C921Y) are viable, allowing the phenotypic characterization of the animals. Loss of OGT catalytic activity leads to impaired O-GlcNAcylation homeostasis in the brain, changes in body size and mass, and microcephaly in OGT^C921Y mice.

To dissect the effects of OGT-CDG variants in vivo, we used a CRISPR/Cas9 genome editing approach in mice to introduce the missense mutation C921Y (Omelková et al., 2023), which has been reported previously and is located in the OGT catalytic domain (Fig. 1A). Briefly, zygotes at 0.5 days post-coitum (dpc) were injected with editing reagents and transferred into pseudo-pregnant female mice. DNA from offspring was genotyped and sequenced to confirm the presence of the OGT^C921Y mutation (Fig. 1B). As Ogt is essential for embryogenesis (Shafi et al., 2000; O'Donnell et al., 2004), we first assessed the viability and mendelian inheritance distribution of offspring generated from heterozygous Ogt^C921Y/+ females crossed to wild-type (WT) males (Ogt^+/y) on the same C57BL/6J genetic background. This generated 48.8% male and 51.2% female pups. These observations suggest no sex-related lethality in the OGT^C921Y mouse line. DNA from offspring was then genotyped and sequenced, revealing that both heterozygous female and hemizygous male offspring could be obtained for this line. We then analyzed the mendelian inheritance distribution of WT (Ogt^+/+ and Ogt^+/y; expected 50%), heterozygous Ogt^C921Y/+ (expected 25%) and hemizygous Ogt^C921Y/y (expected 25%) in animals arising from the crosses. We found that the crosses generated 27.3% WT female, 23.8% Ogt^C921Y/+ female, 23.8% WT male and 25% Ogt^C921Y/y male offspring at the expected mendelian ratio (P=0.29, χ-squared test; n=84) (Fig. 1C). We also investigated the fertility of male hemizygotes carrying the OGT^C921Y mutation, as some individuals affected by OGT-CDG exhibit genital abnormalities (Pravata et al., 2020). We successfully obtained WT and heterozygous Ogt^C921Y/+ animals from WT female and hemizygous Ogt^C921Y/y male breeding pairs, indicating that fertility is not affected in OGT^C921Y mice. The gene encoding OGT is located on the X chromosome in both human and mouse, and as almost all individuals with OGT-CDG are male, only hemizygous Ogt^C921Y/y (hereafter referred to as OGT^C921Y) and WT male animals were used for the following experiments. Taken together, these results show that, contrary to Ogt KO models, animals carrying the OGT^C921Y mutation in the catalytic domain of OGT are viable.

The OGT^C921Y variant affects the catalytic activity of OGT, leading to disruption of O-GlcNAcylation homeostasis in stem cells (Omelková et al., 2023). Therefore, we first investigated the levels of OGT, OGA and O-GlcNAc levels in whole brains from male OGT^C921Y animals (Fig. 2). Global protein O-GlcNAc levels were significantly decreased in OGT^C921Y males compared to those in WT littermates (Fig. 2A,B). In addition, both OGA (Fig. 2A,C) and OGT (Fig. 2A,D) protein levels were significantly reduced in OGT^C921Y animals compared to those in WT mice. To investigate whether the reductions in OGT and OGA protein levels were due to reduced transcription, we performed qPCR analysis on whole-brain tissue extracts from WT and OGT^C921Y mice. We observed a reduction in Oga mRNA levels in OGT^C921Y animals (Fig. 2E), which was accompanied by an increase in Ogt mRNA levels (Fig. 2F), suggesting that the reduction in OGT protein levels is not caused by reduced transcription but occurs at the protein level. Taken together, these results show that the OGT^C921Y mutation is linked to changes in brain O-GlcNAc homeostasis.

Individuals with OGT-CDG commonly present with short stature and low birth weight, and both these features have been observed in people carrying the OGT^C921Y variant (Pravata et al., 2020; Omelková et al., 2023). We therefore evaluated morphometric parameters in OGT^C921Y mice. We observed a significant decrease in both body weight (Fig. 3A) and nose-to-tail length (Fig. 3B) of OGT^C921Y mice compared to that of WT littermates. These quantitative changes were accompanied by a slim appearance of the mutant animals. We next performed EchoMRI imaging to evaluate body composition of the OGT^C921Y mice. We observed a significant reduction in body fat mass in OGT^C921Y mice compared to that of WT animals (Fig. 3C). The OGT^C921Y mice also showed an increase in lean body mass compared to that of WT mice (Fig. 3D). This suggests that the lower body weight observed in the OGT^C921Y mice is due to reduced adipose tissue content as well as short stature. In addition, OGT^C921Y mice displayed lower levels of glycemia (Fig. 3E) compared to those displayed by WT animals, suggesting that the OGT^C921Y mice possess disrupted metabolism. Interestingly, lower glycemia was combined with a significant increase in pancreas weight in OGT^C921Y animals compared to that observed for WT animals (Table 1). Taken together, these results suggest that the OGT^C921Y mutation affects mouse size, weight and fat mass.

Individuals with OGT-CDG present with craniofacial dysmorphias, which are associated with microcephaly in several cases (Pravata et al., 2020). To first investigate whether the OGT^C921Y mutation impacts skull morphology, we acquired measurements and performed microcomputed tomography (microCT) imaging of male OGT^C921Y mice and male WT littermates. Using manual measurements, we observed that OGT^C921Y mice exhibit a significant reduction in skull length compared to that of the WT mice, whereas skull width was found to be similar between the genotypes (Fig. 4A,B). Analysis of the microCT images revealed that OGT^C921Y mice exhibit rounder and smaller skulls compared to those of WT animals (Fig. 4C). Superior and lateral views of the skull microCT images were used for two-dimensional measurement of skull parameters using free landmarks (Fig. 4D). Overall, 72.2% of the linear distances were found to be shortened between 2% and 7% in OGT^C921Y animals compared to the distances recorded for WT littermates (Table S1). Along the rostro-caudal axis, 25.6% of the distances were significantly reduced (Table S1), suggesting mild shortening and deformation of the skull in OGT^C921Y mice. The reduction in skull size was associated with a significant decrease in the absolute brain weight of OGT^C921Y mice relative to the brain weight of WT mice, but this was not accompanied by changes in brain:body weight ratio, owing to the reduced overall body size of the OGT^C921Y mice (Fig. 4E,F), suggesting that OGT^C921Y mice display a microcephaly phenotype. Taken together, these results suggest that the OGT^C921Y mutation leads to mild skull deformation and microcephaly in mice.

We have recently described several variants of OGT that are linked to ID (Pravata et al., 2020). To understand the pathophysiology of this disease, we generated a mouse model carrying a catalytically impaired OGT-CDG variant. Although the C921Y mutation causes reduced OGT catalytic activity in vitro (Omelková et al., 2023), OGT^C921Y animals are viable, allowing the phenotypic characterization of these mice and the investigation of the function of OGT in neurodevelopment. We have shown that the OGT^C921Y mutation leads to a reduction in mass and size, altered O-GlcNAc homeostasis in the brain of the mice, and microcephaly.

OGT^C921Y mice show a reduction in body weight. This was found to be associated with a reduction in body fat mass ratio in OGT^C921Y animals compared to that of their WT littermates. Body weight is determined by a balance between food intake and energy expenditure, combining basal metabolism, thermogenesis and physical activity (Speakman, 2013). Compared to WT mice, OGT^C921Y mice were found to have lower levels of blood glucose as well as an increase in pancreas size. Similarly, specific Ogt deletion in sensory neurons results in reduced body weight and low blood glucose levels in the Nav1.8-Ogt^KO mouse model (Su and Schwarz, 2017). Mice lacking OGT in orexigenic neurons exhibit improved glucose metabolism and are protected from diet-induced obesity (Ruan et al., 2014). Taken together, these observations suggest the involvement of neuronal OGT in controlling basal metabolism that could affect body composition. Interestingly, Oga^KO/+ mice exhibit a lean phenotype, similar to our observations of reduced body fat mass in OGT^C921Y mice (Yang et al., 2015). Oga^KO/+ animals showed increased energy expenditure that was not caused by elevated locomotor activity but by enhanced thermogenesis through an increase in white-to-brown transdifferentiation of adipocytes (Yang et al., 2015). Although possible metabolic defects in OGT^C921Y mice remain to be further investigated, these observations suggest that loss of OGT and/or OGA might cause metabolic changes contributing to the lean phenotype observed in OGT^C921Y mice.

We observed a reduction in brain size in OGT^C921Y mice that was associated with mild skull shortening and deformation. The clinical definition of microcephaly is a reduction in head circumference that implies a decrease in brain size (Zaqout and Kaindl, 2022); we therefore postulate that OGT^C921Y mice exhibit a mild microcephaly phenotype. Although microcephaly was not reported in any of the three brothers carrying the OGT^C921Y variant (Omelková et al., 2023), it has been observed in individuals carrying other OGT-CDG variants including the OGT L254F, A259T and R284P variants (Gundogdu et al., 2018; Selvan et al., 2018; Willems et al., 2017). The development of the brain depends on several processes, including neurogenesis, neuron migration, and the balance between proliferation and differentiation of neural stem cells. Although Ogt has been shown to be important in all these processes (Cheng et al., 2020; Chen et al., 2021; Shen et al., 2021), the mechanisms underlying microcephaly in OGT-CDG remain to be investigated.

Although spontaneous seizure events have been reported for two of the three brothers affected by the OGT^C921Y variant (Omelková et al., 2023), we did not observe such seizure events in OGT^C921Y mice. Spontaneous seizures are difficult to study as they usually last a few seconds and may occur infrequently. Proper monitoring would require continuous video recording associated with electroencephalograms to detect brain activity in free-moving animals (Gu and Dalton, 2017). Such a setup has been used to observe myoclonic seizure events in a mouse model of Rett syndrome, which is an ID disorder caused by variants of MECP2 (Shahbazian et al., 2002). Inducible seizures, triggered either chemically or via sensory stimuli (light/sound), have also been demonstrated in mice carrying Syngap1 and Fmr1 mutations associated with ID disorders (Clement et al., 2012; Musumeci et al., 2000). Increased sensibility to light and sounds has been reported in an individual affected by the OGT^N648Y variant (Pravata et al., 2019b). How OGT^C921Y mice might respond to such procedures remains to be explored. This model will also be useful for investigating the effect of OGT-CDG variants on cognition and behavior in mice during early postnatal development and adulthood. Well-established behavioral paradigms can be used to assess developmental delay, locomotion, learning and memory deficits, and social and compulsive behaviors in order to further validate this model as a translational tool to study OGT-CDG.

We have previously discussed potential mechanisms underlying OGT-CDG symptoms (Pravata et al., 2020), and some of these mechanisms have been assessed in our study. Structural analyses of the OGT^C921Y mutant protein and other catalytically impaired variants have shown that these mutations affect OGT catalytic activity by disruption of the OGT acceptor binding site, leading to altered substrate binding and/or UDP-GlcNAc binding to OGT, resulting in hypo-O-GlcNAcylation of OGT protein substrates in vitro (Pravata et al., 2019a; Omelková et al., 2023; Pravata et al., 2019b). This impaired OGT catalytic activity was found to result in a reduction of global O-GlcNAc levels in the brain of OGT^C921Y mice. Similarly, hypo-O-GlcNAcylation has previously been observed in cell and Drosophila models carrying OGT-CDG variants (Pravata et al., 2019a,b). However, it is not known whether these mutations will affect all OGT protein substrates or only a subset of them. Given that O-GlcNAcylated proteins are particularly abundant in the brain, the loss of O-GlcNAcylation on OGT substrates important for proper brain function may contribute to OGT-CDG. Using a filter-based bioinformatics approach combined with structural and clinical data, we have recently predicted the presence of 38 critical O-GlcNAc sites across 22 neuronal proteins already reported to be linked to ID and developmental delay (Mitchell et al., 2022). This constitutes a list of potential candidate conveyers for OGT-CDG, and further mechanistic dissection of the OGT-CDG mouse model using a range of ‘omics’-based approaches will help to further understand this disorder. Also consistent with previous findings (Pravata et al., 2019a,b), OGA protein and mRNA levels were decreased in brain of OGT^C921Y mice. As OGA has been implicated in intelligence (Savage et al., 2018), and because Oga knockdown leads to microcephaly and hypotonia in mice (Stichelen et al., 2017), loss of OGA might also contribute to the OGT-CDG phenotype. To investigate this hypothesis, adeno-associated virus (AAV) vectors or genetic approaches could be used to elevate OGA levels to evaluate whether this could rescue some of the OGT-CDG phenotypes observed in the OGT^C921Y mouse model. Unexpectedly, OGT protein levels were also reduced in OGT^C921Y mouse brains. This has not been reported previously for cell or invertebrate models of OGT-CDG (Pravata et al., 2020, 2019a). O-GlcNAc levels have been demonstrated to regulate Ogt transcript levels (Tan et al., 2020). However, our data suggest that the decrease in OGT protein levels is not caused by disrupted transcription. Ogt mRNA levels are in fact increased, possibly as a compensatory mechanism to maintain global O-GlcNAcylation levels. OGT protein abundance might be disrupted by either protein instability or misfolding leading to proteolytic processing and/or aggregation. Reduction in OGT stability in vitro has been previously reported for OGT-CDG variants (Pravata et al., 2019a; Willems et al., 2017; Vaidyanathan et al., 2017; Selvan et al., 2018) but not for the OGT^C921Y mutant protein (Omelková et al., 2023; Pravata et al., 2019b). Some OGT-CDG variants cause conformational changes in the protein that could contribute to OGT misfolding (Gundogdu et al., 2018; Pravata et al., 2019b). Although these studies together suggest that OGT-CDG variants could lead to reduced OGT stability, this will require further investigation.

In conclusion, we have successfully generated a mouse model of OGT-CDG that will underpin further investigation of the effects of OGT-CDG variants on brain morphology, brain function and behavior. This model will also be an invaluable starting point to gain insight into OGT-CDG etiology through identification of underlying mechanisms and candidate conveyers of the disease, and will provide a platform for evaluation of potential future treatment strategies.

OGT-CDG mice were produced by microinjection under project licence PPL PB0DC8431 at the Central Transgenic Core of Bioresearch and Veterinary Services at the University of Edinburgh, UK. Female C57BL/6J mice (Charles River UK) were superovulated with 5 IU of pregnant mare serum gonadotropin (PMSG; Prospecbio, hoR-272-b) followed by 5 IU of Chorulon (hCG; National Veterinary Services, 804745) 46 h later, and mated overnight with C57BL/6J stud males (Charles River UK). Zygotes were harvested at 0.5 dpc. Editing CRISPR reagents were centrifuged through a Millipore filter (UFC30VV25) and injected into the cytoplasm of zygotes on a Zeiss Axiovert 100 using a Femtojet Xpert (Eppendorf), Transferman 4R (Eppendorf) micromanipulators, Vacutip (Eppendorf) holding pipettes and Femtotip (Eppendorf) injection needles. Injected zygotes were cultured to two-cell stage and then surgically transferred into pseudo-pregnant Crl:CD1(ICR) or Hsd:ICR (CD-1) females (Charles River UK) that had been mated with vasectomized Crl:CD1(ICR) males (Charles River UK) the night before. Genomic DNA from offspring was genotyped and sequenced to confirm the presence of the OGT^C921Y variant. Editing CRISR reagents, Cas9 nickase (1081062; Integrated DNA Technologies) and primers were purchased from Integrated DNA Technologies, and sequences are listed in Table S2. Founder OGT-CDG mice were crossed to C57BL/6J WT animals (Charles River UK) for further breeding. Animals were housed in ventilated cages with water and food available ad libitum and 12 h/12 h light/dark cycles. All animal studies and breeding were performed on in accordance with the Animal (Scientific Procedures) Act of 1986 for the care and use of laboratory animals. Procedures were carried under UK Home Office Regulation (Personal Project Licence PP8833203) with approval by the Welfare and Ethical Use of Animals Committee of University of Dundee.

Body composition data from 6- to 7-month-old male mice was obtained using the EchoMRI 4in1, in line with the protocol provided by the manufacturer (http://www.echomri.com/Body_Composition_4_in_1.aspx). Basal blood glucose levels from tail vein sampling were measured using Bayer Contour glucose meters and strips.

For MicroCT scanning, 55- to 58-day-old male mice were euthanized and frozen as intact carcasses and defrosted immediately prior to imaging. Carcasses were imaged at 140 kV and 30 μA using a Nikon XTH 225 ST MicroCT scanner at 50 μm resolution. Two-dimensional images were used to generate three-dimensional volumes using 3D slicer (https://www.slicer.org/; Fedorov et al., 2012). Coordinates of 14 landmarks on the skull were recorded from three-dimensional CT images for analysis of the skull morphology. Details of sample genotype were hidden from the researcher performing the landmark analysis.

Brain tissues from 80- to 91-day-old male mice were rapidly dissected, snap frozen in liquid nitrogen and stored at −80°C. Tissues were disrupted in phosphate-buffered saline two times at 5000 rpm for 30 s with 10 s break using a Precellys 24 Touch homogenizer (Bertin Technologies). Homogenates were split in half for further protein and RNA extractions.

Brain homogenates were lysed using 10× RIPA buffer (Cell Signaling Technology), centrifuged at 17,000 g for 20 min at 4°C, and the protein concentration was determined using a Pierce 660 nm protein assay (Thermo Fisher Scientific). Proteins were separated on precast 4-12% NuPAGE Bis-Tris acrylamide gels (Invitrogen) and transferred to nitrocellulose membrane. Membranes were incubated with primary antibodies in 5% bovine serum albumin in Tris-buffered saline buffer with 0.1% Tween-20 overnight at 4°C. Anti-OGA (1:500 dilution; HPA036141, Sigma), anti-O-GlcNAc (RL2) (1:500 dilution; NB300-524, Novus Biologicals), anti-OGT (F-12) (1:1000 dilution; sc-74546, Santa Cruz Biotechnology), mouse anti-lamin B (1:10,000 dilution; 66095-Ig, Proteintech) and rabbit anti-lamin B (1:5000; 12987-1-AP, Proteintech) antibodies were used. Next, the membranes were incubated with IR680- and IR800-labeled secondary antibodies (1:10,000 dilution; 926-32210, 926-32211, 926-68070, 926-68071; LI-COR) at room temperature for 1 h. Blots were imaged using a LI-COR Odyssey infrared imaging system, and signals were quantified using Fiji software (https://fiji.sc/). Results were normalized to the mean of each corresponding OGT^WT replicate set and represented as a fold change relative to OGT^WT.

Total RNA was purified from brain homogenates using an RNAeasy Kit (Qiagen), and then 0.5 to 1 µg of sample RNA was used for reverse transcription with the qScript cDNA Synthesis Kit (Quantabio). Quantitative PCR reactions were performed using the PerfeCTa SYBR Green FastMix for iQ (Quantabio) reagent, in the CFX Connect Real-Time PCR Detection System (Bio-Rad), employing a thermocycle of one cycle at 95°C for 30 s, and then 40 cycles at 95°C for 5 s, 60°C for 15 s and 68°C for 10 s. Data analysis was performed using CFX Manager software (Bio-Rad). Samples were assayed in biological replicates with technical triplicates using the comparative Ct method (Livak and Schmittgen, 2001). The threshold-crossing value was normalized to internal control transcripts (Gapdh, Actb and Pgk1). Primers used for qPCR analysis are listed in Table S3. Results were normalized to the mean of each corresponding OGT^WT replicate set and represented as a fold change relative to OGT^WT.

Statistical analyses were performed with Prism 9. D'Agostino–Pearson, Shapiro–Wilk and Kolmogorov–Smirnov normality tests were performed to verify normality. For data that fulfilled normality requirements, a two-tailed unpaired t-test was used for pairwise comparisons of OGT^WT and OGT^C921Y data. For data sets that did not fulfill normality, a Mann–Whitney test was used for pairwise comparisons of OGT^WT and OGT^C921Y data. The sample size (n) corresponding to the numbers of animals used for each experimental group is indicated in all figure legends.

We thank the Central Transgenic Core of Bioresearch and Veterinary Services at the University of Edinburgh for assistance in generating the mouse OGT^C921Y transgenic line and the animal facility of Dundee University for assistance in animal breeding and general maintenance. We thank Dr Stewart J. Chalmers from Aberdeen University for assistance with MicroCT data collection. We also thank Conor Mitchell for constructive feedback.

All relevant data can be found within the article and its supplementary information.