Nutrient-driven O-GlcNAc in proteostasis and neurodegeneration

Abstract

Proteostasis is essential in the mammalian brain where post-mitotic cells must function for decades to maintain synaptic contacts and memory. The brain is dependent on glucose and other metabolites for proper function and is spared from metabolic deficits even during starvation. In this review, we outline how the nutrient sensitive nucleocytoplasmic posttranslational modification O-linked N-acetylglucosamine (O-GlcNAc) regulates protein homeostasis. The O-GlcNAc modification is highly abundant in the mammalian brain, and has been linked to proteopathies, including neurodegenerative diseases such as Alzheimer’s, Parkinson’s, and Huntington’s. C. elegans, Drosophila, and mouse models harboring O-GlcNAc transferase and O-GlcNAcase knockout (KO) alleles have helped define the role O-GlcNAc plays in development as well as age-associated neurodegenerative disease. These enzymes add and remove the single monosaccharide from protein serine and threonine residues, respectively. Blocking O-GlcNAc cycling is detrimental to mammalian brain development and interferes with neurogenesis, neural migration, and proteostasis. Findings in C. elegans and Drosophila model systems indicate that the dynamic turnover of O-GlcNAc is critical for maintaining levels of key transcriptional regulators responsible for neurodevelopment cell fate decisions. In addition, pathways of autophagy and proteasomal degradation depend on a transcriptional network that is also reliant on O-GlcNAc cycling. Like the quality control system in the endoplasmic reticulum which uses a “mannose-timer” to monitor protein folding, we propose that cytoplasmic proteostasis relies on an “O-GlcNAc timer” to help regulate the lifetime and fate of nuclear and cytoplasmic proteins. O-GlcNAc-dependent developmental alterations impact metabolism and growth of the developing mouse embryo and persist into adulthood. Brain-selective KO mouse models will be an important tool for understanding the role of O-GlcNAc in the physiology of the brain and its susceptibility to neurodegenerative injury.

Graphical abstract

Proteostasis is essential in the mammalian brain where post-mitotic cells must function for decades to maintain synaptic contacts and memory. The brain is dependent on glucose and other metabolites for proper function and is spared from metabolic deficits even during starvation. In this review, we outline how the nutrient sensitive nucleocytoplasmic posttranslational modification O-linked N-acetylglucosamine (O-GlcNAc) regulates protein homeostasis. This cyclic modification is coordinately regulated with other PTMs such as phosphorylation to regulate the required intricacies of cellular processes. Deregulation of PTMs including O-GlcNAc leads to several pathologies that are associated with neurodegeneration.

Introduction

George Cahill, in his groundbreaking article “Starvation in Man” (Cahill, 1970) emphasized that the brain is the last organ to succumb to starvation owing to its strong dependence on glucose as a source of energy. The mammalian brain has evolved such that it has an unusually high requirement for glucose and does not easily tolerate its absence. In humans, the brain accounts for only about 2% of the total body weight, yet it consumes about 20% of glucose-derived energy making it the main consumer of glucose (Erbsloh et al., 1958). In addition, specialized centers in the brain sense central and peripheral glucose levels and regulate glucose metabolism through the vagal nerve as well as neuroendocrine signals. Glucose supply to the brain is under tight regulation by neurovascular coupling and enters the brain from the blood by crossing the blood-brain-barrier through glucose transporters (e.g., GLUT1). Glucose and other metabolites such as lactate are distributed through a highly coupled metabolic network of brain cells. The brain’s dependence on glucose metabolism to maintain neurotransmitter release, axonal transport, and cell survival has important implications for understanding human disease. In diabetics, dangerously high or low glucose can lead to diabetic coma. The high flux of glucose required for maintenance of the central nervous system may also be linked to the formation of reactive oxygen species causing oxidative damage in the brain. Since most neurons are post-mitotic, this is thought to impact the brain disproportionately. Moreover, there is a growing awareness that adult neurogenesis may be more important than once recognized, particularly in the dentate gyrus, hippocampus, and olfactory bulb. Since neurogenesis requires significant amount of energy from glucose, abnormal levels of glucose supply could perturb healthy neurogenesis in these regions.

Neurodegenerative diseases, including Alzheimer’s Disease (AD) have been directly linked with altered brain glucose consumption. It has been estimated that 5.4 million people in the US currently have AD. The risk of AD increases with age, and so by 2050, the Alzheimer’s Association estimates that between 11 and 16 million Americans will have the disease, with one new case appearing every 33 seconds (Association, 2017). The economic cost of AD is already immense and disproportionally high in countries with longer life expectancies. While some estimates suggest that the total worldwide economic cost of dementias represents about 1% of world gross domestic product, the human cost of these diseases is much greater. In this review, we evaluate the accumulating body of evidence suggesting that a signaling pathway that integrates the flux of nutrients including glucose, glutamine, and acetyl-CoA, and terminates in O-linked N-acetylglucosamine (O-GlcNAc) addition to proteins, may contribute to the pathophysiology of neurodegenerative diseases. With a central role in the regulation of gene expression, cellular signaling, protein synthesis, and degradation, the nutrient-responsive O-GlcNAc pathway may represent a unique therapeutic option for diagnosis and management of these disorders. But what makes the brain uniquely sensitive to deregulation of O-GlcNAc metabolism?

A. Protein aggregates in neurodegenerative disease

Human neurodegenerative diseases are associated with deposits of aggregated proteins. These aggregates and their designations are summarized in Figure 1 (right side). As Figure 1 suggests, the proteostasis networks that maintain protein folding involve every aspect of the life cycle of proteins involved. This includes transcription, translation, folding, posttranslational modification, targeting and degradation. It has been noted that many of the proteins that are involved in aggregation in neurodegenerative diseases have been found to be O-GlcNAc modified. The known consequences of these proteins’ modification are discussed below in sections relevant to the individual OGT substrates. However, the question of whether these protein aggregates are pathogenic or simply represent a cellular response to stress is currently unknown. Parkinson’s disease (PD), Alzheimer’s disease (AD), amyotrophic lateral sclerosis (ALS), Huntington’s disease (HD), and other polyglutamine diseases each have characteristic deposits of protein aggregates in the brain (Ross and Poirier, 2005). These deposits can be cytoplasmic, nuclear, or even extracellular. In some cases, this protein aggregation results from a mutation in a disease-related protein, processing of those proteins, or elevation in their levels. Even proteins that are not associated with disease can aggregate in inclusion bodies and cause toxicity.

Figure 1. The potential roles of O-GlcNAcylation in maintaining the cellular proteostasis network and relationship to protein aggregates in neurodegeneration.

O-GlcNAcylation has been shown to be a key regulator of transcription, translation, protein folding and degradation that may be regulated in in neurodegenerative disease. The protein aggregates that are associated with human disease are listed at the right. O-GlcNAcylation can influence aggregation by modulating processes including transcription, protein folding, the cellular quality control, and stress pathways. Acting in consort with other post-translational modifications, O-GlcNAc may influence signaling pathways and the regulation of protein degradation including proteasomal degradation, autophagy, and the cellular stress response.

Neuritic plaques in AD contain mainly Aβ peptide, a proteolytic product of the amyloid precursor protein (APP). AD is also characterized by the accumulation of intracellular aggregates of the microtubule-associated protein tau, termed neurofibrillary tangles (NFTs). Tauopathies such as fronto-temporal dementia with Parkinsonism can be caused by mutations in the tau protein itself. PD, which is characterized by tremor, loss of motor function, and autonomic instability, is caused by degeneration of dopaminergic neurons in the substantia nigra of the midbrain. Point mutations or increased gene expression of the α-synuclein gene cause an autosomal dominant form of PD (Levine et al., 2017), whereas recessive PD can be caused by mutations in the genes encoding several proteins including parkin, DJ-1, or PINK1. Mutations in parkin and PINK1, key regulators of neuronal mitophagy, are associated with cell death correlated with immobile mitochondria unable to be sequestered (Wang et al., 2011). These recessive forms of PD are presumably induced by a loss-of-function of the associated proteins. The aggregates that form in PD are Lewy bodies, cytoplasmic perinuclear inclusion bodies in neurons, in which the α-synuclein protein is a major constituent (Ross and Poirier, 2004; Ross and Poirier, 2005). ALS is characterized by degeneration of motor neurons leading to progressive motor weakness with ubiquitinated protein aggregates present in patient brains. HD is caused by genomic expansion of a triplet CAG repeat coding for polyglutamine near the N-terminus of the huntingtin protein. Protein inclusions containing huntingtin and other proteins are present in regions of the brain that degenerates and there is a good correlation between the extent of triplet expansion and inclusion density. Other polyglutamine diseases such as spino-cerebellar ataxia (SCA) present with similar aggregates containing expanded polyglutamine repeats in the protein Atx-1. The Prion diseases are also associated with amyloid plaques which form both intra- and extra-cellularly by self-propagation of an abnormal protein conformation (Ross and Poirier, 2004; Ross and Poirier, 2005).

Domains common to neurodegenerative protein aggregates

Many of the proteins that are prone to aggregation in neurodegenerative diseases have domains that can be characterized as intrinsically disordered proteins (IDPs). These intrinsically disordered segments are also often the sites of posttranslational modification. Such domains are dynamic structures that interconvert between collapsed and extended structures on a timescale that differs from globular proteins. IDPs are relatively depleted in bulky hydrophobic amino acids and enriched in polar residues and structure-breaking amino acids like Pro and Gly. Aβ is about 20% disordered, tau is greater than 80% disordered, α-synuclein is greater than 90% disordered, and the ataxins range from 50–90% disordered (Uversky, 2015). Not surprisingly, many of the chaperones and other proteins involved in maintaining proteostasis are also extensively disordered (Uversky, 2015). Some of these proteins have an intrinsic propensity to adopt pathological conformations and persistently high concentrations, interaction with chaperones, or point mutations can initiate the aggregation cascade. Disruption of protein folding or protein degradation can contribute to aggregation and posttranslational modifications such as advanced glycation products, deamidation, and phosphorylation may influence the aggregation process (Ross and Poirier, 2005) (Figure 1). O-GlcNAcylation, like glycation, may be elevated under conditions of hyperglycemia and other forms of stress, although this is not always the case (Yang and Qian, 2017). The present review focuses on the many influences of O-GlcNAcylation on neurodegeneration through its role on pathways including protein folding, phosphorylation, and cellular protein degradation pathways. O-GlcNAc transferase, the enzyme responsible for the addition of O-GlcNAc to target proteins, has a propensity to interact with IDPs including many of those proteins that form aggregates in neurodegenerative disease (Trinidad et al., 2012; Bond and Hanover, 2015). In the following sections, we summarize the accumulating evidence for the involvement of O-GlcNAc metabolism in neurodegenerative toxicity. A timeline detailing the sequence of findings leading to our current understanding of the importance of O-GlcNAc cycling in neurodegeneration is shown in Figure 2.

Figure 2. A timeline summarizing the major advancements in understanding O-GlcNAc as a modulator of neurodegeneration.

The timeline was derived from publication dates of these references and references therein: (Torres and Hart, 1984; Holt and Hart, 1986; Davis and Blobel, 1987; Hanover et al., 1987; Holt et al., 1987; Park et al., 1987; Snow et al., 1987; D’Onofrio et al., 1988; Caillet-Boudin et al., 1989; Datta et al., 1989; Inaba and Maede, 1989; Haltiwanger et al., 1990; Starr and Hanover, 1990; Starr et al., 1990; Kearse and Hart, 1991a; Kearse and Hart, 1991b; Lüthi et al., 1991; Chou et al., 1992; Hagmann et al., 1992; Haltiwanger et al., 1992; Reason et al., 1992; Roquemore et al., 1992; Dong et al., 1993; Elliot et al., 1993; Kelly et al., 1993; Murphy et al., 1994; Bailer et al., 1995; Chou et al., 1995; Griffith and Schmitz, 1995; Griffith et al., 1995; Dong et al., 1996; Roquemore et al., 1996; Kreppel et al., 1997; Lubas et al., 1997; Roos et al., 1997; Haltiwanger et al., 1998; Yao and Coleman, 1998a; Yao and Coleman, 1998b; Akimoto et al., 1999; Griffith and Schmitz, 1999; Hanover et al., 1999; Akimoto et al., 2000; Lubas and Hanover, 2000; Ross et al., 2000; Cole and Hart, 2001; Gao et al., 2001; Hanover, 2001; Rex-Mathes et al., 2001; Zachara et al., 2001; McClain et al., 2002; Vosseller et al., 2002; Wells et al., 2002; Zachara et al., 2002; Akimoto et al., 2003; Gao et al., 2003; Hanover et al., 2003; Iyer et al., 2003; Lefebvre et al., 2003a; Lefebvre et al., 2003b; Love et al., 2003; Marshall et al., 2003; Whelan and Hart, 2003; Zhang et al., 2003; Cieniewski-Bernard et al., 2004; Jínek et al., 2004; Khidekel et al., 2004; Liu et al., 2004a; Robertson et al., 2004; Brickley et al., 2005; Gross et al., 2005; Hanover et al., 2005; Lefebvre et al., 2005; Dorfmueller et al., 2006; Forsythe et al., 2006; März et al., 2006; Nandi et al., 2006; Andrali et al., 2007; Deng et al., 2008; Rexach et al., 2008; Bleckmann et al., 2009; Gambetta et al., 2009a; Lazarus et al., 2009; Liu et al., 2009; Sinclair et al., 2009; Yanagisawa and Yu, 2009; Yuzwa and Vocadlo, 2009; Di Domenico et al., 2010; Rexach et al., 2010; Srikanth et al., 2010; Kim, 2011; Lazarus et al., 2011; Smet-Nocca et al., 2011; Yuzwa et al., 2011; Nakamura et al., 2012; Wang et al., 2012; Yu et al., 2012; Yuzwa et al., 2012; Cameron et al., 2013; Diwu et al., 2013a; Diwu et al., 2013b; Hanover and Wang, 2013; Wang and Hanover, 2013; Graham et al., 2014; Skorobogatko et al., 2014; Taylor et al., 2014; Cha et al., 2015; Mao et al., 2015; Marotta et al., 2015; Akan et al., 2016; Gatta et al., 2016; Gong et al., 2016; Lagerlöf et al., 2016; Lagerlöf et al., 2017; Lewis et al., 2017; Olivier-Van Stichelen et al., 2017)

B. The enzymes of O-GlcNAc cycling and roles in neurophysiology

1. Hexosamine signaling pathway in the brain

Besides its function as an energy source, glucose is also an important precursor for the synthesis of downstream metabolites. Through metabolic pathways including glycolysis and the hexosamine biosynthetic pathway (HBP), the brain takes advantage of glucose as its primary energy source and converts this simple sugar to a myriad of critical metabolites. The HBP is an essential player in brain physiology with 2 to 3% of glucose (Simpson et al., 2008) converted to the pathway’s ultimate product, UDP-GlcNAc (Bouché et al., 2004) . The synthesis of UDP-GlcNAc integrates lipid, energy, and nucleotide metabolism making the nucleotide sugar sensitive to cellular metabolite flux (Hanover et al., 2012).

The concentration of nutrient-sensitive UDP-GlcNAc impacts many glycosylation processes, with those that are cyclic arguably the most affected by changes in metabolic flux. Glycosyltransferases utilize nucleotide sugars such as UDP-GlcNAc as substrates to add carbohydrates to proteins that ultimately reside both intra- and extra-cellularly. OGT utilizes UDP-GlcNAc to post translationally modify intracellular serine and threonine residues with O-GlcNAc. Tight regulation of O-GlcNAc cycling relies not only on the concentrations of UDP-GlcNAc and OGT but also on O-GlcNAcase (OGA), the enzyme responsible for the modification’s removal. While modification of some proteins is sub-stoichiometric, others are more-or-less permanently modified by O-GlcNAc supporting a varied role in protein structure, localization, and function In this way, this single monosaccharide post translational modification (PTM) can dynamically respond to changes in glucose-dependent cellular metabolism and modify it’s target proteins thereby potentially acting as an “O-GlcNAc timer” for the processes it influences.

Along with the pancreas, the brain is the most heavily O-GlcNAcylated tissue, consistent with its high glucose consumption. In fact, 40% of all neuronal proteins and 19% of synaptosomal proteins are O-GlcNAc modified. Moreover, mRNAs encoding OGT and OGA are highly enriched in the brain (Kreppel et al., 1997; Lubas et al., 1997; Gao et al., 2001). O-GlcNAc modifies thousands of proteins including those correlated with neurodegenerative disease such as amyloid precursor protein and tau (associated with pathophysiology of AD), α-synuclein (component of Lewy bodies in Parkinson’s Disease, PD)(Spillantini et al., 1997; Wang et al., 2010a), superoxide dismutase (Sprung et al., 2005), and neurofilament proteins (Dong et al., 1993) (involved in amyotropic lateral sclerosis, ALS). How is O-GlcNAc regulated in the brain? Do O-GlcNAc levels change globally or only on a subset of proteins upon neurodegeneration? Which of the many roles of O-GlcNAc are important in neurodegeneration? These are some of the questions that have come under scrutiny and are summarized in (Table 1).

Table I. The major targets and pathways shown to be altered by O-GlcNAc cycling in Neurodegenerative Disease.

The target, manipulation and cell type or organism are listed on the left. The basic finding and reference are in the two right panels. The Table is not an exhaustive compilation of the published literature but is meant to illustrate the approaches that have been taken

Molecule/Protein	Agent Used	Tissue or cell line	Relevant result	Ref.
Tau	OGA inhibitor Thiamet-G	tau mutant (Pro301->Leu) mouse model of AD	Systemic administration of Thiamet-G reduced neurofibrillary tangles (NFT)	(Yuzwa et al., 2012)
Tau, APP		AD mouse model (TAPP) carrying tau (Pro301->Leu) and APP (APPSwe) mutations	Decreased tau phosphorylation and neurite plaques	(Yuzwa et al., 2014)
Tau		Post mortem AD brain frontal cerebral cortex, and cerebellum	O-GlcNAc levels reduced	(Liu et al., 2004)
		Isolated proteins from AD patients brains	Normal tau entangles with hyperphosphorylated tau	(Alonso et al., 1996)
		OGT Knockout mouse brain	Tau phosphorylation increased	(O’Donnell et al., 2004)
Tau, APP or polyglutamine expansion		C. elegans: OGT-1 and OGA-1 mutants	Neurodegeneration worsened in OGA-1 mutants, and ameliorated in OGT-1 mutant	(Wang et al., 2012)
APP		SH-SY5Y neuroblastoma cells	O-GlcNAcylation induced α-secretase mediated cleavage of APP, resulting in neuroprotective sAPPα	(Jacobsen and Iverfeldt, 2011)
Milton		Rat Hippocampal neurons	Increased O-GlcNAcylation decrease mitochondrial motility	(Pekkurnaz et al., 2014)
Ogg1		Murine neonatal cardiac myocytes	O-GlcNAcylation decreased Ogg1 activity that could lead to mDNA damage	(Cividini et al., 2016)
MIBP1		HEK293 cells	MIBP1 activity is suppressed by O-GlcNAcylation	(Iwashita et al., 2012)
Electron Transport Chain (ETC)		Diabetic rat hearts	Increased O-GlcNAcylation could reduce ATP production by inhibiting the activity of ETC complexes	(Ma et al., 2016)
TCA cycle enzymes		Diabetic rat hearts	Increased O-GlcNAc could reduce TCA activity	(Ma et al., 2016)
Insulin Signaling Pathway		3T3-L1 adipocytes	Increased O-GlcNAcylation inhibits insulin signaling pathway activation	(Vosseller et al., 2002)
SNAP-29		HeLa cells	Increased O-GlcNAcylation reduced autophagy by decreasing fusion between autophagosomes and endosome/lysosomes	(Guo et al., 2014)
BCL2		Diabetic db/db mice cardiomyocytes		(Marsh et al., 2013)
Atg7		Mouse brain lysate		(Park et al., 2015)
α-synuclein		In vitro recombinant protein	O-GlcNAcylation prevents α-synuclein aggregation	(Marotta et al., 2015)
Period		D. melanogaster	O-GlcNAcylation of regulate circadian timing by altering dPER subcellular localization	(Kim et al., 2012; Kaasik et al., 2013)
KV7.3 (Kcnq3)		AgRP neurons in mouse	O-GlcNAcylation is essential for excitability of AgRP neurons by modifying Kcnq3	(Ruan et al., 2014)

2. OGT: isoforms and localization

It is intriguing that O-GlcNAc addition is regulated by a single, ubiquitously expressed, essential enzyme, OGT. Elimination of OGT is lethal in most organisms and in dividing cells: generation of Ogt KO stem cells and Ogt KO mice has been unsuccessful. Interestingly, two recent papers have successfully removed OGT in post-mitotic adult neurons, thus opening a new area to examine how O-GlcNAcylation of proteins regulates normal brain function in viable animals.

OGT expression, localization, and regulation

The Ogt gene produces several splice variants differing at the amino (N) terminal region in mammals with splicing and multiple start codons (Hanover et al., 2012; Abramowitz et al., 2014). Indeed, in Drosophila, levels of the genes encoding OGT and a small number of other transcripts are directly modulated by the function of spliceosome (Ashton-Beaucage et al., 2010; Hanover et al., 2012; Abramowitz et al., 2014).

The three canonic isoforms that are produced from alternative start codons in the splice variants of mammalian Ogt are found in the nucleus (ncOGT and sOGT), cytoplasm (ncOGT and sOGT), or mitochondria (mOGT). The largest isoform, ncOGT, is ~110 kDa and contains 12 tetratricopeptide (TPR) repeats at the N-terminus. The structure of the dimeric ncOGT with its 12 TPR repeats is shown in Figure 3A. The TPR repeats form a structure like the armadillo repeats of importin alpha in which Asn residues line a concave surface of the molecule and interact with target proteins as demonstrated by recent crystallographic findings (Lazarus et al., 2013). For example, the TPR repeats of OGT interact with regions of the unstructured domains of nucleoporins (Cordes and Krohne, 1993) suggesting that OGT will also have an affinity for unstructured regions of other substrates (Cordes and Krohne, 1993; Lubas and Hanover, 2000; Bond and Hanover, 2015). The 102 kDa mOGT has a unique mitochondrial targeting sequence and 9 TPR domains. And, the smallest isoform, sOGT, is about 75 kDa with only 2.5 TPR domains (Lazarus et al., 2006; Lazarus et al., 2009). Given that the sOGT isoform lacks the dimerization motif found in ncOGT and mOGT, it is likely to exist in complexes either as a monomer or higher order multimers (Jinek et al., 2004).

Figure 3. The structures of OGT, OGA, and the sites of interaction with some of the binding partners of the enzymes of O-GlcNAc cycling.

The crystal structures pictured are derived from crystal structures 1W3B and 3PE4 (OGT) (Jínek et al., 2004; Lazarus et al., 2011) and 5UN8 and 3ZJ0 (OGA) (Rao et al., 2013; Li et al., 2017). (A) Structure and interaction partners of dimer ncOGT. OGT has many binding partners that are crucial for cellular processes mediating neuronal homeostasis. The substrate UDP-GlcNAc bound to the active site is shown in red. The positions in the structure corresponding the interaction domains between OGT and these molecules are indicated in brackets B) Structure and binding partners of the long isoform of O-GlcNAcase. The O-GlcNAcylated p53 peptide binds in the active site is highlighted in red. The sites of caspase 3 cleavage are indicated by the arrows. This form of OGA has been shown to interact with OGT and histones. For discussion of these interactions, see associated portions of the text.

N-terminal variation in OGT’s TPR domains influences substrate recognition (Lubas and Hanover, 2000) and, therefore, enzyme localization, and cellular activity. Indeed, while all splice variants include the same linker region and the C-terminal catalytic domain responsible for transferase activity, the shortest isoform, sOGT, has not been shown to add O-GlcNAc to proteins in vivo (Lazarus et al., 2006). mOGT targets the inner mitochondrial membrane and displays different substrate specificity from ncOGT (Love et al., 2003; Lazarus et al., 2006). Some literature suggests that ncOGT, rather than mOGT, can catalyze O-GlcNAcylation of the mitochondrial proteome (Trapannone et al., 2016); however, recent work identified that mOGT was required for glycosylation of specific mitochondrial proteins (Sacoman et al., 2017). Interestingly, overexpression of mOGT caused apoptosis in a variety of cell lines (Shin et al., 2011) supporting that tight regulation of OGT and its downstream functions is essential for cellular homeostasis.

Modulation of OGT expression or activity through UDP-GlcNAc concentration, transcription, and post-translational modification not only influences O-GlcNAc addition but also other properties. For example, the OGT C-terminal domain was shown to have an unusual UDP-GlcNAc-dependent protease activity against epigenetic cell cycle regulator HCF-1 (Capotosti et al., 2011). The posttranslational modification of OGT itself is critical for modulating OGT activity in cells, such as neuronal cells. Phosphorylation of serine 3 or 4 by GSK3β is necessary for OGT activity and phosphorylation of Thr144 by AMPK plays a role in OGT substrate selectivity. Interestingly, both GSK3β and AMPK are O-GlcNAc modified by OGT showing a feedback loop in activity modulation (Kreppel et al., 1997; Kaasik et al., 2013; Bullen et al., 2014). Cysteine (S) nitrosylation inhibits OGT activity in resting immune cells and in neuroblastoma cells. S-nitrosylation of OGT is increased by β-amyloid peptide treatment, which results in reduced O-GlcNAcylation and increased tau phosphorylation in human neuroblastoma cells (Ryu and Do, 2011; Ryu et al., 2016)

OGT and its binding partners in the brain

Recent advances in proteomics have accelerated the discovery of protein-interaction networks and OGT has emerged as a common binding partner for proteins implicated in diseases associated with human neurodegeneration. Interactions that have been mapped on the OGT protein are summarized in Figure 3A and described below.

There is evidence that an abnormal cell cycle precedes the other known hallmarks of neurodegenerative diseases and cell cycle aberrations are a known part of AD etiology and will be discussed in more depth in later sections. OGT interacts with proteins required for regulation of gene expression and cell fate decisions including HCF-1, mSin3A, and TETs (Wysocka et al., 2003; Ruan et al., 2012; Deplus et al., 2013). Among these proteins, HCF-1 is a transcriptional co-regulator necessary for cell cycle progression. Mature HCF-1 is O-GlcNAcylated and proteolytically cleaved by OGT. Following its cleavage by OGT, the N and C termini of HCF-1 functions in G1 and M phase of cell cycle, respectively (Capotosti et al., 2011). Moreover, the OGT interaction with HCF-1 regulates gene expression during G1-S progression. In addition, transcriptional regulation through either histone modification or direct modification of DNA appears to be influenced by OGT, which interacts with the histone deacetylase complex through corepressor Sin3A to promote effective, specific gene silencing (Yang et al., 2002). Further, CARM1 histone methyltransferase interacts with OGT and is O-GlcNAc modified for correct CARM1 localization during mitosis (Sakabe and Hart, 2010). Finally, ten-eleven translocation (TET) 5-mC hydroxylases promote DNA demethylation thereby functioning to modulate transcriptional activation and repression. The importance of the OGT and TET interaction is largely unknown, especially in the nervous system where one of the earliest reports about TET mediated hydroxymethylation was discovered (Kriaucionis and Heintz, 2009).

Autosomal dominant cerebellar ataxia is directly linked to an ATTCT pentanucleotide expansion in intron 9 of ATXN10 (SCA10). The ataxin-10 (Atx-10) protein is predicted to have a helical armadillo-like three-dimensional structure like the Huntingtin protein. This motif is thought to regulate protein-protein interactions, including the interaction of Atx-10 with ncOGT in the cytoplasm. In tissues, such as the pancreas, this interaction modulates OGT activity as Atx-10 competes with known ncOGT inhibitor, sOGT. Interestingly, while sOGT expression is high in the pancreas, it is absent in the brain (Andrali et al., 2005; März et al., 2006) suggesting that the Atx-10 interaction with ncOGT in neuronal tissue may alter OGT activity or protein binding in some other way. Indeed, both OGT and Atx-10 are known to modulate cytokinesis through interaction with Aurora-B, which phosphorylates Atx-10. Intriguingly, knockdown of either Atx-10 or OGT causes cytokinesis defects (Slawson et al., 2005; Li et al., 2011). The OGT/Atx-10 protein interaction is promising for understanding these proteins’ role in neuronal maintenance and neurodegeneration. What is learned about this axis may be applied to better understand the role OGT plays in interactions with other proteins containing the aforementioned secondary structure.

Along with the dynamic regulation of the enzyme in neuronal cells and its high level of expression in brain tissue, the OGT gene maps to the locus associated with the Parkinsonian dystonia (DYT3) region in mammals (Németh et al., 1999; Shafi et al., 2000), supporting its foundational importance for ensuring efficacy of neuronal nutrient-sensitive processes. Gene mutations in an important cell cycle related transcription factor, THAP1, have been linked with DYT6 dystonia (Fuchs et al., 2009). THAP1 is an atypical DNA-binding zinc finger protein known to regulate cell proliferation by binding to specific DNA regions and is part of a family of proteins that affect pRB/E2F, proposed to be involved in PD cell death (Houlden et al., 2010). Like OGT, which is highly expressed in the nervous system, THAP1 protein is found in the cerebral cortex, thalamus, substantia nigra and hypothalamus (Zhao et al., 2013). The binding of OGT to THAP1 is critical for proper timing of the G1-S cell cycle progression (Cayrol et al., 2007; Mazars et al., 2010; Zhao et al., 2013) Together, these connections suggest that understanding the interaction between OGT and THAP1 may lead us to better understand dystonias.

OGT, NF-kB, and MIBP1 are evolving as potential theraputic targets for treatment of neuronal pathologies. c-MYC intron binding protein (MIBP1) is highly expressed in postmitotic neurons and its appropriate expression is required for perinatal development of the brain as well as immunological processes. This transcription factor binds to various genomic regions including the first intron of the c-MYC gene and its most prominent binding partner, OGT, is responsible for its O-GlcNAcylation. MIBP1, whose activity is suppressed by O-GlcNAcylation, binds and decreases of the transcription of NF-kB target genes. Loss of the MIBP1 154-amino acid region mapped to bind OGT yields an increase in MIBP1-dependent suppression of NF-kB pathway target genes (Iwashita et al., 2012). Interestingly, NF-kB activation is dependent on O-GlcNAc modification of the c-Rel subunit (Ramakrishnan et al., 2013). Expressed in many cells including neurons in the cortex and hippocampus (O’Neill and Kaltschmidt, 1997), NF-kB is a critical mediator of numerous physiological processes including inflammation and the cell cycle and plays a role in neurological disorders (Shih et al., 2015). OGT, through its interaction with the proteins above, has been shown to be a key regulator of cellular homeostasis via its influence on gene expression. Thus, deregulation of OGT and related mechanisms may be associated with the cascade of events leading to neurodegeneration.

3. OGA: isoforms and localization

OGA was first identified as a hyaluronidase triggering an immune response in a meningioma patient, therefore named as MGEA5 (Meningioma-Expressed Antigen 5), (Heckel et al., 1998). Subsequently, purified OGA was shown to be the enzyme previously described as hexosaminidase C, a glycosidase acting at neutral pH (Braidman et al., 1974). Oga is located on human chromosome 10q24 (mouse Chr 19) (Gao et al., 2001) in a locus associated with AD (Myers et al., 2000) and a predisposition to type II diabetes in the Mexican American subpopulation (Farook et al., 2002; Lehman et al., 2005). While ACEVIEW and other public databases annotate multiple isoforms, only two have been studied extensively and have varied levels of activity in vitro (Kim et al., 2006; Macauley and Vocadlo, 2009). OGA-L (OGA long, 130 kDa) is localized to the cytoplasm. OGA-S (OGA short, 75 kDa) is predominantly found in the nucleus (Comtesse et al., 2001) and associated with lipid droplets (Keembiyehetty et al., 2011) with the short 11- amino acid extension of the linker domain important for targeting to lipid droplets.

Figure Figure 3B depicts a model summarizing the structure of human OGA, which was derived from three X-Ray density maps (Elsen et al., 2017; Li et al., 2017; Roth et al., 2017). The protein has an interlocking dimer interface composed of helices from the opposing monomer. Both OGA isoforms harbor the N-terminal catalytic domain responsible for glycosidase activity, but the C-terminal domains are different. Indeed, while OGA-L was originally published to have C-terminal Histone Acetyl Transferase (HAT) activity in vitro, recent work demonstrates that the domain is a pseudo-HAT as OGA-L does not function this way in vivo (Toleman et al., 2004; Rao et al., 2013). However, the pseudo-HAT domain extends from the interlocking helices and could potentially alter the conformation of the catalytic domains and is critical for binding chromatin. The linker region between catalytic and pseudo-HAT domains of OGA-L is a caspase 3 site cleaved during apoptosis (Butkinaree et al., 2008).

While OGA is thought to interact with at least 90 proteins, (Groves et al., 2017) further studies will be required to understand the consequences of OGA deregulation on individual proteins. Importantly, it has been shown that the two canonic isoforms of OGA are expressed differentially during rat development; OGA-S seems more abundant in the first 19 days and then disappears whereas the OGA-L levels are maintained throughout development (Liu et al., 2012). Further, recent studies knocking out OGA in mice revealed critical roles for OGA in development – including development of the brain (Yang et al., 2012; Keembiyehetty et al., 2015; Olivier-Van Stichelen et al., 2017). Thus, OGA – and by extension O-GlcNAc cycling – plays a critical role in neuronal homeostasis.

C. Genetic approaches targeting O-GlcNAc in neurodegenerative disease

Manipulation of OGT and OGA have been explored in isolated tissues, cultured cells, and model organisms (Summarized in Table 1). To date, the data from different neurodegenerative disease models suggest that elevated O-GlcNAc levels can both increase or attenuate symptoms. These conflicting viewpoints call for a critical review of the literature, examination of experimental conditions, and an appropriate level of weight for findings in animals other than humans. Importantly, proteins associated with neurodegenerative disease including amyloid precursor peptide (APP) (Griffith et al., 1995), tau, and α-synuclein are O-GlcNAc modified (Arnold et al., 1996; Cuervo et al., 2004) with varied consequences. In this section, we focus on results from studies in which OGT and OGA were globally manipulated in whole organisms. These data support that the enzymes of O-GlcNAc cycling and O-GlcNAc itself impact brain development and neuronal plasticity.

1. C. elegans: O-GlcNAc as a regulator of proteostasis and proteotoxicity

OGT-1 and OGA-1 were identified as regulators of proteotoxicity in multiple C. elegans models of neurodegenerative disease (Wang et al., 2012). Phenotypes in transgenic models of tauopathy, amyloid beta-peptide, and polyglutamine expansion were ameliorated in animals lacking OGT-1 while loss of OGA-1 enhanced some phenotypes. These findings appear to contradict some of the findings from rodent studies wherein increases in O-GlcNAc through OGA inhibition appear to improve tau-driven neurodegeneration phenotypes (see section below) (Yuzwa et al., 2008; Yuzwa et al., 2012) . While developmental regulation may explain some differences in findings from mammals to C. elegans, the variation in phenotypes scored and the life stage at which O-GlcNAc was perturbed may also contribute.

Perturbation in O-GlcNAc alters proteostasis in C. elegans (see section on autophagy), a process which is critical for neuronal health. Transcription of proteasome subunit genes is induced in response to proteasome dysfunction caused by pathogen or proteasome inhibitor exposure. This pathway may also be induced by proteotoxic challenge and that response requires SKN-1, a transcription factor related to mammalian Nrf1/2. Intriguingly, Nrf1/2 is negatively regulated by O-GlcNAcylation and is thus an ideal target for pharmacological intervention (Chen et al., 2015).

2. Drosophila: O-GlcNAc and epigenetic regulation of development, cell cycle, and circadian rhythm

The Drosophila ogt(sxc) was discovered as a homeotic gene about 3 decades ago (Ingham, 1984). Subsequent studies underscored its importance as a member of the Polycomb repressor complex (PRC). Polyhomeotic (Ph) is O-GlcNAc modified and in the absence of OGT and, Ph forms large aggregates resulting in a non-functional PRC. Consequently, Hox genes were inappropriately expressed throughout the embryo (Gambetta et al., 2009b; Gambetta and Muller, 2014) resulting in ogt mutant embryos arresting during development.

Epigenetic regulation of gene activation is also partly regulated by O-GlcNAcylation. Trithorax (TRX), Absent Small or Homeotic 1 (ASH1), and SET1 O-GlcNAc modified. In Drosophila lacking OGA, increased O-GlcNAc results in changes in the expression of cell cycle related genes (Akan et al., 2016). Cell cycle dysregulation is an early indicator of neurodegeneration and these findings support O-GlcNAc as critical for appropriate developmental and cell cycle dynamics. In addition, work in Drosophila has revealed a role for O-GlcNAc in regulating circadian rhythm (Kim et al., 2012; Kaasik et al., 2013). Reduction in OGT levels in clock cells shortened circadian rhythm. Drosophila PERIOD (dPER) is O-GlcNAc modified this modification and stabilize and locates dPER in cytoplasm, where it is inactive. Upon reduction in O-GlcNAc levels, dPER could go to nucleus where it functions to repress its own transcription by inhibiting the activity of CLOCK/CYCLE heterodimer (Kim et al., 2012). Based on these results, O-GlcNAcylation seems to regulate proper circadian timing. Such disturbance in daily rhythms involing sleep, activity and hormone release have been linked to human neurodegenerative disease and therefore may represent a possible contributor to neurodegenerative disease progression (Association, 2017).

3. Mouse: neurodegenerative phenotypes associated with perturbation of O-GlcNAc

To decipher the role of O-GlcNAc and the proteins responsible for its cycling, numerous studies have used genetics to alter OGT and OGA expression in mice. Although full body knockout of Ogt and Oga are lethal in mice (Shafi et al., 2000; Keembiyehetty et al., 2015), tissue specific approaches have been fruitful. These studies have allowed researchers to better understand the role of O-GlcNAc cycling in the brain.

Ogt knockout in mouse models

Knockout of Ogt in mouse embryonic stem (ES) cells is lethal requiring knockout of Ogt in selected tissues (Shafi et al., 2000; Hanover et al., 2003; O’Donnell et al., 2004). Using Syn1-Cre driver, O’Donnell et al. triggered Ogt deletion in neuronal cells (O’Donnell et al., 2004). While heterozygote animals exhibited no phenotypes, homozygous knockout of Ogt was rare and animals rarely survived past nursing. Further, those that survived were considerably smaller with reduced motility compared to wild type (WT) littermates. Biochemical analysis demonstrated that tau expression and phosphorylation was increased in animals lacking Ogt in neurons. These findings suggest a possible mechanistic link between O-GlcNAc and neurodegeneration given that tau phosphorylation and O-GlcNAcylation are linked in vivo.

More recently, excitatory neuron-specific knockouts have been generated in mice by multiple groups who took advantage of an αCamKII-Cre driver (Lagerlöf et al., 2016; Wang et al., 2016). These two studies looked at animal development, glucose tolerance, and later physiological phenotypes. Surprisingly, the findings from the studies and interpretations are divergent. Wang, et al. demonstrated that animals without Ogt in excitatory neurons had reduced weight at 7 weeks of age (Wang et al., 2016). In contrast, mice in the related study showed increased weight gain and a significant amount of fat 3 weeks after birth (Lagerlöf et al., 2016). The weight gain in the latter was attributed to satiety defects and could be regulated by food restriction. Major loss of OGT was localized to the paraventricular nucleus (PVN) of the hippocampus and Lagerlof and colleagues concluded that loss of OGT regulates the excitatory synaptic transmission in PVN neurons (Lagerlöf et al., 2016). Conversely, Wang et al. showed significant increases in neuronal cell death by 6 months after Ogt ablation (Wang et al., 2016). This coincided with decreases in brain size, cortical thickness, and neuronal density. Loss of OGT yielded mice that recapitulate neurodegenerative phenotypes with increased cell cycle and immune response, phosphorylation of tau, appearance of tau aggregates, and soluble oligomeric Aβ peptide. Ogt loss also increased anxiety, another characteristic feature of neurodegeneration. While these studies used the same promoter to drive Cre expression, it is important to note that the approaches differed in other details. Lagerlof, et al. used a tamoxifen-inducible system, knocking out the Ogt in αCamKII expressing cells at 6 weeks of age whereas the Wang, et al. knockout of Ogt is initiated earlier (4 weeks) at the onset of αCamKII expression and synaptic maturation (Lagerlöf et al., 2016; Wang et al., 2016). Some common findings emerged. Both studies demonstrated the importance of OGT in excitatory neurons and provide clear evidence for O-GlcNAc cycling involvement in metabolism and neurodegeneration. However, the contrasting findings underscore the importance of taking developmental timing into account when analyzing the results of such knockout models. It is also critically important to examine toxicity and cell viability at every stage in such experiments.

Knockout of OGT in sensory neurons cause axonal defects, leading to thermal and mechanical hyposensitivity (Su and Schwarz, 2017). The phenotype is most likely due to the decrease in axonal endings. Moreover, OGT knockout in AgRP neurons results in a loss in neuronal excitability (Ruan et al., 2014). Possibility of OGT knockout in the nervous system is promised to highlight many essential functions of O-GlcNAc in the cell.

Oga knockout in mouse models

Previously, we generated a full body knockout of Oga (MGEA5) in mice using an oocyte-expressing Cre-line (MMTV-Cre) and demonstrated that Oga KO mice are largely perinatal lethal with few offspring surviving more than 3 weeks (Keembiyehetty et al., 2015). The perinatal lethality was associated with defects in neonatal liver glycogen storage, possibly linked to altered autophagy (Keembiyehetty et al., 2015). A similar perinatal lethality phenotype was observed in non-conditional, genetrap knockout model targeting Oga (Yang et al., 2012). The surviving Oga KO mice we generated were obese, and even heterozygous animals showed evidence of insulin-resistance. In addition to gross structural and developmental defects, Oga KO animals exhibit metabolic deregulation including impaired insulin and glucose homeostasis. A key feature of the Oga heterozygous mice included insulin resistance when challenged with a high fat diet. Due to their short lifespan and poor survival at birth, full body Oga KO animals are difficult to study and tissue-specific knockout models were used for further interrogation.

Using a highly neuron specific Nestin-Cre promoter to knockout Oga in the brain (Olivier-Van Stichelen et al., 2017), we found that Oga brain KO animals show significant increases in brain O-GlcNAc levels but little change in other tissues such as liver, heart, and muscle. Intriguingly, the levels of brain OGT is also diminished suggesting a feedback-regulation of OGT in response to elevated O-GlcNAc levels. The mechanism of this reduction in OGT levels is not known but may be due to changes in chromatin structure or splicing since the decreases occurred principally to the ncOGT isoform. The brain Oga KO animals exhibit several anatomical, behavioral, and metabolic phenotypes. Compared to WT, brain Oga KO animals were smaller, obese, and have a visibly flattened forehead. Closer neuroanatomic analysis revealed enlarged ventricles (lateral ventricles, 3^rd and 4^th), smaller forebrain structures, cortical layering, and a very small olfactory bulb. Examining neurogenesis, we found that brain Oga KO animals showed a dramatic developmental delay associated with persistent neurogenesis long after neurogenesis has ended in WT animals. The Oga brain knockout animals also exhibited striking metabolic imbalances including increased body fat, circulating leptin, triglycerides, and insulin as well as decreased insulin-like growth factor 1 (IGF-1) levels. Consistent with lower levels of GH and IGF-1, the brain Oga KO animals have a pituitary that is much smaller than in WT animals. These findings reinforce that O-GlcNAc cycling is important for maintaining the hypothalamus-pituitary axis and, thus, the maintenance of metabolic homeostasis (Olivier-Van Stichelen et al., 2017). Despite the physiological changes cataloged above, loss of Oga in the brain did not increase neurodegenerative damage or physiological apoptosis (Olivier-Van Stichelen et al., 2017).

Pharmacological findings have suggested that inhibitors of OGA may have therapeutic potential by reducing protein aggregation (Yuzwa et al., 2012; Yuzwa et al., 2014; Hastings et al., 2017). As Oga brain KO animals show reduced levels of OGT in the brain, it is unclear whether interference with O-GlcNAc cycling may provide neuroprotective effects by reducing OGA levels, by lowering OGT levels, or altering protein O-GlcNAcylation on specific substrates. While the brain specific Oga KO has provided essential information about the role of O-GlcNAc in neuronal development and function, neuron-specific Oga KO in a neurodegenerative disease mouse model will be required to examine the potential neuroprotective effects of interfering with O-GlcNAc cycling.

4. O-GlcNAc in neurodevelopment: impact of the intrauterine environment

Human neurodevelopment can be influenced by environmental perturbations in the intrauterine environment. Prenatal development represents a vulnerable period where the fetal and maternal compartments exchange nutrients and neuroendocrine signals. The mammalian placenta plays a critical role in this exchange and environmental factors including nutrients can influence developmental programming. Based on evolutionary and transcriptional evidence, we proposed a role for the enzymes of O-GlcNAc cycling as epigenetic regulators in the intrauterine environment (Love et al., 2010; Hanover et al., 2012). Subsequent work demonstrated that the enzymes of O-GlcNAc cycling are likely to play a crucial role in developmental programming (Howerton et al., 2013; Howerton and Bale, 2014; Nugent and Bale, 2015; Bale, 2016; Olivier-Van Stichelen et al., 2017; Pantaleon et al., 2017). As described above, disruption of OGT and OGA in the brain profoundly impacts brain development and function. Further, mutations in human OGT are associated with X-linked intellectual disability (XLID) providing direct evidence for an important role for this enzyme in neurodevelopment (Bouazzi et al., 2015; Niranjan et al., 2015; Vaidyanathan et al., 2017).

Several studies suggest that maternal stress increases the risk of neurodevelopmental diseases including autism, stuttering, dyslexia, and attention deficit/hyperactivity disorder and that males are more significantly affected than females (reviewed in (Bale, 2016)). These sex differences could be due to hormonal and metabolic influences, sex chromosomes, or a combination of these two factors. By examining the expression of X-linked genes differentially expressed in the placenta itself, OGT has emerged as a major biomarker of placental stress (Howerton et al., 2013; Howerton and Bale, 2014). Intriguingly, the increases observed are far more pronounced in the male offspring-specific placental stress response.

The presence of OGT on the mammalian X-chromosome near the Xist locus (lnRNA triggering X-inactivation) has important implications for OGT expression and subsequent functioning in both males and females (reviewed in (Love et al., 2010; Abramowitz et al., 2014; Olivier-Van Stichelen and Hanover, 2014; Olivier-Van Stichelen et al., 2014a)). OGT is one of the genes on the human X-chromosome with a low ratio of non-synonymous to synonymous substitution rates suggesting that it may represent a clinically significant gene that is subject to negative selection (Ge et al., 2015). In addition, OGT escapes X inactivation in the placenta resulting in transcript levels that are nearly 2-fold higher in female compared to male placenta (Howerton et al., 2013; Howerton and Bale, 2014). Knockout of OGT in the placenta to produce hemizygous and homozygous deletion of OGT in trophoblasts produced offspring with altered hypothalamus-pituitary-adrenal axis function. These changes largely recapitulate the early placental stress phenotype (Howerton and Bale, 2014). Thus, OGT is likely to play an important role in mediating the response to stress in the intrauterine environment, thereby impacting fetus development.

D. O-GlcNAc: A central role in cellular regulation of neuronal health and disease

There are several lines of evidence supporting that appropriate maintenance of O-GlcNAc at both global levels and on individual proteins is critical for mitigating neurodegenerative disease. To understand how changes in O-GlcNAc may impact brain function, researchers have focused on individual proteins associated with neurodegenerative disease noting that disease-associated proteins are O-GlcNAc modified (Griffith et al., 1995). Individual proteins’ roles are impacted by their O-GlcNAc status and thereby influence processes ranging from proteostasis and autophagy to insulin signaling and the cell cycle.

Insulin Signaling

Insulin signaling in the brain is considered a key player in brain homeostasis. The molecular response of the brain to insulin is analogous to the response in other organs. Insulin receptors (IR) are found throughout the brain (Marks et al., 1990) with internal or peripheral secretion of insulin triggering insulin signaling through PI3K activation and transcriptional activity. Both type 1 and 2 diabetic patients have increased risk of neurodegeneration (Kern et al., 2001) and altered expression of diabetes-related genes have been linked to AD pathology (Hokama et al., 2014). With insulin response impaired in AD brains, AD has been categorized as ‘type 3 diabetes’ without hyperglycemia. Moreover, insulin resistance in peripheral tissue reduces insulin uptake, increases amyloid beta levels (Baker et al., 2011), and accelerates the formation of neurite plaques (Matsuzaki et al., 2010).

While other tissues use the insulin pathway mainly for glucose metabolism, brain insulin signaling influences pathology by triggering multiple functions including glucose metabolism, reproduction, proliferation, differentiation, neuroprotection, cognition, and memory. Insulin in the brain does not induce a significant uptake of glucose likely due to low levels of the insulin responsive glucose transporter GLUT4. Nevertheless, brain insulin signaling participates in peripheral glucose homeostasis. Insulin signaling modulates glucose production by the liver (Obici et al., 2002), influences satiety (Kim et al., 2000), and regulates hormone secretion (Tanaka et al., 2000). Multiple studies have demonstrated the importance of insulin signaling for brain development, growth, neuronal outgrowth and maturation, and axon regeneration (Schubert et al., 2003; Toth et al., 2006; Parathath et al., 2008). Similarly, proliferation and differentiation of multipotent neural stem cells are regulated by insulin (Wozniak et al., 1993)

In addition to promoting proliferation, insulin signaling seems to protect neural cells against apoptosis though the PI3K/Akt/mTOR pathway (Ryu et al., 1999) and prevents beta-amyloid induced cell death though an undefined mechanism (Rensink et al., 2004). Cognitive function may also be influenced by insulin signaling in the brain as increasing insulin signaling in the hippocampus improved memory and learning (Zhao et al., 1999). Furthermore, it has been shown that insulin administration eliminates the loss of memory due to ischemic lesions (Voll et al., 1989).

Nutrient-driven O-GlcNAc may be an important regulator of brain insulin signaling. First, secretion of insulin is epigenetically regulated by O-GlcNAcylation in pancreatic beta cells (Durning et al., 2016). Major components of the insulin signaling pathway are O-GlcNAc modified including: the insulin receptor (beta-chain), IRS (Insulin Receptor Substrate), PI3K (Phosphatidylinositol 3 kinase), PDK1 (Phosphoinositide-Dependent Kinase 1), and Akt/PKB (Ball et al., 2006; Gandy et al., 2006; Klein et al., 2009; Whelan et al., 2010). It has been proposed that phosphorylation of IRS1, PI3K, and Akt/PKB activates the insulin signaling pathway while O-GlcNAcylation of the same proteins prevents pathway activation (Holman and Kasuga, 1997). In this model, O-GlcNAc competition with phosphorylation is accentuated by OGT recruitment to the plasma membrane through PPO domain (PIP-binding activity of OGT) (Yang et al., 2008b; Perez-Cervera et al., 2013). However, this proposed mechanism remains controversial as both supporting (Vosseller et al., 2002; Arias et al., 2004; Gandy et al., 2006; Yang et al., 2008b; Klein et al., 2009) and refuting data have been published (Macauley et al., 2008; Macauley et al., 2010). While O-GlcNAc influences insulin signaling, a direct connection with insulin resistance in neurodegenerative diseases has yet to be defined.

Co-translational O-GlcNAcylation and quality control

Quality control of newly synthesized proteins is an essential feature in cell homeostasis. Failure to detect and dispose of misfolded proteins results in proteotoxity and often correlates with disease. One of the key steps in cytosolic quality control is the co-translational ubiquitination of proteins targeted for proteasomal degradation (Bengtson and Joazeiro, 2010). For example, the generation of non-stop mRNA is a phenomenon common across all organisms and a major source of proteotoxicity. Non-stop mRNA encodes aberrant, non-stop proteins sequestrated in the ribosomal complex that cannot be recycled or corrected by a quality control chaperone (Doma and Parker, 2007). Interestingly, O-GlcNAcylation also occurs co-translationally (Starr and Hanover, 1990) and acts with ubiquitination to regulate the stability of the nascent proteins (Zhu et al., 2015b). Previous work supports that O-GlcNAc is a key player in protein stability (e.g., Sp1, beta-catenin) (Han and Kudlow, 1997; Olivier-Van Stichelen et al., 2014b) resulting from direct competition with phosphorylation or through O-GlcNAc modification of the proteasome itself (Sümegi et al., 2003; Zhang et al., 2003). O-GlcNAc addition was initially shown to occur on nascent chains and post-translationally by in vitro translation of the O-GlcNAc target nuclear pore protein p62 (Nup62) (Starr and Hanover, 1990). More recently Zhu et al. demonstrated that co-translational O-GlcNAcylation of Sp1 and Nup62 prevents the nascent proteins from precocious proteasomal degradation (Zhu et al., 2015b). O-GlcNAc addition and removal may then be regarded as an ‘O-GlcNAc timer’ serving a protective role at the time of synthesis on the ribosome and regulating the rate of degradation of nuclear and cytoplasmic proteins. This is similar to the so called ‘Mannose timer’ in the endoplasmic where the action of mannosidase I can trigger proteasomal degradation. It would appear that O-GlcNAc may play an analogous role in the nucleus and cytosol. We therefore suggest that an ‘O-GlcNAc’ timer’ may act in a similar manner to regulate protein stability. If O-GlcNAc addition and removal are both timed events in the life-cycle of the protein, then independent regulation of these activities could alter the stability of the target protein. When altered, O-GlcNAcylation may also significantly contribute to neurodegenerative pathologies.

Cell cycle regulation

Although neurons in the CNS are often viewed as post-mitotic and stable in number, a large body of recent evidence suggests that adult neurogenesis depends upon continued cell division of stem cell precursors. The major sites of adult neurogenesis in humans are two “neurogenic” brain regions. These sites are the subgranular zone (SGZ) in the dentate gyrus of the hippocampus and the subventricular zone (SVZ) of the lateral ventricles where new neurons are generated before migrating through the rostral migratory stream (RMS) to the olfactory bulb to become interneurons (Gage, 2000). The relationship between adult neurogenesis and neurodegenerative disease is complex and has been reviewed elsewhere (Winner et al., 2011; Winner and Winkler, 2015).

The cell cycle is essential for every organism to emerge, develop and survive. Divided in 4 major phases (G0, S, G1 and M), cell cycle progression is regulated by expression and action of cyclins (A, B, D, E) and their partners, the cyclin-dependent kinases (CDKs). In neurodegenerative diseases, there is growing evidence that abnormal cell cycle re-entry precedes other hallmarks of the disease and implicates an aberrant cell cycle in disease progression. Data suggest that at the onset of AD, neurons increase cell cycle re-entry but fail to complete the mitotic phase and thus die (Bonda et al., 2010). AD neurons exhibit elevated cell cycle markers (cyclin D, cdk4, Ki67, cyclin E/cdk4) (McShea et al., 1997; Nagy et al., 1997; Vincent et al., 1997). Indeed, no sign of a complete mitotic phase is found in AD neurons (Ogawa et al., 2003a; Ogawa et al., 2003b).

Inhibition of OGT and OGA prevents cell cycle progression indicating that O-GlcNAc levels are regulated to ensure proper cell cycle progression (Slawson et al., 2005; Dehennaut et al., 2007; Wang et al., 2010b; Drougat et al., 2012). Moreover, OGT and OGA are associated with the midbody and participate in cytokinesis (Slawson et al., 2008). In response to a mitogenic signal, protein O-GlcNAcylation increases as does localization of O-GlcNAc modified proteins to the nucleus (Kearse and Hart, 1991a). Functional studies of O-GlcNAc modified proteins such as p53 (Yang et al., 2006), p27 (Qiu et al., 2017), cyclin D1 (Olivier-Van Stichelen et al., 2012), CDK5 (Ning et al., 2017), c-myc (Chou et al., 1995), beta-catenin (Olivier-Van Stichelen et al., 2014b), NF-kB (Yang et al., 2008a; Kawauchi et al., 2009; Ramakrishnan et al., 2013), FoxM1 (Caldwell et al., 2010), HCF-1 (Capotosti et al., 2011), and Erk1/2 (Jiang et al., 2016) have been critical for furthering the understanding of O-GlcNAc in cell cycle progression. Key AD players are known to be O-GlcNAc modified including APP and tau, which have also been implicated in cell cycle regulation. Both APP and Aβ are mitogenic in vitro (Schubert et al., 1989; Milward et al., 1992). Interestingly, if APP is cleaved by α-secretase facilitated by O-GlcNAcylation, it becomes non-amyloidogenic and neuroprotective sAPPα (Jacobsen and Iverfeldt, 2011). Hyperphosphorylated tau is present in normal mitotic cells and has been linked to CDK activity (Brion et al., 1994; Preuss and Mandelkow, 1998). That O-GlcNAc cycling has been implicated as a key regulator of cell cycle progression and modifies proteins associated with neurodegeneration supports that O-GlcNAc might be involved in the aberrant cell cycles found in neurodegenerative diseases.

Autophagy and mitophagy

Recycling of cellular contents and elimination of damaged organelles/protein aggregates is required to ensure cellular homeostasis. Autophagy, the dynamic lysosomal degradation process, works in concert with endocytic mechanisms to process these proteins and organelles. Autophagy deregulation threatens neuronal health, especially as the brain ages. Mutations in autophagy pathway components are associated with neurodegenerative diseases as cataloged by Nixon (Nixon, 2013). Defects in autophagy induction, cargo recognition and sequestration, and efficient digestion of sequestered substrates yield accumulation of toxic proteins. Indeed, neurons accumulate waste over a lifetime and, while young neurons are effective and efficient at clearing waste, neurons that are compromised in autophagy accumulate protein aggregates and ultimately degenerate.

In multiple neurodegenerative diseases, there is increased autophagy induction (AD and ALS (Boland et al., 2008)), lysosome biogenesis (Cataldo et al., 1994; Cataldo et al., 1995; Cataldo et al., 1996; Ginsberg et al., 2010), and increased transcription of autophagy-related genes (Lipinski et al., 2010). Accumulation of autophagic structures are often noticeable in affected brain tissue (Kegel et al., 2000; Nixon et al., 2005; Boland et al., 2008). In Huntington’s disease, autophagosomes are unusually devoid of cargoes (Nixon, 2013). Mutation of α-synuclein in PD yields aggregates than cannot be degraded by chaperone-mediated autophagy (Cuervo et al., 2004). Recent evidence supports that the O-GlcNAc modification of calpain, responsible for α-synuclein cleavage, inhibits its action (Levine et al., 2017), thereby influencing processing of this key neurodegenerative protein. And, PINK1 and parkin mutations result in mitochondria that cannot undergo mitophagic disposal (Kitada et al., 1998; Valente et al., 2004; Park and Lee, 2006). Neurons in AD show strong phenotypic overlap with lysosomal storage disorders, but these findings have yet to be fully understood (Querfurth and LaFerla, 2010; Platt et al., 2012)

Given that neurons are particularly susceptible to dysfunction in the cellular waste stream, many researchers have endeavored to examine the regulators of autophagy and apoptosis. The dynamic O-GlcNAc is poised to provide crosstalk between autophagy and apoptosis by acting as rheostat to fine-tune these processes by modulating protein folding, localization, stability, etc (Fahie and Zachara, 2016). Loss of OGT and OGA lead to elevated induction of starvation-induced autophagy in the model organism C. elegans (Wang et al., 2012). This study also suggested that O-GlcNAc cycling plays a key role in regulating autophagy flux and proteasome activity when faced with a proteotoxic challenges such as pathological tau accumulation. Several studies in C. elegans noted that OGT is associated with phagosomes and protein O-GlcNAcylation influences autophagosome and lysosome fusion (Wang et al., 2012; Guo et al., 2014).

In vitro work in Neuro2A cells suggests that modulation of O-GlcNAc cycling affects autolysosome formation. Furthermore, the same study demonstrated that in cells and Drosophila, changes in O-GlcNAcylation modulate huntingtin toxicity (Kumar et al., 2014). Modulation of Atg proteins by O-GlcNAcylation is likely to alter autophagy (Jo et al., 2016; Wani et al., 2016b). It is not surprising that O-GlcNAc plays a role in autophagy as key autophagy regulators including Beclin1 (involved in autophagy initiation), SNAP-29 (SNARE complex member), (Guo et al., 2014) BCL2 (Marsh et al., 2013) and Atg7 (Park et al., 2015) are modified by O-GlcNAc. The interaction axis between SNAP-29, Syntaxin-17, and VAMP8 is thought to be controlled by OGT expression in HeLa cells. The authors argue that O-GlcNAcylation of SNAP-29 in both mammalian cells and C. elegans abrogates the complex’s formation thereby decreasing fusion between autophagosomes and endosomes/lysosomes (Guo et al., 2014). Further, they suggest this is intimately involved with nutrient status as starvation influences O-GlcNAcylation of SNAP-29. In Drosophila, a decrease in OGT is associated with an increase of Atg proteins and autolysosomes, and it has been shown that autophagy is regulated through Akt/dFOXO signaling (Park et al., 2015). Defective autophagy induction and processing are associated with accelerated pathology and increased neurotoxicity. Thus, we suggest that O-GlcNAc is poised to influence processing of autophagosome-targeted proteins, and thus neurodegeneration, given its key roles in autophagy.

Apoptosis

In rodent models of neurodegeneration, apoptotic cell death has been convincingly linked to neurodegeneration; although in humans, this connection is less clear (Radi et al., 2014). The regulation of apoptosis is tightly linked to cellular growth signals and, in the central nervous system where many post-mitotic cells must exist for decades, these control mechanisms are extraordinarily important (Frade and Ovejero-Benito, 2015). O-GlcNAcylation has long been known to influence apoptosis by modification of proteins linked to regulation of apoptotic cell death (Omary et al., 1998; Liu et al., 2004b; O’Donnell et al., 2004; Lazarus et al., 2009; Shin et al., 2011). More recently, the mechanisms by which this may occur on targets such as p53 have been examined (Yang et al., 2006; Kawauchi et al., 2009). There are numerous reports of O-GlcNAcylation altering mitochondrial function linked to apoptosis(Shin et al., 2011; Palaniappan et al., 2013; Bond and Hanover, 2015; Zhu et al., 2015a; Wani et al., 2016a; van der Harg et al., 2017). A link between apoptosis and neurodegenerative disease has been made in some studies (Liu et al., 2004b; Zhu et al., 2015a; Chen et al., 2017; Ning et al., 2017). Knockout of Ogt in excitatory neurons led to neurodegeneration with some indications that apoptosis-related genes might be altered (Wang et al., 2016), whereas another study showed no effect on apoptosis or neuron viability (Lagerlöf et al., 2016). Thus, the impact of OGT loss on apoptotic neuronal cell death is unclear. Deletion of Oga in the brain did not significantly increase the number of cortical neurons undergoing apoptosis (Olivier-Van Stichelen et al., 2017). Taken together, these findings suggest that O-GlcNAc may play limited role in neuronal cell death during neurodegeneration and that the modification may serve a protective role.

Mitochondrial movement & metabolism

Mitochondria, the cell’s primary source of ATP, is essential to sensing nutrient status as it utilizes glucose as its major carbon source. Neurons’ complex morphology results in a spatially heterogeneous glucose intake and, consequently, mitochondria need to rapidly re-localize to ensure rapid production of ATP (Detmer and Chan, 2007). Rapid ATP production is crucial for processes including synaptic assembly, transmission, and Ca2+ buffering with motor proteins including kinesin and dynein regulating mitochondrial motility (Schwarz, 2013). Deregulation of mitochondrial trafficking is associated with neurological disorders including AD, ALS, and PD. For example, the GTPase Miro (also called RhoT1/2) interacts with the adaptor protein Milton (also called TRAK1/2 and OIP106/98) thus coupling motor protein complexes to the mitochondria (Glater et al., 2006; Macaskill et al., 2009; van Spronsen et al., 2013). OGT interacts with Milton in both Drosophila and mammals (Iyer et al., 2003; Glater et al., 2006; Brickley et al., 2011). In response to an increase in glucose flux, O-GlcNAcylation of Milton1 (OIP106) decreases mitochondrial motility (Pekkurnaz et al., 2014). Furthermore, knockout of OGT or blockage of UDP-GlcNAc synthesis in both Drosophila and mouse models resulted in 60% of axonal mitochondria to remain stationary (Pekkurnaz et al., 2014). Finally, the Milton-like protein OIP106 forms a complex with OGT and RNA Pol II suggesting that OGT is also targeted to transcriptional machinery thereby potentially regulating downstream targets by O-GlcNAcylation. While Milton O-GlcNAcylation seems to be the main driver of mitochondrial trafficking, tubulin, kinesin, and dynein are also substrates of OGT and their O-GlcNAcylation may also contribute to mitochondrial dynamics (Ji et al., 2011; Ruan et al., 2012; Trinidad et al., 2012). Thus, impaired mitochondrial motility likely impacts the development of neurological disorders and OGT/ O-GlcNAcylation may play a critical role linking nutrient status to mitochondrial motility (MacAskill and Kittler, 2010).

Beyond proteins directly involved in mitochondrial motility, Hu et al. demonstrated that mitochondrial proteins including electron transport chain (ETC) complex members such as NDUFA9 and COX1 are O-GlcNAc modified (Hu et al., 2009). The activity of the ETC complexes was impaired by high glucose treatment mimicking diabetic conditions, with reduced ATP production. The authors interpreted this to suggest that protein O-GlcNAcylation of ETC complex members directly impacts mitochondrial function. Proteins including the TCA cycle enzymes (such as ACO2, IDH3A/3B, OGDH, etc.), redox enzymes (SOD2, PRDX3), and fatty acid oxidation enzymes are also modified by O-GlcNAc (Ma et al., 2016). Surprisingly, the diabetic conditions did not lead to an increase in protein O-GlcNAcylation across the board. In fact, 40% of the detected O-GlcNAc modified sites displayed a 20% reduction in O-GlcNAc levels under high glucose treatment (Ma et al., 2016).

Appropriate levels of protein O-GlcNAcylation are essential for mitochondrial function. Cividini et al. reported that mitochondrial 8-oxoguanine DNA glycosylase (Ogg1) activity is negatively regulated by O-GlcNAc (Cividini et al., 2016). Diabetic mice displayed increased Ogg1 O-GlcNAcylation in cardiac myopathy, leading to decreased Ogg1 activity and that could result in mDNA damage. In addition, O-GlcNAcylation of the voltage dependent anion channel (VDAC) is protective of mitochondrial function preventing the formation of calcium induced mitochondrial permeability transition pore (mPTP) (Ngoh et al., 2008). Upon formation, mPTP disrupts the mitochondrial membrane potential thereby preventing ETC mediated ATP production, triggering events in the apoptosis cascade by releasing pro-apoptotic proteins (Ngoh et al., 2008).

Combined, these studies support that the O-GlcNAc modification must be examined at the single protein – or even single site – level and that the regulation of mitochondrial function by O-GlcNAc is nuanced. This is particularly important for nervous system, and the brain, since most neurons are post mitotic and perturbed mitochondrial function could cause irreversible damage to the neurons. The ubiquity of O-GlcNAc in brain tissue and its dynamic nature on mitochondrial proteins support its role in mitochondrial motility as well as synaptic transmission.

E. Prospects for Imaging and Diagnostics

Modern brain imaging modalities exploit glucose uptake and differential blood flow as surrogates for neuronal activity allowing the monitoring of brain activity and probing neuropathology. While the brain is highly dependent on glucose under normal conditions, neurodegenerative diseases greatly change the brain’s metabolism of glucose. Therefore, glucose metabolism has been a surrogate for monitoring neuronal activity with hypo- and hyper-metabolism diagnostic for diseases (Zhu et al., 2014), including AD and PD (Mergenthaler et al., 2013). Indeed, loss of neurons can result in decreased localized glucose consumption but increased regional consumption elsewhere.

1. O-GlcNAc levels as a surrogate for neurodegenerative damage

Tantalizing evidence suggests that altered brain O-GlcNAc levels correlate with changes in neuronal cell death, neuroinflammation, production of hyperphosphorylated tau, amyloidogenic Aβ peptides, and memory deficits (Wang et al., 2016). A number of studies demonstrate that O-GlcNAcylation is decreased in AD brain tissue (Liu et al., 2004b) compared to control and OGT protein expression is reduced in AD human cortical brain tissue (Liu et al., 2009). The authors suggest that hypometabolism of glucose in AD coupled with the decreased nutrient-sensor O-GlcNAc correlates directly with neurodegeneration in these tissues. However, other work suggests that levels of O-GlcNAc increase in AD brain tissue (Förster et al., 2014). Experimental differences including choice of brain region likely influences the conclusions about global O-GlcNAc levels. Indeed, genetic and pharmacological manipulation in different organisms occasionally yields conflicting results and it is essential to continue profiling how O-GlcNAc levels are altered in neurodegenerative human tissue. There is great promise in being able to visualize O-GlcNAc, OGT, or OGA in patients in real time to give an indication of disease progression. Below, we discuss the use various imaging modalities to monitor glucose consumption and downstream nutrient-sensitive OGA in human patients in real time.

2. Can O-GlcNAc-based imaging improve diagnostics and patient outcomes?

With clinically overlapping features, neurodegenerative diseases including PD, multiple system atrophy, progressive supranuclear palsy, corticobasal degeneration, dementia with Lewy Bodies, AD, and frontotemporal dementia are difficult to differentiate at early stages. To improve patient care, the following strategies are employed: techniques that can assist in early diagnosis, define disease stage, stratify patients, and determine whether chosen therapies are successful. Thus, prediction of which patients will develop neurodegenerative diseases by use of non-invasive imaging techniques is of great interest (Weiner and Khachaturian, 2005) .

Methods including magnetic resonance imaging (MRI) and positron emission tomography (PET) have long been employed by physicians to monitor the structure and function of the brain, respectively. Briefly, MRI is a non-invasive method providing pictures of anatomy. PET imaging takes advantage of bio-available compounds tagged with radionuclides that are metabolized or bind to receptors/sites of drug action. These so-called radiotracers can be used to track a given compound’s biological utilization. Of note, PET scans are increasingly used in combination with imaging that provides anatomic information so that anatomy and function can be interpreted simultaneously.

Based on nuclear magnetic resonance, MRI utilizes magnetic fields, radio waves, and field gradients to generate contrast images as different tissues and fluids provide natural contrast. Topographic distribution of brain tissue can be characterized by neuroimaging and brain atrophy is a clear indicator of AD with loss of brain volume correlating with cognitive deterioration (Fox et al., 1999). While structural imaging can give an indication of pathological features prior to evident clinical symptoms, it is questionable whether this technique can be used early enough to improve outcomes (Johnson et al., 2012). Current optimal uses for MRI include patient stratification and measuring outcomes after treatment. As MRI cannot be used to assess function, functional MRI (fMRI) has become more common. fMRI monitors how the brain responds to an external stimulus or pharmacological agent by measurement of neuronal activity inferred from changes in signal and is often used to evaluate functional connectivity within brain networks (Johnson et al., 2012). This technique has great potential to assess treatment strategies in real-time. However, longitudinal fMRI studies in neurodegenerative patients is challenging and must be further explored to understand pitfalls and its potential impact.

Researchers have long used PET to monitor drug uptake, elimination, and bio-distribution in a living animal. Glucose hypo-metabolism is found in brain regions prone to neurodegeneration, thus profiling glucose utilization and associated molecules of interest to assess disease state and treatment efficacy is increasing in popularity. Several radiotracers have been developed to monitor brain function and the chemical arsenal of PET ligands includes compounds labeled with isotopes such as fluorine-18 (half-life of ~110 min) and carbon-11 (~20 min). For example, the biologically active glucose analog fluorodeoxyglucose (FDG) can be labeled with tracer fluorine-18 (FDG-PET) and used to measure glucose consumption. In patients with AD, FDG-PET measurements correlate with continued decline in cognitive function (Weiner and Khachaturian, 2005) and co-registered MRI and FDG-PET images have equivalent classification values (Santi et al., 2001) . Further, FDG-PET has shown efficacy in distinguishing frontotemporal dementia from AD (Weiner and Khachaturian, 2005) thereby demonstrating its utility beyond other diagnostic tools.

While monitoring FDG retention has proven useful, it is a non-specific indicator of metabolism and can indicate variation in glucose use for a number of reasons beyond neurodegeneration (i.e., ischemia and inflammation). Thus, PET ligands were developed to determine amyloid plaque formation. However, Aβ levels rise and fall depending on the neurodegenerative stage and the extent of plaques does not correlate with defined neurodegenerative stages (McLean et al., 1999) suggesting other targets must be tried. More recently, ligands for monitoring neurofibrillary tangles and neuritic plaques have also been in development and as tau deposition is more strongly linked with cognition in AD may prove more useful in the long-term (Shoghi-Jadid et al., 2002). It is hoped that with increasingly specific ligands and improved sensitivities, PET will increase accuracy of differential diagnosis of neurodegeneration at early clinical stages.

Given that O-GlcNAc appears critical for protein stability and, in some studies, increased O-GlcNAc has been shown to decrease tau aggregation; modulating global brain O-GlcNAc levels has become an attractive therapeutic target. To note, while compounds specifically targeting OGT would be of great interest, given this enzyme’s importance in global nutrient-sensing, targeting OGA has been a more achievable goal. Further, targeting OGT has been challenging because developed analogs often also target and inhibit other transferases, are polar thereby limiting cellular availability, and have been associated with cellular toxicity (Trapannone and Rafie…, 2016). Recently published crystal structures of OGT will give improved insight toward developing ligands (Ma et al., 2013; Janetzko and Walker, 2014). OGA inhibitors and complementary PET ligands to monitor inhibitor efficacy and outcomes have been developed with some rapidity and their use will be explored in the following paragraphs.

Therapeutic OGA inhibitors have been seen in model organism work as well as pre-clinical studies. Use of OGA PET ligands in clinical studies will give an indication of bioavailability (including passage of the blood brain barrier), retention, and potentially the efficacy of therapeutic agents. Companies including Merck and Alectos are pursuing at least one compound related to OGA imaging (Sean M. Smith1, 2016). The chemistries to generate these compounds must be robust, straightforward, and require minimal purification as they must be generated in a nearby cyclotron close to the imaging facility and used while active. The nature of many of the inhibitors heretofore generated gives options for potential clinical imaging modalities, as they are amenable to medicinal chemistry. The production of OGA inhibitors, with varying specificity, has spanned at least two decades (Macauley et al., 2010; Kim et al., 2013). With the publication of improved crystallography data, informed decisions about inhibitor/probe synthesis will only improve options. Thiazoline derivatives have proved of particular interest given that they are orally available, cross the blood brain barrier, and show decreased tau oligomerization in one mouse model (Yuzwa et al., 2012). Whether the data from mouse models will be recapitulated in human patients and ultimately quality of life is improved remains to be seen, but PET ligands targeting OGA, OGT, or O-GlcNAc have the potential to help us to understand treatment regimens.

3. O-GlcNAcase inhibitors targeting neurodegeneration

With accumulating evidence suggesting that the regulation of O-GlcNAcylation is tightly linked to neurodegenerative diseases, pharmaceutical inhibitors of O-GlcNAc cycling have emerged as a potential therapeutic solution to neurodegeneration.

Recent studies on Tau

Human tau is highly phosphorylated with nearly 85 phosphorylation sites and reported to have at least 12 sites of O-GlcNAcylation (Arnold et al., 1996; Liu et al., 2004b). Tau phosphorylation is 2–3 fold higher and O-GlcNAc levels are reduced in AD compared to healthy samples (Alonso et al., 1996; Liu et al., 2004b), as might be expected given the reciprocity between these PTMs (Alonso et al., 1996; Lazarus et al., 2009; Yuzwa et al., 2012). These changes are associated with a self-aggregating, toxic molecule that is detrimental to neuronal microtubule function and neuronal death (Alonso et al., 1996). Furthermore, the conditional mouse brain Ogt KO showed increased tau phosphorylation (O’Donnell et al., 2004) suggesting that these PTMs play key roles in maintaining appropriate levels of soluble protein.

It is interesting to note that regulation of tau modifications is intimately linked to OGT: the kinases GSK3B and CDK5 are responsible for tau phosphorylation and are known OGT substrates (Lubas and Hanover, 2000). GSK3B and CDK5 are part of a complex, in which phosphorylation of tau by CDK5 can prime tau phosphorylation by GSK3B (Mietelska-Porowska et al., 2014). GSK3B can then phosphorylate about 30 sites on tau and its activity is increased in AD. Furthermore, OGT activity is increased by GSK3B mediated phosphorylation under physiological conditions. OGT could then O-GlcNAc modify and increase GSK3B activity, thereby generating a feedback loop by which tau protein solubility is modulated (Yu et al., 2012). While the biological role of O-GlcNAcylation on tau function is still unclear, it has been suggested that O-GlcNAc is necessary to prevent tau from aggregation in mice and in vitro (Yuzwa et al., 2012; Yuzwa and Vocadlo, 2017). How this intricate relationship between OGT, GSK3B and tau plays a role in tauopathies remains to be determined.

With studies supporting that neurodegenerative diseases display lower O-GlcNAcylation in the brain, attempts have been made to increase O-GlcNAc levels in mouse models of neurodegeneration. These studies seek possible use of OGA inhibitors for future therapeutics. In one promising study, researchers have treated a neurodegenerative mouse model with the OGA inhibitor Thiamet-G and found that treatment limits tau aggregation (Yuzwa et al., 2012). In a separate study, Thiamet-G treatment by injection into lateral ventricle in the brain led to a decrease in tau phosphorylation at residues Thr181, Thr212, Ser214, Ser262/Ser356, Ser404 and Ser409, and an increase in tau phosphorylation at Ser199, Ser202, Ser396 and Ser422 in the mouse brain (Yu et al., 2012). In addition, Thiamet-G mediated increase in O-GlcNAcylation also upregulated the GSK3B activity. Since GSK3B is known to phosphorylate tau, its activation by O-GlcNAcylation may explain the increased tau phosphorylation on certain sites (Yu et al., 2012). Long term treatment with OGA inhibitor led to an increased O-GlcNAc and reduced pathologic tau levels (Yu et al., 2012). Another study using the same OGA inhibitor treatment in a tau mutant mouse model (rTg4510) reduced pathological tau phosphorylation levels (Graham et al., 2014). OGA inhibitor was also used in a mouse model with mutations in tau and APP (TAPP) and lead to reductions in levels of both neuritic plaques and amyloidogenic β-amyloid peptides in the TAPP mice (Yuzwa et al., 2014). However, one should consider the possible side effects of systemic use of OGA inhibitor. As we discussed throughout this review and other published excellent reviews about the importance of physiological levels of O-GlcNAc in cellular and organismal biology, increasing O-GlcNAc could lead to pathological conditions in other tissues or organs.

Systemic effects of altered O-GlcNAc cycling: A cautionary Tale

There are several cautionary notes regarding pharmacological manipulation of O-GlcNAc levels. Specifically, acute and chronic changes to O-GlcNAc may yield different cellular outcomes. Acute OGA inhibition correlates with decreased protein phosphorylation but longer-term inhibition does not affect phosphorylation on proteins such as tau suggesting that cells adapt (Zhu et al., 2014). Also, while tau is among the proteins modified by O-GlcNAc, there are thousands of other proteins whose O-GlcNAc status may be modulated by therapeutics targeting of OGT or OGA resulting in unintended consequences. Furthermore, off-target side effects in the brain and other tissues as well as disrupted glucose homeostasis can lead to complications and must be carefully considered for any treatment regimen. Concerns were raised based on the potential insulin resistance induction following increased O-GlcNAcylation. Studies in C. elegans and Drosophila have demonstrated that changes in O-GlcNAc cycling can induce changes in insulin-sensitivity (Hanover et al., 2003; Forsythe et al., 2006; Sekine et al., 2010). Previous studies had also suggested that overexpression of OGT in mice (McClain et al., 2002) or inhibition of OGA in the L1-3T3 cell model (Wells et al., 2003) could induce insulin resistance. However, Macauley, et al. demonstrated that the whole body treatment using OGA inhibitor NButGT does not trigger insulin resistance or alter glucose homeostasis in rats or in adipocytes (Macauley et al., 2008; Macauley et al., 2010). This is presumably due to compensation by other homeostatic physiological mechanisms that cannot occur in genetic experiments or experiments carried out in vitro. Importantly, treatment of mice with NButGT reduced Aβ production both in vitro and in vivo. Low γ-secretase activity mediated by increase of O-GlcNAcylation is suggested as a potential explanation. Since O-GlcNAc is a key epigenetic regulator, alteration in the levels or activities of O-GlcNAc cycling enzymes could have unforeseen effects on transcriptional programs and potentially transgenerational effects (Lewis and Hanover, 2014).

Summary and Conclusions

Genetic, pharmacological, and biochemical evidence suggests that intracellular protein O-GlcNAcylation is emerging as a major pathway contributing to age-dependent neurodegenerative damage. Yet, research in this area remains in its infancy. A summary of the major topics of this review appears in Figure 4. Nutrient-driven O-GlcNAcylation is likely to be a key player in maintaining normal proteostasis in the brain. Intriguingly, mutations in Ogt have been linked to X-linked intellectual deficiency suggesting a major role for O-GlcNAc metabolism in the normal functioning of the brain (Niranjan et al., 2015; Vaidyanathan et al., 2017). The brain’s heavy dependence on glucose metabolism renders postmitotic neurons particularly susceptible to age-dependent changes in proteostasis. Maintenance of synapses, axonal transport, neural mitochondrial metabolism, and trafficking are regulated by O-GlcNAcylation. In addition, nearly all the cellular degradation pathways are either directly or indirectly dependent upon O-GlcNAcylation. The enzymes of O-GlcNAc metabolism are known to partner with chaperones, transcriptional complexes, and other epigenetic regulators. These factors dictate the levels of proteins prone to aggregation in neurodegenerative disease. O-GlcNAcylation also acts on globular domains of kinases and phosphorylated substrates to directly alter signaling cascades. OGT acts co-translationally and post-translationally to modify IDPs prone to aggregation.

Figure 4. A summary of the role of O-GlcNAc in regulating essential brain functions.

UDP-GlcNAc, the end product of the nutrient-sensitive HBP, is dynamically used by OGT to catalyze the addition of O-GlcNAc to substrate proteins. This cyclic modification is coordinately regulated with other PTMs such as phosphorylation to regulate the required intricacies of cellular processes. Deregulation of PTMs including O-GlcNAc leads to several pathologies that are associated with neurodegeneration. Methods for detecting O-GlcNAc and understanding its role in these pathologies continue to improve providing a strong foundation for considering targeted therapeutics based on individual or global protein O-GlcNAcylation and the activities of OGT and OGA.

The close association between O-GlcNAcylation and neurodegeneration provides unique opportunities for both diagnostics and therapeutics. Imaging paradigms based on monitoring O-GlcNAcylation or the enzymes modulating the modification’s addition and removal are currently in development. In addition, small molecule inhibitors of OGT and OGA may prove useful for therapy. These may act by directly preventing the formation of protein aggregates or by altering the signaling cascades leading to the onset of neurodegenerative damage. Yet, since O-GlcNAc metabolism is tightly regulated and ubiquitous, it may be difficult to target the pathway for inhibition using traditional pharmacological means. In addition, targeting therapies to the brain is complicated by the blood-brain barrier limiting the bioavailability of some therapeutic compounds. Finally, since all tissues are likely to exhibit a requirement for O-GlcNAc cycling, selective targeting of the brain may prove difficult. Such off-target effects may limit the usefulness of inhibitors that might otherwise prove useful. Despite these limitations, as one of the key cellular responses to stress and nutrient availability, O-GlcNAcylation represents a promising target for therapeutic strategies.

ABBREVIATIONS

O-GlcNAc

O-linked N-acetylglucosamine

GLUT1

Glucose Transporter type 1

GLUT4

Glucose Transporter type 4

AD

Alzheimer’s Disease

PD

Parkinson’s disease

ALS

Amyotrophic lateral sclerosis

HD

Huntington’s disease

APP

Amyloid precursor protein

NFT

Neurofibrillary tangles

PINK1

PTEN induced kinase 1

SCA

spino-cerebellar ataxia

Atx-1

Ataxin-1

IDPs

Intrinsically disordered proteins

HBP

Hexosamine biosynthetic pathway

UDP

Uridine Diphosphate

PTM

Post translational modification

OGT

O-GlcNAc Transferase

ncOGT

Nuclear and Cytoplasmic OGT

mOGT

Mitochondrial OGT

sOGT

Short isoform of OGT

OGA

O-GlcNacase

OGA-L

OGA Long isoform

OGA-S

OGA Short isoform

KO

Knockout

TPR

Tetratricopeptide Repeat

HCF-1

Host Cell Factor-1

GSK3B

Glycogen Synthase Kinase 3B

AMPK

AMP activated Protein Kinase

TET

Ten-eleven Translocation

CARM1

Coactivator Associated Arginine Methyltransferase 1

Atx-10

Ataxin-10

DYT3

Parkinsonian dystonia

THAP1

THAP Domain containing Protein 1

pRB

Retinoblastoma protein

E2F

Transcription Factor E2F1

NF-kB

Nuclear Factor kappa Beta

MIBP1

c-MYC intron binding protein

MGEA5

Meningioma-Expressed Antigen 5

HAT

Histone Acetyl Transferase

SKN-1

Skinhead 1

Nrf1/2

Nuclear respiratory factor 1/1

PRC

Polycomb repressor complex

Ph

Polyhomeotic

TRX

Trithorax

ASH1

Absent Small or Homeotic 1

SET1

SET domain containing protein 1

Syn1

Synapsin 1

αCamKII

α-calcium–calmodulin-dependent kinase II

MMTV-Cre

Cre Recombinase under mouse mammary tumor virus promoter

PVN

Paraventricular nucleus

IGF-1

Insulin-like growth factor 1

GH

Growth Hormone

XLID

X-linked intellectual disability

Xist

lnRNA triggering X-inactivation

IR

Insulin receptors

PI3K

Phosphatidylinositol 3 kinase

Akt/PKB

Protein Kinase B

mTOR

Mammalian Target of Rapamycin

PDK1

Phosphoinositide-Dependent Kinase 1

IRS1

Insulin Receptor Substrate 1

Sp1

Specificity Protein 1

Nup62

Nuclear pore protein p62

CNS

Central nervous system

RMS

Rostral migratory stream

SVZ

Subventricular zone

SGZ

Subgranular zone

CDK

Cyclin-dependent kinase

SNAP-29

Synaptosome associated protein 29

VAMP8

Vesicle associated membrane protein 8

BCL2

B-cell lymphoma 2

Atg7

Autophagy related 7

ATP

Adenosine triphosphate

TRAK1/2

Trafficking kinesin-binding protein 1

OIP106/98

O-linked N-acetylglucosamine transferase interacting protein 106

NDUFA9

NADH dehaydrogenase 1 aplha subunit A9

COX1

Cyclooxygenase 1

ETC

Electron transport chain

TCA

Tricarboxylic acid

ACO2

Aconitase 2

IDH

Isocitrate dehydrogenase

SOD

Superoxide dismutase

PRDX3

Peroxiredoxin 3

VDAC

Voltage dependent anion channel

mPTP

Mitochondrial permeability transition pore

Ogg1

8-oxoguanine DNA glycosylase

MRI

Magnetic resonance imaging

PET

Positron emission tomography

PTM

Post translational modification

GSK3B

Glycogen synthase kinase 3 beta

CDK5

Cell division protein kinase 5