Introduction

O-linked N-acetyl-β-d-glucosamine (O-GlcNAc) is a dynamic post-translational modification (PTM) of more than 3,000 nuclear, cytoplasmic, and mitochondrial proteins (Carapito et al. 2009; Gurcel et al. 2008; Rexach et al. 2008; Teo et al. 2010; Vosseller et al. 2006; Wang et al. 2010a, b; Wells et al. 2002b; Torres and Hart 1984). While O-GlcNAc appears more common in metazoans, there is evidence for this PTM in simple eukaryotes such as Aspergillus niger (Machida and Jigami 1994) and in some prokaryotes (Shen et al. 2006). O-GlcNAc is cycled on and off serine (Ser) and/or threonine (Thr) residues by just two enzymes: the O-GlcNAc transferase (OGT) that catalyzes the addition of O-GlcNAc and the O-GlcNAcase that removes O-GlcNAc (Fig. 1). Highlighting the importance of O-GlcNAc in cellular homeostasis, deletion of OGT, O-GlcNAcase, and other key enzymes in the hexosamine biosynthetic pathway (HBP) is lethal in mammals (Mio et al. 1999; Greig et al. 2007; Boehmelt et al. 2000a, b; Shafi et al. 2000; Forsythe et al. 2006; Yang et al. 2012).

Fig. 1
figure 1

The hexosamine biosynthetic pathway provides the donor substrate for O-GlcNAcylation and other types of protein glycosylation. Upon entry into the cell, a small percentage of Glucose (2–5 % in adipocytes) is directed into the hexosamine biosynthetic pathway and converted into UDP-GlcNAc via glutamine:fructose-6-phosphate amidotransferase (GFAT), glucosamine-6-phosphate acetyltransferase (EMeg32), N-acetylglucosamine-phosphate mutase (pgm3), and UDP-GlcNAc pyrophosphorylase (uap1). OGT catalyzes the addition of the glycoside to yield O-GlcNAcylated nuclear, cytoplasmic, and mitochondrial proteins, whereas the enzyme O-GlcNAcase catalyzes the removal O-GlcNAc from proteins. O-GlcNAc-modified proteins are involved in many cellular processes including the cellular stress response. HK hexokinase, G6P-I glucose 6-phosphate isomerase, Glc glucose, NAc N-acetyl

Full size image

Intracellular proteins that are O-GlcNAc-modified (O-GlcNAcylated) fall into diverse functional groups. Examples of O-GlcNAcylated proteins include chromatin-associated proteins and histones, transcription factors, ribosomal proteins, proteasomal proteins, cytoskeletal proteins, and many different types of signaling proteins such as kinases and metabolic enzymes (Carapito et al. 2009; Gurcel et al. 2008; Rexach et al. 2008; Teo et al. 2010; Vosseller et al. 2006; Wang et al. 2010a, b; Wells et al. 2002b). O-GlcNAc is thought to regulate these proteins in a manner analogous to protein phosphorylation. O-GlcNAc has been demonstrated to alter numerous protein functions that include: DNA binding and transactivation (Jackson and Tjian 1989; Ozcan et al. 2010; Comer and Hart 1999; Gao et al. 2003; Sayat et al. 2008), protein–protein interactions (Lim and Chang 2010; Ise et al. 2010; Guinez et al. 2010; Guinez et al. 2007; Guinez et al. 2006), protein degradation (Cheng and Hart 2001; Han and Kudlow 1997; Zhang et al. 2003), protein and enzyme activity (Kim et al. 2006b; Rengifo et al. 2007; Zhang et al. 2003; Dias et al. 2009; Bimboese et al. 2011), and protein localization (Sayat et al. 2008; Dudognon et al. 2004) among many others (Hart et al. 2011).

Like phosphorylation, the levels of O-GlcNAc respond to both intracellular and extracellular stimuli including insulin, nutrient levels, and cellular stress (Whelan et al. 2008, 2010; Walgren et al. 2003; Kearse and Hart 1991; Zachara et al. 2004b; Song et al. 2008; Hart et al. 2011), with the latter being the main focus of this review. In response to diverse forms of cellular stress, O-GlcNAc levels are elevated on numerous proteins (Zachara et al. 2004b). Augmenting O-GlcNAc levels promotes survival while suppressing O-GlcNAc levels sensitizes cells to death, suggesting that O-GlcNAc is a key regulator of the cellular stress response (Zachara et al. 2004b). While the exact mechanism(s) by which O-GlcNAc alters protein function leading to cell survival has not been defined, O-GlcNAc is known to regulate many signaling events and pathways, and this review will specifically focus on (1) the phosphoinositide 3-kinase (PI3K)/Akt pathway, (2) heat shock protein (HSP) expression, (3) calcium homeostasis, (4) reactive oxygen species (ROS) generation, (5) mitochondrial dynamics, (6) inflammation, and (7) the interplay with other PTMs. O-GlcNAc modulates these aspects of the cellular stress response to protect against ischemia reperfusion injury, trauma hemorrhage, and neurodegeneration. We will also discuss the interplay between O-GlcNAc and other post-translational modifications, future directions regarding novel technologies for detecting O-GlcNAc, and potential roles for O-GlcNAc in other disease mechanisms.

Biosynthesis of O-GlcNAc

Two enzymes modulate protein O-GlcNAcylation: UDP-N-acetyl-β-d-glucosamine/peptide N-acetylglucosaminyl-transferase (OGT), which catalyzes the addition of O-GlcNAc, and a neutral N-acetyl-β-glucosaminidase (O-GlcNAcase), which is responsible for the removal of O-GlcNAc (Fig. 1). The levels of O-GlcNAc appear to be regulated by five main events: (1) expression and activity of OGT, (2) expression and activity of O-GlcNAcase, (3) the concentration of UDP-GlcNAc, the sugar nucleotide donor of OGT, (4) the availability of protein substrates, and (5) targeting of the enzymes to their substrates.

The hexosamine biosynthetic pathway

The synthesis of UDP-GlcNAc occurs via the hexosamine biosynthesis pathway (HBP; Fig. 1). Upon entering cells, glucose is rapidly converted to glucose-6-phosphate by hexokinase (HK), and subsequently to fructose-6-phosphate by glucose-6-phosphate isomerase. The first step of the HBP is rate limiting and comprises of the conversion of fructose-6-phosphate to glucosamine-6-phosphate by glutamine/fructose-6-phosphate amidotransferase (GFAT). Notably, glutamine is required for this step and can be used to alter cellular O-GlcNAc levels as well as the levels of other sugar metabolites (Marshall et al. 1991). Several subsequent reactions result in the production of UDP-GlcNAc, the sugar nucleotide donor substrate used by OGT and other glycosyltransferases. The HBP accounts for ∼2–5 % of glucose flux in 3T3-L1 adipocytes (Marshall et al. 1991). These and other observations have led some to suggest that O-GlcNAc regulates cellular function in a glucose-dependent manner. Although little is known about the regulation of GFAT, it appears to be inhibited by high concentrations of UDP-GlcNAc (Traxinger and Marshall 1991). Interestingly, free radicals promote GFAT activity suggesting one mechanism by which cellular stress and high glucose can lead to upregulation of UDP-GlcNAc pools (Du et al. 2000). Mutations in genes encoding enzymes that significantly lower O-GlcNAc levels, such as emeg32 (glucosamine-6-phosphate acetyltransferase) and pgm3 (N-acetylglucosamine-phosphate mutase), are embryonic lethal highlighting the importance of O-GlcNAc and other forms of protein glycosylation that rely on UDP-GlcNAc (Boehmelt et al. 2000a, b; Greig et al. 2007).

O-GlcNAc transferase

Unlike many other glycosyltransferases, OGT is a soluble protein that is predominantly localized to the nucleus, mitochondria, and cytoplasm of all tissues studied thus far (Kreppel et al. 1997; Haltiwanger et al. 1992; Lubas et al. 1997; Love et al. 2003; Hanover et al. 2003). Two functional domains characterize OGT: an N-terminal tetratricopeptide repeat (TPR domain) and a C-terminal catalytic domain belonging to the glycogen phosphorylase superfamily (Kreppel et al. 1997; Lubas et al. 1997; Wrabl and Grishin 2001). Recently, the structure of OGT and the TPR domain have been solved (Lazarus et al. 2011; Martinez-Fleites et al. 2008; Jinek et al. 2004). The TPR domain forms an extended alpha-helix similar to importin-α (Jinek et al. 2004), and is important for mediating protein–protein interactions and enzyme activity (Kreppel and Hart 1999; Lubas and Hanover 2000). The catalytic domain contains two Rossman folds, separated by an intervening sequence. Notably, this linker is missing in the bacterial homologs of OGT that do not appear to modify protein substrates (Lazarus et al. 2011).

The gene for OGT maps to Xq13 on the mammalian X chromosome and encodes three well-characterized variants: short OGT, mitochondrial OGT (mOGT), and nuclear/cytoplasmic OGT (Kreppel et al. 1997; Lubas et al. 1997; Love et al. 2003; Hanover et al. 2003). mOGT has an alternative mitochondrial targeting sequence on its N-terminus. Interestingly, overexpression of the mOGT appears to induce apoptosis (Shin et al. 2011). OGT is essential for embryonic stem cell viability and somatic cell function (O’Donnell et al. 2004; Shafi et al. 2000). Deletion of OGT in Arabidopsis (Jacobsen et al. 1996), Drosophila (Sinclair et al. 2009; Gambetta et al. 2009), and mice is lethal (Shafi et al. 2000), and leads to dauer phenotypes in Caenorhabditis elegans (Hanover et al. 2005), suggesting that O-GlcNAc is essential for cellular homeostasis.

The mechanism of OGT substrate specificity is not fully understood; however, deletions in the TPR domain can affect the ability of OGT to O-GlcNAcylate protein and peptide substrates. It is generally thought that protein–protein interactions of OGT form a series of OGT complexes with different substrate specificities (Kreppel and Hart 1999). In support of this hypothesis, OGT has been found to associate with numerous proteins (Cheung et al. 2008) and is targeted to its substrates by p38 mitogen-activated protein kinase (MAPK) under conditions of nutritional deprivation (Cheung and Hart 2008). Moreover, OGT can bind phosphatidylinositol (3,4,5)-trisphosphate, which targets OGT to the plasma membrane during insulin signaling (Yang et al. 2008). These observations may also explain why there is no exact consensus sequence or motif for the addition of O-GlcNAc, although a recent paper has developed a neural network for predicting O-GlcNAc modification sites (Wang et al. 2011).

The activity of OGT appears to be regulated by substrate targeting (discussed above), the levels of UDP and UDP-GlcNAc, and potentially by other post-translational modifications. OGT is both tyrosine phosphorylated (Kreppel et al. 1997; Kreppel and Hart 1999) and O-GlcNAc-modified (Kreppel and Hart 1999; Lubas and Hanover 2000), although the consequence of these modifications has not been defined. Interestingly, OGT is phosphorylated and activated by active Ca2+/calmodulin-dependent kinase (CaMK) IV (Song et al. 2008). Recently, it has been demonstrated that OGT is S-nitrosylated in resting cells and its denitrosylation following induction of the innate immune response results in increased catalytic activity (Ryu and Do 2011). While UDP-GlcNAc levels may alter OGT activity by providing more substrate, there is also evidence that UDP-GlcNAc alters the substrate specificity of OGT. Thus, OGT may glycosylate one group of proteins at low concentrations of UDP-GlcNAc, but at higher concentrations may target a completely different panel of proteins (Kreppel and Hart 1999).

Little is known about how stress affects the activity of OGT, but stress alters glucose uptake and the activity of the HBP (discussed above) presumably augmenting UDP-GlcNAc pools. During heat stress, the activity of OGT is upregulated threefold by currently unknown mechanisms (Zachara et al. 2004b) and OGT translocates to the nucleus (Kazemi et al. 2010). Suggesting yet another mechanism by which stress alters O-GlcNAc levels, stressors such as sodium chloride, ethanol, and arsenite appear to alter the expression of OGT (Zachara et al. 2004b). During glucose deprivation, the expression of OGT is partially dependent on AMP-activated protein kinase, and OGT is targeted to its substrates by p38 MAPK (Cheung and Hart 2008). This results in a dramatic increase in O-GlcNAcylation of the neuronal protein neurofilament H, suggesting a possible mechanism by which defective glucose metabolism in the brain may directly contribute to the loss of axonal structure and stability (Cheung and Hart 2008).

Several assays have been used to detect OGT activity. The most common involves the addition of a tritiated UDP-GlcNAc (3H-UDP-GlcNAc) to a peptide substrate (Haltiwanger et al. 1997). Recently, an effective fluorescence-based substrate analogue displacement assay was developed to assess the activity of OGT (Gross et al. 2005). The spatiotemporal dynamics of O-GlcNAc have also been characterized using a series of genetically based O-GlcNAc FRET (Förster resonance energy transfer) sensors targeted to specific subcellular compartments (Carrillo et al. 2011; Carrillo et al. 2006). For example, a study utilizing the O-GlcNAc FRET sensor found that during serum-stimulated signal transduction rapid increases in O-GlcNAcylation were observed at the plasma membrane and in the nucleus, with a concomitant decrease in O-GlcNAcylation in the cytoplasm (Carrillo et al. 2011).

O-GlcNAcase

O-GlcNAcase is approximately 103 kDa in size and catalyzes the removal of O-GlcNAc from proteins. O-GlcNAcase appears to be predominantly cytoplasmic and like OGT is ubiquitously expressed (Dong and Hart 1994). The gene-encoding O-GlcNAcase was identified as meningioma-expressed antigen 5 (mgea5) and characterized as a hyaluronidase (Heckel et al. 1998). However, further analysis indicated that the product of the mgea5 gene was active at neutral pH, localized to the cytoplasm, and would remove O-GlcNAc from peptides suggesting that it was the O-GlcNAcase biochemically characterized by Dong and Hart (Comtesse et al. 2001; Gao et al. 2001). O-GlcNAcase exists as two isoforms: the full-length isoform with a histone acetyltransferase (HAT) domain in its C-terminus and a short isoform lacking the C-terminal HAT domain.

O-GlcNAcase is serine phosphorylated and O-GlcNAcylated, although the functional significance of these modifications has yet to be elucidated (Gao et al. 2001; Wells et al. 2002a). Like OGT, O-GlcNAcase from bovine brain appears to exist in complexes with other proteins including HSP110, HSC70, amphiphysin, and dihydropyriminidase-related protein-2. Finally, O-GlcNAcase is cleaved by caspase-3 during apoptosis (Wells et al. 2002a). Notably, the cleavage products appear to remain associated with each other and the activity of O-GlcNAcase is unaffected (Butkinaree et al. 2008). Thus far, little is known about the regulation of O-GlcNAcase during the cellular stress response.

While little is known about the regulation of O-GlcNAcase, the activity of O-GlcNAcase can be conveniently assayed in vitro with a synthetic substrate p-nitrophenol-β-GlcNAc (pNP-β-GlcNAc) (Dong and Hart 1994). However, care must be taken to exclude the activities of the lysosomal hexosaminidases (HexA and HexB) in total cell/tissue lysate. Recent studies have shown that fluorogenic substrates exhibit higher sensitivity and can be used to study both O-GlcNAcase isoforms. Both fluorescein di-β-GlcNAc (FDGlcNAc; Kim et al. 2006a) and 4-methylumbelliferyl-2-N-acetyl-2-deoxy-β-d-glucopyranoside have been used successfully (Macauley et al. 2005); however, the latter is tenfold less sensitive than FDGlcNAc (Kim et al. 2006a).

O-GlcNAc and the cellular stress response

In response to injury, cells have the ability to modify their metabolic, transcriptional, translational, and signaling pathways to promote survival, collectively this is known as the “cellular stress response” (Lindquist 1986; Nollen and Morimoto 2002). Recent data suggests that O-GlcNAc is one component of the cellular stress response that is relevant to a variety of models of injury in several cell and tissue types. O-GlcNAc levels become elevated in response to numerous forms of cell stress and tissue injury including: heat stress (Sohn et al. 2004; Zachara et al. 2004b), oxidative stress (Jones et al. 2008; Zachara et al. 2004b), ethanolic stress (Zachara et al. 2004b; Ngoh et al. 2009b), genotoxic stress (doxorubicin, belocin, and UVB irradiation; Zachara et al. 2004b, 2011a; Love et al. 2010), reductive stress (iodoacetamide; Zachara et al. 2004b), ER stress (dithiothreitol and tunicamycin; Zachara et al. 2004b; Ngoh et al. 2009b), hypoxia reoxygenation (Ngoh et al. 2009a; Ngoh et al. 2008; Ngoh et al. 2011), osmotic stress (NaCl, sorbitol, and sucrose; Zachara et al. 2004b; Zou et al. 2007), ATP depletion (sodium arsenite; Zachara et al. 2004b), ischemia reperfusion injury (Champattanachai et al. 2007, 2008; Fulop et al. 2007a, b; Hwang et al. 2010; Jones et al. 2008; Laczy et al. 2010; Liu et al. 2006, 2007; Nagy et al. 2006; Ngoh et al. 2008, 2009a, b, 2011; Pang et al. 2002; Zou et al. 2009), and trauma hemorrhage (Not et al. 2007, 2010; Yang et al. 2006a; Zou et al. 2007, 2009). This response occurs in primary and transformed cells, as well as in tissues in vivo and ex vivo. Suggesting that the addition of O-GlcNAc to proteins is part of a prosurvival signaling program, elevating O-GlcNAc levels prior (Zachara et al. 2004b; Chatham et al. 2008; Ngoh et al. 2008) or immediately following (Liu et al. 2007) cellular injury dramatically improves survival. Conversely, lowering O-GlcNAc levels sensitizes cells and tissues to injury (Zachara et al. 2004b; Ngoh et al. 2009a; Ngoh et al. 2008; Ngoh et al. 2011).

Modulating the levels of O-GlcNAc in a manner consistent with improved survival regulates a number of biological pathways. Some examples include: (1) increased protein solubility during heat stress (Cheung and Hart 2008; Lim and Chang 2006) or Alzheimer’s disease (Yuzwa et al. 2012); (2) reduced calcium overload (Liu et al. 2006, 2007); (3) decreased calpain activation (Liu et al. 2007); (4) altered p38 MAPK activation in response to ischemia reperfusion injury (Fulop et al. 2007b; Cheung and Hart 2008); (5) modulation of proinflammatory cytokines (Huang et al. 2007; Zou et al. 2009); (6) reduced mitochondrial permeability transition pore (mPTP) formation (Jones et al. 2008); (7) increased B-cell lymphoma 2 (Bcl-2) expression and translocation (Champattanachai et al. 2008); and (8) regulation of the expression of a subset of molecular chaperones (Zachara et al. 2004b; Kazemi et al. 2010; Sohn et al. 2004).

Numerous reports have suggested that O-GlcNAc regulates many processes in the heart and vasculature (Champattanachai et al. 2008; Chatham and Marchase 2010; Fulop et al. 2007a; Liu et al. 2007; Ngoh et al. 2011) in models of ischemia–reperfusion (I/R) injury. For example, O-GlcNAc appears to regulate many of the hallmarks of I/R injury such as calcium overload, oxidative damage, ER stress, and mitochondrial aberrations (Chatham and Marchase 2010). Elevating O-GlcNAc levels with glucosamine, PUGNAc (O-(2-acetamido-2-deoxy-d-glucopyranosylidene)amino-N-phenylcarbamate), or thiamet G (TMG, 2-ethylamino-3aR, 6S, 7R, 7aR-tetrahydro-5R-hydroxymethyl-5H-pyrano[3, 2-d]thiazole-6, 7-diol) leads to improved function and reduced tissue death in both ex vivo and in vivo models of I/R injury in the heart (Liu et al. 2007; Liu et al. 2006; Fulop et al. 2007b; Jones et al. 2008). Collectively, these data suggest that the stress-induced elevation of the O-GlcNAc modification is a pro-survival response of cells and tissues (Champattanachai et al. 2007, 2008; Hwang et al. 2010).

The PI3K/Akt pathway

The role of O-GlcNAc in the PI3K/Akt pathway has been extensively studied in the context of nutrient sensing, insulin resistance, and type II diabetes, and readers are directed to pertinent reviews on this topic for more extensive details (Akimoto et al. 2005; Slawson et al. 2010; Buse 2006; Hart et al. 2011). In contrast to these data, the interplay between O-GlcNAc and Akt signaling has been observed to inhibit apoptosis in a number of models including murine pancreatic β-cells (Kang et al. 2008), mouse embryonic fibroblasts (Kazemi et al. 2010), and murine liver (Ku et al. 2010), as discussed below.

Recently, it was demonstrated that O-GlcNAcylation of cytokeratin 18 (K18) is a regulator of Akt1 phosphorylation, which promotes downstream anti-apoptotic and cytoprotective signaling (Ku et al. 2010). Mouse mutants expressing human K18 S30/31/49A mutations, which cannot be glycosylated (K18–Gly−), were more susceptible to liver and pancreatic injury induced by treatment with either streptozotocin (STZ) or combined O-GlcNAcase inhibition and Fas administration (Ku et al. 2010). It is thought that the prosurvival mechanism is O-GlcNAcylation of K18, which leads to its interaction with Akt1 and keratin 8 (K8) ultimately promoting the phosphorylation of Akt1Thr308. Akt is activated by phosphorylation at Thr308 and then phosphorylates substrates such as forkhead box protein (Foxo) and glycogen synthase kinase 3β (GSK3β), thereby enhancing cell survival (Ku et al. 2010; Fig. 2). It is also important to note that Akt itself is dynamically modified by O-GlcNAc at Ser473 (Gandy et al. 2006; Kang et al. 2008). While hyper-O-GlcNAcylation of Akt has deleterious effects in hepatic cell and tissue models of euglycemia (Soesanto et al. 2008) and in pancreatic β-cells under hyperglycemic conditions (Kang et al. 2008), it is unclear if Akt is glycosylated in models of tissue injury.

Fig. 2
figure 2

O-GlcNAc is involved in the cellular stress response. Glucose and glucosamine can enter cells via glucose transporters and are directed towards the HBP for conversion to UDP-GlcNAc. O-GlcNAcylation of key proteins is essential for maintaining cellular homeostasis and promoting survival during stress and injury. Key survival pathways modulated by O-GlcNAc include: (1) the PI3K/Akt pathway, (2) heat shock protein expression, (3) calcium homeostasis, (4) reactive oxygen species regulation, (5) mitochondrial dynamics, (6) inflammation, and (7) PTM interplay. Note that proteins are shown as light blue squares, O-GlcNAc as green circles, and phosphates as orange circles. Arrows denote activation, whereas blunt ends denote inhibition. OGA O-GlcNAcase, OGT O-GlcNAc transferase

Full size image

Heat shock protein expression

One mechanism by which O-GlcNAc is thought to regulate cell survival is by modulating the expression of HSPs, the sentinels of the cellular stress response. Initial reports demonstrated that elevating O-GlcNAc levels augments the heat-induced expression of HSP72 and HSP40 (Zachara et al. 2004b; Sohn et al. 2004). Conversely, in cells in which O-GlcNAc levels had been lowered, by inhibition of the HBP or deletion of OGT, the expression of HSPs was suppressed (Zachara et al. 2004b; Kazemi et al. 2010). Using a real-time polymerase chain reaction (RT-PCR) array of 84 chaperones, Kazemi et al. demonstrated that the expression of 18 molecular chaperones was inhibited when O-GlcNAc levels were lowered by deletion of OGT (Kazemi et al. 2010). These chaperones included HSP72, HSP40 J-domain containing proteins (Dnaja1, Dnaja2, Dnaja3, Dnaja4, Dnajb1, and Dnajb7), the chaperone regulators Bag2 and Bag3, α-crystallin (Cryab, Hspb8), HSP25 (Hspb1), HSP110, HSP90α, and HSP60 (Kazemi et al. 2010). Interestingly, the expression of three proteins was upregulated (Dnajb8, Dnajc5b, and Dnajc5g; Kazemi et al. 2010).

The expression of HSPs is regulated by two key transcription factors: heat shock factor 1 (HSF-1) and Sp1. Several studies have demonstrated that O-GlcNAc regulates both of these transcription factors, directly and indirectly, leading to the regulation of chaperone expression. One example by which O-GlcNAc is thought to regulate HSP expression is by promoting the phosphorylation and inactivation of the kinase GSK3β at Ser9 (Kazemi et al. 2010). In cells with reduced O-GlcNAcylation, increased inhibitory phosphorylation of HSF-1 at Ser303 is observed (Kazemi et al. 2010). Conversely, elevating O-GlcNAc levels suppresses the phosphorylation of HSF-1 at Ser303, promoting both its activation and downstream HSP expression (Kazemi et al. 2010). Recent reports suggest that GSK3β may suppress HSF-1 activity by phosphorylating Ser/Thr residues independent of Ser303 (Batista-Nascimento et al. 2011). Nonetheless, consistent with a model in which increased GSK3β activity suppresses HSF-1 activation in the OGT null, inhibition of GSK3β rescues HSP72 expression in these cells (Kazemi et al. 2010). Further supporting this model, a reduction of Ser9 inhibitory phosphorylation of nuclear GSK3β was observed in cells with lower O-GlcNAc levels (Kazemi et al. 2010). Notably, unlike the study by Gandy et al. (2006), the stress-induced phosphorylation of Akt was unaffected in this model, demonstrating that O-GlcNAc can regulate numerous points upstream of GSK3β.

Glutamine has been demonstrated to protect cells and tissues from diverse forms of cellular stress and tissue injury (Wischmeyer et al. 2001). As glutamine is utilized by the HBP in the generation of UDP-GlcNAc (Fig. 1), it has been postulated that glutamine may work in an O-GlcNAc-dependent manner. In support of this hypothesis, in vivo treatment with glutamine has been shown to elevate O-GlcNAc levels (Singleton and Wischmeyer 2008; Hamiel et al. 2009). Moreover, the ability of glutamine to induce HSP70 is attenuated in cells in which O-GlcNAc levels have been lowered (Hamiel et al. 2009). In heat-stressed cells and a septic mouse model, glutamine treatment appears to promote the nuclear accumulation of HSF-1 and Sp1 in an O-GlcNAc-dependent manner (Singleton and Wischmeyer 2008; Hamiel et al. 2009). Although the mechanisms underlying the above observations remain unclear, it may be explained in part by the observations of Lim and Chang who demonstrated that increased levels of O-GlcNAc are associated with a more soluble pool of Sp1 during thermal stress (Lim and Chang 2006). This may represent altered localization of Sp1, or that Sp1 exhibits increased thermal stability when it is O-GlcNAc-modified (Lim and Chang 2006) similar to Tau (Yuzwa et al. 2012). Together, the data discussed above suggest that O-GlcNAc can regulate the expression of HSPs by regulating both HSF-1 and Sp1.

Calcium homeostasis

Another mechanism by which O-GlcNAc may promote cell survival is by regulating calcium handling. Angiotensin II, an IP3-generating agonist, is a potent stimulator of cardiomyocyte hypertrophy that can elevate cytoplasmic Ca2+ during vasoconstriction, increased blood pressure, and hemodynamic stress. A number of studies have demonstrated that angiotensin II treatment leads to an increase in basal Ca2+ levels in adult and neonatal cardiomyocytes (Kem et al. 1991; Shao et al. 1998; Hunton et al. 2004), with sustained exposure affecting numerous cellular pathways (Meldrum et al. 1996; Liu et al. 2004b; Zhang et al. 2004) and eventually resulting in cell death (Goldenberg et al. 2001; Kajstura et al. 1997). Elevating UDP-GlcNAc and O-GlcNAc levels through treatment with glucosamine or the O-GlcNAcase inhibitor PUGNAc attenuates angiotensin II-induced CCE during I/R injury (Nagy et al. 2006).

Further suggesting that regulation of calcium homeostasis in the heart is modulated by HBP flux and altered O-GlcNAc levels, O-GlcNAc appears to regulate a number of pathways downstream of calcium in models of ischemic injury. Consistent with decreased calcium influx, CaMKII phosphorylation is reduced, as are downstream events such as calpain proteolysis of structural proteins such as α-fodrin (Liu et al. 2007). Moreover, augmented O-GlcNAc signaling, via inhibition of O-GlcNAcase or overexpression of OGT, has been shown to attenuate oxidative stress and calcium overload in cardiomyocytes (Ngoh et al. 2011). Although the mechanism by which this occurs is unknown, it is accompanied by a reduction in ROS release and reduced formation of the mPTP (Ngoh et al. 2011).

While the mechanism by which O-GlcNAc regulates calcium handling in the ischemic heart has not been parsed out, recent studies have demonstrated that several IP3 receptors are O-GlcNAc-modified suggesting a potential molecular mechanism by which O-GlcNAc can modulate Ca2+ flux. The IP3 type I receptor is the principle intracellular calcium release channel in many cell types including neurons and it is regulated by calcium, ATP, and O-GlcNAc. O-GlcNAcylation of the IP3 type I receptor decreased its activity, while the removal of O-GlcNAc restored its function (Rengifo et al. 2007). Following this study, it was demonstrated that the IP3 type III receptor is functionally regulated by O-GlcNAc, except that O-GlcNAcylation enhanced its activity while O-GlcNAc removal negated its function (Bimboese et al. 2011). Thus, O-GlcNAcylation of IP3 type I and III receptors regulates their function(s) in controlling Ca2+ flux and ultimately attenuating Ca2+-dependent apoptosis (Rengifo et al. 2007).

Reactive oxygen species

ROS participate in many intracellular signaling pathways that can ultimately lead to alterations in gene transcription, protein synthesis, and overall cellular function. Stressors such as I/R injury, which induces mitochondrial Ca2+ overload (discussed previously) and ROS generation, favor the formation of the mPTP (discussed in the next section) and ultimately lead to cell death. Augmenting O-GlcNAc levels by adenoviral OGT overexpression or PUGNAc treatment has been shown to attenuate hypoxia and oxidative stress-induced ROS production. Conversely, suppression of O-GlcNAcylation by adenoviral O-GlcNAcase overexpression exacerbated ROS levels (Ngoh et al. 2011). One mechanism by which O-GlcNAc is thought to suppress ROS levels is by promoting the expression of antioxidant enzymes by stimulating the activity of Foxo1 and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α; Housley et al. 2009). For example, it was demonstrated that induction of PGC1α in response to ROS leads to its O-GlcNAcylation and activation by OGT. Subsequently, the OGT-PGC1α complex is targeted to and hyper-O-GlcNAcylates Foxo1 (Fig. 2) to promote increased transcription of ROS responsive enzymes, such as glutathione peroxidase 1 and superoxide dismutase 2, which protect against oxidative damage (St-Pierre et al. 2006; Housley et al. 2008, 2009). While the above studies were performed in hepatocytes, subsequent studies in a model of myocardial damage demonstrated that catalase mRNA levels were augmented above baseline following inhibition of O-GlcNAcase (PUGNAc), while overexpression of O-GlcNAcase reduced baseline catalase mRNA levels (Ngoh et al. 2011). Notably, in contrast to the studies discussed here, O-GlcNAc is thought to promote ROS generation in models of hyperglycemia and glucose toxicity (Goldberg et al. 2011), although the molecular basis for this paradox remains to be defined (Lima et al. 2012).

Mitochondrial dynamics

The mitochondrion is the powerhouse of the cell and is responsible for ATP production, maintenance of cellular byproducts, and overall regulation of cellular homeostasis. Altered mitochondrial dynamics leads to activation of the mitochondrial death pathway, which ultimately results in formation of the mPTP. Pore formation is activated by calcium surplus and ROS generation (previously discussed, Fig. 2), and is responsible for nonspecifically allowing small molecules (<1.5 kDa) to enter and exit the mitochondrial matrix (Crow et al. 2004). One postulated mechanism for the cytoprotective nature of O-GlcNAc is through the regulation of mitochondrial function (Ngoh et al. 2008; 2010, 2011; Jones et al. 2008) by reducing calcium overload, dampening ROS production, and preventing mPTP formation (Ngoh et al. 2008, 2011; Jones et al. 2008). It has been suggested that O-GlcNAc may regulate the mPTP directly by modifying and modulating the voltage-dependent anion channel (VDAC; Ngoh et al. 2008; Jones et al. 2008; Fig. 2). For example, deletion of OGT led to reduced O-GlcNAcylation of VDAC, sensitizing cells to mitochondrial membrane potential collapse and mPTP formation (Ngoh et al. 2008; Jones et al. 2008). Notably, formation of the mPTP is also promoted by active GSK3β (Juhaszova et al. 2004; Miura and Miki 2009), which is inhibited in some models by stress-induced O-GlcNAcylation (Kazemi et al. 2010; Ku et al. 2010).

In addition to modulating VDAC, O-GlcNAc has been demonstrated to modulate Bcl-2 and the electron transport chain. Mitochondrial localization of Bcl-2 has been postulated to inhibit mPTP opening (Tsujimoto et al. 2006; Tsujimoto 2003), protect against a loss in mitochondrial membrane potential (Tsujimoto et al. 2006; Tsujimoto 2003; Murphy et al. 2005), prevent cytochrome c release (Tsujimoto et al. 2006; Tsujimoto 2003; Murphy et al. 2005), and attenuate mitochondria-mediated apoptosis (Tsujimoto et al. 2006; Tsujimoto 2003; Murphy et al. 2005). One study demonstrated that elevation of O-GlcNAc levels in hydrogen peroxide-treated cells, an ex vivo model of ischemia–reperfusion injury, led to increased mitochondrial Bcl-2 translocation (Champattanachai et al. 2008). There is some debate surrounding this finding, as PUGNAc did not appear to alter I/R-induced increases in mitochondrial Bcl-2 levels (Champattanachai et al. 2008). Finally, elevating levels of glucose in neonatal cardiomyocytes resulted in augmented O-GlcNAc levels and subsequently reduced function of mitochondrial electron transport complexes I, III, and IV (Hu et al. 2005), however, its role in stress has not been studied yet. Interestingly, overexpressing O-GlcNAcase to reduce O-GlcNAc levels can reverse the effects of reduced mitochondrial function when glucose levels are elevated (Hu et al. 2005).

Inflammation

Trauma hemorrhage injury, like infection, induces an acute inflammatory response necessary for tissue repair. However, a hyperinflammatory response can compromise healthy tissue, leading to cell and tissue death (Nathan 2002; Jarrar et al. 1999). The inflammatory response is driven by secreted tissue-derived cytokines, such as tumor necrosis factor-α (TNF-α) and interleukin 6 (IL-6), that are responsible for inflammation in the damaged area. Recent studies have shown that augmenting O-GlcNAc levels prior to and following trauma hemorrhage injury improves survival, and this protective phenotype is associated with reduced expression of TNF-α and IL-6 and improved recovery of multiple tissues and blood pressure (Chatham et al. 2008; Not et al. 2007, 2010; Yang et al. 2006a).

O-GlcNAc has also been demonstrated to suppress acute inflammation in other models of injury presumably promoting cell survival (Xing et al. 2008; Pathak et al. 2012; Not et al. 2007). For example, in a rat model of arterial injury, elevating O-GlcNAc levels by administration of glucosamine or PUGNAc-attenuated inflammatory mediator expression, leukocyte infiltration, and neointima formation (Xing et al. 2008). Reduced inflammation was attributed to decreased TNF-α-induced phosphorylation of the nuclear factor κB (NFκB) p65 subunit and subsequent inhibition of NFκB transcriptional activation and signaling (Xing et al. 2008). Additionally, augmenting O-GlcNAc levels reduced the expression of a number of chemokines (cytokine-induced neutrophil chemoattractant-2β and monocyte chemotactic protein-1) and adhesion molecules (vascular cell adhesion molecule-1 and P-selectin; Xing et al. 2008).

Proinflammatory cytokines, such as TNF-α and interleukin 1 (IL-1), activate downstream signals in a manner mediated by TGF-β-activated kinase I (TAK1), also known as mitogen-activated protein kinase kinase kinase 7 (Sakurai et al. 2000; Wang et al. 2001). A recent study showed that O-GlcNAcylation of TAK1-binding protein (TAB1) modulates TAK1-mediated cytokine release when induced by IL-1 and osmotic stress (Pathak et al. 2012). Utilizing wild-type and O-GlcNAc-deficient mutant TAB1 (Ser395) in TAB1−/− mouse embryonic fibroblasts, this study demonstrated that the O-GlcNAc modification of TAB1 at Ser395 is essential for full TAK1 autophosphorylation at Thr187 and activation upon stress, as well as the subsequent downstream activation of NFκB and production of IL-6 and TNF-α (Pathak et al. 2012).

Interplay between O-GlcNAc and other PTMs

One mechanism by which O-GlcNAc is thought to regulate proteins is by directly or indirectly regulating other post-translational modifications including phosphorylation, nitrosylation, and ubiquitination.

Phosphorylation

Phosphate (PO4 3−) and GlcNAc (C8H15NO6) moieties are both covalently and reversibly added in an O-linkage to Ser/Thr residues in a manner known as phosphorylation and O-GlcNAcylation respectively. There is extensive crosstalk between these two modifications and this dynamic interplay is implicated in regulating numerous signaling networks (Wang et al. 2007, 2008, 2010b; Hart et al. 2011). For a subset of proteins (Cheng and Hart 2001; Kelly et al. 1993; Comer and Hart 2001; Du et al. 2001), O-GlcNAc and phosphorylation appear to compete for the same Ser/Thr residue, such as on the proto-oncogene c-myc (Chou et al. 1995). Alternatively, for numerous proteins O-GlcNAc can sterically hinder nearby phosphorylation sites (Dias et al. 2009; Yang et al. 2006b; Housley et al. 2008; Wang et al. 2010b; Tarrant et al. 2012), for example in the CK2 protein where O-GlcNAcylation of Ser347 prevents its phosphorylation at nearby amino acid Thr344 (Tarrant et al. 2012). The idea that these modifications are reciprocal comes from data demonstrating that phosphorylation of subsets of proteins is suppressed when global O-GlcNAc levels are elevated (Griffith and Schmitz 1999; Lefebvre et al. 2001; Wang et al. 2008). One elegant study demonstrated that in vitro O-GlcNAcylation of the C-terminal domain of RNA polymerase II blocked its phosphorylation and vice versa (Comer and Hart 2001). Suggesting that this interplay also alters protein function in vivo, additional studies have demonstrated that tau is hyper-phosphorylated in a brain OGT knockout (O’Donnell et al. 2004). However, readers should be cautioned as the reciprocal relationship between O-GlcNAc and phosphorylation exists only for a subset of proteins; while for others, there is no direct association and proteins can be both phosphorylated and O-GlcNAcylated simultaneously, such as insulin receptor substrate-1 (Ball et al. 2006).

In addition to cross-talk associated with site occupancy, there is cross talk between the enzymes responsible for O-GlcNAcylation and phosphorylation. OGT and O-GlcNAcase are known to associate with and modify kinase and phosphatase enzymes (Khidekel et al. 2004; Wang et al. 2010b; Wells et al. 2004; Slawson et al. 2008; Dias et al. 2009, 2012; Cheung et al. 2008). For example, CaMKIV is inhibited by O-GlcNAc modification within its ATP-binding region, and de-O-GlcNAcylation and subsequent phosphorylation is required for CaMKIV activation (Dias et al. 2009). Several proteomic analyses have been performed to further delineate the interplay between O-GlcNAc and phosphorylation in cell cycle progression. One of these studies demonstrated that the kinase Aurora B, which regulates mitotic progression, forms a transient complex with OGT and O-GlcNAcase in M phase of the cell cycle and colocalizes with OGT to midbodies during telophase and cytokinesis (Slawson et al. 2008). Inhibition of Aurora B leads to diminished midbody localization of OGT and an increase in global O-GlcNAc levels (Slawson et al. 2008). This study also demonstrated that vimentin, a substrate for Aurora B and OGT during mitosis, exhibited alterations in its O-GlcNAcylation and phosphorylation patterns upon modulation of OGT or O-GlcNAcase activity (Slawson et al. 2008). A second study identified 141 previously unreported O-GlcNAc sites on proteins that are involved in spindle assembly and cytokinesis, many of which are the same site as or are in close proximity to sites of phosphorylation (Wang et al. 2010b). Finally, modulation of O-GlcNAc levels was found to alter the phosphorylation status of proteins localized to the mitotic spindle and midbody, for example upon OGT overexpression, cyclin-dependent kinase 1 becomes hyperphosphorylated thereby reducing its activity and subsequently decreasing the phosphorylation status of its target proteins (Wang et al. 2010b).

Endothelial nitric oxide synthase (eNOS) generates nitric oxide and stimulates vasodilation by inhibiting smooth muscle contraction and platelet aggregation, with these functions reported to be impaired in diabetes-associated erectile dysfunction. One study demonstrated that in rats treated with alloxan to induce diabetes, eNOS was hyper-O-GlcNAcylated and hypo-phosphorylated at Ser1177 (Musicki et al. 2005). Phosphorylation of eNOS at this site is mediated by Akt phosphorylation and this study demonstrated that the phosphorylation of both Akt and eNOS was decreased in the diabetic model, implicating that O-GlcNAcylation at Ser1177 is one potential mechanism for eNOS inactivation and thereby diminished vasodilation (Musicki et al. 2005).

Other PTM roles

There is emerging evidence for the interplay of O-GlcNAc with other post-translational modifications such as protein ubiquitination and nitrosylation, although it is important to note that the exact mechanisms by which these modifications interact are still being elucidated.

A growing body of evidence suggests that O-GlcNAc may be a regulator of protein degradation. This includes data demonstrating that enhanced O-GlcNAcylation is associated with an extended half-life of proteins, such as with Sp1 (Han and Kudlow 1997) and the murine β-estrogen receptor (Jiang and Hart 1997; Cheng and Hart 2001). At least three mechanisms have been postulated to underlie these observations. Firstly, O-GlcNAc can block the phosphorylation of Pro-Glu-Ser-Thr domains in the murine β-estrogen receptor thereby preventing its degradation (Cheng et al. 2000; Cheng and Hart 2001). Secondly, as discussed below, data suggests that O-GlcNAc may regulate ubiquitination. Thirdly, O-GlcNAc is implicated in regulating the proteasome directly, as highlighted by studies showing that that numerous proteins in the proteasome are O-GlcNAc-modified and that enhanced O-GlcNAcylation may inhibit proteasome function (Zhang et al. 2003; Sumegi et al. 2003).

One study demonstrated that both ubiquitination and O-GlcNAcylation increase upon heat-shock treatment and that these PTMs can occur on some proteins concomitantly (Guinez et al. 2008). O-GlcNAc-modified proteins were not observed to be stabilized following proteasome inhibition, and ubiquitination could be enhanced or reduced by increasing or decreasing O-GlcNAc levels respectively (Guinez et al. 2008). It was also speculated that an E1 ubiquitin-activating enzyme interacts with Hsp70 only in its O-GlcNAcylated form (Guinez et al. 2008). An agonistic relationship between these two modifications is suggested such that O-GlcNAc may alter the activity of E1 enzymes to modulate stress-induced ubiquitination (Shrikhande et al. 2010; Shimura et al. 2001; Fujiki et al. 2011). It has also been shown that O-GlcNAcylation of a protein may facilitate its subsequent ubiquitination. For example, O-GlcNAcylation of histone H2B at Ser112 facilitates monoubiquitination at Lys120 to regulate transcription (Fujiki et al. 2011).

S-nitrosylation is a cysteine modification that is found and maintained on OGT in resting cells and S-nitrosylated OGT has greatly reduced catalytic activity compared to the native form of the protein (Ryu and Do 2011). OGT becomes denitrosylated in macrophages treated with lipopolysacharides to trigger an innate immune response, which leads to increased catalytic activity and hyper-O-GlcNAcylation of proteins (Haberhausen et al. 1995). While other PTMs such as acetylation, methylation, sumoylation, and many others have pivotal roles in regulating protein function and signal transduction, the interplay of these modifications with O-GlcNAcylation has yet to be defined.

Molecular mechanisms of disease

There are many facets of the cellular stress response whose misregulation can contribute to neurodegenerative disease pathophysiology, and O-GlcNAc is implicated in the etiology of some of these diseases. Notably, the OGT gene is on the X chromosome at position Xq13.1, the same locus associated with dystonia–parkinsonism syndrome (Haberhausen et al. 1995), and O-GlcNAcase maps to chromosome 10 near the locus associated with late-onset Alzheimer’s disease (10q24.1–q24.3; Bertram et al. 2000).

Alzheimer’s disease

Alzheimer’s disease (AD) is a form of dementia that is characterized by reduced brain function and cognitive deficits, where both genetic and environmental factors play a role in pathogenesis (Hardy and Allsop 1991). AD phenotypes are often the result of unfolded or aggregated proteins, aberrant signaling, and oxidative stress. While there are many mechanisms by which these events are regulated, the most well-studied proteins involved in AD are amyloid β precursor protein (APP), amyloid β (Aβ), and tau (Hardy and Allsop 1991). Misregulation of APP, Aβ plaque formation, tau phosphorylation, and tangle formation dominate the cascade of events leading to AD and ultimately to neuronal death (Dong et al. 2004).

AD transgenic mouse models are characterized by Aβ plaque formation and this leads to decreased cell proliferation and defective contextual memory (Kang et al. 2007). APP, the precursor protein to Aβ, is O-GlcNAc-modified although the importance of this modification remained elusive for many years (Jacobsen and Iverfeldt 2011). It was later demonstrated that inhibiting O-GlcNAcase with PUGNAc or reducing the expression of O-GlcNAcase (siRNA) increased the amount of APP O-GlcNAcylation and non-amyloidogenic α-secretase processing (Jacobsen and Iverfeldt 2011). This has further downstream effects such as increasing levels of the neuroprotective sAPPα fragment, which reduces Aβ secretion and suggests a protective role for O-GlcNAc (Kim et al. 2012). Finally, treatment with the O-GlcNAcase inhibitor NButGT in a mouse model of AD resulted in reduced γ-secretase activity and subsequent attenuation of Aβ plaque production and inflammation in vivo, further highlighting the important protective role of O-GlcNAc under conditions of stress and neurodegeneration (Rissman et al. 2007, 2012).

Tau was demonstrated to be directly O-GlcNAc-modified (Liu et al. 2004a; Arnold et al. 1996) and it was later shown that O-GlcNAcylation negatively regulates site-specific phosphorylation of tau in the human brain (Liu et al. 2004a). Notably, increased phosphorylation of Tau has been implicated in the formation of the toxic tangles associated with neurogeneration. Recent data have demonstrated that inhibition of O-GlcNAcase, which leads to an increase in O-GlcNAc levels, in a murine model of AD led to increased O-GlcNAcylation of tau, decreased tau aggregation, and diminished neuronal cell death. Together, these data highlight the protective role of O-GlcNAc and suggest that O-GlcNAcase could be a therapeutic target for slowing AD progression (Yuzwa et al. 2012). Interestingly, low glucose uptake and metabolism during aging has been postulated to lead to decreased O-GlcNAcylation of key proteins such as Tau, potentially exacerbating the AD phenotype during aging (Yuzwa et al. 2012).

Parkinson’s disease

Parkinson’s disease (PD) is a brain disorder characterized by the degeneration of midbrain dopaminergic neurons, which is often accompanied by the formation of Lewy bodies (abnormal protein aggregates; Henchcliffe and Beal 2008). Together, this leads to improper signaling and loss of muscle function. Genetic alterations in the nucleus have been linked to disrupted mitochondrial morphology and function that can then modulate protein activity, trigger apoptosis, or lead to accumulation of misfolded or damaged proteins in PD pathogenesis (Henchcliffe and Beal 2008).

Aggregation of the α-synuclein protein or mutations in the α-synuclein gene (A53T) which lead to an impairment of mitochondrial dynamics, morphology, and movement have been linked to rare inherited forms of PD (Polymeropoulos et al. 1997; Spillantini et al. 1997; Xie and Chung 2012). One study identified a 22 kDa O-glycosylated form of α-synuclein in a protein complex with parkin (an E3 ubiquitin ligase) and UbcH7 (an E2 ubiquitin-conjugating enzyme). These studies suggested that the glycosylated form of α-synuclein was bound and ubiquitinated by parkin (Shimura et al. 2001). On the contrary, a subsequent study postulated that α-synuclein is modified by O-GlcNAc and that its nonglycosylated form interacts with parkin in embryonic hippocampal cells, although the specific type of glycosylation was never confirmed (Kim et al. 2003). Recently, the O-GlcNAc modification status of α-synuclein was mapped to Thr72 (Wang et al. 2010a), which has been demonstrated to prevent its aggregation (Marotta et al. 2012). Further data (discussed previously) highlights the interplay between O-GlcNAcylation and ubiquitination to regulate protein stability and degradation, although a direct link between these phenomena and PD has yet to be elucidated.

Methods to study O-GlcNAcomics

In this section, we will discuss the current methodologies used to modulate O-GlcNAc levels, enrich O-GlcNAc-modified proteins, and identify sites of O-GlcNAcylation.

In vitro and in vivo modulation of O-GlcNAc levels

O-GlcNAc levels can be directly or indirectly altered in vitro and in vivo by modulating the activity or expression of OGT (Gross et al. 2005, 2011; Dorfmueller et al. 2011), O-GlcNAcase (Dong and Hart 1994; Haltiwanger et al. 1998; Liu et al. 2002; Dorfmueller et al. 2006; Whitworth et al. 2007; Macauley et al. 2008; Knapp et al. 2007; Beer et al. 1990; Yuzwa et al. 2008; Konrad et al. 2002; Laczy et al. 2010), or other key enzymes in the HBP (Buse et al. 2002; McClain 2002; Patti et al. 1999; Ross et al. 2000; Marshall et al. 1991; Traxinger and Marshall 1991). Techniques such as over expression of OGT (Zachara et al. 2004b), O-GlcNAcase (Slawson et al. 2005), and GFAT (Chen et al. 1997; Marshall et al. 2004; James et al. 2000) by both transient transfection and viral transduction have been successful, as has reducing the expression of these enzymes using RNA interference (Zachara et al. 2004b; Ngoh et al. 2009a; Hsieh et al. 2012). OGT has also been overexpressed using a tetracycline-inducible OGT stably expressed in HeLa cells, although for effective overexpression a histone deacetylase inhibitor was included with the tetracycline for appropriate regulation of OGT (Marshall et al. 2003).

Murine models and the OGTF/Y,mER-Cre-2A-GFP cell line

Several mouse models and cell lines exist in which the levels of O-GlcNAc can be modulated. Notably, in murine models deletion of OGT (Shafi et al. 2000; O’Donnell et al. 2004), EMeg32 (Boehmelt et al. 2000b), or pgm3 (Greig et al. 2007) leads to embryonic lethality, whereas deletion of O-GlcNAcase leads to perinatal lethality (Yang et al. 2012). A number of hypomorphic alleles of pgm3 have been characterized, leading to cells and mice with different concentrations of UDP-GlcNAc (Greig et al. 2007). Unlike deletion of EMeg32, deletion of OGT is lethal in isolated embryonic fibroblasts (O’Donnell et al. 2004). To overcome this challenge, researchers have immortalized cells in which the first two exons of OGT are flanked by loxP recombination sites (O’Donnell et al. 2004; Kazemi et al. 2010). These cells (OGTF/Y) have been stably transfected with a Cre recombinase estrogen receptor chimera (Cre-ERT2-GFP) that is induced by 4-hydroxytamoxifen (4-OHT; (OGTF/Y,mER-Cre-2A-GFP); Kazemi et al. 2010). Addition of 4-OHT activates Cre recombinase leading to the deletion of OGT.

Modulating O-GlcNAc levels using inhibitors of OGT, O-GlcNAcase, and the HBP

A large number of inhibitors have been developed for blocking the activity of O-GlcNAcase in both cell culture and animal models (Dong and Hart 1994; Haltiwanger et al. 1998; Liu et al. 2002; Dorfmueller et al. 2006; Whitworth et al. 2007; Macauley et al. 2008; Knapp et al. 2007; Beer et al. 1990; Yuzwa et al. 2008; Konrad et al. 2002), and here we will focus on those that are commercially available. PUGNAc was the first O-GlcNAcase inhibitor isolated and inhibits O-GlcNAcase at a K i of 52 nM in vitro (Haltiwanger et al. 1998; Dong and Hart 1994). PUGNAc is suitable for use in cells (10–100 μM, 4–18 h) and in animals (50 mg/kg, 4–12 h) (Jones et al. 2008); however, prolonged use (>36 h) can lead to cell cycle defects (Slawson et al. 2005). While widely used, recent evidence demonstrates that PUGNAc can also inhibit other lysosomal glycosidases including HexA and HexB (Macauley et al. 2005; Ficko-Blean et al. 2008), suggesting that PUGNAc may have effects on other cellular pathways aside from O-GlcNAcylation. STZ has also been widely used, but only inhibits O-GlcNAcase at very high concentrations (Toleman et al. 2006; Gao et al. 2000; Liu et al. 2002; Okuyama and Yachi 2001; Roos et al. 1998). It is important to note that STZ is also a DNA-alkylating agent and can release nitric oxide, making it difficult to determine whether the observed effects are due to O-GlcNAcase inhibition or STZ toxicity (Kwon et al. 1994). Alloxan has been suggested to inhibit O-GlcNAcase (Lee et al. 2006), but has been reported to also inhibit OGT and as such should be used with extreme caution (Konrad et al. 2002; Macauley et al. 2005). More recently, TMG has been introduced, and unlike PUGNAc, TMG is effective in crossing the blood–brain barrier and does not appear to inhibit HexA or HexB (Yuzwa et al. 2008).

Several OGT inhibitors have also been isolated that work well in vitro (Gross et al. 2005), and thus far these have been useful in isolated neonatal cardiomyocytes (5 μM; Ngoh et al. 2008) and breast cancer cells (200 μM; Caldwell et al. 2010). A more recent study has focused on synthesizing novel OGT inhibitors with structural analogues to UDP-GlcNAc/UDP (Dorfmueller et al. 2011). While the synthesized compounds inhibit OGT in the micromolar range, it should be noted that they are inactive when used on living cells (Dorfmueller et al. 2011). Another recent study developed a synthetic carbohydrate precursor, 2-acetamido-1,3,4,6-tetra-O-acetyl-2-deoxy-5-thio-α-d-glucopyranose, which can be successfully converted by the cell to produce UDP-5SGlcNAc in cell culture and in vitro models (Gloster et al. 2011). UDP-5SGlcNAc is not an efficient substrate for OGT, but excludes binding of UDP-GlcNAc therefore inhibiting OGT function (Gloster et al. 2011). It is important to note that these inhibitors are not yet commercially available.

There are numerous commercially available chemicals for supressing flux through the HBP (Buse et al. 2002; McClain 2002; Patti et al. 1999; Ngoh et al. 2010; Ross et al. 2000; Marshall et al. 1991; Traxinger and Marshall 1991), which reduced UDP-GlcNAc levels and thus the O-GlcNAc modification. Treatment with exogenous glucose, glutamine, or glucosamine leads to increased O-GlcNAc levels by increasing flux through the HBP (Traxinger and Marshall 1991), although it is important to note that these compounds enter numerous pathways in the cell and observed effects may not be due to changes in O-GlcNAcylation. Azaserine and DON (6-diazo-5-oxo-l-norleucine) are GFAT inhibitors, and treatment with these compounds decreases flux through the HBP thereby reducing O-GlcNAc levels (Marshall et al. 1991). Despite their common usage, both of these GFAT inhibitors do not act specifically on the HBP and may have effects on other enzymes and cellular pathways, especially those involved in glutamine metabolism.

Isolation, purification, and detection of O-GlcNAc and O-GlcNAc-modified proteins

Our understanding of the role of O-GlcNAc has been hampered due to the lack of specific and sensitive tools. Many challenges arise as the most commonly used analytical protein techniques, including gel electrophoresis (GE), liquid chromatography (LC), and mass spectrometry (MS), often fail to detect O-GlcNAc. This is due to a number of factors such as (1) the addition of a monosaccharide of O-GlcNAc does not generally affect the migration of the glycoprotein during either 1D- or 2D-GE; (2) the O-GlcNAc modification is rapidly hydrolyzed by hexosaminidases during protein isolation unless appropriate inhibitors are included; (3) the β-O-glycosidic linkage is chemically labile and is rapidly released under conditions of mild acid or base; (4) due to its lability, O-GlcNAc is usually lost at the ion source of electrospray ionization (ESI) or during ionization by matrix-assisted laser desorption ionization (MALDI) mass spectrometry; and (5) ionization of peptides modified by O-GlcNAc is suppressed during mass spectrometry and as such need to be enriched. Here, we will review the most robust techniques for the detection and purification of O-GlcNAc-modified proteins, as well as techniques for defining sites of O-GlcNAcylation.

Detecting O-GlcNAc-modified proteins by [3H]-galactose labeling

The gold standard for the detection of O-GlcNAc-modified proteins has been UDP-Gal:GlcNAc β-1,4-galactosyltransferase (GalT1) labeling, which transfers 3H-galactose from UDP-[3H]-Gal to any terminal GlcNAc residue (Torres and Hart 1984; Roquemore et al. 1994). One advantage of this technique is the generation of a radiolabeled sugar (3H-Gal-R) that can be followed in subsequent analyses. As GalT1 will transfer galactose to any terminal GlcNAc reside, it is important to treat samples with peptide N-glycosidase F (PNGase F) to enzymatically remove N-glycans prior to 3H-galactose labeling. Recently, this technique has also been adapted to incorporate an unnatural N-azidoacetylgalactosamine (GalNAz) onto proteins (see section Click-iT).

Immunopurification with antibodies and lectins

In contrast to protein phosphorylation where a range of pan- and site-specific phospho-antibodies are commercially available, only a few such tools exist for the O-GlcNAc modification. These antibodies fall into two classes: those that can be considered pan-specific and others that are site specific (Zachara et al. 2011c). The two most commonly used pan-specific antibodies for detecting O-GlcNAc-modified proteins are CTD110.6 (Comer and Hart 2001), a mouse IgM antibody raised against the C-terminus of RNA Pol II, and RL2 (Snow et al. 1987), a mouse IgG antibody raised against O-GlcNAc-modified components of the nuclear pore complex. Recently, a number of monoclonal IgG antibodies have been introduced for detecting O-GlcNAc-modified proteins including 1F5.D6(14), 18B10.C7(3), and 9D1.E3(10) (Teo et al. 2010). These antibodies were raised against a three-component immunogen-containing epitope regions from the casein kinase α-subunit 15, a mouse major histocompatibility complex class II-restricted helper T-cell epitope, and a Toll-like receptor 2 agonist (Teo et al. 2010). Like RL2, these antibodies are mouse IgGs and appear to only recognize O-GlcNAc in certain three-dimensional environments. In addition to the antibodies discussed previously, antistreptococcal monoclonal mouse IgG antibodies have been commonly used for detecting proteins containing GlcNAc residues. These antibodies were raised against streptococcal group A carbohydrates, which is composed of a polyrhamnose backbone with GlcNAc side chains (Turner et al. 1990).

There has only been a small cohort of studies that have incorporated a top-down shotgun proteomics approach for identifying O-GlcNAc-modified proteins using the antibodies previously discussed. These studies have been hampered because none of the discussed antibodies appear to bind O-GlcNAc-modified peptides. As such, proteins that interact with O-GlcNAc-modified proteins are also identified, and in most cases O-GlcNAc-modified peptides are suppressed in the MS making O-GlcNAc modification sites difficult to map. Nonetheless, a number of studies have been performed highlighting proteins that are potentially O-GlcNAc-modified in response to cellular stress and tissue injury (Jones et al. 2008; Teo et al. 2010; Zachara et al. 2011c), and the O-GlcNAcylation status of a number of proteins has been confirmed by independent techniques (Table 1). One recent study used 1F5.D6(14), 18B10.C7(3), and 9D1.E3(10) to immunoprecipitate O-GlcNAc-modified proteins followed by large-scale shotgun proteomics to identify more than 200 differentially expressed glycoproteins from HEK293 cells and rat livers responding to trauma hemorrhage and resuscitation (Teo et al. 2010). Another recent study relied on Stable Isotope Labeling with Amino acids in Cell culture (SILAC; Ong et al. 2002; 2003) in combination with CTD110.6 immunoprecipitation (Zachara et al. 2011a). Here, Cos-7 cells were isotopically labeled in SILAC media (13C6 l-arginine and 13C6 15N4 l-arginine) and treated with heat shock or an inhibitor of O-GlcNAcase, PUGNAc (Zachara et al. 2011a). Labeled proteins were combined in equal ratios and O-GlcNAc-modified proteins were immunoprecipitated using CTD110.6 immobilized to agarose and subsequently identified by LC-MS/MS (Zachara et al. 2011a). Numerous proteins with diverse functions were identified, including nuclear factor 90, RuvB-like 1, RuvB-like 2, and several COPII vesicle transport proteins. Many of these proteins bind double-stranded DNA-dependent protein kinase (DNA PK) or double-stranded DNA breaks, suggesting a role for O-GlcNAc in regulating DNA damage signaling or repair (Zachara et al. 2011a).

Table 1 A summary of O-GlcNAc-modified proteins, their biological significance, and the methods by which they were detected. This table highlights the wide range of techniques and tools for the enrichment and detection of O-GlcNAc-modified proteins, including but not limited to O-GlcNAc antibodies, click-iT methods, immunoprecipitation, and proteomics tools
Full size table

Lectin affinity chromatography

Wheat germ agglutinin (WGA) is a commonly used plant lectin for enriching and detecting O-GlcNAc-modified proteins (Zachara et al. 2004a, 2011b; Vosseller et al. 2006). WGA has weak affinity for single GlcNAc residues, but its affinity is dramatically increased for GlcNAcs that are clustered (Finlay et al. 1987; Lundquist and Toone 2002; Lee and Lee 2000). The drawback of WGA as a tool for O-GlcNAc enrichment and detection is its additional recognition of sialic acid (NeuAcα(2–3)) residues (Monsigny et al. 1980). Treating samples initially with PNGase F to remove N-glycans, and by performing nuclear and cytoplasmic protein extractions (Zachara et al. 2011b) can resolve these shortcomings. Succinylation of WGA (sWGA) increases its specificity for GlcNAc as it ablates its reactivity to sialic acid (Monsigny et al. 1980). However, succinylation reduces the affinity of sWGA for O-GlcNAc and thereby decreases its utility for immunoprecipitation.

Lectin weak affinity chromatography (LWAC) utilizes the affinity of WGA in an extended column format (>3 m in length) with a small diameter (∼1 cm) to enrich O-GlcNAc-modified peptides (Vosseller et al. 2006). This chromatography technique employs low flow rates and an isocratic elution buffer containing low concentrations of GlcNAc (Vosseller et al. 2006). LWAC has been adapted successfully for shotgun characterization of O-GlcNAc-modified proteomes from postsynaptic density preparations (Vosseller et al. 2006), Arabidposis thaliana (Xu et al. 2012), and mouse embryonic stem cells (Myers et al. 2011).

Click-iT and other chemically based strategies to detect O-GlcNAc-modified proteins

Click chemistry is a chemoenzymatic strategy for quickly and reliably synthesizing compounds of interest. By this method, target substrates are labeled with a probe of interest that can be derivatized to detect or enrich proteins and peptides. This technique has been adapted for the detection and enrichment of O-GlcNAc-modified proteins and peptides. Sugars containing either ketones or azido groups are incorporated into O-GlcNAc-modified proteins using one of two techniques: (1) a mutant GalT1 (Y289L) with an enlarged active site is used to add UDP-GalNAz to terminal GlcNAc residues similar to the GalT1 labeling discussed above (Khidekel et al. 2003); and (2) in cells by metabolic labeling with a peracetylated azido-GlcNAc substrate (Vocadlo et al. 2003; Sprung et al. 2005), as many of the enzymes in the HBP will tolerate unnatural sugars. However, the kinetics of O-GlcNAz removal from proteins by O-GlcNAcase is significantly reduced. These unnatural sugars can be derivatized further to incorporate a biotin tag, which facilitates the detection and enrichment of O-GlcNAc-modified proteins and peptides (Wang et al. 2010a). Alternatively, this method has been used to incorporated polyethylene glycol moieties (Rexach et al. 2010), which results in a molecular weight shift (∼2 kDa) for each GlcNAc residue found on a protein and thus makes it possible to estimate the stoichiometry of modification. Additionally, this method has been used to incorporate fluorescent tags, allowing researchers to quantitatively measure and image intracellular O-GlcNAc-modified proteins in vivo (Clark et al. 2008). Although these established methods provided great advancements for the detection of O-GlcNAc-modified proteins, the weak nature of metabolic labeling with peracetylated azido-GlcNAc has led to the development of better technologies. Subsequent studies have shown that GalNAz can be converted to UDP-GalNAz and epimerized to UDP-GlcNAz by mammalian biosynthetic enzymes, which is then be added to cells by OGT (Boyce et al. 2011). Proof-of-principle experiments highlighting this metabolic labeling demonstrated that numerous proteins are O-GlcNAc-modified, laying a solid framework for future studies to visualize and characterize dynamic O-GlcNAc-mediated signaling events (Boyce et al. 2011).

Mass spectrometry

Mass spectrometry-based methods have become widely and successfully employed for studying proteins and proteomes as well as post-translational modifications such as phosphorylation and methylation using collision-induced dissociation (CID) in tandem mass spectrometry (MS/MS). However, detection and site mapping of O-GlcNAc-modified peptides by mass spectrometric techniques is challenging. This is largely due to the labile nature of the β-linkage in the gas phase, which is lost prior to fragmentation of the peptide backbone and often prevents correct peptide identification, localization of the O-GlcNAc site, and consequently relative quantification. More importantly, distinguishing O-GlcNAc peptides from a complex mixture is problematic because ion intensities of O-GlcNAc peptides are often suppressed when compared to unmodified peptides. Despite these pitfalls, the generation of diagnostic fragment ions of O-GlcNAc-modified peptides, for example [M + H]1+ of m/z 204.1 for GlcNAc oxonium ions (Carr et al. 1993; Huddleston et al. 1993), can be used to distinguish O-GlcNAc-modified peptides from complex mixtures. By optimizing fragmentation energy to release O-GlcNAc (m/z 204.1), O-GlcNAc-modified peptides can be detected by ESI-MS, which has been employed in mapping O-GlcNAcylation sites in human cytomegalovirus tegument basic phosphoprotein (UL32) to serine 921 and serine 952 (Greis et al. 1994).

In contrast to CID, electron capture dissociation and electron transfer dissociation (ETD) are alternative fragmentation methods used to preserve more labile modifications such as phosphorylation, methylation, acetylation, glycosylation, nitrosylation, and sulfation, and allow for direct mapping of peptide/protein modifications (Mikesh et al. 2006; Syka et al. 2004; Udeshi et al. 2007, 2008; Wu et al. 2007; Wang et al. 2010a; Sobott et al. 2009). ETD fragmentation cleaves along the peptide backbone between the Cα-N producing c and z ions while still maintaining peptide side chains and modifications (Syka et al. 2004; Sobott et al. 2009). ETD-MS has been utilized to identify dynamically O-GlcNAcylated proteins and their modification sites in excitatory neurons in the brain (Khidekel et al. 2007), the cell adhesion protein paxillin (Kwak et al. 2010), transcription activation of Foxo1 in response to glucose (Housley et al. 2008), examination of the extensive crosstalk between phosphorylation and O-GlcNAcylation during cytokinesis (Wang et al. 2010b), and many other processes (Chou et al. 1995). Although the emergence of ETD-MS has proven to be a successful technique, again enrichment of O-GlcNAcylated peptides is ideal for effective sequencing and identification of these peptides by MS.

Future directions

Phosphorylation and other PTMs have historically been the exclusive focus of studies on cellular signal transduction events. As discussed in this review, O-GlcNAc is emerging as a PTM that should also be considered as an equally important contributor and regulator of signaling pathways, especially those involved in regulating the cells response to stress and injury. Previously, it has proven difficult to detect O-GlcNAc and O-GlcNAc-modified proteins. The work of numerous groups has resulted in the development of alternative approaches that solve many of the caveats associated with studying this labile protein modification. The application of these techniques will answer many of the remaining questions: Which proteins are O-GlcNAc modified in response to stress?; Are the proteins modified in a stress- and/or tissue-specific manner?; How does the O-GlcNAc modification alter the function of these proteins at a molecular level?; How are OGT and O-GlcNAcase regulated during stress and injury?; and importantly, how is the O-GlcNAc-mediated stress response misregulated in disease?