Cross-talk between Two Essential Nutrient-sensitive Enzymes

Background: OGT and AMPK collectively target hundreds of intracellular signaling processes, but no study has addressed whether they regulate each other. Results: AMPK activity mediates the substrate selectivity of OGT, and O-GlcNAcylation modulates the activity of AMPK. Conclusion: There is significant cross-talk between the O-GlcNAc and AMPK systems. Significance: OGT and AMPK may synergistically regulate numerous nutrient-sensitive processes essential for life. Nutrient-sensitive pathways regulate both O-GlcNAc transferase (OGT) and AMP-activated protein kinase (AMPK), cooperatively connecting metabolic homeostasis to regulation of numerous intracellular processes essential for life. Similar to phosphorylation, catalyzed by kinases such as AMPK, O-GlcNAcylation is a highly dynamic Ser/Thr-specific post-translational modification of nuclear, cytoplasmic, and mitochondrial proteins catalyzed exclusively by OGT. OGT and AMPK target a multitude of intracellular proteins, with the net effect to protect cells from the damaging effects of metabolic stress. Despite hundreds of studies demonstrating significant overlap in upstream and downstream signaling processes, no study has investigated if OGT and AMPK can directly regulate each other. We show acute activation of AMPK alters the substrate selectivity of OGT in several cell lines and nuclear localization of OGT in C2C12 skeletal muscle myotubes. Nuclear localization of OGT affects O-GlcNAcylation of numerous nuclear proteins and acetylation of Lys-9 on histone 3 in myotubes. AMPK phosphorylates Thr-444 on OGT in vitro; phosphorylation of Thr-444 is tightly associated with AMPK activity and nuclear localization of OGT in myotubes, and phospho-mimetic T444E-OGT exhibits altered substrate selectivity. Conversely, the α- and γ-subunits of AMPK are O-GlcNAcylated, O-GlcNAcylation of the γ1-subunit increases with AMPK activity, and acute inhibition of O-GlcNAc cycling disrupts activation of AMPK. We have demonstrated significant cross-talk between the O-GlcNAc and AMPK systems, suggesting OGT and AMPK may cooperatively regulate nutrient-sensitive intracellular processes that mediate cellular metabolism, growth, proliferation, and/or tissue function.


Nutrient-sensitive pathways regulate both O-GlcNAc transferase (OGT) and AMP-activated protein kinase (AMPK), cooperatively connecting metabolic homeostasis to regulation of numerous intracellular processes essential for life. Similar to phosphorylation, catalyzed by kinases such as AMPK, O-GlcNAcylation is a highly dynamic Ser/Thr-specific post-translational modification of nuclear, cytoplasmic, and mitochondrial proteins catalyzed exclusively by OGT. OGT and AMPK target a multitude of intracellular proteins, with the net effect to protect cells from the damaging effects of metabolic stress.
Despite hundreds of studies demonstrating significant overlap in upstream and downstream signaling processes, no study has investigated if OGT and AMPK can directly regulate each other. We show acute activation of AMPK alters the substrate selectivity of OGT in several cell lines and nuclear localization of OGT in C2C12 skeletal muscle myotubes. Nuclear localization of OGT affects O-GlcNAcylation of numerous nuclear proteins and acetylation of Lys-9 on histone 3 in myotubes. AMPK phosphorylates Thr-444 on OGT in vitro; phosphorylation of Thr-444 is tightly associated with AMPK activity and nuclear localization of OGT in myotubes, and phospho-mimetic T444E-OGT exhibits altered substrate selectivity. Conversely, the ␣and ␥-subunits of AMPK are O-GlcNAcylated, O-GlcNAcylation of the ␥1-subunit increases with AMPK activity, and acute inhibition of O-GlcNAc cycling disrupts activation of AMPK. We have demonstrated significant cross-talk between the O-GlcNAc and AMPK systems, suggesting OGT and AMPK may cooperatively regulate nutrient-sensitive intracellular processes that mediate cellular metabolism, growth, proliferation, and/or tissue function.
O-linked ␤-N-acetylglucosamine (O-GlcNAc) 2 is a highly dynamic, inducible, and reversible post-translational modification of serine and threonine residues found on over 2500 intracellular proteins involved in the regulation of virtually all core processes associated with cellular metabolism, growth, proliferation, and function. O-GlcNAc is essential for life and dysregulated O-GlcNAc cycling is involved in the pathogenesis of disorders associated with the metabolic syndrome (e.g. cardiovascular disease and type 2 diabetes) and certain types of cancer (1,2). O-GlcNAc cycles on and off Ser/Thr residues by the concerted actions of two enzymes: O-GlcNAc transferase (OGT) and O-GlcNAcase (OGA), respectively. O-GlcNAcylation is a nutrient-sensitive process heavily influenced by flux through the hexosamine biosynthetic pathway (HBP). De novo synthesis of uridine diphosphate N-acetylglucosamine (UDP-GlcNAc), the high-energy donor substrate for OGT, is heavily dependent on glucose flux through the HBP. Glucose, glutamine, acetyl-CoA, and UTP are all necessary intermediates in UDP-GlcNAc synthesis, suggesting that flux into the HBP may be a mechanism by which intracellular nutrient status is linked to O-GlcNAc-specific regulation of hundreds of signaling processes (1,2).
Similar to O-GlcNAcylation, regulation of AMP-activated protein kinase (AMPK) is a vital nutrient-sensing process that mediates cellular and whole body energy homeostasis via dozens or hundreds of protein targets and/or signaling pathways (3). AMPK is a heterotrimeric complex comprising an ␣ (catalytic), and ␤ and ␥ (regulatory) subunits, with the ␥-subunit binding the adenine nucleotides AMP, ADP, and ATP. Each subunit is encoded by multiple genes (␣1, ␣2, ␤1, ␤2, ␥1, ␥2, ␥3) generating at least 12 heterotrimeric combinations. AMPK is activated by 5Ј-AMP and ADP but inhibited by ATP, making it extremely sensitive to intracellular AMP:ATP and ADP:ATP ratios, and thus activated when intracellular energy levels are depleted. AMPK is also responsive to multiple extracellular nutritional and hormonal signals that regulate food intake, energy expenditure, proliferation, and cell viability (3). AMP promotes AMPK activity by two distinct mechanisms, of which the first is also caused by ADP: 1) by promoting increased net phosphorylation of a critical threonine residue (Thr-172) located in the activation loop of the ␣-subunit, and 2) by allosteric activation of the phosphorylated kinase. In general, activation of AMPK promotes catabolic and inhibits anabolic processes, with net effects to replenish or conserve intracellular ATP. Depending on tissue/cell type, AMPK promotes glucose uptake, glycolysis, and mitochondrial biogenesis, while inhibiting gluconeogenesis, and fatty acid (FA), triglyceride, cholesterol, RNA, and protein synthesis (3). AMPK has thus emerged a strong drug target for disorders associated with the metabolic syndrome and certain types of cancer.
OGT and AMPK collectively target thousands of downstream proteins that vary widely based on biological or experimental parameters and tissue/cell type. A better understanding of non-canonical pathways that mediate contextually dependent spatial, temporal, and/or substrate specific activity of OGT and AMPK is needed. Collectively, data suggest significant overlap in OGT and AMPK signaling processes: First, many metabolic stresses that regulate OGT also regulate AMPK (e.g. glucose deprivation, ER stress, and H 2 O 2 -induced mitochondrial stress). Second, regulation of both OGT and AMPK is associated with protection of cells from the damaging effects of these metabolic stresses. Last, many downstream targets of AMPK are also O-GlcNAcylated; e.g. CREB-regulated transcription coactivator 2 (CRTC2 or TORC2), cellular tumor antigen (p53), and peroxisome proliferator-activated receptor ␥ coactivator 1-␣ (PGC1-␣) (4 -6). Despite thousands of studies on O-GlcNAcylation and AMPK, no study has interrogated the potential for direct dynamic cross-talk between OGT and AMPK. Here, we report that activation of AMPK alters the substrate selectivity of OGT in numerous cell lines and nuclear localization of OGT in differentiated skeletal muscle myotubes, affecting global O-GlcNAcylation of nuclear proteins and acetylation of Lys-9 (K9) on histone 3 in myotubes. AMPK phosphorylates Thr-444 (T444) on OGT in vitro; phosphorylation of Thr-444 is tightly associated with AMPK activity and increased nuclear localization of OGT in myotubes, and phospho-mimetic Thr-444 to Glu (T444E) mutant OGT exhibits altered substrate selectivity. Conversely, both ␣and all 3 ␥-subunits of AMPK are O-GlcNAcylated, O-GlcNAcylation of the ␥1-subunit directly correlates with AMPK activity, and acute inhibition of O-GlcNAc cycling blunts activation of AMPK by physiological and pharmacological stimuli. Collectively, our results demonstrate significant cross-talk between the O-GlcNAc and AMPK systems, suggesting OGT and AMPK may cooperatively regulate nutrient-sensitive intracellular processes essential for life and the pathogenesis of metabolic diseases.
Confocal Microscopy-Proliferating cells were plated on glass coverslips coated with poly-D-lysine (Sigma). To image differentiated C2C12 myotubes, C2C12 cells were plated and differentiated on Thermanox plastic coverslips (Thermo Scientific). Cells were washed, fixed in 4% paraformaldehyde, and processed as published (13). Permeabilized cells were blocked in 5% BSA or 10% goat serum, incubated with antibodies specific for OGT (AL28), OGA (345) (14), H3 K9Ac, and/or H3 K27me3 (Abcam), incubated with the appropriate secondary antibodies (anti-IgG 405-, 488-, 555-, or 647-conjugated Alexa Fluors; Invitrogen) and mounted. For DNA staining, cells were briefly incubated in DAPI (1 g/ml) before being mounted. Images were captured on a laser scanning Zeiss LSM510-Meta confocal microscope maintained by the Johns Hopkins University microscope core facility. Quantification was performed double blinded (i.e. all images were captured and analyzed using coded samples). Confocal images were acquired as a z-stack of five 1-m slices (spanning a 2.8-m range). Projections are one representative 1-m slice. Quantification was performed on the most optimal z-stack for each nucleus. For nuc OGT / cyto OGT measurements, ROI units (2 m) were placed in the nucleus and immediately adjacent cytoplasm and quantified (ImageJ ROI analysis) to calculate the nuclear-to-cytoplasmic ratio of OGT immunofluorescence. For all quantified data of histone modifications, images were captured in one session using identical microscope settings and nuclear H3 K9Ac or K27me3 immunofluorescence pixel intensities were quantified using ImageJ ROI analyses. Nuc OGT /Cyto OGT distribution curves were generated by counting the number of nuclei exhibiting nuclear-to-cytoplasmic ratios of OGT immunofluorescence that fall within incremental data subsets of 0.2 ranging the maximal dynamic range of the microscope, from 0 -4 (e.g. 0 -0.2, 0.21-0.4, etc). The midpoint for each data subset (e.g. 0.5 for subset 0.41-0.6) was plotted against the percentage of nuclei exhibiting ratios within the respective data subset and displayed as a smoothed line plot. At least five images were captured/sample for all experiments.

AMPK Activity Assays and in Vitro Cold-or [␥ 32 ]Phosphate/ [ 3 H]GlcNAc
Labeling-AMPK activity assays were performed on -␣1 or -␣2 immunoprecipitates as previously described (8,15), and data are presented as fold differences in mean specific activity (nanomol of ␥-32 P incorporated into SAMS peptide/ min) normalized to respective controls Ϯ S.E. For in vitro [␥ 32 ]phosphate or cold ATP labeling, 1 or 500 g of ncOGT was incubated with 1 or 200 g of respective AMPK complexes in the presence of 40 M AMP, 5 mM MgCl 2 , and 3 Ci protein samples were processed and immunoblotted as described above. For in vitro [ 3 H]GlcNAc-labeling reactions, 5 g of each respective AMPK complex was incubated with 1 g of ncOGT, 2 Ci of UDP[ 3 H]GlcNAc, and 0.5 units of Calf intestinal alkaline phosphatase (CIP) in 20 l of 50 mM HEPES (pH 7.4) overnight at 4°C on a rotator. Protein samples were subsequently processed as described (16).
Mass Spectrometry-Similar to above, ncOGT and AMPK complexes were incubated with ATP, separated by SDS-PAGE, and gel bands corresponding to the molecular weight of ncOGT were excised for analysis by mass spectrometry. Digestion conditions were similar to those described previously (19). Briefly, excised bands were pH equilibrated in 100 mM NH 4 HCO 3 , incubated in 10 mM DTT, 55 mM iodoacetamide, 100 mM NH 4 HCO 3 to reduce/carbamidomethylate Cys residues, and digested in 12 ng/l trypsin (Promega) overnight to obtain peptides amenable to mass analysis. Tryptic peptides were extracted from gel slices using alternating washes of 100 mM NH 4 HCO 3 and 50% acetonitrile/5% formic acid, followed by a final dehydration in 100% acetonitrile. Tryptic peptides were dried to completion using a Speedvac Concentrator (Savant) and reconstituted in 0.1% acetic acid in water for mass analysis.
For mass spectrometric analyses, a portion of the digest was pressure loaded onto a precolumn (360 m o.d. ϫ 75 m i.d.) packed with C18 reverse-phase resin (5-20 m diameter, 120 Å pore diameter, YMC Co., LTD), desalted with 0.1% acetic acid in water, and connected to an analytical column packed with C18 reverse-phase resin (5-m diameter, 120 Å pore diameter, YMC Co., LTD) equipped with an integrated electrospray emitter tip (20). Mass analyses were completed on an in-house modified, electron transfer dissociation (ETD)-enabled LTQ-Orbitrap and LTQ-FT-ICR-MS (Thermo Scientific). High resolution MS scans were acquired in the Orbitrap or FT-ICR cell whereas low resolution, data-dependent collision-activated dissociation (CAD) and ETD MS/MS data were acquired in the ion trap using previously described acquisition methods, HPLC gradient and instrument parameters (21). Data analysis was aided using the Open Mass Spectrometry Search Algorithm (OMSSA) (version 2.1.1) to search against a constructed ncOGT (gi 68067509, Accession O15294, NCBI) database using the following variable modifications: oxidation of Met, phosphorylation of Ser, Thr, Tyr, and carbamidomethylation of Cys (22). The remaining OMSSA search parameters have been described previously (21). Peptide MS/MS spectral data pertaining to ncOGT was manually validated by visual inspection of pertinent CAD and ETD spectra. The surface representation of the crystal structure of hOGT 4.5 complexed with UDP and CKII3K peptide (Protein DataBank ID: 3PE4) (23) was constructed using MacPyMOL Molecular Graphics System (DeLano Scientific).
Phospho-Thr-444 OGT Custom Antibody Design and Purification-Antibody specific for the phosphorylated Thr-444 residue on OGT was custom designed, synthesized, and provided by Pacific Immunology (Ramona, CA). Briefly, naked [C-IPEAIASYR-T-ALKLKPD] and phospho-[C-IPEAIASYR-T(PO3)-ALKLKPD] peptides were synthesized and HPLC purified. Phospho-peptide was coupled to KLH and used as an antigen in 2 separate rabbits. Antibody purification was per-formed on bleed 11, as previously described (24), with minor modifications. Six 1-ml aliquots of serum were diluted up to 30 ml in 10 mM Tris (pH 7.5), and purified 2ϫ over naked peptide coupled to a Sulfo-link agarose bead column (Thermo Scientific; provided by Pacific Immunology). Flow-through was collected and stored on ice. The naked-peptide column was then washed with 20ϫ bed volumes of 10 mM Tris (pH 7.5) plus 500 mM NaCl, 10ϫ bed volumes of 100 mM glycine (pH 2.5), 15ϫ bed volumes of 10 mM Tris (pH 8.8), 10ϫ bed volumes of 100 mM triethylamine (pH 11.5), and 20 -30ϫ bed volumes of 10 mM Tris (pH 7.5). Flow-through was subsequently purified over the naked-peptide column for a third time. Flow-through from this last purification was then purified 2ϫ over the phosphopeptide Sulfo-link column (Thermo Scientific; provided by Pacific Immunology). The phospho-peptide column was subsequently processed as published (24), and antibody aliquots were stored at Ϫ20°C.
Statistical Analyses-Quantification of immunoblotted protein band pixel intensities (Ϯ S.E.) and densitometric lane profiles were generated using ImageJ software. All data are representative of at least three separate experiments (n ϭ 1 or 2/treatment/experiment), unless otherwise noted. Statistical significance was assessed using ANOVA and standard Student t-tests.

Nuclear Localization of OGT Is Tightly Associated with AMPK Activity in Differentiated C2C12 Mouse Skeletal Muscle
Myotubes-To address the potential of direct signaling between the O-GlcNAc and AMPK systems, we investigated whether acute physiological or pharmacological metabolic stresses known to regulate AMPK also affect protein O-GlcNAcylation. We examined the effect of 2 h of glucose deprivation, nutrient/growth factor deprivation (absence of all nutrients except 25 mM glucose (i.e. no serum, vitamins, amino acids, sodium pyruvate, or L-glutamine)) and AICAr treatment (an AMP-mimetic and routinely used activator of AMPK) on the localization and activity of OGT, OGA, and AMPK in differentiated C2C12 myotubes, the undifferentiated myoblastic cell substrata on which myotubes fuse, and various proliferating cell lines (e.g. Hek293A, HepG2, HeLa, and C2C12). Consistent with previous studies (13,14), in undifferentiated myoblastic cells and all proliferating cell lines studied OGT is primarily localized in the nucleus while OGA is primarily localized in the cytoplasm (Fig. 1, A and B). In contrast, there is a significant increase in cytoplasmic localization of OGT in differentiated myotubes, while OGA remains primarily cytoplasmic (Fig. 1C). These data are consistent with what is observed in other terminally differentiated cell types (e.g. rat cerebellar cortex, mouse 3T3-L1 adipocytes, and rat cardiac muscle) (25)(26)(27). Collectively, our data and previous work suggest that there are distinct functional roles for OGT in differentiated versus proliferating cells.
Under basal conditions, OGT exhibited a broad bell curveshaped percent distribution of nuclear-to-cytoplasmic ratios of OGT immunofluorescence (nuc OGT /cyto OGT ) in myotubes. This was due to the high degree of variability in nuclear local- . E, lysates from differentiated C2C12 cells exposed to the same conditions in parallel were immunoblotted (WB) as indicated. Mean Nuc OGT /Cyto OGT and densitometric values of phospho-over total AMPK and ACC (ϮS.E.) are normalized to Ctrl. #, ##, and * denote statistical significance of p Ͻ 0.01, p Ͻ 0.0001, and p Ͻ 1 ϫ 10 Ϫ28 versus Ctrl. and Veh., respectively. †, † †, and ‡ denote statistical significance of p Ͻ 0.01, p Ͻ 0.001, and p Ͻ 1 ϫ 10 Ϫ37 versus NGF.D., respectively.
ization of OGT both between myotubes and within the same myotube ( Fig. 2A). In general, there was relatively equal nuclear and cytoplasmic localization of OGT, with a mean nuc OGT / cyto OGT ratio of 1.06 and over 90% of the nuclei exhibiting nuc OGT /cyto OGT ratios between 0.5 and 1.5 (Fig. 2, B--D). Two hours of glucose deprivation, a mild physiological activator of AMPK, increased nuclear localization of OGT. Treatment with AICAr, a more specific pharmacological activator of AMPK (15), induced a dramatic increase in nuclear localization of OGT, with 75% of the nuclei exhibiting nuc OGT /cyto OGT ratios greater than 1.5 (Fig. 2, B-D). An effect similar to what was observed with glucose deprivation was observed after just 30 min of AICAr treatment (mean nuc OGT /cyto OGT value of 1.46 Ϯ 0.53, p Ͻ 1 ϫ 10 Ϫ11 ). In contrast, nutrient/growth factor deprivation (in the presence of 25 mM glucose) induced a dramatic decrease in nuclear localization of OGT, with 70% of the nuclei exhibiting nuc OGT /cyto OGT ratios less than 0.5 (Fig. 2,  B-D). AICAr rescued nutrient/growth factor deprivation-in-duced decreases in nuclear localization of OGT (Fig. 2, B-D). AMPK activity in lysate from differentiated C2C12 cells treated in parallel directly correlated with nuclear localization of OGT in myotubes (assessed by immunoblotting for phospho-Thr-172 AMPK-␣ and phospho-Ser-79 on acetyl-CoA carboxylase (ACC, a ubiquitous downstream target of AMPK)) (Fig. 2E). Importantly, activation of AMPK had no effect on the enzymatic activity of OGT, the localization of OGT in proliferating cells and undifferentiated myoblastic cells, or the localization of OGA in all cell lines. Our data demonstrate that nutrient-sensitive nuclear localization of OGT is tightly associated with AMPK activity in myotubes.
Nuclear Localization of OGT Correlates with O-GlcNAcylation of Nuclear Proteins and Acetylation of Lys-9 on Histone 3 in Myotubes-OGT has a myriad of downstream targets and functional roles in the nucleus (2,6). We therefore wanted to confirm that altered localization of OGT by nutrient deprivation and AICAr is consistent with global changes in O-GlcNAcylation of nuclear proteins, and altered regulation of a nuclear process previously reported to be regulated by OGT. Immunoblots of nuclear extracts from differentiated C2C12 cells deprived of nutrients and growth factors or treated with AICAr demonstrated a strong association between O-GlcNAcylation of nuclear proteins and AMPK activity (Fig.  3A). Notably, biochemical analyses of lysate from differentiated C2C12 cells include cellular extract from the unresponsive myoblastic substrata in which OGT is primarily nuclear (Fig.  1B). The observed 1.6-fold difference in nuclear OGT assayed by immunobloting (Fig. 3A), versus approximate 4-fold difference in nuclear OGT assayed by microscopy in myotubes under the same conditions (Fig. 2) exemplifies this concept. The observed changes in O-GlcNAcylation of total nuclear extract Two ETD spectra were averaged to obtain this spectrum. The predicted monoisotopic cЈ and zЈ-type fragment ion masses are listed above and below the peptide sequence, respectively. The predicted average mass for the doubly charged z 16 ion is shown. All fragment ions identified within the spectrum are labeled while the corresponding predicted ion masses are underlined within the peptide sequence. Reduced charge species resulting from electron capture without dissociation are labeled within the spectrum with neutral losses from these species contained by heavy brackets. Species that fall within the prescribed 3 Da isolation window are represented by , whereas those denoted with an asterisk correspond to reduced charge species resulting from a coeluting, doubly charged species. B, wild-type (WT), phosphorylated active wild-type (pT172-WT), or phosphorylated kinase-dead mutant (pT172-K45R) recombinant AMPK-␣1␤1␥1 complexes were incubated with recombinant O-GlcNAc transferase (ncOGT) in the presence of [␥-32 P]ATP (an autoradiograph ([␥ 32 ]phosphate; top panels) of the same gel stained with G250 Coomassie Blue (total protein; bottom panels)). C, location of Thr-444 (yellow) on a surface representation of the crystal structure of hOGT 4.5 complexed with UDP and CKII peptide (stick structure in the substrate-binding cleft of OGT). hOGT 4.5 contains a truncated TPR domain (4.5 TPR units; gray) connected to the catalytic domain (N-terminal catalytic (N-cat; blue), intervening (Int-D; green) and C-terminal catalytic (C-cat; red) subdomains) via a transitional helix (H3; purple) which includes a predicted nuclear localization signal (NLS; orange). Pivoting about the hinge (represented as the last 6 amino acids of TPR 12 and first 6 amino acids of TPR 13 in teal) is postulated to facilitate access of a wide variety of substrates to the substrate-binding cleft of OGT. D, OGT activity assays of 2 separate purifications of recombinant WT-and T444E-ncOGT. Assays performed on no CKII peptide (No ppt.) were included as negative controls.
are likely representative of a more significant difference in O-GlcNAcylation of nuclear proteins specifically in myotubes.
To confirm that nuclear localization of OGT in myotubes affects a nuclear process previously reported to be regulated by OGT in other cell lines (e.g. histone modifications) (2, 6), we used confocal microscopy to investigate the effect of nutrient/ growth factor deprivation or AICAr on acetylation of Lys-9 (K9Ac) and tri-methylation of Lys-27 (K27me3) on histone 3 (H3). H3 K9Ac levels were significantly decreased with nutrient/growth factor deprivation and increased with AICAr (Fig.  3B). H3 K9Ac levels were highly correlated with nuclear localization of OGT, not only between myotubes, but within the same myotube for each condition (Fig. 3B), demonstrating that acetylation of Lys-9 on histone 3 is tightly linked to the presence of OGT in the nucleus. There was no effect of nutrient/growth factor deprivation or AICAr on H3 K27me3 levels. Our data demonstrate that acute metabolic stresses that alter AMPK activity affect nuclear localization of OGT, O-GlcNAcylation of nuclear proteins and acetylation of Lys-9 on H3 in myotubes.
AMPK Phosphorylates Thr-444 on OGT in Vitro, and Phosphorylation of Thr-444 Is Tightly Linked to AMPK Activity and Increased Nuclear Localization of OGT in Myotubes-Many plausible mechanisms could account for the effects of nutrient/ growth factor deprivation and AMPK activity on the localiza-tion of OGT. To assess if phosphorylation of OGT by AMPK is one possibility, we incubated recombinant nucleocytoplasmic OGT (ncOGT) with phosphorylated fully active recombinant AMPK (pT172-WT) in the presence of [␥-32 P]ATP. Using tandem mass spectrometry, we identified Thr-444 as the residue on OGT phosphorylated by AMPK in vitro (Fig. 4, A and B). Full-length OGT is composed of a C-terminal catalytic domain and 13 N-terminal tetratricopeptide repeats (TPR). Based on a combination of experimental data and computational molecular modeling, Lazarus et al. proposed that OGT dramatically pivots about a "hinge" region located between TPR repeats 12 and 13, facilitating access of a wide variety of potential targets to the substrate binding cleft of OGT (23). A predicted nuclear localization signal (NLS) (2) is located directly adjacent to the "hinge" region. Thr-444 is located at the C-terminal distal end of TPR 12, in close proximity to the "hinge" region ( Fig. 4C) congruent with the possibility that phosphorylation of Thr-444 by AMPK could regulate both the nuclear localization and substrate selectivity of OGT. Consistent with our data, there was no difference in the enzymatic activity between wild-type (WT) and phospho-mimetic T444E-ncOGT (Fig. 4D).
To determine if Thr-444 regulates AMPK-induced nuclear localization of OGT in cell culture, we investigated the localization of overexpressed GFP-tagged WT and Thr-444 to Ala (T444A) mutant OGT in various proliferating cell lines and differentiated C2C12 cells. In contrast to what is observed with endogenous OGT expression (Fig. 1), overexpressed OGT is primarily cytoplasmic in proliferating cells and virtually excluded from the nucleus in myotubes (data not shown 3 ). The mechanisms driving disrupted localization of overexpressed OGT in cell culture remains unknown, but is consistent with previous literature in which a very small increase in OGT expression (2-3-fold) has profound effects on multiple signaling pathways (13). These data indicate that studies investigating AMPK-mediated regulation of overexpressed OGT in cells are not physiologically relevant.
To assess if activation of AMPK could alter phosphorylation of Thr-444 on endogenous OGT in cell culture, we designed a custom antibody specific for phosphorylated Thr-444 on OGT. Phospho-Thr-444 OGT immunoblots of recombinant WT, T444A mutant, and T444E phospho-mimetic ncOGT incubated with or without phosphorylated active recombinant (pT172-WT) or phosphorylated kinase-dead mutant (pT172-D157A) AMPK in the presence of ATP confirmed antibody specificity for phospho-Thr-444 over naked OGT (Fig. 5A). Although specific, technical caveats render this antibody unsuitable for immunofluorescence, or immunoblotting of cell lysate and OGT immunoprecipitates. However, we observed an approximate 3-fold increase in the amount of phospho-Thr-444 OGT immunoprecipitated from nuclear extracts of differentiated C2C12 cells treated with AICAr versus nutrient/ growth factor deprivation (Fig. 5B). We expect this effect would have been-more significant if it were possible to perform these immunoprecipitations on lysate from isolated myotubes, decreasing noise from the unresponsive myoblastic substrata. Collectively, our data indicate nuclear localization of OGT could be mediated by direct phosphorylation of Thr-444 by AMPK in myotubes.

The Substrate Selectivity of OGT Is Altered by Highly Specific Activation of AMPK in Proliferating Cells and Phospho-mimetic
Mutation of Thr-444 (T444E) in Vitro-Although there was no effect any of the conditions tested on the localization of OGT in proliferating cells, because the location of Thr-444 is near the proposed "hinge" region, we speculated there might be an effect on the substrate selectivity of OGT. There are over 2500 O-GlcNAcylated proteins, making analysis of global changes in the substrate selectivity of OGT by conventional one-dimensional SDS-PAGE not suitable. We exploited two-dimensional electrophoresis (2DE) to separate protein by both mass and charge, facilitating a more accurate global analysis of altered O-GlcNAcylated protein status in the cell.
Acute treatment (0.5 and 2 h) with a highly specific small molecule activator of ␤1-subunit harboring AMPK complexes, A-769662 (28), induced global changes in protein O-GlcNAcylation in proliferating Hek293T cells (Fig. 6). Similar results were observed in HepG2 and proliferating C2C12 cells. Activation of AMPK using AICAr also altered global protein O-GlcNAcylation in the cytosolic fraction of lysate from differentiated C2C12 cells (Fig. 7, A and B). Although nonspecific effects are unlikely when using AICAr within a short time frame (i.e. within 30 min) (15), siRNAmediated knockdown of AMPK in the presence of AICAr, or the use of a highly specific activator of AMPK (e.g. A-769662) would provide more conclusive evidence supporting AMPKdependent regulation of OGT in myotubes. Two technical caveats prevent use of these approaches in differentiated C2C12 cells: One, viral infection of C2C12 myotubes requires 72 h of incubation in high-titer lentivirus (Ͼ200 MOI), affecting O-GlcNAcylation irrespective of AMPK activity. Two, ␤2-subunit harboring AMPK complexes are primarily expressed in differentiated C2C12 myotubes (Fig. 7C), rendering studies with A-769662 in these cells not feasible. Despite these caveats, data discussed thus far occurs within a time scale consistent with the potential for direct signaling between AMPK and OGT (i.e. 0.5 to 2 h), and recent data suggest that O-GlcNAcylated protein substrate specificity is largely mediated by OGT, versus OGA (29).
To determine if the effects of AMPK activation on the substrate selectivity of OGT could be mediated by phosphorylation of Thr-444, we investigated differences in enrichment for WTversus T444E-ncOGT interacting proteins from lysate of Hek293T cells incubated in A-769662 for 0 or 2 h. Significant global changes in O-GlcNAcylated proteins bound to WTversus T444E-ncOGT (Fig. 8, A and B) demonstrate placement of a large highly charged moiety on residue Thr-444 (analogous to phosphorylation) is sufficient to induce dramatic changes in the substrate selectivity of OGT. A significant increase in enrichment for AMPK from A-769662 conditioned lysate (Fig. 8C) is congruent with all other data presented demonstrating increased interaction between OGT and activated AMPK.  (Fig. 9A). Overexpressed ␣1, ␣2-, ␥1-, ␥2-, and ␥3-subunits of AMPK were O-GlcNAcylated in Hek293A cells. Interestingly, the ␣-subunits were only modified in complexes harboring the ␥1-subunit (Fig. 9B). The endogenous ␣-subunit of AMPK was O-GlcNAcylated and co-immunoprecipitated with the ␤1and ␤2-subunits in Hek293 and HeLa cells (Fig. 9, C and  D), consistent with what was observed with overexpressed protein. Importantly, increased O-GlcNAcylation of the ␣-subunit after 6 h of treatment with GT, an inhibitor of OGA (Fig. 9D) (30), suggests O-GlcNAc cycles on and off the ␣-subunit under basal conditions, and thus may play a role in regulating AMPK during maintenance of cellular homeostasis.
To assess the possibility of a functional role for O-GlcNAc on AMPK, we examined whether O-GlcNAcylation of the ␣or ␥-subunits is altered by treatment with AICAr or glucose deprivation. Treatment of Hek293A cells with AICAr induced a 2-fold increase in O-GlcNAcylation of the ␥1-subunit of AMPK (Fig. 10A). There is also a progressive increase in O-GlcNAcylation of the ␥1-subunit after 3, 6, and 12 h of glucose deprivation (Fig. 10B). Consistent with previous publications (16), global increases in OGT and O-GlcNAcylated protein was only observed after 12 h of glucose deprivation. O-GlcNAcylation of the ␥1-subunit by recombinant ncOGT was increased on phosphorylated active wild-type (pT172-WT) and constitutively active (T172D) AMPK, but not on phosphorylated kinase-dead (pT172-K45R) AMPK in vitro Acute treatment with a highly specific inhibitor of OGA, TMG (31) inhibited glucose deprivation-induced phosphorylation of Thr-172 on AMPK and blunted phosphorylation of Ser-79 on ACC (Fig. 11, A and B). Inhibition of OGA increased basal phosphorylation of AMPK after prolonged (8 h) treatment, but there was no effect on phosphorylation of ACC (Fig. 11B) suggesting this may be a compensatory mechanism for global dampening of phosphorylated active AMPK. Pre-treatment of differentiated C2C12 cells for 2 h with GT (a less specific inhibitor of OGA compared with TMG) blunted AICAr-induced increases in the kinase activity of AMPK-␣1 and -␣2 immunoprecipitates (Fig. 11C), but had no effect on phosphorylation of Thr-172. We speculate supraphysiological activation of AMPK with AICAr, which leads to very rapid and dramatic increases in intracellular AMP, overcomes the effect of acute inhibition of O-GlcNAc cycling on phosphorylation of Thr-172. The observed reduction in immunoprecipitated AMPK-␣1 kinase activity is therefore independent of phosphorylation of Thr-172, and is congruent with O-GlcNAc dampening the activity of phosphorylated active AMPK. Control data collectively demonstrated; One, global elevation of O-GlcNAcylated proteins for all experiments using TMG or GT confirmed inhibition of OGA. Two, treatment with TMG or GT for up to 48h had Primary antibody incubated without cell lysate (ab), and lysate incubated with nonspecific IgG (IgG) were included as negative controls.
no effect on the basal kinase activity of AMPK. Three, although the effect of GT on AICAr-induced increases in ␣2 activity was not statistically significant, activity of ␣1 complexes was consistently 8-fold higher than ␣2 complexes in differentiated C2C12 cells (data not shown). 3 All data using GT has been replicated with TMG, suggesting that all effects discussed above are specific to inhibition of OGA, versus off-target effects of either inhibitor used.
In conclusion, our data can be summarized in five main points collectively supporting the potential for ubiquitous cross-talk between the O-GlcNAc and AMPK systems in numerous cell lines: 1), nuclear localization of OGT is dynamically regulated by acute metabolic stresses that strongly corre-late with AMPK activity, affecting O-GlcNAcylation of nuclear proteins and acetylation of K9 on H3 in skeletal muscle myotubes. 2), acute activation of AMPK alters the substrate selectivity of OGT in several proliferating cell lines and the cytoplasm of myotubes. 3), AMPK phosphorylates Thr-444 on OGT in vitro and phosphorylation of this residue is associated with AMPK activity and altered nuclear localization and substrate selectivity of OGT. 4), both ␣and all 3 ␥-subunits of AMPK are dynamically O-GlcNAcylated. 5), acute inhibition of O-GlcNAc cycling blunts activation of AMPK by physiological and pharmacological stimuli.

DISCUSSION
OGT and AMPK each have a multitude of downstream targets that vary widely based on tissue/cell type and biological or experimental parameters. A better understanding of other orders of regulation that may mediate contextually dependent spatial, temporal, and substrate specific activity of OGT and AMPK, independent of their respective upstream canonical pathways, is needed. Because of the considerable overlap in upstream regulation and downstream functional consequences of OGT and AMPK activity, we wanted to assess the possibility of direct cross-talk between these two enzymes as a novel alternative nutrient-sensitive process that mediates their respective activity and/or substrate specificity. We demonstrated acute regulation of AMPK activity directly correlates with altered nuclear localization of OGT in myotubes, affecting O-GlcNAcylation of numerous nuclear proteins and K9 acetylation on H3 (generally associated with transcriptional activation). Our results indicate this could be mediated by direct phosphorylation of Thr-444 on OGT by AMPK. McGee et al. demonstrated that AMPK phosphorylates and induces nuclear export of the transcriptional repressor, histone deacetylase 5 (HDAC5), increasing H3 K9Ac levels in differentiated primary human myoblasts (32). Because our data indicate that H3 K9 acetylation is dependent on nuclear localization of OGT in C2C12 myotubes, it would be interesting to investigate if AMPK-induced nuclear export of HDAC5 is dependent on O-GlcNAcylation of AMPK, HDAC5, or some other co-activator. Further investigation of how O-GlcNAc-dependent regulation of the "histone code" mediates gene transcription in response to processes that regulate AMPK in skeletal muscle (e.g. exercise) is a priority for future studies. Such studies could provide vital insight into the molecular mechanisms underlying exercise/mechanical loading-induced global changes in gene transcription that affect skeletal muscle adaptation and plasticity (e.g. increased metabolic capacity and/or mass).
Although we observed no effect of AMPK activation on the localization of OGT in proliferating cells, there was a significant effect on the substrate selectivity of OGT in both proliferating cells, and the cytoplasm of myotubes. Our data indicate these effects could also be mediated by phosphorylation of Thr-444 by AMPK. We postulate that the highly distinctive nutritional demand of non-dividing versus rapidly dividing cells underscores a potentially unique functional role for OGT in terminally differentiated tissue in general. Consistent with this hypothesis, Whelan et al. demonstrated significant nuclear-tocytoplasmic translocation of OGT upon acute treatment with insulin in differentiated 3T3-L1 adipocytes (26), an effect that is not observed in undifferentiated 3T3-L1 adipocytes or other insulin responsive proliferating cell types. Furthermore, Ramirez-Correa et al. demonstrated that O-GlcNAcylation of myofilament (cytoplasmic) proteins affects contractile function in cardiac muscle (27), suggesting cytoplasmic OGT may have a distinct functional role in skeletal muscle as well. Collectively, previous publications and our data highlight the broader potential significance for ubiquitous AMPK-mediated regulation of the substrate specificity of OGT in numerous cell types.
To confirm the possibility of cross-talk, we demonstrated AMPK is dynamically O-GlcNAcylated on its ␣and ␥-subunits and acute treatment of cells with inhibitors of OGA blunts activation of AMPK by physiological and pharmacological stimuli. Due to the considerably large number of O-GlcNAcylated proteins in a cell, there may be several indirect mechanisms that could account for the effect of acute (2 h) inhibition of OGA on AMPK activation. However, most reported effects of global perturbation of O-GlcNAc cycling on various intracellular processes occur on a 6 -24 or even 48-h time scale (1). This suggests that the effects we observed are likely due to altered O-GlcNAc cycling on AMPK itself. Notably, current tools used to study O-GlcNAc in cell culture affect all intracellular O-GlcNAcylated proteins, making investigation of the functional role of O-GlcNAc on a specific protein quite difficult without the use of site-specific mutant analyses. Due to substoichiometric O-GlcNAcylation of recombinant AMPK by ncOGT in vitro, and insufficient amounts of material from AMPK immunoprecipitates, current limitations to detection of O-GlcNAcylated residues by mass spectrometry (1) have prevented the identification of which residues are O-GlcNAc modified on AMPK. Despite these technical caveats, we have dem- Densitometric quantification of phospho-over total AMPK or ACC immunoblots are represented as graphs or bold numbers normalized to Ctrl., respectively. C, kinase activity assays were performed on AMPK-␣1 and -␣2 immunoprecipitates of lysates from differentiated C2C12 cells treated with vehicle (C or Ctrl.), AICAr (A; 1 mM, 30 min), or pre-incubated in GT (10 M, 2 h) prior to AICAr treatment (GTϩA). Right panels, representative control immunoblots of lysate used for kinase activity assays. All quantification represent mean values Ϯ S.E. *, **, ***, and **** denote statistical significance of p Ͻ 0.05, p Ͻ 0.01, p Ͻ 0.001, and p Ͻ 0.0001, respectively, for TMG versus Ctrl for each time point. ## denotes a statistical significance of p Ͻ 0.01 when comparing the 0h versus 6h time points in TMG-treated cells.
onstrated that highly specific perturbation of O-GlcNAc cycling for a short period of time significantly affects activation of AMPK, and that AMPK is dynamically modified by O-GlcNAc. Our data imply that global O-GlcNAc cycling is important for activation of AMPK, and strengthen the hypothesis of O-GlcNAc being a nutrient-sensitive post-translational modification that can regulate AMPK activity independently of AMP/ADP or Ca 2ϩ .
In summary, this work establishes a strong foundation for future more targeted interrogation of the potential for OGT and AMPK to participate in synergistic regulation of numerous nutrient-sensitive intracellular processes essential for cellular metabolism, growth, proliferation, and/or tissue function.