Review

The Hexosamine Signaling Pathway: Deciphering the "O-GlcNAc Code"

See allHide authors and affiliations

Science's STKE  29 Nov 2005:
Vol. 2005, Issue 312, pp. re13
DOI: 10.1126/stke.3122005re13

Abstract

A dynamic cycle of addition and removal of O-linked N-acetylglucosamine (O-GlcNAc) at serine and threonine residues is emerging as a key regulator of nuclear and cytoplasmic protein activity. Like phosphorylation, protein O-GlcNAcylation dramatically alters the posttranslational fate and function of target proteins. Indeed, O-GlcNAcylation may compete with phosphorylation for certain Ser/Thr target sites. Like kinases and phosphatases, the enzymes of O-GlcNAc metabolism are highly compartmentalized and regulated. Yet, O-GlcNAc addition is subject to an additional and unique level of metabolic control. O-GlcNAc transfer is the terminal step in a "hexosamine signaling pathway" (HSP). In the HSP, levels of uridine 5′-diphosphate (UDP)-GlcNAc respond to nutrient excess to activate O-GlcNAcylation. Removal of O-GlcNAc may also be under similar metabolic regulation. Differentially targeted isoforms of the enzymes of O-GlcNAc metabolism allow the participation of O-GlcNAc in diverse intracellular functions. O-GlcNAc addition and removal are key to histone remodeling, transcription, proliferation, apoptosis, and proteasomal degradation. This nutrient-responsive signaling pathway also modulates important cellular pathways, including the insulin signaling cascade in animals and the gibberellin signaling pathway in plants. Alterations in O-GlcNAc metabolism are associated with various human diseases including diabetes mellitus and neurodegeneration. This review will focus on current approaches to deciphering the "O-GlcNAc code" in order to elucidate how O-GlcNAc participates in its diverse functions. This ongoing effort requires analysis of the enzymes of O-GlcNAc metabolism, their many targets, and how the O-GlcNAc modification may be regulated.

O-GlcNAc–Dependent Signaling: The First Two Decades

The general characteristics of O-GlcNAcylation have been reviewed (16). O-GlcNAc was discovered in the early 1980s in studies in which a soluble β-1-4 galactosyltransferase was used as a probe of lymphocyte plasma membranes (7). Originally, O-GlcNAc was thought to be a cell-surface modification similar to other endomembrane carbohydrate modifications such as O-linked mucins, but it soon became clear that it was intracellular (811). Nucleoporins were identified as prominent intracellular targets modified by O-GlcNAc (8, 1014). The lectin wheat germ agglutinin (WGA) binds to these proteins and inhibits nuclear transport through the nuclear pore complex (15, 16). Although this inhibition by WGA led to speculation about the role of O-GlcNAc in nuclear transport, it now appears unlikely that O-GlcNAc is required for this process (1719). Chromatin is also modified by O-GlcNAc (20), as are many transcription factors (2128). The enzymes of O-GlcNAc metabolism have been identified, cloned, and characterized (2937). Interest also began to center around the role of O-GlcNAc in the nervous system (3842), and more recently in cellular signaling (1, 5, 34, 4346). Genetic model systems in Drosophila and Caenorhabditis elegans are now in place to evaluate O-GlcNAc metabolism in the context of the whole organism. To facilitate the identification of new O-GlcNAc glycoproteins, a new generation of immunologic (11, 13, 14, 47) and chemical tools have been created (4852). Thus, the field now has the proper tools to begin to decipher the O-GlcNAc–modified proteome and to explore the function of this modification in eukaryotic cell biology.

A Survey of O-GlcNAc–Modified Proteins

Methods for O-GlcNAc detection

Several analytical procedures exist for the analysis of O-GlcNAc. The initial tool utilized galactosyltransferase in a radiochemical assay. Galactosyltransferase transfers galactose from the donor UDP-Gal (uridine diphospho-N-acetylgalactosamine) to terminal N-acetylglucosamine residues. Ser/Thr-linked O-GlcNAc is a substrate for this reaction, and this was the method used to identify the linkage using radiochemically labeled UDP-Gal (7, 10). The main limitation of this technique lies in the accessibility of the GlcNAc residues to the galactosyltransferase. The lack of selectivity of the reaction to O-GlcNAc may also pose a problem; any terminal GlcNAc is a potential target.

A number of plant lectins bind to terminal GlcNAc residues, and these have been used to detect the linkage. The mostly widely used of these, WGA or succinylated WGA, detects O-GlcNAc on target proteins (10). However, the sensitivity of the method is limited so that only proteins with multiple O-GlcNAc residues are detected readily. Like the other methods used to detect O-GlcNAc, detection with lectins is not selective, and to date, no O-GlcNAc–specific lectin has been identified.

Antibodies have been developed that react with O-GlcNAc in the context of protein structure. These antibodies recognize glycosylated nuclear pore proteins (13, 14, 47, 53, 54) or the C-terminal domain (CTD) of RNA polymerase II (55). Other antibodies have been shown to exhibit some specificity for O-GlcNAc, including one directed against streptococcal antigens (53, 56). All of the antibodies are somewhat restrictive in their target specificity and may require more than one site of modification.

Chemical approaches have also been developed to analyze O-GlcNAc residues. Alkaline β-elimination is useful for generating radiochemically labeled O-GlcNAc moieties (7, 10) and can be used to introduce a radiochemical label if borotritide is used instead of borohydride in the reduction stage of the reaction. A variation on this general method, useful for mass spectrometry, relies on mild β-elimination followed by Michael addition with dithiothreitol (BEMAD) (52). We recently developed a purely chemical means of detection that involves the incorporation of an azido derivative of GlcNAc (GlcNAz) into target proteins. The enzymes of O-GlcNAc metabolism (OGT and O-GlcNAcase) tolerate analogs of their natural substrates in which the N-acyl side chain has been modified to a bio-orthogonal azide moiety (GlcNAz). These O-azidoacetylglucosamine–modified proteins can be covalently derivatized with various biochemical probes at the site of protein glycosylation by using the Staudinger ligation. This strategy should have general application for both the identification of O-GlcNAc–modified proteins and mapping protein target sites that bear O-GlcNAc (51). Cells can be metabolically labeled or the reaction can be performed in vitro with recombinant OGT (51). (Fig. 1) Because UDP-GlcNAc is incorporated into several classes of glycoconjugates, specificity must be demonstrated with properly controlled experiments.

Fig. 1.

Incorporation of GlcNAz into cells allows detection of O-GlcNAc–modified proteins. Biosynthesis of O-GlcNAc–modified proteins. GlcNAz (azido derivative of GlcNAc) can be taken up and used in a manner similar to other hexosamines. Uptake: Exogenously added Ac4GlcNAz diffuses into the cell and is deacetylated through the action of intracellular esterases, and then enters into the salvage pathway. Incorporation: the salvage pathway bypasses the de novo pathway through the action of GlcNAc kinase (GNK) and also generates GlcNAz-6-phosphate. This intermediate is converted to GlcNAz-1-phosphate through the action of phospho-N-acetylglucosamine mutase (AGM1). N-Acetylglucosamine-1-phosphate is converted to the end product of the hexosamine biosynthetic pathway, UDP-GlcNAz, through the action of UDP-GlcNAc pyrophosphorylase (AGX1). OGT transfers the saccharide moiety of UDP-GlcNAz to protein substrates within the cell. Detection: O-azidoacetylglucosamine-modified proteins are covalently derivatized with various biochemical probes at the site of protein glycosylation (circle). Hexosaminidase C (O-GlcNAcase) acts to cleave the glycosidic linkage of posttranslationally modified proteins to liberate the protein and GlcNAz.

A final chemical method involves a combination of galactosyltransferase and chemical modification. In this approach, an unnatural ketone group in UDP-galactose is accommodated by a genetically manipulated galactosyltransferase. Detection is accomplished using avidin-biotin (48). Coupled with modern analytical techniques, these evolving chemical methods should allow O-GlcNAc sites on target proteins to be mapped with increasing sophistication.

General features of the O-GlcNAc proteome

Analysis of the mammalian proteins known to be modified by O-GlcNAc—the O-GlcNAcproteome—although not yet complete, has revealed a diverse set of proteins engaged in numerous cellular functions (Fig. 2). O-GlcNAc–modified proteins are usually phosphoproteins and are often components of macromolecular complexes such as transcription complexes or nuclear pores. Proteins engaged in functions ranging from carbohydrate metabolism, signaling, transcription and translation, and the stress response are well represented in the O-GlcNAc proteome (Table 1). In this sense, O-GlcNAc is much like phosphorylation. Transcription/translation forms most O-GlcNAc substrates. This fact may reflect an important role for O-GlcNAc in these processes or may reflect the tremendous amount of current study in these areas. It is clear from Fig. 2 and Table 1 that OGT modifies a variety of substrates. The challenge lies in determining what role this modification plays in these diverse cellular functions.

Fig. 2.

O-GlcNAc proteome. The known O-GlcNAc substrates are divided into groups according to substrate function. Most of the substrates identified to date are involved in gene regulation. This may reflect a bias toward analysis of transcription factors and proteins involved in cellular regulation.

Hexosamine signaling pathway: An overview

The hexosamine signaling pathway may serve as a nutrient sensor. The concentrations of UDP-GlcNAc are sensitive to fluctuating glucose concentrations. These changes are transmitted throughout the cell to target proteins by the enzymes of O-GlcNAc cycling (OGT and O-GlcNAcase). The hexosamine signaling pathway is currently envisioned to consist of two parts: synthesis of UDP-GlcNAc and transfer and removal of O-GlcNAc. The first of these pathways, UDP-GlcNAc biosynthesis, involves several key enzymes with GFAT (ℒ-glutamine:𝒟-fructose-6-phosphate amidotransferase) presumed to be the rate-limiting enzyme in this pathway. This enzyme is also feedback inhibited by UDP-GlcNAc, the end product of hexosamine biosynthesis. The precursors for UDP-GlcNAc synthesis are nutrient derived: the amino acid glutamine; acetyl–coenzyme A (CoA) derived from free fatty acid oxidation; glucose; and uridine. Glutamine is a nonessential amino acid and the most abundant amino acid in muscle tissue. Acetyl-CoA is produced by fatty acid oxidation in mitochondria. Glucose enters the cell and is metabolized to glucose-6-phosphate, fructose-6-phosphate, glucosamine-6-phosphate, and finally to UDP-GlcNAc. The levels of UDP-GlcNAc are responsive to nutrient levels, suggesting that hexosamine biosynthesis could serve as a cellular sensor of nutrient levels. Fluctuating UDP-GlcNAc levels and differentially targeted OGT isoforms could then lead to discrete modifications of various cellular targets, potentially causing a change in the target’s function. In many ways, this pathway could be likened to adenosine 3′,5′-monophosphate (cyclic AMP)–dependent kinase in which target phosphorylation is regulated by a metabolite derived from a distinct metabolic pathway. In an analogous manner, OGT responds to concentrations of UDP-GlcNAc and glycosylates its substrates. UDP-GlcNAc levels may also serve to regulate OGT directly as OGT auotglycosylates itself (30, 31, 34). In addition, there may be an autoregulatory loop in which OGT and O-GlcNAcase modify one another.

The enzymes of hexosamine synthesis

Several enzymes play key roles in hexosamine synthesis (1, 46). Upon entering the cell, glucose is converted to glucose-6-phosphate. Glucose-6-phosphate then participates in glycogen synthesis, or after conversion to fructose-6-phosphate, enters the glycolytic pathway. Between 2 and 5% of the fructose-6-phosphate formed is directed to the hexosamine biosynthetic pathway, where it is converted to glucosamine-6-phosphate by the rate-limiting enzyme GFAT. The next enzyme in the pathway, the acetyltransferase Emeg32, uses acetyl-CoA to convert the glucosamine-6-phosphate to N-acetylglucosamine-6-phosphate (57, 58), which is rapidly converted to uridine 5′-diphospho-N-acetylglucosamine (UDP-GlcNAc). GFAT is thought to be a cytosolic enzyme, whereas Emeg32 is a membrane-associated enzyme (57). GFAT is regulated by feedback inhibition by UDP-GlcNAc and phosphorylation in mammalian cells; this inhibition by UDP-GlcNAc is competitive with respect to fructose-6-phosphate. GFAT is thought to be the rate-limiting enzyme in hexosamine biosynthesis, but Emeg32 is critical for maintaining proper intracellular concentrations of UDP-GlcNAc (58).

The enzymes of O-GlcNAc metabolism

Two highly evolutionarily conserved enzymes mediate the addition and removal of O-GlcNAc: O-linked GlcNAc transferase (OGT) and O-GlcNAcase, respectively. Coding sequences homologous to the OGT gene have been identified from Archaea to humans (2931). Budding yeast apparently lack the O-GlcNAc modification and have little feedback control over UDP-GlcNAc synthesis. The absence of OGT in budding yeast may reflect the yeast’s dependence on UDP-GlcNAc as a precursor for chitin synthesis for the rigid yeast cell wall. In other organisms, the organization of OGT is quite conserved. The molecule consists of an N-terminal segment with targeting information, a tetratricopeptide repeat (TPR) domain, a linker region, and a C-terminal catalytic domain (Fig. 3). The number of TPRs can vary widely, from 1 or 2 to more than 16. The catalytic domain is also conserved. A brief explanation of these domains provides a useful introduction to how this enzyme functions in the hexosamine signaling pathway.

Fig. 3.

The hexosamine signaling pathway. The hexosamine signaling pathway is divided into two parts: UDP-GlcNAc synthesis and the addition and removal of O-GlcNAc. The synthesis of UDP-GlcNAc is responsive to flux through several nutrient pathways. Each pathway is color-coded to denote its contribution in UDP-GlcNAc synthesis. Addition and removal of O-GlcNAc by OGT and OGA, along with examples of modified substrates, is also shown.

The N-terminus of OGT appears to be important for subcellular targeting in certain instances. One isoform of mammalian OGT is targeted to the mitochondrion by a short noncanonic targeting sequence, MTS (Fig. 4). Other isoforms are present in the nucleus and cytoplasm and lack this N terminus. (See also "Subcellular targeting of O-GlcNAc enzymes" below.)

Fig. 4.

Domain structure of O-linked GlcNAc transferase (OGT). OGT contains 9 to 12 TPR repeats, a linker domain, followed by the well-conserved, C-terminal catalytic domain. OGT is encoded by a single gene on the X chromosome. The 23 exons are shown along with the alternatively spliced isoforms below. Promoter 1 (P1) serves at the promoter for ncOGT. This is the largest isoform (116 kD) and is found in the nucleus and cytoplasm. Promoter 2 (P2) and the use of an alternative start site in exon 5 produce mOGT. This isoform targets mitochondria. The shortest isoform, sOGT, is generated by an alternative start site in exon 10. Both ncOGT and mOGT contain distinct sequences at their N termini, indicated by red and green, respectively. The N terminus of mOGT contains a mitochondrial targeting sequence (MTS). The nine TPRs in mOGT are identical to the last nine TPRs in ncOGT. Similarly, the three TPRs of sOGT are identical to the last three TPRs in mOGT and ncOGT. All isoforms contain identical catalytic domains.

TPRs are 34–amino acid repeats that facilitate protein-protein interactions. TPRs are found in a wide range of proteins. The OGT TPR motif consists of 3 to 16 repeats in the N-terminal half of the protein. Sequence alignment of TPR domains reveals a consensus sequence defined by a pattern of small and large amino acids. TPR motifs are widespread, ranging from bacteria to humans, and are involved in a variety of biological processes, including cell cycle regulation, transcriptional control, mitochondrial and peroxisomal protein transport, protein folding, and response to heat shock and cellular stress. Several x-ray structures of TPR proteins are now in the in the RCSB (Research Collaboratory for Structural Bioinformatics) protein data bank www.rcsb.org/pdb/, including one of the three TPRs in protein phosphatase 5. This structure revealed that the TPR adopts a helix-turn-helix arrangement, with adjacent TPR motifs packing in a parallel fashion, producing a spiral of repeating antiparallel α helices with a packing angle between the helices of ~24° within a single TPR. The structure has a right-handed superhelical shape with a radius of ~50 Å. Two protein surfaces are generated; the inner concave surface is formed mainly by residues on helices A, and the other surface is formed by residues from both helices A and B. Studies with model peptides suggest that peptide binding to TPRs is facilitated by both short- and long-range interactions (5964).

The structure of the TPR domain of nuclear/cytoplasmic OGT (ncOGT) revealed that the TPR domain of human ncOGT is homodimeric and contains 11.5 TPR repeats (18). The repeats form an elongated superhelix. This structure is shown in Fig. 5. The concave surface of the superhelix is lined by absolutely conserved asparagines, in a manner reminiscent of the peptide-binding site of importin-α. On the basis of this structural similarity, we proposed that OGT might use an analogous molecular mechanism to recognize its numerous cellular targets.

Fig. 5.

X-ray structure of the TPR domain of ncOGT and the basis of substrate recognition. The TPR domain of OGT forms a homodimer of right-handed superhelices. The dimer is maintained through mainly hydrophobic interactions. The concave surface of the superhelix is lined with asparagine residues. Like the peptide-binding surface of importin-α, this contiguous stretch of asparagines may allow OGT to interact with substrates. A movie showing a 360° rotation of the OGT homodimer about the homodimer interface is also available as supporting online material (http://stke.sciencemag.org/cgi/content/full/sigtrans;2005/312/re13/DC1).

The central domain of OGT is predicted to be a flexible linker between the TPR domain and the catalytic domain. In C. elegans, this linker contains a putative nuclear localization sequence, but this sequence is not conserved in mammals and is unlikely to function in nuclear targeting (30). Indeed, this region is one of the most divergent segments of the highly conserved OGT molecule, both in terms of length and amino acid sequence.

The catalytic domain of OGT is predicted to contain two Rossman folds on the basis of analysis of related enzymes that use UDP-GlcNAc and related nucleotides (2, 65). The presence of these folds was suggested to place OGT in the glycogen synthetase family of enzymes (65). Site-specific mutants indicate that catalytically important residues are clustered around the nucleotide binding pocket of the enzyme. Deletion of regions of the catalytic domain destroys the enzymatic activity of OGT (34).

The β-O–linked N-acetylglucosaminidase termed the O-GlcNAcase was originally identified as a hyaluronidase associated with meningioma and thus called MGEA5 (meningioma expressed antigen 5) (66). The MGEA5 gene resides on chromosome 10 at position 10 q24.1-q24.3 (66) and encodes at least two alternatively spliced transcripts, which are widely expressed in mammalian tissues (Fig. 6). The longer of these was originally designated O-GlcNAcase (36, 37). However, the shorter of the two isoforms also has O-GlcNAcase activity, and each of the isoforms is differentially targeted.

Fig. 6.

Domain structure of the O-GlcNAcase isoforms. A single gene on chromosome 10 encodes O-GlcNAcase. The two main isoforms are shown. O-GlcNAcase contains a hyaluronidase (yellow) domain at the N-terminal end. The acetyltransferase domain (blue) is localized at the C-terminal end and is not found in the shorter V-O-GlcNAcase. This shorter isoform also contains a distinct stretch of amino acids at the C-terminal end (red). The domain structures are shown above and were drawn in Cn3D and are based on a bee venom hyaluronidase [Protein Data Bank (PDB): 1FCV] and the acetyltransferase domain from GCN5 (PDB: 1Z4R). Arrow denotes location of bound hyaluronic acid tetramer (substrate).

The N-terminal segment of the O-GlcNAcase was originally identified as a hyaluronidase domain because of its sequence similarity to the C. elegans "hyaluronidase" (66). The secondary structure of this domain shows that this segment of the molecule adopts the TIM barrel fold common to many enzyme families. This structure is formed by alternating α helices and β sheets, such that the β sheets form an inner barrel that is protected from solvent by the surrounding α helices. However, it is not clear whether this segment is solely responsible for O-GlcNAcase activity (67). The precise roles of individual residues in the active site await structural determination.

The C-terminal segment of the major isoform of O-GlcNAcase shares many features of the GCN5 family of acetyltransferases (1, 66, 67). This domain of O-GlcNAcase may have intrinsic histone acetyltransferase (HAT) activity (68). However, the HAT domain is not a complete acetyltransferase domain and might require accessory proteins for its activity. Demonstration of this activity is important because O-GlcNAc has been implicated in transcriptional repression. Indeed, OGT interacts with the mSin3A transcriptional repression complex (69). An appealing possibility is that OGT and O-GlcNAcase each participate in the process of chromatin remodeling associated with transcriptional repression via histone deacetylase (HDAC) and HAT activities, respectively. The shorter variant of O-GlcNAcase lacks the acetyltransferase domain but contains a short amino acid extension.

Cellular Aspects of O-GlcNAc Metabolism

Subcellular location of O-GlcNAc–modified targets

The subcellular distribution of O-GlcNAc has been examined in a number of ways, including enzymatic labeling of subcellular fractions and antibody labeling in fixed cells. Amounts of O-GlcNAc differed widely in various subcellular fractions. These are listed in order of decreasing relative abundance: nuclear membrane > nuclei > rough microsomes > golgi > cytosol > mitochondria (9, 70).

Subcellular targeting of O-GlcNAc enzymes

The genomic organization of OGT is well conserved in mammals. Human OGT is located at Xq13.1, spans ~43 kb, and contains 23 exons (71). Multiple OGT isoforms are generated through alternative splicing and multiple start codons within this single gene (35, 71). Three main isoforms are generated, each with distinct N termini and identical C-terminal catalytic domains (Fig. 4). These N termini contribute to the differential localization of these isoforms (70). The isoforms are named for their intracellular localization and size: ncOGT (nuclear/cytoplasmic); mOGT (mitochondrial); and sOGT (short isoform). Exons 1 to 4 and 6 to 23 produce ncOGT. ncOGT is the largest isoform (116 kD) with 12 TPR domains followed by the common catalytic domain. It is found in both the nucleus and cytoplasm. ncOGT has been directly implicated in proteasomal inhibition (72), transcriptional repression (69), and stress tolerance (7375). mOGT is generated by an alternative start site in exon 5. This alternative start yields a transcript with a much longer 5′UTR than that of ncOGT. The N terminus of mOGT contains the mitochondrial targeting information, followed by a membrane-spanning α helix. These domains localize mOGT to the inner mitochondrial membrane (70). An internal start in exon 10 generates a transcript encoding the shortest isoform (sOGT). sOGT contains two TPRs followed by the catalytic domain. Like ncOGT, it is localized throughout the cell. A screen for antiapoptotic factors identified sOGT as one of three classes of proteins that could protect cells from growth factor withdrawal–mediated apoptosis (76). The multiple OGT isoforms, each with unique targeting sequences, enable OGT to interact with a variety of substrates in a regulated manner. This differential localization of multiple OGT isoforms would allow for the transmission of nutrient status to various cellular compartments (Fig. 7).

Fig. 7.

Regulation of hexosamine signaling by enzyme localization. Nutrients entering cells are metabolized by enzymes residing in the cytosol, mitochondrion, and on the surface of lipid droplets; information must be communicated back to the cell surface to regulate nutrient transport and to the nucleus for regulation of the genes encoding these metabolic enzymes. The differential localization of OGT could act as an integrator of these metabolic signals. By modifying targets in the nucleus, mitochondria, and cell surface, OGT may modulate transcription and transport, nutrient transport, and mitochondrial metabolism and apoptosis.

O-GlcNAcase (MGEA-5) is a predominantly cytoplasmic and nuclear enzyme (66, 68, 77, 78). Alternative splicing generates a shorter isoform of MGEA-5 that also has O-GlcNAcase activity (79). This shorter isoform of MGEA-5 was reported to be nuclear (78).

O-GlcNAc in Intracellular Signaling

The glycosylation catalyzed by OGT differs in many fundamental ways from endomemembrane glycosylation reactions. It takes place in the cytosol and nucleus and undergoes dynamic addition and removal. In these compartments, it is present in a milieu dominated by the kinase-dependent signaling cascades common to many aspects of cellular physiology. In this section, we will review the evidence suggesting that O-GlcNAc may itself participate in cellular signaling that is mediated by kinases and phosphatases.

Signaling in plants: SPINDLY and SECRET AGENT

Early evidence that OGT-like proteins may participate in cellular signaling came from studies in Arabidopsis. Unlike mammals, which have a single gene encoding mammalian OGT, plants appear to have two separate genes for OGT (80). Information from Genebank suggests that these proteins are found throughout angiosperms. These proteins appear to have roughly the same organization as does mammalian OGT, with the N-terminal half containing varying numbers of TPRs, followed by the catalytic domain at the C terminus. SPINDLY (SPY) was the first plant OGT identified and contains 10 TPR domains (81). The second OGT homolog, SECRET AGENT (SEC), has 12 TPR domains and is slightly more similar in sequence to mammalian OGT in the C-terminal catalytic domain than is SPY. Although a single deletion in either spy or sec is tolerated, the spy and sec double mutant is lethal. Specifically, gametogenesis and embryogenesis are affected (80). These data suggest a conserved, essential role for O-GlcNAc in plants.

Plant OGT, like its mammalian counterpart, appears to influence several signaling pathways. Before the proteins were cloned and expressed, several spy alleles had been identified genetically on the basis of their ability to disrupt gibberellin signal transduction (82). Gibberellins (GAs) are plant hormones that control a number of processes in plant development including seed germination, leaf expansion, stem elongation, flower initiation, and flower and fruit development (83). Spy alleles yield tall, slender plants consistent with constitutive activation of the GA pathway, suggesting that SPY serves normally as a repressor of GA signaling. Not all of the spy phenotypic characteristics can be attributed to GA disregulation, suggesting that SPY also may influence other signaling pathways. SPY has been implicated in the cytokinin pathway (84), a pathway responsible for the reprogramming of terminally differentiated tissue to support new growth. Briefly, cytokinin binds to its receptor histidine protein kinase at the cell surface and initiates a phospho-relay that ends in the nucleus, triggering a subset of transcriptional activators and repressors. Genetic experiments using two spy alleles suggest that SPY is a positive regulator of this pathway. The role of SPY in both the GA and cytokinin pathways might allow for integration and cross-talk between these two pathways.

O-GlcNAc modification alters substrate activity

Although many proteins serve as substrates for OGT and bear the O-GlcNAc modification, it has been difficult to show the exact role of that modification in protein activity. Recently, a few examples of O-GlcNAc modification altering a substrate’s activity have emerged (Fig. 8). Hyperglycemia inhibits eNOS activity, and the Akt phosphorylation site of eNOS at residue 1177 is glycosylated and thus underphosphorylated under hyperglycemic conditions (85). Inhibiting the rate-limiting enzyme in the hexosamine biosynthetic pathway restored phosphorylation at this site and eNOS activity. The O-GlcNAc modification and impairment of eNOS that occurs in the diabetic state may contribute to atherosclerosis (85) and erectile dysfunction associated with diabetes (86). The ubiquitin-proteasome pathway is another example in which O-GlcNAc inhibits substrate activity. O-GlcNAc modification of the 26S proteasome prevents proteolysis by inhibiting the adenosine triphosphatase activity of the proteasome (72). Of the substrates identified to date, transcription factors make up most of the proteins modified by OGT (Fig. 2), raising the possibility that O-GlcNAc could function in control of transcription or translation (or both). The transcription factor pancreatic duodenal homeobox1 (PDX-1) is a key regulator in the developing pancreas and also activates the transcription of many genes involved in glucose homeostasis. It is modified by O-GlcNAc at two major sites (87). O-GlcNAc modification is associated with increased DNA binding activity of PDX-1 and increased insulin secretion. In this final example, activity appears to be enhanced by O-GlcNAc modification. Much like phosphorylation, O-GlcNAc modification can serve as both a positive and a negative regulator of substrate activity.

Fig. 8.

O-GlcNAc modification alters substrate activity. Three examples of substrate modification with O-GlcNAc changing the activity of the substrate are shown. Both eNOS (86) and the 26S proteasome (72) are inhibited by O-GlcNAc addition. In contrast, the DNA binding activity of Pdx-1 is enhanced with O-GlcNAc addition (87).

O-GlcNAc and phosphorylation

One aspect of hexosamine signaling by O-GlcNAc that remains poorly understood is the precise mechanism by which the addition of O-GlcNAc might function in signal transduction. One possibility is that O-GlcNAc, which occurs on Ser or Thr residues, functions simply to block phosphorylation sites, producing a mutual exclusivity between the modifications (88). This hypothesis has been extended to explain the coexistence of differentially modified isoforms of the target proteins. In the case of protein phosphorylation, some modified sites gain the ability to interact with proteins bearing domains such as WW, WD, and 14-3-3 domains. Evidence suggests that O-GlcNAc may also recruit proteins having lectinlike characteristics (89, 90). These proteins are members of the HSP70 heat shock family of proteins (89). Another protein of this type that apparently binds to O-GlcNAc is the cellular prion protein binding partner CBP70 (91). These findings suggest that O-GlcNAc may be an equal partner with phosphorylation when the modifications coexist. Active recruitment by phosphorylated and glycosylated Ser or Thr sites within target proteins of other effector proteins is an attractive model to explain the many biological functions of O-GlcNAc and the complex interrelationship between glycosylation and phosphorylation.

O-GlcNAc and the insulin signaling pathway

Strong evidence that O-GlcNAcylation plays a role in mammalian signaling pathways derives from work on mammalian insulin resistance. Tissue culture lines have proven useful for examining insulin signaling and are often studied in primary adipocytes or in the 3T3 L1 cell line (92, 9396). When differentiated, 3T3 L1 cells have an adipose-like phenotype and are highly responsive to insulin, which modulates surface expression of glucose transporters (97102). Hexosamine synthesis has a role in the phenomenon called insulin resistance (93, 103105). The desensitization of the glucose transport system to insulin requires three components: glucose, insulin, and glutamine. Increased concentrations of UDP-GlcNAc have a role in the process of desensitization to insulin. The monosaccharide glucosamine, by bypassing feedback inhibition of the enzyme GFAT, leads to elevated concentrations of UDP-GlcNAc and insulin resistance, whereas inhibitors of GFAT prevent insulin resistance. Several molecular properties of OGT indicate that the transferase may be the terminal step in the hexosamine signaling pathway: OGT is enriched in the pancreas; OGT can interact with many substrates via its TPR domain; and OGT activity is modulated by UDP-GlcNAc concentrations (30). Additionally, several groups have found increased concentrations of UDP-GlcNAc and increasing amounts of O-GlcNAc correlated with insulin resistance (43, 44, 106111). Therefore, the role of O-GlcNAc in insulin resistance and diabetes mellitus is being explored in humans and in model organisms.

O-GlcNAc and response to stress: OGT as a catalytic chaperone

OGT may have a role in cellular stress responses (7375). Exposure of cells to heat shock leads to an increase in the abundance of O-GlcNAc modifications (73, 75). Inhibiting O-GlcNAc addition leads to the increased susceptibility of these cells to heat-induced stress (73, 75). Many heat shock proteins are modified by O-GlcNAc, including hsp70 and lens crystallins, yet it is still unclear how O-GlcNAc performs a protective role during the stress response (73, 112, 113).

An interesting relationship exists between stress-induced pathways and insulin signaling (114121). This has been largely worked out in the nematode C. elegans, in which the insulin-like signaling pathway has a role in life-span extension and response to stress (19, 114, 117121). One of the downstream transcriptional targets of the insulin-like signaling pathway in the worm is the gene encoding the C. elegans O-GlcNAcase (termed oga-1), along with many heat shock proteins (see below) (122). The nematode model system may be useful in examining the relationship between hexosamine signaling and cellular responses to stress.

O-GlcNAc and Human Diseases

A growing body of evidence suggests that deregulation of O-GlcNAc metabolism may play a role in human disease. Because of its potential cellular functions as a catalytic chaperone, nutrient sensor, and proteasome inhibitor, O-GlcNAc has been suggested to have a role in mammalian insulin resistance associated with diabetes mellitus and with neurodegeneration.

The O-GlcNAcase gene: A NIDDM susceptibility locus?

In humans, the gene encoding the O-GlcNAcase (MGEA5) has been investigated as a NIDDM (noninsulin-dependent diabetes mellitus, type 2 diabetes) susceptibility locus. In Pima Indians, who have the world’s highest prevalence of NIDDM, two variants were detected, but there was no association with parameters of insulin resistance or diabetes in 1300 Pimas. The mutations in MGEA5 are unlikely to contribute to NIDDM in this population (123). The MGEA5 gene was also investigated in the Mexican American population (124). Twenty-four single nucleotide polymorphisms (SNPs) were identified by sequencing in 44 subjects. Association tests indicated significant linkage of a previously unknown SNP with diabetes (P = 0.0128, relative risk = 2.77) and age at diabetes onset (P = 0.0017). The associated SNP was located in intron 10 of MGEA5, producing an alternate stop codon. The variance attributed to this SNP accounted for ~25% of the logarithm of odds. Thus, this variant within the MGEA5 gene may increase risk of diabetes in Mexican Americans. Unlike the Pima Indian group described above, the Mexican American population consists of an admixed population of Native and European Americans, whereas the Pimas are primarily of Native American descent. The SNPs exhibiting association with diabetes traits in the Mexican American individuals were found in intronic regions and were not examined in the Pima study. Therefore, the MGEA5 gene is a likely NIDDM susceptibility locus present at least in the Mexican American population.

The Goto-Kakizaki (GK) rat is a useful model for mammalian NIDDM (125). The GK rat exhibits fasting hyperglycemia, altered insulin secretion, and insulin resistance. Genetic analysis of the GK rat has revealed a number of loci responsible for the phenotype, with a major locus on chromosome 1 termed Niddm1. This region encodes an insulin degradation enzyme (IDE) and is adjacent to MGEA5 (126). The O-GlcNAcase gene exhibits synteny between the rat, mouse, and human IDE genes. The proximity of the genes suggests that O-GlcNAc may have a role in the phenotype. Consistent with these findings, increased concentrations of O-GlcNAc and of OGT have been detected in corneas of GK rats (127). The GK rat model may be useful for studying the role of O-GlcNAc in NIDDM.

Transgenic mice provide a useful model in which to study the physiological role of hexosamine signaling. (Fig. 9). Mice have been generated that overexpress glutamine:fructose-6-phosphate amidotransferase (GFAT) (128130) in muscle and fat under control of a Glut4 promoter (128, 131). Because GFAT is the rate-limiting enzyme in the hexosamine biosynthetic pathway, overexpression of GFAT will lead to production of more UDP-GlcNAc, and consequently more O-linked glycosylation by OGT. Using these mice, one can test the consequences of enhanced O-glcNAcylation on glucose homeostasis. A 2.4-fold increase in GFAT activity was observed in muscle of the transgenic mice; this led to weight-dependent hyperinsulinemia in random-fed mice (128). Overexpression of GFAT in the liver of transgenic mice produced enhanced glycogen storage, hyperlipidemia, obesity, and impaired glucose tolerance (130). When expressed in pancreatic β cells, GFAT activity in islets of transgenic mice was increased compared with that of littermate controls. The increased GFAT activity led to 1.4- and 2.1-fold increased pancreatic insulin content in 2- and 10-month-old transgenic mice, respectively. Fasting insulin concentrations in the blood were 1.6-fold higher than in littermate controls. Thus, increased hexosamine flux caused by GFAT overexpression in the β cell results in hyperinsulinemia, insulin resistance, and (in males) mild type 2 diabetes (129).

Fig. 9.

Transgenic mice designed to test aspects of the hexosamine-signaling pathway. A variety of transgenic mice have been generated. The promoter, gene, target tissue, and phenotype generated by these mice are shown.

A more direct test of the hypothesis that hexosamine signaling influences NIDDM came with the overexpression of OGT in muscle and fat of transgenic mice (46). Transgenic overexpression of OGT to modify proteins with O-GlcNAc produces the type 2 diabetic phenotype. Even modest overexpression of an isoform of O-GlcNAc transferase in muscle and fat resulted in insulin resistance and hyperleptinemia. These data support the proposal that OGT participates in a hexosamine-dependent signaling pathway that is linked to insulin resistance and leptin production (46). In vitro studies using NIH 3T3 L1 cells and the inhibitor O-(2-acetamido-2-deoxy-𝒟-glucopyranosylidene)amino N-phenyl carbamate (PUGNAc) in adipocytes also demonstrated a role for O-GlcNAc in insulin resistance (44). These animal models and in vitro systems support a model in which O-GlcNAc may modify the insulin signaling pathway.

The hexosamine signaling pathway functions in a genetically amenable organism, C. elegans (19). Animals with a putative null allele of OGT in C. elegans are both viable and fertile. Although nuclear pore proteins of the homozygous deletion strain are devoid of O-GlcNAc, nuclear transport of transcription factors appears to be normal. However, the OGT mutant exhibits metabolic changes manifested in an approximately threefold increase in trehalose concentrations and glycogen stores with a concomitant approximately threefold decrease in triglycerides levels. In nematodes, a highly conserved insulin-like signaling cascade regulates macronutrient storage, longevity, and dauer formation (Fig. 10). Consistent with a role of OGT in insulin-like signaling in C. elegans, the OGT knockout suppresses dauer larvae formation induced by a temperature-sensitive allele of the insulin-like receptor gene daf-2. Thus, OGT induces a kind of insulin resistance in the nematode. These findings demonstrate that OGT modulates macronutrient storage and dauer formation in C. elegans and provide a genetic model for examining the role of O-GlcNAc in cellular signaling and insulin resistance.

Fig. 10.

A C. elegans model of hexosamine signaling. Schematic of the highly conserved, insulin-like signaling pathway is shown above. The pathway is initiated by binding of insulin-like ligands to DAF2 (insulin receptor). This binding activates a series of kinases and culminates with the suppression of DAF-16 (foxo transcription factor). The pathways regulated by this signaling cascade are listed on the right. Of note, the gene for O-GlcNAcase appears to be regulated by insulin signaling (122). The life cycle of C. elegans is shown below. When nutrients are limiting (insulin signaling is depressed), the nematode enters into a dauer stage. Then when conditions are favorable, the nematode reenters the cycle and matures into an adult worm. Signaling through the insulin pathway can then be measured by the propensity for the worms to enter dauer. Using this assay for insulin signaling, we found that knockout of ogt-1 suppressed dauer formation (19).

Table 1.

Proteins modified by O-GlcNAc.

O-GlcNAc and neurodegenerative disease

Several lines of evidence suggest that O-GlcNAc signaling may have a role in neurodegenerative disease. The transcripts encoding OGT and O-GlcNAcase are found in high abundance in the brain (30, 66), and the O-GlcNAcase was originally identified as an autoantigen associated with meningioma (66). OGT transcripts have been localized to the cerebral cortex of the rat brain and are particularly enriched in Purkinje cells (132). Transcripts encoding the longest OGT isoform are enriched in the motor cortex, and in the CA3 region of the hippocampus and the dentate gyrus (133). In light of this intriguing localization, it is useful to summarize the evidence suggesting a role for O-GlcNAc metabolism in nervous system function and in human diseases involving neurodegeneration.

In Alzheimer’s disease, affected neurons accumulate protein aggregates consisting of β-amyloid and the microtubule-associated protein tau. Both the amyloid precursor protein and tau are modified by O-GlcNAc (40, 134, 135). The degree of O-GlcNAc modification of these proteins appears to be altered in Alzheimer’s disease (40, 136138). Other neuronal proteins such as AP3 (adaptor protein 3) are also extensively modified by O-GlcNAc (41, 42, 139). Evidence also suggests that these clathrin-associated proteins might be differentially glycosylated in Alzheimer’s disease (42, 139). Another hypothesis suggests that intraneuronal aggregation could be due to interference with proteasomal function (77). O-GlcNAc modification of proteasomes inhibits proteasome function (72), and aggregates in Alzheimer’s disease contain ubiquitin (77).

Genetic studies also suggest a link between Alzheimer’s disease and O-GlcNAc metabolism. A candidate locus for late-onset Alzheimer’s disease maps to a region of chromosome 10 that is near the gene encoding O-GlcNAcase and insulin degrading enzyme (IDE) (10q24.1-q24.3) (140, 141). As mentioned above, the O-GlcNAcase gene has been implicated in NIDDM (124). Alteration of the O-GlcNAcase gene could lead to changes in the abundance of O-GlcNAcase or its activity that would render susceptible patients particularly prone to changes in flux through the hexosamine signaling pathway.

The locus responsible for dystonia-Parkinsonism syndrome (DYT3) was mapped to human chromosome X at Xq13.1 (142, 143). This is very near the position of OGT at Xq13.1. In light of this proximity and its biological properties, OGT is being actively pursued as a candidate gene for X-linked dystonia-Parkinsonism.

In addition to the evidence accumulating of a role in human disease, a number of model systems have been established that suggest a role for O-GlcNAc in neurodegeneration (136138, 144146). Many of these studies involve rodent models of the disease. Food deprivation of mice for 1 to 3 days progressively enhances tau phosphorylation in the hippocampus and to a lesser extent in the cerebral cortex. This phosphorylation was reversible by refeeding for 1 day. One possible interpretation of these findings is that nutrient-sensing pathways such as the hexosamine signaling pathway may function in the excessive phosphorylation of tau observed in Alzheimer’s disease. Glucose uptake in the brain of Alzheimer’s patients is reduced and may lead to other metabolic changes. OGT knockout mice exhibited changes in tau phosphorylation (146). Other model systems that may facilitate examination of the hexosamine signaling pathway are Drosophila (147152) and C. elegans (147, 151154). Human tau and its mutated variants create an uncoordinated phenotype in C. elegans (153, 154). Null alleles of OGT (19) and O-GlcNAcase in C. elegans may allow examination of the role of O-GlcNAc addition in the formation of tau and β-amyloid aggregates in a genetically amenable system.

Summary and Conclusions

Deciphering the O-GlcNAc code: Importance of model systems

Current approaches attempting to understand the role of O-GlcNAc signaling in biological systems are somewhat similar to the efforts undertaken since the late 1930s to understand phosphorylation [reviewed in (155)]. Only in the 1970s and 1980s was the true significance of protein phosphorylation revealed (155). Likewise, this field took some time to mature from the initial identification of enzymes involved in O-GlcNAc metabolism to an understanding of how these enzymes are regulated. As with phosphorylation, it is likely that O-GlcNAc performs a large number of functions that can only be understood when biological phenomena are analyzed in isolation. One important tool for this kind of analysis is the use of model systems in which genetics is more fully developed than has been possible in mammals. Although knockouts of the O-GlcNAc transferase result in stem cell and embryonic lethality (33, 35, 146), it may be possible to study the function of the enzyme using conditional knockout approaches or in model systems such as C. elegans where OGT appears somewhat dispensable (19). Ongoing work on O-GlcNAc metabolism in genetically amenable organisms such as Drosophila and C. elegans will likely provide important clues to the cellular functions of the enzymes involved in hexosamine signaling.

Implications for systems biology

Bioinformatics approaches suggest that both OGT and O-GlcNAcase are coexpressed with other cellular components involved in cellular signaling such as kinases and phosphatases (156). Computational approaches are useful in analyzing highly conserved signaling pathways such as the hexosamine-signaling pathway. At present, it is impossible to precisely predict sites of attachment of O-GlcNAc by computational approaches, although one site (www.cbs.dtu.dk/services/YinOYang/) attempts to do this. The difficulties in such predictions stem in part from the limited number of target proteins identified to date. Other uncertainties are whether some O-GlcNAc attachment sites may represent "primer" sites allowing recruitment of OGT or glycosylation of adjacent sites. Even more uncertain is the effect of O-GlcNAc addition on phosphorylation of the target protein. Available resources such as www.cbs.dtu.dk/services/NetPhos/ have begun to attempt such an analysis.

Concluding remarks

We have assembled an overview of an emerging signaling pathway terminating in the addition of O-GlcNAc. This pathway links biosynthesis of hexosamines from nutrient sources to posttranslational modification of target proteins such as transcriptional components and nuclear pores to kinases. Hexosamine signaling is likely to intersect many other signaling pathways involving protein phosphorylation. Hexosamine signaling may also have hitherto unappreciated roles in diseases such as neurodegeneration, diabetes mellitus, and cancer. Our current vision of this pathway is quite incomplete and evolving; pieces of the puzzle are missing, and we don’t yet have the broad outlines of the finished picture. What is clear is that this highly evolutionarily conserved pathway acts through a limited number of enzymes, which have recently come under close scrutiny. The hope is that by deciphering the role of these enzymes in cell physiology, new signaling paradigms may emerge.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
View Abstract

Navigate This Article