Role of O-linked N-acetylglucosamine protein modification in cellular (patho)physiology

Abstract

In the mid-1980s, the identification of serine and threonine residues on nuclear and cytoplasmic proteins modified by a N-acetylglucosamine moiety (O-GlcNAc) via an O-linkage overturned the widely held assumption that glycosylation only occurred in the endoplasmic reticulum, Golgi apparatus, and secretory pathways. In contrast to traditional glycosylation, the O-GlcNAc modification does not lead to complex, branched glycan structures and is rapidly cycled on and off proteins by O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA), respectively. Since its discovery, O-GlcNAcylation has been shown to contribute to numerous cellular functions, including signaling, protein localization and stability, transcription, chromatin remodeling, mitochondrial function, and cell survival. Dysregulation in O-GlcNAc cycling has been implicated in the progression of a wide range of diseases, such as diabetes, diabetic complications, cancer, cardiovascular, and neurodegenerative diseases. This review will outline our current understanding of the processes involved in regulating O-GlcNAc turnover, the role of O-GlcNAcylation in regulating cellular physiology, and how dysregulation in O-GlcNAc cycling contributes to pathophysiological processes.

CLINICAL HIGHLIGHTS.

The modification of proteins by sugars is one of the most common posttranslational modifications of proteins. Such modifications were believed to occur only on extracellular and secreted proteins and to consist of large branching structures comprising different sugar molecules. In the mid-1980s, a new modification was identified which consisted of a single N-acetylglucosamine moiety (O-GlcNAc) attached to serine and threonine residues of nuclear and cytoplasmic proteins. Since its discovery, O-GlcNAc modification of proteins has been shown to affect numerous cellular functions, and changes in O-GlcNAc levels have been implicated in a wide variety of diseases. The goal of this review is to summarize our current knowledge of O-GlcNAc biology and its contribution to normal physiology and disease.

1. INTRODUCTION

1.1. Brief History of O-GlcNAc

The modification of proteins by carbohydrates, otherwise known as protein glycosylation, is the most common posttranslational modification of proteins and occurs in all cells and organisms (1). In the early 1900s, there was considerable speculation that carbohydrates were important parts of the structure of proteins, but the technology was lacking to provide definitive evidence. It was not until the 1960s and 1970s that a better understanding of the structure and function of these complex glycans on proteins started to emerge. Through the mid-1980s, the consensus was that protein glycosylation was restricted to extracellular proteins that originated in the endoplasmic reticulum (ER), Golgi apparatus, and secretory pathway (1). However, in 1984, Torres and Hart (2) designed a study to characterize terminal N-acetylglucosamine (GlcNAc) residues on the surface of lymphocytes. Unexpectedly, they demonstrated that the majority of these terminal residues were localized inside the cell, and that rather than part of extended glycan structures, they existed as a single O-linked GlcNAc monosaccharide.

Two years later, Holt and Hart (3) characterized the cellular distribution of O-GlcNAc-modified proteins in rat liver cells, demonstrating that while they were found in nearly every cellular compartment, they were particularly enriched in the cytoplasm and nucleus. In 1987, a monoclonal antibody to rat liver nuclear pore complex (clone RL2), appeared to primarily recognize O-linked O-GlcNAc groups (4). Hanover et al. (5) also reported nuclear pore proteins were modified by O-GlcNAc; however, the functional consequence of this modification was unclear at that time. During the same period, Holt et al. (6) identified both serine (Ser) and threonine (Thr) residues as the primary residues modified by O-GlcNAc. The fact that O-GlcNAcylated proteins were especially enriched in the nucleus raised the possibility that the modification could be involved in protein transport into the nucleus; however, the observation that cytoskeletal proteins in erythrocytes, which lack a nucleus, were modified by O-GlcNAc suggested other functions for this modification (7). In 1992, the protein responsible for adding O-GlcNAc to proteins, O-GlcNAc transferase (OGT) was purified (8), but it was not until 1997 that the gene encoding OGT was identified, revealing a glycosyltransferase that was unrelated to any other previously known glycosyltransferases (9).

In contrast to traditional protein glycosylation, it was quickly established that O-GlcNAc modifications occurred rapidly and reversibly (10), suggesting the existence of an N-acetyl-glucosaminidase(s) responsible for its removal from proteins. Dong et al. (11) purified an O-GlcNAcase that was distinctly different from lysosomal hexosaminidases, in that it was localized in the cytosol and was optimally active at a neutral, rather than acidic, pH. In 2001, O-GlcNAcase was cloned and recognized to be identical to a previously known hexosaminidase C of unknown function, which specifically cleaved O-GlcNAc but not O-linked N-acetylgalactosamine (O-GalNAc) from glycopeptides (12). O-GlcNAcase was subsequently shown to have an identical sequence to a previously identified protein from meningioma patients called meningioma expressed antigen 5 (MGEA5) (13).

Over the decades since its discovery, O-GlcNAc modified proteins have been identified in all metazoans, some bacteria, protozoa, and viruses, but to date, not in yeast (14). Recent studies have suggested that in yeast, the addition of O-linked mannose (O-Man) on nuclear and cytoplasmic proteins might play a similar role to O-GlcNAc (15). In support for the necessity of O-GlcNAc signaling in mammals, germline deletion of OGT in mice has been shown to be embryonically lethal (16), and deletion of OGA results in perinatal mortality in mice (17). Furthermore, mammalian OGT and OGA are ubiquitously expressed, and proteins of every functional class have been shown to be subject to O-GlcNAcylation (FIGURE 1A) (18). Since its discovery, O-GlcNAcylation has been shown to contribute to numerous cellular functions, including signaling, protein localization and stability, transcription, chromatin remodeling, mitochondrial function, and cell survival (FIGURE 1B) (14). Given its diverse roles, it is not surprising that dysregulation in O-GlcNAcylation has been implicated in a wide range of pathophysiological processes, such as diabetes, diabetic complications, cancer, cardiovascular, and neurodegenerative diseases (19–24).

FIGURE 1.

A: O-linked N-acetylglucosamine acylated (O-GlcNAcylated) proteins belong to many different classes of proteins responsible for regulating diverse cellular processes. Some of the largest classes of proteins include those in regulating metabolism, transcription, and translation, as well as structural proteins. B: O-GlcNAcylated proteins are present in numerous cellular compartments, including the nucleus, cytosol, and mitochondria. Cytosolic domains of membrane proteins are also O-GlcNAcylated, as well as proteins involved in autophagy and proteosomal degradation of proteins, chaperone proteins, vesicle proteins, and numerous cytosolic proteins and enzymes. This figure is based, in part, on information presented in Chapter 19, The O-GlcNAc Modification, Essentials in Glycobiology, 3rd ed. (14). ER, endoplasmic reticulum.

1.2. Differences between O-GlcNAc and Traditional Glycosylation

Until the paradigm-changing study by Torres and Hart (2), protein glycosylation was thought to be limited to extracellular and excreted proteins. These proteins are processed via the ER-Golgi pathway, which contains a large number of glycosyltransferases that are responsible for creating N- and O-linked glycan structures. N-glycans are attached to proteins as asparagine residues via an N-glycosidic bond; whereas, O-glycans are attached to Ser or Thr residues. These proteins are subject to processing and maturation by numerous glycosyltransferases, leading to stable elongated and branched structures comprising a number of different monosaccharides. Glycosyltransferases are estimated to account for at least 2% of the human genome (25). The importance of the tight regulation of this process is highlighted by the fact that mutations in genes related to glycosylation are associated with more than 100 human genetic diseases, which are frequently associated with intellectual disabilities, as well as abnormalities in most organ systems (25).

Key distinguishing features of the O-GlcNAc modification are that 1) with few exceptions, it occurs primarily on nuclear and cytoplasmic proteins; 2) it consists of a single monosaccharide; 3) it is dynamic and rapidly reversible; 4) it is catalyzed by a single unique O-GlcNAc transferase, and 5) it is removed by a glycohydrolase that is specific for the removal of O-GlcNAc. It is of note that only very recently have mutations in the OGT gene been linked to human disease, such as intellectual disability (26–29). How mutations of OGT lead to disease and what specific functions are perturbed remain to be determined. In light of the growing recognition of the importance of O-GlcNAc-modified proteins in regulating cellular homeostasis—and that it is likely as abundant a modification as phosphorylation—it may seem surprising that it had not been identified earlier. One reason for this is that proteins modified by O-GlcNAc do not exhibit changes in mobility during gel electrophoresis, because of their low molecular weight and because unlike O-phosphate, it is uncharged. Moreover, despite its abundance, it exhibits low stoichiometry, estimated to be 5–10% of a specific site. In addition, the presence of hexosaminidases can remove O-GlcNAc from proteins unless they are specifically inhibited.

1.3. Identification of O-GlcNAcylated Proteins

The most widely used approach to detect O-GlcNAcylated proteins are pan-specific O-GlcNAc antibodies, the most commonly being RL2 and CTD110.6, but there are also a number of other commercially available O-GlcNAc antibodies as listed in TABLE 1 These can be used to provide a semiquantitative assessment of overall changes in O-GlcNAc levels via immunoblot or distribution of O-GlcNAc via immunohistochemistry in tissues and cells (FIGURE 2). The limitations of these antibodies include the fact that they have epitope specificity, and their selectivity for high- versus low-abundance proteins. Furthermore, a continuing limitation in the study of O-GlcNAcylated proteins is the lack of commercially available site-specific O-GlcNAc antibodies, although several have been developed for specific studies (79–82). Because of the lack of site-specific antibodies to determine whether a protein of interest is an O-GlcNAc target, the simplest approach is to immunoprecipitate the protein of interest followed by an O-GlcNAc immunoblot; however, because of differences in epitope specificities, a negative result with a single antibody does not preclude the possibility that a protein is modified. Moreover, because of the possibility of cross-reactivity with other sugars, if a positive result is obtained, a number of additional control experiments should be considered, such as preincubation with free GlcNAc to outcompete the antibody. Additionally, their cross-reactivity with N-linked modifications varies by detection conditions (83). A number of methodological reviews on different approaches for studying O-GlcNAcylated proteins have been published that provide valuable technical details (44,45, 84,85).

Table 1.

Tools for use in studying O-GlcNAc levels in cells and tissues

Antibodies*	Characteristics and Limitations	Commercial Availability
CTD110.6 (IgM) (30)	Monoclonal antibody raised against RNA polymerase II subunit 1C terminal domain. Reportedly less dependent on protein structure than other antibodies, thereby, recognizing more proteins. Relatively low-binding affinity, therefore, biased toward more highly abundant proteins.	Multiple sources
HGAC39 (IgG) (31–33)	Monoclonal antibody raised against streptococcal group A carbohydrate (GAC) demonstrated to recognize O-GlcNAcylated proteins.	No
HGAC85 (IgG) (33)	Monoclonal antibody raised against streptococcal group A carbohydrate (GAC) demonstrated to recognize O-GlcNAcylated proteins. Not as widely used as CTD110.6 or RL2. Some of the most recent studies have used it in ChIP-chip assays.	Multiple sources
MY95 (IgG) (34,35)	Originally generated as an antinuclear pore complex antibody, subsequently shown to recognize O-GlcNAc modification.	No
RL2 (IgG) (4)	Because of epitope specificity, having been raised against nuclear pore protein, it recognizes only a subset of O-GlcNAc proteins. Relatively low binding affinity; therefore, it is biased toward more highly abundant proteins. This is the case for all pan-O-GlcNAc antibodies.	Multiple sources
1F5.D6(14) (IgG) (36)	Appears to recognize a wider range of O-GlcNAcylated proteins, including those below 50 kDa that are not usually identified by RL2 or CTD110.6.	Millipore
9D1.E4(10) (IgG) (36)	Appears to recognize a wider range of O-GlcNAcylated proteins, including those below 50 kDa, which are not usually identified by RL2 or CTD110.6.	Millipore
18B10.C7(3) (IgG) (36)	Appears to recognize a wider range of O-GlcNAcylated proteins, including those below 50 kDa, which are not usually identified by RL2 or CTD110.6.	Millipore

Other Identification Methods	Characteristics and Limitations	Commercial Availability
Agrocybe aegerita GlcNAc-specific lectin (AANL) (37)	AANL, also reported as AAL2, is a useful tool for enrichment and identification of O-GlcNAcylated proteins and peptides.	Not known
Click-IT O-GlcNAc Enzymatic Labeling System (38–40)	Chemoenzymatic labeling of proteins resulting in O-GlcNAc residues being replaced by azido-modified galactose (GalNAz), followed by chemoselective ligation azide. Resulting proteins can be detected via Western blot using a using a variety of alkyne-modified chemical probes. This approach is not epitope specific, which has advantages over pan-O-GlcNAc antibodies; however, more sample processing steps are required.	Multiple sources
Galactosyl transferase/ [3H]-galactose (41)	Results in incorporation of 3H into terminal GlcNAc residues on proteins allowing for detection by autoradiography. Can be very time intensive because of low sensitivity of 3H.	Multiple sources
Wheat germ agglutinin (WGA) and succinylated WGA (sWGA) (41)	Both WGA and sWGA can be used for immunoblotting. WGA identifies all terminal GlcNAc residues, as well as sialic acid. sWGA reduces affinity for sialic acid. Lack of specific for O-GlcNAc requires careful interpretation.	Multiple sources

Enrichment Strategies	Characteristics and Limitations	Commercial Availability
β-elimination followed by Michael addition of dithiothreitol (BEMAD) (42)	BEMAD relies on the β-elimination of phosphate or O-GlcNAc under basic conditions followed by Michael addition using dithiothreitol or a biotin-thiol probe.
Chemi-enzymatic labeling (43)	Using the same approach described for BEMAD, it replaces O-GlcNAc with GalNAz, combined with a biotin- or streptavidin-cleavable linker can be used to enrich O-GlcNAcylated peptides. A range of linkers have been developed that have different properties.
Immunoprecipitation (36, 44)	The most widely used antibodies, RL2 and CTD110.6, do not perform well for immunopurification. Although not as widely used, the Millipore antibodies listed above have been reported to be effective for IP.
WGA-agarose (45)	Because of the lack of specificity for O-GlcNAc, other glycoproteins will be copurified. This can be minimized by pretreatment to remove the N- and O-linked glycans.	Multiple sources

GFAT Inhibitors	Characteristics and Limitations	Commercial Availability
Azaserine (46)	Glutamine analog used to decrease HBP flux by inhibiting GFAT; however, it lacks specificity as it can inhibit other pathways that utilize glutamine (47). In the absence of any other small molecule GFAT inhibitors, it continues to be widely used, but interpretation of results needs to be cautious.	Multiple sources
6-diazo-5-oxo-L-norleucine (DON) (46)	Glutamine analog used to decrease HBP flux by inhibiting GFAT; however, it lacks specificity as it can inhibit other pathways that utilize glutamine (48). In the absence of any other small molecule GFAT inhibitors, it continues to be widely used, but interpretation of results needs to be cautious.	Multiple sources

OGA Inhibitors	Characteristics and Limitations	Commercial Availability
α-GlcNAc thiolsulfonate (49)	OGA inhibitor exhibiting selectivity of short OGA compared to long OGA.	No
Gluco-nagstatin (50,51)	Based on natural product, nagstatin, which is a potent inhibitor of β-hexosaminidase. Inhibits OGA but is more effective for of β-hexosaminidase.	No
GlcNAcstatins (50, 52,53)	Family of OGA inhibitors, with high degree of potency and selectivity.	Abmole Bioscience, Inc.
NAG-thiazoline (50, 54)	Inhibitor of OGA with much greater selectivity over other hexosaminidases than PUGNAc.	No
NButGT (50, 55,56)	OGA inhibitor based on NAG-thiazoline scaffold. Less potent that NAG-thiazoline but even more selective over other hexosaminidases.	No
PUGNAc (57)	Widely used cell permeable OGA inhibitor, but also inhibits hexosaminidase A and B inhibitor. Limited aqueous solubility; usually dissolved in DMSO prior to dilution in aqueous solutions for biological studies.	Multiple sources
Streptozotocin (50, 54, 58,59)	A GlcNAc analog initially thought to inhibit OGA. Has major off target effects and more recent studies have reported that it does not inhibit OGA. Not recommended for use.	Multiple sources
Thiamet-G (50, 60)	Derivative of NButGT, with greater stability, water soluble, and orally available.	Multiple sources

OGT Inhibitors	Characteristics and Limitations	Commercial Availability
Alloxan (61,62)	A weak OGT inhibitor with numerous off target effects. Requires millimolar concentrations to lower cellular O-GlcNAc levels. Not recommended for use.	Multiple sources
Benzyl-2-acetamido-2-deoxy-α-d-galactopyranoside (BADGP)(63,64)	Although BADGP decreases O-GlcNAc levels in millimolar range it lacks specificity for OGT. Inhibits other O-Glycosyltransferases. Not recommended for use.	Sigma
L01 (63)	Effective in reducing cellular O-GlcNAc levels in 10–100-µM range.	No
OSMI-1 (65–68)	Reduces O-GlcNAc levels in cells at ∼50 µM. Shown to be specific for OGT compared with other glycosyltranserases.	Sigma
OSMI-3, OSMI-4 (69)	On the basis of OSMI-1 scaffold, but with greater potency. Reduces cellular O-GlcNAc levels in ∼5-µM range.	ProbeChem
ST045849 (TT04) (63, 66, 70–73)	Identified via high throughput screen. Specificity for OGT compared to other glycosyltranserases unclear.	TimTech
ST060266 (70, 74)	Identified via high-throughput screen. Specificity for OGT compared to other glycosyltranserases unclear.	TimTech
ST078925 (70, 72, 75)	Identified via high-throughput screen. Specificity for OGT compared to other glycosyltranserases unclear.	TimTech
UDP-5SGlcNAc; (61, 76)	Effective in vitro inhibitor of OGT but lacks cell permeability.	No
2‐deoxy‐2‐N‐hexanamide‐5‐thio‐d‐glucopyranoside (5SGlcNHex); (77)	Shown to be effective in lowering tissue O-GlcNAc levels following in vivo administration.	No
4-methoxyphenyl 6-acetyl-2-oxo-2,3-dihydro-1,3-benzoxazole-3-carboxylate (78)	A cell permeable, irreversible inhibitor of OGT.	No
5-thioglucosamine (5SGlcNAc) (76)	The peracetylated form of 5SGlcNAc crosses the cell membrane and converts to UDP-5SGlcNAc, which binds to the active site of OGT. Reduces O-GlcNAc levels in cells in the 10-100-µM range.	No

AAL2, Agrocybe aegerita GlcNAc-specific lectin; CTD, COOH-terminal domain; GFAT, d-fructose-6-phosphate amidotransferase; GlcNAc, N-acetylglucosamine moiety; HBP, hexosamine biosynthesis pathway; IP, immunoprecipitation; NAG, N-acetylglucosamine; O-GlcNAc, O-linked N-acetylglucosamine; OGT, O-GlcNAc transferase; RL2, rat liver nuclear pore complex; UDP, uridine diphosphate-azido-modified galactose.

FIGURE 2.

Overview of different approaches for the identification of O-linked N-acetylglucosamine acylated (O-GlcNAcylated) proteins, including galactosyltransferase labeling; immunopurification and chemoenzymatic labeling were combined with LC-MS/MS. Further details are found in TABLE 1. IP, immunoprecipitation; UDP, uridine diphosphate-azido-modified galactose.

Another challenge in the identification and characterization of O-GlcNAcylated proteins is that O-GlcNAc-modified peptides are frequently not detected using traditional collision-induced dissociation mass spectrometry (MS), as O-GlcNAc is very labile and is usually lost during collision-induced fragmentation. The development of electron transfer dissociation (ETD) MS techniques (86,87) significantly enhanced the identification of O-GlcNAc modification sites, since ETD does not usually result in the loss of the O-GlcNAc moiety from the peptide; however, the problems of ion suppression and low stoichiometry remain. To overcome these limitations, it is necessary to enrich the sample for O-GlcNAcylated peptides (88). One approach is to use a traditional immunoprecipitation with a single anti-O-GlcNAc antibody, as described in several studies (36, 89). On the other hand, because of the limitations of epitope specificity, there is a concern that only a subset of O-GlcNAc proteins will be identified. One way to overcome that limitation is to use a combination of antibodies; for example, Lee et al. (90) developed a G5-lectibody resin column that consisted of four different O-GlcNAc antibodies and the lectin wheat germ agglutinin (WGA) (FIGURE 2). Another enrichment approach is chemoenzymatic labeling (38, 44, 91), which involves using uridine diphosphate-azido-modified galactose (UDP-GalNAz) and a mutant galactosyltransferase to tag O-GlcNAc moieties with a reactive azide group. This is followed by the addition of a biotin group attached to a cleavable linker, using a copper-based azide-alkene cycloaddition (84). The resulting modified proteins or peptides can then be affinity purified using an avidin column. A number of different linkers have been described, including a UV-cleavable linker (92,93) and, more recently, a hydrazine-sensitive linker (94). A key advantage of the latter approach is the resulting tag at the O-GlcNAc site allows for more efficient fragmentation. There are several different versions of this technique, and refinements continue to be developed focused on improving the efficiency of the release of proteins/peptides from the avidin column, as well as simplifying the underlying chemistry (43, 95,96).

Compared with most other PTMs, our understanding of how O-GlcNAcylation regulates protein function and cellular physiology remains limited, although our knowledge is rapidly growing. The development of increasingly selective and specific small-molecule inhibitors of OGT and OGA, as summarized in TABLE 1, have helped advance our knowledge of the role of O-GlcNAc in cellular function. These new tools have helped stimulate research in O-GlcNAc biology, as illustrated in FIGURE 3. About 20 years ago, there were less than 20 articles published per year, but this has been steadily increasing, and in 2018, it reached ∼225 (or more than the first 20 years combined). As this becomes a more widely recognized area of biology, it is likely to grow more rapidly, given the increasing appreciation of its importance in regulating key physiological processes combined with its contributions to the development of diverse pathologies. Several thousands of proteins have been identified as O-GlcNAc targets (44), and the number continues to increase, as new techniques are developed. Many new O-GlcNAc sites are identified via high-throughput MS; consequently, the biological function of the modification is often not known. In TABLE 2, we provide a list of some key proteins, divided via cellular location or function, which are validated O-GlcNAc targets in which the specific modification site has been identified. In TABLE 3, the physiological effects of gain or loss of function of OGT and OGA in mammals are summarized.

FIGURE 3.

Number of O-linked N-acetylglucosamine acylated (O-GlcNAc) publications by year annotated by key events in O-GlcNAc biology from its initial discovery. Relevant citations are all included in the main text. CTD, COOH-terminal domain; ETD, electron dissociation transfer hOGA, human O-GlcNAcase; hOGT, human O-GlcNAc transferase; KO, knockout; OGA, O-GlcNAcase; OGT, O-GlcNAc transferase; RL2, rat liver nuclear pore complex.

Table 2.

List of selected O-GlcNAcylated proteins and their modification sites

Transcription Factors and Transcriptional Regulators
Protein Symbol	O-GlcNAc Modification Sites*	Effects on Protein Function	Citations
C/EBPβ	Ser-180, Ser-181	Regulates both the phosphorylation and DNA binding activity of C/EBPβ.	(97)
cMyc	Thr-58	Reduces phosphorylation of Ser-62 and Thr-58 and increases protein stability.	(98,99)
CREB	Ser-40	Represses both basal and activity-dependent transcription.	(100)
ERα	Ser-10, Thr-50, Thr-575	May regulate protein turnover.	(101,102)
ERβ	Ser-16	May regulate protein turnover.	(103)
ERRγ	Ser-317, Ser-319	Enhances receptor activity.	(104)
FOXO1	Thr-317, Ser-550, Thr-648, Ser-654	Increases target gene transcription.	(105)
FXR	Ser-62	Enhances FXR gene expression and protein stability.	(106)
KEAP1	Ser-104	O-GlcNAc is required for the efficient ubiquitination and degradation of Nrf2.	(107)
Sp1	Ser-491, Ser-612, Thr-640, Ser-641, Ser-698, Ser-702	Inhibits transcriptional activation.	(108,109)
LXRα/β	Ser-49	Increases transactivation.	(110–112)
Oct4	Thr-116, Thr-225, Ser-236, Ser-288/889/890, Ser-335, Ser-349, Thr-351, Thr-352, Ser-355, Ser-359	O-GlcNAc increases transcriptional activity.	(113)
P27	Ser-2	Suppresses protein interactions.	(114)
p53	Ser-139	Reduces phosphorylation and stabilizes protein.	(115)
Per2	Ser-566, Ser-580, Ser-653, Ser-662, Ser-668, Ser-671, Thr-734, Thr-965, Ser-983, Thr-1180	O-GlcNAc increases its suppressor activity.	(116)
PGC-1α	Ser-334, Ser-333	Enhances stability and upregulated downstream genes.	(117,118)
PPARγ	Thr-54	Reduces its transcriptional activity.	(119)
SIRT1	Ser-549	O-GlcNAcylation increases deacetylase activity, promotes cytoprotection under stress.	(120)

Insulin Signaling and Other Metabolic Proteins
Protein Symbol	O-GlcNAc Modification Sites*	Effects on Protein Function	Citations
Akt1/2	Ser-126, Ser-129, Thr-305, Thr-308, Thr-312, Ser-473	Decreases phosphorylation and activity.	(121–126)
CRTC2	Ser-70, Ser-171	Increased O-GlcNAc leads to nuclear translocation and promotes gluconeogenesis.	(127)
GAPDH	Thr-227	Increases nuclear translocation.	(128)
GFAT1	Ser-243	AMPK phosphorylation of Ser-243 reduced GFAT activity.	(129,130)
GSK3β	Ser-9, Thr-38, Thr-39, Thr-43	Decreased activity.	(125, 131–133)
G6PDH	Ser-84	Increases activity.	(134)
IRS1	Ser-914, Ser-1009, Ser-1036, Ser-1041	Attenuates insulin-mediated phosphorylation of IRS1.	(135,136)
PDH (E1)	Ser-13, Ser-15, Ser-134, Ser-232	Higher O-GlcNAc levels associated with greater activity.	(137,138)
PDH (E2)	Ser-7, Ser-15, Ser-239, Ser-411	Remains to be determined.	(137–139)
PDK2	Ser-110	Remains to be determined.	(138)
PFK1	Ser-529	O-GlcNAc levels increased in hypoxia and decreases activity.	(140)
PKM2	Thr-405, Ser-406	Reduces activity and increases nuclear translocation.	(141)

Contractile and Cytoskeletal Proteins
Protein Symbol	O-GlcNAc Modification Sites*	Effects on Protein Function	Citations
αB-crystallin	Thr-170	Regulates stress induced translocation.	(142,143)
β-Actin (Skeletal Muscle)	Ser-52, Ser-155, Ser-199, Ser-232, Ser-323, Ser-368	O-GlcNAcylation of Ser-199 may regulate elongation of actin filaments.	(144–147)
Keratin18	Ser-29, Ser-30, Ser-48	Increased O-GlcNAc promotes cell survival by activating Akt.	(148,149)
Myosin heavy chain(Skeletal Muscle)	Ser-1097, Ser-1299, Ser-1708, Ser-1920	Reduced myosin Ca2+ sensitivity.	(145,146)
Myosin heavy 6 (Cardiac)	Thr-35, Thr-60, Ser-172, Ser-179, Ser-196 Ser-392, Ser-622, Ser-626, Ser-644, Ser-645, Ser-749. Ser-880, Ser-1038, Ser-1148, Ser-1189, Ser-1200, Ser-1308, Ser-1336, Ser-1470, Ser-1597, Thr-1600, Thr-1606, Ser-1638, Thr-1697, Ser-1711, Ser-1777, Ser-1838, Ser-1916	Reduced myosin Ca2+ sensitivity.	(147, 150)
MYL1	Ser-45, Thr-93, Thr-164	Remains to be determined.	(147, 150)
MYL2	Ser-15	Remains to be determined.	(147)
Synapsin I	Ser-55, Thr-56, Thr-87, Ser-436, Ser-516, Thr-524, Thr-562, and Ser-576	Regulation of synaptogenesis.O-GlcNAc of Thr-87 regulates localization of synapsin I.	(151,152)
Tau	Ser-208, Ser-238, Ser-400, Ser-692	Loss of O-GlcNAc induces hyperphosphorylation.	(153–155)
TnI	Ser-150	Remains to be determined.	(147)
TnT	Ser-190	Increased O-GlcNAc reduces Ser208 phosphorylation.	(156)
Tropomyosin α1	Ser-87, Ser-123, Ser-186, Ser-206	Remains to be determined.	(150)

Oxidative Phosphorylation and Other Mitochondrial Proteins
Protein Symbol	O-GlcNAc Modification Sites*	Effects on Protein Function	Citations
ATP5A	Ser-76, Thr-432	Decreased activity.	(138, 157,158)
ATP5B	Ser-106, Ser-128	Decreased activity.	(138)
DRP1	Thr-585, Thr-586	Associated with increased mitochondrial fragmentation and decreased mitochondrial membrane potential.	(159)
Milton	Ser-447, Ser-829, Ser-830, Ser-938	Attenuates mitochondrial motility.	(160)
NDUFA9	Ser-156, Ser-230	Impaired complex activity.	(138, 161)
Prohibitin	Ser-121	Decreases phosphorylation.	(138,139, 162)
VDAC1	Thr-2, Ser-240, Ser-260	Increased O-GlcNAc attenuates mitochondrial Ca2+ uptake.	(138, 163)
VDAC2	Ser-2	Increased O-GlcNAc contributes to mitochondrial dysfunction and apoptosis.	(139, 164)

Kinases, Phosphatases, and Other Signaling Molecules
Protein Symbol	O-GlcNAc Modification Sites*	Effects on Protein Function	Citations
CaMKII	Ser-279, Ser-280	Activates enzyme.Increases NOX2 ROS production	(165)(166)
CaMKIV	Thr-57, Ser-58, Ser-137, Ser-189, Ser-344, Ser-345, Ser-356	Activates enzyme.	(167)
CDK5	Ser-46, Thr-245, Thr-246, and Ser-247	Decreases activity.	(168)
CK2α	Ser-347	Attenuates interaction with Pin1 and facilitates proteasomal degradation.Reduces CKII phosphorylation decreasing its stability. May also affect substrate selectivity.	(131, 169,170)
GRASP55	Ser-389, Ser-390,Thr-403, Thr-404,Thr-413	O-GlcNAcylation attenuates autophagy.	(171)
IKKβ	Ser-733	Increases activity.	(172)
PKC-α, β,δ, ε, θ, ζ	Numerous (see citations for details)	Function not well defined and could be isoform specific. In some cases, likely competes with phosphorylation and decreases enzyme activity.	(173–175)
PTP1B	Ser-104, Ser-201, Ser-386	Lower O-GlcNAc levels leads to lower phosphatase activity.	(176)
RACK1	Ser-122	Promotes stability and interaction with PKCβII.	(177,178)
RIPK3	Thr-467	Prevents dimerization and limits inflammation.	(179)
SNAP29	Ser-2, Ser-61, Thr-130, Ser-153	O-GlcNAcylation attenuates autophagy.	(180)
ULK1,2	Thr-613, Thr-635, Thr-726, Thr-754,	O-GlcNAc increases AMPK-mediated phosphorylation of ULK, leading to increased activity.	(92, 181,182)
YAP1	Ser-109, Thr-241	Ser-109 O-GlcNAc Disrupts interaction with upstream kinase LATS1. Thr-241 O-GlcNAc increases activity by improving stability.	(183,184)
Miscellaneous
Protein Symbol	O-GlcNAc Modification Sites	Effects on Protein Function	Citations
α-synuclein	Thr-64, Thr-72, Thr-75, Thr-81, Ser-87	Increased O-GlcNAcylation associated with decrease in aggregation; however, this has primarily been identified in in vitro studies.	(185)
β-catenin	Ser-23, Thr-40, Thr-41, Thr-112	Promotes its recruitment to the plasma membrane and its binding to E-cadherin and regulates stability.	(186,187)
CREB-binding protein	Ser-147, Ser-2360	Remains to be determined.	(188)
eIF4A	Ser-322, Ser-323	Regulates formation of translation initiation complex.	(189)
eNOS	Ser-1177	Impairs activity.	(190)
FNIP1	Ser-938	Attenuates phosphorylation and increases proteasomal degradation.	(191)
HCF-1	Thr-490, Ser-569, Ser-620, Ser-622, Ser-623, Thr-625, Ser-685, Thr-726, Thr-779 Thr-801, Thr-861, Thr-1143, Thr-1273, Thr-1335, Thr-1743, Thr-1238, Thr-1241	O-GlcNAcylation is a signal for its proteolytic processing	(188, 192)
HSP90	Ser-434, Ser-452, Ser-461	Remains to be determined	(193)
PLB	Ser-16	Inhibits phosphorylation.	(194)
TAB1	Ser-395	Required for full activation of TAK1	(81)
TAB2	Ser-166, Ser-350, Ser-354, Thr-456	May control the activity of IL-1 signaling pathway	(95, 188)
TAB3	Ser-408	Mediates cell migration via activation of NF-κB,	(195)

^*The proteins included in this table were selected primarily on the basis of their inclusion within the main body of the article. In addition to the citations listed in Table 2, other resources that were used included PhosphoSite Plus, www.phosphosite.org (196), Dias et al. (197), Ma et al. (138,139). CKII, casein kinase II; FXR, farnesoid X receptor; GFAT, l-glutamine: d-fructose-6-phosphate amidotransferase; IRS1, insulin receptor substrate 1; LATS1, large tumor suppressor kinase 1; NOX2, NADPH phagocyte oxidase isoform; Nrf2, nuclear factor E2–related factor-2; ROS, reactive oxygen species; O-GlcNAc, O-linked N-acetylglucosamine; TAK1, transforming growth factor-β activated kinase 1; ULK, Unc-51-like autophagy activating kinase 1.

Table 3.

Physiological roles of OGT and OGA in mammals

Genotype	Phenotype	Citations
Human
Ogt human mutation L254F or E1974H	Intellectual disability.	(26,27, 29, 198)
Rat
Oga Goto-Kakizaki (GK) rat mutation	Diabetes.	(199)
Systemic inhibition of OGA by thiamet G in wildtype rat	Impaired novel place recognition.	(200)
Mouse
AgRP-cre::Ogtf/f	Ogt ablation in AgRP neurons. Obese due to lack of browning of white fat.	(201)
CaMKII-cre::Ogtf/f	Constitutive Ogt ablation in forebrain neurons. Loss of body and brain weight and significant neurodegeneration almost as early as birth, decreased lifespan.	(202)
CaMKII-creER::Ogtf/f	Inducible Ogt ablation in forebrain neurons. Obese due to overeating after acute deletion of Ogt.	(203)
γ-F-crystallin-rtTA:: dnOGA	OGA overexpression in the eye. Premature cataracts.	(204)
MCK-rtTA::dnOGA	OGA overexpression in the skeletal muscle. Skeletal muscle atrophy.	(205)
MMTV-cre::Ogaf/f	Constitutive Oga ablation in mammary gland. Obesity, in females.	(17)
MHC-cre::Ogtf/f	Constitutive Ogt ablation in cardiomyocytes. Postnatal lethality, dilated hearts, and signs of heart failure.	(206)
Oga -/-	Germline knockout. Perinatal lethality.	(17, 207)
Oga +/-	Germline heterozygote. Lean not due to overeating or higher energy expenditure, associated with higher RER.	(207)
Ogt -/-	Germline knockout. Embryonic stem cell lethality.	(16, 208)
Systemic inhibition of OGA by thiamet G in Aβ and Tau overexpression models	Decrease pathology.	(60, 209)

AgRP, Agouti-related protein.

In the following sections, we have provided an overview of our current understanding of the regulation of O-GlcNAcylation, its role in regulating cellular physiology, as well as its potential role in pathophysiology of diseases, including diabetes, cardiovascular disease, and cancer.

2. REGULATION OF O-GlcNACYLATION

2.1. Hexosamine Biosynthesis Pathway

Uridine diphosphate-GlcNAc (UDP-GlcNAc) is the common substrate for all amino sugars involved in the synthesis of glycoproteins and proteoglycans. As the end product of the hexosamine biosynthesis pathway (HBP), UDP-GlcNAc integrates multiple metabolic pathways and has long been considered an important nutrient signaling pathway partially regulated by substrate availability (FIGURE 4) (210–212). l-glutamine: d-fructose-6-phosphate transaminase (GFPT, EC 2.6.1.16), often referred to as l-glutamine: d-fructose-6-phosphate amidotransferase (GFAT), is the rate limiting enzyme of the HBP, catalyzing the transfer of the amide group from glutamine to fructose 6-phosphate, leading to the synthesis of glucosamine 6-phosphate (213). In mice and humans, there are two different isoforms of GFAT (GFAT1, GFAT2) encoded by two different genes, located on different chromosomes, and each isoform has markedly different tissue distribution (213, 214). In tissue from the central nervous system, GFAT2 expression predominates over GFAT1, whereas, in most other tissues, GFAT1 is more highly expressed (214). A splice variant of GFAT1, known as GFAT1Alt, has been identified, and it appears to be predominantly expressed in skeletal muscle and exhibits a higher K_m for fructose-6-phosphate and a lower K_i for UDP-GlcNAc than those for GFAT1 (213, 215). In mammals, GFAT exists as a tetramer, and its activity is highly dependent on the availability of both glutamine and glucose (213). Both glucosamine-6-phosphate and UDP-GlcNAc are potent allosteric inhibitors of mammalian GFAT (FIGURE 5) (214). Up to 20 different phosphorylation sites have been identified in GFAT; however, the responsible kinases and function(s) of the majority of them are unknown. GFAT1 and GFAT2 are regulated by cAMP-dependent protein kinase (PKA) phosphorylation on Ser-205 and Ser-235, respectively (213, 216,217); both AMPK and calcium/calmodulin-dependent kinase (CaMK) II phosphorylate GFAT1 at Ser-243 (FIGURE 5) (129, 218). There are contradictory reports on the effects of phosphorylation on GFAT activity, which may be due to isoform-specific differences. In the heart, a number of studies indicate that AMPK phosphorylation of GFAT1 reduces its activity (219); however, it remains to be determined whether this is also the case in other cells and tissues. GFAT is also transcriptionally regulated (FIGURE 5) by specificity protein 1 (Sp1) (220) and activating transcription factor 4 (ATF4) (221). Multiple GFAT1 missense mutations leading to loss of activity and reduced O-GlcNAc levels have been linked to neuromuscular transmission defects (222). Despite reports of several putative GFAT inhibitors (213, 223, 224), the regulation of HBP flux via small-molecule inhibitors of GFAT has been limited to the glutamine analogs azaserine and 6-diazo-5-oxo-l-norleucine (DON), which are limited because of their lack of specificity (TABLE 1).

FIGURE 4.

Schematic of UDP-GlcNAc and O-GlcNAc synthesis. Glucose enters the cell via the glucose transporter system where it is rapidly phosphorylated by hexokinase (HK) and converted to fructose-6-phosphate by phosphoglucoseisomerase (PGI). Fructose-6-phosphate is subsequently metabolized to glucosamine-6-phosphate by l-Glutamine: d-fructose-6-phosphate amidotransferase (GFAT), which requires glutamine. Glucosamine-6-phosphate is converted to N-acetylglucosamine-6-phosphate by glucosamine 6-phosphate N-acetyltransferase (Emeg32), utilizing acetyl-CoA. Phosphoacetylglucosamine mutase (Agm1) converts N-acetylglucosamine-6-phosphate to N-acetylglucosamine-1-phosphate. The synthesis of uridine-diphosphate-N-acetylglucosamine (UDP-GlcNAc) is catalyzed by UDP-N-acetylglucosamine pyrophosphorylase (Uap1), which consumes uridine triphosphate (UTP). UDP-GlcNAc is the substrate for (O-GlcNAc transferase (OGT) leading to the formation of O-linked β-N-acetylglucosamine (O-GlcNAc)-modified proteins. β-N-acetylglucosaminidase (OGA) catalyzes the removal of O-GlcNAc from the proteins. GlcNAc can reenter the HBP via two salvage pathways: 1) via N-acetylglucosamine kinase (NAGK) to generate N-acetylglucosamine 1-phosphate and 2) involving the conversion by N-acetylgalactosamine kinase (GALK2) of N-acetyl-galactosamine to N-acetylgalactosamine 1-phosphate and UDP-N-acetylgalactosamine, with subsequent conversion by an epimerase to UDP-GlcNAc. Glucosamine, which enters the cell via the glucose transport system and can be phosphorylated by hexokinase (HK) to form glucosamine 6-phosphate thereby bypassing GFAT. The kinases that have been identified as regulating GFAT, OGT, and OGA are indicated; additional details may be found in the text (see FIGURE 6 and TABLE 3). AMPK, AMPK-activated protein kinase; CaMKII, calcium/calmodulin (Ca²⁺/CaM) dependent protein kinase II; CHK1, checkpoint kinase-1; CK2, casein kinase 2; EPI, epimerase; GalNAc, glucosamine fructose-6-phosphate amidotransferase; GFAT, glucosamine fructose-6-phosphate amidotransferase; GlcNAc, N-acetylglucosamine; GlcNAc-1P, N-acetylglucosamine-1-phosphate; GSK3β, glycogen synthase kinase 3b; HEX, hexokinase; IRS1, insulin receptor substrate-1; PPi, pyrophosphate; UDP-GalNAc, uridine diphosphate N-acetylgalactosamine.

FIGURE 5.

Regulation of glucosamine fructose-6-phosphate amidotransferase (GFAT). GFAT activity is regulated at several levels including substrate availability, feedback inhibition by uridine diphosphate N-acetylglucosamine (UDP-GlcNAc) and glucosamine-6-phosphate (GlcN-6-P), as well as post-translational modifications by phosphorylation, acetylation, succinylation, and ubiquitination. The only kinases that have been identified as phosphorylating GFAT are PKA, AMPK, and calcium/calmodulin (Ca²⁺/CaM) dependent protein kinase II (CaMKII). Sp1, ATF4, and XBP1s have been shown to regulate GFAT at the transcriptional level. The database phosphosite.org and the citations contained therein were used to identify known posttranslational modification sites of GFAT (196). ATF4, activating transcription factor 4; Lys, lysine; Sp1, specificity protein 1; XBP1s, X-box binding protein 1.

Glucosamine 6-phosphate N-acetyltransferase (GNPNAT, GNA1), known as Emeg32 in mice, uses acetyl-CoA to convert glucosamine 6-phosphate to N-acetylglucosamine 6-phosphate (225), which is subsequently isomerized by phosphoglucomutase (PGM) to N-acetylglucosamine-l-phosphate. The final step in the HBP is conversion of N-acetylglucosamine-1-phosphate to UDP-GlcNAc, which is catalyzed by UDP-N-acetylglucosamine pyrophosphorylase (UAP1), also known as UDP-N-acetylhexosamine pyrophosphorylase (226). In addition to its de novo synthesis, UDP-GlcNAc can also be generated by two salvage pathways (FIGURE 4) (227). In one pathway, GlcNAc is phosphorylated by N-acetylglucosamine kinase to generate N-acetylglucosamine 6-phosphate. A second pathway involves the conversion of N-acetyl-galactosamine to N-acetylgalactosamine-1-phosphate and UDP-N-acetylgalactosamine, with subsequent conversion by an epimerase to UDP-GlcNAc (FIGURE 4). The relative contribution of the salvage pathways to total UDP-GlcNAc synthesis is not known. That deletion of GNPNAT gene is embryonically lethal (228) suggests that de novo synthesis is likely the predominant pathway. On the other hand, while deletion of EMeg32 substantially decreased UDP-GlcNAc levels, they were not eliminated. Further, the loss of EMeg32 did not lead to major changes in N- and O-glycosylation, whereas, O-GlcNAc levels were markedly suppressed, suggesting that salvage pathways were sufficient to maintain ER- and Golgi-mediated glycosylation, but not O-GlcNAcylation (228). Alternatively, under conditions in which UDP-GlcNAc levels are limiting, N- and O-glycosylation are prioritized over O-GlcNAcylation.

It is frequently stated that 2–5% of glucose entering a cell is metabolized via GFAT and the HBP. The origin of this statement is from a study using cultured adipocytes in which metabolic flux through glycolysis and the HBP was not directly measured, and the fraction of glucose metabolized by the HBP was inferred from other measurements (46). Moreover, energy metabolism, including glycolysis, varies widely from quiescent cells in culture to high-energy demands of the heart and brain. Consequently, the relative flux of glucose via glycolysis and HBP will also vary widely. There are very few studies that have directly measured the flux of glucose via the HBP, relative to other glucose-utilizing pathways, using radio- or stable-isotope techniques, either in cell culture or intact organs. In 2017, Gibb et al. (229), using ¹³C₆-glucose labeling in cultured neonatal cardiomyocytes, concluded that glucose is metabolized more by the HBP than the pentose phosphate pathway, suggesting that glucose utilization via the HBP could be considerably greater than previous estimates; however, this was under the low-energy demands of cell culture, as well as being in neonatal cardiomyocytes which would have different energetic demands from those in an adult heart. Olsen et al. (230) have recently developed an LC/MS technique using ¹³C-labeled substrates to measure the rate of glucose metabolism via the HBP at the same time as glycolytic flux. In the isolated perfused working heart, an HBP flux of ∼2.5 nmol/g protein/min was measured, which represented 0.003–0.006% of the glycolytic flux. This is several orders of magnitude lower than earlier estimates, most likely because of the high metabolic rate of the heart. Moreover, when increasing glucose concentration from 5 to 25 mM, changes in HBP flux relative to glycolysis occurred as a result of changes in glycolytic, not HBP, flux rates. This illustrates the limitation of evaluating HBP flux as a fraction of glycolysis, as well as the assumption that HBP flux will be similar across biological systems with widely varying metabolic demands. While, much remains to be understood about of the regulation of HBP, the implementation of new techniques will enable direct measures of HBP flux in diverse biological systems.

2.2. O-GlcNAc Transferase

O-GlcNAc transferase (OGT; uridine diphospho-N-acetylglucosamine:polypeptide β-N-acetylglucosaminyltransferase; EC 2.4.1.255) is a soluble glycosyltransferase primarily located in the cytoplasm and nucleus, which is responsible for using UDP-GlcNAc to modify proteins with O-GlcNAc (211, 231–234). OGT is located on the X-chromosome and encodes a multidomain protein containing multiple tetratriopeptide repeats (TPRs) in the NH₂-terminal domain and two catalytic domains in the COOH-terminal region (FIGURE 6A) (211, 231–234). OGT is highly conserved, across all metazoans, and with the exception of zebrafish (237, 238), a single gene encodes OGT. Alternative splicing results in three mammalian isoforms of OGT, which differ primarily by the number of TPRs. A form of OGT, epidermal growth factor domain-specific OGT (EOGT), has also been identified in the ER, but it shares little homology with other forms of OGT. In contrast to OGT, it can also elongate O-GlcNAc into complex glycans (239), as reviewed in detail elsewhere (240, 241).

FIGURE 6.

Structure and posttranslational modifications of human OGT (A) and OGA (B). Phosphorylation sites with identified kinases are shown as red circles, and all known phosphorylation sites are listed below the figures. O-GlcNAc sites are shown as blue squares, and acetylation sites are shown as green circles. AT, acetyl transferase; CaMKII, calcium/calmodulin (Ca²⁺/CaM) dependent protein kinase II; CD, catalytic domain; CHK1, checkpoint kinase-1; GSK3β, glycogen synthase kinase 3β; hOGT, human O-GlcNAc transferase; hOGA, huma O-GlcNAcase; Int-D, intervening domain; IRS1, insulin receptor substrate-1; LC, low complexity region; NLS, nuclear localization sequence; TPR, tetratricopeptide repeats. The databases https://www.phosphosite.org and http://www.phosphonet.ca/, and the citations contained therein were used to for known phosphorylation sites and acetylation modification sites(196). Citations for sites where kinases have been identified are included in the text; additional resources include Lundby, et al. (235), Levine and Walker (232) and Roth et al. (236). Other PTMs not shown include ubiquitination and sumoylation.

The full length or nucleocytoplasmic OGT (ncOGT; 110 kDa) and short OGT (sOGT; 78 kDa) have up to 13.5 and 4 TPRs, respectively; the precise number of TPRs is species specific (232). The mitochondrial OGT (mOGT; 90 kDa) has a mitochondrial targeting sequence at the NH₂ terminus of the protein and 9 TPRs. All three OGT isoforms have the same two catalytic domains, as well as a putative phosphatidylinositol (3,4,5)-trisphosphate (PIP₃)-binding domain (242); however, studies of the crystal structure of human OGT-UDP-peptide complex were unable to confirm the presence of the PIP₃ domain (90). OGT is a member of the GT-B glycosyltransferase superfamily, and it is the only glycosyl transferase to have a ∼120 amino acid sequence in the middle of the catalytic domain, known as the intervening domain (Int-D) (232, 234, 243). The function of this region of the protein remains unclear, although it contains a large number of basic residues, suggesting contacts with negatively charged partners, which may affect cellular localization or protein-protein interactions (232, 243).

The catalytic domain of OGT represents less than half of the total molecular weight of the protein with the remainder of the protein primarily consisting of the TPRs (234). TPRs comprise 34 amino acid motifs clustered in repeats that form α-helical superspirals and have been proposed to play a key role in determining OGT interactions with target substrate proteins (234, 244). The importance of the TPR domain in regulating OGT substrate selectivity was demonstrated by the observation that modification of only two aspartate residues in the TPR substantially changed OGT substrate preference (245). Although most OGT substrates require the presence of the TPR domains for O-GlcNAcylation to occur, some proteins only interact with the catalytic region (131, 232, 234). OGT exists as a homodimer (243, 244, 246), and the TPR regions are also required for dimerization; however, in vitro disruption of this interaction did not change OGT activity (247). The function of OGT dimerization is unclear; however, it could serve to stabilize interactions with specific substrates. For example, mutations in the TPR 6–7 region prevent OGT dimerization, as well as reduce O-GlcNAc levels of nuclear pore glycoprotein p62 (Nup62) (244).

A definitive consensus sequence for O-GlcNAcylation has not been identified; consequently, although our knowledge of OGT structure and its molecular interactions with substrates has improved, our knowledge of the regulation of OGT function and the mechanisms underlying its substrate specificity remains limited (231). Reports demonstrating that interactions between UDP-GlcNAc and the peptide play a role in orientation of the peptide relative to the active site suggest that the active site contributes to substrate selection (248). Moreover, preference by OGT for Ser/Thr residues that have prolines and branched-chain amino acids in close proximity, resulting in an extended peptide orientation, also suggests that the active site imposes some degree of sequence constraint and, thus, substrate selection (249). However, perhaps the most widely accepted view is that OGT substrate selection is largely determined by binding to the TPR domains. This is supported by the fact that the TPR domains are essential for O-GlcNAcylation of most proteins, as well as the fact that specific substrates interact with specific TPR regions. For example, in brain tissue, ataxin-10 (Atx-10) binds to TPRs 6–8 (246), whereas mSIN3A, ten-eleven translocation (TET) 2/3, and trafficking kinesin protein 1 (TRAK1) all require TPRs 1–6 (232). Moreover, mutations of just two aspartate residues in the TPR domain changed OGT activity as well as target specificity (245). Structural studies have revealed a hinge region around TPRs 12–13, which could influence the access of protein substrates to the active site of OGT (232, 234). There is also evidence that under some conditions, sOGT can act as a negative regulator of ncOGT (233, 246). The importance of the TPR domain is reflected in the observations that missense mutations in this region result in decreased OGT function and neurodevelopmental abnormalities (29, 250).

OGT activity and substrate recognition can also be regulated by phosphorylation of OGT on Ser, Thr, and tyrosine (Tyr) residues (233). Almost 20 different phosphorylation sites have been reported on OGT, many identified by large-scale proteomic studies (FIGURE 6A). Although the function of many of these sites remains unknown, as do the kinase(s) responsible for their phosphorylation, a few have been characterized. For example, insulin increases Tyr phosphorylation of OGT via activation of the insulin receptor (IR), resulting in increased OGT activity (251). Glycogen synthase kinase (GSK)-3β phosphorylates OGT on Ser-3/4, leading to increased OGT activity (116), and AMPK phosphorylates Thr-444, resulting in changes in subcellular localization and substrate binding targets (252). In addition, OGT is phosphorylated on Ser-20 by checkpoint kinase 1 (Chk1), leading to stabilization of OGT, which is required for cytokinesis (253). Ser-20 on OGT is also a target for CaMKII, which increases its activity (181). O-GlcNAcylation of Ser-389, located in the TPR domain regulates OGT nuclear localization (254). Serines 3 and 4 on OGT are also sites of O-GlcNAcylation (116); however, the function of this modification remains to be determined. Recent work that has focused on some of these phosphorylation sites has identified that in sOGT, mutation of either Thr-12 or Ser-56 to an alanine significantly altered substrate binding by over 500 proteins (255). OGT is also acetylated on multiple residues (FIGURE 6A) (235). While the effects of this modification are not known, the fact that two of the sites occur within one of the catalytic domains suggests they could regulate OGT activity in some manner.

Targeting of specific proteins by OGT can also occur via its interaction with adaptor or scaffold proteins, which recruit substrates on OGT. For example, in the liver during fasting host cell factor 1 (HCF1) targets OGT to peroxisome proliferator-activated receptor γ coactivator 1α (PGC-1α), where increased O-GlcNAcylation increases PGC-1α stability and upregulates gluconeogenic genes (118). Other such interactions include p38 MAPK, which recruits OGT to neurofilament-heavy polypeptide (NFH) (256), REV-ERBα, which prevents OGT degradation (65), and OGA, which increases interactions between OGT and pyruvate kinase isoform M2 (PKM2) (257). The potential for an OGT-OGA interaction raises the possibility of an “O-GlcNAczyme” complex, which would be consistent with rapid and reversible changes in O-GlcNAcylation (258). Insulin treatment results in translocation of OGT from the nucleus to the cytosol and plasma membrane, and changes in nutrient availability also lead to redistribution of OGT between the nucleus and cytosol (242, 251). The three OGT isoforms also exhibit differences in their subcellular localization, and along with differences in the TPR regions between the isoforms, different proteome subsets can result; however, this has only been demonstrated with in vitro studies (169).

In addition to its role as a glycosyl transferase, OGT also exhibits protease activity, although, to date, only one proteolytic substrate, the transcriptional coregulator HCF1, has been identified (259, 260). To fully function as a coregulator HCF1 is required to undergo proteolytic cleavage; however, the mechanism by which this occurred remained elusive until 2011 when two reports demonstrated that OGT played a key role in this process (259, 260). These studies demonstrated that both UDP-GlcNAc and OGT were required for HCF1 proteolysis, consistent with the notion that O-glycosylation of HCF1 was necessary. It was initially thought that OGT had a specific protease active site; however, subsequent studies demonstrated that the COOH-terminal of HCF1 binds to the TPR domain of OGT, and the cleavage region in the glycosyltransferase active site was similar to that for regular glycosylation substrates (261). Moreover, these studies also showed that UDP-5SGlcNAc, which binds to the OGT-active site in the same manner as UDP-GlcNAc, but is resistant to glycosyltransferase, inhibited HCF1 proteolysis. The role of OGT mediated cleavage of HCF1 remains unclear; however, it could represent a link between cellular metabolism and the regulation of cell cycle, which is supported by reports that the OGT/HCF1 complex itself is an important regulator of glucose metabolism (118).

2.3. O-GlcNAcase

O-GlcNAcase (OGA, EC 3.2.1.169) is a hexosaminidase that was first characterized in 1994 by Dong and Hart (11) following purification from rat spleen; it has also been referred to as NCOAT, nuclear cytoplasmic O-GlcNAcase and acetyltransferase (262), GCA (263), MEA5, and MGEA5 (264). Dong and Hart (11) demonstrated that it was distinct from lysosomal hexosaminidases by virtue of its subcellular localization to the cytosol and nucleus, its neutral pH optimum, and its selectivity for GlcNAc over GalNAc. Subsequent sequence identification and cloning demonstrated that OGA was identical to a previously identified protein, MGEA5, found in meningioma patients (264). The human OGA gene is located on chromosome 10 and encodes a protein with an NH₂-terminal glycosyl hydrolase domain and a COOH-terminal domain that demonstrates sequence homology to histone acetyl transferase (AT) proteins (FIGURE 6B). The glycosyl hydrolase domain has two low-complexity or disordered regions [i.e., repeats of single amino acids or short amino acid motifs (265)], on either side. The larger low-complexity region is followed by the stalk domain, which connects to the COOH-terminal domain (231, 266). Between the catalytic domain and the COOH-terminal region is a noncanonical caspase cleavage site (267). It has been reported that OGA exhibited AT activity (199, 268); however, this has not been replicated by others (269). Recent structural studies found that human OGA (hOGA) lacks the residues necessary for binding to acetyl-CoA and described this region as a “catalytically incompetent pseudo-AT domain” (270). Alternative splicing results in the generation of a short OGA (sOGA), which lacks the pseudo-AT region and has a different 15-amino acid COOH-terminal region (271). sOGA has lower hexosaminidase activity than hOGA in vitro (56, 272) and appears primarily localized to the surface of lipid droplets (273), although its specific function remains to be elucidated.

For many years, full-length hOGA proved resistant to crystallization, consequently, most of the initial structural and mechanistic information were derived from bacterial homologs of OGA (274). In 2017, three independent studies reported the crystal structure of hOGA using different, catalytically functional, truncated versions of the protein (236, 275, 276). A key and unexpected finding in all three studies was that hOGA formed an obligate homodimer in which the stalk domain of one monomer is positioned over the catalytic site of the other monomer (266). One consequence of this arrangement is the creation of a substrate binding site, comprising conserved hydrophobic residues, which supports the notion that the dimer is the active form of OGA. Further structural analysis suggests OGA may preferentially remove O-GlcNAc from certain sites, suggesting that OGA may be an equal partner with OGT in regulating O-GlcNAc turnover (277). Moreover, it has also been reported that a number of specific residues on OGA contribute to its interactions with different peptide substrates, which has implications for differential regulation of O-GlcNAcylation on different proteins (278,279). While the active site of OGA is now better characterized than ever before, other regions, including the low-complexity region and the pseudo-AT domain remain poorly understood (266).

Similar to OGT, OGA is also subject to both phosphorylation and O-GlcNAcylation. At least 20 different Ser, Thr, and Tyr phosphorylation sites have been mapped by MS in both the glycosyl hydrolase and pseudo-HAT domains (FIGURE 6B); however, the effects of these modifications on OGA activity have not been determined. Interestingly, the Ser-405 O-GlcNAcylation site on OGA, is in the central low-complexity region, which is also the region where OGA interacts with OGT (266). Thus, O-GlcNAcylation of this residue plays a role in the regulation of the interaction between OGT and OGA. OGA is also acetylated in the stalk domain at Lys-599 (235).

2.4. Maintenance of O-GlcNAc Homeostasis

Changes in cellular O-GlcNAc levels occur in response to a diverse array of physiological and pathological stimuli, and dysregulation of O-GlcNAc homeostasis has been linked to a wide array of diseases, as discussed in detail later. Deletion of the OGT gene is embryonically lethal (208), and in cell culture, mouse embryonic fibroblasts (MEFs) die around 4–5 days after OGT is knocked out (280). Although OGA knockout mice survive to birth, the majority die before weaning (17). Chronic increases in O-GlcNAc levels induced by overexpression of a dominant negative OGA (dnOGA) in a tissue-specific manner results in apoptosis in skeletal muscle (205) and altered metabolism in the heart (281). Therefore, it is evident that maintenance of O-GlcNAc homeostasis is essential for the normal physiological function of cells and tissues. This has led to the concept that there is an optimal range of O-GlcNAcylation that supports normal physiological processes and that outside that range, either too high or too low, cellular dysfunction results. Consequently, it has been proposed that to keep within this optimal range, OGT and OGA work together to form a “buffering” system that can respond to moderate changes in O-GlcNAcylation (257).

The prevailing wisdom for many years was that the primary mechanism regulating cellular O-GlcNAc levels was nutrient availability, so that under conditions of nutrient excess, such as hyperglycemia, O-GlcNAc levels increased, and if glucose levels dropped O-GlcNAc levels would decrease. This concept is consistent with the fact that glucosamine, which bypasses GFAT, leads to uncontrolled synthesis of UDP-GlcNAc and increased O-GlcNAc levels (24, 271, 282, 283). This also enables glucosamine to be used pharmacologically to increase UDP-GlcNAc synthesis and O-GlcNAc levels. Acute changes in glucose availability either by increasing exogenous glucose or via insulin-stimulated increases in glucose uptake can have little or no effect on cellular O-GlcNAc levels (284). This is also consistent with the report that a five-fold increase in glucose had no effect on HBP flux in the isolated perfused heart (230). However, such responses likely vary greatly, depending on cell type and duration of treatments (285). In addition, OGT activity is very sensitive to UDP-GlcNAc concentrations, exhibiting multiple apparent K_m values for UDP-GlcNAc under varying UDP-GlcNAc concentrations (247). Consequently, changes in UDP-GlcNAc concentrations can directly influence global as well as regional OGT activities and O-GlcNAc levels.

However, while nutrient availability as a regulatory mechanism is a valuable concept, it does not address the mechanisms that underlie widely differing rates of changes in O-GlcNAcylation that occur in response to different stimuli. For example, depolarization of neuroblastoma cells dramatically increased OGT activity, reaching a peak within 1 min, leading to increased O-GlcNAc levels (286), and stimulation of human neutrophils leads to an approximately five fold increase in O-GlcNAc within 30 s, returning to normal levels over the next 5–10 min (287). On the other hand, stress, such as ischemia (288) or heat shock occur over periods of minutes to hours, induced increases in O-GlcNAc levels (289). Conversely, in certain disease states, such as cancer (21, 290), diabetes (283, 290), or cardiac hypertrophy (291,292), O-GlcNAc levels are chronically elevated, and the mechanisms that lead to this resetting of steady-state O-GlcNAc levels outside of the normal range remains poorly understood (293). As discussed above, GFAT, OGT, and OGA are all modified by phosphorylation, indicating that these posttranslational modifications may contribute to the dynamic regulation of O-GlcNAcylation.

Given that HBP flux and O-GlcNAc levels are modulated, in part, by nutrient availability, it is perhaps not surprising that O-GlcNAc levels can also be regulated by nutrient-regulating hormones. Insulin is probably the most studied of these, and it has been shown that insulin treatment recruits OGT from the nucleus to the plasma membrane, where it is phosphorylated by the IR (242, 251, 294). This phosphorylation increases OGT activity and leads to O-GlcNAcylation of insulin receptor substrate-1 (IRS1) and AKT, resulting in attenuation of insulin signaling. Treatment of HepG2 cells with the adipokine leptin resulted in an approximately twofold increase in overall O-GlcNAc levels within 15 min with parallel increases in GFAT protein levels (295). However, the mechanism by which leptin stimulated GFAT expression and O-GlcNAc levels was not identified. Ghrelin, a hormone that is released in response to fasting, increases O-GlcNAc levels in hypothalamic appetite-stimulating agouti-related protein (AgRP) neurons, leading to their increased firing rate (201). In the liver, glucagon, which increases in response to fasting, increased OGT phosphorylation and O-GlcNAc levels in a CaMKII-dependent manner (181).

In addition to nutrient-dependent hormones, G protein-coupled receptor agonists have also been shown to increase cellular O-GlcNAc levels. For example, activation of the endothelin (ET)_A receptor with ET-1 resulted in a time-dependent increase in O-GlcNAc levels in vascular cells, and the subsequent downstream signaling by ET-1 was dependent on the increase in O-GlcNAc levels (296, 297). Phenylephrine (PE), primarily an α-receptor agonist, also increased O-GlcNAc levels in cardiomyocytes, and the increase in O-GlcNAc was required for subsequent activation of PE-mediated signaling pathways (298, 299). One explanation for PE-dependent increases in O-GlcNAc levels was an increase in GFAT expression (298, 299). Another study suggested that the increase was mediated by the Ca²⁺ dependent CaMKII/calcineurin pathway (298).

The transcriptional regulation of OGT and OGA has been understudied; however, there are a number of reports indicating that they regulate each other and that O-GlcNAcylation itself might contribute to their transcriptional regulation (257), further discussed in Sect. IIIA and IIIB. For example, in OGT knockout MEFs the time course of loss of OGT was paralleled by a decrease in OGA (280). Increasing O-GlcNAc levels via OGA inhibition demonstrated that OGA transcription was O-GlcNAc dependent (300). Other studies have reported that low O-GlcNAc levels contribute to increased OGT transcription (68). The strongest evidence to date of mutual regulation of OGT and OGA was by Qian et al. (301), who reported that overexpression of OGA resulted in increased OGT transcription, whereas knockdown OGA with siRNA significantly decreased OGT levels. Using promoter luciferase reporters for both OGT and OGA, they clearly demonstrated reciprocal transcriptional regulation (301). They also showed that OGA cooperated with p300m, a histone acetyltransferase and the transcription factor CCAAT/enhancer-binding protein-β (C/EBP-β) to promote OGT transcription (301). E2F transcription factor 1 (E2F1), which contributes to the activation of many genes, was found to be a repressor of both OGT and OGA (302). It has been shown that E2F1 activity itself might be regulated by O-GlcNAcylation, illustrating another mechanism by which O-GlcNAc can regulate the transcription of OGT and OGA (303). The OGT promoter contains a TATA box, which likely facilitates OGT transcription, whereas the OGA transcription is not dependent on a TATA box (302). In macrophages, Cullin 3 (CUL3), an E3 ubiquitin ligase, has been reported to downregulate OGT expression in a nuclear factor E2-related factor-2 (Nrf2)-dependent manner (304).

Another potential mechanism for maintaining O-GlcNAc homeostasis is alternative splicing of both OGT and OGA, resulting in intron retention, which is regulated by changes in O-GlcNAc levels (68, 305). Consequently, when O-GlcNAc levels are high, nuclear retention of OGT due to intron retention increases, whereas low-O-GlcNAc levels decrease this process; thereby, decreasing or increasing OGT protein, respectively (68, 305). Conversely, low O-GlcNAc levels increase nuclear retention of OGA (305). It is noteworthy that these changes occur relatively rapidly and, as such, represent a potentially important mechanism in O-GlcNAc-mediated regulation of O-GlcNAc homeostasis (305). A number of microRNAs (miRs), including miR-101, 200a/b, 423-5p, 501-3p, 539, and 619-3p, have also been shown to regulate O-GlcNAc levels by targeting OGT (306–309) or OGA (310). Consequently, although our knowledge of the transcriptional regulation of OGT and OGA is improving, the physiological role of these pathways remains to be determined.

One of the more puzzling aspects of O-GlcNAc homeostasis is that glucose deprivation leads to a marked increase in overall cellular O-GlcNAc levels. This was first reported in HepG2 cells where it was observed that 12 h following the removal of glucose, there was approximately eight-fold increase in O-GlcNAc levels, which was accompanied by a ∼40% decrease in UDP-GlcNAc levels, suggesting that increased HBP flux was not contributing to the elevated O-GlcNAcylation (311). A later study came to a similar conclusion, as the addition of 50–100 µM glucosamine blocked the O-GlcNAc increase, resulting from glucose deprivation (312). Contrary to this finding was a report that glycogen degradation triggered by glucose deprivation provided the substrate for O-GlcNAcylation (313). Moreover, ATF4, a regulator of the unfolded protein response (UPR), increased GFAT1 expression in response to glucose deprivation, and ATF4 inhibition or knockdown prevented the increase in O-GlcNAc and GFAT1 (221). They also suggested that ATF4, along with another UPR-related protein X-box binding protein 1 (XBP1), mediated the steady-state levels of GFAT1 (221). This is consistent with an earlier report demonstrating that stress-induced increases in O-GlcNAc were mediated via XBP1 increases in GFAT1 protein levels (314). Glucose deprivation increased mRNA levels of both OGT and OGA; however, this was not associated with an increase in the levels of either protein (312). Interestingly, extracellular Ca²⁺ was required for this increase in O-GlcNAc levels, and inhibition of CaMKII blunted the response to glucose deprivation (312).

Historically, nutrient availability was considered to be the primary regulator of cellular O-GlcNAc homeostasis; however, it is increasingly evident that it is only one of many factors. Our knowledge of transcriptional regulation is growing, but it remains limited, and the role of phosphorylation in regulating GFAT and OGT activity is improving; however, the role of phosphorylation in regulating OGA activity is underexplored. Several lines of evidence suggest that Ca²⁺-mediated activation of CaMKII contributes to O-GlcNAc homeostasis, which would be consistent with rapid agonist-induced increases in O-GlcNAc being independent of nutrient availability or transcriptional regulation of GFAT1, OGT, and OGA.

2.5. Cross Talk Between O-GlcNAcylation and Phosphorylation

As the dynamic nature of O-GlcNAcylation was recognized, there was speculation that it might play a regulatory role in protein function. Once it was recognized that O-GlcNAc modified Ser and Thr residues, which are also potential sites of phosphorylation, there was increasing consideration about possible interactions between O-GlcNAc and phosphorylation on proteins (211, 293). One concept that gained early popularity was the possibility of reciprocity between O-GlcNAcylation and phosphorylation. In other words, that a specific residue on a protein could be modified by either O-GlcNAc or phosphorylation became more widely known as the “Ying-Yang” hypothesis (315). There are, indeed, several proteins that support this concept, such as cMyc (Thr-58) (98), estrogen receptor (ER)-β (Ser-16), and endothelial nitric oxide synthase (eNOS) (Ser-1177) (190). Alternatively, modification of adjacent sites can negatively interact with each other, for example, histone deacetylase 4 (HDAC4) is O-GlcNAcylated at Ser-642, and this blocks CaMKII-mediated phosphorylation of Ser-632 (316). Consequently, it is becoming increasingly clear that the cross talk between O-GlcNAc and phosphorylation is much more complex than first thought (FIGURE 7). A high-throughput proteomic analysis estimated that in only 8% of O-GlcNAcylated proteins was the same residue modified by phosphorylation (182). Some proteins are modified by both O-GlcNAc and phosphate, but not at the same sites; for example, increases in phosphorylation of Thr-200 of CaMKIV decreases overall O-GlcNAcylation at several sites, including Ser-189, and conversely, prevention of Thr-200 phosphorylation increased CaMKIV O-GlcNAcylation (167). In myosin light chain 1, the O-GlcNAc-modified sites are Thr-93 and Thr-164, whereas the phosphorylation sites are at Thr-69 and Ser-200 (147).

FIGURE 7.

Extensive crosstalk between phosphorylation and O-GlcNAcylation. A, i: example of phosphorylation and O-GlcNAcylation of the same or adjacent sites that prevent simultaneous modifications. and A, ii: modification by both phosphorylation and O-GlcNAcylation, as well as O-GlcNAcylation enhancing phosphorylation via for example increased binding to adaptor proteins. B, i: example of a kinase that is activated when phosphorylated and inactivated by O-GlcNAcylation. B, ii: a kinase that can be modified by both O-GlcNAc and phosphorylation involving complex interactions between OGT/OGA and kinases/phosphatases. C: O-GlcNAcylation also regulates phosphorylation via modification of phosphatases, and OGT activity is directly regulated by kinase phosphorylation. GSK3β, glycogen synthase kinase 3β; OGA, O-GlcNAcase; OGT, O-GlcNAc transferase; PPT, protein phosphatase.

The intricate relationship between O-GlcNAcylation and phosphorylation was demonstrated by a study in which GSK3β was inhibited and O-GlcNAcylation levels of specific proteins were quantified. Of the 45 O-GlcNAcylated proteins identified, 10 exhibited an increase in O-GlcNAcylation, whereas 19 showed a decrease (317). In another study in which O-GlcNAc levels were increased by inhibition of OGA, increased phosphorylation was observed on 148 sites and lower phosphorylation was observed at 280 sites (318). A comparison between wild-type and OGT-null cells identified 232 phosphosites that were upregulated and 133 phosphosites that were downregulated 36 h after OGT deletion (319). Interestingly O-GlcNAc modifications have been reported to occur in clusters (182), a phenomenon that is also observed with phosphorylation (320). Therefore, it is evident that the cross talk between these two posttranslational modifications is complex, and it is currently not possible to predict a priori how they will interact on individual proteins. However, specific motifs have been identified, which when phosphorylated, strongly inhibit O-GlcNAcylation and vice versa; whereas, sites that are phosphorylated by proline-directed kinases do not appear to be subject to O-GlcNAcylation (321).

Recent studies have demonstrated that OGT and OGA form complexes with both kinases and phosphatases (322–324) demonstrating that changes in O-GlcNAcylation and phosphorylation could be occurring simultaneously on a target protein (FIGURE 7A). There is also a growing list of kinases that have been shown to be O-GlcNAcylated and that this modification directly affects their function (see TABLE 2 and FIGURE 7B). An analysis of glycoproteomic data sets reported that more than 100 kinases contain identified O-GlcNAcylation sites, emphasizing the importance of O-GlcNAc in regulating kinase function (325). A study of synaptic proteins found that kinases were more frequently a target for O-GlcNAcylation than other proteins; however, the specific modification sites were frequently outside the catalytic domain of the kinases in question (182). Thus, protein O-GlcNAcylation can alter phosphorylation either via direct modification of phosphorylated proteins, but also by regulation of kinases that are responsible for phosphorylation. Although there is less known about O-GlcNAcylation of phosphatases, protein tyrosine phosphatase 1B (PTP1B) has been shown to be O-GlcNAcylated at Ser-104, Ser-201, and Ser-386, resulting in increased enzymatic activity (176).

One specific example of the complex regulation between kinases and O-GlcNAcylation is CaMKIV, which is also known to activate OGT (286) (FIGURE 7C). Activation/deactivation of CaMKIV is rapid and tightly regulated beginning with displacement of protein phosphatase 2α (PP2A) and subsequent phosphorylation of Thr-200 by CaMKK; subsequent inactivation, involves the reassociation with PP2A and reduction in phosphorylation. Dias et al. (167) found that CaMKIV contained at least five O-GlcNAc modification sites and that during its activation, O-GlcNAc levels rapidly decreased as the interaction with OGA increased. Following inactivation, CaMKIV O-GlcNAc levels returned to normal, suggesting that OGT was recruited to CaMKIV; however, a direct interaction was not identified (167). Mutation of Thr-200 to alanine, thereby, preventing phosphorylation, resulted in an increase in CaMKIV O-GlcNAcylation; conversely, mutation to glutamate to mimic phosphorylation led to lower O-GlcNAc levels, demonstrating a direct interaction between Thr-200 and O-GlcNAcylation. Of the five O-GlcNAc sites identified, three Ser-189, Thr-57, and Ser-58, when mutated to alanine to prevent O-GlcNAcylation, markedly reduced Thr-200 phosphorylation. The double-mutant T57A/S58A resulted in no measurable kinase activity (167). Conversely, the S189A mutant, not only dramatically reduced O-GlcNAc levels, but also increased basal kinase activity. O-GlcNAcylation of CaMKII has also been shown to increase its kinase activity (165).

AMPK is another key example of a kinase that both regulates O-GlcNAc levels, as well as being regulated by O-GlcNAc itself, as reviewed in detail (219). As discussed earlier, GFAT is phosphorylated on Ser-243 by AMPK, resulting in decreased GFAT activity and lower O-GlcNAc levels (130, 218). Moreover, activation of AMPK leads to increased GFAT phosphorylation and decreased O-GlcNAc levels (299). AMPK also targets OGT by phosphorylating Thr-444; however, rather than alter its activity, this phosphorylation leads to changes in OGT cellular localization and substrate specificity (252). Interestingly, although OGA has not been shown to be a target for AMPK, deletion of the AMPKα2 isoform results in lower OGA protein levels and increased O-GlcNAc levels, without changes in either OGT or GFAT (299). Bullen et al. (252), demonstrated that all α- and γ-subunits of AMPK are O-GlcNAcylated and further that activation of AMPK increased O-GlcNAcylation of the γ₁-subunit. They also reported that inhibition of OGA attenuated physiological and pharmacological activation of AMPK.

Although Ser and Thr residues are often the focus of discussions of cross talk between O-GlcNAcylation and phosphorylation, it is also clear that Tyr phosphorylation and O-GlcNAcylation also influence each other (326). An analysis of a small set of O-GlcNAcylated proteins concluded that >60% of them were also Tyr phosphorylated (326). The potential importance of interactions between O-GlcNAc and Tyr phosphorylation was first recognized, when it was demonstrated that insulin stimulates Tyr phosphorylation of OGT via the IR, thereby increasing OGT activity (251). A subsequent study revealed that increased O-GlcNAc levels reduced Tyr phosphorylation of IRS-1 (327). Studies on prohibitin, reported that O-GlcNAcylation attenuated its Tyr phosphorylation (328). Using peptides that contained residues that could be both Tyr phosphorylated and O-GlcNAcylated, Ande et al. (328) showed that while O-GlcNAcylation reduced Tyr phosphorylation, an increase in Tyr phosphorylation actually enhanced O-GlcNAcylation. Another study using peptide microarrays concluded that Tyr phosphorylation may have a greater effect on the regulation of O-GlcNAcylation than O-GlcNAcylation on phosphorylation (329).

Here, the focus has been on interactions between phosphorylation and O-GlcNAcylation; however, O-GlcNAcylation also interacts with other PTMs, although these are less widely studied (257). For example, both OGT and OGA have been found to be ubiquitinated, and O-GlcNAc has been shown to stabilize proteins by inhibiting their ubiquitination (330, 331). RelA/p65, a member of the nuclear factor κ-light-chain enhancer of activated B cells (NF-κB) family of transcription factors, is both O-GlcNAcylated (Thr-305, Ser-319, Ser-337, Thr-352, and Ser-374) and acetylated (lysine-310). Acetylation is required for full transcriptional activity of RelA/p65, and O-GlcNAcylation at Thr-305 and Thr-315 promotes its acetylation at lysine-310 (332, 333). In addition, both OGT and OGA are also acetylated, although the effect of this modification on their function is not known (FIGURE 6) (235). Of note, HDAC4 has been shown to be modified by O-GlcNAc on Ser-642, providing further evidence of direct interactions between O-GlcNA cylation and acetylation (316). As further discussed in the next section, cross talk between acetylation and O-GlcNAcylation has also been implicated in epigenetic regulation (334).

3. O-GlcNACYLATION AND CELLULAR FUNCTION

In the prior sections, we discussed the discovery of O-GlcNAcylation, its unique place in glycobiology, and the numerous pathways that impact its regulation. In this section, we will focus on the cellular functions of regulation by O-GlcNAcylation, including modulation of gene expression at both the levels of transcription, as well as via epigenetic mechanisms, followed by specific examples of how O-GlcNAcylation impacts cellular signaling (e.g., insulin and calcium), metabolism (e.g., mitochondria), and survival (e.g., autophagy).

3.1. Transcription

Gene expression is controlled at a number of levels (FIGURE 8). First, at the level of DNA sequence are response elements that recruit transcriptional regulators that include numerous families of transcription factors. However, the regulation of gene expression is intricately controlled beyond proper recruitment of transcriptional machinery to the promoter regions of genes that contain these response elements. In particular, the direct modification of transcription factors and the transcriptional machinery can be either inhibitory or activating in altering RNA levels. One of the first molecular functions to show biological relevance of O-GlcNAc-mediated regulation is modification of these transcription factors (210). The early studies started with an initial focus on O-GlcNAcylation of the ubiquitously expressed Sp1, a zinc finger transcription factor that binds GC-rich motifs of many promoters (FIGURE 8A, I). Work by Jackson and Tjian (335) found that increased O-GlcNAcylation on Sp1 from either Drosophila or human cells increased its transcriptional activity. The mechanism of this transcriptional activation was identified by Kudlow and colleagues (336, 337), who reported that the O-GlcNAc modification protected Sp1 from proteasomal degradation. In turn, this allowed O-GlcNAc modification of Sp1 to act as a nutritional checkpoint for times of inadequate nutrients, hypoglycosylation, and resource sparing through decreased transcription (336). Subsequent studies on the mechanisms of O-GlcNAc action on transcriptional regulation also found that within this same factor, Sp1 could be O-GlcNAcylated at its activation domain to suppress transcriptional activity (109), suggesting multiple and opposing roles for this single modification depending on context. These mechanistic insights provided important clues as to how diabetes, for example, could disrupt cellular processes of signaling via decreased Sp1 transcriptional activation. In the context of diabetes, O-GlcNAc modification of Sp1 was further shown to modify transcriptional activity because of a dynamic interplay with its phosphorylation status, which further alters Sp1 subcellular compartmentalization and activity (338) and may impact transcriptional regulation of mitochondrial function (339). Additionally, Sp1 O-GlcNAcylation may simply interfere with its interaction with other transcription factors, such as Elf-1 (340), NF-Y (341), Oct1 (342), as well as Sp3 and Sp4 (343), highlighting the diverse set of mechanisms by which O-GlcNAcylation of transcription factors mediate gene expression.

FIGURE 8.

O-linked β-N-acetylglucosamine (O-GlcNAc) regulation of gene expression and epigenetics. A: examples of transcriptional regulation mediated by O-GlcNAc with additional details and references highlighted in the text. These examples include both inhibitory and activating roles of O-GlcNAc (blue square with G). A, I: for the transcription factor Sp1, one O-GlcNAc modification blocks an activation site to inhibit transcription, while a different O-GlcNAc site inhibits proteasomal degradation increasing transcription. A, II: a separate set of examples of this dual role of O-GlcNAc-mediated regulation is at ChREBP, which, under normal glucose levels, leads to O-GlcNAc modification (blue square) to enhance 14-3-3 binding decreasing transcriptional activity, while under high glucose levels, additional posttranslational regulation by phosphorylation (red circle with P) leads to a different O-GlcNAc modification and augments activity. A, III: multiple roles are shown for both direct O-GlcNAc modification of RNA polymerase II (RNAP II), as well as the contribution of uridine diphosphate-azido-modified galactose (UDP)-GlcNAc hydrolysis to enhance transcriptional activity. A, IV: multiple roles are shown for different O-GlcNAc modifications to either inhibit ubiquitination (gray square with U) to increase complex stability or other modifications that may impact nuclear cytoplasmic import/export of cargo (e.g., RNA). B: examples of epigenetic regulation mediated by O-GlcNAc with additional details and references highlighted in the text. B, I: highlights the multiple direct (i.e., O-GlcNAc of histone proteins) and indirect (i.e. interaction and modification of other histone modifiers), such as EZH2 to impact histone methylation (black square with M) and SIN3a to impact histone acetylation (purple square with A). B, II: highlights the interaction of O-GlcNAc transferase (OGT) with the ten-eleven translocation (TET) enzymes impacting DNA 5 hydroxymethylation (5hmC) which, in turn, can lead to DNA demethylation. B, III: a few examples are highlighted from the text of noncoding RNA [ncRNA; e.g., long-noncoding RNA (lncRNA) and microRNA (miR)] regulation of O-GlcNAc enzymes.

Although much of the early work was done on O-GlcNAc modification of Sp1, a number of subsequent studies have identified a continuously growing list of transcription factors with a direct O-GlcNAc modification and functional consequence. This includes additional key transcriptional regulators of insulin signaling and metabolism, such as brain and muscle ARNT-like 1 (BMAL1) (344), carbohydrate-responsive element-binding protein (ChREBP) (345), Forkhead Box O1 (FOXO1) (105, 346), liver X receptor (LXR)α (111, 347), PGC-1α (117), and peroxisome proliferator-activated receptors γ (PPARγ) (119), as well as a growing number of transcription factors involved in cancer: GLI (348), HIC1 (349), LXRα/β (350), and NF-κB (333), and many others (TABLE 2). In the case of ChREBP (FIGURE 8A, II), O-GlcNAcylation can again have both activating, as well as inhibitory roles (345). In the context of high glucose conditions, parallel phosphorylation enhances O-GlcNAc levels to maintain transcriptional activity, which was consistent with earlier findings showing higher hepatic ChREBP O-GlcNAcylation and transcriptional activity in diabetic mice (351). However, under normal glucose conditions ChREBP is O-GlcNAc modified at a different Ser residue, which increases its interaction with other factors, such as 14-3-3, resulting in nuclear exclusion and decreased transcriptional activity (345). These findings demonstrate the dynamic, and sometimes opposing, roles that O-GlcNAcylation can have on transcriptional regulation. Although not comprehensive, these examples, and numerous others, solidify our knowledge that modification of transcription factors by O-GlcNAcylation is responsible for changing transcriptional activity through altered DNA binding, localization, stability, and interaction with other coregulators (352).

The second level of transcriptional regulation is on the transcriptional machinery itself. Although the earliest studies were focused on the modification of the transcription factors, it was also recognized by the early 1990s that RNA polymerase II (RNAP II) is a direct target of protein O-GlcNAcylation at its COOH-terminal domain (CTD) (353). Subsequently, this observation was expanded into the idea that RNAP II posttranslational regulation occurred as mutually exclusive states of phosphorylation and O-GlcNAcylation to establish different functional states of this transcriptional regulator (354) (FIGURE 8A, III). Later, Ranuncolo et al. (355), defined the functional significance of this O-GlcNAc cycling on RNAP II as a critical regulatory circuit of the assembly of the transcriptional preinitiation complex (PIC), a mechanism that was further extended to suggest that the hydrolysis of UDP-GlcNAc may actually serve as a high-energy donor to facilitate PIC formation and the elongation step (356,357). Owing to the repetitive nature of the CTD domain of RNAP II, the O-GlcNAcylation of RNAP II additionally allows for a series of highly heterogeneous glycoforms, providing the potential to regulate transcription in response to fluctuating cellular conditions (358). This mechanism has subsequently been suggested to perform a nutrient-sensing role to buffer transcriptional activity to match environmental metabolic demands, as well as developmental milestones (32).

Other transcriptional machinery shown to be modified by O-GlcNAcylation include TATA-binding proteins (TBP) (359), topoisomerase I (Topo I) (360), and nucleoporins (NUPs) of the nuclear pore complex (NPC) (361) (FIGURE 8A, IV). Although the roles of O-GlcNAcylation on TBP or Topo I are generally understudied, the presence of O-GlcNAc on NUPs is extensive and has been widely studied (5,6, 362). In the case of TBP, it has been suggested that O-GlcNAc directs the cycling of this protein on and off promoter regions during the regulation of transcription. While the regulation of Topo I by O-GlcNAc appears to directly mediate DNA relaxation and, therefore, transcriptional accessibility to genes. On the other hand, modification of the NPC by O-GlcNAcylation has growing evidence to support a direct role in regulation of nucleocytoplasmic transport (363). Furthermore, the relatively high steady-state levels of O-GlcNAcylation on components of the NPC appear to preserve the integrity of the complex by preventing its ubiquitination and degradation (364). Extending beyond this role in trafficking, components of both UDP-GlcNAc synthesis and protein O-GlcNAcylation appear to involve the NPC toward regulation of speckle and paraspeckle formation, suggesting additional roles in gene expression (365), as it relates to chromatin structure.

3.2. Epigenetics and Chromatin Remodeling

Epigenetics is defined as the changes that occur above the genome to alter transcriptional regulation in a manner responsive to other events (366); in this way, it is a logical continuation of the O-GlcNAc-mediated control of gene expression described above. More specifically, epigenetics encompasses three areas (FIGURE 8B): 1) histone modifications (e.g., O-GlcNAcylation and acetylation), 2) DNA modifications (e.g., 5-mC and 5-hmC), and 3) noncoding RNA (e.g., miR, lncRNA). Direct O-GlcNAc modification of histones combined with the cross talk between epigenetics and the machinery involved in O-GlcNAcylation (e.g., OGT, OGA) have revealed this to be another important level of O-GlcNAc regulation as discussed here.

Sakabe and colleagues (367) were the first to demonstrate that the histone proteins themselves were directly O-GlcNAcylated. Despite some evidence questioning the relative abundance of direct O-GlcNAc modification of histones in mammalian cells (368), a robust and growing number of studies have mapped histone O-GlcNAcylation (80, 369–373). The role of this histone modification is still being determined, with recent evidence suggesting that it mediates cell cycle progression by suppressing histone H3 phosphorylation (369, 373). Alternatively, O-GlcNAcylation of histone H2B appears to facilitate ubiquitin ligase binding, and further histone modification in a mechanism proposed to initiate transcriptional activation (80). Additional sites of O-GlcNAcylation have been identified on histone H2B with unknown consequences (370). While still other histone variants have been identified with O-GlcNAcylation sites, including histone H2A (371,372), a modification that is in parallel with H2A phosphorylation and inverse to H2A acetylation. Clearly, more work in this area needs to be completed as O-GlcNAcylation has been identified on all four core histone proteins (374).

Another interesting component of the influence O-GlcNAcylation has on histones, is its interaction with other posttranslational mechanisms of histone modification (FIGURE 8B, I). Specifically, it was shown that OGT directly interacts with the histone deacetylase and transcriptional corepressor, SIN3a (375). This and other observations directly connected histone O-GlcNAcylation with acetylation, an observation that was subsequently shown to be regulated by both pathological conditions, as well as physiological stresses such as exercise (376). A second potential link between these two posttranslational modifications came from the identification of a putative histone acetyltransferase activity in OGA (199), an observation that has subsequently come under debate (270). However, the link between these two modifications continues to be explored with a more recent study finding that OGT interacts with the histone acetyltransferase nonspecific lethal (NSL) to alter histone H4 acetylation through complex stabilization (377). The influence of O-GlcNAc on other posttranslational modifications continues to expand and provide novel insights into transcriptional regulation with its interaction with protein methylation. Specifically, OGT can O-GlcNAcylate the histone methyltransferase enhancer of Zeste homolog 2 [EZH2, (378)], which, in turn, regulates gene expression involved in skeletal muscle insulin sensitivity (379), tumor suppression (380), as well as neuronal memory formation (381). It is clear that we are only at the beginning of uncovering the cross-talk between O-GlcNAcylation and the histone code.

DNA modification by 5-methylcytosine was one of the first epigenetic modifications recognized in the early 50s (382), with direct enzymatic regulation identified by the early 60s (383). We now have a relatively robust understanding of the addition of methylation to CpG sites within the genome (384); however, DNA demethylation is more diverse and less well understood (385). In 2009, two articles were published on the discovery and characterization of TET, methylcytosine-dioxygenase, enzymes responsible for active DNA hydroxymethylation (5hmC) and demethylation (386, 387). Subsequently, it was shown that OGT interacted with TET protein providing a connection between DNA modifications and O-GlcNAcylation (388–391) (FIGURE 8B, II). The role by which this interaction influences gene expression varies by the context and TET family member. For example, TET2 interacts with OGT to mediate O-GlcNAcylation of histone H2B (388). However, when OGT interacts with TET1, DNA 5hmC appears to increase (391). Interestingly, these interactions can be part of a larger complex, including TET2/TET3/OGT and HCF1, further bringing in regulation of histone methylation to mediate transcriptional activation (389). OGT, in turn, can O-GlcNAc modify each of the three TET enzymes, reducing phosphorylation, which could alter their activity. Therefore, this interaction between O-GlcNAc and DNA modifications appears to be an important additional area of study that is beginning to provide important insight linking a number of epigenetic mechanisms.

The last, and least explored, area connecting O-GlcNAcylation and epigenetics is that of noncoding RNA (ncRNA) (FIGURE 8B, III). Although ncRNA was described for decades as a having a passive role as a messenger between DNA and protein, this idea has been proven incorrect, as ncRNAs are now well known to play active roles in transcriptional regulation (392). A unique mechanism by which O-GlcNAc machinery is directly regulated in an epigenetic manner is through long noncoding RNA (lncRNA). The first example of this relates to the location of the OGT gene on the X chromosome. Specifically, the XIST lncRNA is localized to the inactive X chromosome (Xi), and its presence regulates OGT levels in females but not males (393). As mentioned above, it was recently shown that a novel splice variant of OGT, nuclear OGT retained-intron (OGT-RI), may function as a nuclear ncRNA to regulate O-GlcNAc homeostasis (68). In addition to regulating O-GlcNAcylation, lncRNA can also be regulated by O-GlcNAcylation. For example, it was shown that under high-glucose conditions O-GlcNAcylation of p65 can activate the lncRNA for hyaluronan synthase 2 (HAS2), a naturally occurring antisense transcript (HAS2-AS1), leading to increased HAS2 to regulate hyaluronan synthesis (394). This mechanism provides an additional example by which O-GlcNAcylation can regulate, as well as be regulated by ncRNA.

At the other end of the spectrum, miRs have also been implicated in the regulation of O-GlcNAcylation. One of the first examples of this was in a study of failing heart in which Methusamy et al. (310) identified miR-539 as being induced. This miR induction, in turn, suppressed OGA protein levels, resulting in higher total protein O-GlcNAcylation. However, in hepatocarcinoma cells, miR-24-1 was shown to be lower in those cell lines with higher metastasis potential (395). The authors showed that this partly occurs though binding of miR-24-1 to the 3'-UTR of the transcript for OGT, which regulates levels of O-GlcNAcylation and stability of the oncoprotein c-Myc. In contrast, high glucose decreases levels of miR-200a and miR-200b in endothelial cells, resulting in less binding of these miRs to OGT mRNA and increased OGT protein levels associated with increased O-GlcNAcylation (308). Further, treating diabetic mice (db/db) with miR-200a and miR-200b mimics leads to decreased OGT levels that could reduce O-GlcNAcylation and inflammation. A final example of the link between miR and O-GlcNAc came from a screen that identified miR-501-3p for its potential to reduce OGT protein levels, decrease infectivity of hepatitis C virus, and decrease liver disease progression and development of cancer (306).

3.3. Insulin Signaling

In the late 1980s, it was established that insulin resistance induced by different interventions was associated with a decrease in translocation of the insulin-sensitive glucose transporter, GLUT4. In cultured adipocytes, it was shown that neither elevated glucose nor insulin alone could reduce insulin sensitivity; however, together, they resulted in a marked decrease in maximal insulin response (46). Additional studies demonstrated that glutamine was required for the development of decreased insulin sensitivity, and subsequently, Marshall et al. (46) demonstrated that increased flux through the HBP was a factor in this process. Prolonged treatment with glucosamine in the presence of insulin, decreased basal and insulin-stimulated glucose uptake, and reduced plasma membrane GLUT4 levels, providing further evidence that products from the HBP regulated insulin signaling (396). This was supported by Patti et al. (397), who found that glucosamine infusion in rats resulted in impaired insulin-stimulated glucose uptake and glycogen synthesis in skeletal muscle. In the same study, they showed that this was associated with a glucosamine-dependent increase in O-GlcNAc levels of IRS1/2. Moreover, overexpression of GFAT in skeletal muscle and adipocytes resulted in peripheral insulin resistance (398). Together, these studies provided the first indications that insulin signaling may be regulated via an HBP-mediated increase in O-GlcNAcylation.

Increasing overall O-GlcNAc levels in 3T3-L1 adipocytes, by inhibiting OGA with PUGNAc, resulted in impaired insulin-stimulated glucose uptake, with no changes in insulin-mediated increase in IR-β or IRS2 phosphorylation (133). On the other hand, insulin-induced increases of AKT and GSK3β phosphorylation were attenuated following PUGNAc treatment. While AKT and GSK3β were not found to be O-GlcNAcylated, IRS1 and β-catenin were both modified in a PUGNAc-dependent manner (133). Subsequent studies have shown that AKT is also a target for O-GlcNAcylation and this attenuates its function (121, 126). Several studies have questioned the role of O-GlcNAcylation in the development of insulin resistance. For example, lowering O-GlcNAc levels via overexpression of OGA or knockdown of OGT did not mitigate hyperglycemia-induced insulin resistance in adipocytes (399). In addition, the use of more specific OGA inhibitors than PUGNAc, such as NBuGt and 6-Ac-Cas (599, 600), failed to recapitulate earlier observations of insulin resistance seen with PUGNAc, although they did confirm that IRS-1 was a target for O-GlcNAcylation (600). Buse et al. (399) concluded that increased O-GlcNAcylation was just one of many factors involved in insulin resistance and was not necessarily required. On the other hand, in the setting of normoglycemia, a reduction of O-GlcNAc in the liver in vivo via overexpression of O-GlcNAcase significantly increased AKT activity (400).

Although the precise role of increased O-GlcNAc in contributing to cellular insulin resistance remains unclear, there is consensus over the fact that insulin treatment stimulates Tyr phosphorylation of OGT (FIGURE 9A) (242, 251). Insulin stimulation of adipocytes resulted in marked increased in Tyr phosphorylation of OGT, resulting in an increase in OGT activity and an association between OGT and IR. Insulin treatment also resulted in translocation of OGT from the nucleus to the cytoplasm (251). Another study reported that insulin triggered the translocation of OGT from the nucleus to the plasma membrane, which was facilitated by PIP₃ binding to OGT (242). As a result, it was proposed that OGT contained a PIP₃-binding domain; however, structural studies have not been able to confirm the presence of such a domain (90). The recruitment of OGT to the plasma membrane, resulted in increased OGT phosphorylation and activity, with subsequent increases in O-GlcNAcylation of IRS1, AKT, and other downstream targets of insulin signaling (242). PTP1B, has long been considered to be a major mechanism in attenuating insulin signaling through the dephosphorylation of Tyr residues in the activation loop of the insulin receptor (405). Interestingly, PTP1B has been reported to be O-GlcNAcylated, leading an increase in enzyme activity, thereby potentially contributing to a reduction in insulin signaling (FIGURE 9A). These findings suggest that OGT translocation and O-GlcNAcylation of target proteins are part of a feedback mechanism in which sustained activation of insulin signaling is suppressed in an insulin-dependent manner.

FIGURE 9.

Examples of O-linked β-N-acetylglucosamine (O-GlcNAc) regulation of cellular signaling pathways. A: Insulin signaling. On initiation, in response to insulin, there is autophosphorylation of the insulin receptor (IR) and subsequent phosphorylation and activation of Akt, FOXO and GSK3β and their downstream signaling pathways. On termination, insulin triggers rapid translocation of O-GlcNAc transferase (OGT) from the nucleus to the plasma membrane, where it is phosphorylated by the IR, which increases OGT activity, leading to the subsequent O-GlcNAcylation of insulin receptor substrate 1 (IRS1), AKT, FOXO, and GSK3β, thereby attenuating activity at multiple steps in the insulin signaling cascade. The protein tyrosine phosphatase 1B (PTP1B) is responsible for decreasing phosphorylation of IR and attenuating insulin signaling. It is also O-GlcNAcylated, which increases its activity contributing to a further reduction in insulin signaling. It has been reported that translocation of OGT to the plasma membrane is via binding to PIP₃ (242), which is generated in response to activation of insulin signaling; however, to date, structural studies have not revealed a phosphatidylinositol-3-phosphate (PIP₃)-binding motif in OGT(90). B: Calcium signaling. Intracellular Ca²⁺ increases in response to diverse number of agonists, either as a result of Ca²⁺ release from ER/SR alone or via activation of plasma membrane Ca²⁺ channels as a result of ER/SR Ca²⁺ release (store-operated Ca²⁺ entry, SOCE). This increase in Ca²⁺ leads to activation of calmodulin (CaM), resulting in the phosphorylation and activation of both CaMKII and IV, which are responsible for regulation of numerous cellular processes. CaMKII also phosphorylates OGT, increasing its activity, and in a feedback manner, OGT O-GlcNAcylates CaMKII/IV reduces their activities. Ca²⁺-CaM also activates the calcium-dependent phosphatase calcineurin, which is responsible for regulating diverse cellular processes, and it has been reported that increased OGT expression and O-GlcNAcylation is sufficient to activate calcineurin-mediated transcription pathways (298). Influx of extracellular Ca²⁺ has been shown to contribute to the stress-induced increases in O-GlcNAc levels (312). Therefore, it is possible, that Ca²⁺-CaM and/or calcineurin regulate OGT (or OGA) activities; however, this has yet to be demonstrated experimentally. Multiple contractile proteins are O-GlcNAc modified (TABLE 2), and the increases in O-GlcNAc levels that can occur in diseases, such as diabetes, reduces Ca²⁺ sensitivity of some contractile proteins (147, 150). Phospholamban (PLB) and SERCA are responsible, in part, for the reuptake of Ca²⁺ into the sarcoplasmic reticulum (SR), contributing to muscle relaxation. Increases in O-GlcNAc levels decrease the activities of both PLB and SERCA directly or indirectly, which could be a contributing factor to impaired myocardial relaxation (i.e., diastolic function) that occurs with diabetes. The endoplasmic reticulum (ER)/SR transmembrane protein STIM1, which plays a key role in regulating SOCE, is also a target for O-GlcNAcylation and increases in O-GlcNAc levels impairs its function and attenuates SOCE (401). Increasing O-GlcNAc levels attenuates mitochondrial Ca²⁺ overload, although the precise mechanisms are unclear(163, 402). CaMKII phosphorylation of the mitochondrial Ca²⁺ uniporter (MCU) potentiates mitochondrial Ca²⁺ overload (403); therefore, as O-GlcNAcylation of CaMKII decreases its activity, this may represent a potential protective mechanism. However, others have questioned the role of CaMKII in the regulation of mitochondrial Ca²⁺ uptake by the MCU (404).

One limitation of many studies examining the role of O-GlcNAc in insulin signaling is that they have been mostly in cultured adipocytes; consequently, it may differ in other insulin-sensitive cells and tissues. Nevertheless, there is no doubt that key elements of the insulin signaling pathway, including IRS1/2, PDK1, AKT, and GSK3β (See TABLE 2) are all targets for O-GlcNAcylation and that in all cases, increasing O-GlcNAcylation suppresses their activities (290).

3.4. Calcium Signaling

There is growing recognition of reciprocal regulation between Ca²⁺ signaling and protein O-GlcNAcylation, perhaps best exemplified by the regulation of CaMKIV activity via O-GlcNAcylation and OGT activity regulated by CaMKIV-mediated phosphorylation (FIGURE 9B) (167). Depolarization of neuroblastoma cells resulted in a rapid increase in OGT activity and O-GlcNAc levels that was shown to be mediated by CaMKIV phosphorylation of OGT (286). An early indication of the potential for Ca²⁺/O-GlcNAc cross talk was the identification of seven O-GlcNAc modification sites on synapsin I that were clustered around its regulatory phosphorylation sites, and O-GlcNAcylation of these sites reduced the affinity of CaMKII for synapsin I (151). In the heart, CaMKII has also been shown to be O-GlcNAcylated on Ser-279, and this results in autonomous activation of CaMKII, which, in the setting of diabetes, was linked to increased arrhythmias (165). In liver, similar to the observations in neuroblastoma cells, CaMKII phosphorylated OGT, increasing O-GlcNAc levels, which subsequently activated autophagy (181). Given the diverse array of cellular functions that are regulated by the CaMK family of proteins (406), it is likely that many more links with O-GlcNAcylation remain to be discovered.

The nuclear factor of activated T cells (NFAT) family of transcription factors, are widely distributed and contribute to the regulation of numerous processes, including the immune system, cardiac and skeletal muscle, and brain. Ca²⁺-dependent activation of calmodulin, activates the phosphatase calcineurin which rapidly dephosphorylates NFAT, resulting in its nuclear translocation and activation. In neonatal cardiomyocytes NFAT translocation is initiated by hypertrophic agonists, such as angiotensin II (ANG II) or PE, and hyperglycemia was found to inhibit this translocation in an HBP-dependent manner (407). A subsequent study demonstrated that increasing O-GlcNAc levels attenuated the ANG II-induced increase in cytosolic Ca²⁺ (408). More recently, it was reported that activation of O-GlcNAc signaling was required for NFAT translocation in cardiomyocytes, and this could be blocked with the calcineurin inhibitor cyclosporin A (298). Thus, O-GlcNAcylation may have regulatory roles in both the initiation of NFAT signaling, as well as its inhibition (FIGURE 9B). Another protein that has been established as a key player in Ca²⁺-dependent activation of NFAT, particularly in the immune system, is stromal interacting molecule 1 (STIM1), via its role in regulating the store operated calcium entry pathway (SOCE) (409). STIM1 is positively and negatively regulated by phosphorylation, and increases in O-GlcNAc increased basal phosphorylation, but attenuated activation-dependent increases in phosphorylation (401). STIM1, is a highly conserved protein, considered to be a core component of mammalian Ca²⁺ signaling (409); consequently, its functional regulation by O-GlcNAc could have far reaching consequences.

In skeletal and cardiac muscle, Ca²⁺ also plays a central role in contraction of the myofilaments, via Ca²⁺ binding to key proteins, as well as regulating Ca²⁺ release and uptake by the ER/SR (FIGURE 9B) (410). Skeletal muscle myosin was the first contractile protein that was shown to be O-GlcNAcylated, with all isoforms being modified (411). Subsequent studies identified actin and myosin light chain (MLC) 1,2 as O-GlcNAc targets and found that in skinned muscle fibers, acute increases in O-GlcNAc levels reduced Ca²⁺ sensitivity, suggesting a possible role for O-GlcNAc in regulating skeletal muscle contractility (146). Myofilament proteins from cardiac muscle were also found to be O-GlcNAcylated, and in addition to those identified in skeletal muscle, troponin I (TnI) was modified at Ser-150, which is a phosphorylation site that regulates Ca²⁺ sensitivity (147). Pharmacological increases in O-GlcNAc levels reduced Ca²⁺ sensitivity, consistent with the earlier skeletal muscle study (146).

Uptake of cytosolic Ca²⁺ into the ER/SR is an important mechanism for regulating Ca²⁺ signaling, as well as muscle contraction, and this is controlled by sarcoplasmic/endoplasmic reticulum calcium ATPase (SERCA) and its inhibitor phospholamban (PLB). Phosphorylation of PLB attenuates its inhibitory effect, thereby, facilitating more rapid uptake of Ca²⁺ into the ER/SR by SERCA. PLB is a target for O-GlcNAcylation, and Ser-16 was found to be the most likely target (194), although to date, this has not been confirmed by MS. Increased PLB O-GlcNAc levels, either by inhibition of OGA or under conditions of hyperglycemia, reduced PKA-mediated phosphorylation of PLB Ser-16 (194); conversely, knockdown of OGT significantly increased PLB phosphorylation (FIGURE 9B). Increased O-GlcNAc levels, were associated with lower SERCA activity and a greater association between SERCA and PLB (194). It was concluded that the increase in PLB O-GlcNAcylation could be a factor in the slower reuptake of Ca²⁺ by the ER/SR in the heart that occurs with diabetes. SERCA expression in the heart, has been shown to be decreased under conditions of increased O-GlcNAc levels, and this was attributed to an increase in O-GlcNAcylation of the transcription factor Sp1 (263). While Yokoe et al. (194), reported that SERCA was not an O-GlcNAc target, others have shown that it is O-GlcNAcylated (412, 413), although the functional consequence of this modification remains to be identified.

Another important contributor to cellular Ca²⁺ signaling is the inositol 1,4,5-trisphosphate (InsP₃)-receptor, which is localized to the ER/SR membrane and is activated by InsP₃ generated in response to a variety of extracellular stimuli. In C₂C₁₂ myotubes, increased O-GlcNAc levels attenuated bradykinin-induced production of IP₃ and associated InsP₃R-mediated Ca²⁺ release, which was associated with O-GlcNAcylation of PLC-β1 (414). The InsP3 receptor type 1 (InsP₃R-1) itself is O-GlcNAcylated under basal conditions and this could be increased following treatment with the OGA inhibitor PUGNAc (415). Moreover, decreases in basal O-GlcNAc levels resulted in a marked increase in channel opening probability, and conversely, increasing O-GlcNAc levels reduced channel opening probability. These findings demonstrated a potential role for O-GlcNAc in regulating InsP₃R-1 under physiological conditions (FIGURE 9B) (415). A subsequent study from the same group found that InsP₃R-2 was not O-GlcNAcylated, and its function was unaltered by global changes in O-GlcNAc levels (416). Interestingly, however, they found that InsP₃R-3 was O-GlcNAcylated, but that changes in its O-GlcNAc levels had the opposite effect to those observed with InsP₃R-1. InsP₃R is ubiquitously expressed, although the different isoforms exhibit tissue-specific differences in their function (417); consequently, there is a clear need for a better understanding of the role of O-GlcNAc in their regulation.

It is clear that O-GlcNAcylation provides a key link between nutrient and Ca²⁺ signaling contributing to the regulation of the majority of key Ca²⁺ signalling pathways (FIGURE 9B). What is also becoming increasingly evident is that O-GlcNAc levels are regulated in a Ca²⁺-dependent manner, as exemplified by CaMKII/IV-mediated phosphorylation of OGT, resulting in increased activity and high O-GlcNAc levels (181, 286). Stress induced increases in cellular O-GlcNAc levels have also been shown to be dependent on extracellular Ca²⁺ and activation of CaMKII (312). The role of Ca²⁺ in regulating GFAT activity is poorly understood, even though it can be phosphorylated by CaMKII, and to date, the role of Ca²⁺ in regulating OGA activity is unknown.

3.5. Metabolism and Mitochondrial Function

As discussed in sect. 2.1, substrate availability plays a key role in regulating HBP flux and O-GlcNAcylation; however, what is less well known is that O-GlcNAcylation contributes to the regulation of metabolism at multiple levels. For example, almost every enzyme in glycolysis has been shown to be O-GlcNAc modified from GLUT4 to pyruvate dehydrogenase (FIGURE 10A) (418). In addition, O-GlcNAcylation of glucose-6-phosphate dehydrogenase regulates pentose phosphate pathway activity, and GSK3β O-GlcNAcylation regulates glycogen synthesis. In the heart, increased flux through the HBP has been linked to greater fatty acid oxidation, potentially via O-GlcNAcylation of the fatty acid transporter, CD36 (419,420). Similarly, in adipocytes, activation of the HBP and increasing O-GlcNAc levels also stimulated fatty acid oxidation (421). Interestingly, a splice variant of OGA, sOGA, is associated with lipid droplets, further supporting a connection between O-GlcNAcylation and the regulation of lipid metabolism (273). In addition to direct regulation of metabolic fluxes, O-GlcNAcylation also contributes to the transcriptional regulation of metabolism via O-GlcNAcylation of multiple transcription factors, including PGC-1α, FoxO3, and CREB (100, 105, 117, 598).

FIGURE 10.

Metabolism and mitochondrial O-GlcNAcylation. A: majority of enzymes in glycolysis and glucose metabolism are targets for O-GlcNAcylation (418). ALD, aldolase; ENO, enolase; GAPDH, glyceraldehyde-6-phosphate dehydrogenase; GFAT, glutamine fructose-6phosphate amidotransferase; GP, glycogen phosphorylase; G6PDH, glucose-6-phosphate dehydrogenase; GS, glycogen synthase; HEX, hexokinase; LDH, lactate dehydrogenase; PFK, phosphofructokinase; PDC, pyruvate dehydrogenase complex; PGI, phosphoglucoisomerase; PGK, phosphoglycerate kinase; PGM phosphoglucomutase; PK, pyruvate kinase; UDP-GP, UDP-glucose pyrophosphorylase. B: wide range of different types of mitochondrial proteins that are O-GlcNAcylated. C: map of the O-GlcNAc modification sites on proteins that are central to the regulation of energy metabolism and mitochondrial function. Data largely based on a number of recent proteomic studies (138,139). Blue squares denote O-GlcNAc, whereas numbers inside blue squares indicate number of O-GlcNAc sites on the protein.

The identification of a splice variant of OGT, with a mitochondrial targeting sequence (422), suggested that mitochondrial proteins may be targets for O-GlcNAcylation. However, early reports indicated that there was little, if any, O-GlcNAcylation in mitochondria. It was proposed that the lack of mitochondrial O-GlcNAc was due to the lack of substrate, as there was no clear mechanism for UDP-GlcNAc transport into the mitochondria (423). Nevertheless, a number of studies suggested that alterations in cellular O-GlcNAc levels could alter mitochondrial function and particularly their response to stress. For example, increasing O-GlcNAc levels in cardiomyocytes with PUGNAc attenuated the loss of mitochondrial membrane potential induced by oxidative stress, and this was associated with O-GlcNAcylation of voltage-dependent anion channel (VDAC)1 (163). Both glucosamine treatment and OGT overexpression increased O-GlcNAc levels and reduced ischemia-reperfusion-induced injury in isolated cardiomyocytes (402). The same treatments attenuated hydrogen peroxide-induced mitochondrial membrane potential and enhanced recruitment of the anti-apoptotic protein BCL2 to the mitochondria (402). Subsequent studies suggested BCL2 itself might be a target for O-GlcNAcylation (424).

Although acute activation of O-GlcNAc levels protected mitochondria from oxidative stress, increases in O-GlcNAcylation associated with hyperglycemia were found to have adverse effects on mitochondrial function. For example, in pancreatic β-cells, hyperglycemia increased O-GlcNAc levels on the chaperone heat shock protein (HSP) 60, interfering with its binding to the proapoptotic factor BCL2 Associated X, Apoptosis Regulator (BAX), resulting in BAX translocation to the mitochondria and subsequent cytochrome-c release (425). Hyperglycemia resulted in impaired activity of mitochondrial complexes in neonatal cardiomyocytes, which could be reversed by overexpression of OGA. High glucose levels also increased O-GlcNAcylation on subunits of mitochondrial complexes, which was reduced following overexpression of OGA (161). Hyperglycemia also increased O-GlcNAcylation of dynamin-related protein 1 (DRP1), increasing mitochondrial fragmentation and decreasing mitochondrial membrane potential (159). In neurons, hyperglycemia impaired mitochondrial motility in an O-GlcNAc-dependent manner, due to O-GlcNAcylation of MILTON1 (also known as TRAK1), a trafficking protein that is essential for mitochondrial movement (160). 8-oxoguanine DNA glycosylase (Ogg1), is involved in mitochondrial DNA repaired and is O-GlcNAcylated in response to hyperglycemia, decreasing its activity and possibly contributing to increased mtDNA damage (426).

In 2009, Hu et al. (161), identified O-GlcNAcylation on a number of mitochondrial oxidative phosphorylation complex proteins, which was associated with decreased complex I, II, and IV activity in mitochondrial of cardiomyocytes exposed to high glucose. Cao et al. (157) identified 11 O-GlcNAcylated proteins from rat liver mitochondria, which included enzymes in the TCA cycle, as well as those involved in ATP synthesis (FIGURE 10B). It remained unclear, however, as to how mitochondrial proteins could be modified by O-GlcNAc. In 2015, transport of ³H-UDP-GlcNAc into cardiac mitochondria was reported and the pyrimidine nucleotide carrier (PNC1) was identified as the UDP-GlcNAc mitochondrial transporter (427). Immunogold labeling and live cell imaging were also used to demonstrate the presence of OGA localized to the mitochondria, demonstrating for the first time that an active O-GlcNAc cycle was present in the mitochondria (427). Subsequent proteomics studies identified 86 mitochondrial proteins as O-GlcNAc targets that were involved in a diverse array of mitochondrial functions, including the TCA cycle, oxidative phosphorylation, fatty acid oxidation, and calcium regulation (FIGURE 10C) (138, 139). Others have been unable to detect mOGT isoform in cells or tissues, and ncOGT was reported to be the isoform responsible for O-GlcNAcylation of mitochondrial proteins (428). It has also been suggested that mOGT regulates mitochondrial stress responses, while ncOGT is responsible for the regulating mitochondrial bioenergetics (429). Thus, additional work remains to elucidate the fundamental mechanisms regulating O-GlcNAc cycling in the mitochondria, as well as understanding the physiological role of O-GlcNAcylation in regulating mitochondrial function.

3.6. Cell Survival/Autophagy

Pharmacological studies of OGT or OGA inhibition in INS-1 and βTC-6 cells have linked high O-GlcNAcylation with cell death and increased O-GlcNAcylation with reduced phosphorylation of Ser-473 on AKT1 as the potential mechanism (122). In human embryonic kidney 293 and HeLa cells, hyper-O-GlcNAcylation by overexpression of OGT induced apoptosis, which was associated with increased AKT O-GlcNAcylation and decreased AKT phosphorylation (124). Hyper-O-GlcNAcylation is also evident in response to cerebral ischemia (124). In a model of diabetic retinopathy, increased O-GlcNAcylation of NF-κB was associated with retinal ganglion cell death (430). Recent studies also have shown that O-GlcNAcylation on Ser-56 and Ser-57 of c-Fos is increased in 5XFAD mouse model of Alzheimer’s disease, as well as neuroblastoma cells exposed to amyloid beta (Aβ), this increased its stability and interaction with c-JUN, increased target gene Bim expression, and promoted cell death (431). Increased O-GlcNAc levels have also been reported to exacerbate acetaminophen-induced liver injury, whereas lower O-GlcNAc levels, resulting from liver-specific OGT knockout reduced the degree of injury (432). Increased cardiomyocyte apoptosis observed in a rat model of diabetes was linked to an overall increase in O-GlcNAc levels, as well as increased O-GlcNAcylation and reduced phosphorylation of the proapoptotic protein BCL2-associated agonist of cell death (or BAD) (433).

In contrast to the increase in O-GlcNAc levels, leading to cell death, Zachara et al. (289), demonstrated that exposure of cells to a variety of stressors induced an increase in O-GlcNAc that they showed was cytoprotective and represented an endogenous cell survival response. Subsequent studies revealed a link between the increase in O-GlcNAc and cell survival with the greater induction of heat shock protein expression via inactivation of G3K3β (280). A number of proteins involved in the DNA repair pathway have also been shown to be regulated either directly or indirectly by O-GlcNAcylation (89, 319, 434–436). O-GlcNAcylation is required for the normal regulation of DNA damage pathways, and OGT is recruited to sites of DNA damage (434). Additional studies found that pharmacologically increasing O-GlcNAc levels was cytoprotective, particularly in the setting of acute oxidative stress or ischemia-reperfusion (163, 402, 437). There have been several reports demonstrating a strong correlation between O-GlcNAc levels and tissue injury, with low O-GlcNAc levels associated with increased injury (437–439). The specific mechanisms by which the increase in O-GlcNAc is protective remains to be determined; however, there are indications that it attenuates Ca²⁺ overload, protects mitochondria against oxidative stress, and attenuates proinflammatory responses (163, 402, 440–443).

Autophagy is an important cell survival mechanism, and a role for O-GlcNAc in regulating autophagy is starting to emerge. Recent studies have shown Unc-51-like autophagy activating kinase 1 (ULK1) (444), mammalian target of rapamycin (mTOR) (445), hypoxia-inducible factor 1α (HIF1 α) (446), synaptosome-associated protein 29 (SNAP29), tubulin polymerization-promoting protein (TPPP), Golgi reassembly-stacking protein of 55 kDa (GRASP55) (180, 447–450), all of which play important roles in the regulation of autophagy, are modified by O-GlcNAcylation. Increased O-GlcNAcylation of SNAP29 and GRASP55 resulted in decreased autophagy (FIGURE 11) (180, 450). In C. elegans, an ogt mutation elevated autophagy during development and knockdown of OGT in mammalian cells also promoted autophagy (180). Reduced OGT was associated with increased formation of the SNARE complex, which includes SNAP29 and increased O-GlcNAcylation of SNAP29-inhibited autophagy at the step of autophagosome maturation (180). Impaired autophagic signaling in the diabetic heart was associated with increased O-GlcNAc levels and O-GlcNAcylation of BCL2 and Beclin 1 (424). Arsenic-induced inhibition of autophagic flux was shown to be mediated by O-GlcNAcylation of SNAP29, and this was prevented by reducing O-GlcNAc levels by knockdown of OGT (451). Transfection of SNAP29 containing an O-GlcNAc site mutant attenuated arsenic inhibition of autophagy, further substantiating the role of SNAP29 O-GlcNAcylation in regulating autophagy (451).

FIGURE 11.

O-linked β-N-acetylglucosamine (O-GlcNAc) regulation of cell survival and autophagy. A role of O-GlcNAc in cell survival is evidenced by O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA) knockout/knockdown and overexpression studies, although the identities of specific modified proteins in promoting cell death or survival are largely unclear. Among others, AKT, NF-κB, and c-Fos have been shown to be O-GlcNAcylated, and their O-GlcNAcylation is associated with cell death processes. In terms of pathology, Alzheimer’s disease-associated tau phosphorylation has been shown to be attenuated by enhancement of its O-GlcNAcylation. Although there are also studies demonstrating that increased O-GlcNAcylation can enhance proteotoxicity, conceivably by altering the degradation of O-GlcNAcylated forms of specific proteins. The involvement of O-GlcNAcylation in autophagy is best demonstrated in studies on SNAP29 and GRASP55, whose O-GlcNAcylation attenuates autophagic flux.

A mutagenesis study has shown that single mutations of GRASP55 at Ser-389, Ser-390, Thr-403, Thr-404, and Thr-413 decreased its O-GlcNAcylation, while a single mutation at Ser-380 increased its O-GlcNAcylation, whereas, mutating all Ser-389, Ser-390, Thr-403, Thr-404, and Thr-413 sites or these five sites in combination with Ser-380 both abolished its O-GlcNAcylation (FIGURE 11) (450). Notably, these sites are not the previously identified phosphorylation sites, and mutating known phosphorylation sites did not affect GRASP55 O-GlcNAcylation. During normal conditions, GRASP55 is O-GlcNAcylated, and in response to glucose deprivation, it is de-O-GlcNAcylated and interacts with LC2-II and lysosomal associated membrane protein 2 (or LAMP2), activating autophagy through enhancing autophagosome-lysosome fusion (450).

In neuroblastoma SH-SY5Y cells, O-GlcNAcylation increased autophagy, and this was linked to O-GlcNAcylation of ATG4B, which enhanced its proteolytic activity (452). However, the O-GlcNAcylation sites on ATG4B were not identified, and it remains to be determined whether ATG4 O-GlcNAcylation is necessary or sufficient for activation of autophagy (452). Interestingly, OGT interacted with ATG4B and downregulation of OGT decreased ATG4B. Pharmacological inhibition of OGA activated mTOR and inhibited autophagic flux in primary neurons (453), while in glia/primary neuron-mixed cultures, a similar intervention enhanced autophagic flux without affecting mTOR (454). In vivo treatment with the OGA inhibitor, Thiamet G, increased autophagic flux in the brain (454). Taken together, there is growing evidence that there is cross talk between autophagy and protein O-GlcNAcylation and that autophagic proteins are targets for O-GlcNAcylation; however, the precise role of O-GlcNAc in regulating autophagy remains unclear (FIGURE 11).

Conceivably, O-GlcNAcylation of a subset of proteins may also change their recognition and/or degradation by autophagy; therefore, as a result, inhibition or activation of autophagy will change the ratio of O-GlcNAcylated versus un-O-GlcNAcylated species of specific proteins. Additionally, activation or inhibition of autophagy may change the levels and the intracellular localization of OGA and OGT, or autophagy adaptor proteins that recognize specific autophagy substrates. Such an example has been demonstrated in breast cancer cells in which rapamycin treatment decreased O-GlcNAc levels by decreasing OGT level (455). These assumptions need to be critically examined by experiments to be substantiated. Specific effects of autophagy activation or inhibition at distinct steps on specific protein species and functional consequences need to be defined.

3.7. Circadian Clock

The circadian clock, a transcriptional mechanism that results in ∼24-h cycles of many physiological processes, has been identified in almost all cell types (456). In mammals, there is a central clock located in the suprachiasmatic nucleus (SCN), and there are peripheral clocks located in all other tissues and cells. The central clock is sensitive to light and regulates systemic rhythms in activity, body temperature, hormone levels, immune function, and blood pressure, whereas peripheral clocks regulate tissue-specific processes. Dyssynchrony between the SCN and external cues, as well as between the SCN and peripheral clocks, which occur as a result of shift work, jet lag, or poor sleep, have been linked to an increased risk of metabolic disease, including Type 2 diabetes, cardiovascular diseases, and cancer (457). A number of studies have demonstrated that there is complex cross talk between O-GlcNAc cycling and regulation of the circadian clock.

In the heart there were time-of-day-dependent changes in overall O-GlcNAc levels, as well as OGT and OGA protein levels that were not present when the cardiomyocyte circadian clock was deleted (344). In the same study, the authors showed that BMAL1, a circadian clock-regulating transcription factor, was also O-GlcNAcylated in the heart. It was also demonstrated that pharmacologically increasing O-GlcNAc levels resulted in a phase shift in the SCN clock (344). This was the first indication that not only were cellular O-GlcNAc levels regulated by the circadian clock, but also that acute changes in O-GlcNAcylation directly affected the circadian clock. Berthier et al. (65) found that REV-ERBα, a transcription factor that integrates circadian clock and energy metabolism, interacted with OGT preventing its degradation and, thus, increasing OGT activity, providing a potential mechanism by which the circadian clock can regulate cellular O-GlcNAc levels.

In Drosophila, upregulation of OGT increases the duration of circadian rhythms, whereas OGT knockdown shortened them (458). This was associated with changes in O-GlcNAcylation of the Drosophila PERIOD protein (dPER) and stabilization of dPER. Several studies further demonstrated a key role for O-GlcNAc in regulating the mammalian circadian clock (116, 330, 459). Central components of the circadian clock, including BMAL1, PER2, and CLOCK, are all targets for O-GlcNAcylation, and O-GlcNAc modification stabilized them, at least in part, by preventing their ubiquitination (330). There was also a consistent observation that reduced O-GlcNAc levels shortened, while increased O-GlcNAc lengthened, the duration of circadian rhythms. PER2 was shown to be O-GlcNAcylated in a region that regulates sleep phase in humans, competing with phosphorylation in the same region (116). Modulation of O-GlcNAc levels in the liver by either overexpression or knockout of OGT changed the diurnal rhythms of circulating glucose levels (330). Together, these studies demonstrate that OGT, OGA, and O-GlcNAcylation are key components of the circadian clock and, as such, contribute to the regulation of daily rhythms in physiological processes in central and peripheral clocks. Moreover, since dysregulation in O-GlcNAc cycling is associated with chronic diseases, this might also contribute to circadian dysregulation seen in response to stress and disease (460).

4. O-GlcNACYLATION IN HEALTH AND DISEASE

The wide-ranging effects of O-GlcNAcylation on protein and cellular function combined with its integration of nutrient availability and diverse signaling pathways make it ideally positioned to coordinate cellular response to physiological and pathological stimuli. In this section, we will summarize what is known about the role of O-GlcNAc involvement in normal physiology and in pathologies of diseases.

4.1. Development

In Xenopus and Drosophila, there are marked changes in the O-GlcNAc proteome during embryogenesis, including changes in the extent of O-GlcNAcylation on specific proteins, as well as the number of proteins that are O-GlcNAc targets (461, 462). Key regulators of embryonic development in Drosophila are O-GlcNAcylated, and disruption of O-GlcNAc cycling resulted in abnormal development (463, 464). Polycomb group (PcG) proteins play an important role in developmental regulation in Drosophila and mammals. In Drosophila, there is overlap between O-GlcNAc modification sites and PcG protein binding sites, indicating a role for OGT in regulating PcG gene silencing, further supporting a critical connection between O-GlcNAc cycling and developmental regulation (465).

In mice, the loss of EMeg32, which catalyzes the conversion of glucosamine-6-phosphate to glucosamine N-acetyl-6-phosphate, is embryonically lethal at day 7.5 (228). EMeg32 deletion, significantly decreased UDP-GlcNAc levels. This primarily leads to loss of O-GlcNAc protein modification rather than Golgi-mediated glycosylation, suggesting that O-GlcNAcylation plays an essential role in development. This is supported by the observation that OGT gene deletion is embryonically lethal at the blastocyst stage (∼day 4.5) and that this also results in embryonic stem cell death (208). To determine the effects of a loss of OGT in specific cell types O’Donnell et al. (16) used Cre-recombinase approaches to delete OGT in T cells, fibroblasts, and neurons. They found that loss of OGT resulted in T-cell apoptosis and growth arrest of fibroblasts. Selective ablation of OGT in neurons resulted in much smaller pups that did not survive beyond 10 days and was associated with marked increases in phosphorylation of tau (16). Cardiomyocyte-specific deletion of OGT (cmOGT KO) resulted in marked postnatal lethality, with only 12% surviving to weaning age. The surviving cmOGT KO mice exhibited signs of hypertrophy and heart failure characterized by dilation and accompanied by increased fibrosis and apoptosis (206). The constitutive KO of OGA, resulted in embryonic developmental delay and very high perinatal mortality, which was associated with mitotic defects (466). Keembiyehetty et al. (17) extended these observations, demonstrating that following OGA deletion, only 3% of pups survive to weaning, with the majority dying soon after birth. However, no structural defects were observed at either E16.5 or E18.5 in OGA KO animals, and surviving KO pups were almost indistinguishable from WT littermates. They observed profound hypoglycemia in KO mice, suggesting that perinatal lethality was likely a result of impaired glycogen metabolism. Collectively, these studies demonstrate the importance of normal O-GlcNAc cycling in developmental regulation.

Maternal stress, whether it is due to overnutrition or undernutrition, drug or alcohol consumption, or other factors, is known to increase the risk of neurodevelopmental and metabolic disorders (467). In a mouse model of early prenatal stress, placental OGT levels were significantly decreased; moreover, a placenta-specific reduction in OGT adversely affected early neurodevelopment, which was more pronounced in males compared to females (468,469). Severe hyperglycemia in rats, resulted in structural and functional changes in the placenta that was associated with increased O-GlcNAc levels in endothelial and trophoblast cells (470). In a model of maternal corticosterone exposure, Pantaleon et al. (471) demonstrated that there was an increase in OGT levels and O-GlcNAcylation of AKT in male, but not female, placentae. Exposure to corticosteroids in utero also resulted in depression-like behavior that was more pronounced in males, possibly due to an OGT-mediated effect on mitochondrial motility (472). One mechanism contributing to sex-specific differences in placental OGT could be due to increased resilience in females via increased OGT-mediated stabilization of EZH2 and higher levels of the epigenetic modification H3K27me3 (473). The observations of sex-specific differences in OGT and O-GlcNAc in the placenta provide a potential mechanism underlying sexual dimorphic responses to maternal stress.

An important consideration regarding the role of OGT and O-GlcNAc in developmental regulation is the fact that in all mammals, OGT is present on the X-chromosome: XqD in mice, and Xq13 in humans (208, 474). One consequence of this is that when maternally inherited, female heterozygous OGT KO is embryonically lethal, whereas when paternally inherited, mice are viable (475). Moreover, because of X-inactivation, female heterozygous OGT KO mice would be expected to see a reduction in OGT by ∼50%; however, this is not the case, suggesting a selection bias in early development against cells expressing the inactivated maternal allele (208). OGT expression in most tissues is equal between males and females, indicating that it is subject to silencing by X-inactivation of one of the two X-chromosomes in females (475). The fact that OGT is an X-linked gene raises the possibility that it could be a factor in sex differences in susceptibility to certain diseases. This could occur when a mutation occurs on one of X-chromosomes, which in females could be silenced, thereby, reducing, or even eliminating risks for certain diseases (FIGURE 8B, III). An example of this is reflected in the recent reports of several OGT mutations associated with X-linked intellectual disability (28,29, 476). In contrast, in two female twins, a wild-type OGT was silenced via X-inactivation, leading to expression of an OGT mutation in the catalytic domain, resulting in intellectual disability and developmental delays (28). Incomplete X-inactivation or X-reactivation could also be a factor in diseases that are known to predominantly affect women, including autoimmune diseases, such as lupus (475). Indeed, hypomethylation of X-linked genes, including OGT, was observed in women with active lupus compared to inactive disease or healthy women (477). The degree of overexpression of OGT mRNA and protein was directly related to the severity of disease, and both were higher in women with lupus compared to men with lupus, consistent with a contribution from the inactive X-chromosome (477). It has also been proposed that OGT reactivation could contribute to cancer progression in women (475).

4.2. Obesity and Diabetes

As discussed above, early studies linked increased HBP flux to the possible development of insulin resistance, and other reports demonstrated a close relationship between O-GlcNAcylation and insulin signaling. Consequently, excessive HBP flux and sustained increases in O-GlcNAc levels have been implicated in the development of metabolic disease. For example, overexpression of GFAT in skeletal muscle and adipose tissue resulted in systemic hyperinsulinemia and insulin resistance, and its overexpression in the liver caused obesity, dyslipidemia accompanied by insulin resistance, and in older animals, overt diabetes (478). A selective increase of GFAT in pancreatic β-cells resulted in hyperinsulinemia and insulin resistance (478). These findings support the concept that the HBP contributes to systemic metabolic regulation, but that under conditions of excess, it leads to metabolic dysfunction characteristic of insulin resistance and diabetes. Single-nucleotide polymorphisms in the GFAT2 gene are linked to Type 2 diabetes and increased risk of diabetic complications, such as diabetic nephropathy (479,480). In leukocytes, GFAT2 gene expression was found to be lower in prediabetic Caucasian subjects compared with controls, and this was even more pronounced in diabetic individuals; GFAT1 was not different between groups (481). How these changes related to GFAT1/2 protein levels or activities is not known.

OGT overexpression in the liver also leads to insulin resistance, supporting the idea that the effects of GFAT overexpression are most likely the result of increased O-GlcNAcylation (242). Additional support for a role of O-GlcNAc cycling in metabolic regulation comes from the observation that hepatic knockdown of OGT resulted in a decreased expression of gluconeogenic genes and improved glucose homeostasis in diabetic mice (118). Moreover, OGA polymorphism has been linked to Mexican Americans with increased risk for diabetes (482–484). Both OGT and OGA mRNA levels were significantly lower in diabetic individuals compared to both prediabetic and control subjects (481); it is unclear what factors are driving these changes or how they translate to changes in protein levels or activity. Muscle O-GlcNAc levels were significantly increased in patients with Type 2 diabetes compared with controls, regardless of insulin levels; however, there were no changes in protein levels of GFAT, OGT, or OGA (379). Skeletal muscle-specific OGT knockout increased overall energy expenditure, markedly increased insulin sensitivity, and reversed insulin resistance associated with a high-fat diet-induced insulin resistance (379, 485). Modest overexpression of OGT in skeletal muscle and adipose tissue resulted in insulin resistance and increased circulating leptin levels, despite having no effect on body weight, fat pad weight, or fasting blood glucose levels. O-GlcNAc signaling has also been implicated in regulating whole body metabolism in response to overnutrition by modulating macrophage proinflammatory activity (486). Collectively, these findings demonstrate the importance of O-GlcNAc cycling in multiple tissues in the regulation of whole body metabolic homeostasis.

OGT deletion specifically in pancreatic β-cells resulted in severe hyperglycemia, hypoinsulinemia, and β-cell apoptosis (487). Thioredoxin-interacting protein (TXNIP) is widely recognized as playing a key role in β-cell dysfunction and death, factors that underlie both Type 1 and Type 2 diabetes (488). Hyperglycemia increases O-GlcNAcylation of TXNIP in β-cell lines and in islets from diabetic rodents, which was associated with increased inflammatory processes (489). The dnOGA variant described above, which lacks OGA activity, was initially identified in the Goto-Kakizaki rat and is associated with a nonobese diabetic phenotype (199), as well as significantly increased O-GlcNAcylated proteins in the pancreas and islets (490). Increased O-GlcNAc levels have also been implicated in diabetic complications associated with many organ systems, diabetic cardiomyopathy, nephropathy, and retinopathy (20, 491–493).

Supporting the importance of brain protein O-GlcNAcylation in sensing nutrients and signaling satiety are the observations that forebrain neuron-specific OGT knockout results in overeating-dependent obesity (203). These data suggest that a lack of neuronal O-GlcNAcylation leads to excessive food intake, which can contribute to obesity. In further support of the importance for tight regulation of O-GlcNAcylation, OGT ablation in AgRP neurons, promotes white adipose tissue browning and protects mice against diet-induced obesity and insulin resistance (201). These data suggest that a lack of AgRP neuron O-GlcNAcylation sends signals that attenuate white adipose tissue browning. OGA deletion specifically in the brain resulted in a substantially elevated body fat, fatty liver, and metabolic dysfunction, including increased leptin, triglycerides, and insulin (494). These findings suggest that the hypothalamic-pituitary axis, which plays a key role in metabolic regulation in the brain, was adversely affected by the loss of OGA. Global OGA haploinsufficiency (Oga^+/−) results in decreased body weight, decreased inguinal subcutaneous lipid content, with increased respiratory exchange ratio, suggesting increased glucose utilization, with no change in food intake or activity, yet also with higher energy expenditure (17). In response to a high-fat diet, only female Oga^+/− mice demonstrated increased weight gain. On the other hand, it has been reported that Oga^+/− mice were also resistant to high-fat diet-induced obesity, and they had enhanced differentiation of subcutaneous white adipose tissue-derived stromal vascular fraction cells into energy-consuming brown-like adipocytes (207). These observations indicate a complex regulation by OGT and OGA that involve multiple central and peripheral mechanisms.

4.3. Neurological Function and Diseases

It has been shown that cytosolic OGT activity is 10 times higher in brain compared to muscle, adipose, heart, and liver (495), suggesting significant involvement of O-GlcNAcylation in nervous system function. Using postsynaptic density preparation of mouse brain, Vosseller et al. (448) and Shapira et al. (496) identified ∼500 unique O-GlcNAc-modified proteins, including vesicle docking and fusion protein, Bassoon, and post-synaptic cytoskeletal structure protein, ankyrin G. Additional proteins involved in synaptic plasticity, synaptic vesicle trafficking, and axonal branching have been identified as O-GlcNAc targets (100, 152, 448, 497). Other examples of O-GlcNAcylation play a role in neuronal function, which include: 1) inhibition of cyclic-AMP response element binding protein (CREB) activity by O-GlcNAcylation of Ser-40 (100); 2) altered mitochondrial dynamics by O-GlcNAcylation of the motor adaptor protein Milton (160); and 3) altered hippocampal synaptic plasticity by GluA2 O-GlcNAcylation (200). It is estimated that ∼40% of neuronal proteins and 19% of synaptosome proteins are O-GlcNAcylated (182). It has also been shown that O-GlcNAcylated proteins, OGT, and OGA are abundant in widespread regions of the brain, especially the synapses and postsynaptic density preparations (100, 200, 448, 498–500). O-GlcNAcylated proteins have been found in multiple cell types in the brain, including pyramidal cells, GABAergic interneurons, and astrocytes in the hippocampus (200), as well as Purkinje cells in cerebellar cortex (501). These data are consistent with O-GlcNAc signaling playing an important role in neuronal function.

As discussed above, OGT mutations have been linked to intellectual disability (26,27, 29, 198). For example, Parkinsonian-dystonia (DYT3) has been mapped to the X chromosomal region that includes the Ogt locus (502), although whether the OGT mutation is causative for DYT3 has yet to be demonstrated. Interestingly, OGA is localized to chromosome 10(10q24), a region associated with late-onset Alzheimer’s disease (AD) (503). Protein O-GlcNAcylation was shown to be decreased in postmortem AD brains compared with controls, and this was associated with hyperphosphorylation of tau (504). Pharmacological (60, 505) or genetic (505) interventions that enhance O-GlcNAcylation have been shown to increase nonamyloidogenic α-secretase processing, resulting in increased levels of the neuroprotective sAPPα fragment and decreased Aβ secretion. Hyper-O-GlcNAcylation induced by an OGA inhibitor, decreased levels of Aβ40 and Aβ42 peptides in the brain, decreased plaque formation, and improved cognition in the 5xFAD amyloid-β mouse model (55). Pharmacological approaches to inhibit OGA have been used in tau overexpression models and were found to increase tau O-GlcNAcylation and decrease neurodegenerative phenotypes (60, 209). On the basis of these studies, clinical trials are currently ongoing to determine OGA inhibition in treatment of AD (506, 507); however, controversy remains as to whether lower O-GlcNAc levels contribute to AD. This was highlighted in a quantitative proteomics study, which found that in AD brains, 12 peptides exhibit ed decreased, while 119 peptides exhibited increased, O-GlcNAcylation compared with controls (508). Consequently, the long-term impact of globally increasing O-GlcNAc levels by inhibiting OGA as a treatment for AD remains unclear.

In contrast to increasing O-GlcNAc levels as a potential therapeutic approach for AD, other studies suggest that increased O-GlcNAc levels may lead to impaired neuronal function. For example, in a Caenorhabditis elegans model of neurodegeneration, the OGA inactive mutant, which is associated with increased O-GlcNAc levels, was shown to increase proteotoxicity (509). Hyper-O-GlcNAcylation also induces neuronal apoptosis in vitro, is associated with downregulation of AKT Ser-308 and Ser-473 phosphorylation, and correlates with greater tissue damage in an in vivo model of cerebral ischemia (124). Increased O-GlcNAcylation of NF-κB has also been found to be associated with retinal ganglion cell death in a model of diabetic retinopathy (430). There is also an increase in protein O-GlcNAcylation in postmortem brains of patients with Parkinson’s disease (PD) (453). α-Synuclein, a component of Lewy bodies that is involved in the pathogenesis of PD, is also O-GlcNAcylated, and this could affect α-synuclein degradation. Other O-GlcNAc-modified proteins associated with neurodegenerative diseases include tau and amyloid precursor protein (APP), involved in AD (93, 510), as well as superoxide dismutase (SOD)—a protein involved in amyotrophic lateral sclerosis (511)—and gigaxonin, mutations of which cause giant axonal neuropathy (512).

Since α-synuclein aggregation is the main component of Lewy bodies appearing in PD, as well as the nonamyloid component in AD (513–517), its modification by O-GlcNAcylation is drawing attention. In vivo proteomic studies have identified up to nine different O-GlcNAc modification sites, several within the region required for aggregation (i.e., residues 61–95) (185). Earlier studies demonstrated that α-synuclein O-GlcNAcylation at Thr-64 and Thr-72 in rat brain, and at Ser-87 in humans (92,93, 518) and in vitro O-GlcNAcylation at Thr-72 or Ser-87 decreased aggregation propensity and toxicity in cultured cells (519,520). Additional in vitro studies demonstrated that O-GlcNAcylation of Thr-72, Thr-75, or Thr-81 inhibited the nucleation step of aggregation, whereas O-GlcNAc modification of Ser-87 had little effect (185). Simultaneous O-GlcNAc modification of α-synuclein at three sites—Thr-72, Thr-75, and Thr-81—completely abrogated it aggregation (185). On the other hand, studies in primary cultured neurons found that increasing O-GlcNAc levels by inhibiting OGA increased the total levels of α-synuclein (453). It is possible that O-GlcNAcylation of α-synuclein may change its turnover or degradation; however, whether O-GlcNAcylated or un-O-GlcNAcylated α-synuclein is turned over more slowly remains unclear (453). Thus, although it is clear that α-synuclein is subject to O-GlcNAcylation, the impact of this modification on its aggregation and turnover remains to be understood. In light of the importance of α-synuclein in neurodegenerative diseases, a more comprehensive understanding of how O-GlcNAc levels impact α-synuclein homeostasis is important and would be a significant new finding at the intersection of glycobiology and neurodegenerative diseases.

As discussed in sect. 3, there is increasing evidence that O-GlcNAcylation is involved in regulating quality control of proteins and organelles through its effects on autophagy. Therefore, in addition to a direct effect on the accumulation and toxicity of proteins, such as α-synuclein, APP, tau, and SOD, their insufficient clearance due to impaired autophagy may also contribute to the pathogenesis associated with PD, AD, and dementia with Lewy bodies (DLB) (521–525). TPPP is a prime candidate linking O-GlcNAc to AD, DLB, and PD, because it has been found in Lewy bodies (526) and because it has been shown to promote α-synuclein secretion by inhibition of autophagosome-lysosome fusion (447). In primary neurons, OGA inhibition has been shown to attenuate autophagic flux (453), providing a further link between increased O-GlcNAc levels and neuronal dysfunction. Thus, the importance of O-GlcNAcylation in directly modulating toxic protein function and accumulation and indirectly modulating toxic protein accumulation by modulating autophagy will need to be further investigated.

In addition to O-GlcNAc’s roles in neurodegenerative diseases, there is evidence that acute changes in O-GlcNAc levels are involved in regulating learning and memory in vitro and in vivo via O-GlcNAcylation of the GluA2 subunit of the AMPAR receptor (200). Acute increases in O-GlcNAc levels decreased excitability of hippocampal neurons and attenuated excitatory synaptic transmission, which was also mediated by GluA2 containing AMPAR receptors (527). Increases in O-GlcNAc also decreased intrinsic neuronal excitability, suggesting that it could regulate overall excitation/inhibition balance and neuronal output (528). In Drosophila, loss of OGA activity altered locomotion and deficits in learning, suggesting that OGA and O-GlcNAc may play a conserved role in cognitive function (529). OGA^+/− mice, which have chronically elevated O-GlcNAc levels, exhibited antidepressant-like behavior, which was associated with reduced inhibitory synaptic transmission in the medial prefrontal cortex, suggesting a possible link between altered O-GlcNAc levels in the brain and depression (530).

In the context of chronic sensorimotor perturbation, such as hindlimb unloading, overall O-GlcNAc levels were unchanged; however, O-GlcNAcylation of synapsin I in the cytosol was increased, suggesting a possible role for O-GlcNAc in regulating sensorimotor synaptic plasticity (531). OGT and O-GlcNAc have also been implicated in memory consolidation and fear conditioning via the epigenetic regulation of genes by EZH2 (43). In a C. elegans model of neuronal injury and regeneration, both reducing and increasing O-GlcNAc levels were found to enhance regeneration via two distinct mechanisms (532). Lower O-GlcNAc levels increased regeneration via an AKT-dependent increase in glycolysis, whereas the positive effects of higher O-GlcNAc levels were mediated by changes in mitochondrial function. The beneficial effect of acute increases in O-GlcNAc levels was also shown in seizure models, where pharmacological increase in O-GlcNAc attenuated hyperexcitability both in vitro and in vivo (533). Epilepsy was shown to decrease OGT and O-GlcNAc levels in human brains and in brains from a rat model of epilepsy; moreover, increasing O-GlcNAc levels by OGA inhibition decreased seizure duration, as well as epileptic spike events (534).

4.4. Cardiovascular Function and Diseases

The first studies of O-GlcNAcylation in the cardiovascular system identified the small heat shock protein αB-crystallin and the transcription factor Sp1 as O-GlcNAc targets in the heart and vascular smooth muscle cells, respectively (143, 336). Both studies emphasized the dynamic nature of this modification and its potential role in regulating the function of these proteins. Dong and Hart (11) reported that the heart exhibited measurable OGA activity, and a subsequent study showed that OGT activity was higher in the rat heart than many other tissues, including liver and adipose (535). However, GFAT activity, while measurable, was much lower in heart than most tissues (535). Brownlee and colleagues (190, 536) demonstrated in endothelial cells that hyperglycemia increased O-GlcNAc levels on Sp1 and eNOS and concluded that this could contribute to the endothelial dysfunction that occurs in diabetes.

In cardiomyocytes, hyperglycemia resulted in a prolonged Ca²⁺ transients, which was mimicked by increasing HBP flux with glucosamine or inhibition of OGA (263). Interestingly, overexpression of OGT had no effect under normal conditions and did not exacerbate the effects of hyperglycemia; however, OGA overexpression blunted the effects of hyperglycemia on Ca²⁺ transients (263). A subsequent study found that OGA overexpression in the diabetic heart improved cardiomyocyte contractility, and this was associated with a reduction in overall O-GlcNAc levels and the normalization of SERCA protein levels (537). Others have suggested that increased O-GlcNAcylation of PLB, a major regulator of SERCA activity, may contribute to impaired contractility that occurs with diabetes (194). Diabetes is associated with electrical abnormalities, including prolonged QT interval, which are associated with increased risk or cardiac arrhythmias (538). Increased O-GlcNAcylation of both CaMKII (165) and the voltage-gated sodium channel Nav1.5 (539) have been implicated as contributors to increased arrhythmogenesis in diabetes. Interestingly, O-GlcNAcylation of CaMKII has also been shown to be required for increased NOX2-mediated ROS production induced by hyperglycemia (166).

It is widely accepted that diabetes leads to a sustained increase in cardiac O-GlcNAcylation and that this sustained increase in cardiac O-GlcNAcylation has been implicated in the adverse effects of diabetes on the heart (540–542). This is commonly believed to be a direct consequence of increased availability of glucose; however, this has yet to be definitively confirmed. In addition to higher cardiac O-GlcNAc levels, 2 wk after Type 1 diabetes was induced, Hu et al. (537) observed increased mRNA and protein levels of both OGT and OGA, although surprisingly there was no change in activities of either enzyme. Another study reported biphasic changes in cardiac O-GlcNAc levels with increases at 2 wk after the initiation of diabetes, followed by a subsequent decrease to normal levels at 4 wk, with a second increase after 16 wk of diabetes (543). However, mRNA levels of OGT, OGA, and GFAT1 increased only in the 16-wk diabetic group, and only OGT protein levels were elevated at that time point. Thus, while diabetes is clearly associated with increased cardiac O-GlcNAc levels, the mechanisms underlying this response remain unclear.

The adverse effects of diabetes on the heart has also been linked to alterations in metabolism characterized by decreased carbohydrate oxidation and increased dependence on lipids for energy production. As discussed in sect. 3.3, multiple components involved in insulin signaling are O-GlcNAcylated, and increased O-GlcNAcylation is found to attenuate insulin signaling. This could be a contributing factor to the reduction in carbohydrate metabolism seen in diabetes; however, this has not yet been examined in the heart. The increase in lipid metabolism in rodent models of diabetes has been attributed, at least in part, to increased plasma membrane levels on the fatty acid transporter FAT/CD36 (544). In the isolated perfused heart, the addition of glucosamine, which acutely increases cardiac O-GlcNAc levels, also increases fatty acid oxidation and leads to an increase in plasma membrane levels of FAT/CD36 (419). FAT/CD36 was found to be O-GlcNAcylated and was associated with OGT (419). The link between O-GlcNAcylation, fatty acid oxidation, and FAT/CD36 translocation has been confirmed by others (545). It is also worth noting that the addition of physiological concentrations of glutamine increased fatty acid oxidation in the perfused heart, which occurred via activation of the HBP and an increase in plasma membrane FAT/CD36 levels (420). O-GlcNAcylation of E2 and E3 subunits of PDH in the heart, has been linked to an increase in PDH activity, which was attenuated under conditions of impaired branched-chain amino acid metabolism (137). Studies have also found a link between O-GlcNAcylation and cardiac ketone body metabolism, with increased O-GlcNAc levels associated with decreased ketone body oxidation (281, 546). A large number of enzymes in glycolysis are also O-GlcNAcylated (547), although the effect of this modification on regulating glucose metabolism in the heart is unknown. Consequently, O-GlcNAcylation may contribute to the regulation of cardiac metabolism under normal physiological conditions, as well as contribute to metabolic dysregulation in diabetes.

Impaired mitochondrial function is also a common feature in the heart in response to diabetes, and O-GlcNAcylation of mitochondrial proteins has been shown to reduce mitochondrial oxidative metabolism and contribute to increased mitochondrial fission (159, 161). Others have also reported alterations in mitochondrial O-GlcNAcylation in the heart in diabetes (427). It should be noted however, that whether O-GlcNAcylation mediates the effects of hyperglycemia on mitochondrial function has been questioned (548). While the O-GlcNAcome of cardiac mitochondria is diverse (138), our understanding of how it is regulated is limited. Mitochondrial uptake of UDP-GlcNAc in cardiomyocytes is regulated by PNC1, but it is not known whether transport is mediated by primarily by UDP-GlcNAc availability or whether it is determined by regulation of PNC1 activity by allosteric inhibitors or activators such as Ca²⁺ or by regulation of expression levels by insulin and IGF-1 (549). It also remains to be determined whether O-GlcNAc cycling rates in the mitochondria are similar or different than those in the nucleus or cytosol. The mechanisms by which O-GlcNAcylation is targeted to specific proteins and residues in the mitochondria are also unknown.

Cardiac O-GlcNAc levels are also increased in hypertrophy, heart failure, and aging, although the role of O-GlcNAcylation in the development of cardiac dysfunction has not been as extensively studied as in the context of diabetes. As noted above, in diabetes, it is assumed that O-GlcNAc levels in the heart increase because of high glucose levels increasing HBP flux. The mechanism leading to high O-GlcNAc levels in other pathological conditions is less clear, as there is no primary underlying metabolic abnormality. In a rat model of pressure overload-induced hypertrophy, in addition to an increase in O-GlcNAc levels, OGT, and OGA mRNA, and protein levels were significantly increased (291). Similar changes in O-GlcNAc, OGT, and OGA were found in biopsies from patients with aortic stenosis compared with samples from nonhypertrophied hearts (291). The increase in OGT could lead to the increase in O-GlcNAc, and the increase in OGA could be an adaptive response to try and lower O-GlcNAc levels; however, the cause and effect of these changes are unknown. Interestingly, acute hypertrophic stimuli increased cardiomyocyte O-GlcNAc levels, which was linked to increased expression of OGT protein, as well as increased GFAT phosphorylation and protein levels (298, 299). These studies demonstrated that the initial increase in O-GlcNAc levels was necessary and sufficient to activate hypertrophic signaling in cardiomyocytes. Further support for a role of the HBP and O-GlcNAcylation in regulating cardiac hypertrophy was provided by studies reporting that GFAT1 overexpression stimulates growth of neonatal cardiomyocytes in culture and accelerates cardiac remodeling in a model of pressure overload-induced hypertrophy; conversely, GFAT1 deletion blunted adverse remodeling (550). Preliminary reports also demonstrate that chronic overexpression of OGT in the heart leads to a dilated cardiomyopathy, in the absence of any additional stress (551).

In addition to its oncogenic role, cMyc also regulates cardiac hypertrophy, where it is reexpressed in response to hypertrophic stimuli, and its repression decreases the degree of hypertrophy (552). Inducible, cardiomyocyte overexpression of cMyc resulted in cardiac hypertrophy that was associated with an overall increase in cardiac O-GlcNAc levels, and cMyc KO attenuated the response to hypertrophy and reduced O-GlcNAc levels (553, 554). OGT deletion either before or after pressure overload hypertrophy resulted in worse outcomes, providing further support for O-GlcNAc being required for normal hypertrophic signaling (555). It was suggested that O-GlcNAcylation of PKA was a possible mechanism by which O-GlcNAc mediates the effects of hypertrophy. Interestingly, cardiac specific overexpression of LXRα was protective against pressure overload-induced hypertrophy, and this was associated with induction of cytoprotective mechanisms via O-GlcNAcylation of the transcription factors GATA4 and MEF2c (556).

Sustained exercise is also associated with cardiac remodeling and hypertrophy that is typically considered to be beneficial. Interestingly, in contrast to pathological hypertrophy, hypertrophy induced as a result of exercise swim-training reduced cardiac protein O-GlcNAcylation under normal conditions (557), as well as in the context of diabetes (558). This was associated with a decrease in OGT protein and OGT, OGA, and GFAT2 mRNA. A single 15-min bout of exercise resulted in a decrease in nuclear, but not cytosolic, cardiac O-GlcNAc levels, which was associated with a reduced interaction between OGT and the REST chromatic repressor (559). Surprisingly, treadmill exercise for 1 and 4 wk increased cardiac O-GlcNAc levels in hearts from db/db mice, independent of changes in OGT and OGA protein levels (376). The mechanism by which exercise affects O-GlcNAc levels is not known; however, in cultured neonatal cardiomyocytes, exogenous NAD⁺ resulted in a time- and dose-dependent decrease in O-GlcNAc levels that was not accompanied by changes in OGT or OGA protein levels (344).

Although considerable attention has been paid to the adverse effects of increased O-GlcNAc levels in chronic cardiac diseases, a number of studies have demonstrated that acute pharmacological increases in O-GlcNAc levels decrease cell and tissue injury and improve cardiac functional recovery (163, 402, 437, 442). Other studies have shown that decreases in O-GlcNAc levels are associated with increased sensitivity of cardiomyocytes to oxidative stress (438). It is also worth noting that well-established cardioprotective strategies, including ischemic preconditioning and glucose–insulin–potassium therapy, have also been shown to increase cardiac O-GlcNAc levels (560–563), although the extent to which this contributes to their protective effects is not known.

Protein O-GlcNAcylation has also been shown to regulate vascular function with sustained increases associated with impaired endothelial cell and vascular smooth muscle cell function, as seen with diabetes. Increased vascular O-GlcNAc levels have also been implicated in increased vascular reactivity, thereby, directly contributing to the development of hypertension. Of note, pharmacologically increasing O-GlcNAc levels in normal vascular smooth muscle also increases vascular reactivity (173, 564). ET-1, which contributes to acute and long-term changes in vascular function, also directly increases vascular O-GlcNAc levels in an ET_A and ET_B receptor-dependent fashion (173, 296). The effects of O-GlcNAc in regulating vascular reactivity appear to be mediated via the RhoA/Rho-kinase pathway (297); however, the direct link between O-GlcNAc and the RhoA/Rho-kinase pathway has yet to be identified. In addition to regulating vascular reactivity, acute increases in O-GlcNAc have been shown to be protective by reducing endoluminal injury and subsequent infiltration of inflammatory mediators (565, 566). Increasing O-GlcNAc levels also blunted the adverse effects of TNF-α on endothelial function via suppression of inducible nitric oxide synthase expression (567). Vascular dysfunction is also a hallmark of idiopathic pulmonary arterial hypertension, which has been associated with alterations in glucose metabolism. Barnes et al. (568) found that in tissues and cells from IPAH patients that GFAT, OGT, and O-GlcNAc levels were all increased and that knockdown of OGT attenuated the proliferative phenotype of pulmonary arterial smooth muscle cells.

4.5. Skeletal Muscle

Skeletal muscle is distinctly different from cardiac muscle in both its origin (569) and its regulation by O-GlcNAcylation (570). Within healthy muscle, there are higher levels of O-GlcNAcylation in proteins in the slow oxidative soleus muscle, compared to fast glycolytic extensor digitorum longus muscle (411). Furthermore, this difference in O-GlcNAcylation was lost in the soleus muscle following mechanical unloading and muscle wasting. These authors identified 14 O-GlcNAc-modified proteins involved in signal transduction, metabolism, and contraction. Over a decade later, using more sophisticated enrichment and MS techniques, this list was expanded to 342 O-GlcNAcylated skeletal muscle proteins (571). In contrast to the earlier study, they were also able to identify specific O-GlcNAc protein modification sites and solidified the role of O-GlcNAcylation in normal sarcomere cytoarchitecture.

Consistent with the earlier suggestion of a link between protein O-GlcNAcylation levels and the muscle fiber type, skeletal muscle-specific knockout of OGT (mOGT-KO) reduced O-GlcNAc levels, increased Myh7 fiber expression, and reduced oxidative phosphorylation gene expression (379). Interestingly, this mOGT-KO protected mice from high-fat diet-induced glucose and insulin intolerance through a proposed systemic signaling by IL-15 that affects adipose tissue and whole body metabolism. This later finding speaks to an interesting O-GlcNAc-mediated signaling from muscle to influence normal physiology. In support of this skeletal muscle signaling, a separate group, also using an mOGT-KO model, identified the loss of OGT as sufficient to activate AMPK to enhance glucose utilization (485), a link that was previously identified by in vitro studies (252). In contrast to the OGT knockout models, increasing total muscle protein O-GlcNAcylation by overexpression of a naturally occurring dominant-negative OGA leads to muscle atrophy and increased mortality (205). Consequently, it is clear that O-GlcNAcylation of skeletal muscle proteins is an important molecular signaling pathway that affects both muscle and the entire organism.

Beyond the role of O-GlcNAcylation for basal physiology and muscle signaling, external stimuli can further alter skeletal muscle protein O-GlcNAc levels. For example, exercise treadmill-training, but not acute exercise, is sufficient to increase protein O-GlcNAcylation (572). These changes in protein modification appear to be most robust on cytoplasmic proteins, further supporting a role in cellular signaling. Somewhat surprisingly, myocardial infarction-induced heart failure had no effect on skeletal muscle protein O-GlcNAcylation, despite decreases in levels of both OGT and OGA protein (572). However, the latter could be explained by the 6-wk duration of the study. Subsequent work showed that in high-fat diet-induced diabetes, skeletal muscle protein O-GlcNAcylation was increased, and using both in vitro and in vivo studies, Wang et al. (573) suggest that this is directly linked to transcriptional reprogramming of mitochondria by promoting PGC-1α O-GlcNAcylation and degradation. These studies support that stimuli, such as exercise, require intact O-GlcNAc signaling as part of the adaptations seen in skeletal muscle.

As discussed in sect. 3.3, increasing HBP flux decreased skeletal muscle insulin sensitivity in normoglycemic but not diabetic conditions (574), which was related to UDP-GlcNAc levels in skeletal muscle glucose transporter 4 (GLUT4)-containing vesicles (575). Somewhat paradoxically, either short-term (16 h) or long-term (8 mo) treatment of rodents with the OGA inhibitor NButGT, increased skeletal muscle total protein O-GlcNAcylation without significantly altering glucohomeostasis (576). Further evidence for this nutrient signaling to skeletal muscle protein O-GlcNAcylation came from a study examining the combination of glutathione depletion and acute exercise training. As mentioned above, long-term exercise altered skeletal muscle, but acute exercise alone was not sufficient to produce significant changes. However, when intracellular glutathione was depleted in rats by diethyl maleate treatment before treadmill running, OGT and total protein O-GlcNAc levels were increased (577). The authors provided additional information to suggest that the mechanism of this protein O-GlcNAcylation was driven by the cellular redox state, providing a potential explanation of the exercise-induced adaptation to cellular oxidative stress. Together, these studies suggest a unique mechanism by which the source and means by which flux through the O-GlcNAcylation pathway occurs dictate its physiological consequences.

4.6. Liver

The liver is responsible for regulating whole body energy homeostasis in a continuously changing metabolic environment. Following a meal, when blood glucose levels are typically in excess, the liver takes up glucose and stores it as glycogen, whereas when glucose falls below normal, glycogenolysis is activated, resulting in glucose export to the blood. Once glycogen stores are depleted, gluconeogenesis will be activated, resulting in glucose generation from noncarbohydrate sources and when gluconeogenesis becomes limiting, such as during a prolonged fast, the liver will activate ketogenesis, facilitating the conversion of fatty acids to ketone bodies. All of these pathways are tightly regulated at the transcriptional, translational, and posttranslational levels. Insulin plays a central role in regulating liver metabolism, and as discussed in sect. 3.3, most of the components of the insulin signaling pathway are O-GlcNAcylated, and this attenuates their activity; thus, this represents a mechanism for attenuating insulin signaling in response to sustained increases in insulin (578). However, hepatic overexpression of OGT resulted in insulin resistance and increased genes associated with gluconeogenesis while decreasing lipogenic gene expression (242). Changes in O-GlcNAc cycling in the liver either by OGT overexpression or knockdown altered gluconeogenic gene expression and circulating glucose levels (330). In the liver, glucokinase plays a central role in controlling glycogen synthesis and glycolysis, thereby contributing to the regulation of glucose homeostasis. Glucokinase has been shown to be a target for O-GlcNAcylation, and higher O-GlcNAc levels increase glucokinase protein levels, most likely by increasing its stability (579). This represents a potential mechanism for the short-term increase in glucokinase activity, in response to changes in glucose levels.

In response to fasting, glucagon activates gluconeogenesis, in part, by triggering the translocation of transducer of regulated CREB 2 (CRTC2) to the nucleus. In response to refeeding, circulating insulin is increased, and inhibits this program by increasing the degradation of CRTC2. In diabetes, high circulating levels of glucose and insulin resistance cause high levels of hepatic gluconeogenesis, further contributing to hyperglycemia. Both high glucose and glucosamine increased CRTC2 nuclear translocation, and this was associated with increased O-GlcNAcylation of CRTC2 at Ser-70 and Ser-171 (127), which are key phosphorylation sites that keep CRTC2 sequestered in the cytoplasm. Increased O-GlcNAcylation of PGC-1α protects it from degradation, thereby, also promoting gluconeogenesis (118). An OGT/HCF-1 complex is required to O-GlcNAcylate PGC-1α, and the knockdown of either OGT or HCF-1 improved glucose homeostasis in Type 2 diabetic db/db mice. Both the LXRα receptor and the liver transcription factor ChREBP play key roles in regulating hepatic glucose and lipid metabolism, and both have been shown to be O-GlcNAcylated, with the levels of O-GlcNAcylation increasing in response to hyperglycemia (347, 351). Both refeeding and diabetes lead to increased LXRα O-GlcNAcylation, which, in turn, activates sterol regulatory element-binding protein 1c (SREBP-1c), a key regulator of lipogenesis (347). The nuclear receptor REV-ERBα is also involved in regulating hepatic metabolism, and a REV-ERBα/OGT complex has been shown to control SREBP-1c transcription by stabilizing OGT levels and decreasing AKT phosphorylation (65). ChREBP also interacts with OGT, resulting in increased O-GlcNAcylation and leading to stabilization of ChREBP and increased transcriptional activity toward both glycolytic and lipogenic genes (351, 580). Collectively these findings demonstrate the important role of O-GlcNAcylation in regulating hepatic metabolism in physiological and pathological conditions.

4.7. Cancer

Aberrant energy metabolism is a well-established hallmark of all forms of cancer, typically characterized by a high demand for glucose. Glutamine is also an essential nutrient that facilitates the rapid growth that occurs in cancer cells (581). The high demand for glucose and glutamine is due, in part, to their contributions to energy production; however, glutamine also contributes to a wide range of biosynthetic pathways (582). Glucose and glutamine metabolism also converge at GFAT, the first step in the HBP, leading to UDP-GlcNAc synthesis (FIGURE 4). Studies have linked increased O-GlcNAc levels to the increased survival, progression, invasion, and metastasis of many cancers (21, 210, 583, 584). In addition to increased availability of substrates for the HBP, oncogenes such as KRAS and c-Myc have been shown to increase HBP flux and O-GlcNAcylation via upregulation of GFAT (583). Disease-free survival and overall survival outcomes are significantly worse in prostate and lung cancer patients with high OGT and O-GlcNAc levels (74, 564, 584, 585), and knockdown of OGT reduces tumor growth and cancer cell proliferation (583, 584).

As discussed in sect. 3.1, O-GlcNAcylation is widely recognized as playing an important role in transcriptional regulation, and this represents one mechanism underlying its involvement in the progression of cancer. Numerous oncogenes and tumor suppressor genes are targets for O-GlcNAcylation, including p53, cMyc, HIF1α, FoxM1, cyclin D₁, and others (290, 583, 586). Almost invariably, increased O-GlcNAcylation increases their activation, either directly or indirectly, typically resulting in increased stability, translocation, and transcription (583, 584). For example, GSK3β-mediated phosphorylation at Thr-58 on cMyc is required for its degradation; thus, mutations of this residue prevent phosphorylation, reducing its degradation, converts cMyc into a potent oncogene (583, 584). O-GlcNAcylation of this same residue markedly increases cMyc stability, whereas OGT inhibition or knockdown decreases cMyc levels (66). UDP-N-acetylglucosamine pyrophosphorylase-1-like-1 (UAP1L1), which shares ∼60% homology with UAP1 (FIGURE 4), but lacks its enzymatic activity, is increased in hepatocellular carcinoma and is associated with poor prognosis (587). UAP1L1 appears to interact directly with OGT, thereby contributing to increased overall O-GlcNAc levels in hepatoma cells and increased O-GlcNAcylation of cMyc. A complex between unconventional prefoldin RPB5 interactor (URI), the protein phosphatase 1γ (PP1γ), and OGT also regulates cMyc levels by its O-GlcNAcylation, thereby contributing to its role in tumorigenesis (588). In some cases, O-GlcNAcylation of cMyc or HIF1α, contributes to the increases in glycolysis and glutamine metabolism, while in others, hyper-O-GlcNAcylation increases cell proliferation, cell survival, or invasion (583, 584). In breast cancer cell lines, hyperglycemia upregulates chemoresistance pathways via O-GlcNAcylation of Hedgehog transcription factors GLI1 and GLI2 (348), which could be a factor in the poorer outcomes of breast cancer patients that also have diabetes. Changes in methylation also contribute to cancer progression, and this can be directly affected by O-GlcNAcylation. For example, O-GlcNAcylation of EZH2 increases H3K27me3 forming a feedback loop that promotes metastasis (307). In human colon cancer stem cells, cross talk between O-GlcNAcylation and DNA methylation of the promoter region of MYBL1, a transcriptional activator, was shown to regulate tumor cell growth both in vitro and in vivo (589).

Increases in O-GlcNAcylation have also been associated with the metabolic reprogramming that occurs in cancer. This is due, in part, to transcriptional regulation of metabolism, but it can also occur via direct O-GlcNAcylation of metabolic enzymes. For example, PFK1, which catalyzes the conversion of fructose 1,6-bisphosphate from fructose-6-phosphate is subject to allosteric regulation at Ser-529 by fructose-2,6-bisphosphate (F-2,6-BP). O-GlcNAcylation of Ser-529 on PFK1 inhibits glycolysis, redirecting glucose into the pentose phosphate pathway (PPP), resulting in greater cell growth and increased resistance to oxidative stress. Conversely, blocking this modification decreased cell proliferation and tumor formation (140). In addition, glucose-6-phosphate dehydrogenase, the rate-limiting enzyme of the PPP is O-GlcNAcylated at Ser-84, increasing its activity, and further activating the PPP and increasing tumor growth (134). While most glycolytic enzymes have been shown to be O-GlcNAc targets (547), the functional consequences of many of these modifications are not known. GAPDH is O-GlcNAcylated at Thr-227, which blocks interactions between GAPDH monomers, and this has been linked to its increased nuclear translocation (128), where it has been reported to influence gene expression and telomere integrity (590). The roles of nonglycolytic functions of GAPDH in cancer remains to be elucidated. Pyruvate kinase M2 (PKM2), which plays a key role in metabolic reprogramming in cancer, is O-GlcNAcylated at Thr-405 and Ser-406, and this was shown to increase glucose utilization and glycolysis, whereas, blocking PMK2 O-GlcNAcylation attenuated tumor growth in vivo (141).

5. CONCLUSIONS

The initial discovery in 1984 of nuclear and cytosolic proteins modified by a single O-GlcNAc moiety (2) overturned widely held assumptions in glycobiology and initiated an entirely new area of research that encompasses diverse areas of cell biology, physiology, and pathophysiology. Many of the early studies examined the role of O-GlcNAc in regulating cellular physiology in the context of diabetes and hyperglycemia, conditions in which O-GlcNAc levels are chronically elevated on multiple proteins and in tissues that are adversely affected by diabetes. As our knowledge of the regulatory role of O-GlcNAc grew, it was recognized that O-GlcNAcylation also contributed to the complex coordination of cellular signaling under normal physiological conditions. The intricate relationship between O-GlcNAcylation and phosphorylation, as well as other posttranslational modifications, is slowly being unraveled. Although protein O-GlcNAc levels are undoubtedly regulated, in part, by nutrient availability, there is increasing recognition of multiple layers of regulation, including transcriptional regulation of GFAT, OGT, and OGA, as well as kinase-mediated regulation of these same proteins contributing to maintaining O-GlcNAc homeostasis. In addition, OGT binds to numerous partner proteins, which affects its stability, localization, and targets for O-GlcNAcylation. It is also increasingly clear that Ca²⁺ is another component of the O-GlcNAc regulatory network, although the specific nature of this relationship remains to be identified. Thus, although the concept of O-GlcNAc levels as a “rheostat”, reflecting nutrient availability (210, 257), is still applicable under some circumstances, it is only one of several mechanisms underlying the regulation of O-GlcNAcylation.

While research in O-GlcNAc biology and physiology continues to expand (FIGURE 3), challenges remain with regard to more rapid growth. As recently as 2019, more than 30 years after its discovery, Gerald Hart, the founder of O-GlcNAc biology, wrote that “Despite the efforts of many laboratories, this field is still in its infancy.” (210). One important limitation that has yet to be overcome is the commercial availability of site-specific O-GlcNAc antibodies, although a few such antibodies have been published for a few proteins (79–82). There have been great improvements in techniques to map O-GlcNAc modification sites using mass spectrometry. While their successful application often remains beyond the scope of many proteomic cores, efforts are continuing to improve these techniques (95). The pharmacological modulation of O-GlcNAc levels was hampered for many years by the absence of well-characterized, selective, high-affinity inhibitors of OGT and OGA. The development of the thiazoline family of OGA inhibitors by Vocadlo and colleagues in 2005 (54) substantially helped move the field forward. The subsequent development of Thiamet-G (60), a highly selective, high-affinity OGA inhibitor, which is orally available, crosses the blood-brain barrier, and rapidly increases cellular O-GlcNAc levels at concentrations of <1 µM, has been of particular value. Not until 2015 were OGT inhibitors developed with the necessary level of specificity and affinity; it is anticipated that the wider availability of these compounds will be of great value to the field (67, 69). Recent reports of other new OGT inhibitors that are effective in decreasing O-GlcNAc levels in vivo are also encouraging (77).

Conceptually, perhaps one of the biggest challenges is understanding how only a single OGT and single OGA can regulate thousands of different proteins, particularly given the lack of a clear consensus sequence. The comparison with phosphorylation is inevitable, where many hundreds of kinases and phosphatases are responsible for controlling the phosphorylation of specific proteins; thus, it can be difficult to see how OGT and OGA achieve their specificity. However, most kinases have multiple downstream targets and phosphorylation of individual targets are determined by the specific stimulus, cellular location, as well as other factors, including scaffolding proteins or other interacting proteins. AMPK, which regulates diverse cellular processes, is an excellent example, with 64 validated downstream targets and potentially many more (591, 592). The 14-3-3 protein family, which regulates diverse protein kinase signaling pathways by binding to specific phosphoserine/phosphothreonine motifs (593), is also an O-GlcNAc binding protein, with the O-GlcNAc binding site overlapping with its phosphorylation binding motif (594). This raises the possibility of 14-3-3 proteins playing an important role in regulating O-GlcNAc signaling cascades, as well as potentially integrating O-GlcNAc and phosphorylation signaling pathways.

It has been estimated that OGT could have as many as 800 interacting partners, a large number of which are linked to transcription and metabolism (118). Although less studied, OGA has been reported to have at least 90 binding partners, many of which change in response to stress (595). Although the biological significance of many of these interactions have yet to be characterized, these findings clearly demonstrate potential mechanisms for achieving selective targeting and removal of O-GlcNAc from proteins. There is undoubtedly considerable scope for studies to better understand the regulation and function of interacting partners of both OGT and OGA. Another important goal for future studies is understanding the temporal and spatial changes in cellular O-GlcNAc levels, this will include improving our knowledge of functional consequences of many of the PTMs on OGT and OGA. New technologies are being developed to better understand OGT and OGA substrate recognition, O-GlcNAc sensors targeted to specific intracellular targets, and nanobodies that can be used to direct OGT to specific protein targets (596, 597), will help improve our understanding of the role of O-GlcNAc in physiological regulation of cellular function.

The importance of O-GlcNAc modification of proteins in regulating cellular and organismal physiology was slow to emerge, due, at least in part, to the lack of basic tools However, as we have discussed here, protein O-GlcNAcylation is now recognized as playing an important role in the physiology of most organ systems, including specific processes, such as contractility in cardiac and skeletal muscle, gluconeogenesis in the liver, hippocampus-based learning and memory, vascular function, as well as universal processes, such as mitochondrial function, metabolic regulation, autophagy, calcium signaling, transcription, and epigenetics. Therefore, with the continued development of new pharmacological and molecular tools, the opportunity for future research into the fundamental physiological role of protein O-GlcNAcylation, as well as its contribution to the progression of numerous diseases, remains wide open.

GRANTS

This work is supported in part by National Institutes of Health Grants P30AG050886 (to J. Z.), R56AG060959, I01BX004251, and R01HL142216 (to J.Z. and J.C.C.), R21HL152354 (to J.C.C and A.R.W), and R01HL133011 (to A.R.W.).