|Home | About | Journals | Submit | Contact Us | Français|
The enzymes of O-GlcNAc cycling couple the nutrient-dependent synthesis of UDP-GlcNAc to O-GlcNAc modification of Ser/Thr residues of key nuclear and cytoplasmic targets. This series of reactions culminating in O-GlcNAcylation of targets has been termed the Hexosamine Signaling Pathway (HSP). The evolutionarily ancient enzymes of O-GlcNAc cycling have co-evolved with other signaling effecter molecules; they are recruited to their targets by many of the same mechanisms used to organize canonic kinase-dependent signaling pathways. This co-recruitment of the enzymes of O-GlcNAc cycling drives a binary switch impacting pathways of anabolism and growth (nutrient uptake) and catabolic pathways (nutrient sparing and salvage). The Hexosamine Signaling Pathway (HSP) has thus emerged as a versatile cellular regulator modulating numerous cellular signaling cascades influencing growth, metabolism, cellular stress, circadian rhythm, and host-pathogen interactions. In mammals, the nutrient-sensing HSP has been harnessed to regulate such cell-specific functions as neutrophil migration, and activation of B-cells and T-cells. This review summarizes the diverse approaches being used to examine O-GlcNAc cycling. It will emphasize the impact O-GlcNAcylation has upon signaling pathways that may be become deregulated in diseases of the immune system, diabetes mellitus, cancer, cardiovascular disease, and neurodegenerative diseases.
Organisms have evolved a robust network of signaling pathways allowing them to distinguish sources of food from pathogens (immunity), regulate the uptake and utilization of food (metabolism), and adapt to nutrient availability (gene expression). One of the most evolutionarily ancient of these pathways is the nutrient-sensing addition of O-GlcNAc to target proteins [1-5]. Many reviews have appeared in the past few years detailing many aspects of O-GlcNAc metabolism [1, 2, 5-7]. In addition, several excellent reviews have focused on methods of detection of O-GlcNAc [8, 9]. This review will focus on the genetic, molecular genetic and chemical genomic dissection of the HSP in an attempt to expose the molecular logic linking the many intracellular signaling pathways influenced by O-GlcNAc cycling.
Before embarking on a detailed analysis of mechanism, the likely involvement of O-GlcNAc cycling in human disease must be emphasized. Based on large-scale genetic associations, the gene encoding the enzyme that removes the O-GlcNAc modification has recently been shown to be a susceptibility locus for type-2 diabetes in the Mexican-American population . Additionally, O-GlcNAc transferase (OGT) is encoded on chromosome X and has emerged as a candidate gene for X-linked Dystonia Parkinsonism [11, 12]. Numerous other studies have linked O-GlcNAc cycling to the proteotoxicity associated with neurodegeneration [5, 13, 14, 14-19], type-2 diabetes mellitus [20-24], and cardiovascular disease [25-30]. It is likely that alterations in O-GlcNAc cycling may also contribute to the pathology of other human diseases, including cancer [1, 2, 5], and diseases of immune function [31, 32], thus making the HSP a potentially important drug target.
In animals, two highly conserved enzymes are responsible for the cycling of O-GlcNAc on Ser/Thr residues of target proteins. The addition of O-GlcNAc is mediated by the glycosyltransferase, OGT. This enzyme was first identified in human, rat and C. elegans [22, 33]. In addition, the transferases from Arabidopsis and Giardia have also been identified [4, 34]. The enzyme mediating O-GlcNAc removal is the O-GlcNAcase, encoded in mammals by the MGEA5 locus [35, 36]. This enzyme has now been characterized in human, mouse, rat, and C. elegans. In all animals studied to date, single genes encode OGT and the O-GlcNAcase. Arabidopsis appear to be an exception and contains two genes encoding OGT. The following sections will detail the protein expression, localization, chemical and structural biology of these enzymes of O-GlcNAc cycling.
In mammals, OGT is encoded on chromosome X (Xq13) in close proximity to the Xist locus involved in X-inactivation [37-39]. From publicly available databases (such as BioGPS), it is now clear that the transcripts encoding mammalian OGT gene are very highly expressed in T-cells, B-cells, and macrophages. It is also present in moderately high levels in pancreatic ß cells and in the central nervous system. Low level, ubiquitous expression is found in other organs. The gene encodes a number of splice variants producing protein products with distinct amino termini targeting them either to the nucleus, cytoplasm or mitochondrion  (See Fig 1A, left panel). These variants range in molecular mass from about 70 kDa to about 110 kDa, and contain between 2 and 13 tetratricopeptide (TPR) repeats at their amino terminus. The enzyme variants all have a common catalytic domain (Fig 1A). Extensive mutagenesis analysis of the catalytic domain identified both activating and inactivating mutations [3, 15, 41, 42]. A number of mutations have also been molecularly identified in the Arabidopsis OGT homologs Spindly and Secret Agent by forward genetic approaches [43, 44]. However, the divergent coding sequences between the plant and animal OGT make comparison of catalytically important residues difficult. The other domains of OGT have come under similar scrutiny. When the structure of the TPR region of the nucleocytoplasmic form of OGT was solved, it revealed that the enzyme exists as a superhelical assembly that dimerizes using important contacts within the TPR repeats  (Fig 1B, right panel). Mutagenesis of this dimerization motif inhibited enzyme activity . These findings likely suggest that formation of oligomers of OGT may be required for full catalytic activity.
The structure of the catalytic domain of a bacterial homolog of OGT has recently been solved by two groups [46, 47], now allowing a model of human OGT to be produced by combining the human TPR structure  with the bacterial OGT homolog structures [46, 47] (Fig 1B). It should be noted that the bacterial structure lacks the “linker region” separating the catalytic domains in human OGT , and is catalytically inactive with respect to macromolecular substrates. Although this linker could have significance for the regulation of OGT, the overall fold is unlikely to be altered by its omission. The structure of the bacterial OGT has revealed a number of important clues as to how OGT may function to recognize substrate (Fig 1B). The catalytic domain adopts the glycosyltransferase-B fold-a metal independent catalytic center . The catalytic center is indicated between the two lobes of the catalytic domain in Fig 1B, as indicated in the two projections of the structure (right and left, Fig 1B.). The TPR residues form a binding groove interacting with the nucleotide sugar-binding site. The structure confirms previous speculations that the TPR residues were key components of the substrate-binding pocket [42, 48]. It extends these observations by pointing to the central importance of the arrangement of catalytic and TPR domains. The superhelical TPR repeats were shown to have an asparagine ladder motif similar to that present in the nuclear import receptor, importin . This asparagine ladder is likely to serve as a binding surface for the interaction of target peptides. The TPR repeats are also the site of interaction with many of the binding partners of OGT discussed in later sections (Fig 1B, left).
The other significant finding emerging from the crystal structure of OGT was the localization of the C-terminal PIP3 binding domain to a helix at the ‘top’ of the head of the bi-lobed catalytic domain (Fig 1B, right and left). Since it is present in a solvent exposed volume, it is now rather easy to imagine this domain interacting with membrane inositol-phosphate (PIP3) lipids, temporarily anchoring the OGT catalytic domain to the plane of the PIP3-rich membrane. Recruitment of OGT would enhance access to the key substrates similarly recruited to the cell surface by the activation of tyrosine kinase receptors (see below). A more detailed analysis of OGT structure-function can be found in the article by Davies in this volume.
When first identified as a neutral pH hexosaminidase , it became clear that O-GlcNAcase had been previously cloned as an autoantigen associated with meningioma designated MGEA5 [5, 36]. Like OGT, transcripts derived from the O-GlcNAcase gene are highly expressed in immune cells and in the central nervous system. It is also widely expressed at lower levels in other tissues, including endocrine tissues. The gene encodes at least two alternatively spliced (an probably many more) isoforms varying in molecular mass from about 72 kDa to over 100 kDa. The longer of these variants has, at its C-terminus, a histone acetyltransferase (HAT) domain [5, 35, 36](Fig. 2 A, long OGA). The shorter variant has a much smaller C-terminal domain (Fig. 2 A, short OGA). Based on similarity to hyaluronidases, the catalytic domain was proposed to be restricted to the N-terminal domain . This conclusion was initially questioned , but was subsequently confirmed directly by mutagenesis [50, 51]. The HAT domain has been suggested to be enzymatically active, but histone acetyltransferase activity has not yet been independently verified [50, 52]. Each of the isoforms contains a domain which has been shown to interact with OGT as indicated in Fig. 2A .
Recent progress in understanding the catalytic mechanism of the O-GlcNAcase active site was greatly facilitated by the solution of the crystal structure of putative O-GlcNAcase homologs from Clostridium perfrigens and the human gut endosymbiont, Bacteroides thetaiotaomicron [54, 55] (Fig 2B). The bacterial O-GlcNAcase homolog is organized as ‘TIM-barrels’ with a deep active site pocket consistent with binding of O-GlcNAc-modified peptides. Mutagenesis and kinetics studies on these homologs show that the bacterial enzymes behave similarly to their human counterparts, operating via a previously proposed ‘substrate-assisted’ catalytic mechanism involving the acetamide group of GlcNAc [51, 56]. A central domain of human O-GlcNAcase has been shown to interact with human OGT (indicated as OGT binding) . The solved bacterial structures lack this ‘OGT binding domain’ as well as the ‘HAT’ domain that are features of the human enzyme. It has been shown that the HAT domain of O-GlcNAcase interacts with histone tails (H3 and H4), and as mentioned previously, has been suggested to serve as a HAT under these conditions. However, the inability of some laboratories to demonstrate intrinsic HAT activity using recombinant O-GlcNAcase suggests a contaminating enzyme could account for the observed activity. The HAT-like domain is modeled in Fig. 2B using the related structure of Tetrahymena Gcn5 histone acetyltransferase (PDB: 5GCN) .
Although O-GlcNAc was originally identified on the surface of T-lymphocytes , it shortly became clear that most of the O-GlcNAc in cells resides in intracellular compartments [58-62]. In mammals, the intracellular targeting of the enzymes of O-GlcNAc cycling appears to be a critically important aspect of the diverse biological processes in which they are involved.
The nucleus was particularly enriched in the O-GlcNAc modification and the nuclear pores were the most obvious of the glycosylated substrates [58-62]. For example, high levels of O-GlcNAc found on the nucleoporin, Nup62, results in this protein standing out as the immunodominant band in cellular extracts probed with O-GlcNAc-specific antibodies. The antibodies raised against O-GlcNAc vary slightly in their binding avidities and their requirement for polypeptide backbone. GlcNAc will compete for binding of some of the antibodies with their targets (eg., CTD 110.6), while it does not compete for binding of others (eg., RL2). Thus, competition controls should be used, when possible, in studies employing antibodies to avoid false positives in target identification. A large number of cytoplasmic proteins are now known to be substrates. These include structural proteins such as the cytokeratins, myosin signaling kinases, transcription factors, and ribosomal proteins. More recent analysis of OGT splice variants has revealed a larger isoform of OGT, termed ncOGT, and the short isoform (sOGT) are both distributed between the nucleus and cytoplasm (See Fig 1A) [40, 63]. Although a sequence in the central portion of human OGT was suggested to be its nuclear targeting signal, this short basic sequence has not been demonstrated to be essential for nuclear targeting of OGT . The longer form of the O-GlcNAcase has a nuclear targeting domain in its C-terminal ‘HAT’ domain. Since O-GlcNAcase and OGT interact, it is formally possible that they enter the nucleus in a complex. OGT might also ‘piggyback’ into the nucleus bound to partners such as p38 MAPK. Alternatively, OGT may use an alternative nuclear transport pathway. OGT exhibits structural similarities to importin α , and like importin α, OGT could directly interact with an importin β for nuclear import.
Early studies showed that mitochondria contained few, if any, O-GlcNAc modified substrates [58-62]. However, one of the isoforms of OGT (mOGT) is targeted to the mitochondrion by a unique N-terminal sequence [40, 63] (See Fig 1A). On the basis of these findings, the mitochondrial form of OGT was proposed to regulate mitochondrial function, including apoptosis. Interestingly, the shortest isoform of OGT, which is nuclear and cytoplasmic (Fig. 1 A) has been shown to be an anti-apoptotic protein in a screen for anti-apoptotic genes using leukemia cells . Recent work suggests that O-GlcNAc metabolism in the mitochondria may be of great importance in many cell types including the heart and cardiovascular system [29, 30, 40, 65-69]. The effects of the O-GlcNAc modification on the physiology and pathology of the heart will be discussed in another of the articles in this volume. It is not yet clear if the O-GlcNAcase becomes localized to the mitochondrion since no obvious mitochondrial signal is present in either isoform of O-GlcNAcase. The shorter isoform of O-GlcNAcase (Fig. 2, B) was originally thought to be a nuclear isoform. It has more recently been found to be associated with the endoplasmic reticulum and lipid droplets , and could potentially enter the mitochondria by an unusual vesicular route.
The enzymes of O-GlcNAc cycling, unlike many carbohydrate-acting enzymes, are soluble proteins, and are particularly amenable to production in vitro using E. coli expression systems. Recombinant enzymes have been used for structural work, enzymology, and chemical biology approaches. The isoforms of OGT were expressed individually, and the two longest of these (termed ncOGT, and mOGT, Fig. 1A) were shown to be enzymatically active [15, 41, 45]. The shorter isoform (sOGT, Fig. 1A) did not glycosylate any known substrate but showed nucleotide hydrolase activity and was a dominant negative inhibitor of the other two enzymes in vitro. The other isoforms of OGT show differential abilities to glycosylate substrates. For example, O-GlcNAcase and Tau are substrates for ncOGT as expected. Surprisingly, however, the src-family Yes tyrosine kinase was a substrate only for the mOGT . The shorter isoform sOGT was recently used in a high throughput screen, based on fluorescence anisotropy, which identified some initial lead compounds that function as inhibitors of OGT . These inhibitors are moderately complex heterocyclic compounds with no obvious ‘pharmacophore’ emerging. Obviously, additional high potency inhibitors are essential for moving forward on the function of OGT in higher metazoans where genetics are not feasible. Further high throughput assays are underway in a number of laboratories.
Bacterial expression of the O-GlcNAcase has also led to studies that have been quite fruitful. The human, and C. elegans enzymes have each been expressed in active, recombinant form [5, 35, 50-53, 72-75]. Both isoforms of the mammalian O-GlcNAcase are enzymatically active (Fig. 2, A) [51, 76]. The O-GlcNAcase activity has been detected using a number of assays including p-Nitrophenyl-β-D-GlcNAc, and more recently, using a sensitive fluorescent derivative fluorescein di-(N-acetyl-beta-D-glucosaminide) (FDGlcNAc). This latter substrate was devised and synthesized in efforts to increase sensitivity of O-GlcNAcase detection. Extension of the acetyl-moiety of this probe, also allowed the production of an O-GlcNAcase-specific substrate . This general principle pioneered in a study with GlcNAc-thiazolines , and was subsequently used by a number of laboratories for inhibitor design, producing a number of selective agents [56, 72, 78-84]. These approaches will be further detailed in other articles in this issue.
Identification of the enzymes of O-GlcNAc cycling has led to the identification of protein binding partners. The physical interactions of OGT that are the best characterized and are summarized in Figure 1B. However since O-GlcNAcase can associate with, and is modified by OGT [15, 53, 85], both activities may, under certain circumstances, be recruited by the OGT binding partners. For OGT, two-hybrid screens have identified numerous proteins that interact with the TPR region [85, 86]. The proteins identified in these screens include OIP106 and GRIF-1. These proteins are members of a family of proteins with the so-called ‘HAP’ (Huntington Associated Protein) motif. Proteins containing this motif are associated with vesicle and organelle movement in neurons. The Drosophila Milton protein, involved in mitochondrial movement, is another member of this class of proteins. OGT was also shown to interact with mSin3A , HCF-1 , and the human dosage compensation complex  consistent with a role for OGT in the control of gene expression. OGT directly binds to several classes of enzymes. For example, OGT interacts with protein phosphatase PP1 , a regulatory subunit of PP1, MYPT1 (a myosin phosphatase), as well as the histone methyltransferases CARM1 and MLL5 . The interaction of OGT with this diverse and numerous list of factors reflects the growing awareness that OGT is involved in complex and robust regulatory paradigms, touching many aspects of cellular physiology.
Other probable interacting partners for OGT have arisen from two-hybrid analysis in invertebrates. In C. elegans, two-hybrid analysis indicates an interaction of the C-terminus of OGT-1 with the MAPK, PMK-1 . The interaction of the C-terminus of OGT with p38 MAPK has now been shown in mammals . Intriguingly, this interaction increases with p38 MAPK activation during glucose deprivation. The catalytic activity of OGT is not altered by glucose deprivation, but p38 MAPK serves to recruit OGT to specific targets, including neurofilament H. In Drosophila, OGT has been shown to interact with 26 other proteins by two-hybrid analysis . These interacting partners include park (parkin), eyg (eye-gone), Stam (signal transducing adaptor molecule, CSN4, (COP9 signalosome subunit 4), sns (sticks-and-stones), dys (dystotrophin-like protein 1), and the ubiquitin-ligase ara-2 (ariadne 2). As seen in other systems, these probable interactions suggest that OGT may be involved in a number of unexplored signal transduction pathways altering numerous biological processes in the fly. Verification of these interactions and continued development of reagents to examine the Drosophila HSP will enable these approaches.
The hexosamine biosynthetic pathway, as it is now understood, is shown in Figure 3; the enzymes listed are given for both mammals and the genes of C. elegans (in parenthesis). Historically, Marshall and his colleagues first described the impact of hexosamine biosynthesis on desensitization of the insulin-signaling pathway [95-97]. These studies revealed that desensitization of insulin-responsive glucose transporters in adipocytes required glucose, insulin and glutamine. The authors speculated that hexosamine biosynthesis was the cause, and confirmed this by using inhibitors of the rate-limiting enzyme in hexosamine synthesis, glutamine: fructose-6-phosphate amidotransferase (GFAT). Desensitization was also induced by glucosamine, a metabolite bypassing GFAT, and causing increased production of UDP-GlcNAc [95, 96]. Of course, the products of the hexosamine biosynthetic pathway were known to be used by many intracellular enzymes, and reside in many intracellular compartments (See Figure 3). The challenge facing the field then became deciding which of the many metabolic byproducts of UDP-GlcNAc were responsible for altering insulin sensitivity.
When OGT was first molecularly identified [22, 33], similarities emerged with known signaling molecules including yeast SSN6 and the Arabidopsis protein Spindly , involved in gibberelin signaling. Based on these similarities and the nutrient sensing potential for OGT, O-GlcNAc was proposed to play a key role in diabetes and insulin resistance . An ongoing discussion about the role of the hexosamine biosynthetic pathway in signaling still continues; two broad theories have been advanced. The first suggests that altered insulin signaling is the result of changes in O-GlcNAc cycling (shown in Fig. 3) [23, 24]; the second theory suggests that altered N-glycosylation can impact insulin signaling [98, 99]. These theories are not mutually exclusive, of course, and it is probable that both pathways contribute to insulin resistance in vitro. For the purposes of this review, the role of O-GlcNAc cycling in altering signaling will be emphasized, but any discussion of the findings must also take into account the complexities of cellular sugar nucleotide utilization.
Following the initial observation suggesting that O-GlcNAc may play a key role in insulin resistance [23, 24], much support for this idea has emerged [20, 100-109]. Many of the key substrates in the insulin-signaling pathway are O-GlcNAc modified, yet the mechanism of O-GlcNAc-mediated insulin resistance is still not precisely known.
The mouse has proven to be a useful model for understanding the physiological impact of the HSP in mammals. Among the first experiments to address the problem in the context of a whole animal were studies involving overpression of GFAT, the rate-limiting enzyme in hexosamine synthesis in several target tissues [110-113]. These mouse studies showed that over expression of GFAT lead to peripheral insulin resistance, and leptin secretion in a manner predicted by the early studies on preadipocytes. A direct link to O-GlcNAc metabolism was made in a study in which O-GlcNAc transferase (OGT) was over expressed in muscle and fat of transgenic mice . These transgenic animals exhibited elevated insulin and leptin levels, and were insulin resistant as evidenced by the hyperinsulinemic-euglycemic clamp procedure. Parallel studies in rat L1-3T3 preadipocytes demonstrated that inhibition of the O-GlcNAcase also induced insulin resistance . Taken together, these experiments strongly suggested a link between the hexosamine biosynthetic pathway, O-GlcNAc cycling, and mammalian insulin resistance. More recent evidence has questioned this link since specific inhibitors of O-GlcNAc cycling may not produce the insulin resistance phenotype . Thus, genetic loss and gain of function studies to address this key issue have become essential.
The enzymes of hexosamine synthesis have yet to be carefully examined in mouse loss of function models except for a knockout of EMeg32, the acetyltransferase involved in assembly of the nucleotide sugar UDP-GlcNAc  (See Figure 3). Homozygous null embryos lacking EMeg32 die at embryonic day 7.5 and are developmentally delayed. In vitro differentiated EMeg32 −/− ES cells show reduced proliferation and mouse embryonic fibroblasts (MEFs) derived from knockout embyos exhibit defects in proliferation and adhesiveness. These defects could be reversed by stable expression of EMeg32 or by nutritional restoration of intracellular UDP-GlcNAc levels. The knockout results in reduced O-GlcNAc modifications of cytosolic and nuclear proteins. Interestingly, the EMeg32 knockout MEFs exhibited activated PKB/AKT and are resistant to apoptosis. Thus, EMeg32-dependent UDP-GlcNAc levels influence cell cycle progression and susceptibility to apoptotic stimuli . The impact of the knockout was attributed to the decreased levels of O-GlcNAc addition.
Mouse knockouts of the enzymes of O-GlcNAc cycling have been more problematic. OGT is an X-linked gene, and since most embryonic stem cells are XY, early attempts to generate knockout mice resulted in stem cell lethality [38, 63]. Subsequent studies circumvented this problem by using Cre-recombinase technology, but knocking out OGT in early development is embryonic lethal. Tissue-specific knockouts of OGT have been generated in a variety of cell lineages including thymocytes, neurons and fibroblasts , resulting in T-cell apoptosis, neuronal tau hyperphosphorylation, and fibroblast growth arrest, respectively. However, the general cytotoxic effects observed in the targeted tissues make interpretation of these results difficult. Efforts are underway to generate hypomorphic alleles of OGT that might alter O-GlcNAc levels but not lead to embryonic lethality. A conditional knockout of O-GlcNAcase has not yet been reported.
The pathway of insulin signaling leading to transcriptional programming is now well understood, in large part, because of genetic analysis carried out in the nematode Caenorhabditis elegans and the fruitfly, Drosophila melanogaster [14, 14, 74, 74, 116-122]. Genetic analysis in these organisms allowed an ordering of the series of reactions following binding of insulin-like peptides to the insulin receptor. These pathways are summarized in Figure 4. Much of what was learned in the worm, and the fly, directly translates to the mammalian signaling pathways. For example, in C. elegans, the insulin-signaling pathway has been genetically linked to the dauer diapause, longevity, innate immunity and the cellular stress response (Fig. 4) [118-122]. The mammalian insulin signaling pathway is now known to be highly similar to those present in C. elegans and D. melanogaster [14, 74, 123, 124].
As an approach to understanding the role of O-GlcNAc cycling in mammalian insulin resistance, the HSP has recently been examined in the worm and fly. In Drosophila, the OGT locus was recently demonstrated to correspond to a previously identified gene termed ‘Super Sex Combs’ (Sxc) . This locus was identified in 1984 as a regulator of homeotic gene expression. Several Sxc alleles displayed anterior and posterior transformations . Loss of Ogt/Sxc function suggests a role in Polycomb-dependent transcriptional repression (See later section). The O-GlcNAcase of the fly has been disrupted using P-element insertion proximal to its promoter . This fly strain is viable and fertile. These genetic and molecular tools should prove invaluable for sorting out the role of hexosamine signaling and insulin signaling in Drosophila. The study of the Drosophila melanogaster HSP is currently ongoing in several laboratories.
Unlike the knockouts of OGT and O-GlcNAcase in mammals, null alleles of these enzymes in C. elegans are viable and fertile, facilitating genetic analysis. In addition, the C. elegans system is amenable to genetic analysis for signaling pathways such as insulin-like signaling (See Fig. 4), TGF-beta signaling, and MAPK signaling using a large collection of previously identified mutants and reverse genetics [121, 128, 129].
Knockout of the O-GlcNAc transferase (encoded by ogt-1) in C. elegans is viable and fertile providing a biological platform to study the impact of the modification on cell signaling and metabolism [14, 74]. Interestingly, the nuclear pore proteins, normally highly decorated with O-GlcNAc, completely lacked O-GlcNAc, yet no impact on nuclear transport of key transcription factors was observed in these null animals. As summarized in Figure 4, metabolic changes were detected in the ogt-1(ok430) knockout animals including 3-fold increases in glycogen and trehalose, and a dramatic (3 fold) reduction in lipid stores. Pathways regulated by insulin-like signaling in the worm, dauer (diapause) and longevity were also examined in these studies. The C. elegans dauer is a nutritionally responsive diapause induced upon crowding and starvation and triggered by a released pheromone termed daumone. The dauer pathway requires the concerted action of the C. elegans insulin-signaling pathway (Fig. 4), and TGF- ß-like pathways. The study employed a temperature sensitive allele of the insulin-like receptor (daf-2(e1370)) (Fig. 4) that allows controlled inactivation of insulin signaling, scored as the extent of dauer larvae formation. By generating a double mutant harboring daf-2(e1370) and ogt-1(ok430), the impact of O-GlcNAc modification on insulin-like signaling was examined. The ogt-1 knockout was shown to greatly diminish dauer larvae formation in the sensitized daf-2 mutant background suggesting that the wild-type function of the ogt-1 gene is inhibition of insulin signaling downstream of daf-2. Inactivation of the ts allele of daf-2 normally greatly extends the lifespan of C. elegans, but the ogt-1(ok430) knockout suppresses this lifespan extension (in preparation), once again arguing that the normal function of OGT is to blunt insulin signaling . The ogt-1 knockout strain was also more susceptible to UV and osmotic stress than the wildtype strains (Love and Hanover, in preparation). These findings are summarized in Fig. 4, bottom right. Thus, the C. elegans studies of OGT have shown that OGT plays a role in regulating the process of dauer formation, stress and longevity, processes intimately linked to intracellular signaling and nutrient status.
These observations were extended when a null allele of the O-GlcNAcase (oga-1(ok1207)) was characterized in C. elegans . This null allele was shown to be enzymatically inactive in vitro and has a similar metabolic phenotype as the ogt-1 knockout animals. Namely, the worms have elevated trehalose and glygogen levels and decreased neutral lipid stores. When insulin signaling was examined by the assays described above, the oga-1(ok1207) knockout showed an insulin-resistant phenotype, suggesting that the normal function of the O-GlcNAcase is to relieve the desensitization of insulin signaling created by O-GlcNAc transferase. Other phenotypes of the O-GlcNAc cycling mutants included changes in phosphorylation profiles, and strikingly increased levels of glycogen synthase kinase 3 (GSK3) .
Identification of O-GlcNAc cycling mutants in C.elegans has also led to the inclusion of ogt-1 and oga-1 in unbiased high-throughput screening efforts using RNA interference to create genetic interaction networks. One such study examined some of the most important signaling pathways in C. elegans including growth factor signaling, Ras-signaling, Wnt signaling, Notch signaling and DNA damage repair . This screen was restricted to C. elegans LGIII (Chromosome III) fortuitously the chromosomal location of ogt-1. The findings of that global interaction study are consistent with a role for ogt-1 in pathways distinct from the insulin-like signaling pathway. In particular, genetic interactions of ogt-1 with the β-catenin-encoding gene bar-1 and the FGF, encoded by let-756, were detected. The interaction module associated with bar-1 and ogt-1 is of particular interest since it was confirmed to play a role in the regulation of fat storage and metabolism. This finding is also consistent with earlier observations that ogt-1 mutants have dramatically lower neutral lipid stores than do wild-type animals (See also Fig. 4). In mammals, β-catenin has been shown to be O-GlcNAc modified, and the function of the protein is altered by the modification. [24, 131]. β-catenin is a key transcriptional modulator interacting with the TCF family of architectural transcription factors and triggering key developmental switches including those involved in triggering formation of T-cells and B-cells in adaptive immunity.
The genetic interaction of ogt-1 with let-756, encoding a FGF homolog, further suggests that O-GlcNAc cycling may participate in a similar module involved in the regulation of growth. The let-756 gene is involved in numerous genetic interactions in C. elegans and, like the interaction of ogt-1 and bar-1, the interaction of ogt-1 with let-756 is likely to be quite complex. Although the insulin-like pathway was examined in this high throughput screen (daf-2), no genetic interaction was observed with ogt-1 as measured by a slow growth phenotype. This demonstrates the limitations of specific assays and highlights the need for multiple approaches in dissecting and defining the many roles of O-GlcNAc cycling.
One way of visualizing the nutrient-sensing HSP is to view it as the calm ‘eye’ in the violent ‘storm’ of intracellular signaling. The HSP sensor ‘calmly’ detects nutrient availability communicating that information to the signaling ‘storm’ circling about it. The combined forces of the signaling cascades that respond to nutrient status in metazoans are not unlike a revolving storm, constantly moving, yet adapting to a changing environment. Some nutrient-response systems respond to specific signals, while others react to more general nutrient stimuli. For example, AMPK responds to the AMP/ATP ratio, while mTOR responds to both amino acid levels and insulin signaling. Perhaps the most general sensor of cellular nutritional status is the nutrient-integrating HSP and the enzymes of O-GlcNAc cycling. For this reason, the HSP may be placed at the center, or ‘eye’, of a swirling nutrient-sensing ‘storm’ (Figure 5). It is clear that significant crosstalk occurs between phosphorylation and O-GlcNAc addition. A recent study examined 711 phosphopeptides and found that approximately half of the sites of phosphorylation cycle; of these, elevated GlcNAcylation resulted in lower phosphorylation at 280 sites and caused increased phosphorylation at 148 sites . Thus the interplay between these signaling modifications is complex. A coherent model of how O-GlcNAc cycling may serve to modulate intracellular signaling in such a complex system must take into account a number of variables. These include the nutrient-sensing capabilities of hexosamine synthesis, the intracellular localization and interactions of the O-GlcNAc cycling enzymes, the molecular targets, other homeostatic mechanisms and signaling cascades, and the final biological output. In the robust signaling networks in which O-GlcNAc cycling is likely to act, this is a daunting task. However, as is clear from the evidence presented here, some generalizations emerge. These generalizations are summarized in Figure 5. The HSP appears to be a central regulator of the cellular response to feast and famine.
A highly simplified diagram of the complex interaction between the Insulin-AKT, MAPK, mTOR and AMPK pathways is shown in Figure 5. OGT is likely to interact with these canonic nutrient-sensing pathways at many points as indicated below. First, there is growing evidence that tyrosine kinase receptors may directly interact with OGT. OGT itself is tyrosine phosphorylated  and auto-glycosylated [22, 33, 42]. In addition, insulin treatment of 3T3-L1 adipocytes stimulates both the tyrosine phosphorylation and the catalytic activity of OGT. This study also demonstrated that a subset of OGT coimmunoprecipitates with the insulin receptor. OGT also interacts with other tyrosine kinases including the Src kinase family member, Yes . A recent bioinformatic analysis of known O-GlcNAc modification sites and known tyrosine phosphorylation sites revealed potential association between these two post-translational modifications in various proteins including the protein prohibin. This has led to speculation that O-GlcNAc modification and tyrosine phosphorylation may act as a novel and previously unidentified binary switch adding additional complexity to cell signaling pathways .
In addition to directly interacting with tyrosine kinases, OGT is recruited to sites of PIP3 formation during growth factor signaling, suggesting that the many signaling cascades triggered by PI3K are likely to be influenced by O-GlcNAc signaling . Growth factor receptors are key activators of PI3K. At the cell surface, growth factor receptor occupancy would trigger PI3K leading to PIP3 formation and OGT recruitment (Fig. 5). Factors co-recruited to PIP3, such as AKT and PDK, would actively compete with OGT for the modification of key Ser/Thr residues, thus competition with OGT could effectively modulate the action of these kinase cascades. The nutrient sensing capabilities of the OGT catalytic domain would allow real-time monitoring of nutrient status to be directly communicated to the canonic growth factor signaling cascade. Activated AKT modifies many transcription factors whose nuclear concentrations may be modulated by interaction with the 14-3-3 family of binding proteins. By blocking phosphorylation at AKT sites, OGT can oppose the cytoplasmic sequestration leading to transcriptional activation. Notable examples of this kind of regulation are the insulin-responsive Foxo family transcription factors  and the gluconeogenic regulator CRCT2 . In vitro studies have demonstrated the capacity of OGT to function over a wide range of UDP-GlcNAc concentrations making the enzyme ideally suited for a role as a nutrient sensor [42, 48].
Another critically important interaction occurs between OGT and p38 MAPK (Fig 5). The MAP kinases are responsive to stress stimuli, including cytokines, ultraviolet irradiation, heat shock and osmotic shock; the enzymes are, therefore, involved in many aspects of cellular physiology and apoptosis. The interaction of MAPK p38 with the C-terminus of OGT occurs at a site overlapping the PIP-interacting domain, or PPO (PIP-binding activity of OGT) . It is likely that the binding of p38 MAPK kinase would block interaction of PIP3 with the PPO binding site. From a regulatory perspective, this makes sense. The cellular stress response is triggered by numerous environmental insults including heat shock, oxidative stress, starvation, and proteotoxicity. The MAPK signaling cascade is critical for these responses and typically acts counter to growth signals. A balance exists between growth and response to stress in organisms from yeast to man. Recruitment of OGT to different sites of action, depending on the activation state of either PI3K or MAP kinase, may act as a kind of binary switch, providing a nutrient-sensitive balance between growth control and the cellular stress response (Fig. 5).
AMPK is a highly conserved sensor of cellular energy status that becomes activated by AMP under conditions where intracellular ATP levels drop. AMPK responds to this energy stress by blunting growth and biosynthetic pathways, partly through its inhibition of the rapamycin-sensitive mTOR (mTORC1) pathway (Fig. 5) [137, 138]. The mammalian TOR (mTOR) pathway is also a key regulator of cell growth and proliferation and its deregulation is associated with human diseases, including cancer and diabetes . The mTOR pathway serves to integrate the status of nutrients and cellular energy, and growth factors. In response to these upstream effectors, the mTOR pathway regulates autophagy, ribosome biogenesis, translation and metabolism. Recent work has identified two distinct mTOR-containing complexes termed mTORC1 (containing raptor) and mTORC2 (containing rictor). These complexes, along with TSC1/2, Rheb, and AMPK, function as upstream regulators of mTOR (See below). Growth factors like insulin, act through both AKT-dependent and extracellular signal-regulated kinase (ERK)-kinase dependent phosphorylation to suppress the GTPase activity of the TSC complex. This suppression leads to full activation of GTP–Rheb, which is the principal activator of the mTOR–raptor complex. Nutrients and growth factors act in concert to activate this pathway, which is, in turn, essential for conveying pro-growth and pro-survival signals (Fig. 5). In contrast, depletion of nutrients or energy stores inhibits mTOR through independent activation of the TSC either by direct phosphorylation by AMP kinase (AMPK) in response to falling ATP or via REDD1 induction during hypoxia . Potential cross talk between mTOR activation and the HSP could modulate many of the regulatory functions of this signaling pathway.
The interaction of OGT with AMPK is also complex and has great potential importance in signaling (Fig 5). AMPK-dependent processes up-regulate the levels of OGT transcript upon glucose deprivation . In addition, evidence has been presented that OGT may directly modify AMPK  leading to activation of AMPK activity in response to hexosamine flux. While this has not been directly examined in C. elegans, it is intriguing that AMPK (aak-1) mutants blunt dauer formation in manner very similar to ogt-1 mutants [14, 74, 142]. These findings all point to a concerted role for the nutrient-sensing AMPK and OGT-dependent pathways in the regulation of dauer formation and insulin-like signaling.
In single-cell eukaryotes and in C. elegans, TOR is the dominant regulator of overall mRNA translation (Fig. 5). It has become clear that in higher metazoans, mTOR controls the expression of mRNAs that are important to cell growth. Although there are no reports of direct interactions between OGT and mTOR to date, similarities certainly exist in the manner in which they may interact with the cellular translational machinery. The net result of mTOR regulation is either to stimulate protein synthesis and growth when activated or to blunt translation in response to stress and starvation. mTOR regulates translation by altering ribosome biogenesis, S6 kinase activation, activation of eIF4G and depressing the activity of eIF2. OGT has been also shown to impact the translational machinery at several different levels. Many ribosomal proteins are modified by O-GlcNAc . O-GlcNAc modifies an eIF-2 associated 67-kDa protein (p67) that protects the eIF-2 alpha-subunit from eIF-2 kinase, thus promoting protein synthesis . Both eIF-2 and eIF5 were identified as O-GlcNAc modified in a proteomics screen for O-GlcNAc modified proteins  using azido-GlcNAc probes . These probes allow protein identification, but not stoichiometry to be directly determined, and much work remains to understand the extent of modification of these key substrates. Undoubtedly, many other translation factors are modified by O-GlcNAc at levels that are below the sensitivity of detection in these preliminary studies.
Protein translation may also be regulated by mRNA triage by specialized structures termed stress granules (SG). Recent findings also point to a role for O-GlcNAcylation in mRNA triage and translation arrest by SGs . These authors showed that OGT and upstream components of hexosamine signaling are required for SG and processing body (PB) assembly. O-GlcNAc-modified proteins were found to be components of SGs, but not PBs, including numerous ribosomal proteins. These findings strongly suggest that O-GlcNAc modification of the translational machinery is required for aggregation of untranslated messenger ribonucleoproteins into SGs. Thus, the regulation of translation in response to stress is mediated both by the mTOR and hexosamine signaling pathways. Finally, as will be detailed below, OGT may play a role in modulating autophagy, a process also regulated by mTOR in response to stress and starvation.
Sirtuins, or so-called class III histone deacetylases (HDACs), are protein deacetylases/ADP ribosyltransferases acting on a wide range of cellular targets [147-150]. These target proteins reside in the nucleus, cytoplasm, and mitochondria and are post-translationally modified by either acetylation (via SIRT1, -2, -3 and -5) or by ADP ribosylation (via SIRT4 and -6) . Sirtuins play a key role in regulating lifespan extension in Drosophila and C. elegans and also are regulators of lifespan extension via calorie restriction in mammals. Finally, the sirtuins may regulate cellular responses to stress and prevent propagation of damaged DNA.
Where it has been directly examined, O-GlcNAc addition often seems to oppose the action of the sirtuins, like SirT1, on protein function (Fig. 5). An example of this is the action of the enzymes on the key glucagon-responsive transcription factor CRCT2. CRTC2 was shown to be O-GlcNAc modified at sites normally phosphorylated to keep CRTC2 in the cytoplasm. Decreasing amounts of O-GlcNAc on CRTC2, by overexpression of the deglycosylating enzyme O-GlcNAcase, blocked the effects of glucose on gluconeogenesis . In a parallel study, this laboratory showed that the histone acetyltransferase p300 and sirtuin 1 (SirT1) act on CRCT2 to modulate its activity. p300 increased hepatic CRTC2 activity by acetylating it; these effects were attenuated during late fasting, when CRTC2 was down regulated by SIRT1-mediated deacetylation. Disrupting SIRT1 activity, by liver-specific knockout of the Sirt1 gene or by administration of a SIRT1 antagonist, increased CRTC2 activity and glucose output, whereas exposure to SIRT1 agonists reduced them . A similar series of observations have been made with the enzyme endothelial nitric oxide synthase (eNOS). This enzyme is O-GlcNAcylated at a critical AKT site (serine 177) that is normally phosphorylated to activate the enzyme. The presence of O-GlcNAc at this site is associated with decreased NO production in hyperglycemic diabetic patients and is suggested to play a key role in triggering erectile dysfunction in this patient population [152, 153]. In contrast, SIRT1 acts to deacetylate eNOS, leading to a stimulation of eNOS activity and increasing endothelial NO. These studies, and others, suggest that OGT-mediated glycosylation and sirtuin-dependent deacetylation may play opposing roles in situations of nutrient excess or starvation.
As mentioned previously, OGT interacts with components of the machinery involved in mitochondrial movement in neurons, binding to the protein GRIF1, the mouse homolog of Milton in Drosophila [154-156, 156-160]. The transport of mitochondria to specific neuronal locations is critical for maintaining cellular energy demands and for buffering calcium. It is not yet clear whether the nutrient-sensing capabilities of OGT are brought to bear on regulating the axonal transport of mitochondria, but its presence in those complexes is provocative. Coupled with the finding that an isoform of OGT may be targeted to mitochondria [40, 63], these data suggest that the hexosamine signaling pathway may directly influence such critical mitochondrial functions as energy metabolism, fission and fusion, and apoptosis (Fig. 6).
A recent proteomics study has identified OGT as an integral part of the Drosophila phagosome . Phagocytic cells have critical functions in remodeling tissues during embryogenesis and are central effectors of immune defenses. During phagocytosis, large objects are internalized into 'phagosomes'. These are the organelles in which immune processes such as microbial destruction and antigen presentation are initiated. A related process occurs when intracellular components are engulfed, the resulting structure is called an autophagosome and the process is dubbed ‘autophagy.’ This is a process regulated by nutrient availability; activity of AKT and mTOR inhibit the formation of autophagosomes. However, stress and starvation induce the formation of autophagosomes through a process involving mTOR inhibition and activation of a type III PI-3 kinase. First, by analogy from its recruitment to PIP3 in the plasma membrane, OGT may also be directly recruited to the forming autophagosome. This interaction of OGT with the preautophagosome is expected to occur with newly formed PIP3 resulting from Type III PI3K activation (Fig. 6). Second, by limiting AKT activation OGT may serve to blunt mTOR activation-a critical step in autophagy. In this way, the nutrient sensing hexosamine-signaling pathway has a least two potential means to directly communicate with the forming autophagy machinery.
OGT is also potentially a component of the machinery regulating endosome trafficking [162, 163](Fig. 6). GRIF1 and Trak1, two OGT interacting proteins, interact with hepatocyte growth factor-regulated tyrosine kinase substrate (Hrs), an essential component of the endosomal sorting machinery. Although a direct role for OGT in modulating endosome to lysosome trafficking via these Trak1/GRIF1 complexes has yet to be shown, such regulation by a nutrient sensor is likely. Given the central role of O-GlcNAc cycling as a cellular nutrient sensor, a role in processes essential for nutrient acquisition might well be expected.
In addition to endocytosis, phagocytosis and autophagy, the other major cellular degradation pathway is the ubiquitin-ligase-proteasome system (Fig 6). The main function of the proteasome is to degrade unneeded or damaged proteins by proteolysis. A large body of evidence has accumulated suggesting that the HSP modulates the 26S proteasome [164-170]. These data suggest that O-GlcNAc normally serves to blunt the action of the proteasome, either through direct modification of the AAA ATPase rpt2, [167, 170] or possibly by interfering with activating kinases . There is also some recent evidence suggesting that protein ubiquitination is modulated by O-GlcNAc and that the ubiquitin-activating enzyme E1 is O-GlcNAc modified . Modulation of proteasomal activity is yet another example of how the nutrient-sensing hexosamine pathway may be coupled to modulate anabolic and catabolic metabolism intracellularly.
Even the earliest studies identifying O-GlcNAc targets hinted at a role for O-GlcNAc cycling in modulating gene expression [58, 59, 61]. Some of the sites at which O-GlcNAc may be important for gene expression are summarized Fig. 6. A more extensive treatment of this intriguing topic is presented in the article by Hart in this volume. The high relative concentration of O-GlcNAc in the nucleus and on nuclear pores suggested a regulatory role. In addition, components of transcription complexes were shown to bear O-GlcNAc residues [171, 172] and O-GlcNAc was shown to decorate Drosophila polytene chromosomes . A more detailed study then showed that the CTD domain of RNA polymerase II was modified and a model was proposed which suggested that O-GlcNAc might regulate transcription directly [174-176]. Since these early findings, a large number of papers have appeared highlighting the role of O-GlcNAc in transcription [1, 53, 87, 89, 131, 135, 177-181]. A detailed discussion of these many recent advances is beyond the scope of this review, but they point to a central role for O-GlcNAc in the regulation of transcription. Several models are in place to examine this. In C. elegans the impact of null alleles of the O-GlcNAc cycling enzymes can be examined with respect to their roles in altering chromatin structure and transcription in a living organism. In Drosophila, the Sxc/Ogt locus plays an essential role in regulating Polycomb-dependent transcriptional repression.
O-GlcNAc may also alter chromatin structure, mRNA splicing and the nuclear export machinery. Some evidence suggests that RNA binding proteins bear O-GlcNAc but no mechanistic studies have emerged testing this hypothesis [143, 145]. It is clear that OGT and nuclear pore components are part of the dosage compensation complex involved in X-inactivation in mammals . X-inactivation is triggered by an RNA dependent mechanism involving an X-inactive specific transcript (Xist) that encodes a large non-coding RNA responsible for mediating the specific silencing of the mammalian X chromosome from which it is transcribed . The dosage compensation complex, which includes OGT, and the acetyltransferase hMOF (human homolog of Male on First), is then recruited to silence chromatin on the inactive X. The role, if any, of OGT in mediating this process is currently under further study. The impact of O-GlcNAc on mRNA export has not been directly examined, but the modification has not emerged as a central player in this process. However, O-GlcNAc is likely to play a role in mRNA stability for some transcripts. As mentioned above, recent evidence points to a role for OGT in mediating the formation of SGs and PBs . These intracellular structures cooperatively regulate the translation and decay of messenger RNA.
O-GlcNAc may also have an impact on gene expression post-transcriptionally at the level of nuclear import, RNA export and mRNA stability (Fig. 6). Although the nuclear pore is a major site of O-GlcNAc addition, the role of the modification on nuclear protein import remains unclear. Using Xenopus laevis, this has been examined directly; the findings suggest that nuclear pore proteins in which O-GlcNAc was altered by addition of terminal galactose still supported nuclear import . In addition, C. elegans strains lacking O-GlcNAc cycling enzymes do not appear to have changes in the rate or extent of transcription factor nuclear localization [14, 74]. Other studies have pointed to a role for O-GlcNAc in triggering nuclear import of target proteins in response to O-GlcNAcylation[184-186]. Whether this is a primary function of the modification or a secondary consequence of O-GlcNAc cycling is an open question. Alternatively, the presence of O-GlcNAc on nuclear pores may simply reflect the recruitment of OGT to the Pore Complex during the formation of perinuclear heterochromatin associated with transcriptional repression.
It has been suggested that chitin, a polymer of GlcNAc, is the most abundant biological polymer on earth, surpassing even plant-derived cellulose. Hexosamine biosynthesis itself appears to be highly evolutionarily conserved (See Figure 3). GlcNAc and glucosamine are features of the cell wall of both gram-negative and gram-positive bacteria and amino sugars feature prominently in the glycans of protists, plants and animal species. In animals, the synthesis of UDP-GlcNAc from biosynthetic precursors by the enzymes EMeg32 and GFAT (Figure 3) is a highly regulated process. GFAT is feedback inhibited by the end product UDP-GlcNAc, suggesting tight metabolic control of hexosamine biosynthesis and UDP-HexNAc levels . In budding yeast, where a chitin-rich cell wall must be synthesized using UDP-GlcNAc, GFAT does not appear to be feedback inhibited by the product of hexosamine synthesis. Presumably, the hexosamine signaling pathway has become deregulated in yeast to allow for the formation of the chitin-rich cell wall. Curiously many yeast appear to lack OGT coding sequences (see below).
Bioinformatic inspection of known genomes and sequences suggests that the TPR domain of OGT bears some similarity to an essential domain in the yeast glucose repressor SSN6, but the yeast protein apparently lacks a catalytic domain. It has also become clear from these analyses that the Arabidopsis spindly protein is highly related to animal OGT, placing OGT in both plant and animal kingdoms. This will be discussed in greater detail in the Article by Olszewski in this volume. The animal and plant sequences, coupled with subsequent genome sequencing efforts, have allowed inspection of the likely evolution of the HSP. The sophistication of that evolutionary analysis itself continuously evolves, as more and more genomes are sequenced. Previous studies have considered the evolution of the enzymes of O-GlcNAc cycling, and sequences from certain bacteria and fungi were related in primary sequence to the OGT from human and C. elegans . In this early study, it was suggested that either OGT arose quite early in evolution and was selectively lost or that lateral transfer could explain the appearance of the enzyme in such divergent forms of life. Following this suggestion, the pathogenic bacterium Listeria was shown to encode a bifunctional OGT (GmaR) that regulates flagellar motility through anti-repression . This is a particularly interesting example, since bacterial flagella are important both for bacterial colonization of the host and for host recognition of the bacteria by the innate immune system. Many pathogens negatively regulate flagella production after infection, presumably as a defense against surveillance by the host. The Listeria GmaR protein O-GlcNAc modifies the flagellum, while also serving as a temperature-dependent repressor of flagella production by an O-GlcNAc-independent mechanism. This may represent an example of convergent evolution, where O-GlcNAc addition is catalyzed by an enzyme with distinct properties from the norm. This suggests that bacteria have used the OGT coding modules in a unique way to circumvent host defense mechanisms.
A recent detailed analysis of OGT evolution suggests that OGT-related proteins appear in many forms of bacteria, fungi, protists, plants, and of course, invertebrate and vertebrate animals . These recent findings suggests that lateral transfer is unlikely to account for these many occurrences, and points toward a more evolutionarily ancient origin of the enzyme with selective loss in some current species.
The evolution of the O-GlcNAcase is somewhat more puzzling. Enzymes clearly related to O-GlcNAcase are found in most animals and homologs are suggested to occur in some bacteria, but have yet to be identified in many organisms with homologs of OGT. The most notable absence is the lack of a clear O-GlcNAcase homolog in higher plants where two OGT-like enzymes exist. Of course, the O-GlcNAcase catalytic domain has many similarities with chitinases and hyaluronidases making unambiguous identification of more remotely related orthologs difficult. However, the suggestion that the two enzymes may not always coexist as cycling partners is an intriguing finding. Particularly in bacteria, this phenomenon may be partially explained by the use of the enzymes as toxins. Toxins are the subject of another article in the volume by Aktories. The bacterial OGT-related enzyme that has been structurally characterized (see above) has only 2 full TPR repeats and is, therefore, similar to the inactive sOGT human isoform which is a dominant negative inhibitor of human OGT and a known anti-apoptotic gene product in humans [15, 64]. If OGT-related and O-GlcNAcase-related bacterial toxins evolved from common ancestors of O-GlcNAcase and OGT, selective loss of one or both could be tolerated. They may be retained in bacteria only because of their usefulness in dealing with animal and plant hosts where O-GlcNAc cycling does occur. Alternatively, organisms may have evolved independent means of removing O-GlcNAc modified protein other than cycling, such as selective proteolysis. This might explain why OGT appears to be more widespread in evolution than O-GlcNAcase. Do these evolutionary relationships reflect a kind of host-pathogen ‘Arms-Race,’ in the use of O-GlcNAc cycling modules? Answers to this important question must await a more detailed analysis of the significance of O-GlcNAc cycling in divergent forms of life.
What is the molecular logic dictating the targets and effects of O-GlcNAc cycling? The cellular response to feast or famine is mediated by the concerted action of a variety of key signaling pathways including the Sirtuins, AMP-kinase, mTOR, and Hexosamine Signaling Pathways (Figures 5 and and6).6). In turn, these pathways interact with, and serve to modulate, homeostatic mechanisms such as the Insulin signaling, and MAP kinase-signaling cascades (Figs. 4--6).6). The HSP is of particular interest since it is responsive to cellular levels of amino acids, sugars and ATP (Fig. 3). The HSP is an evolutionarily ancient pathway that has presumably coevolved with other canonic nutrient-sensing modules (Fig 4--6).6). Growing evidence suggests that O-GlcNAc, utilizing nutrient responsive levels of UDP-GlcNAc, modulates intracellular signaling by its covalent attachment to key components of kinase-dependent signaling cascades (Figs. 4--6).6). The enzymes of O-GlcNAc cycling are recruited to their sites of action by the same activation mechanism (PI-3 kinase) triggering insulin and many other signaling cascades (Figures 4--6).6). Thus, the HSP impacts insulin signaling and other pathways by directly responding to nutrient availability. Genetic evidence further suggests that HSP, via O-GlcNAc, serves as a modulator of transcription, translation, and protein stability (Figure 6). The two key enzymes in this process, OGT and O-GlcNAcase, have thus emerged as promising drug targets. The pathways impacted by the nutrient-responsive HSP modulate key physiological processes deregulated in metabolic syndrome (stress, innate immunity, and metabolism). In fact, the O-GlcNAcase gene is a known diabetes susceptibility locus in Mexican Americans. A ‘vicious cycle’ exists in such populations; children of mothers with diabetes show increased risk for developing the disease due to unknown epigenetic factors in the intrauterine environment. One of the current hypotheses currently being pursued is that O-GlcNAc cycling integrates metabolic information, potentially leading to epigenetic reprogramming in the intrauterine environment. The HSP appears to be a central node in a complex web regulating of pathways of nutrient acquisition, stress response, immunity and metabolism. Future work in many laboratories will be required to understand the global effects of the hexosamine-signaling pathway and its role in human disease.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.