Metabolic Inhibitors of O‐GlcNAc Transferase That Act In Vivo Implicate Decreased O‐GlcNAc Levels in Leptin‐Mediated Nutrient Sensing

Abstract O‐Linked glycosylation of serine and threonine residues of nucleocytoplasmic proteins with N‐acetylglucosamine (O‐GlcNAc) residues is catalyzed by O‐GlcNAc transferase (OGT). O‐GlcNAc is conserved within mammals and is implicated in a wide range of physiological processes. Herein, we describe metabolic precursor inhibitors of OGT suitable for use both in cells and in vivo in mice. These 5‐thiosugar analogues of N‐acetylglucosamine are assimilated through a convergent metabolic pathway, most likely involving N‐acetylglucosamine‐6‐phosphate de‐N‐acetylase (NAGA), to generate a common OGT inhibitor within cells. We show that of these inhibitors, 2‐deoxy‐2‐N‐hexanamide‐5‐thio‐d‐glucopyranoside (5SGlcNHex) acts in vivo to induce dose‐ and time‐dependent decreases in O‐GlcNAc levels in various tissues. Decreased O‐GlcNAc correlates, both in vitro within adipocytes and in vivo within mice, with lower levels of the transcription factor Sp1 and the satiety‐inducing hormone leptin, thus revealing a link between decreased O‐GlcNAc levels and nutrient sensing in peripheral tissues of mammals.

The dynamic modification of nucleocytoplasmic proteins with b-O-linked N-acetylglucosamine (O-GlcNAc) is present in all multicellular eukaryotes. [1] Thea ttachment of GlcNAc (1, Figure 1) to hydroxy groups of serine or threonine residues of hundreds of target proteins is catalyzed by the glycosyltransferase O-GlcNAc transferase (OGT), which uses uridine diphospho-N-acetylglucosamine (UDP-GlcNAc, 2)a s ad onor sugar substrate.T he glycosidase O-GlcNAcase (OGA) cleaves O-GlcNAc off from proteins.
O-GlcNAc is implicated in various diseases including cancer, neurodegeneration, cardiovascular disease,a nd obesity. [2] Notably,O -GlcNAc levels respond to nutrient availability both within cells and in vivo in animal models. [1b] Tr ansgenic mice overexpressing OGT in fat or muscle tissue exhibit elevated serum leptin and insulin levels in addition to insulin resistance. [3] Furthermore,d eletion of the gene encoding OGT from neurons of the paraventricular nucleus (PVN) within the hypothalamus of mice results in uncontrolled eating. [2b] Notably,O GA inhibitors that are active in vivo have helped gain insight into the roles of increased O-GlcNAc levels in mammals.S trikingly,O GA inhibitors sometimes Figure 1. The hexosamine biosynthetic pathway (HBP) and the GlcNAc salvage pathway yield UDP-GlcNAc ( 2). Glutamine fructose amidotransferase( GFAT) catalyzes the conversion of glucosamine(5)i nto glucosamine-6-phosphate (GlcNH 2 -6PO 4 ; 5-6PO 4 ). Acetylation of 5-6PO 4 by glucosamine acetyltransferase (GAT) yields GlcNAc-6PO 4 , an intermediate also produced by GlcNAc kinase (GNK) in the GlcNAc salvage pathway.The sequential action of phosphoglucosamine mutase (AGM) and GlcNAc pyrophosphorylase (AGX) on GlcNAc-6PO 4 leads to the formation of UDP-GlcNAc(2). 5SGlcNAc (3)can be similarly salvaged and converted into UDP-5SGlcNAc (4). yield different results from those made using genetic approaches to increase O-GlcNAc levels, [4] perhaps because OGT and OGA also have non-catalytic roles.Unfortunately, similar studies regarding the influence of decreased O-GlcNAc on mammalian physiology are lacking because there are no OGT inhibitors suitable for use in vivo.Given the emerging roles of OGT in nutrient sensing and other processes,i nhibitors of OGT that can be used as research tools in vivo are of high interest. [5] One approach to decreasing O-GlcNAc levels using small molecules has been to use broad-spectrum amidotransferase inhibitors such as 6-diazo-5-oxo-l-norleucine (DON), which promiscuously blocks all amidotransferases that biosynthesize many cellular metabolites including UDP-GlcNAc. Highthroughput screening has been pursued to deliver hits that can be improved upon. [5b,d] These leads,h owever, show modest cellular activity and limited solubility. [5b] They also exhibit offtarget cellular toxicity. [5b] Another approach [5c] has been to generate aGlcNAc analogue,2-acetamido-2-deoxy-5-thio-ad-glucopyranose (5SGlcNAc, 3), which in its per-O-acetylated form (Ac 4 5SGlcNAc, 3-OAc) can diffuse across the plasma membrane.W ithin cells, 3-OAc is de-O-acetylated and assimilated via the GlcNAc salvage pathway ( Figure 1) to generate UDP-5SGlcNAc (4), which is ac ompetitive OGT inhibitor (K i = 8 mm). 3-OAc lowers cellular O-GlcNAc levels (EC 50 = 0.8-5 mm). This metabolic OGT inhibitor is not toxic but suffers from poor solubility in aqueous solution. Indeed, while often used in cell studies, [6] solubilizing 3-OAc requires high concentrations of DMSO,m aking it incompatible for dosing of mammals.A ccordingly,t oe xplore the roles of decreased O-GlcNAc levels in organismal physiology,c ompounds that act in vivo are of high interest.
To generate atool compound for inhibiting OGT in vivo, we synthesized ap anel of water-soluble analogues of 5SGlcNAc (3)p ossessing various N-acyl substituents (7-17; Figure 2a,b and Scheme S1 in the Supporting Information).
We reasoned that these hydrophobic N-acyl groups would confer ab alance between hydrophilicity and lipophilicity, making them water-soluble yet able to diffuse into cells. Furthermore,a lthough the substrate specificity of N-acetylglucosamine-6-phosphate de-N-acetylase (NAGA) is not known, [7] we speculated that compounds 7-17,o nce phosphorylated, might be substrates for this recently identified enzyme ( Figure 2b). Action of NAGA on phosphorylated 7-17 would lead to formation of ac ommon intermediate, 5-thioglucosamine-6-phosphate (5-6PO 4 ), which can be assimilated by the hexosamine biosynthetic pathway (HBP). However,d irect entry of 7-17 into the GlcNAc salvage pathway cannot be ruled out as the specificity of NAGA is unknown, and bulkier N-glycolyl (GlcNGc)-and N-azidoacetyl (GlcNAz)-containing analogues enter the HBP. [8] To measure the potential of 7-17 and their per-Oacetylated congeners to block OGT activity in cells,w e treated cells with these compounds.A nalysis of O-GlcNAc levels of cell lysates showed that only some deprotected analogues were effective in reducing O-GlcNAc levels (Figure 2c and Figure S1). Of these,the promising 2-hexanamide derivative (5SGlcNHex, 12)s howed dose-dependent decreases in O-GlcNAc comparable to those achieved with Ac 4 5SGlcNAc (3-OAc;F igure S2).
We next tested whether these compounds were directly activated as UDP-linked analogues or if they were metabolized by NAGA. We therefore analyzed the pool of nucleotide sugars from cells treated with compounds 7, 8,a nd 12 by capillary electrophoresis (CE;F igure 3). Electropherograms for nucleotide sugars from cells treated with 7, 8,a nd 12 revealed two new peaks with CE mobilities matching those of the chemoenzymatically prepared standards UDP-5SGlcNAc (4)a nd its epimer UDP-5SGalNAc (Figure 3a). Extracts from cells treated with per-O-acetylated 5SGlcNH 2 (6-OAc, Figure S3) also yielded the same two unnatural nucleotide sugars in cells.T hese data suggest that these 5SGlcNR analogues are processed within cells to the common intermediate 2-amino-2-deoxy-5-thioglucopyranose 6-phosphate (6-PO 4 )a nd then assimilated by the HBP to form UDP-5SGlcNAc (4).
To clarify the metabolic processing of compounds 7, 8,and 12,w eu sed recombinant enzymes of the mammalian HBP ( Figure 1). Compounds 7, 8,a nd 12 were all phosphorylated by GNK (Figure 3b), yet none of the phosphorylated products were substrates for AGMe xcept for 7.H owever, the nucleotide sugar produced in vitro from 7 by the combined HBP enzymes had ad ifferent mobility than nucleotide sugar UDP-5SGlcNAc (4), which we detected in cells treated with compound 7 ( Figure S4). We accordingly tested whether NAGA converted the corresponding 5-thiosugar-6-PO 4 derivatives of 7, 8,a nd 12 into 5SGlcNH 2 -6-PO 4 (6-6PO 4 ). We confirmed this scenario by digestion of 12-PO 4 with recombinant human NAGA. Electropherograms obtained by CE analysis of the processing of 12-PO 4 by NAGA showed that this compound was converted into 5SGlcNH 2 -6-PO 4 (6-PO 4 ;F igure 3c). These data suggest that the 5SGlcNR analogues 7-17,i ncluding 7, 8,a nd the most potent derivative 5SGlcNHex (12), are phosphorylated within cells by GNK and then deacylatedb yN AGA, which we found to display remarkable substrate tolerance.T he resulting common intermediate,5SGlcNH 2 -6-PO 4 (6-PO 4 ), is then assimilated via the HBP to form UDP-5SGlcNAc (4), which leads to metabolic inhibition of OGT and decreased O-GlcNAc levels in tissues.
Previous efforts to use Ac 4 5SGlcNAc (3-OAc) in vivo in mammals failed because of its poor aqueous solubility.W e therefore evaluated whether these new water-soluble metabolic OGT inhibitors could be used in vivo.A ccordingly,w e dosed mice by intraperitoneal (IP) delivery with our most cell-active compound 12.Aconcentration-dependent decrease in spleen O-GlcNAc levels was observed upon treatment with 12,w ith apparent effects at even 3mgkg À1 (Figure S5 a). Compound 12 (300 mg kg À1 )d ecreased the spleen O-GlcNAc levels over time with maximal inhibition by 8h,w hich was maintained for at least 48 h ( Figure S5 b). Compound 12 reduced O-GlcNAc levels in the kidneys,lungs, fat, pancreas,h eart, spleen, and muscle tissue but not in the blood or brain (Figure S5 c). Mice injected with high doses of 12 (300 mg kg À1 )b ecame lethargic,w hich is ak nown consequence of low leptin levels. [9] As econd injection of 300 mg kg À1 on as econd day caused mice to be moribund, and we discontinued treatment. We therefore dosed mice IP with alower dose of 12 (50 mg kg À1 , n = 3) and found reduced O-GlcNAc levels in various tissues 16 ha fter dosing but no effects on brain, liver,pancreas,and kidney (Figure 4a,b and Figure S6 a-c). TheO-GlcNAc levels had returned to baseline levels after 16 h (Figure 4a). Metabolic inhibitor 12 therefore acts in vivo to reversibly lower O-GlcNAcylationi nv arious tissue types.
Notably,o thers have shown that Ac 4 5SGlcNAc (3-OAc) appears to be able to inhibit other glycosyltransferases in cell lines. [5b] We found, however, by lectin blot assessment that 12 induced no apparent changes in other forms of protein Figure 3. 5SGlcNR analoguesa re sequentially processed in cells by GlcNAc kinase (GNK) and, likely, N-acetylglucosamine-6-phosphate-de-N-acetylase (NAGA) to generate 5SGlcNH 2 -6PO 4 (6-PO 4 ). a) Nucleotide sugar analysis of cells treated with selected inhibitors indicated that all of the 5SGlcNR analoguest ested (3-OAc, 7, 8,and 12)w ere converted within cells into two new nucleotide sugars, indistinguishable from synthetically prepared UDP-5SGlcNAc (4)a nd UDP-5SGal-NAc. b) 5SGlcNR analoguesw ere phosphorylated by GNK in vitro. c) 5SGlcNR-6-phosphates were not substrates for AGM or AGX but were readily hydrolyzed in vitro by NAGA to produce 5SGlcNH2-6PO 4 (6-PO 4 ). . Each independent sample was tested in duplicate; symbols denote statistical significance (*p < 0.05, **p < 0.01, ***p < 0.001 compared to control;Student's t-test). glycosylation in all tissues tested ( Figure S7) even at adose of 300 mg kg À1 after 48 h, which is consistent with arecent report on the use of the parent compound Ac 4 5SGlcNAc (3-OAc) in cells. [10] Curiously,w eo bserved that mice treated with either low or high doses (50-300 mg kg À1 )o f12 exhibited engorged stomachs that were full of rodent chow after 16 h( Figure S8). As lethargy and excessive food consumption (hyperphagia) are seen in mice deficient in leptin signaling, [11] we measured the serum leptin concentration of animals treated with 50 mg kg À1 of 12 (Figure 4b). Their leptin levels dropped transiently to aminimum at 16 h (Figure 4b), consistent with the engorged stomachs. [9,12] Theleptin levels were also lower in mice treated with ah igh dose of 12 (300 mg kg À1 ;F igure S9). We therefore hypothesized that OGT inhibition might impair leptin production ( Figure S10).
To test this hypothesis,w eu sed 3T3-L1 adipocytes and found that these cells secreted less leptin when treated with either 12 (Figure 4d)o rA c 4 5SGlcNAc (3-OAc;F igure S11) without affecting glucose uptake.N otably,t he transcription factors C/EBP-a,-b,a nd Sp1 all regulate leptin production. Compound 12 lowered the levels of these known O-Glc-NAcylated proteins in adipocytes (Figures S9) and in mice ( Figure S6 d). These collective data are consistent with OGT acting as anutrient sensor in aprocess coupling HBP flux to leptin levels ( Figure S10).
We envision that future studies to detail dosing regimens with 12 will be important to explore the effects of decreased O-GlcNAc levels on various physiological processes.Experiments will also be needed to detail the mechanistic links between reduced O-GlcNAc levels due to in vivo OGT inhibition, impaired leptin secretion, and apparent hyperphagia. However,our data provide the first direct correlation between decreased O-GlcNAc levels and impaired leptin production in vivo.O ur findings are in accord with studies showing that reduced O-GlcNAc lowers levels of Sp1 in cells [6f, 13] and that OGT plays ap ivitol role in fat tissues. [14] Notably,t hese data also suggest that regulation of leptin by O-GlcNAc is bidirectional in vivo as overexpression of OGT, [3] knockout of OGA, [15] and metabolic upregulation of UDP GlcNAc levels [16] all increase leptin expression.
In summary,w eh ave described convenient new compounds to inhibit OGT in cells and in vivo.S trikingly, inhibition of OGT with 5SGlcNHex (12)p rovides support for the hypothesis that reduced O-GlcNAc levels signal impaired nutrient supply in mammals.W ee xpect that these observations will stimulate activity in the development of additional metabolic OGT inhibitors.Moreover,weenvision that this chemical strategy of metabolic OGT inhibition will serve as av aluable complement to using genetic methods in various rodent models to accelerate studies to understand the in vivo roles of O-GlcNAc in mammals.

Experimental Section
Please see the Supporting Informationf or full experimental procedures.All experiments carried out on animals were approved by the University Animal Care Committee of SFU.