The emerging link between O-GlcNAcylation and neurological disorders

Abstract

O-linked β-N-acetylglucosaminylation (O-GlcNAcylation) is involved in the regulation of many cellular cascades and neurological diseases such as Alzheimer’s disease (AD), Parkinson’s disease (PD), and stroke. In the brain, the expression of O-GlcNAcylation is notably heightened, as is that of O-linked N-acetylglucosaminyltransferase (OGT) and β-N-acetylglucosaminidase (OGA), the presence of which is prominent in many regions of neurological importance. Most importantly, O-GlcNAcylation is believed to contribute to the normal functioning of neurons; conversely, its dysregulation participates in the pathogenesis of neurological disorders. In neurodegenerative diseases, O-GlcNAcylation of the brain’s key proteins, such as tau and amyloid-β, interacts with their phosphorylation, thereby triggering the formation of neurofibrillary tangles and amyloid plaques. An increase of O-GlcNAcylation by pharmacological intervention prevents neuronal loss. Additionally, O-GlcNAcylation is stress sensitive, and its elevation is cytoprotective. Increased O-GlcNAcylation ameliorated brain damage in victims of both trauma-hemorrhage and stroke. In this review, we summarize the current understanding of O-GlcNAcylation’s physiological and pathological roles in the nervous system and provide a foundation for development of a therapeutic strategy for neurological disorders.

Introduction

Glycosylation is the covalent attachment of glycans to proteins or lipids with an O- or N- (or occasionally a C- or S-) linkage, usually associated with complex polysaccharides with branches. The spectrum of glycosylation’s activities expanded when O-linked β-N-acetylglucosaminylation (O-GlcNAcylation) was first found in murine lymphocytes during the early 1980s [1, 2]. This event was recognized as a single monosaccharide’s attachment to serine and threonine of nucleocytoplasmic proteins without subsequent elongation [3]. After decades of study, more than one thousand proteins bearing O-GlcNAcylation were identified in species such as viruses, bacteria, filamentous fungi, insects, plants, and animals. Protein O-GlcNAcylation acts much like phosphorylation, which is maintained in a delicate and dynamic homeostatic state in vivo and mediates cellular signaling transduction by modulating protein function [4]. This scenario plays an essential role in the process of cell signaling, protein homeostatic maintenance, and gene expression [5–8]. The interaction between protein O-GlcNAcylation and phosphorylation was highly emphasized in an early report that proposed this bipartisan activity as “Yin-Yang theory” [9, 10]. Crosstalk between protein O-GlcNAcylation and other protein translational modifications, such as protein acetylation, ubiquitination, and sumoylation, was also found in recent years [11–13]. Moreover, protein O-GlcNAcylation is commonly referred to as a nutrient sensor for cells owing to its close connection with the hexosamine biosynthesis pathway (HBP) (Fig. 1). The end product of HBP is UDP-GlcNAc, which is used as a donor substrate for O-linked N-acetylglucosaminyltransferase (OGT) and regulates OGT’s enzymatic activity, in turn providing substrate selectivity, as well as regulating O-GlcNAclyation levels in cells [14, 15]. The cellular UDP-GlcNAc level is restricted not only by glucose flux but also, mainly, by amino acid flux. Lipid metabolism and nucleotide metabolism are also involved [16] (Fig. 1). As the essential role of O-GlcNAcylation in the regulatory networks of cellular biological process came to light, its importance grew as a regulator of cell signaling and as a bridge between extracellular stimuli and cellular stress responses.

Fig. 1.

Hexosamine biosynthetic pathway (HBP). Glucose flux enters mainly the glycolysis pathway and glycogen synthesis pathway, as well as using the pentose phosphate pathway. Approximately 5% of glucose flux is redirected into the HBP. Glucose is converted into UDP-GlcNAc by a series of enzymes including l-glutamine:fructose-6-phosphate amidotransferase (GFAT), glucosamine-6-phosphate acetyltransferase (Emeg32), phosphate-acetylglucosamine mutase (mutase) and UDP-GlcNAc pyrophosphorylase (pyrophosphorylase), which is a sugar donor to the O-GlcNAcylation reaction. OGT catalyzes transfer of the O-GlcNAc moiety to serine or threonine residues of proteins, whereas OGA catalyzes removal of the O-GlcNAc group. Using the end product of HBP, O-GlcNAcylation is linked with carbohydrate, protein, lipid, and nucleotide metabolism. OSMI-1 and ST045849 are OGT inhibitors; Thiamet G and PUGNAc are OGA inhibitors

Protein O-GlcNAcylation and O-GlcNAc cycling enzymes are widely distributed in brain tissues; in particular, OGT is highly expressed at neuronal synapses. As a unique metabolic pathway that responds to conditions of cellular stress, i.e., oxidative, osmotic and chemical stress, O-GlcNAcylation is notably responsive [17]. Accordingly, changes in its level may exert a protective effect during acute stress conditions [18, 19]. Aberrant O-GlcNAcylation abolishes this protection and promotes the progression of multiple diseases, the choice of which depends on the cellular context. In this review, we will summarize the physiological and pathological roles of O-GlcNAcylation in neurological disorders, as well as the implications of exploiting this phenomenon for therapeutic development.

Protein O-GlcNAcylation: general overview

Functional consequences of O-GlcNAcylation

Differing from classical N- or O-linked glycosylation, which transpires mainly during protein biosynthesis in the endoplasmic reticulum and Golgi apparatus, O-GlcNAcylation occurs in the milieu of a great diversity of proteins localized in nuclei, cytoplasm, and mitochondria. Some of those proven to bear O-GlcNAcylation include Nup62, RNA polymerase II, c-Myc, Sp1, AMPK, HIF-1α, and AKT.

As a monosaccharide attachment, O-GlcNAcylation presents a technical challenge with respect to identification and site mapping of protein owing to the long-time lack of appropriate research tools. The classical methods of biochemistry were initially used to detect protein O-GlcNAcylation, involving H³-labeled UDP-Galactose which is transferred to the O-GlcNAc moiety of proteins by β1-4-galactosyltransferase (GalT). Autoradiography was then applied to identify the modified proteins [1]. Subsequently, a series of antibodies against O-GlcNAcylation were developed, although only a few of those reported were site specific [20–24]. Among those, CTD110.6 and RL2 were widely used for immunoblotting, immunofluorescence, and immunohistochemistry [20, 25, 26]. Recently, a fluorescein-conjugated antibody became commercially available for flow cytometry detection [27]. Now, after adapting biological mass spectrometry (MS) for this purpose, several methods based on affinity chromatography or chemical/chemoenzymatic derivatization were applied to enrich O-GlcNAcylated proteins and peptides before sequence analysis by MS. Combined with several MS detection approaches, such as electron-transfer dissociation (ETD)-MS/MS, varying number of proteins with O-GlcNAcylation and O-GlcNAcylation sites were identified in mammalian cells [28–31]. An earlier proteomic research identified 30 O-GlcNAcylated proteins from an extract of rat brain and provided site mapping information using Michael addition with dithiothreitol (BEMAD) technique followed by affinity chromatography and liquid chromatography (LC)–MS/MS [32]. Khidekel and colleagues developed a strategy of quantitative MS-based proteomics, namely quantitative isotopic and chemoenzymatic tagging, to monitor the dynamics of O-GlcNAcylation in rat brains after external stimulation [33]. Using a modified chemical/enzymatic photochemical cleavage method for enrichment, a total of 458 O-GlcNAc sites in 195 proteins were identified in the samples from mouse brains [34]. The astonishing result that followed was obtained by combining lectin weak-affinity chromatography and reversed-phase liquid chromatography (RPLC)–ETD-MS/MS, with which a total number of 1750 O-GlcNAc sites was identified in the synaptosomes from mouse brains [35]. These results prove that about 10–20% of proteins are O-GlcNAcylated, whereas 40–60% of proteins are phosphorylated. Moreover, quantitation methods based on MS have been developed and favor elucidating the function of O-GlcNAcylation in cellular processes [36, 37].

The biological effects of O-GlcNAcylation on proteins have been studied extensively in diverse settings, and emerging evidence suggests that O-GlcNAcylation impacts numerous protein functions. Akin to N-glycans, O-GlcNAcylation is proposed to stabilize proteins by enhancing their resistance to thermal unfolding or aggregation. O-GlcNAcylations are mostly found in intrinsically disordered regions of proteins, which are apt to be misfolded and clumped in aggregates. In that context, the protective role of O-GlcNAcylation against misfolding and aggregation, including such neural proteins as tau and β-amyloid, is detailed below. As an example of this point, less protein aggregation in CHO cell lines was found after transfection with an OGT overexpression construct [38]. In contrast, the thermal stability of Sp1 was decreased by β-N-acetylglucosaminidase (OGA) overexpression in cells [39]. Additionally, the transcriptional activator β-catenin also became stabilized by O-GlcNAcylation in cancer cells, dependent on nutrient conditions [40].

Regarding proteins with catalytic activity, O-GlcNAcylation influences mainly their enzymatic activity and substrate specificity. Protein O-GlcNAcylation of endothelial nitric oxide synthase (eNOS), which is responsible for nitric oxide (NO) production in cultured bovine aortic endothelial cells and human coronary artery endothelial cells, inhibits its phosphorylation by Akt and suppresses its ability to generate NO [41, 42]. Similarly, the rate-limiting enzyme of glycolysis, phosphofructokinase 1 (PFK1) is O-GlcNAcylated at Ser529 in response to hypoxia to redirect glucose flux into the pentose phosphate pathway, contributing to the biosynthesis of biomacromolecules in cancer cells [43]. Moreover, O-GlcNAcylation on co-activator-associated arginine methyltransferase 1 regulates its substrate specificity in addition to its stability, dimerization, and cellular localization [44].

Of its many qualities, the ability of O-GlcNAcylation to influence the binding of proteins to their ligands is particularly noteworthy. Signal transducer and activator of transcription (Stat) 5a belongs to a protein family composed of seven isoforms involved in mediating cellular responses to cytokines, hormones, and growth factors [45]. Stat5a was found to bear O-GlcNAcylation at Thr92 in HC11 mammary epithelial cells. However, substitution of Thr92 with an alanine ablates the glycosylation of Stat5a as well as its translational activity due to a loss of interaction with CREB-binding proteins [46]. Yin Yang 1 (YY1) is a zinc finger transcription factor that participates in the regulation of development [47]. Upon elevated O-GlcNAcylation induced by glucose, the formation of protein complexes by YY1 and hypophosphorylated retinoblastoma protein Rb was disrupted, and then YY1 was free to bind to DNA promoters [48].

Interplay between O-GlcNAcylation and other post-translational modifications

Interactions between O-GlcNAcylation and phosphorylation have been examined extensively for decades. The dynamic balance of O-GlcNAcylation and phosphorylation modulates the activity of various proteins in cells. Caenorhabditis elegans-bearing ogt-1 or oga-1 showed dramatically altered patterns of phosphorylation [49]. Elevation of O-GlcNAcylation in mammalian cells also affected the phosphorylation status of proteins. Proteomic studies identified about 400 proteins with both phosphorylation and O-GlcNAcylation including chaperone, cytoskeletal and regulatory proteins, metabolic enzymes, kinases, transcription factors, and RNA-processing proteins [50]. O-GlcNAcylation sites commonly occur at the same sites or near to those of phosphorylation. Because of the proposal that the O-GlcNAc moiety provides spatial resistance to the addition of phosphate, the theory was called “Yin-Yang” and related investigations on many proteins, including tau protein and CAMKII, have been performed with results that support this hypothesis. However, recent proteomic studies provided opposing evidence to overthrow this theory [35]. In summary, the apparent ability of O-GlcNAcylation to suppress phosphorylation may derive from a regulatory effect on kinase and phosphatase, rather than direct hindrance to the addition of phosphate.

Other post-translational modifications are also involved in the interplay with O-GlcNAcylation such as ubiquitination. For example, inhibition with glucosamine or O-(2-acetamido-2-deoxy-d-glucopyranosylidene) amino-N-phenylcarbamate (PUGNAc) promotes the global increase of ubiquitination [23]. In contrast, O-GlcNAcylation on β-catenin decreases its ubiquitination, which may be caused by direct competition. Moreover, O-GlcNAcylation of p53 at Ser149 could block phosphorylation at Thr155, leading to resistance against degradation induced by ubiquitination [51]. Regarding the complexity of crosstalk among these and other post-translational modifications, further research is needed to explore their relationship with O-GlcNAcylation.

O-GlcNAc cycling enzymes

Unlike the foregoing modifications, O-GlcNAcylation on myriad proteins is regulated by a unique pair of enzymes, OGT and OGA, which are responsible for the addition and removal of O-GlcNAc, respectively [52]. Both enzymes are highly conserved from C. elegans to humans, which proved to be of immense importance in understanding vertebrate physiology. OGT and OGA are distributed extensively in many cell types of humans and other higher eukaryotes. Compared with other organs, the pancreas and brain express greater amounts of OGT and OGA [53, 54], and both enzymes are sensitive to glucose levels in plasma.

O-linked N-acetylglucosaminyltransferase (OGT)

OGT is encoded by a single gene in Xq13 of the human genome, which is spliced and translated into three variable isoforms: nucleocytoplasmic OGT (ncOGT), mitochondrial OGT (mOGT), and short OGT (sOGT) [55] (Fig. 2a). At the N terminus of OGT is a tetratricopeptide repeat (TPR) domain differing in the three isoforms and proposed to interact with other partners, as well as to form self-trimerization. The catalytic domain of OGT localizes at the C-terminus and is linked with the TPR domain by a putative nuclear localization sequence (NLS) [53, 56]. Variable numbers of TPR repeats provide plasticity for OGT isoforms to modulate their substrate specificity [57, 58]. Interactions with partners or substrates such as OGA [59] and p38 MAPK [60] are associated with the regulation of OGT’s activity and localization [61, 62], affecting its recruitment toward target proteins such as RNA polymerase II, transcription complexes, and neurofilament H.

Fig. 2.

Structures of OGT and OGA isoforms. a Ogt allele is transcribed and spliced into three isoforms: nucleocytoplasmic OGT (ncOGT), mitochondrial OGT (mOGT) and short OGT (sOGT). These isoforms share similar structures, composed of tetratricopeptide (TPR) repeats, putative nuclear localization sequence (NLS) and catalytic domains and varying in the number of TPR repeats. b Two variants of OGA are nucleocytoplasmic OGA (ncOGA) and short-form OGA (sOGA). By comparing with sOGA, ncOGA contains a domain that is putatively responsible for histone acetyltransferase (HAT) activity

OGT is highly expressed in the brain, consequently leading to an augmented content of O-GlcNAcylated proteins. OGT mediating O-GlcNAcylation engages in an intense crosstalk with phosphorylation of neuronal proteins and plays a significant role in myriad cellular processes including transcriptional regulation, signaling, proteasomal degradation, and organelle trafficking. Aberrant O-GlcNAcylation-induced chaos in these pathways is linked to neuronal dysfunction. An early genetic research project established a putative link between an ogt allele and a rare neurological disorder, which was designated as a form of Parkinson-dystonia [63]. Another genome-wide associated study performed in peripheral blood mononuclear cells from patients with multiple sclerosis revealed an alteration in their expression of OGT compared with healthy volunteers, indicating that O-GlcNAc cycling may participate in the pathogenesis of multiple sclerosis [64]. As an X-linked gene, ogt allele expression in male and female victims may differ and yield a bias toward the appearance of a disease such as systemic lupus erythematosus [65]. OGT is an active player in the modulation of neuronal functions under physiological and pathological conditions.

β-N-Acetylglucosaminidase (OGA)

OGA is encoded by a single gene of meningioma-expressed antigen 5 (MGEA5) on chromosome 10 [66], which is spliced into two variants: nuclear and cytoplasmic isoform (ncOGA) and a short-form version (sOGA) (Fig. 2b). ncOGA is the long form and consists of a catalytic domain at the N terminus and a putative histone acetyltransferase (HAT) domain at the C terminus, as well as a link region between the two parts. In contrast, sOGA shares identical domains but lacks the HAT region. In the structure of OGA, caspase 3 recognizes the site between the catalytic domain and link region during apoptosis, whereas the enzymatic activity of OGA does not change after cleavage [67]. Like OGT, OGA is widely expressed in all tissues, highly conserved throughout evolution and abundant in the pancreas and brain [54, 68, 69]. The region of the mgea5 locus is reported to be associated with late-onset AD [70, 71].

As is well known from research on neurological disorders, OGA inhibitors are widely used to increase the O-GlcNAcylation in cultured cells and animal models. A series of studies uncovered a number of OGA inhibitors, such as PUGNAc, 1,2-dideoxy-2′-propyl-alpha-d-glucopyranoso-[2,1-d]-Delta 2′-thiazoline (NButGT) and Thiamet G. PUGNAc was first described as an inhibitor of another β-N-acetylhexosaminidase and later recognized as an OGA inhibitor [72, 73]. Although PUGNAc inhibits OGA, at low doses it also inhibits human lysosomal hexosaminidase B [73, 74]. After a period of exploration, the highly effective and selective inhibitor named NButGT was screened based on previous knowledge. NButGT showed a relatively potent 800-fold selectivity for OGA over the lysosomal β-hexosaminidases. Experiments in vivo and in vitro proved that NButGT treatment more effectively increased O-GlcNAcylation than a change in ganglioside levels [75–77]. Afterward, NButGT was modified to generate the compound Thiamet G, which is highly potent and selective for human OGA. Further, Thiamet G is extraordinarily stable when distributed in solution, capable of penetrating the blood–brain barrier and suitable for administration to cell cultures and intact animals [24, 78–83]. All these chemical agents offer the much needed opportunity to manipulate O-GlcNAcylation levels in cells and animal models of neurological diseases.

O-GlcNAcylation, OGT, and OGA in brain

O-GlcNAc is widely distributed in almost all types of cells and tissues in metazoan animals. The content of transcript-encoding OGA is enlarged in brain tissue, whereas that encoding OGT are highly expressed in the brain and pancreas [53, 54, 84]. With highly active OGT compared with peripheral organs, most proteins in the brain are modified with O-GlcNAc [85, 86]. In fact, several studies focused on the dynamics and topographic distribution of O-GlcNAcylation and O-GlcNAc cycling enzymes have linked their fluctuant levels to development of the rat brain [87, 88]. Monitoring of the O-GlcNAcylation levels in rat brains from embryos to 2-year-old indicated a decrease of O-GlcNAcylation level after birth and a subsequent stable level. The expression of OGT and OGA splicing isoforms was then compared during development of the rat brain, as well as the enzymatic activity of OGT and OGA. After birth, the rats gained greater activity of OGT even as the activity of OGA simultaneously changed in an opposite trend. Immunohistochemical staining of rat brains showed an extensive distribution of O-GlcNAcylation as well as OGT and OGA in every type of neuron across all the regions of these specimens [87]. Related research indicated an upregulation of O-GlcNAcylation in the cerebellar cortex and hippocampus, in which OGT was highly expressed [83, 85, 89, 90]. The ubiquitous distribution of O-GlcNAcylation in the cytoplasm and nuclei suggests a tight correlation between O-GlcNAcylation and neuronal functions.

Early studies showed that the neuron-specific mutation of OGT led to neuronal apoptosis [91]. Recent evidence now indicates that O-GlcNAcylation may be involved in neuronal apoptosis by modulating the activity of p53 or cyclin-dependent kinase 5 (CDK5) [89, 92]. In addition, upregulated O-GlcNAcylation of NF-κB was described as an arch-criminal for promoting apoptosis of retinal ganglion cells in a diabetic retinopathy model [93]. Other experimental data suggested that OGT was involved in the process of mitochondrial transport within neurons [94] and was deemed as a molecular mechanism in the regulation of excitatory synaptic function and synapse maturity [95, 96]. In the C. elegans model of tauopathy, amyloidosis and polyglutamine expansion, an alteration of O-GlcNAcylation by direct mutagenesis significantly modulated the severity of each phenotype [94]. Finally, since the O-GlcNAcylation level and the expression of O-GlcNAc cycling enzymes fluctuated during development of the CNS and became dysregulated as part of neurological disorders, O-GlcNAcylation was deemed to be responsible for the senile decline of brain functional recovery and could be a potential target for therapeutic development of neurological disorders [97, 98].

O-GlcNAcylation in neurological disorders

Alzheimer’s disease (AD)

AD, as the most prevalent of all neurodegenerative diseases, afflicts millions of patients with memory loss and an impaired recognition ability [86, 99]. At the histopathological level, extracellular senile plaques containing β-amyloid peptide (Aβ), intracellular neurofibrillary tangles of tau proteins, and neuronal loss are the major signatures of AD.

Increasing amounts of evidence has convinced researchers to follow the hypothesis that glucose uptake and utilization are impaired in AD, analogous to that in other metabolic diseases. During the pathogenesis and development of AD, glucose metabolism and supply of energy to the brain are both downregulated accordingly [100–104]. Many studies of AD models (transgenic mice and rats) have indicated that the amyloid-associated pathophysiologic process is closely linked to abnormal glucose uptake in the brain. In fact, Aβ toxicity is deemed the villain leading to glucose hypometabolism [105]. Similarly, in the human body, functionality study has shown that neurodegeneration may relate to glucose hypometabolism [106]. Like type 2 diabetes, AD may develop from an insulin deficiency and dysregulated signal transduction that impair cerebral glucose metabolism. Clinical data also support the possibility that insulin resistance results in an increased risk of dementia [107–112], indicating a similar potential in AD. In brains from patients with AD, insulin receptors are profuse [113], although insulin levels in cerebral spinal fluid (CSF) and/or the CSF/plasma insulin ratio are downregulated [114]. Additionally, the insulin signaling pathway in AD brains typically loses the ability to respond to insulin/insulin-like growth factor stimulation [115], and the expression level and activity of several key proteins such as Akt and GSK-3β lessen [112, 116]. Furthermore, the presence of GLUT1 and GLUT3 predominately expressed in neurons declines in brains of AD patients [108], possibly resulting from the decrease of GLUT3 localization to the plasma membrane caused by insulin deficiency. Collectively, emerging evidence emphasizes the importance of glucose hypometabolism and insulin signaling deficiency as essential factors in the etiology of AD, although the precise mechanism is yet to be understood.

To fill the information gap between impairment of glucose metabolism in the brain and the development of AD, O-GlcNAcylation is assigned the role of a promising molecular link that acts as a nutrient sensor with a close association with glucose metabolism [86, 117]. A decrease of O-GlcNAcylation level in AD brains is proposed to result from impaired glucose metabolism induced by Aβ toxicity and type 2 diabetes. The subsequent loss of O-GlcNAcylation protection leads to the progression of AD. Now, because of its well-known protective effect during the cellular stress response [18], more and more research devotees are inclined to support the likelihood that the elevation of O-GlcNAcylation is a promising therapeutic target for managing AD.

The two protein aggregates considered as historical hallmarks of AD are amyloid plaques and neurofibrillary tangles. Amyloid plaques are formed as the byproducts of extracellular amyloid precursor protein (APP) digestion by β- and γ-secretases, namely Aβ peptide, and intracellular neurofibrillary tangles are composed of hyperphosphorylated tau proteins (Fig. 3). The latter are much better indicators for the cognitive deficit of AD.

Fig. 3.

Molecular cascade of tau and Aβ pathology. Normal tau proteins bind to microtubules to stabilize the complex of α-tubulins and β-tubulins. When kinases such as GSK-3β, CDK5, and MAPK are activated under the conditions of diseases, tau is phosphorylated and detaches from the microtubule to form oligomers, and subsequently aggregates as paired helical filaments (PHFs) and further as neurofibrillary tangles (NFTs). Inhibition of OGA by Thiamet G (TMG) evokes the O-GlcNAcylation at Ser400 of tau, which may suppress the phosphorylation at Ser396 and S400 of tau within a short period of administration and directly prevent the aggregation of tau. On the other hand, transmembrane amyloid precursor protein (APP) is cleaved by β- and γ-secretase to produce soluble amyloid precursor protein β (sAPPβ) and Aβ, and release the APP intracellular domain (AICD). Aβ could self-organize to aggregate and form amyloid plaques. Inhibition of OGA may reduce the activity of γ-secretase to diminish the secretion of Aβ

O-GlcNAcylation and tau

Tau protein is a microtubule-associated protein (MAP) encoded by a single MAPT gene in humans, and that gene has six variants differing in the length of their N-terminal projection region and the number of microtubule-binding repeats in the C terminus. Tau proteins are highly soluble and intrinsically disordered, interacting with various partners as a hub in the cellular protein–protein interaction network [118]. Phosphorylation occurs on many different residues of tau proteins, and its dysregulation is the most prevalent driving factor associated with neurotoxicity. In this situation, the ratio of phosphorylated tau is increased about six- to eightfold above that at physiological conditions [119]. Hyperphosphorylation of tau leads to its detachment from the microtubule, subsequently increasing the tangle of microtubules as well as free tau proteins in the cytoplasm [120–122]. Accumulations of free tau proteins begin to form phosphorylated helical filaments, as reported from research performed in vitro and in vivo [123–125] (Fig. 3). Data from studies in transgenic mice in which specific kinases were inhibited have proven that hyperphosphorylation of tau is of critical significance in its toxicity [126, 127]. Intriguingly, soluble small oligomers of tau are found to be the toxic species, whereas the aggregations of tau as neurofibrillary tangles are non-toxic in the brain [128, 129]. Therefore, the massive aggregation of tau proteins is proposed as a pathologic cell stress response.

O-GlcNAcylation on tau proteins, when first discovered in samples of bovine tissues, indicated a stoichiometry of about 4 moles of O-GlcNAc per mole of tau proteins [130]. Subsequent studies then surveyed the precise residues of O-GlcNAcylation on tau proteins. Four sites on human tau proteins have been confirmed as Thr123, Ser208, Ser400, and Ser409/Ser412/Ser413 [81, 131]. O-GlcNAcylation at Ser400 is also found on tau proteins from mice and rats [81, 132, 133]. At an early stage of exploration on this topic, O-GlcNAcylation was thought to be a component in the regulation of tau activity and a mutual interaction with phosphorylation. However, several later efforts identified the reciprocity between phosphorylation and O-GlcNAcylation on tau proteins. Gong et al. described the O-GlcNAcylation of tau in humans and showed the antagonistic effect of O-GlcNAcylation against phosphorylation of tau at different residues in cultured cells and rat brain slices. Surprisingly, no O-GlcNAcylation was found on aggregated tau proteins in samples from AD patients [134]. Experimental data from fasting mice and a triple transgenic mouse model of AD indicated an increase of phosphorylation on tau [135, 136], and blocking of protein phosphatase led to a decrease of O-GlcNAcylation [137]. Tau phosphorylation was also increased in transgenic mice subjected to neuron-specific OGT deletion [91]. Additionally, Aβ-induced S-nitrosylated OGT led to the hyper-O-GlcNAcylation of tau protein in neurons [138].

Overall, these discoveries have stimulated efforts to explore the ability of elevations in O-GlcNAcylation to provide protection against tau toxicity in AD. Many kinds of tauopathy mouse models have been generated by genetic manipulation, including JNPL3 mice [139], Tau.P301L mice [140] and Tg4510 mice [128]. In JNPL3 mice, the administration of Thiamet G for several months suppressed the aggregation of tau, reduced the amount of neurofibrillary tangles in the brain and protected against neuronal loss [81]. A striking revelation was that an increase of O-GlcNAcylation at Ser400 directly inhibited tau aggregation in vitro rather than affecting the conformation of tau by suppression of its phosphorylation, an outcome that was recently confirmed [141]. Treatment of Tg451 mice with Thiamet G for 4 months resulted in a similar aftermath. Notably, a monoclonal antibody against Ser400 O-GlcNAcylation of tau was generated and applied for use in this study [82]. Thiamet G was also used for the Tau.P301L model. A 2.5-month-long treatment with Thiamet G decreased the distribution of neurons bearing neurofibrillary tangles, offset weight loss, improved motor defects, and lengthened survival rates as well [142]. The controversial item in this model was that tau proteins were not modified by O-GlcNAc, when applying a methodology that differs from that of previous studies.

To summarize briefly, increasing tau O-GlcNAcylation by the inhibition of OGA protects tau proteins against aggregation independent of phosphorylation, although the results from previous similar studies are still controversial. A putative proposal to explain the difference is that acute administration of an OGA inhibitor suppressed the activity of kinases and phosphatases, which would be responsible for phosphorylation and dephosphorylation of tau proteins. After a long term of perturbation, cells may adapt to these changes. On the other hand, sustained action of kinases could overcome the barrier of O-GlcNAc blockade to restore the phosphorylation level. Overall, these studies provide solid evidence that chronic treatment with OGA inhibitor ameliorates neuronal loss and motor defects by elevating O-GlcNAcylation at Ser400 and possibly other residues, shedding light on the development of a therapeutic strategy for AD [86].

O-GlcNAcylation and APP

Amyloid plaques are typical lesions in the brain of a patient with AD where Aβ peptides have aggregated [143]. Aβ peptides are produced by the cleavage of APP by β- and γ-secretase, whereas soluble APP is generated by cleavage by α- and γ-secretase. It is well accepted that dysregulation of Aβ processing and generation of Aβ peptides are the earliest phenomena to initiate the pathogenesis of AD, apparently with the aid of other factors [144, 145].

As noted previously, Aβ toxicity may lead to impaired glucose metabolism in the brain, but type 2 diabetes could be the accelerant in AD. APP was reported to be O-GlcNAcylated decades ago, although this information was not brought to the forefront at that time [146, 147]. Much later, application of the OGA inhibitor, PUGNAc, increased O-GlcNAcylation and visibly reduced the production of Aβ [148, 149]. Recent study using another OGA inhibitor, NButGT [150], in the 5xFAD Aβ mouse model also supported the protective effect of long-term treatment with OGA inhibitor that inhibited Aβ toxicity. As a result, the levels of Aβ40 and Aβ42 decreased as did plaque formation and neuroinflammation followed by improved cognition. In vitro, experiments with CHO cells expressing the Swedish mutation in APP also showed a decrease of that protein’s C-terminal fragment. Presumably, these effects result mainly from the suppression of γ-secretase by O-GlcNAcylation [150]. However, the precise sites of O-GlcNAcylation on γ-secretase have not yet been mapped, and how O-GlcNAcylation functions in the regulation of γ-secretase remains to be explored (Fig. 3). In bigenetic TAPP mice that combine the pathologic effects of amyloid and tau, the outcome of long-term OGA inhibition induced by Thiamet G was also described. The prevention of cognitive defects resulted from OGA inhibition, along with decreases of Aβ levels and plaque numbers in the brain, as well as the stabilization of tau [151]. Interestingly, Ryu et al. reported that O-GlcNAcylation could be elevated by Aβ-induced S-nitrosylation of OGT, whereas hyper-O-GlcNAcylation within cells was protective against Aβ neurotoxicity [138]. Although these studies provide convincing evidence of the protection that OGA inhibition provides against Aβ toxicity, exactly how the long-term treatment of OGA inhibitors affects the processing of APP remains open to further validation.

Additionally, impairments were noted in several other cellular signaling pathways of AD, in which key proteins were modified with O-GlcNAc. Protein kinase A catalytic subunits (PKAcs) were reported to be O-GlcNAcylated at both α and β subunits, which positively regulate their enzymatic activity. A decrease of O-GlcNAcylation during AD downregulates the activity of PKAcs, leading to the suppression of PKA-CREB signaling, which causes deficiencies in learning and memory [152].

Parkinson’s disease (PD)

The common neurodegenerative disorder, PD, occurs in approximately 1% of persons over 65 years old with no bias as to race or geographical location. The characteristic clinical syndrome includes muscle rigidity, bradykinesia, and resting tremor, accompanied by the secondary syndromes of cognitive dysfunction and postural instability [153, 154]. The pathophysiology of PD is attributed to the loss of dopaminergic neurons leading to a disability of motor function-related nerves in the central nervous system, along with the accumulation of Lewy bodies in surviving neurons [155, 156].

α-Synuclein is thought to be of critical importance in the etiology of PD, although the precise underlying basis of this disorder remains ambiguous. Genetic research has implicated an association between α-synuclein and familial PD [157, 158]. Further, several mutations in α-synuclein, such as A30P, A53T, and E46K, were reported as causes of inherited PD [157, 159–162]. During the pathogenesis of PD, α-synuclein undergoes a conformational change and forms dimer and oligomers, subsequently aggregated as fibrils that are toxic to cells and model organisms [163–165]. A series of post-translational modifications such as phosphorylation [166–171], ubiquitination [172, 173], and glycosylation [34, 132, 174] are proposed to be involved in this process. Shimura et al. described a protein complex composed of parkin and UbcH7, as well as an O-glcosylated α-synuclein, which implies the existence of a linkage between the O-glcosylated form of α-synuclein and parkin [174]. However, an opposing result was also cited indicating that parkin interacts with the nonglycosylated form of α-synuclein in the embryonic hippocampal cell line, H19-7. Initially, when specific antibodies against O-GlcNAc were applied to identify the immunoprecipitated proteins, glycosylation on α-synuclein was not confirmed [175]. However, several years later, the O-GlcNAcylation site on α-synuclein was established as Thr72 in humans and located in a region that is resistant to protein degradation [132]. On its homologous protein in murine neurons, O-GlcNAcylation sites were mapped to Thr53, Thr64, and Thr72, but to Ser87 in human erythrocytes [10, 34]. Marotta et al. first used a chemically synthesized peptide containing O-GlcNAcylation to validate the proposal that O-GlcNAcylation at Thr72 could inhibit α-synuclein aggregations. They found the full length of α-synuclein with or without O-GlcNAcylation and experimental data confirmed the inhibitory effect of O-GlcNAcylation at Thr72 on fibril formation, as well as the reduction of α-synuclein’s toxicity for living cells [176]. Recently, Matthew’s group identified a less inhibitory effect of O-GlcNAcylation at Ser87 of α-synuclein against protein aggregation using a strategy of synthetic protein chemistry [177] (Fig. 4). Although a vivid picture of PD’s etiology is yet to come, blocking of α-synuclein aggregation by O-GlcNAcylation could be a potential target for therapeutic control of PD. Inhibitors of OGA might serve as the agents, since they have a proven capacity to raise the O-GlcNAcylation level in the brain [82, 151].

Fig. 4.

Schematic diagram of α-synuclein aggregation. Genetic mutations of α-synuclein are associated with the pathologic effects of PD. Misfolded α-synuclein is susceptible to forming dimers and oligomers, as well as aggregating fibrils, leading to the formation of Lewy bodies, which are toxic for neurons. Strikingly, a single O-GlcNAcylation at Thr72 or Ser 87 of α-synuclein inhibits the aggregation and reduces the toxicity of aggregated α-synuclein for adjacent cells

Several other clues hint that the link between O-GlcNAc and PD is factual. A recent report asserted that tyrosine 3-monooxygenase is modified by O-GlcNAc, which is responsible for dopamine secretion. Stimulation of N-propanoylmannosamine (an unnatural precursor of neuraminic acid) downregulated the O-GlcNAcylation of tyrosine 3-monooxygenase and promoted its activation, leading to an increase of dopamine secretion [178]. Since this outcome might be controversial with respect to previous reports, further validation is warranted.

Stress and stroke

O-GlcNAcylation is believed to be an important mitigating factor of cells’ stress response. An early report showed the prolonged survival times of rats undergoing experimentally induced hemorrhages after suppression of stress-induced hyperglycemia [179]. At first, this protection was proposed to result from either the physical effect of maintaining homeostasis [180, 181] or the provision of an adequate energy supply [182]. Glucose flux through HBP was highlighted afterward because the production of UDP-GlcNAc was the end product. Other experimental data suggested that an elevated O-GlcNAcylation level induced the increased survival of mammalian cells ex vivo after exposure to stress [18]. However, O-GlcNAc modification has been closely associated with vascular ischemic/reperfusion injury [183]. In an animal model of trauma hemorrhage, intraperitoneal injection of glucosamine induced an increase of O-GlcNAcylation level, improved blood flow in critical organs, and decreased the production of inflammatory cytokines [184]. Subsequently, PUGNAc, an inhibitor of OGA, was used as an alternative agent to upregulate the O-GlcNAcylation level in a rat model of trauma hemorrhage and similar results were obtained [185]. In other studies, an increase of O-GlcNAc levels again improved cardiac function and perfusion of critical organ systems as well as reducing the amount of inflammatory cytokines in plasma [186–189]. Treatment with glucosamine (GlcN) and PUGNAc during resuscitation dramatically decreased the trauma-hemorrhage-induced increase of TNF-α, IL-6, NF-κB, IκB-α phosphorylation and NF-κB DNA binding activity in cardiac tissues [190]. Outcomes from all these studies support the concept that protection is offered by O-GlcNAcylation under conditions of physiologic stress.

Elsewhere, a dramatic increase of O-GlcNAc modification in the cortex was recorded after stroke [191, 192]. Supposedly, an elevation of O-GlcNAc modification was associated with neuronal apoptosis in the early stage of stroke and the extent of recovery after ischemic injury. Moreover, Hwang et al. provided evidence that treatment with GlcN could inhibit the production of inflammatory cytokines and activation of microglia in a rat model of stroke [193]. As an alternative way to increase O-GlcNAcylation, our group pursued the application of a highly selective, blood–brain barrier-permeable inhibitor of OGA, Thiamet G, in mice manipulated to model ischemic stroke [194]. We then confirmed the neuroprotective effect of Thiamet G, a novel inflammation antagonist, in mice imbued with transient middle cerebral artery occlusion. Experimental data showed that Thiamet G significantly reduced infarct size and diminished the neurological deficits and lessened motor coordination impairment in subjects of experimental stroke. The results achieved in vivo and confirmed in vitro indicate that the neuroprotective effect of Thiamet G is associated predominantly with a shift of microglia polarization in an NF-κB-dependent manner (Fig. 5).

Fig. 5.

Proposed molecular events after ischemia and reperfusion. After stimulating ischemia and performing reperfusion, the O-GlcNAcylation level rises dramatically in a short period of time. O-GlcNAcylation at Thr308 and Ser473 of Akt suppresses its phosphorylation and activation. Decreased Akt activity results in the loss of phosphorylation of Bad at Ser136, leading to its increased binding with the Bcl-2 family of proteins. This action downregulates the formation of protein complexes composed of Bcl-2 proteins and BAX. Subsequently, more BAXs attack the mitochondrial membrane to activate caspase-3, which causes apoptosis of neurons. Inhibition of OGA by Thiamet G (TMG) accelerates this process. On the contrary, the NF-κB pathway is activated, and the complex of p65 and p50 translocates into nuclei to initiate transcription of target genes including pro-inflammatory cytokines, iNOS and COX-2. Increase of O-GlcNAcylation on p65 by pharmacological intervention with glucosamine (GlcN) and TMG inhibits transcription initiation and suppresses the classical activation of macrophage/microglia, resulting in their shift of phenotype into M2. BAX, BCL2-associated X, apoptosis regulator; I/R, ischemia/reperfusion; OGT, O-linked N-acetylglucosaminyltransferase; OGA, β-N-acetylglucosaminidase; IKK, inhibitor of nuclear factor κB kinase; IκB, inhibitor of nuclear factor κB; “G” indicates the moiety of GlcNAc; TMG, Thiamet G

After ischemic stroke, apoptosis is trigged by ischemia-induced stress, leading to neuronal cell death and preliminary tissue damage. At the initial stage of ischemic stroke, apoptosis appears around the penumbra. Afterwards, the ischemic area expands as ischemic time lengthens, and numerous cells in the central area of ischemia die of necrosis [195, 196]. In a mouse model of ischemic stroke, O-GlcNAcylation was found to promote apoptosis by attenuating phosphorylation/activation of AKT and the BCL2-associated agonist of cell death (Bad). In the process, both Thr308 and Ser473 of AKT were modified with O-GlcNAc as identified using co-immunoprecipitation and mutagenesis techniques. However, O-GlcNAcylation-induced apoptosis was attenuated by the overexpression of AKT in cultured cells. A negative correlation between AKT phosphorylation and O-GlcNAcylation in brain tissues of ischemic mice became evident by immunoblotting. These results indicate that cerebral ischemia induces a rapid increase of O-GlcNAcylation that promotes apoptosis through down-regulation of AKT activity and provide a novel mechanism with which O-GlcNAcylation regulates ischemia-induced neuronal apoptosis through AKT signaling [191] (Fig. 5). Similarly, a preliminary study performed in a model of intracerebral hemorrhage (ICH) in vitro indicated that O-GlcNAcylation of CDK5 protected against neuronal apoptosis by suppressing the p53 pathway [92]. An observation showed that O-GlcNAcylation of neuronal nitric oxide synthase (nNOS) slowly increased at 3 h after ischemic stroke in a rat model of middle cerebral artery occlusion. This result indicated a possible association of O-GlcNAcylation with excitotoxicity induced by brain injury. Subsequent experiments in vitro hinted that O-GlcNAcylation of nNOS leads to neuronal death by modulating the interaction between OGT and nNOS [197]. This work provides new evidence on the role of O-GlcNAcylation in the pathology of ischemic stroke and contributes to the development of a useful therapeutic strategy.

Interestingly, in a parallel study focused on the change of protein post-translational modifications including O-GlcNAcylation, ubiquitination and SUMO conjunction in young and aged mice subjected to surgery to induce middle cerebral artery occlusion, results suggested that impaired O-GlcNAcylation in brains of aged mice may be responsible for slow recovery compared with that of young mice [192]. This indicated that protein O-GlcNAcylation is associated with functional recovery after ischemic stroke and is proposed as an age-related post-translational modification. Further investigation may provide more information about the perturbation of O-GlcNAc cycling during the onset and progression of stroke.

Other neurological disorders

In addition to the diseases mentioned above, the impact of protein O-GlcNAcylation has been extended. Wang and co-workers found that a decrease of O-GlcNAc cycling reduced neuronal loss in a model of Huntington’s disease (HD) in transgenic mice expressing a fusion human Huntington protein with an expansion of polyglutamines and vice versa [94]. Recently, Jonathan et al. confirmed that mutant Huntingtin gene-induced disruption of nucleocytoplasmic transport was responsible for HD-like pathology in multiple models of HD, and an increase of O-GlcNAcylation by Thiamet G rescued nucleocytoplasmic trafficking defects and alleviated neurotoxicity in HD [198]. These data suggest that hyper-O-GlcNAcylation is neuroprotective, which is consistent with foregoing research. Further investigations are still requisite to elucidate the role of O-GlcNAcylation in the etiology and pathogenesis of HD. In a genome-wide association analysis of a Drosophila melanogaster model of trauma brain injury, results also indicated that O-GlcNAc cycling enzymes and HBP are associated with glucose flux after brain injury [199].

Amyotrophic lateral sclerosis (ALS), another typical neurodegenerative disease with adult onset, is characterized by a loss of neurons, particularly motor neurons in the central nervous system. ALS of humans and mice is closely linked to dysregulation of protein kinase function and an increase of phosphorylated products [200, 201]. Under the condition of ALS, an essential protein, neurofilament, is reputedly hyperphosphorylated, a status that is also modified by O-GlcNAc [202, 203]. As is usual on other proteins, such as tau, O-GlcNAcylation and phosphorylation of neurofilaments are mutually reciprocal [204]. In a rat model of ALS, a decrease of O-GlcNAcylation in the spinal cord proved to be associated with the pathogenesis of ALS and acted against phosphorylation on neurofilament-M, whereas the inhibition of OGA attenuated the tangle formation intrinsic to this disease [22]. Shan and co-workers used NButGT to identify a similar result in mice carrying a mutation in superoxide dismutase 1 (SOD1) [77]. These experiments suggest that upregulation of O-GlcNAcylation by intervention with OGA inhibitors is a promising therapeutic method for ALS.

Multiple sclerosis, a chronic autoimmune and inflammatory disease of the central nervous system, is characterized by demyelinating lesions and progressive axon loss [205]. Recently, the remarkable impact of O-GlcNAc modification on immunity and inflammation was identified [206]. Several studies have provided preliminary clues that O-GlcNAc modification is a participant in the pathogenesis of autoimmune diseases [65, 207, 208]. However, no further revelations have yielded the precise role of O-GlcNAc modification in CNS inflammation. Yet we identified microRNA-15b as a key regulator of T helper 17 cell-associated effects in a mouse model of multiple sclerosis; those effects modulated the transcriptional regulation of retinoic acid-related orphan receptor γT through O-linked N-acetylglucosamine glycosylation of NF-κB by targeting OGT [209]. These data indicate that O-GlcNAcylation supposedly contributes to the etiology and pathogenesis of multiple sclerosis.

Conclusions

In this review, we highlight the physiological and pathological functions of protein O-GlcNAcylation in the nervous system. Compared with the long-established effects of phosphorylation, O-GlcNAcylation is an emerging actor in the theater of neurological diseases. The “playbook” of these performances and related studies is just beginning to open. As an indispensable link between metabolism and neuronal activity, protein O-GlcNAcylation is a putatively essential instrument in the development of neurological diseases. A growing body of evidence from research studies performed in vivo and in vitro show that O-GlcNAcylation fluctuates and changes, and O-GlcNAc cycling enzymes depend as well on the cellular context and stress conditions in neurological disorders that underlie the capability for learning and memory. Although researchers have mapped the critical O-GlcNAcylation site on tau protein to Ser400 and deciphered its biological effect during the pathogenesis of AD, most experimental data provide a description of changes in O-GlcNAcylation status rather than in the molecular mechanisms beneath the surface. Our challenge is to figure out solutions to these problems in view of the multiple factors involved in the development of neurological disease.

In the near future, emerging research studies in this area will provide more precise information about the regulation of O-GlcNAcylation and specificity of O-GlcNAcylation. We await understanding of O-GlcNAcylation in the nervous system and neurological disorders, so as to build a complete network of metabolic and neuronal activity that restores health in the aftermath of neurological damage. Moreover, pharmacological intervention with inhibitors of O-GlcNAc cycling enzymes in the experimental setting may shed light on the “sweet spot” of developing therapeutic methods for alleviating neurological diseases.

Abbreviations

Aβ

β-Amyloid peptide

AD

Alzheimer’s disease

ALS

Amyotrophic lateral sclerosis

AICD

Amyloid precursor protein intracellular domain

APP

Amyloid precursor protein

Bad

BCL2-associated agonist of cell death

BAX

BCL2 associated X

BEMAD

Michael addition with dithiothreitol

CDK5

Cyclin-dependent kinase 5

CSF

Cerebral spinal fluid

eNOS

Endothelial nitric oxide synthase

ETD

Electron-transfer dissociation

GalT

β1-4-Galactosyltransferase

GFAT

l-Glutamine:fructose-6-phosphate amidotransferase

GlcN

Glucosamine

HAT

Histone acetyltransferase

HBP

Hexosamine biosynthetic pathway

HD

Huntington’s disease

ICH

Intracerebral hemorrhage

LC

Liquid chromatography

IKK

Inhibitor of nuclear factor κB kinase

IκB

Inhibitor of nuclear factor κB

I/R

Ischemia/reperfusion

MAP

Microtubule-associated protein

MGEA5

Meningioma expressed antigen 5

mOGT

Mitochondrial OGT

MS

Mass spectrometry

NButGT

1,2-Dideoxy-2′-propyl-alpha-d-glucopyranoso-[2,1-d]-Delta 2′-thiazoline

ncOGA

Nuclear and cytoplasmic OGA

ncOGT

Nucleocytoplasmic OGT

NFTs

Neurofibrillary tangles

NLS

Nuclear localization signal

NO

Nitric oxide

OGA

β-N-Acetylglucosaminidase

nNOS

Neuronal nitric oxide synthase

O-GlcNAcylation

O-linked β-N-acetylglucosaminylation

OGT

O-linked N-acetylglucosaminyltransferase

PD

Parkinson’s disease

PKAcs

Protein kinase A catalytic subunit

PFK1

Phosphofructokinase 1

PHFs

Paired helical filaments

PUGNAc

O-(2-Acetamido-2-deoxy-d-glucopyranosylidene) amino-N-phenylcarbamate

RPLC

Reversed-phase liquid chromatography

sAPPβ

Soluble amyloid precursor protein β

SOD1

Superoxide dismutase 1

sOGA

Short-form OGA

sOGT

Short OGT

Stat

Signal transducer and activator of transcription

TPR

Tetratricopeptide

YY1

Yin Yang 1