Cross-talk between GlcNAcylation and phosphorylation: roles in insulin resistance and glucose toxicity

Abstract

O-linked β-N-acetylglucosamine (O-GlcNAc) is a dynamic posttranslational modification that, analogous to phosphorylation, cycles on and off serine and/or threonine hydroxyl groups. Cycling of O-GlcNAc is regulated by the concerted actions of O-GlcNAc transferase and O-GlcNAcase. GlcNAcylation is a nutrient/stress-sensitive modification that regulates proteins involved in a wide array of biological processes, including transcription, signaling, and metabolism. GlcNAcylation is involved in the etiology of glucose toxicity and chronic hyperglycemia-induced insulin resistance, a major hallmark of type 2 diabetes. Several reports demonstrate a strong positive correlation between GlcNAcylation and the development of insulin resistance. However, recent studies suggest that inhibiting GlcNAcylation does not prevent hyperglycemia-induced insulin resistance, suggesting that other mechanisms must also be involved. To date, proteomic analyses have identified more than 600 GlcNAcylated proteins in diverse functional classes. However, O-GlcNAc sites have been mapped on only a small percentage (<15%) of these proteins, most of which were isolated from brain or spinal cord tissue and not from other metabolically relevant tissues. Mapping the sites of GlcNAcylation is not only necessary to elucidate the complex cross-talk between GlcNAcylation and phosphorylation but is also key to the design of site-specific mutational studies and necessary for the generation of site-specific antibodies, both of which will help further decipher O-GlcNAc's functional roles. Recent technical advances in O-GlcNAc site-mapping methods should now finally allow for a much-needed increase in site-specific analyses to address the functional significance of O-GlcNAc in insulin resistance and glucose toxicity as well as other major biological processes.

the addition of a single O-linked β-N-acetylglucosamine (O-GlcNAc) moiety (GlcNAcylation) to the hydroxyl groups of serine and/or threonine residues of target proteins is an inducible, reversible, and dynamic posttranslational modification (PTM). Since its discovery in the early 1980s (44, 97), O-GlcNAc has been found to have an abundance and protein distribution similar to phosphorylation, but thus far only about 600 GlcNAcylated proteins have been identified. Unlike traditional glycosylation, the O-GlcNAc modification is nearly exclusively located on cytoplasmic and nuclear proteins and is not elongated into more complex glycan structures, such as those found on extracellular or luminal domains of membrane or secreted glycoproteins. O-GlcNAc is involved in a wide range of biological processes, ranging from cell cycle progression and transcription to signal transduction and metabolism (38, 93). Emerging evidence suggests that elevated GlcNAcylation of proteins contributes to glucose toxicity and is strongly associated with hyperglycemia-induced insulin resistance, two major hallmarks of type 2 diabetes (see recent review in Ref. 19). This review will provide a critical evaluation of three broad areas and experimental approaches used to support O-GlcNAc's putative roles in the pathogenesis of diabetes: 1) regulation of hexosamine biosynthetic pathway (HBP) flux, 2) regulation of O-GlcNAc cycling enzymes, and 3) O-GlcNAc site-specific regulation of protein function. Existing data highlight the need for a significant increase in research efforts that focus on defining O-GlcNAc's site-specific regulation of metabolically relevant protein functions.

The HBP and Insulin Resistance

Since the 1950s, there have been more than 1,200 papers published linking hexosamine metabolism to diabetes. More than a decade ago, in vitro studies in rat adipocytes provided direct evidence supporting a link between flux through the HBP and the development of insulin resistance (74). Since then, multiple in vitro and in vivo studies in rodents have demonstrated that chronic elevated flux through the HBP may represent one mechanism by which hyperglycemia can lead to insulin resistance (for recent review, see Ref. 10).

There are two major nutrient inputs (glucose and glucosamine) and one rate-limiting enzyme [glutamine:fructose-6-phosphate amidotransferase (GFAT)] that represent three distinct ways to regulate the amount of flux through the HBP (Fig. 1A). First, ∼2–5% of total intracellular glucose (depending upon the cell type) enters the HBP (40, 74, 98; for review, see Ref. 8) and is ultimately converted into uridine diphosphate-N-acetylglucosamine (UDP-GlcNAc), the immediate donor substrate for GlcNAcylation. Second, the rate-limiting step in this process, conversion of fructose 6-phosphate to glucosamine 6-phosphate, is catalyzed by GFAT. Third, glucosamine can also enter the HBP directly, bypassing the rate-limiting step, and be “salvaged” into glucosamine 6-phosphate by glucosamine kinase or hexokinase. Therefore, there are three routinely used experimental approaches to evaluate the roles of hexosamine flux and subsequent increases in UDP-GlcNAc levels in the development of insulin resistance: 1) exposure to chronic hyperglycemia, 2) exposure to glucosamine, and 3) perturbation of GFAT activity. The following will give a brief critical review of the conclusions drawn using various combinations of these three experimental approaches.

Fig. 1.

Flux through the hexosamine biosynthetic pathway (HBP; A), regulation of O-linked β-N-acetylglucosamine (O-GlcNAc) cycling enzymes O-GlcNAc transferase (OGT)/O-GLcNAcase (OGA) in mediating insulin resistance (B), and Ca²⁺ cycling (C). Increased protein GlcNAcylation by 1) excess HBP flux due to elevated glucose, overexpression of glutamine:fructose-6-phosphate amidotransferase (GFAT), or glucosamine and 2) overexpression of OGT or inhibition of OGA {by O-(2-acetamido-2-deoxy-d-glucopyranosylidene)-amino-N-phenylcarbamate (PUGNAc) or 1,2-dideoxy-2′-methyl-d-glucopyranoso[2,1-d]-2′-thiazoline (GT)} contributes to the establishment of insulin resistance. Overexpression of OGT in cardiomyocytes impairs Ca²⁺ cycling, whereas OGA overexpression promotes proper Ca²⁺ cycling. Glucosamine-6-P, glucosamine 6-phosphate; Fru-6-P, fructose 6-phosphate; Gln, glucosamine; DON, doxynorleucine; UDP-GlcNAc, uridine 5′-diphospho-N-acetylglucosamine. Red arrows represent increases in respective metabolites. Blue arrows represent unknown mechanisms.

It has been exhaustively demonstrated that adipocytes and muscle cells exposed to chronic high glucose levels in the presence of insulin develop insulin resistance. Interestingly, this phenotype is prevented when adipocytes are incubated with a potent inhibitor of GFAT, doxynorleucine, and can be restored with subsequent glucosamine treatment (74, 76). These studies suggest that flux through the HBP is not only a mechanism but may also be necessary for the development of hyperglycemia-induced insulin resistance. Glucosamine was shown to be much more potent than glucose and reversible in a dose-dependent manner in its ability to induce the insulin-resistant phenotype (41, 74). Sustained exposure to high glucose or glucosamine (both of which cause elevated UDP-GlcNAc levels) results in impaired insulin-stimulated glucose transporter 4 (GLUT4) translocation and subsequent reduced insulin-stimulated glucose uptake in muscle cells without affecting insulin receptor or GLUT4 protein expression levels (10, 41). One potential problem with some of these earlier studies is the possible depletion of ATP induced by large amounts of exogenous glucosamine (47). However, more recent papers, which carefully controlled for ATP levels, still demonstrated that glucosamine is a potent inducer of insulin resistance (41, 88). Collectively, these studies suggest that a strong association exists between elevated intracellular UDP-GlcNAc levels and the occurrence of insulin resistance in vitro and in vivo.

However, a direct causal relationship between elevated UDP-GlcNAc levels and the development of insulin resistance still remains to be established and fully understood. Studies have shown that, although both high glucose and glucosamine treatments lead to increases in UDP-GlcNAc and O-GlcNAc levels, they also can lead to oxidative and endoplasmic reticulum stress (67, 111, 116), both of which have been shown to cause chronic inflammation and insulin resistance (30). This combined with the idea that O-(2-acetamido-2-deoxy-d-glucopyranosylidene)-amino-N-phenylcarbamate (PUGNAc) and/or overexpression of O-GlcNAc transferase (OGT) may increase O-GlcNAc to a level typically not achievable by high glucose suggests that caution should be used when drawing conclusions regarding functional roles of O-GlcNAc using these methods. These limitations also suggest that, to completely understand O-GlcNAc and its functional roles, site-specific studies are required, and currently these types of studies are severely lacking. Furthermore, earlier studies in other systems demonstrate that UDP-GlcNAc is a potent feedback inhibitor of GFAT (62), and therefore, the nucleotide sugar could also provide an important negative feedback mechanism to regulate HBP flux. As a consequence, our knowledge regarding regulation of HBP can be considered limited, since it remains unclear as to how up to millimolar concentrations of UDP-GlcNAc accumulate inside the cell. The presence of this possible feedback regulatory mechanism, as well as the other side effects of high glucose and glucosamine treatment, suggests that other mechanisms are likely involved in the establishment of insulin resistance in addition to elevated HBP flux.

Regulation of O-GlcNAc Cycling Enzymes and Insulin Resistance

As stated above, excess flux through the HBP results in elevated UDP-GlcNAc levels and is strongly associated with insulin resistance (Fig. 1A). In turn, O-GlcNAc transferase (OGT) catalytic activity is highly sensitive to UDP-GlcNAc concentrations (64). Thus the global extent of GlcNAcylation is highly sensitive to flux through the HBP (11, 75, 83, 89, 119), suggesting that nutrient flux through the HBP may also regulate O-GlcNAc cycling on proteins (Fig. 1B). Since O-GlcNAc is abundant on signaling proteins, including those in the insulin-signaling pathway, and has a complex interplay with phosphorylation, one likely mechanism for HBP-dependent insulin resistance is increased GlcNAcylation, which might antagonize phosphorylation-dependent insulin signaling (11, 103).

Two enzymes regulate the cycling of O-GlcNAc, OGT (uridine diphospho-N-acetylglucosmine:polypeptide β-N-acetyl- glucosaminyltransferase) (33, 35) and O-GlcNAcase (OGA; β-N-acetylglucosaminidase) (20, 26). Using UDP-GlcNAc as a donor substrate, OGT catalyzes the transfer of a single N-acetylglucosamine monosaccharide (GlcNAc) to the hydroxyl groups of serine and/or threonine residues of target proteins, whereas OGA catalyzes the removal of O-GlcNAc from proteins. The mechanism by which these two enzymes can dynamically regulate O-GlcNAc cycling in response to certain stimuli is currently not well understood. Understanding how OGT and OGA are regulated will provide insight into the regulatory mechanisms by which OGT and OGA and subsequent GlcNAcylation are involved in insulin signaling and insulin resistance. The following section briefly discusses how perturbation of these two enzymes' expression and/or activities are linked to insulin resistance.

Regulation of OGT.

There are multiple studies providing direct evidence for regulation of OGT activity in playing a role in diabetes. A recent report demonstrated that OGT- and O-GlcNAc-modified protein levels are increased in the pancreatic islets of diabetic rats (1). Studies in OGT-overexpressing transgenic mice, in which overexpression was targeted to muscle and fat tissues, demonstrated that even a moderate overexpression of OGT (∼20%) was sufficient to cause insulin resistance (76). Additionally, overexpression of OGT in liver resulted in an impairment of insulin-responsive genes and the establishment of insulin resistance (118). Likewise, C. elegans expressing an OGT^−/−-null allele exhibited dramatic changes in metabolism and impairment of dauer larva formation by the insulin-like receptor gene daf-2 (36), providing a unique model to study OGT function in insulin signaling and resistance.

Increased GlcNAcylation also appears to be important in the pathogenesis of diabetic complications such as diabetic cardiomyopathy. Overexpression of OGT in cardiomyocytes induced a phenotype common to diabetes-associated cardiac dysfunction, hyper-GlcNAcylation, and impaired calcium cycling via transcriptional inhibition of sarcoplasmic reticulum Ca²⁺ ATPase (16). This phenotype could be reversed by ectopic expression of OGA, lowering global GlcNAcylation levels. In a recent study, OGT was shown to be phosphorylated and activated by calmodulin-dependent protein kinase IV (CaMKIV) in vivo and in vitro (94). It was further demonstrated that membrane depolarization promoted OGT activation and elevated total protein GlcNAcylation in neuroblastoma NG-108-15 cells, but interestingly, inhibition of either voltage-gated calcium channels or CaMKIV could ablate this effect (94). Taken together, these results suggest that OGT activity can be regulated by calcium flux and CaMKIV-dependent phosphorylation.

OGT regulates multiple cellular processes, including transcription, translation, signaling, protein localization, and cytoskeletal reorganization (38, 69, 110). OGT-knockout embryonic stem cells are not viable, highlighting the importance of O-GlcNAc in vital cellular processes (90). OGT has been purified, characterized, and is highly conserved in a number of organisms ranging from C. elegans to humans (63, 70). The OGT gene maps to a region on chromosome Xq13 that has been linked to Parkinson's disease (79). Unfortunately, the full crystal structure of OGT has not been solved, which would provide vital insight into the mechanisms regulating GlcNAcylation of proteins. In addition to its regulation by UDP-GlcNAc, OGT's specificity and activity toward specific protein substrates is controlled by protein-protein interactions with its ∼11 tetratricopeptide repeat (TPR) protein docking domains (7, 64, 71). Recently, the crystal structure of the TPR domains of OGT was reported and found to be remarkably similar to the TPR domains of importin-α (52), a protein that serves as a transport protein to carry proteins across the nuclear envelope (29, 43). Thus, it is likely that OGT substrate targeting is regulated in a manner similar to the regulation of importin-α-mediated nuclear transport.

In the past, functional studies focusing on inhibition of OGT enzymatic activity and/or transcriptional suppression of OGT (e.g., by siRNA) have been limited due to a lack of potent in vivo inhibitors and the protein's long half-life. However, recent studies have shown that siRNA-targeted knockdown of OGT can be achieved (13, 86). Some micromolar OGT inhibitors have been developed, and new strategies are actively being pursued to develop novel, specific, and potent OGT inhibitors. One such strategy has been published recently (31). Although alloxan, a uridine analog that is often used to induce diabetes in animals, has been used to inhibit OGT (13, 61), it unfortunately is a very nonspecific inhibitor of many enzymes using uridine-containing substrates and is chemically reactive with many cellular components, including DNA (96).

Regulation of OGA.

OGA was first described as hexosaminidase C in cytosolic extracts (9), later purified and characterized (20), and subsequently cloned and sequenced (26). Similar to OGT, perturbations of O-GlcNAc cycling by modulating OGA expression and activity has further demonstrated a role of O-GlcNAc in mediating insulin resistance (Fig. 1B). Inhibition of OGA by PUGNAc (91) causes an increase in O-GlcNAc levels and leads to insulin resistance in 3T3-L1 adipocytes (103). Adenovirus OGA overexpression in diabetic hearts showed improved calcium cycling (48), whereas the exact opposite phenotype was observed with OGT overexpression (Fig. 1C). Genetic studies have linked a mutation in the OGA gene [meningioma-expressed antigen 5 (MGEA5)] to the susceptibility of diabetes in the Mexican American population (66). Finally, C. elegans expressing an OGA^−/−-null allele exhibited a phenotype metabolically similar to that of human type 2 diabetes (24).

OGA is expressed in all tissue types examined and has a tissue distribution similar to OGT, but unlike OGT, which is mostly in the nucleus, OGA is localized mostly in the cytoplasm (26). However, on average, ∼20% of OGT is within the cytoplasm, and an equivalent proportion of OGA is found within the nucleus. The OGA gene was found to be identical to MGEA5, which was initially identified to be associated with human meningiomas (42). The OGA gene also maps to a region on chromosome 10q24.1, which has been linked to Alzheimer's disease (6), sometimes referred to as “diabetes type 3” (17). As with OGT, the mechanisms by which OGA is regulated in response to certain stimuli are not well understood. OGA may be regulated during apoptosis via specific cleavage by caspase-3; however, the cleavage does not seem to have an adverse effect on OGA activity in vitro (107) and may play other roles in either the localization or targeting of the enzyme. Currently, the crystal structure of OGA has not been solved. As with OGT, the crystal structure of OGA would provide structural insights into the mechanism by which OGA may regulate O-GlcNAc cycling and insulin resistance.

The discoveries of OGA inhibitors have greatly aided in the study and understanding of O-GlcNAc, but their use can be associated with some drawbacks. PUGNAc, while successfully inhibiting OGA to increase O-GlcNAc levels, may also cause the inhibition of various lysosomal hexosaminidases, therefore complicating the interpretation of results as being a direct result of elevated GlcNAcylation. Additionally, 1,2-dideoxy-2′-methyl-d-glucopyranoso[2,1-d]-2′-thiazoline [NAG-thialozine (GT)] was found to be a potent inhibitor of OGA and is now preferred over PUGNAc (60, 116) but also displays considerable activity toward β-hexosaminidase (73). Recent studies have also indicated that these commonly used inhibitors may have distinctly different effects on different isoforms of OGA (59). Consequently, a series of GT analogs that display considerable selectivity toward OGA have been developed. One GT analogue in particular, 1,2-dideoxy-2′-propyl-d-glucopyranoso[2,1-d]-2′-thiazoline, displayed a 3,100-fold selectivity toward OGA over β-hexosaminidase (73). Additionally, GlcNAcstatin (a glucoimidazole) was also developed and was shown to have a high selectivity for the inhibition of OGA (100,000-fold) over that of hexosaminidases (21). Although reported in the literature (61, 69), the GlcNAc analog streptozotocin (STZ) is, in fact, a poor inhibitor of OGA (38) and has many other side effects that are toxic to cells.

Evidence for a direct role of O-GlcNAc in insulin resistance.

Although multiple studies demonstrate a strong association between insulin resistance and perturbation of O-GlcNAc pathways, a direct causal role for O-GlcNAc in the development of insulin resistance remains controversial. Elevation of O- GlcNAc levels causes insulin resistance and impaired glucose uptake in multiple cell types (76, 103). Therefore, it has been proposed that blocking nutrient-dependent increases in GlcNAcylation might also prevent nutrient-induced insulin resistance. However, Robinson et. al. (86) recently reported that chronic reduction of protein O-GlcNAc levels by overexpressing OGA does not prevent insulin resistance upon sustained treatment with glucose and low-dose insulin. In these studies, although global GlcNAcylation levels were decreased, specific effects of OGA overexpression on the GlcNAcylation of proteins within the insulin-signaling pathway were not determined. Since isoforms or holoenzyme complexes of OGA are known to have different inhibitor sensitivities and perhaps different substrate specificities, future studies will be needed to examine the effect of OGA expression on the GlcNAcylation of individual site(s) on these key signaling proteins. Robinson et. al. (86) also note that OGA expression and its activity was mainly cytosolic, confirming a previous report demonstrating that overexpression of OGA has limited nuclear expression (26). To overcome this technical shortcoming, these investigators took another approach using siRNA-targeted knockdown of OGT to determine whether decreased or inhibited GlcNAcylation could prevent insulin resistance. Similarly, an ∼90% knockdown of OGT expression and consequent global decrease in GlcNAcylation levels (to a similar degree) did not prevent the development of insulin resistance in 3T3-L1 adipocytes (86), which appears to contradict the conclusions of early studies in this same cell type (74, 75). Molecular interpretation of these results will also require a study of the effects of OGT RNA interference on GlcNAcylation of specific site(s) on key insulin signaling proteins. Very recent data suggest that GlcNAcylation does indeed play a role in insulin resistance mediated by OGT targeting to the plasma membrane via phosphoinositides to increase GlcNAcylation of key components within the insulin-signaling pathway (118). Thus, existing data strongly suggest that elevated O-GlcNAc can itself induce insulin resistance, but clearly, many other molecular events contribute to insulin resistance (10, 86).

Regulation of OGT/OGA targeting and complex formation in insulin resistance.

Unlike the presence of 518 kinase genes in the human genome (73), the myriad nucleocytoplasmic proteins are specifically O-GlcNAc modified by only one single highly conserved catalytic subunit, OGT (63). Thus, OGT targeting and specificity are quite different than those for kinases (50). Current data indicate that OGT is targeted to its substrates via transient complex formation with other polypeptides, analogous to regulation of RNA polymerase II promoter recognition specificity. This model is supported by structural data showing that OGT has two distinct domains, a COOH-terminal catalytic domain similar to that of glycogen phosphorylase (115) and an NH₂-terminal domain containing multiple TPRs involved in mediating protein-protein interactions (7). OGT and OGA are often in the same complexes (113), suggesting tight reciprocal regulation to prevent a futile cycle. Yeast two-hybrid studies have uncovered many putative OGT-targeting proteins, including the OGT-interacting protein p106 (OIP106, also known as TRAK1) (49). OIP106 is GlcNAcylated, coimmunopurifies with OGT, and targets OGT to transcription complexes. With regard to insulin signaling, recent studies demonstrate that, under high glucose conditions, NeuroD1a, a key transcription factor controlling insulin synthesis, interacts with OGT, but in contrast, under low glucose conditions, NeuroD1a interacts with OGA (4). Additionally, it was demonstrated that GlcNAcylated NeuroD1a has increased nuclear localization (4). OGT and protein phosphatase 1 form transient complexes (Ref. 108 and Slawson C and Hart GW, unpublished observations), suggesting that in some cases the same enzyme complexes may regulate both the removal of O-phosphate and addition of O-GlcNAc to target proteins. This highlights a very important concept for how site-specific O-GlcNAc/O-phosphate interplay may mediate phosphorylation-mediated events (i.e., insulin-signaling cascades).

Dynamic interplay between O-GlcNAc and O-phosphate in insulin resistance.

Many studies have established that the O-GlcNAc modification may be regulated in part by its interplay with phosphorylation and vice versa (38, 92). O-GlcNAc is similar to O-phosphate in terms of its abundance and presence on serine and/or threonine residues of nuclear and cytoplasmic proteins. Most O-GlcNAc-modified proteins are also known phosphoproteins. Given their similarities, it is not surprising that one modification may spatially, or physically, regulate the presence of the other.

Wang et. al. (105) recently demonstrated that several proteins had increased GlcNAcylation in response to glycogen synthase kinase-3 inhibition. Okadaic acid (a phosphatase inhibitor) treatment increased global phosphorylation while concomitantly decreasing global GlcNAcylation in various cell lines (65). Conversely, treatment with PUGNAc (an inhibitor of OGA) increased global GlcNAcylation and decreased phosphorylation in 3T3-L1 adipocytes (103). These, among other data, have led to the “ying-yang” hypothesis (108) whereby O-GlcNAc may be competitive and/or reciprocal to O-phosphate in the regulation of certain phosphorylation-mediated events (e.g., insulin-signaling cascades). Using various mass spectrometric techniques, a number of O-GlcNAc-modified proteins have had their O-GlcNAc sites mapped (Supplemental Table 1; Supplemental data for this article are available online at the American Journal of Physiology-Endocrinology and Metabolism website). On some proteins, O-GlcNAc and O-phosphate have identical sites of modification. For example, Thr⁵⁸ of c-Myc (15, 54), Ser^111/112 of SV large T-antigen (77), Ser¹⁶ of estrogen receptor-β (14), and others reciprocally carry either the O-GlcNAc or the O-phosphate modification at the same hydroxyl moiety (Supplemental Table 1). Furthermore, as in the case of the tumor antigen p53, an O-GlcNAc modification adjacent to a known regulatory phosphorylation site can affect phosphorylation of that residue and subsequent protein function (117). It is therefore reasonable to postulate that elevated O-GlcNAc levels [as observed in diabetic rats (1, 39)] may blunt insulin's phosphorylation-dependent signaling pathways, contributing to insulin resistance (Fig. 2). This hypothesis is supported by a recent study demonstrating that, upon insulin stimulation, phosphatidylinositol 3,4,5-triphosphate recruits OGT to the plasma membrane, leading to GlcNAcylation and altered phosphorylation of key signaling molecules in the insulin-signaling pathway (i.e., Akt and IRS-1) and attenuation of insulin signal transduction (118). In the future, site-specific studies will be required to definitively assess the functional roles that O-GlcNAc may play in attenuating insulin signal transduction. However, conventional approaches at elucidating the functional role of a PTM on a specific residue (e.g., site-directed mutagenesis) cannot be employed if the same residue harbors either the O-GlcNAc or O-phosphate modification.

Table 1.

O-GlcNAc site-mapped proteins that have metabolic relevance

Protein Name	Protein Function	Species	Tissue	O-GlcNAc Peptides	Residues	Metabolic Relevance	Ref. No.
α-crystallin B chain	Structural component of the lens	Rat	Lens, Heart	EEKPAVTgAAPK	164–174	Glucose, oxygen/reactive oxygen species metabolism	87
c-Myc	Transcription factor, oncogene	Human	Sf9 insect cells	FELLPTg/pPPLSpPSR	53–65	May regulate LDHA among other metabolic genes	15
CRTC2/TORC2	Transcription coactivator for CREB1	Human	HEK-293 cells	SHYGGSg/pLPNVNQIGC SSg/pDSgALHTSVMNPNP WSTFRPRTgSgSNASRISG*	65–79 170–184 310–326	Regulator of hepatic gluconeogenesis and glucose output	18
FoxO1	Transcription factor	Human	Fao cells	TVSTMPHTSgGMNRLTQ QSFPHSVKTTgTHSWVSgG	542–557 639–655	Transcription regulator of multiple glucose-responsive genes	46
IRS-1	Regulator of insulin signaling	Rat cDNA	HEK-293 cells	APPPSSgTASASASV	1,031–1,044	Insulin receptor/insulin-like growth factor receptor signaling	5
NCOA1/SRC-1	Transcription coactivator for nuclear receptors, histone acetyl transferase	Rat	Brain	INPSVNPGISpPAHGVTR INPSVNPGISpPAHGVTgR	386–402 386–402	Major steroid hormone receptor coactivator (interacts with STAT transcription factors)	57
OGA	N-acetyl-d-glucosaminidase	Rat	Brain	QVAHSgGAK	401–408	Chronic inhibition of OGA (w/PUGNAc) leads to insulin resistance	56
p53	Tumor suppressor	Human	MCF-7 cells	CPVQLWVDSgTPPPGTpR	141–156	Mediates cellular stress response to glucose starvation	117
Piccolo	Scaffolding protein	Mouse	Brain	SCTAQQPATgTgLPEDR*	2,851–2,875	Involved in insulin secretion	101
Ponsin	SH3 domain containing adaptor protein	Mouse	Brain	TPVDYIDLPYSgSgSgPSR*	1,189–1,204	Required for insulin-med glucose transport, insulin receptor signaling	101
Rab3 GDP/GTP exchange protein	GTP/GDP exchange factor for Rab3 subfamily G proteins	Rat	Brain	SSSSTTASSSPSTIVHGA HSEPADSTEVGDK	699–729	Involved in regulation of MAP kinase activity	56
Spectrin β2	Axonal/presynaptic cytoskeletal protein	Rat	Brain	HDTSASTQSTPASSR HDTSASTQSTPASSgR	2,305–2,319 2,310–2,324	Interacts with calmodulin in a calcium-dependent manner	57
		Mouse	Brain				101

O-GlcNAc, O-linked β-N-acetylglucosamine; CRTC2/TORC2, transducer of regulated cAMP response element-binding protein (CREB)2; HEK, human embryonic kidney; LDHA, lactate dehydrogenase A; FoxO1, forkhead box protein O1; IRS-1, insulin receptor substrate-1; NCOA1/SRC-1, nuclear/steroid receptor coactivator-1; OGA, O-GlcNAcase; PUGNAc, O-(2-acetamido-2-deoxy-d-glucopyranosylidene)-amino-N-phenylcarbamate; MCF-7, human breast adenocarcinoma; O-GlcNAc-mapped sites are in bold font. Published phosphosites (as cited by the UniProtKB protein database) are in italics.

^*Technical limitations restricted identification of the exact location of the O-GlcNAcylated residue to 1 of the underlined residues. The O-GlcNAc site of Rab3 GDP/GTP was not explicitly determined, but the O-GlcNAc was localized to the peptide listed.

Fig. 2.

O-GlcNAc-modified proteins directly involved in glucose-responsive transcription and signaling events. See main text for detailed description. IRS-1, insulin receptor substrate-1; GS, glycogen synthase; Sp1, transcription factor Sp1; PDX-1, pancreatic duodenal homeobox-1; NeuroD1a, neurogenic differentiation 1; STZ, streptozotocin; GLUT4, glucose transporter 4. Red arrows represent an increase or decrease in respective metabolites.

O-GlcNAc in Glucose-Responsive Transcription and Signaling Events

Multiple proteins involved in glucose-responsive and insulin-dependent transcription and signaling events have been identified as being O-GlcNAc modified (10, 19). However, because studies identifying exact O-GlcNAc sites followed by site-specific functional analyses are extremely limited, the functional significance of the O-GlcNAc modification on these proteins has not yet been clearly defined. This is highlighted by the fact that of the 73 proteins that have had their O-GlcNAc site(s) mapped (see Supplemental Table 1), there are few site-directed mutagenesis or site-specific functional followup studies reported to date. Several studies have shown that when Thr⁵⁸ of c-Myc is mutated to a nonhydroxyamino acid, this transcription factor becomes an oncoprotein (3, 32). However, since Thr⁵⁸ is O-GlcNAc modified in nongrowing cells (54) and phosphorylated by glycogen synthase kinase-3β kinase in growing cells, the functional significance of each PTM remains unclear. Additionally, the majority of proteins whose O-GlcNAc sites have been mapped (>80%) were isolated from brain and spinal cord, not from other tissues relevant to diabetes and metabolism (i.e., muscle, liver, and fat). Of all proteins O-GlcNAc site-mapped to date, only two [nuclear pore glycoprotein p62 and forkhead box protein O1 (FoxO1); Supplemental Table 1] have been mapped in a tissue that is metabolically directly responsive to insulin (i.e., liver).

GlcNAcylation of proteins regulating glucose-dependent insulin expression.

OGT- and O-GlcNAc-modified protein levels are increased in the pancreatic islets of diabetic rats (1). Furthermore, PUGNAc-induced elevation of GlcNAcylation resulted in a decrease in glucose-dependent insulin secretion in isolated pancreatic islets (1). Thus, chronic elevation of GlcNAcylated proteins may cause dysregulation of pancreatic insulin expression and secretion, contributing to hyperinsulinemia and the development of insulin resistance.

Of the three transcription factors known to regulate glucose-dependent insulin gene expression and secretion in β-cells, two are known to be O-GlcNAc modified, NeuroD1a and pancreatic duodenal homeobox-1 (PDX-1) (see review in Ref. 12). Under elevated glucose conditions, GlcNAcylation of NeuroD1a and PDX-1 are increased, resulting in their translocation to the nucleus accompanied by an increase in DNA binding and glucose-dependent insulin synthesis (β-cell; Fig. 2) (4, 25). Interestingly, it has been reported that under identical conditions PDX-1 also becomes phosphorylated, with similar downstream effects (72), suggesting that both phosphorylation and GlcNAcylation of PDX-1 may synergistically regulate insulin synthesis. The demonstration that elevation of GlcNAcylation by PUGNAc “decreases” glucose-dependent insulin secretion may seem counterintuitive, since an increase in GlcNAcylation of NeuroD1a and PDX-1 was shown to “increase” glucose-dependent insulin secretion. However, these seemingly opposite results highlight the need for more site-specific studies to elucidate the functional role of O-GlcNAc modifications on certain residues in regulating NeuroD1a and PDX-1 activity vs. the effect global elevation of O-GlcNAc has on regulating these proteins' activities. Currently, it is unknown whether the third transcription factor known to regulate glucose-dependent insulin gene expression, Maf1, is O-GlcNAc modified; however, it has been demonstrated that its expression is regulated by glucose and GlcNAcylation of unknown proteins (99). These data, as summarized above (β-cell; Fig. 2), suggest that, through modification of insulin transcription factors (NeuroD1a, PDX-1, and possibly Maf1), O-GlcNAc is involved in insulin transcription. However, it remains to be proven after O-GlcNAc site mapping and site-directed mutagenesis whether O-GlcNAc displays site-specific functionality on NeuroD1a, PDX-1, and possibly Maf1.

GlcNAcylation of proteins regulating glucose-responsive transcription/signaling events.

Hyper-GlcNAcylation of the transcription factors Sp1, FoxO1, and the transcriptional coactivator CRTC2 or TORC2 [transducer of regulated cyclic adenosine monophosphate response element-binding protein (CREB) 2] have all been linked to high glucose-induced expression of gluconeogeneic genes and thus may contribute to glucose toxicity (18, 45, 46, 51, 104). Hyperglycemic condition results in increased GlcNAcylation of Sp1 at multiple sites (51, 104) and is directly correlated with its DNA-binding and transcriptional activities (53, 106) regulating multiple glucose responsive genes associated with diabetes (muscle/fat cell; Fig. 2) (22). More specifically, recent studies have demonstrated that high glucose-induced GlcNAcylation of Sp1 enhances the transcription of plasminogen activator inhibitior-1 (PAI-1), which plays a role in diabetic cardiovascular disease (22, 27, 28). High glucose-induced GlcNAcylation of Sp1 and subsequent expression of PAI-1 were inhibited by siRNA-targeted knockdown of OGT, overexpression of a dominant-negative OGT, and overexpression of OGA (28). Furthermore, PUGNAc treatment under normal glucose conditions increased both Sp1 GlcNAcylation and PAI-1 expression (28). To date, the O-GlcNAc site(s) of Sp1 has not been determined, but accumulating evidence highly suggests that Sp1 transcriptional activity can be regulated directly by its O-GlcNAc modification status independent of adverse effects due to elevated glucose and/or HBP flux.

Although O-GlcNAc site-specific functional analyses of Sp1 are needed, evidence suggestive of O-GlcNAc/O-phosphate interplay on Sp1 (34) highlights the current challenges faced when using classical site-directed mutagenesis to determine the functional role of a particular modification (i.e., O-GlcNAc) that can reside on the same residue as another modification (i.e., O-phosphate) or is regulated by a neighboring residue containing either modification. As mentioned above, GlcNAcylation of CRTC2 and FoxO1 have been shown to mediate high glucose-induced hepatic gluconeogenesis independently of insulin signaling (18, 45, 46). Because O-GlcNAc site-specific functional analyses have been performed for both CRTC2 and FoxO1, a more detailed description of these studies will be discussed in the next section. Taken together, data on GlcNAcylation-dependent regulation of Sp1, CRTC2, and FoxO1 all suggest that O-GlcNAc plays a pivotal role in glucose-responsive gene expression and high glucose-induced hepatic glucose output, independent of insulin signaling, and thus may mediate glucose toxicity.

Endothelial nitric oxide synthase (eNOS) and glycogen synthase (GS) represent two additional GlcNAcylated glucose-responsive proteins. Endothelial dysfunction is an abnormality that may accompany insulin resistance and diabetes but not necessarily be a direct cause of it (84). Under normal conditions, insulin properly regulates eNOS function via the insulin receptor (IR)-IR substrate (IRS-1)-phosphatidylinositol 3-kinase pathway (120). Under hyperglycemic conditions, eNOS function is impaired (Fig. 2) (23). Interestingly, under these same conditions, GlcNAcylation of eNOS is increased, whereas phosphorylation at Ser¹¹⁷⁷ is decreased (21), suggesting that there may be some O-GlcNAc/O-phosphate interplay. These results also suggest that the O-GlcNAc site of eNOS may be Ser¹¹⁷⁷, but this remains to be definitively proven by mass spectrometry techniques, such as those briefly described later. Furthermore, elevated GlcNAcylation of eNOS was shown to inhibit its subsequent activation by Akt (21), suggesting that the site of GlcNAcylation is at or near the Akt phosphorylation site. GlcNAcylation of eNOS has been implicated in diabetic erectile dysfunction (80). Glycogen synthase (GS) has been demonstrated to be O-GlcNAc modified in 3T3-L1 adipocytes (82). GlcNAcylation of GS was demonstrated to reduce its ability to be activated by insulin (muscle/fat cell; Fig. 2) (81, 82). Diabetic STZ-treated mice had reduced GS activity and increased GlcNAcylation of GS (81, 82). This suggests that O-GlcNAc is involved in some aspect of regulating GS function.

The above represents a review of a number of glucose-responsive proteins identified as being O-GlcNAc modified (e.g., NeuroD1a, PDX-1, Sp1, CRTC2, FoxO1, eNOS, and GS) with data suggestive of O-GlcNAc having a functional role on these proteins. However, these studies have also highlighted the need for more site-specific O-GlcNAc functional studies. Hopefully, such studies will help to elucidate the importance of this modification on specific residues in regulating proteins involved in mediating the many aspects of insulin resistance and the pathogenesis of type 2 diabetes.

Identification of O-GlcNAcylated residues on metabolically relevant proteins.

Of the 75 proteins that have had their O-GlcNAc site(s) mapped or localized, (Supplemental Table 1), 12 proteins were chosen as having established metabolic relevance and are listed in Table 1. One of these proteins, IRS-1, is directly involved in mediating insulin signaling by acting as a major intracellular docking/adaptor protein linking insulin-dependent receptor autophosphorylation events to downstream signaling cascades, resulting in GLUT4 translocation (muscle/fat cell; Fig. 2). Mass spectrometric studies along with site-directed mutagenesis have confirmed that Ser¹⁰³⁶ is a major site of GlcNAcylation of IRS-1 (5). A UniProtKB database search did not reveal any known phosphorylation sites located within five amino acids of this O-GlcNAc site (Table 1). Therefore, a direct O-GlcNAc/O-phosphate interplay at Ser¹⁰³⁶ may not be a functional possibility in the regulation of IRS-1. However, it is important to note that in the three-dimensional structure of a protein, two seemingly distant residues could be in close proximity. In any case, the functional significance of GlcNAcylation of Ser¹⁰³⁶ on IRS-1 remains to be elucidated.

Two other metabolically relevant proteins that have had their O-GlcNAc sites mapped are CRTC2 and FoxO1. As stated in the previous section, high glucose-induced GlcNAcylation of CRTC2 and FoxO1 mediates hepatic gluconeogenesis. However, in contrast to IRS-1, a site-specific functional role for these modifications has been established. Under normal conditions, CRTC2, a CREB transcriptional factor coactivator for CREB, is phosphorylated at both Ser⁷⁰ and Ser¹⁷¹, causing it to be sequestered in the cytoplasm by 14-3-3 proteins. However, it was recently demonstrated that, under high glucose conditions, CRTC2 becomes GlcNAcylated at Ser⁷⁰ and Ser¹⁷¹, resulting in its nuclear translocation and transcription of hepatic gluconeogeneic genes (i.e., glucose 6-phosphatase and phosphoenolpyruvate carboxykinase; Fig. 3) (18). Mutations at these two residues disrupted the CRTC2:14-3-3 protein interaction and resulted in the nuclear localization of the CRTC2 mutants (18). This result is consistent with the increasing data that suggest that a functional and dynamic interplay exists between GlcNAcylation and phosphorylation. In a separate study, it was demonstrated that hepatic FoxO1 GlcNAcylation is increased in STZ-treated rats (45, 46). FoxO1 also controls the expression of gluconeogeneic enzymes glucose-6-phosphatase and phosphoenolpyruvate carboxykinase. Under hyperglycemic conditions, FoxO1 is GlcNAcylated on Thr³¹⁷, resulting in an increase in its transcriptional activity (Fig. 3) (45, 46). Mutation of Thr³¹⁷ to an alanine decreased FoxO1 activation during a luciferase reporter assay in response to high glucose (46). Furthermore, insulin blocked high glucose-induced GlcNAcylation and subsequent activation of FoxO1 (45, 46). Taken together, these studies suggest that high glucose-induced GlcNAcylation of CRTC2 and FoxO1 and subsequent regulation of hepatic gluconeogenesis may play a role in glucose toxicity, especially under an insulin-resistant setting. Additional studies are needed to determine whether CRTC2 and FoxO1 act synergistically in their functional regulation of gluconeogenesis.

Fig. 3.

O-GlcNAc site-specific regulation of transducer of regulated cAMP response element-binding protein (CREB)2 (CRTC2; a.k.a TORC2) and forkhead box protein O1 (FoxO1) in mediating hepatic gluconeogenesis and “glucose toxicity.” High glucose-induced expression of gluconeogeneic genes is dependent on HBP flux and in part by GlcNAcylation of FoxO1 and CRTC2. Red arrows represent adenoviral overexpression of OGT and OGA. AZA, azaserine (GFAT inhibitor); 14-3-3, 14-3-3 protein family members.

Although there are 75 proteins with their O-GlcNAc sites at least localized to a peptide, more than 600 proteins have been identified as O-GlcNAc modified, 13% of which can be directly related to metabolic pathways (69). Therefore, a relatively small subset of these O-GlcNAc-identified proteins (<15%) have had their O-GlcNAc sites mapped. Furthermore, a majority of the O-GlcNAc site-mapped proteins (>80%) were isolated from brain/spinal cord tissues (where O-GlcNAc is abundant) vs. other metabolically relevant tissues (i.e., muscle, liver, pancreas, or fat). O-GlcNAc is most abundant in the islets of the pancreas, where it appears to serve as a nutrient sensor to regulate insulin synthesis (1, 2, 4, 25, 60); however, few O-GlcNAc sites from this tissue have been studied. It is therefore of crucial importance that proteomic analyses be performed on proteins isolated from more metabolically relevant tissues to not only increase the number of known O-GlcNAc-modified proteins but also identify the sites of modification on these proteins within a metabolic context.

As discussed, a major lack in knowledge regarding O-GlcNAc function is due to limited studies that have directly confirmed a site-specific functional role for O-GlcNAc on a given protein. In addition, site-directed mutagenesis studies are not feasible for cases where O-GlcNAc may occur at the same residue as phosphorylation, making site-specific functional studies of O-GlcNAc at these sites very difficult. However, in cases in which there is no “suggested/predicted” interplay with phosphorylation, site mapping followed by site-directed mutagenesis and functional studies would be extremely useful in determining a functional phenotype for O-GlcNAc. Various methods have been developed that have greatly aided in the success of mapping O-GlcNAc sites of proteins. Some of these methods are briefly highlighted in the next section (for recent review, see Ref. 85).

Approaches to O-GlcNAc Site Mapping

Progress in the study of GlcNAcylation has been slow, due primarily to difficulties associated with its detection and analysis. Unlike phosphate, GlcNAc is uncharged and does not result in an observable shift of the modified protein during electrophoresis, even on high-resolution two-dimensional gels. Additionally, O-GlcNAc is substoichiometric, like phosphorylation, and is rapidly lost via glycosidase action upon cellular damage. In addition, O-GlcNAc is much more labile during ionization and in the gas phase in mass spectrometry, and detection of GlcNAc peptides is suppressed in the presence of unmodified peptides. The lack of any recognizable consensus primary sequence for O-GlcNAc further complicates its study. Fortunately for the field, recently, several chemical/enzymatic, immunological, and analytical techniques have been developed, with each having advantages and limitations as described (112).

Chemoenzymatic techniques coupled with a variety of mass spectrometric methods have been developed and are essential in O-GlcNAc site mapping of proteins (55, 95, 101, 107). Quantitative proteomics studies have utilized modifications of the BEMAD (mild β-elimination followed by Michael addition with dithiothreitol) method to differentially label O-GlcNAc-modified peptides with dithiothreitol (101, 109). It was discovered that the enzymes responsible for the O-GlcNAc modification could use analogs of their natural substrate UDP-GlcNAc in which the N-acetyl group bears an N-azido group (100). In cells treated with N-azidoacetylglucosamine (GlcNAz), GlcNAz is incorporated onto proteins (100). The O-GlcNAz-modified proteins could then be derivatized (i.e., probes, tags, etc.) to reveal their glycoslyation site(s). Because O-GlcNAc is lost during collision-activated dissociation methods, the above techniques coupled with ion trap mass spectrometry with electron transfer dissociation (78) and Fourier transform mass spectrometry with electron capture dissociation (102) have greatly aided in the detection and site mapping of O-GlcNAc-modified proteins.

Recently, a strategy has been developed that allows the dynamics of O-GlcNAc to be monitored using quantitative-based MS proteomics (i.e., electron transfer dissociation) (56). The method, termed quantitative isotopic and chemoenzymatic tagging, utilizes a selective and chemoenzymatic approach to tag O-GlcNAc-modified proteins isotopically, whereby detectable changes in GlcNAcylation can be monitored.

Conclusion and Future Directions

O-GlcNAc playsvarious roles in almost all vital cellular processes. Specifically, data summarized above suggest that alterations in total protein GlcNAcylation (either by flux through the HBP or modulation of O-GlcNAc-cycling enzymes) and site-specific protein GlcNAcylation are both influential in insulin secretion, the pathogenesis of insulin resistance, and the development of other diabetic complications. Although there are many studies providing direct evidence for perturbation of total GlcNAcylation levels altering glucose-responsive signaling pathways, the GlcNAc field is relatively deficient in studies focusing on the site-specific role of the O-GlcNAc modification in regulating protein function. Mapping O-GlcNAc sites on proteins represents a critical next step to investigate O-GlcNAc's functional roles. Furthermore, to date more than 80% of the proteins/peptides that have had their O-GlcNAc sites mapped are from brain/spinal cord tissue. Proteomic studies done in other metabolically relevant tissues (i.e., muscle, adipose, pancreas, and liver) would be advantageous for gaining a better understanding of the role GlcNAcylation plays in metabolism, insulin signaling, and the development of type 2 diabetes. A wealth of information is known about O-GlcNAc; however, given its obvious complexity, the field still has much to learn about O-GlcNAc's regulation, its interplay with O-phosphate, and the underlying and/or causal roles it may have in establishing insulin resistance, glucose toxicity, and the onset of diabetes.

GRANTS

Research conducted in the corresponding author's laboratory was supported in part by National Institutes of Health (NIH) Grants R01-DK-061671, R33-DK-071280, R37-HD-13563, and R01-CA-42486 and NIH contract N01-HV-28180. G. W. Hart receives a share of the royalties received by Johns Hopkins University on sales of the CTD 110.6 antibody. Terms of this arrangement are managed by Johns Hopkins University.