Protein O-GlcNAcylation and the regulation of energy homeostasis: lessons from knock-out mouse models

O-GlcNAcylation was discovered in 1984 by GW. Hart as a new type O-linked glycosylation [1], that was unexpectedly found in the cytosolic and nuclear cell compartments. O-GlcNAcylation has gained considerable interest in biomedical sciences in the last two decades, as it regulates most cellular processes and has been involved in several important human pathologies, including metabolic [2], cardiovascular [3] and neurogenerative diseases [4], as well as cancer [5]. This intriguing reversible post-translational modification regulates a wide range of cellular proteins (more than 4000 proteins identified to date) in a manner reminiscent of phosphorylation [6, 7]. The O-GlcNAc cycling system is apparently simplest than the phosphorylation signalling network, as only two enzymes, OGT and OGA, cloned in 1997 [8, 9] and 2001 [10] respectively, regulate the addition and removal of this sugar from proteins. However, this apparent simplicity brings with itself considerable difficulties for the understanding of its mechanism and specificity of action, both at cellular and whole organism levels [7].

In recent years, numerous studies have tried to explore these issues using OGT and OGA knock-out mouse models. Because O-GlcNAcylation is a post-translational modification tightly dependent on nutritional conditions, and particularly on glucose availability, it plays a determinant role in the regulation of energy metabolism and in metabolic diseases such as diabetes and obesity. In this review, we have analysed the recent literature reporting the consequences of OGT or OGA-KO in mice, but focusing more specifically on those models that have brought important insights into our understanding of the role of O-GlcNAcylation in the regulation of energy homeostasis. We more specifically discuss the apparent contradiction arising from results indicating that whereas increased O-GlcNAcylation level associated with high glucose conditions have deleterious effects, impaired O-GlcNAcylation also markedly alters the regulation of energy metabolism.

O-GlcNAcylation, a highly dynamic post-translational modification

The regulation of protein activity by post-translational modifications is central to cell signaling mechanisms. Phosphorylation is certainly one of the most studied post-translational modification. However, many other modifications (glycosylation, sumoylation, acetylation, ubiquitinylation, nitrosylation, palmitoylation, ADP-ribosylation, hydroxylation, etc.) are also present on proteins and control their activity, stability and/or subcellular localization. Among these, a particular glycosylation, O-GlcNAcylation, is becoming increasingly important in many fields of biology and biomedical research [6, 7].

Unlike canonical glycosylations, that take place in the endoplasmic reticulum and the Golgi apparatus and mainly concern secreted proteins or extra-cellular domains of membrane proteins, O-GlcNAcylation corresponds to the addition of N-Acetylglucosamine (GlcNAc) on serines and threonines of cytosolic, nuclear and mitochondrial proteins [6, 7]. Like phosphorylation, O-GlcNAcylation is a reversible modification that controls the activity, the stability and the subcellular localization of proteins, as well as their interactions with different cellular partners. Furthermore, O-GlcNAcylation can modulate protein phosphorylation on serines or threonines, either by competing with phosphorylation for the same amino acid or by up-regulating or down-regulating phosphorylation of an adjacent amino acid [11]. However, unlike phosphorylations-dephosphorylations, which are regulated by a myriad of kinases and phosphatases, only two enzymes, OGT (O-GlcNAc transferase) and OGA (O-GlcNAcase) control the addition or removal of the GlcNAc molecule on proteins [12]. OGT and OGA are highly conserved during evolution [8,9,10]. O-GlcNAcylation is found in most living organisms, with the notable exception of yeast, in which an equivalent has been revealed recently (O-manosylation of cytosolic and nuclear proteins) [13].

The puzzling question of substrate specificity

OGT comprises on its N-terminal side a domain containing tetra-tricopeptide repeats (TPR), known to be involved in protein–protein interactions, and in its C-terminal region, the glycosyl-transferase catalytic activity (Fig. 1).

Fig. 1figure 1

O-GlcNAc-cycling enzymes. A Three OGT isoforms, two nucleocyplasmic (ncOGT and sOGT) and one mitochondrial (mOGT), are generated by alternative splicing of the OGT messenger RNA (MTS Mitochondria Targeting Sequence). OGT comprises a N-terminal domain containing tetra-tricopeptide repeats (TPR) involved in protein–protein interactions and a C-terminal domain with glycosyltransferase activity. The PPO domain (Phophosphatidylinositol Phosphate binding domain of OGT), located in the C-terminal part, is involved in the recruitment of the OGT to the plasma membrane, allowing its interaction with insulin signaling proteins. CD1, CD2 Catalytic domains 1 and 2. ID intervening domain. B Two OGA isoforms, long (L-OGA) and short (S-OGA) are generated by alternative splicing of the OGA messenger RNA. L-OGA comprises a hexosaminidase domain in its N-terminal region and a pseudo-histone acetyltransferase domain (HAT) in its C-terminal region. S-OGA does not have the pseudo-HAT domain and seems to be addressed to the mitochondria via a 15 specific amino acids sequence located in the C-terminal region. C O-GlcNAcylation is a highly dynamic process regulated by the different OGT and OGA isoforms in different cell compartments

More than 4000 proteins have now been identified as targets for O-GlcNAcylation. A challenging question in the field has been to figure out how a single enzyme, OGT, can accurately regulate such a large panel of substrates in various biological situations.

A certain level of specificity could be achieved by spatial compartmentation of OGT. Thus, 3 different OGT isoforms, arising from mRNA alternative splicing, have been described [14]: a long, nucleocytoplasmic isoform (ncOGT), a short isoform (sOGT) also localized in the cytosol and the nucleus, and an isoform addressed to the mitochondria (mOGT). This spatial compartmentation will restrict the access of the different isoforms of OGT to subsets of specific substrates, for instance intramitochondrial proteins for mOGT [15, 16].

In addition, re-localization of OGT in the vicinity of specific subset of substrates upon hormonal stimulation has been described. Thus, the C-terminal region of OGT comprises a phosphoinsositide binding domain (denominated PPO domain for Phophosphatidylinositol Phosphate binding domain of OGT) which has been involved in the recruitment of OGT at the plasma membrane upon insulin stimulation [17] (see below).

To our knowledge, the potential involvement of PPO in the recruitment of OGT to other cellular membranes has not been investigated. Indeed, increased PIP3 production on endomembranes, including the cytosolic surface of the endoplasmic reticulum and Golgi apparatus, have been observed under different stimuli [18]. Whether the PPO of OGT could also serve for the recruitment of the enzyme to substrates localized on the surface of endomembranes constitutes an important question for future investigations.

As OGT enzymatic activity lacks a strict consensus sequence for substrate recognition, it has been proposed that part of substrate specificity is provided by structural motifs outside of the active site, through the TPR domains located in the N-terminal side of the enzyme. According to this model, different TPR may interact with different substrates, permitting the access of the substrate to the active site of the enzyme. Thus, the deletion of the 3 first TPR of ncOGT had no effect on its activity towards peptide substrates, but totally inhibited its activity towards certain proteins such as caseine kinase II and nucleoporine p62 [19], demonstrating the importance of specific TPR domains for the activity of the enzyme towards specific substrates.

An additional level of specificity may arise from the fact that OGT can also interact with its substrates through adapter molecules, which will recruit substrates to the enzyme in a cell specific and/or context dependent manner. According to this hypothesis, OGT will behave as the catalytic subunit of large transient complexes, where adapter molecules are able to target OGT to its substrates. Thus, in glucose-deprived neuroblastic Neuro-2a cells, p38 MAP kinase targets OGT to the neurofilament protein NF-H, inducing its O-GlcNAcylation and increasing its solubility [20]. In hepatocytes, under fasting conditions (see below), the host cell factor C1 (HCF-1) recruits OGT to O-GlcNAcylate peroxisome proliferator-activated receptor coactivator-1α (PGC-1α) and increase its stability [21], whereas PGC-1α itself has been shown to target OGT to FoxO1 [22].

In addition, post-translational modification of OGT itself can also affect its activity and/or substrate selectivity. Thus, in 3T3L1 adipocytes, tyrosine-phosphorylation of OGT upon insulin stimulation increases its activity, resulting in an increased O-GlcNAcylation of Stat3 [23]. In neuronal NG-10815 cells, potassium chloride-induced depolarization promotes phosphorylation of OGT by the serine-threonine kinase CaMKIV (Calcium/calmodulin-dependant protein kinase IV), resulting in an increase in OGT activity and subsequent stimulation of AP1 transcriptional activity [24]. On the other hand, in liver cells, glucagon-induced calcium signalling promotes OGT phosphorylation by CaMKII, which in turn promotes activation of autophagy in adaptation to starvation [25] (see below).

Two isoforms of OGA [26], long (L-OGA) and short (S-OGA), are also produced by alternative splicing (Fig. 1). L-OGA, more extensively studied than S-OGA, is found mainly in the cytosol and in the nucleus. It comprises the N-Acetyl-glucosaminidase domain in its N-terminal side, and, in its C-terminal side, a pseudo-Histone Acetyl-transferase domain, whose function remains poorly understood.

The less studied S-OGA has only the N-Acetyl glucosaminidase domain and a 15 amino acids C-terminal sequence specific to this isoform. Conflicting results were initially reported regarding the subcellular localization of S-OGA: this isoform was first described as localized in the nucleus of glioblastoma cells [26], then  on the surface of lipid droplets in HeLa cells [27]. However, it has been shown recently that S-OGA is addressed to the mitochondria and is involved in the production of reactive oxygen species (ROS) in this organelle [28]. This strongly supports the notion that O-GlcNAcylation plays an important role in the regulation of oxidative stress [29], at least in part through regulation of mitochondrial ROS homeostasis.

OGT and OGA can both be modified by O-GlcNAcylation, suggesting that cross-regulation can occur between the two enzymes [30]. Moreover, ncOGT and L-OGA have been shown to physically interact within a molecular complex denominated O-GlcNAczyme, and this interaction appears to play an important role in their biological functions [31, 32]. Whether such a complex also exist between mOGT and S-OGA to regulate their biological function in mitochondria remains to be determined.

The hexosamine biosynthesis pathway

The substrate used by OGT to O-GlcNAcylate proteins is UDP-N-Acetylglucosamine, produced in the hexosamine biosynthetic pathway (HBP).

A fraction (2–3%) of the glucose entering the cell is directed to the HBP (Fig. 2). After isomerisation of glucose 6-phosphate into fructose-6-Phosphate in the initial step of glycolysis, fructose-6-phosphate is converted to glucosamine-6-phosphate by the glutamine:fructose-6-phosphate amidotransferase (GFAT), the rate-limiting step of the pathway. After a series of reactions, UDP-N-acetylglucosamine (UDP-GlcNAc) is produced. OGT uses UDP-GlcNAc as a substrate to O-GlcNAcylate nuclear, cytosolic and mitochondrial proteins [12].

Fig. 2figure 2

O-GlcNAcylation depends on the energy status of the cell. The hexosamine biosynthesis pathway results in the formation of UDP-GlcNAc, which is used by OGT to O-GlcNAcylate proteins. UDP-GlcNAc is at the crossroads of different cellular metabolisms (glucose, amino acids (glutamine), fatty acids (acetyl-CoA) and nucleotides (UTP)) and therefore reflects the energy status of the cell. HK Hexokinase, GPI Glucose-6-phosphate isomerase, GFAT Glutamine-6-P amidotransferase, GNA1 Glucosamine 6-P acetyltransferase, PGM3 Phosphoacetylglucosamine mutase, UAP1 UDP-N-Acetylhexosamine pyrophosphorylase 1, OGT O-GlcNAc transferase, OGA O-GlcNAcase

GFAT, which catalyses the rate-limiting step of the pathway, is subjected to transcriptional and post-translational regulation. Two GFAT isoforms, GFAT 1 and 2, encoded by different genes (GFPT1 and GFPT2), have been described [33, 34]. GFAT1 and GFAT2 have different tissue distribution: GFAT1 mRNA are predominantly expressed in pancreas, placenta and testis, whereas GFAT2 mRNA are expressed in the central nervous system [34]. Differences in the regulation of their enzymatic activities by protein kinase A have been described [35]. However, differential regulation of GFAT1 and GFAT2 expression in different cell types and/or experimental conditions have been poorly explored. As a rate-limiting enzyme, changes in GFAT expression and/or activity are likely to affect UDP-GlcNAc levels [36] and protein O-GlcNAcylation [37].

The hexosamine biosynthetic pathway, which is closely dependent on glucose metabolism, also integrates several other metabolisms: the metabolism of glutamine, which provides the amine function, but also of acetyl-CoA and UTP (Fig. 2). OGT is therefore considered as a metabolic sensor that regulates the activity of proteins according to the energy state of the cell.

It must be noted that changes in intracellular concentrations in UDP-GlcNAc may also participate in the regulation of OGT activity and specificity. Indeed, OGT possesses three different Km for UDP-GlcNAc [38] suggesting that some proteins could be substrates for OGT at low UDP-GlcNAc concentrations, whereas other substrates would be modified only when UDP-GlcNAc concentration in the cell is above a certain threshold. This may constitute an additional mechanism by which OGT could affect the activity of different subsets of proteins, according to metabolic and environmental conditions.

In addition, it should be kept in mind that UDP-GlcNAc is also used for canonical complex glycosylation in the endoplasmic reticulum and Golgi apparatus, and therefore, changes in the activity of the HBP may also affect cellular homeostasis through alteration of complex glycosylation of proteins. Surprisingly enough, cross-talks between canonical and non-canonical glycosylations remain poorly investigated [39].

O-GlcNAcylation and regulation of glucose homeostasis

Daily nutritional transitions from fasted to fed state imply that important hormonal and metabolic regulations operate to maintain blood glucose concentration within a narrow range. Insulin and glucagon, respectively produced by β- and α-cells of the endocrine pancreas, are two key hormones involved in the regulation glucose homeostasis. In the fasted state, decreased insulin and increased glucagon concentrations in the blood promote liver glucose production, first by glycogenolysis (glycogen hydrolysis), and once glycogen stores are depleted, by gluconeogenesis (synthesis of glucose from lactate, pyruvate and amino acids). This permits endogenous glucose release by the liver under fasting conditions. Upon feeding, increase in blood glucose concentration stimulates insulin and inhibits glucagon secretion by the pancreas. Increased insulin and decreased glucagon inhibit glucose production by the liver, while insulin promotes glucose uptake, metabolism and/or storage in muscle, liver and adipose tissue.

Insulin resistance, defined as a decreased efficiency of insulin action on its target tissues, is a major feature of metabolic diseases, leading to impaired glucose clearance and chronic hyperglycaemia.

It has long been known that chronic hyperglycaemia per se has deleterious effects on various tissues, resulting in further deterioration of insulin sensitivity, impaired insulin secretion, and the development of overt diabetes. Enhanced protein O-GlcNAcylation, resulting from increased glucose availability to feed the HBP pathway, has been suggested as one of the mechanisms underlying the toxic effect of glucose [40]. Indeed, disturbances in protein O-GlcNAcylation have been observed in situations of chronic hyperglycemia, and many studies have shown a role for this modification in the phenomenon of glucotoxicity and in the development of diabetic complications [2]. Consistent with this notion, transgenic animals overexpressing OGT or GFAT in the β cells of the pancreas, the liver, or in muscles and adipose tissue, develop phenotypes reminiscent of obesity or type 2 diabetes (hyperinsulinemia, hyperleptinemia, glucose intolerance, insulin resistance) [41, 42]. Moreover, various studies have indicated that O-GlcNAcylation of different signaling intermediates, such as IRS1 and Akt, negatively regulate insulin signaling and contribute to insulin resistance [17, 43]. An important article has elucidated the mechanism by which OGT regulates proteins involved in early steps of insulin signalling [17]. Upon insulin stimulation, phosphatidyl-inositol 3-phosphate (PIP3) production by PI-3 kinase induces the recruitment of OGT, from its nuclear and cytosolic localisation to the plasma membrane. The recruitment of OGT at the plasma membrane through its PPO (Fig. 1A) then favours O-GlcNAcylation of insulin signalling proteins, resulting in attenuation of the signal. Thus, whereas this mechanism may constitute a physiological negative feed-back loop to regulate insulin signaling in the normal situation, under chronic hyperglycemic conditions, an exacerbated O-GlcNAcylation of insulin signaling intermediates would induce a decrease in insulin sensitivity, leading to the establishment of a vicious circle and the aggravation of the hyperglycemia.

In addition to altering early steps of insulin signaling, O-GlcNAcylation activates transcription factors involved in the regulation of carbohydrate metabolism. Excessive production of glucose through liver gluconeogenesis is a major cause of fasting hyperglycemia in diabetes. FoxO1, a major transcription factor involved in the regulation of genes coding for gluconeogenic enzymes, is O-GlcNAcylated and activated under high glucose condition, resulting in increased expression of various gluconeogenic genes [44, 45]. Other transcriptional regulators, including PGC-1α and the CREB regulated transcription coactivator 2 (CRTC2), were also shown to be activated by O-GlcNAcylation, thereby promoting gluconeogenesis in liver cells [22, 46]. O-GlcNAcylation of the transcription factor ChREBP also promotes its transcriptional activity, resulting in increased expression of lipogenic genes and steatosis, thereby contributing to glucolipotixicity in the diabetic liver [47].

Moreover, O-GlcNAcylation appears to be involved in several alterations associated with chronic hyperglycemia, such as cardiovascular and renal complications [2].

Thus, under conditions of hyperglycemia or high nutrient availability, exacerbated protein O-GlcNAcylation can have deleterious effects on several aspects of the regulation of energy metabolism.

However, it should be noted that there is not always a direct relationship between nutrient availability and O-GlcNAcylation level. For instance, in liver cells, HCF-1 recruits OGT to O-GlcNAcylate PGC-1α, thereby facilitating deubiquitination and stabilization of PGC1α. It was shown that O-GlcNAcylation of PGC-1α peaks at 5 mM glucose, whereas it decreased at lower and higher glucose concentrations [21]. Moreover, the same group observed that fasting induces an increase in O-GlcNAcylation and activation of Ulk proteins, promoting liver autophagy in adaptation to starvation [25]. This effect was mediated by a glucagon-induced signaling pathway involving stimulation of calcium/calmodulin-dependent kinase II, increased activity of OGT and Ulk1/2 O-GlcNAcylation, and induction of liver autophagy to maintain energy homeostasis. Another group also described a key role for increased O-GlcNAcylation upon starvation in the liver [48]. Gonzalez-Rellan et al. observed that OGT protein is markedly increased in liver of fasted mice, associated with increased O-GlcNAcylation and stabilization of p53, increased binding of p53 to the PCK1 promoter and activation of gluconeogenesis. These examples point to the complex role of O-GlcNAc in the regulation of energy metabolism during adaptation to nutritional changes (for a review, [49]).

In agreement with these observations, studies using OGT and OGA knockout mice revealed that both excess and insufficient O-GlcNAcylation can have adverse effects on the control of energy homeostasis.

OGT gene is located on chromosome X, and its invalidation is lethal at a very early stage of embryonic development [50]. Moreover, tamoxifen-inducible global OGT knockdown in adulthood induces a lethal phenotype 4 weeks after induction of the deletion, demonstrating the importance of O-GlcNAcylation for adult mouse survival [51]. To clarify the role of O-GlcNAcylation in vivo, various mouse models with tissue-specific OGT deletion have been developed. OGA is also essential for embryonic development, and its deletion leads to neonatal mortality, indicating the importance of fine-tuned regulation of O-GlcNAc level on proteins [52, 53]. As discussed below, some interesting studies have been performed on hemizygous animals with OGA haplo-insufficiency, but only few models with tissue specific OGA deletion have been developed to date [54, 55].

OGT and OGA knock-out mice models have brought important insights into the role of O-GlcNAcylation in various physiological and pathological processes. However, in the following paragraphs, we will restrict our discussion to those models that have enlighten the role of O-GlcNAcylation in the regulation of energy metabolism (Fig. 3).

Fig. 3figure 3

Lessons from OGT and OGA knock-out mice. A Tissue-specific OGT-KO mice have brought important insights into the role of O-GlcNAcylation in the regulation of energy homeostasis. B Only few tissue-specific OGA-KO mice have been developed to date, and contradictory results have been obtained for the regulation of energy homeostasis in two different models of mice with global OGA haploinsufficiency

Lessons from OGT knock-out miceOGT knock-out in the central nervous system

Since OGT is considered as a nutritional sensor, several studies have focused on the potential role of OGT in the central regulation of food intake. AgRP orexigenic neurons are highly activated during the fasting period, in order to stimulate food intake via the inhibition of anorexigenic neurons [56]. Ghrelin is a hormone released upon fasting by the empty stomach that stimulates hunger by activating AgRP neurons in the hypothalamus. Ruan et al. observed that OGT expression as well as protein O-GlcNAcylation levels are increased in AgRP neurons during fasting or upon Ghrelin injection [57], providing an additional example of a situation where protein O-GlcNAcylation and nutrient availability are not correlated. These authors developed mice with selective OGT knockout in these neurons. OGT knockout increases energy expenditure by inhibiting the suppressive activity of these neurons on the browning of white adipose tissue. Moreover, AgRP neurons are known to be involved in the central regulation of hepatic gluconeogenesis, and AgRP neurons-OGT KO mice have decreased hepatic expression of gluconeogenic genes, and a reduction of blood glucose levels during pyruvate and glucose tolerance tests. Consistent with the activated thermogenic program, these mice are protected against high fat diet-induced obesity and insulin resistance. Altogether, this study revealed that in AgRP neurons, OGT has a critical role in suppressing thermogenesis in response to fasting, allowing energy sparing when food intake is reduced, while promoting glucose production through liver gluconeogenesis [57].

Lagerlof et al. investigated the effect of tamoxifen-inducible OGT deletion in αCaMKII-expressing neurons [58]. These neurons are mainly located in the paraventricular nucleus (PVN). In this nucleus, food intake activates αCaMKII neurons. Loss of OGT in these neurons completely blocks their activation in response to food intake. Mice knocked out for OGT in these neurons develop hyperphagia leading to increased fat mass and rapid weight gain compared to control mice. This result suggested that in PVN, OGT has a role in the regulation of satiety [58]. Another group [59] independently developed a very similar mice model and confirmed that OGT-KO in αCaMKII-expressing neurons resulted in increased feeding behaviour and obesity within 4 weeks following induction of OGT deletion, associated with higher insulin and leptin levels, insulin resistance and neuronal loss in the hypothalamus. Interestingly, by investigating these mice on a longer period of time, they observed that the insulin resistance reversed 2–3 month after induction of the deletion, and that the mice surprisingly displayed increased insulin sensitivity 9.5 months after neuronal OGT-KO.

Together, these studies [57,58,

留言 (0)

沒有登入
gif