Critical observations that shaped our understanding of the function(s) of Intracellular Glycosylation (O-GlcNAc)

Abstract

Almost 100 years after the first descriptions of proteins conjugated to carbohydrates (mucins), several studies suggested that glycoproteins were not restricted to the serum, extracellular matrix, cell surface, or endomembrane system. In the 1980’s, key data emerged demonstrating that intracellular proteins were modified by monosaccharides of O-linked β-N-acetylglucosamine (O-GlcNAc). Subsequently, this modification was identified on thousands of proteins that regulate cellular processes as diverse as protein aggregation, localization, post-translational modifications, activity, and interactions. In this Review, we will highlight critical discoveries that shaped our understanding of the molecular events underpinning the impact of O-GlcNAc on protein function, the role that O-GlcNAc plays in maintaining cellular homeostasis, and our understanding of the mechanisms that regulate O-GlcNAc-cycling.

Intracellular glycosylation, the early years

The first clues that protein glycosylation could occur independent of the endoplasmic reticulum (ER) and Golgi apparatus emerged in the 1970s (Figure 1). Key data included labeling of nuclear and cytoplasmic proteins with lectins including Concanavalin A (Con A) and Wheat germ agglutinin (WGA), metabolic labeling of cells with radiolabeled carbohydrates followed by subcellular fractionation, and the identification of hexosaminidases that localized to the cytoplasm (Reviewed in Hart et al., 1989[1]). In 1984, Torres and Hart reported a robust series of experiments demonstrating that lymphocyte proteins were modified by a βGlcNAc-residue through an O-glycosidic bond, and that these residues were not extended into more complex glycans (O-GlcNAc)[2]. A series of control experiments, performed to demonstrate that single O-linked βGlcNAc-residues were not an experimental artifact, unexpectedly led to the conclusion that the majority of the O-GlcNAc-residues modified intracellular proteins. Specifically, lymphocytes were labeled with (³H)Glucosamine prior to labeling of intact or permeabilized cells with β−1–4Galactosyltransferase (Gal-T) and UDP-(¹⁴C)Galactose. Analysis of the β-eliminated product (generates an alditol) revealed that in intact lymphocytes 8% of the labeled glycan was the disaccharide (¹⁴C)Galctoseβ1–4(³H)GlcNAcitol; whereas, in permeabilized lymphocytes, 60% of labeled glycan was the disaccharide (¹⁴C)Galctoseβ1–4(³H) GlcNAcitol [2]. Over the next three years, numerous studies demonstrated unequivocally that nuclear pore proteins were modified by O-GlcNAc, as well as other cytosolic and nuclear proteins [3–11]. Thousands of proteins have been now demonstrated to be modified by O-GlcNAc and these proteins fall into broad functional classes that include: DNA and RNA binding proteins, epigenetic modulators, ribosomal proteins, proteasomal proteins, kinases, lysine deacetylases, metabolic enzymes, and cytoskeletal proteins [12].

Figure 1.

Timeline of critical observations that shaped our understanding of the O-GlcNAc modification.

While initially studied in mammals, O-GlcNAc is a common modification of metazoan and plant proteins [12,13]. There are sporadic reports of proteins modified by a single O-linked β-GlcNAc-residues in eubacteria [14,15], as well as protists and fungi [16–19]. However, the relationship of these glycan modifications to O-GlcNAc is unclear due to the following observations: Firstly, single O-linked β-GlcNAc-residues modify some epidermal growth factor (EGF)-repeat containing cell surface proteins. Extracellular O-GlcNAc (eO-GlcNAc) is catalyzed by an ER-resident enzyme (UniProtKB Q5NDL2) that is unrelated to the O-GlcNAc transferase (OGT, the enzyme that catalyzes the addition of intracellular O-GlcNAc (UniProtKB O15294)) [20,21]. Secondly, the detection of O-GlcNAc in protists is challenging as many approaches used to detect O-GlcNAc also cross-react with single O-linked αGlcNAc-residues, which are a common modification of cell surface proteins (for example, Dictyostelium discoideum [22]), as well as abbreviated N-linked glycans (for example, N-linked di-N-acetyl-chitobiose of Plasmodium falciparum [23,24]). Thirdly, vascular plants and mosses have two homologs of OGT: Spindly (SPY) and Secret Agent (SEC) [18,25]. Phylogenetic analyses demonstrates that sequences related to SPY and SEC are not found in archaea, many protists (Eimeria, Entamoeba, Leishmania, Plasmodium, Spironucleus, Theileria, Trichomonas, and Trypanosoma), and several yeasts (Saccharomyces, Schizosaccharomyces, Candida, and Cryptococcus) [18]. Lastly, the aforementioned phylogenetic analysis partitions OGT sequences into SPY- or SEC-like, with the metazoan OGTs classed as SEC-like [25]. A recent study demonstrates that SPY, isolated from Arabidopsis thaliana, is a fucosyltransferase [26]. These data suggest that an alternative glycan may be used in some eukaryotes, especially those whose genomes contain orthologs of SPY. Furthermore, while there is no ortholog of either SEC or SPY in Sacharomyces cerevisiae, recent data suggests that this organism may utilize mannose to modify nuclear and cytosolic proteins [27].

1987: Identification of the first O-GlcNAc reactive antibody and other tools for studying O-GlcNAcylation

Why did O-GlcNAc remain undetected for so long? Proteins modified by O-GlcNAc present several challenges to researchers that include: 1) O-GlcNAc does not typically alter the migration of a protein on an SDS-PAGE gel; 2) Unlike phosphorylation, O-GlcNAc does not impact the isoelectric point of proteins; 3) the linkage between amino acids and O-GlcNAc is labile in many forms of mass spectrometry; and 4) the ionization of glycopeptides is often suppressed during mass spectrometry [12,28]. To overcome these and other challenges, researchers have developed a toolkit for the detection, visualization, and site-mapping of O-GlcNAc-modified proteins.

Raised against nuclear pore proteins, RL2 was the first antibody characterized that detected the O-GlcNAc-modification [5,11]. While generally considered a pan-specific antibody, the epitope of RL2 includes elements of the protein backbone. Additional pan-specific antibodies have been developed and characterized, including CTD110.6, 18B10.C7(3), 1F5.D6(14), 9D1.E4(10), and HGACs 39, 49, and 85 [29–32]. While these antibodies have less dependence than RL2 on the protein backbone, they appear to have overlapping specificities and in some cases bind terminal GlcNAc residues of complex-glycans as well as eO-GlcNAc [30–36]. A handful of site-specific O-GlcNAc-reactive antibodies have been developed, aiding the study of c-Myc, Histone H2, Sirtuin 1 (SirT1), Tau, and Tak-1 binding protein 1 (TAB1) [37–42]. Lectins such as WGA are also useful; however, WGA recognizes all terminal N-acetylglucosamine- and sialic acid-residues. Binding to the latter can be suppressed by succinylation; however, this modification reduces the affinity of WGA for GlcNAc [43]. Additional lectins used to analyze O-GlcNAc include: Agrocybe aegerita lectin (AANL or AAL2), Psathyrella velutina lectin (PVL), Nicotiana tabacum agglutinin (Nictaba), and Helix pomatia agglutinin (HPA) [44–47]. The utility of lectins and antibodies for the detection and analysis of O-GlcNAcylated proteins is dependent on controls. Suggested controls include competition with free GlcNAc, removal of N-linked glycans with Peptide-N-Glycosidase F (PNGase F), on-blot β-elimination, or modulation of O-GlcNAc cycling enzyme expression or activity [34,48]. While useful, treatment of proteins with a β-N-hexosaminidase does not typically discriminate between O-GlcNAc and terminal GlcNAc residues displayed on other glycans.

Gal-T, which was originally used to detect O-GlcNAc [2], remains the gold standard for the detection and characterization of O-GlcNAc modified proteins. Gal-T is reported to only modify βGlcNAc residues providing linkage confirmation, and the incorporation of a labelled Galactose provides a useful tag for identifying proteins as well as glycosylation sites [49]. This approach has been adapted to modify O-GlcNAc with unnatural sugars (azido or ketone sugars) [50–52] and relies on the use of a Gal-T with an expanded active site (Y289L) [50,53]. Subsequently, unnatural sugars are derivatized using the Staudinger-ligation or cyclo-addition to introduce bioorthogal groups (colloquially referred to as Click Chemistry) that facilitate the detection, visualization, enrichment, and quantitation of O-GlcNAc-modified proteins and peptides [50–52,54–58]. Metabolic labeling can also be utilized to incorporate unnatural sugars [54–59], although this approach is less efficient and can result in the non-specific labeling of cysteine residues [60]. The latter data reinforces the selection of appropriate controls and the need for careful annotation of spectra.

As discussed briefly above, enrichment of O-GlcNAc-modified proteins and peptides is often required prior to site-mapping. A number of approaches have been developed to facilitate enrichment including affinity matrices [33,61,62] and techniques that derivatize O-GlcNAc providing an affinity handle [50,51,54,58,59,63–66]. The development of electron transfer dissociation (ETD) and electron-capture dissociation (ECD) tandem-mass spectrometry, in which the glycosidic bond is stable [67,68], has promoted the identification of thousands of O-GlcNAc-modification sites [13,50,51,54,56,58,59,61,62,65,66,69–72]. Many of these sites are collated in databases stored at PhosphoSitePlus [73], OGTSite [74], MS-Viewer [75], or dbOGAP [76].

1988: The first sites of O-GlcNAcylation

The initial sites of O-GlcNAcylation were mapped using a combination of labeling with Gal-T and Edman degradation or fast-atom bombardment mass-spectrometry [4,63,77–85]. The resulting data revealed that sites of O-GlcNAcylation were often rich in serine, threonine, valine and proline and that these sequence motifs resembled those used by proline-directed mitogen activated protein kinases [86]. On some proteins O-GlcNAc was mapped to known phosphorylation sites (for example, Thr58 of c-Myc) [77,84,87]. Collectively, these data led to the Yin-Yang hypothesis, which proposed that one molecular mechanism by which O-GlcNAc impacted protein function was by blocking phosphorylation of key residues either directly (same-site) or by steric hindrance (nearby-site) [86]. This hypothesis was further supported by data demonstrating that: 1) the use of phosphatase inhibitors could depress O-GlcNAcylation [88–90]; 2) for RNA polymerase II, the addition of a single GlcNAc residue to a 70 amino acid fragment of the C-terminal domain could block phosphorylation in vitro, and reciprocally, that phosphorylation could block O-GlcNAcylation [91]; and 3) the enzymes which cycle the O-GlcNAc-modification can be found in complex with phosphatases and kinases, suggesting that some serine and threonine residues are rapidly cycled between phosphorylated and O-GlcNAcylated [92,93].

Initial reports in a compiled database of O-GlcNAc-modified proteins, dbOGAP, reported that 42 of 357 sites had the potential for same-site reciprocity, suggesting that Yin-Yang hypothesis was relevant to a subset (~10%) of O-GlcNAcylated proteins [76]. Several high-throughput studies confirm that reciprocity is not as common as originally proposed. Trinidad and co-workers identified 1750 and 16,500 sites of O-GlcNAcylation and phosphorylation, respectively, on proteins isolated from murine synaptosomes. Only 7% of glycosylation sites were also found to be phosphorylation sites [71]. Supporting these data, Wang and co-workers mapped phosphorylation and O-GlcNAcylation sites on proteins isolated from an enrichment of mitotic spindles and mid-bodies. Similar to the aforementioned study, only 6% of 141 identified glycosylation sites were identified as phosphorylation sites. Notably, augmenting OGT expression in this model decreased phosphorylation (by more than 50%) on 17% of proteins, and counterintuitively, enhanced phosphorylation on 7% of proteins (by more than 200%) [94]. A number of additional studies suggest that the relationship between O-GlcNAc and phosphorylation may be more complex than originally thought. Quantitative phosphoproteomics of OGT wild-type and null cells identified 5529 phosphorylation sites, of which 232 phosphosites were upregulated and 133 downregulated in the absence of O-GlcNAc [95]. Finally, a number of kinases or kinase-regulating proteins have been demonstrated to be modified and regulated by O-GlcNAc [40,96–101].

1990: Purification and Cloning of the O-GlcNAc transferase

The existence of nuclear and cytoplasmic glycosylation challenged the “glycosylation dogma” as in 1984 there were no biosynthetic or transport pathways that would result in the synthesis or accumulation of nuclear and cytoplasmic glycoproteins. Studies focused on the synthesis of nuclear pore protein (Nup) p62 demonstrated that O-GlcNAcylation occurred within 5 min of synthesis and prior to incorporation into the nuclear membrane [7]. Newly synthesized and glycosylated Nup p62 was associated with the post-microsomal fraction, strongly suggesting that glycosylation occurred in the cytosol. In 1990, Haltiwanger and co-workers used sequence information from Nup p62, band 4.1, and p65 to develop a synthetic peptide (YSDSPSTST) that was rapidly O-GlcNAcylated when incubated with washed reticulocyte membranes [102,103]. These studies led to the purification and cloning of a UDP-N-acetylglucosamine:peptide N-acetylglucosaminyl-transferase (EC 2.4.1.255), better known as the O-GlcNAc transferase or OGT [104,105]. OGT is comprised of two major domains: an N-terminal tetratricopeptide repeat (TPR) domain and a C-terminal catalytic domain that contains two Rossmann-like folds [104–106]. The TPR domain, which in full length OGT contains 13 TPR repeats, forms an elongated superhelical structure that mediates dimerization (Isoform 1 – 1034 and Isoform 3 – 1046 amino acids). Molecular dynamics simulations suggest that there is a hinge between the TPR and catalytic domains that is capable of significant motion and likely critical for substrate docking [106]. There are at least four isoforms of OGT (UniProtKB - O15294; [107]), which differ in the number and composition of the TPR repeats and localize to the nucleus, cytoplasm, and mitochondria of mammalian cells [104,105,108,109].

OGT has a preference for protein sequences that are intrinsically disordered and that have certain amino acids in close proximity to the glycosylation site, including proline and valine residues [110]. This preference for certain amino acids is explained by the crystal structure, which revealed that OGT interacts with amino acids in the acceptor protein/peptide at residues +1, +3, and −1 in relation to the glycosylation site [106]. In spite of these data, no consensus motif which dictates glycosylation has been identified and this is likely due to OGTs requirement for multi-site interactions with substrates that extend beyond the active site. Underpinning this conclusion are the following observations: First, deletion of the N-terminal 6 TPR repeats of OGT blocks the ability of OGT to modify some protein substrates such as Nup p62 [111,112], but not peptide substrates [111,113]. Second, the RNA polymerase II peptide (YSPTSPS) is not a significant substrate for OGT until there are at least 5 repeats [91], in spite of the observation that shorter peptides, such as one derived from casein kinase 2 (CK2; PGGSTPVSSANMM), are robust substrates [113]. Third: mitochondrial OGT (Isoform 2) and full-length OGT (Isoforms 1 and 3) differ only in the composition of their TPR domains [108,109], yet, they appear to have overlapping substrate specificity when assayed against full-length proteins in vitro. For example, full-length OGT modifies Nup p62, O-GlcNAcase, CK2, and Tau, whereas mitochondrial OGT modifies Nup p62, CK2, and Yes-Kinase [114]. Fourth, incubation of OGT (Isoform 1) with just the TPR domain inhibits O-GlcNAcylation of a protein substrate, but not the CK2 peptide [115]; Finally, structural studies of OGTs TPR-domain highlighted an asparagine ladder in the elongated superhelix that is phylogenetically conserved and predicted to be important for substrate binding [116]. By tethering substrate-sequences to the OGT catalytic domain, Rafie and co-workers were able to crystalize OGT in complex with peptides derived from TAB1, collapsin response mediator 2 protein (CRMP2) and CK2. These studies demonstrated that the TPR domain of OGT has numerous interactions with substrates and that mutation of 5 asparagine residues (5N5A) within the asparagine ladder blocks these interactions, as well as O-GlcNAcylation of substrates [110]. Reduced glycosylation of proteins in cell lysates expressing OGT 5N5A, as well as on protein microarrays O-GlcNAcylated with OGT 5N5A, further support these data [117].

The multi-site interaction of OGTs TPR domain with substrates strongly suggests that protein interactors or post-translational modifications may impact substrate selectivity and/or activity of OGT. OGT is O-GlcNAc-modified, acetylated, phosphorylated, S-nitrosylated and ubiquitinated. The impact of only a subset of these modifications has been defined. In fine-tuning the circadian clock in hepatocytes, glycogen synthase kinase 3β (GSK3β) phosphorylates OGT at sites including Ser03 and Ser04 (Isoforms 1 and 3) resulting in activation of OGT [118]. In muscle cells, phosphorylation at Thr454 (isoform 3; Thr444 in isoform 1) is catalyzed by AMP-dependent protein kinase (AMPK) and augments OGT nuclear localization in glucose-deprived muscle cells [98]. Two recent studies demonstrate that OGT is phosphorylated at Ser20, a residue that is only found in OGT isoform 3. The first study demonstrated that Ser20 phosphorylation by checkpoint kinase 1 (Chk1) is essential for localization of OGT to midbodies during cytokinesis in HeLa cells [99], and the second study determined that phosphorylation by calcium/calmodulin dependent kinase (CaMK) II altered the pattern of O-GlcNAcylation in hepatocytes [119]. Three recent studies have identified several sites of O-GlcNAcylation on OGT isoform 4 [120], two sites at the N-terminus (Ser03, Ser04) of OGT (Isoform 1 and 3)[118], and one site at Ser389 on OGT (Isoform 1) [121]. The latter site has been implicated in promoting nuclear import of OGT in HeLa cells. Finally, OGT is basally S-nitrosylated in macrophages, inhibiting catalytic activity, and is activated by de-nitrosylation in response to lipopolysaccharide stimulation [122]. A number of studies have identified OGT-interacting proteins [115,123–132]. Collectively, these studies highlight roles for p38 Map kinase and Ataxin-10 in substrate targeting and/or activity of OGT [123,127]. In contrast, trafficking kinesin-binding protein (Trak)-1 (formerly OIP106), Trak-2 (formerly OIP98; also known as Milton/Grif-1), Chk1, ten-eleven translocation (TET), Host-cell factor-1 (HCF1), mSin3A, and Aurora kinase are thought to impact the localization of OGT [93,99,126,129–134]. Milton/GRIF-1 also interact with Kinesin, suggesting that OGT plays a role in mediating organelle transport [131,135,136].

1994: Identification of the O-GlcNAcase

Short term activation of T-lymphocytes with Con A, combined with Gal-T labeling, demonstrated a rapid and transient reduction in O-GlcNAcylation of cytosolic proteins [137]. Further, pulse-chase studies demonstrated that the half-life of O-GlcNAc on αB-Crystallin, cytokeratin 8 (K8), and cytokeratin 18 (K18) was less than that of the protein backbone [138,139]. Together, these data suggested that O-GlcNAc was dynamically added and removed from proteins in a manner analogous to protein phosphorylation. Thus, there must be an O-GlcNAc-eraser in addition to the O-GlcNAc-writer (OGT). Mammalian cells express at least four hexosaminidases, which include the lysosomal hexosaminidases (Hex) - Hex A and Hex B, as well as two neutral hexosaminidases - Hex C and Hex D. The latter has recently been characterized and will cleave both pnitrophenol (pNP)-GlcNAc and pNP-GalNAc, but has a strong preference for pNP-GalNAc [140]. In 1994, Dong and co-workers isolated and characterized a hexosaminidase from rat spleen that, unlike the lysosomal hexosaminidases, had a neutral pH optima, was not inhibited by free GalNAc, did not cleave pNP-GalNAc, localized to the nucleus and cytoplasm, and efficiently cleaved O-GlcNAc from peptide substrates [141]. Subsequently, O-GlcNAcase (OGA; EC 3.2.1.169), or Hex C, was cloned and given the gene name Mgea5 based on its homology to a hyaluronidase [142–145]. The Human Genome Organization (HUGO) has renamed the gene to Oga to reflect biochemical and genetic data that demonstrates that the Mgea5 protein product is critical for the removal of O-GlcNAc [142,143,146–148]. Like OGT, OGA localizes to the cytoplasm, mitochondria, and nucleus; however, the degree of nuclear localization appears to be cell line/tissue dependent [141–143,149,150].

Close inspection of the OGA sequence demonstrates homology between human OGA (hOGA) and OGA of Drosophila melanogaster (55%) and Caenorhabditis elegans (43%) [142]. Orthologs of OGA are present in bacteria, but appear absent from protists, plants, and fungi (with the exception of Histoplasma) [18]. Sequence analysis demonstrates that OGA (residues 1–916; UniProtKB O60502) has an N-terminal domain with homology to hyaluronidases (residues 60–366) and a C-terminal domain (residues 707–916) with homology to histone acetyltransferases [142,151–155]. The intervening sequence (residues 367–707) is comprised of a stalk domain followed by a region of low complexity region (residues 396–553) [156]. A shorter isoform (677 amino acids) of OGA has also been characterized, and only differs from the full-length isoform in the C-terminal 15 amino acids [144,145]. The catalytic residues are found in the N-terminal domain [157,158]; however, early structural evidence from the Oceanicola granulosus OGA demonstrated that residues from the intervening sequence participate in forming a substrate-binding groove [156]. Strikingly, mutational analysis of the human hOGA identified a number of residues in the intervening sequence that differentially impacted deglycosylation of TAB1, Cyclic AMP-responsive element-binding protein (CREB) 1, and Forkhead box (Foxo) 1, without effecting hydrolysis of the pseudosubstrate 4-methylumbelliferyl GlcNAc [156].

OGA is cleaved by Caspase 3 at Asp413, which localizes to an unconserved region of the intervening sequence [142,143,159]. While the biological impact of this cleavage is uncharacterized, the resulting N- and C-terminal domains of OGA remain associated. Interestingly, the N-terminal domain is not active unless co-expressed with the C-terminal domain [159]. Three recent structures of the N-terminal domain of hOGA provide molecular insight into this curious observation [160–162]. Each study addressed the challenge of crystalizing hOGA by removing the intrinsically disordered regions of OGA, as well as the C-terminal domain, while retaining residues from the N-terminal domain (residues 60–400, 11–396, or 14–400) and the intervening sequence (residues 552–704, 535–712, or 544/554–705). These studies revealed that OGA forms an unusual arm-in-arm homodimer that is mediated by a helix located in the intervening sequence. Dimerization was essential for activity, and in accordance with the structure of the O. granulosus OGA, the exchanged helix contributes to the substrate-binding cleft.

Analysis of the apparent molecular weight of OGA in bovine brain using Size-exclusion chromatography suggests that it exists in a protein complex. In spite of numerous purification steps, OGA co-purifies with 6 proteins (Tata-binding protein interacting protein-120, heat shock protein 110, heat shock cognate 70 (HSC70), cullin, amphiphysin, calcineurin and dihydropyrimidinase related protein-2) [142,143]. Thus, like OGT, OGA may be regulated by protein-protein interactions. Numerous interactors have been identified including Fatty Acid Synthase, whose association inhibits OGA activity [149]. OGA is often found in a complex with OGT [93,163–165]; the association of these two proteins with proteins such as the GATA-1·FOG-1·Mi2β repressor complex, targets O-GlcNAc cycling to promoters [166]. OGA and OGT are also targeted to midbodies in complex with Aurora B Kinase and Protein Phosphatase 1. Supporting a role for targeting dynamic cycling of O-GlcNAc to specific subcellular loci, overexpression of either OGT or OGA results in cytokinesis defects [93,167]. Finally, OGA is post-translationally modified by acetylation, phosphorylation, O-GlcNAcylation, sumoylation, and ubiquitintion [51,168–174]. However, the role of these post-translational modifications has not been defined.

2000: The critical role of O-GlcNAc in cell and tissue homeostasis

Efforts to generate OGT null embryonic stem (ES) cells via homologous recombination revealed that OGT was X-linked and that ES cells required at least one functional copy of OGT for survival. Using conditional mutagenesis, in combination with the Cre-deletor ZP3-Cre, Shafi and co-workers demonstrated that OGA was essential for mouse embryogenesis with lethality observed at E4.5 [175]. Theoretically, as a result of X-inactivation, female heterozygous knock-out mice should be a mosaic for OGT expression. Notably there is no difference in OGT expression or activity in heterozygous mice when compared to wildtype mice [176]. Underpinning this phenotype, recent studies suggest that the ogt gene may escape X-inactivation [177]. Subsequently, the authors inactivated OGT in select tissues. Deletion of OGT in T-lymphocytes (Lck-Cre) results in an reduction of peripheral thymocytes, a result of lymphocyte apoptosis [176]. The use of the cre-deletor Syn1-Cre, which drives expression in neuronal cells during development (detectable at E12.5), resulted in a reduction of litter size (50%). Surviving mice failed to develop normal locomotor activity or nurse and died within 10 days of birth [176]. Lastly, OGT was inactivated in mouse embryonic fibroblasts. Three different studies have demonstrated that immortalized mouse embryonic fibroblasts lacking a functional OGT allele are not viable, with the timing of cell death dependent on the method by which Cre-was introduced [148,176,178].

The data discussed above suggested that OGT was essential for cell and thus tissue viability. However, subsequent work has demonstrated that deletion of OGT is not always lethal [118]. There are several possibilities for the discrepancies in phenotypes: firstly, O-GlcNAc may be more essential in differentiating or rapidly-dividing cells; secondly, the impact of OGT deletion may highlight different metabolic or signaling requirements of cells and tissues; and lastly, as observed with both inactivation of OGT from the heart and pancreas, some differences may reflect the time in development at which OGT was deleted or the overall timeframe of the deletion [179–181]. Nonetheless, these studies have highlighted roles for OGT in regulating heart function (heart specific; α-myosin heavy chain promoter) [180,181], circadian rhythm (whole body; inducible) [118], nerve conduction and neuromuscular dysfunction (P₀-Cre mice) [182], decreased PKA-CREB signaling (neurons, CamK2α-Cre) [183], cold-induced thermogenesis (Brown Fat; Ucp-1-Cre) [184], defects in thermal sensitivity (DRG sensory neurons; Nav1.8-Cre) [185], resistance to acetaminophen-induced liver injury (hepatocytes) [186], altered feeding behavior (PVN neurons; Nav1.8-Cre or brn3a-CreERT2) [187], neuronal activity and browning of white fat (AgRP neurons; AgRP-Cre+) [188], pancreatic beta cell function [179], starvation induced autophagy (Adenovirus (AAV) Cre) [119]; skeletal muscle glucose metabolism (HSACre/+; HSA-rtTA/TRE-Cre) [189]; sensing placental stress (B6-CYP19-Cre) [190,191], and lung tumorigenesis (AAV Cre) [192].

Unlike deletion of OGT, ablation of OGA expression in transformed mouse embryonic fibroblasts is not lethal [148]. However, deletion of OGA in murine models results in lethality either late in gestation or immediately after birth [146,147]. Several phenotypes underlie these observations; OGA null MEFs are characterized by mitotic defects and genomic instability [146]. Using a different strategy to inactivate OGA, parallel studies identified low circulating glucose levels, reduced glycogen stores, defects in insulin signaling, and significant changes in transcription [147]. This genetic model has subsequently been used to assess the impact of O-GlcNAc cycling on colorectal cancer formation [193], energy expenditure and tissue browning [194], myogenesis [195], and neurodevelopmental timing and metabolism [196].

In C. elegans, loss-of-function ogt-1 and oga-1 animals are viable [197–199]. Interestingly, the ogt-1 and oga-1 null worms phenocopy each other, demonstrating decreased storage of triglycerides and increased glycogen and trehalose stores; however, the latter phenotype is exaggerated in ogt-1 null animals. Changes in nucleotide sugar biosynthesis may underpin this phenomenon; steady-state levels of UDP-HexNAc and UDP-glucose are substantially elevated in ogt-1 null animals and to a lesser extend in the oga-1 null. Changes in nucleotide sugar biosynthesis are accompanied by an elevation in the transcription of key enzymes in the hexosamine biosynthetic pathway (HBP), as well as trehalose metabolism. These data suggest that O-GlcNAc cycling is critical for the regulation of macronutrient storage.

OGT and OGA have also been inactivated in Drosophila melanogaster. The oga^Del.1 mutant flies demonstrate a semi-penetrant oogenesis defect. In spite of changes in their transcriptional program, oga^Del.1 flies are otherwise viable and fertile [200]. OGT is encoded by the Polycomb group (PcG) gene super sex combs (sxc). PcG factors are required for long-term repression of hox genes. Unlike other PcG proteins, sxc alleles are recessive and cause lethality at the pupal stage. The location of the mutations in the sxc alleles, as well as complementation experiments with catalytically dead OGT constructs, confirm that OGTs glycosyltransferase activity is required for hox repression. O-GlcNAcylation of the Polyhomeotic (Ph) protein is thought to stabilize this protein preventing it from aggregating, thus enabling Ph protein to be a functional component of the polycomb repressive complex 1 (PRC1).

2002: Feast or Famine, the role of O-GlcNAcylation in nutrient sensing

The HBP generates UDP-GlcNAc, the nucleotide sugar utilized by OGT, from glycolytic intermediates (Figure 2) [201]. These data, combined with observations reporting that O-GlcNAc levels are responsive to changes in extracellular glucose levels, suggested that the HBP and protein O-GlcNAcylation may act as a nutrient sensor [104,202–204]. This model was extended to suggest that dysregulation of O-GlcNAcylation may contribute to insulin resistance and glucose toxicity [202,205]. Underpinning this model are key observations reporting that changes in metabolite flux (glucosamine), as well as genetic manipulation of enzymes within the HBP, induce insulin resistance in cell culture and animal models [202,206–208]. In 2002, direct evidence linking aberrant O-GlcNAcylation to the development of insulin resistance was reported. In sum, inhibition of OGA with O-(2-Acetamido-2-deoxy-D-glucopyranosylidenamino) N-phenylcarbamate (PUGNAc) led to insulin resistance in 3T3-L1 adipocytes and skeletal muscle, and overexpression of OGT in muscle and fat led to insulin resistance and hyperleptinemia in mice [209–211]. Viral overexpression of OGT in the liver induces insulin resistance [212], whereas a liver specific knockout of OGT improves glucose homeostasis in diabetic mice [128]. Further supporting this model, elevated levels of O-GlcNAc have been observed in numerous diabetic models [189,209,213–215]. Several studies suggest that the role of O-GlcNAc cycling in the pathophysiology of diabetes is more complicated than originally thought. Changes in cell physiology resulting from elevation of extracellular glucose have been attributed to osmotic effects, and elevation of O-GlcNAc with the OGA inhibitor Thiamet-G in animal models does not recapitulate the data discussed above which utilized PUGNAc [216,217].

Figure 2. The biosynthesis of O-GlcNAc and UDP-GlcNAc.

O-GlcNAc is cycled by just two enzymes; the O-GlcNAcase (OGA) catalyzes the removal of O-GlcNAc, whereas the O-GlcNAc transferase (OGT) catalyzes the addition of O-GlcNAc. OGT utilizes the nucleotide sugar UDP-GlcNAc, which is synthesized from glycolytic intermediates by the hexosamine biosynthetic pathway (HBP). Critical enzymes include: Hexokinase (Hxk), Glucose-6-phosphate isomerase (Gpi), Glutamine-fructose-6-phosphate aminotransferase 1 (Gfpt1), Glucosamine-phosphate N-acetyltransferase 1 (Gnpnat1), phosphoglucomutase 3 (Pgm3/Agm1), and UDP-N-acetylglucosamine pyrophosphorylase 1 (Uap1/2).

The potential of O-GlcNAc to regulate insulin signaling and gluconeogenesis is highlighted by O-GlcNAcylation of proteins such as the insulin receptor (IR), insulin receptor substrate (IRS), PI3K, AKT/PKB, Foxo, PGC1α, and cyclic adenosine monophosphate response element-binding protein 2 (CRTC2) [128,218–222]. Insulin stimulation induces phosphorylation of the IR prior to O-GlcNAcylation, suggesting a model in which O-GlcNAc acts to attenuate insulin signaling. Such a model is supported by trafficking of OGT to the plasma membrane upon insulin stimulation and the observation that elevating O-GlcNAc promotes phosphorylation of IRS1 at Ser307 and Ser632/635 [212,223]. The latter phosphorylation events have been correlated with attenuation of insulin signaling [212]. In sum, these data suggest that prolonged O-GlcNAcylation could acts as a break on the insulin signaling pathway contributing to insulin resistance. Three lines of evidence suggest that hyper-O-GlcNAcylation may promote gluconeogenesis and thus potentiate hyperglycemia. First, binding of PGC1α is thought to target OGT to the Foxo transcription factors, resulting in increased transcriptional activity and expression of gluconeogenic genes [125,224,225]. Second, HCF1 was demonstrated to recruit OGT to PGC1α, which resulted in enhanced O-GlcNAcylation and stabilization of PGC1α [128]. Third, 14-3-3 proteins sequester CRTC2 in the cytoplasm by binding phosphorylated Ser171. O-GlcNAcylation of CRTC2 can be induced in primary hepatocytes at Ser70 and 171 by exposure to glucose or glucosamine. Enhanced O-GlcNAcylation correlates with increased nuclear accumulation of CRTC2 and gluconeogenic gene expression [222]. In diabetic animals, the levels of CRTC2 O-GlcNAcylation were enhanced, as was gluconeogeneic gene expression. The latter could be rectified by overexpression of OGA [222].

The use of genetic models (discussed above) strongly suggests that O-GlcNAc plays a role in mediating either feeding or the subsequent utilization of metabolites [183,185,187,197–199]. Neurons within the arcuate nucleus of the hypothalamus regulate both feeding and peripheral metabolism. Deletion of OGT from a subset of the neurons within this region, the Orexigenic neurons expressing agouti-related protein (AgRP)/neuropeptide Y (NPY), leads to a browning of white fat and protection from diet-induced obesity [188]. Inactivation of OGT in DRG neurons results in mice that weigh less than their littermate controls and exhibit enhanced glucose tolerance [185]. Similarly, deletion of OGT from skeletal muscle suppresses body weight, predominantly impacting fat mass. Notably, these mice also display enhanced energy expenditure and reduced basal insulin levels. Consistent with the work of Ruan and co-workers, deletion of OGT protected from high-fat induced insulin-resistance [189]. In contrast to these data, and speaking to the complex role of O-GlcNAc in regulating metabolism, deletion of OGT from the αCamKII-neurons of the paraventricular nucleus (PVN) leads to enhanced body weight, a result of increased food intake [187].

2004: Dynamic O-GlcNAcylation, a survival signal

Increased availability of glutamine, glucosamine, and glucose has been reported to promote cell survival in diverse models of physiological or environmental injury. Inhibition of the HBP attenuates the impact of these metabolites on cell death, suggesting that O-GlcNAc or other glycoconjugates affect survival decisions [226–231]. In 2004, we demonstrated that heat shock resulted in a rapid, dose-dependent, increase in O-GlcNAcylation on numerous proteins. Furthermore, enhancing O-GlcNAcylation promoted the synthesis of heat shock proteins, as well as survival [232]. Since this time, these data have been recapitulated in diverse models of physiological and environmental injury that include oxidative stress [232,233], osmotic stress [40,232], hypoxia reoxygenation injury [234], ER Stress [231,232,235], irradiation [232,236], DNA damage [237], unfolded proteins [232], glucose deprivation [34,119,123], trauma hemorrhage [238], and ischemia reperfusion injury [233,239–242]. These studies have been performed in transformed and primary cells, as well as physiological models, that include cardiomyocytes (heart), endothelial cells (vein, arteries), myoblasts (muscle), neurons (brain), hepatocytes (liver), tubule cells (kidney), macrophages and T-cells (immune), epithelial cells (lung, ovary, mesothelium, colon), and fibroblasts [243]. Together, these data suggest that stress-induced cycling of O-GlcNAc is a conserved response of mammalian cells and tissues to injury that promotes survival.

A number of studies have identified proteins on which O-GlcNAc is cycled in response to injury, either ad hoc or using high-throughput approaches [30,33,237]. The range of proteins identified suggests that O-GlcNAc will impact many cellular pathways, including transcription and epigenetics, translation, nuclear transport, signal transduction, metabolism, and cell shape. Functional studies support this assertion. O-GlcNAc has been demonstrated to reduce protein aggregation (discussed below) [244–246], suppress ER Stress [231,235], and augment autophagy [119]. The impact of O-GlcNAc on heat shock protein expression has been attributed to changes in GSK3β activity and the stability of the transcription factor Sp1 [178,232,247]. In models of oxidative stress and hypoxia reoxygenetion injury, O-GlcNAc has been demonstrated to maintain mitochondrial homeostasis, suppress calcium overload, preserve mitochondrial membrane potential, and inhibit the formation of the mitochondrial permeability transition pore [233,235,248,249]. O-GlcNAc promotes flux into the pentose phosphate pathway, which ultimately impacts glutathione levels, by activating glucose-6-phosphate dehydrogenase and inhibiting phosphofructokinase (PFK1) [250,251]. SirT1 is activated by O-GlcNAc in response to oxidative, genotoxic, and metabolic stress, resulting in deacetylation of key proteins such as p53 [252]. Lastly, O-GlcNAcylation of Keratin 8 and 18 potentiates AKT signaling (discussed below) [101]. In addition to providing insight into the molecular events regulated by O-GlcNAc, these studies have demonstrated that stress-induced O-GlcNAc cycling is more complicated than originally proposed. In models of oxidative stress, O-GlcNAcylation can be decreased on a sub-set of proteins [33]. These data lead to an amendment of the original model – that is “Stress-induced O-GlcNAc cycling is one component of the mammalian stress response; preventing deglycosylation of key proteins acts in concert with enhanced glycosylation of other proteins to prevent apoptosis/necrosis and to promote survival” [243].

2010: In vivo insight into the impact of O-GlcNAc on protein structure and function

How does one define the impact of O-GlcNAcylation on protein function at a molecular-level? Unlike protein phosphorylation, there is no amino acid that can mimic a glycosylated amino acid. Reducing the utility of mutational analysis, as discussed above, OGT has multisite interactions with its substrates and ~10% of glycosylation sites may be impacted by prior phosphorylation (discussed above). These challenges necessitate the use of a range of techniques to probe the impact of O-GlcNAcylation on protein function. Such approaches include use of proteins co-expressed with OGT in bacteria to generate glycoproteins for in vitro studies [244], chemical cross-linking to identify proteins within complexes with O-GlcNAcylated proteins [253], and the selective incorporation of O-GlcNAc onto residues for in vitro studies [100,254,255]. Collectively, these studies have highlighted roles for O-GlcNAc in mediating protein-protein interactions, protein structure and stability, post-translational modifications, and enzyme activity.

In 2010 the first mouse models engineered to express a hypo-O-GlcNAcylated protein emerged, providing insight into the physiological role of K8 and K18 glycosylation. Ku and co-workers generated mice expressing a mutant of K18 which were incapable of O-GlcNAcylation at Ser30, 31, and 49 (Gly⁻), as well as a phospho-mutant of K18 (Ser53Ala), a mutant which disrupts formation of keratin filaments (Arg90Cys), and wild-type K18. The K18 Gly⁻ mice phenocopy K18 null mice, exhibiting hepatocyte apoptosis and pancreatic necrosis when challenged with streptozotocin or FAS-ligand and PUGNAc. Basally, there were no gross differences in Keratin filament formation; however, expression of K18 Gly⁻ reduced phosphorylation of Akt1 at Thr308, as well as Protein Kinase C θ at Thr538 [101]. Glycosylation of K18 and K8 appeared necessary to form a complex with AKT, which promoted phosphorylation and activation of AKT.

The possibility of proteins which “read O-GlcNAc” in a manner analogous to Src 2 homology (SH2) domains, which dock to phosphorylated tyrosine residues, was raised in 1994. Felin and co-workers isolated two carbohydrate binding proteins (CBP), CBP22 and CBP70, using GlcNAc-affinity chromatography [256]. Subsequently, three CBPs of 70, 65, and 55kDa were isolated using a similar approach from nuclear and cytosolic lysates. Lefebvre and colleagues demonstrated that these CBPs could bind O-GlcNAcylated proteins using a blot overlay assay. Critically, binding could be blocked by pre-labeling O-GlcNAcylated proteins with Gal-T. Proteomic analysis identified CBP70 as HSC70 [257]. To address the impact of O-GlcNAc-dependent interactions in vivo, Yu and co-workers introduced a GlcNAc analog, GlcNDAz, that will crosslink to nearby proteins after photoirradiation [253]. Using this approach, the authors sought to identify proteins adjacent to O-GlcNAc residues in the nuclear pore complex, which had previously been demonstrated to be inhibited by WGA binding [258]. Enrichment of GlcNDAz-crosslinked-nuclear pore proteins with mAB414, followed by mass spectrometry determined that the O-GlcNAcylated residues of NUP358 were in close proximity to Transportin 1. GlcNDAz was subsequently used to verify the potential of 14-3-3 γ to interact with O-GlcNAcylated proteins in vivo. 14-3-3 γ was isolated from an affinity experiment using O-GlcNAcylated peptides as bait, which also identified 14-3-3 (α, β, δ, ζ, η, τ), α-enolase, and the ErbB3-binding protein as O-GlcNAc readers. Structural analysis revealed that O-GlcNAcylated peptides bind to 14-3-3 γ proteins in the same amphipathic groove as phosphorylated peptides. Mutagenesis demonstrated that Asp178 and Val181 are critical for binding to O-GlcNAcylated peptides [259].

Two studies have addressed the impact of O-GlcNAc on histone-protein interactions using neoglycoconjugates. Both studies took advantage of technology that enables the conversion of a cysteine residue to dehydroalanine, which can in turn be reacted with a GlcNAc-thiol to generate site-specific neoglycoconjugates [254,255]. This approach works well for histones whose sequences typically do not contain cysteine. Neoglycoconjugates of Histone 2A, glycosylated at residue Thr 101, were used to demonstrate that O-GlcNAcylation destabilized nucleosome formation. Underpinning this phenotype, the O-GlcNAcylation site lies at the interface where the Histone 2A and Histone 2B dimer interacts with Histone 3/Histone 4 tetramer [254]. Raj and colleagues used neoglycoconjugates of Histone 2B, glycosylated at Ser112, as bait to identify proteins that preferentially associated with glycosylated nucleosomes (Histone 3/Histone 4 tetramer associated with 2 Histone 2A/Histone 2B dimers). These studies suggest that O-GlcNAcylation of Histone 2B recruits the FACT complex to nucleosomes (facilitates chromatin transcription) [255]. Histone Ser112 O-GlcNAcylation has also been implicated in anchoring a histone 2B ubiquitin ligase (Bre1A/B), and results in monoubiquitination at Lys130 [260].

Initial sequence analysis of O-GlcNAc-modified proteins suggested that glycosylation sites occurred in PEST (Pro, Glu, Ser, Thr rich) sequences, that when phosphorylated targeted proteins for degradation [78,261]. Glycosylation was proposed to interfere with the functioning of these sequences, thus promoting the stability or half-life of proteins. Indeed, elevated cellular O-GlcNAcylation is associated with increased expression or stability of Sp1 [262], estrogen receptor β (ERβ) [263], plakoglobin [264], p53 [265], FoxoM1 [266], Snail [267], carbohydrate response element binding protein [268], BMAL/Clock (add Li) [269], nuclear pore proteins [270], nucleotide-binding oligomerization domain-containing protein 2 (Nod2) [245], c-Myc [192,271], Yes associated protein (YAP) [272], Fibrillarin (FBL), nucleolar protein (NOP) 56, and NOP58 [273]. The latter was highlighted in a proteomics screen which not only assessed the dynamics of O-GlcNAcylation, but identified proteins on which the stability or half-life was regulated by O-GlcNAc [273]. Recent data suggests that this phenotype may manifest at translation, where O-GlcNAc has been demonstrated to protect nascent polypeptides chains from premature proteasomal degradation [274]. Although other possibilities exist; O-GlcNAc has been implicated in inhibiting the proteasome [275–277], as well as regulating enzymes involved in protein ubiquitination [260,265,272,278].

One example of OGT and an Ubiquitin regulating enzyme collaborating to regulate protein function involves Ubiquitin specific protein 7 (USP7), a deubiquitinating enzyme. USP7 was isolated as an interacting protein which forms a stable complex MML5, a histone 3 lysine 4 (H3K4) methyltransferase, and OGT [279]. Disruption of this complex, by suppressing OGT expression, leads to a reduction in MML5 expression that can be suppressed with the proteasome inhibitor MG132. Thus, this study demonstrates that OGT regulates MML5 expression and presumably methylation of histones [279]. Indeed, Sakabe and co-workers demonstrated that overexpression of OGT could enhance methylation of Histone 3 at Arginine 17 [280]. In addition to MML5, protein arginine methyltransferase 4 (also known as Carm 1) is modified and regulated by O-GlcNAc. Here O-GlcNAcylation in the C-terminal domain alters substrate specificity through an unknown mechanism [124,280,281].

O-GlcNAcylation has been implicated in regulating the activity of enzymes, activating enzymes such as SirT1 [252], CamK II [282], and G6PD [250]. In contrast, O-GlcNAcylation of CamK IV [97], and phosphofructokinase 1 (PFK1)[251] is associated with reduced activity. O-GlcNAc has been proposed to inhibit activating phosphorylation (CamK IV)[97], sterically hinder the binding of allosteric activators (PFK1) [251], inhibit the binding of an autoinhibitory domain (CamK II) [282], and alter the Km of enzymes (SirT1/G6PD) [250,252]. Notably, glycosylation of G6PD at Ser84 activates this enzyme, which reflects the decreased Km of glycosylated G6PD for NADP⁺ as well as increasing the dimerization of the glycosylated enzyme [250]. Thus these data support themes discussed above – that O-GlcNAc can both promote and inhibit protein interactions and suggest that O-GlcNAc may impact protein structure.

Much of the initial work investigating the potential for O-GlcNAc to impact protein structure has been performed using peptides and their glycosylated analogs. Simanek and co-workers demonstrated that O-GlcNAc could induce a β-turn into a peptide (SYSPglycoTSPSYS) that when unmodified assumes a random coil [283]. Similarly, O-GlcNAc also induces a β-turn into intrinsically disordered peptides [284] derived from murine ER β [263]. Circular dichroism and NMR demonstrate that peptides from the proline-rich domain of Tau tend to form a polyproline helix which is stabilized by O-GlcNAcylation, in contrast to phosphorylation [285]. The impact of O-GlcNAc and phosphorylation on peptides that form α-helices are similar, both stabilizing α-helices when close to the N-terminus and destabilizing α-helices when introduced onto interior residues [286]. The impact of O-GlcNAc on protein structure has been studied on Tau; O-GlcNAcylated protein was obtained by co-expressing Tau with OGT in E. coli. Tau lacking the N-terminal domain (244–441) has a reduced propensity to aggregate when modified by O-GlcNAc [244]. O-GlcNAc is similarly implicated in stabilizing Nod2, α-Synuclein, and TAB1 [244–246]. The latter data is thought underpin the protective role that O-GlcNAc plays in neurodegenerative disease [287].

2011: Additional catalytic activities of OGT

OGT glycosylates and interacts with HCF1, a chromatin associated protein important for cell cycle control [133,288]. HCF1 is comprised of an N-terminal region that is important for transactivation and a C-terminal region that consists of an acidic region and a series of fibronectin-like repeats. Linking the N- and C-terminal regions are six-tandem repeats, also known as HCF1_PRO repeats, cleavage of which is required for full function of HCF1. Suppression of O-GlcNAcylation results in the accumulation of full length HCF1 suggesting that proteolytic maturation of HCF1 is dependent on OGT [134,289]. In vitro studies using recombinant OGT and a HCF1 cleavage reporter demonstrated that OGT was likely the HCF1 protease. This conclusion was reinforced by the requirement of UDP-GlcNAc for proteolysis, as well as OGT catalytic activity [134,289,290]. HCF1_PRO repeats contain two regions essential for proteolysis: a threonine rich region essential for OGT binding and a cleavage site which contains a conserved Cys-Glu-Thr sequence. Cleavage occurs between the cysteine and glutamate residues and proceeds by the transient generation of a glutamyl-sugar ester – explaining the requirement of UDP-GlcNAc [290,291]. These data suggest that changes in the expression of OGT or UDP-GlcNAc concentrations may impact cellular function in a manner dependent on HCF1. Indeed, OGT-dependent changes in HCF1 proteolysis have been implicated in pulmonary arterial hypertension [292] and HSV immediate early gene expression [134]. Notably, cleavage of HCF1 by OGT is restricted to vertebrates. HCF1 is not cleaved in C. elegans, and in D. melanogaster, Taspase1 cleaves HCF1.

2017: Congenital disorders of O-GlcNAcylation

Several studies have determined that mutations in OGT lead to a congenital disorder of glycosylation known as X-linked intellectual disability [293–296]. Mutations in the OGT TPR domain (Δ155–177; Leu254Phe, Arg259Thr, Arg284Pro, Ala319The, Glu339Gly; isoform 3) do not impact dimerization of OGT, but do alter thermal stability [294,295]. While these mutations have a modest impact on OGT activity, Crisper/Cas9 genome editing of embryonic stem cells did not reveal changes in the steady state levels of OGT, OGA, or O-GlcNAc [295]. Highlighting the importance of the TPR-domain, expression of each OGT mutant led to upregulation and downregulation of different transcripts. Notably, a core group of transcripts were altered by all mutations, whereas as subset of transcripts were impacted by one or more mutants [293,295]. The molecular mechanisms underlying these data have not been defined, however, there are numerous possibilities. Notably, O-GlcNAc has been demonstrated to modify numerous transcription factors, RNA polymerase II, epigenetic regulators, and the TET proteins which catalyze the hydroxylation of DNA 5-methylcytosine to 5-hydroxymethylcytosine [297].

OGA is encoded on a region of chromosome (10q) linked to the development of type II diabetes [298,299]. While single nucleotide polymorphisms have been identified within the oga allele, the impact of these polymorphisms on OGA activity and/or expression has not been determined. However, two alternative splice variants of OGA have been cloned from Sprague-Dawley and Goto Kakizaki rats. The latter is a non-obese rodent model that spontaneously develops type II diabetes early in life. Both variants are reported to be inactive [151], and consistent with this observation elevated levels of O-GlcNAc are found in the corneas of Goto Kakizaki rats [300]. A number of other mutants of OGT and OGA have been identified in the OGT and OGA coding sequence of cancer cells. Again, the impact of these mutations on the activity, expression, localization, and substrate targeting of either enzyme have not been assessed [301].

The Future

The modification of proteins by O-GlcNAc is implicated in regulating a host of cellular and physiological processes (reviewed in this issue and elsewhere), including: cell cycle, mitochondrial structure and function, transcription and epigenetics, nutrient sensing and metabolic dysfunction, metabolism, circadian rhythm, autophagy, immune signaling, and differentiation. Given these data, it is not surprising that dysregulation of O-GlcNAcylation is implicated in the etiology of a host of diseases, that include amyotrophic lateral sclerosis, Alzheimer’s disease, Parkinson’s disease, heart failure, hypertension, Type II diabetes, many forms of cancer, and congenital disorders of glycosylation [12,287,297,302–306]. The challenge is delineating if changes in O-GlcNAcylation are involved in disease initiation or as a result of disease progression. Improvements in site-mapping, new tools for modulating O-GlcNAc, and techniques that enable a site-specific understanding of the molecular events regulated by O-GlcNAc are central to resolving this challenge and developing a detailed understanding of the roles that this enigmatic carbohydrate modification plays in cellular physiology.

Abbreviations

AMPK

AMP-dependent protein kinase

5N5A

5 Asparagine 5 alanine mutant

CamK

Calcium/calmodulin dependent kinase

CBP

Carbohydrate binding protein

Chk1

Checkpoint kinase

CK2

Casein Kinase 2

Con A

Concanavalin A

CREB

Cyclic AMP-responsive element-binding protein

CRTC2

cyclic adenosine monophosphate response element-binding protein 2

EGF

Epidermal growth factor

eO-GlcNAc

extracellular O-linked N-acetylglucosamine

ER

Endoplasmic Reticulum

ERβ

estogren receptor β

Foxo

Forkhead box

G6PD

Glucose-6-phosphate dehydrogenase

Gal-T

β−1–4Galactosyltransferase

GSK3 β

glycogen synthase kinase 3β

HBP

Hexosamine biosynthetic pathway

HCF1

Host cell factor 1

Hex

Hexosaminidase

hOGA

Human O-GlcNAcase

HSC70

heat shock protein cognate 70

IR

insulin receptor

IRS

insulin receptor substrate

K8 and K18

Keratin 8 and Keratin 18

NOP

Nucleolar protein

Nup

nuclear pore protein

O-GlcNAc

O-linked β-N-acetylglucosamine

OGA

O-GlcNAcase

OGT

O-GlcNAc transferase

PcG

Polycomb-group

PH

polyhomeotic

PFK1

Phosphofructokinase 1

pNP

pnitrophenol

PRC1

polycomb repressive complexes

PUGNAc

O-(2-Acetamido-2-deoxy-D-glucopyranosylidenamino) N-phenylcarbamate

SirT1

Sirtuin 1

TAB1

Tak-1 binding protein 1

TET

Ten-eleven translocon

TPR

Tetratricopeptide repeat

Trak

trafficking kinesin-binding protein

USP7

Ubiquitin specific protein 7

WGA

Wheat germ agglutinin

YAP

Yes-associated protein