Loss of O-GlcNAcase catalytic activity leads to defects in mouse embryogenesis

O-GlcNAcylation is an essential post-translational modification that has been implicated in neurodevelopmental and neurodegenerative disorders. O-GlcNAcase (OGA), the sole enzyme catalyzing the removal of O-GlcNAc from proteins, has emerged as a potential drug target. OGA consists of an N-terminal OGA catalytic domain and a C-terminal pseudo histone acetyltransferase (HAT) domain with unknown function. To investigate phenotypes specific to loss of OGA catalytic activity and dissect the role of the HAT domain, we generated a constitutive knock-in mouse line, carrying a mutation of a catalytic aspartic acid to alanine. These mice showed perinatal lethality and abnormal embryonic growth with skewed Mendelian ratios after day E18.5. We observed tissue-specific changes in O-GlcNAc homeostasis regulation to compensate for loss of OGA activity. Using X-ray microcomputed tomography on late gestation embryos, we identified defects in the kidney, brain, liver, and stomach. Taken together, our data suggest that developmental defects during gestation may arise upon prolonged OGA inhibition specifically because of loss of OGA catalytic activity and independent of the function of the HAT domain.

O-GlcNAcylation is an essential post-translational modification that has been implicated in neurodevelopmental and neurodegenerative disorders. O-GlcNAcase (OGA), the sole enzyme catalyzing the removal of O-GlcNAc from proteins, has emerged as a potential drug target. OGA consists of an Nterminal OGA catalytic domain and a C-terminal pseudo histone acetyltransferase (HAT) domain with unknown function. To investigate phenotypes specific to loss of OGA catalytic activity and dissect the role of the HAT domain, we generated a constitutive knock-in mouse line, carrying a mutation of a catalytic aspartic acid to alanine. These mice showed perinatal lethality and abnormal embryonic growth with skewed Mendelian ratios after day E18.5. We observed tissue-specific changes in O-GlcNAc homeostasis regulation to compensate for loss of OGA activity. Using X-ray microcomputed tomography on late gestation embryos, we identified defects in the kidney, brain, liver, and stomach. Taken together, our data suggest that developmental defects during gestation may arise upon prolonged OGA inhibition specifically because of loss of OGA catalytic activity and independent of the function of the HAT domain.
O-GlcNAcylation and the O-GlcNAc cycling enzymes are critical for normal development in several organisms. In Drosophila, ogt, also known as super sex combs (sxc), is essential for viability and correct segment identity specification (15,16). Ogt KO leads to impaired embryonic growth and cell viability in zebrafish (17). Deletion of either Ogt or Oga is lethal in mice (18)(19)(20). In mammals, OGT and OGA are ubiquitously expressed, and they regulate the development of several tissues (21)(22)(23). O-GlcNAcylation modification is particularly abundant in the mammalian brain (24)(25)(26), and both O-GlcNAc cycling enzymes are important for normal development and function of the central nervous system (27)(28)(29). Very recently, missense OGT mutations have been identified and characterized in patients affected with intellectual disability (ID) in association with developmental delay and brain anomalies (30). ID is a neurodevelopmental disorder that affects 1 to 2% of the population and is characterized by impaired cognitive function and adaptive behavior (31). Phenotypic characterization of several OGT variants gave rise to a new syndrome of congenital disorder of glycosylation named OGT-CDG; however, the mechanisms underlying the patient brain phenotypes remain unknown (30). Interestingly, a reduction of OGA expression was observed in several mouse embryonic stem cell (mESC) lines carrying OGT-CDG mutations (32)(33)(34)(35), suggesting that altered OGA expression may be associated with reduced OGT activity in some mutations and might contribute in part to the neuronal phenotype in these patients. OGA has been shown to contribute to proper brain function using several animal models. In flies, loss of OGA activity affects cognition and synaptic morphology (36). Viable heterozygous Oga +/− mice exhibit learning and memory impairment (29). In addition, a genome-wide association study using 14 independent epidemiological human cohorts has associated SNPs in OGA with intelligence and cognitive function (37), suggesting a role of OGA in human cognitive function.
Deregulation of O-GlcNAcylation has also been linked to neurodegenerative disease. However, it appears difficult to define a general neuroprotective or neurodegenerative role for O-GlcNAcylation. O-GlcNAc protein levels are found reduced in Alzheimer's disease (AD) while being increased in Parkinson's disease postmortem human brain tissues (38,39). Many proteins associated with these diseases are O-GlcNAcylated, including tau, α-synuclein, and β-amyloid precursor proteins allele expresses full-length OGA with a single D285A missense mutation in the GH domain (the mutation site is in orange). B, sequences of OGA WT and OGA D285A 278 to 294 residues highlighting the D285A site are shown in green/orange. C, schematic representation of the 10 kb Oga D285A transgene targeting exon 4 to 8 in the Oga gene used to generate Oga D285A mouse embryonic stem cells (mESCs). Recombinant cell colonies, encoding FRT-flanked puromycin resistance gene, were isolated and injected into blastocysts. Highly chimeric progenies were crossed with Flp deleter (C57BL/6-Tg(CAG-Flpe) 2Arte) mice to eliminate puromycin resistance. The locations of the enzymatic restriction sites, primers, and probes are also shown. D, southern blot of WT and four targeted mESC clones (named B-C06, B-C11, B-D05, and B-E02) with the cag probe shows correct homologous recombination at the 5' side (upper panel) and 3' side (lower panel) and single integration in all clones. The expected molecular weight bands for the targeted (T) allele are shown. E, PCR of WT (40). Several studies have shown that reduction of O-GlcNAc levels leads to neurodegeneration. For example, the reduction of O-GlcNAcylation in forebrain excitatory neurons leads to progressive neurodegeneration associated with neuronal cell death, neuroinflammation, increased levels of hyperphosphorylated tau, and β-amyloid peptides in mice (41). Moreover, it has been shown that O-GlcNAc at specific sites reduced α-synuclein aggregation and cell toxicity using synthetic protein methodology (42). However, other studies have also suggested a role of increased O-GlcNAcylation in neurodegeneration. Increase of O-GlcNAc modification in the mouse hippocampus and neuronal precursor cells is associated with neuronal apoptosis (24,43). In addition, elevation of O-GlcNAc levels through pharmacological OGA inhibition causes an increase of α-synuclein accumulation and a decrease of autophagic flux prior to neuronal cell death in rat primary neurons (38). Considering the number of O-GlcNAc-modified proteins in the brain, it is likely that the role of this protein modification depends on each specific substrate and the cellular, tissue, and disease contexts.
Nevertheless, emerging evidence suggests that elevation of O-GlcNAc levels through pharmacological inhibition of OGA may be a relevant therapeutic strategy for the treatment of AD. OGA inhibition through chronic treatment with OGA inhibitors reduces tauopathy phenotypes in several mouse AD models (44)(45)(46)(47). It has been hypothesized that O-GlcNAcylation of tau could prevent its phosphorylation and aggregation (48). Although a reduction in tau hyperphosphorylation and aggregation have been observed after OGA inhibition, it remains uncertain whether it is mediated directly through O-GlcNAcylation as increase in tau O-GlcNAcylation in vivo has only been observed in one study (44). It may be possible that the observed benefits are mediated through other O-GlcNAcylated substrates as global elevation of O-GlcNAcylation is achieved upon OGA inhibition. Considering that O-GlcNAcylation is important in regulating several biological functions, it is imperative to understand the molecular mechanisms and physiological consequences of prolonged OGA inhibition.
Despite the critical role of OGA, our understanding about how OGA activity participates in development and pathological process remains limited. OGA is a 103 kDa multidomain protein (Fig. 1A); it consists of an N-terminal O-GlcNAc hydrolase catalytic domain with sequence homology to glycoside hydrolase family 84, a middle, mostly disordered, "stalk" domain and a C-terminal domain showing sequence homology to GCN5-related histone acetyltransferases (HATs) (49). Initially, OGA was reported to possess HAT activity (50), but other studies failed to reproduce this observation (51)(52)(53). OGA most likely contains a pseudo-HAT domain because it lacks key amino acids required for acetyl coenzyme A binding (52) required for histone acetylation. Thus, it has been established that OGA only possesses OGA catalytic activity (52). Phenotypic comparison of Oga KO and catalytically dead Oga D133N Drosophila alleles, however, revealed differences in behavior and neuronal phenotypes that suggest nonenzymatic functions of the OGA protein backbone (36). These additional functions could be potentially dependent on the HAT domain whose function still remains unknown.
OGA is indispensable for late embryonic development in mammals (19,54). In two Oga KO (Oga KO ) mouse models, embryos exhibited a reduction in size, and neonates died within 48 h after birth (19,54). Although the precise causes of death remain unknown, the lethal phenotype was associated with abnormal lung histology marked with a reduction in prealveolar space in 18.5 dpc embryos (19), hypoglycemia, and impaired glycogen deposition in the liver of neonates (54). Mouse Oga KO studies using mouse embryonic fibroblasts or liver tissues from homozygous animals assigned these phenotypes to different cellular defects, specifically to mitotic abnormalities, cytokinesis failure (19), and altered metabolic signaling (54,55). However, it remains unknown if these phenotypes are specifically because of the loss of OGA enzymatic activity or loss of the OGA protein.
To investigate the role of OGA catalytic activity in development and disease, we generated a knock-in Oga mouse model (Oga D285A ), where a key residue for catalytic activity, D285, responsible for O-GlcNAc moiety binding and hydrolysis, is mutated to alanine (56). Therefore, the Oga D285A animals lack OGA activity while still expressing full-length OGA protein. This genetic strategy can model chronic inhibition of OGA activity in mouse, allowing the evaluation of the longterm effects of OGA inhibition and the investigation of the potential role of the HAT domain of OGA in vivo. We show that loss of OGA activity leads to perinatal lethality, organ defects, and tissue-specific disruption in O-GlcNAcylation homeostasis in embryos. These results highlight the essential role of the OGA domain during development independently of the HAT domain. Although the use of this model in adult animals is limited because of the lethality observed at the homozygous level, a similar strategy could be used to generate inducible mouse model carrying the Oga D285A mutation that will allow the future investigation of long-term adverse effects of prolonged OGA inhibition during adulthood in vivo.

Loss of OGA catalytic activity leads to perinatal lethality
To dissect the role of OGA catalytic activity in mammalian development, we designed an Oga constitutive knock-in mouse model where OGA enzymatic activity is abolished (Fig. 1, A and B). Previous studies have shown that OGA Asp285 (D285) is essential for both O-GlcNAc hydrolysis and O-GlcNAc binding (56)(57)(58), whereas mutation to alanine causes nearly 100% loss of catalytic efficiency in vitro (59). We introduced the OGA D285A mutation into mESCs by homologous recombination (Fig. 1C). Correct single integration of the point mutation into mESCs was determined by Southern blotting (Fig. 1D) and PCR genotyping (Fig. 1E). Clones positive for the mutation were used for blastocyst injections to generate an Oga D285A mouse line lacking OGA enzymatic activity. DNA from positive offspring was sequenced to confirm the presence of the D285A mutation (Fig. 1F).
O-GlcNAc homeostasis is altered in Oga D285A mice in a tissuespecific manner Although homozygous Oga D285A mice are not viable, the heterozygous Oga D285A/+ animals survived to adulthood. Previous studies have revealed increased protein O-GlcNAcylation in heterozygous Oga KO/+ hippocampus (29) and liver (54). We tested O-GlcNAc levels in the Oga D285A/+ mice to assess O-GlcNAc homeostasis. Mouse brain tissue obtained from 55to 63-day-old WT and heterozygous Oga D285A/+ adult animals (n = 6 per group) were analyzed using an antibody specific for O-GlcNAcylated proteins. Protein O-GlcNAcylation was comparable in WT (mean ± SD fold change to WT, 1.1 ± 0.5 fold change) and heterozygous Oga D285A/+ samples (0.9 ± 0.7 fold change) (Fig. 3A). It has been established that OGT and OGA protein levels are sensitive to changes in protein O-GlcNAcylation in mammalian cells (34,60). We next investigated such possible compensatory mechanisms by assessing in Oga D285A/+ (OGT: 0.6 ± 0.2 fold change) compared with WT (OGT: 1.0 ± 0.7 fold change) did not reach statistical significance (t test, p = 0.164). Our data indicate that heterozygous loss of OGA catalytic activity is compensated by altered OGA protein levels in adult heterozygous Oga D285A/+ animals to maintain normal level of protein O-GlcNAcylation (Fig. 3, B and C). Next, we investigated the effect of the D285A mutation on protein O-GlcNAcylation in homozygous Oga D285A embryos. Previous mouse Oga KO studies showed global elevation of O-GlcNAc level defects in brain and liver (54,55). Thus, brain and liver samples isolated from Oga D285A 15.5 dpc mouse embryos (n = 4 per group) were subjected to Western blotting. This revealed an increase of O-GlcNAcylation in brain tissues from homozygous Oga D285A/D285A embryos (3. In the brain, a twofold increase of OGA protein levels and an approximately threefold reduction of OGT protein levels were detected in homozygous Oga D285A/D285A samples (OGA: 1.9 ± 0.4, OGT: 0.3 ± 0.03 fold change) compared with WT brain samples (Fig. 4, B and C). OGT and OGA protein levels appeared similar in WT samples (OGA: 1.0 ± 0.1, OGT: 1.0 ± 0.3 fold change) and heterozygous Oga D285A/+ samples (OGA: 1.1 ± 0.1, OGT: 0.7 ± 0.1 fold change). Next, we investigated Oga and Ogt mRNA levels. No detectable differences in brain Oga mRNA levels were observed between the three genotypes (Oga +/+ : 1.0 ± 0.8, Oga D285A/+ : 0.8 ± 0.4, and Oga D285A/D285A : 0.8 ± 0.3 fold change) (Fig. 4D). However, reductions of Ogt mRNA levels were apparent in homozygous Oga D285A/D285A (0.5 ± 0.1 fold change) and heterozygous Oga D285A/+ (0.6 ± 0.2 fold change) embryos, respectively, compared with WT samples (1.0 ± 0.3 fold change), although the differences between heterozygous and WT samples did not reach statistical significance (p = 0.35, Kruskal-Wallis multiple comparisons test) (Fig. 4D). In the embryo liver, an increase in OGA protein level was detected in homozygous Oga D285A/D285A samples (OGA: 2.2 ± 0.98) compared with WT liver samples. Although a reduction in OGT protein level (OGT: 0.6 ± 0.2) was apparent, the difference did not reach statistical significance (p = 0.51, Kruskal-Wallis multiple comparisons test). No difference in OGA and OGT protein levels was observed between heterozygous (OGA: 0.7 ± 0.2, OGT: 0.5 ± 1.1) and WT embryos (OGA: 1.0 ± 0.5, OGT: 1.0 ± 0.7) (Fig. 4, B and C). An increase of Oga mRNA levels was apparent in homozygous Oga D285A/D285A (3.0 ± 1.0) compared with WT samples (1.0 ± 0.2) and heterozygous Oga D285A/+ embryos (0.8 ± 0.2). No detectable differences in liver Ogt mRNA levels were observed between the three genotypes (Oga +/+ : 1.0 ± 0.3, Oga D285A/+ : 0.7 ± 0.2, and Oga D285A/D285A : 0.7 ± 0.3 fold change) (Fig. 4D). Taken together, these results suggest that loss of OGA catalytic Microcomputed tomography reveals widespread organ defects in Oga D285A embryos We next performed an unbiased in-depth analysis of growth and morphology on 18.5 dpc mouse embryos to identify organ defects that could explain perinatal lethality specific to loss of OGA catalytic activity. We employed X-ray microcomputed tomography (microCT) to capture the anatomy of intact whole embryos. In total, 70 anatomical structures were scored in WT, heterozygous, and homozygous embryos (n = 8 per group). Analysis of microCT imaging revealed anatomical abnormalities in 26 regions across all genotypes (Table S1). Two WT embryos showed a single abnormality, one in the brain and one in the heart. Three heterozygous Oga D285A/+ embryos exhibited morphological defects in eight regions, including heart, stomach, and kidney (Table S1). The majority of the abnormalities were observed in homozygous Oga D285A/D285A embryos with seven samples revealing defects in 21 areas (Table 1). For these regions, all WT Oga +/+ embryos appeared normal, whereas morphology of the stomach lumen (two cases) and kidney (one case) was affected in heterozygous Oga D285A/+ embryos. We decided to focus on essential organs that were affected in at least four Oga D285A/D285A samples, namely the kidneys, liver, stomach, and brain ventricles. Half of the embryos showed abnormalities in the kidneys exhibiting dilated renal pelvis that (brain: threefold reduction, **p = 0.01; liver: twofold reduction, p = 0.51) and heterozygous Oga D285A/+ samples (brain: 2.2-fold reduction, p = 0.233; liver: 1.8fold reduction, p > 0.999). D, quantification of Oga mRNA levels indicated no differences for all genotypes in the brain, whereas in the liver, Oga mRNA levels are increased in homozygous Oga D285A/D285A sample compared with WT control (threefold increase, p = 0.188) and heterozygous Oga D285A/+ samples (3.8-fold increase, *p = 0.013). Quantification of Ogt mRNA levels indicated that homozygous Oga D285A/D285A (twofold reduction, *p = 0.033) and heterozygous Oga D285A/+ (1.5-fold reduction, p = 0.35) samples show a decrease in Ogt mRNA levels compared with WT control in the brain, whereas no differences in Ogt mRNA levels were observed between all genotypes in the liver. OGA, O-GlcNAcase; OGT, O-GlcNAc transferase. Table 1 List of abnormalities found in heterozygous Oga D285A/+ and homozygous Oga D285A/D285A embryos The number of affected embryos (n = 8 per genotype) for each abnormality is reported. Green indicates no defects found, and yellow-red gradient indicates that abnormal features were detected.
Loss of OGA activity leads to defects in mouse embryogenesis developed to advanced unilateral hydronephrosis in two Oga D285A/D285A embryos (Fig. 5A). Enlarged intralobular space in the liver was observed in 75% of Oga D285A/D285A animals, whereas in the two most severe cases, it was associated with reduced left/right and caudal lobes (Fig. 5B). The lumina of the stomach appeared reduced in half of the Oga D285A/D285A embryos (Fig. 5C). In addition, there were abnormalities in the fourth brain ventricles in five cases suggesting developmental defects in the brain (Fig. 6A).
To explore possible volume differences in the affected organs, 3D segmentation was performed. We detected a reduction of volume for the liver, kidney, stomach lumen, and brain in the  representative microCT images of axial sections of the abdomen region and 3D volume renderings of the right kidney (in red) from a WT, heterozygous Oga D285A/+ , and homozygous Oga D285A/D285A 18.5 dpc embryos. The renal pelvis appeared dilated in the Oga D285A/D285A embryos (yellow arrow) and is associated with advanced unilateral hydronephrosis in two animals (yellow asterisk). Quantification of the volume of the right kidney showed no difference in kidney volume between all genotypes after normalization to whole embryo body. B, representative microCT images of axial sections and 3D volume renderings (in red) of the liver from a WT, heterozygous Oga D285A/+ , and homozygous Oga D285A/D285A 18.5 dpc embryos. The intralobular spaces from the homozygous Oga D285A/D285A appeared enlarged as shown with yellow arrow compared with WT control. Quantification of the volume of the liver showed no difference in liver size between all genotypes after normalization to whole embryo body. C, representative microCT images of axial sections of the abdomen region and 3D volume renderings of the stomach lumen (in red) from WT, heterozygous Oga D285A/+ , and homozygous Oga D285A/D285A 18.5 dpc embryos. The stomach lumen from the homozygous Oga D285A/D285A mice appeared reduced as shown with yellow arrow compared with WT and heterozygous Oga D285A/+ embryos. Quantification of the volume of the stomach lumen showed a possible reduced size of the lumen in homozygous Oga D285A/D285A compared with heterozygous Oga D285A/+ (p = 0.102) and WT (p = 0.243) embryos after normalization to whole embryo body although this did not reach statistical significance.

Discussion
Previous studies have established that Oga is essential for mammalian development (19,54). However, it is not known whether this is linked to OGA activity or functions of the other OGA domains. Similar to the Oga KO mouse models, the loss of OGA catalytic activity leads to reduced growth and perinatal lethality in our OGA catalytic-deficient model, suggesting that it is the loss of OGA enzymatic function that causes lethality and alteration in mouse embryonic development. In the Oga KO + / +  Figure 6. Oga D285A/D285A embryos show enlarged brain ventricles. Data were analyzed using one-way ANOVA with Tukey's multiple comparisons test, n = 8 for all genotypes. The scale bars for the grayscale sections are represented. A, representative microcomputed tomography images of axial sections of the brain and 3D volume renderings of the ventricular system from a WT, heterozygous Oga D285A/+ , and homozygous Oga D285A/D285A 18.5 dpc embryos. B, quantification of the volume of the brain showed no differences between all embryo genotypes after normalization to whole embryo body. C, quantification of the volume of the brain ventricles showing possible increases of the ventricular system in heterozygous Oga D285A/+ (p = 0.281) and homozygous Oga D285A/D285A embryos (p = 0.210) compared with WT control embryos after normalization to brain volume. A Dixon's Q test identified the smallest value (circled) in homozygous Oga D285A/D285A group as an outlier. Once this value is excluded from the analysis, the difference between Oga D285A/D285A and the WT control becomes statistically significant (**p = 0.002) using one-way ANOVA with Tukey's multiple comparisons test.
Loss of OGA activity leads to defects in mouse embryogenesis models, defects in lungs and liver tissues were observed in the homozygous embryos using classic histological techniques.
Here, we used microCT to investigate whether other organ defects could arise from loss of OGA activity. We showed that Oga D285A/D285A pups, from the same mouse strain used for the Oga KO models, display various organ-specific defects with varying penetrance among the embryos that were not previously described. The most prominent abnormalities were found in the kidney exhibiting dilatation of the renal pelvis and advanced hydronephritis, the latter is associated with kidney infection and kidney failure (61) posing a significant risk of postnatal death (62). Defects in the liver, stomach, and brain were observed in parallel, and accumulation of abnormalities in several organs could contribute to perinatal lethality in the Oga D285A/D285A mice. Loss of OGA activity resulted in increased global O-GlcNAc levels in homozygous Oga D285A/D285A embryos supporting that the D285A mutation impairs OGA activity in mouse. Increased protein O-GlcNAcylation has been associated with several chronic pathologies, including cancers, osteoarthritis, diabetes, diabetic vascular dysfunction, and diabetic nephropathy (63)(64)(65)(66)(67). Hyperglycemia is a hallmark of diabetes, and high glucose levels induce elevation of global O-GlcNAcylation by increased flux through the hexosamine biosynthetic pathway. Raised O-GlcNAc levels were found in tissues from diabetic animals and humans (68)(69)(70) as well as in pups of diabetic female mice (71). Several studies showed that O-GlcNAcylation contributes to hyperglycemia-induced tissue damage in heart and kidney of adult patients (69). Furthermore, maternal hyperglycemia causes developmental delay associated with congenital defects, also called as diabetic embryopathy, affecting the kidneys, central nervous system, heart, and skeletal system among others (72,73). Inhibition of OGT blocked the negative impact of hyperglycemia or glucosamine supplementation on blastocyst formation, cell number, and apoptosis during mouse embryogenesis (74). Reduction of global O-GlcNAcylation through OGT inhibition in diabetic pregnant mice in vivo reduced neural tube defects in embryos (75). In our Oga D285A model, elevated O-GlcNAc levels independent of hyperglycemia were also associated with anomalies during mouse development, demonstrating a direct link between excess of O-GlcNAcylation and developmental defects.
Interestingly, we observed mild enlargement in the brain ventricles of both Oga D285A/+ and Oga D285A/D285A embryos indicating that the phenotypes observed previously in brainspecific Oga KO mice (55) are caused by the loss of catalytic activity of OGA. Ventricle enlargement, frequently associated with neurodevelopmental conditions including ID, originates from defects in neurogenesis, proliferation, or ciliary development (76)(77)(78). In the brain-specific Oga KO model, an imbalance between neuronal proliferation and disturbed neurogenesis was detected in neonates (55) that can be linked to hippocampal-dependent spatial learning and memory defects observed in adult heterozygous Oga KO/+− . A common hallmark in OGT-CDG mutations associated with ID is the reduction in OGA expression, suggesting a possible mechanism that might contribute in part to the pathogenesis of ID (30,(32)(33)(34)(35). Our model could be used to further examine the role of OGA and its activity in neurodevelopment and OGT-CDG pathogenesis.
Loss of OGA activity resulted in increased global O-GlcNAc levels accompanied with an increase in OGA protein levels and a decrease in OGT protein levels, providing further evidence for a compensatory molecular response to maintain O-GlcNAc homeostasis in the brain in both Oga D285A/+ adults and Oga D285A/ D285A embryos. Although similar upregulation was observed for O-GlcNAc and OGA levels in the liver of embryos, OGT expression remained unchanged suggesting a tissue-specific response. Previous reports have shown that pharmacological inhibition of OGA activity led to an increase of Oga mRNA expression and OGA protein levels (60,79,80). Similarly, we detected increased Oga mRNA and OGA protein levels in liver samples from Oga D285A/D285A embryos. The primer set used to detect Oga mRNA levels flank exon 1, allowing the evaluation of mature Oga mRNA levels. The data suggest that transcriptional upregulation of Oga mRNA levels occurs to produce more proteins to compensate for loss of catalytic activity. However, the levels of Oga mRNA remained unchanged in the brain of Oga D285A/D285A embryos suggesting that OGA regulation in this tissue may occur at the post-transcriptional level. Interestingly, OGA is itself O-GlcNAcylated, which could provide a mechanism for such post-transcriptional regulation (81,82).
The levels of OGT protein were reduced in the brain. Similarly, OGT protein levels were found decreased in association with reduction of OGA protein expression in models of two OGT-CDG mutations (32,33). We showed that reduction of OGT protein levels was also associated with a reduction of Ogt mRNA levels. Taken together, these data indicate possible transcriptional control of OGT expression influenced by OGA activity. OGAmediated regulation of Ogt transcription through cooperation with the HAT p300 and transcription factor CCAAT/enhancerbinding protein β has been described in vitro in WT primary mouse hepatocytes (83). In addition, elevated O-GlcNAc because of pharmacological inhibition of OGA promotes the retention of Ogt RNA intron 4 through the Ogt intronic splicing silencer leading to Ogt mRNA degradation (84). The primer set we used to evaluate the levels of Ogt mRNA levels flank exons 4/5, allowing the detection of the mature form of Ogt mRNA (85). Further experiments will be needed to evaluate whether the reduction of mature Ogt mRNA levels we observed in the brain is due to less Ogt mRNA produced or because of an increase in intronic retention leading to Ogt mRNA degradation. Together with our data, these studies suggest that loss of OGA activity induces alteration of Ogt transcription to maintain in vivo O-GlcNAc homeostasis in the brain but not in the liver during gestation. Although the precise mechanisms associated with OGA/OGT regulation in the Oga D285A mice remain to be explored, our results indicate that mechanisms for O-GlcNAc homeostasis maintenance may differ from one tissue to another or only occur in some tissues. A similar strategy could be used to generate a mouse model carrying an inducible Oga D285A allele to investigate the physiological consequences of prolonged OGA inhibition during adulthood.
A number of studies suggest that elevation of O-GlcNAcylation through pharmacological inhibition of OGA could provide therapeutic benefit for chronic neurological conditions, such as temporal lobe epilepsy (86), AD (44)(45)(46)(47), and amyotrophic lateral sclerosis (87). The degree of OGA inhibition needed to achieve the desired biological response while minimizing any potential side effects is an important consideration. In the brain, at least 80% of OGA inhibition is necessary to achieve global protein O-GlcNAcylation elevation in vivo (44). This is in accordance with our results showing that heterozygous Oga D285A/+ mice displayed unchanged global O-GlcNAc levels. Although heterozygous Oga D285A/+ mice are viable and do not develop overt developmental phenotypes, complete loss of OGA activity in homozygous Oga D285A/D285A mice led to elevated O-GlcNAc levels, perinatal lethality, and abnormal development. The phenotypes observed in the Oga D285A/D285A mice suggest the possibility of adverse on-target side effects upon prolonged inhibition of OGA. Although the use of adult and embryo tissues is a different context and no apparent toxic effects have been reported to date because of prolonged exposure of OGA inhibitors in adult animals, the investigations are mainly limited to the brain (44,47,88), and the effects of long-term OGA inhibition in other tissues remain unclear.
Two splice variants of human OGA exist: (1) a full-length nucleocytoplasmic isoform that contains the N-terminal glycoside hydrolase domain together with the C-terminal HAT domain and (2) a short nuclear isoform that lacks the HAT domain and shows lower OGA activity compared with the fulllength isoform (89,90). OGA undergoes caspase-3 cleavage giving rise to two fragments that individually lack OGA catalytic activity. However, the two fragments can reassemble and restore fully functional OGA in cells (51). This suggests that the C-terminal part of OGA that includes the HAT domain may be important for catalytic activity. Although it remains unknown whether this domain possesses any additional functions, phenotypic comparison of Oga D285A and Oga KO mouse models reveals that the OGA domain is essential for late embryonic development and perinatal survival.
In summary, we generated a mouse model lacking OGA catalytic activity, providing a genetic model for prolonged inhibition of OGA in vivo. This model will be useful to study the role of O-GlcNAcylation in development and to understand the potential function of the noncatalytic OGA domains. We showed that loss of OGA activity leads to perinatal lethality, organ defects, and tissue-specific disruption in O-GlcNAcylation homeostasis in mice.

Experimental procedures
Generation of Oga D285A knock-in mice Oga D285A knock-in (C57BL/6NTac-Mgea5 tm3592(D285A)Arte ) mice were generated by Taconic Artemis GmbH via insertion of the constitutive knock-in allele and subsequent Flpe-mediated deletion of a puromycin resistance cassette (Fig. 1C). First, Oga D285A mESCs were created via insertion of the knock-in allele via homologous recombination in C57BL/6NTac (Art B6 3.6) mESCs. The targeting vector coding for a 10-kbsequence spanning exon 4 to 8 of the Oga gene was electroporated into mESCs. The construct also contained a flippase recognition target-flanked puromycin resistance cassette in the intronic region between intron 6 and 7 allowing for subsequent isolation of recombinant clones (Fig. 1C). Correct homologous recombination and single integration at both 5' and 3' sides in mESCs clones was validated using the cag probe that detects a region located within the puromycin selection cassette. The genomic DNA was digested with either SphI for the 5' side or EcoRV for the 3' side and analyzed by Southern blot. The presence of the D285A mutation and the single integration was determined by PCR using 10328_1 and 10328_2 primers followed by sequencing using the 10610_135 primer.
Then, 10 to 15 cells from three selected mESC colonies bearing the D285A mutation were injected into 3.5-days blastocysts from BALB/c female mice (BALB/cAnNTac; Taconic Artemis GmbH) in Dulbecco's modified Eagle's medium supplemented with 15% fetal calf serum under mineral oil. After recovery, 48 to 51 blastocysts were transplanted into nine pseudopregnant NMRI females (BomTac:NMRI; Taconic Artemis GmbH). Ten highly chimeric progenies (>50%) were selected based on their coat color and bred with flippase (Flp) deleter C57BL/6 mice (C57BL/6-Tg(CAG-Flpe)2Arte; Taconic Artemis GmbH) to eliminate the puromycin cassette. Eight pups heterozygous for the D285A mutation were used as colony founders. Primers used for genotyping and validation are listed in Table S2.

Animal husbandry and genotyping
Founder Oga D285A/+ heterozygous mice were crossed to C57BL/6J WT animals (Charles River UK Limited) inhouse. The line was initially bred by intercrossing heterozygous animals, then maintained by backcrossing to C57BL/6J background for two generations. Animals were housed in standard holding cages with water and food available ad libitum and 12/12 h light/dark cycles throughout the study. All animal studies and breeding were performed in accordance with the Animal (Scientific Procedures) Act of 1986 for the care and use of laboratory animals. Procedures were carried under United Kingdom Home Office Regulation with approval by the Welfare and Ethical Use of Animals Committee of University of Dundee.
Genotyping was performed by diagnostic PCR using Thermococcus kodakaraensis Hot Start DNA polymerase (EMD Millipore) on genomic DNA isolated from ear-notch biopsy with 10609_129 and 10609_130 primers (Table S2) that amplified 395 base pair fragment of the knock-in and 320 base pair fragment of the WT allele. Animals were genotyped after weaning at 23 to 37 days old.

Weight measurement
Weights of the embryos were determined after fixation on analytical scale. Excess liquid from the surface of the embryos was removed with paper towels prior to measurement. Statistical significance was calculated using one-way ANOVA with Tukey's post hoc test.

CpOGA treatment
Mouse embryo brain and liver lysates lacking GlcNAcstatin G were treated with recombinant clostridium perfringens O-GlcNAcase (CpOGA) to test the specificity of the anti-O-GlcNAc antibody (RL2). Recombinant glutathione-S-transferase-tagged CpOGA was expressed and purified as described earlier (91). Lysates containing 60 μg protein at 0.85 μg/ml concentration were incubated with 10 μg recombinant WT CpOGA for 60 min at 30 C. The reaction was stopped by addition of SDS loading buffer containing 1% 2mercaptoethanol and boiling the samples for 5 min.

Real-time quantitative PCR
Total RNA was purified using RNAeasy Kit (Qiagen), and then 1000 ng 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 (BioRad), 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 (BioRad). Samples were assayed in biological quadruplicates with technical triplicates using the comparative Ct method. The threshold-crossing value was normalized to internal control transcripts (Gapdh, Actb, and Pgk1). Primers used are listed in Table S3. Results were normalized to the mean of each corresponding WT replicates set and represented as a fold change relative to WT. Statistical significance was evaluated using one-way ANOVA with Kruskal-Wallis multiple comparisons test on ΔCt values obtained for each biological replicate.

Sample preparation and imaging parameters for X-ray microCT
Crosses between heterozygous Oga D285A/+ female and Oga D285A/+ male mice were set up, and plug formation in the females was monitored to determine day 0.5 after fertilization. Pregnant females were sacrificed 18 days after plug formation with increasing concentration of carbon dioxide, and the femoral artery was severed. Embryos of 18.5 dpc were dissected into 6-well plate containing ice-cold PBS (pH 8). Embryos were washed twice with ice-cold PBS, and small tail samples were collected for genotyping. Embryos were fixed in ice-cold 4% paraformaldehyde in phosphate buffer solution for 7 days on roller at 4 C. Fixed embryos were stored in 1% paraformaldehyde in phosphate buffer until shipment and further processing. At MRC Harwell, samples were rinsed in distilled water, then placed in 15 ml 50% Lugol's solution prepared in MilliQ distilled water, and kept at room temperature on the rocker, in vials wrapped in tinfoil. About 50% Lugol's solution was changed every other day for 14 days to achieve contrasting. Embryos were rinsed with MilliQ distilled water for a minimum of 1 h before they were embedded in 1% w/v Iberose-High Specification Agarose (AGR-500) made in MilliQ distilled water and left at room temperature to set for 2 h. 3D imaging of samples was performed using a Skyscan 1172 microCT scanner (Bruker) with aluminium filter setting, scanning with 70 kV, the image pixel size set to 5 μm, employing a rotation step 0.25 while achieving 180 rotation along the anterior-posterior axis.

Image processing
Acquired images were reconstructed using NRecon Reconstruction (Bruker) software. Data were further cropped, and TIFF stacks were generated using Harwell Automated Recon Processor (HARP; Harwell) software. Image series were converted into .nrrd file format and viewed using open-source Fiji-based 3D Viewer and on 3D Slicer (version 4.10.1; https://www.slicer.org/), an open-source medical image processing and visualization system. Segmentation and 3D volume rendering for detailed morphological analysis were performed with the ITK-SNAP medical image segmentation tool (version 3.6.0 (92)). To eliminate bias, image stacks were renamed, and the genotype was blinded for the analysis. Statistical significance was calculated using one-way ANOVA with Tukey's multiple comparisons test. The Dixon's Q test was performed for outlier identification in the sample groups according to Rorabacher (93). The experimental Q value (Q exp ) was calculated from raw data using Q exp = (x 2 − x 1 )/(x n − x 1 ), where x 1 is the value of interest, x 2 , the closest value to x 1 in the data set, and x n the highest value in the data set. Q exp was next compared with the Q critical value (Q crit ) of 0.526 for n = 8 for a confidence level of 95% using the Q-table. Data that displayed a Q exp higher than the Q crit of 0.526 were considered as an outlier at a confidence level of 95%.

Data availability
All data are contained in the article.

Supporting information-This article contains supporting information.
Acknowledgments-We thank our departmental support teams for their assistance (Medical Research Council Genotyping and Tissue Culture teams) and the School of Life Sciences Biological Services (all resource units) for the essential management, maintenance, and husbandry of mice. We thank Olawale Raimi for expressing and purifying clostridium perfringens O-GlcNAcase reagent and Andrew T. Ferenbach for assistance with sequencing data and molecular reagents.