Complex N-glycan and Metabolic Control in Tumor Cells

Golgi B1,6N-acetylglucosaminyltransferase V (Mgat5) produces B1,6GlcNAc-branched complex N-glycans on cell surface glycoproteins that bind to galectins and promote surface residency of glycoproteins, including cytokine receptors. Carcinoma cells from polyomavirus middle T (PyMT) trans-genic mice on a Mgat5 À/À background have reduced surface levels of epidermal growth factor (EGF) and transforming growth factor-B (TGF-B) receptors and are less sensitive to acute stimulation by cytokines in vitro compared with PyMT Mgat5 +/+ tumor cells but are nonetheless tumorigenic when injected into mice. Here, we report that PyMT Mgat5 À/À cells are reduced in size, checkpoint impaired, and following serum withdrawal, fail to down-regulate glucose transport, protein synthesis, reactive oxygen species (ROS), and activation of Akt and extracellular signal-regulated kinase. To further characterize Mgat5 +/+ and Mgat5 À/À tumor cells, a screen of pharmacologically active compounds was done. Mgat5 À/À tumor cells were comparatively hypersensitive to the ROS inducer 2,3-dimethoxy-1,4-naphthoquinone, hyposensitive to tyrosine kinase inhibitors, to Golgi disruption by brefeldin A, and to mitotic arrest by colcemid, hydroxyurea, and campto-thecin. Finally, regulation of ROS, glucose uptake, and sensitivities to EGF and TGF-B were rescued by Mgat5 expression or by hexosamine supplementation to complex N-glycan biosynthesis in Mgat5 À/À cells. Our results suggest that complex N-glycans sensitize tumor cells to growth factors, and Mgat5 is required to balance responsiveness to growth and arrest cues downstream of metabolic flux.


Introduction
Oncogenic gain-of-function mutations in intracellular mediators of Ras/extracellular signal-regulated kinase (Erk) and phosphatidylinositol 3-kinase (PI3K)/Akt signaling stimulate cell proliferation and metabolism, thereby increasing growth autonomy by relaxing the requirements for extracellular growth factors.However, tumor cells remain dependent on extracellular cues, notably cytokines and substratum adhesion to polarize the cytoskeleton and promote epithelial-to-mesenchymal transition (EMT), a feature of invasive carcinomas (1).Epidermal growth factor (EGF) family of growth factors binds receptor tyrosine kinases (RTK) and activates PI3K/ Erk signaling, glucose uptake, cell proliferation, and microfilament remodeling (2).TGF-h binding to ThRII promotes Par6-Smurf1dependent degradation of RhoA in focal regions of polarized cells, which contributes to microfilament remodeling and the mesenchymal morphology (3).TGF-h activation of the canonical Smad2/ Smad3 transcription factor pathway induces extracellular matrix production and, at high levels, inhibits cell cycle progression (4,5).A balanced stimulation of RTK/Erk/PI3K and ThR/Smad2 pathways is required to sustain growth and avoid cell cycle arrest induced by the latter (6).
Galectin-3 binds to the N-glycans on cell surface receptors, which opposes their loss to endocytosis and enhances sensitivities to cognate ligands (7).The galectins are a family of h-galactosidebinding proteins that cross-link glycoproteins with avidities dependent on N-glycan structure and number (8).The number of N-glycans is a feature defined by the protein sequence of each glycoprotein, whereas N-glycan structures are determined by the Golgi N-glycan processing pathway and metabolite supply to sugarnucleotide pools.Notably, fructose-6P, glutamine, and acetyl-CoA supply the hexosamine pathway for de novo UDP-N-acetylglucosamine (UDP-GlcNAc) biosynthesis, a rate-limiting substrate for N-acetylglucosaminyltransferases I, II, IV, and V [encoded by Mgat1, Mgat2, Mgat4a/b, and Golgi h1,6N-acetylglucosaminyltransferase V (Mgat5); refs.9, 10].These medial Golgi enzymes initiate antennae; their substitution by h1,4galactosyltransferase resulting in 0 to 4 (Galh1,4GlcNAch1) antennae with proportionate increases in affinity for galectin-3.The physical and kinetic characteristics of the N-glycan branching pathway cooperate with the number of N-glycans to differentially regulate growth [e.g., EGFR receptor (EGFR)] and arrest (e.g., TGF-h) receptors at the cell surface downstream of metabolic flux to UDP-GlcNAc biosynthesis (8).
Mammary tumor cells derived from polyomavirus middle T (PyMT) transgenic mice on a Mgat5 À/À background are less responsive to insulin-like growth factor (IGF), EGF, platelet-derived growth factor (PDGF), fibroblast growth factor (FGF), and TGF-h, with a reduced galectin-3 binding and shift receptors from the cell surface to the endosomes (7).The PyMT oncoprotein is a cytosolic adaptor protein that enhances PI3K/Akt-dependent membrane remodeling (11), which tends to reduce surface residency of glycoprotein receptors when the galectin lattice is weak.PyMT Mgat5 À/À mice display reduced metastasis in vivo and tumor cells retain an epithelial morphology in culture, but expression of Mgat5 from a retroviral vector in PyMT Mgat5 À/À tumor cells restores surface receptors, EMT, and the metastatic phenotype (7).Tumors arise in PyMT Mgat5 À/À mice with a longer latency, but growth is observed in all mammary fat pads and often accelerated in the later stages, suggesting that there are genetic or epigenetic compensatory mechanisms (12).
Because RTKs are under tonic suppression by phosphatase in endosomes (13,14), we examined the possibility that redoxdependent suppression of these phosphotyrosine phosphatases (PTP) might promote ligand-independent growth signaling in PyMT Mgat5  (16)(17)(18).Oxidation of Cys 124 in the active site of PTEN is reversible by thioredoxin/NADH (17).At focal sites of RTK activation, Rac1 is recruited to the NADPH-dependent oxidase (NOX) complex, which catalyzes H 2 O 2 production (18).PDGF (19), EGF (20), and stromelysin (matrix metalloproteinase-3) stimulate Rac1/H 2 O 2 and promote EMT in tumor cells (21).However, most of the reactive oxygen species (ROS) in mammalian cells are generated in the mitochondria (22).Elevated ROS in BCR-ABLtransformed cells are dependent on oncogenic activation of PI3K and mammalian target of rapamycin (mTOR), glucose uptake, and oxidative respiration (23).Overexpression of Mox1, a NADPH oxidase that catalyzes H 2 O 2 production in a constitutive manner, is also transforming in NIH-3T3 cells (24).Herein, we show that glucose uptake and mitochondrial ROS production promotes growth signaling in PyMT Mgat5 À/À tumor cells, compensating in part for cytokine insensitivity.Moreover, supplementation to complex N-glycan branching by Mgat5 expression or via the hexosamine pathway to UDP-GlcNAc promotes sensitivity to cytokines and control of glucose metabolism in tumor cells.
Cell cycle fluorescence-activated cell sorting.For serum starvation experiments, 5 Â 10 5 cells were plated per 10-cm tissue culture dish in fresh medium and cultured overnight, then washed with 1Â PBS, and incubated for the indicated times with serum-free DMEM.For GlcNAc titration, subconfluent cells were conditioned for 48 h in DMEM, 10% fetal bovine serum (FBS), and GlcNAc at the indicated concentrations.At the indicated times, cells were trypsinized, washed in PBS before fixation with 80% ethanol, and held at 4jC until completion of experiment.Cells were washed in PBS and then in PBS, 12% Triton-X, and 0.12 mmol/L EDTA, followed by incubation with DNase-free RNase A (1 Ag/mL) at 37jC for 30 min, and then stained with 50 Ag/mL propidium iodide for 1 h in the dark.FACSCalibur and cell cycle profile analysis with ModFitLT version 3.0 software were used to analyze the cells.Tumor cell sizes were compared using forward scatter (FSC) measurements on the fluorescence-activated cell sorting (FACS) and with a Coulter counter.
Glucose transport and ATP.Cells were seeded in triplicate at 2.5 Â 10 5 cells per well in six-well plates.Following incubation at 37jC, the cells were washed twice with assay buffer [Krebs-Ringer solution: 116 mmol/L NaCl, 5.4 mmol/L KCl, 0.8 mmol/L MgSO 4 , 1.0 mmol/L CaCl 2 , 25 mmol/L Tris-HCl, 0.2% bovine serum albumin, 1.0 mmol/L NaH 2 PO 4 , and 2.0 mmol/L Na 2 PO 4 (pH 7.45)].Then, assay buffer (1 mL/well) was added to each well and the cells were incubated at 37jC for 30 min, and then 1 mL of [ 3 H]2deoxy-D-glucose (2 ACi/mL) in assay buffer was added to each well.The cells were incubated at 37jC for 20 min and washed four times with ice-cold assay buffer, and 0.5 mL of ice-cold 5% trichloroacetic acid (TCA) was added to each well.After 30 min on ice, 0.4 mL of TCA supernatant was removed and [ 3 H]2-deoxy-D-glucose was quantified by scintillation counting.ATP in cells at a density of 1,000 cells per well in 96-well plates (Corning 96 Flat White) was measured in triplicate, using the CellTiterGlo Luminescent Cell Viability Assay kit (Promega).
ROS measurements.Cells were seeded in triplicate at a density of 1,000 cells per well in 96-well plates (Corning 96 Flat Opaque).For drug incubations, cells were pretreated for 30 min with 10 Amol/L carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP; Sigma) or for 2 h with 5 mmol/L N-acetyl-cysteine (NAC; Sigma).The wells were washed with PBS, and 100 AL of a 1 Amol/L H 2 DCFDA (Molecular Probes) in PBS were added.The probe diffuses into the cells and, on oxidation, becomes fluorescent and trapped in the cell.The plate was incubated at 37jC for 30 min and fluorescence intensity of the oxidized probe at the bottom of each well was measured with excitation at 485-20 nm and emission at 530-25 nm using the Analyst HT plate reader (Molecular Devices) and the Criterion Host software (LJL Biosystems).Background fluorescence in the absence of probe was subtracted, and a standard curve for oxidation was generated using hydrogen peroxide.Results are plotted as the mean F SD of triplicate determination, and experiments were repeated thrice.
Translation rates.Cells were seeded in DMEM plus 10% FBS in triplicate at 2.5 Â 10 5 cells per well in six-well plates.At the given time, the medium was discarded and the cells were washed twice with PBS.Each well was equilibrated with DMEM methionine-free medium for 30 min at 37jC, then medium was removed, and DMEM methionine-free medium with 20 ACi [ 35 S]methionine was added.The cells were treated for 30 min at 37jC.Following the incubation, the medium was removed and each well was washed twice with cold PBS, and then 200 AL of lysis buffer were added to each well.The cells were scraped from each well and then transferred to Eppendorfs.The samples were centrifuged at 14,000 Â g for 10 min at 4jC.The supernatants were removed and 10% TCA was added on ice for 30 min, and then proteins were trapped on GF-C filters (Whatman).The filters were washed thrice with 10% TCA and then washed once with 96% ethanol.The filter papers were air dried and [ 35 S]methionine incorporation was quantified by scintillation counting.
Nuclear translocation of p-Erk and Smad2/3.Cells plated in 96-well plates at 1,000 cells per well for 24 h were serum starved for a further 24 h and then stimulated with TGF-h1, EGF, or FBS.Cells were fixed after 40 min of TGF-h stimulation to measure Smad2/3 nuclear translocation and cells were fixed after 5 min of EGF stimulation to measure p-Erk nuclear translocation.Cells were fixed for 10 min with 3.7% formaldehyde at 20jC, washed with PBS plus 1% FBS, and permeabilized using 100% methanol for 2 min.The cells were washed thrice and blocked in PBS plus 10% FBS overnight at 4jC.Mouse anti-Smad2/3 (Transduction Laboratories) or mouse p-Erk1/2 (Thr 202 /Tyr 204 ; M-8159) was added at 1:1,000 in PBS plus 10% FBS for 2 h at 20jC.To measure nuclear phospho-p53 (Ser 15 ), cells were stained with antibodies (Cell Signaling Technology).The cells were washed thrice with PBS plus 1% FBS.Secondary antibody Alexa Fluor 488labeled antimouse Ig (Molecular Probes) was added at 1:1,000 with Hoechst (1:2,000) for 1 h at 20jC.After washing thrice, the plates were scanned using the ArrayScan automated fluorescence microscope (Cellomics, Inc.).The nuclear staining intensity and cytoplasmic staining intensity were determined individually for 200 cells per well, and cytoplasmic staining intensity was subtracted from nuclear staining intensity for each cell.The mean F SE (n = 200) was generally <4% at each assay point.

Cancer Research
Cancer Res 2007; 67: (20).October 15, 2007 96-well plates were cultured in DMEM and 1% FCS with each compound in the Sigma LOPAC(1280) library of pharmacologically active compounds at a final concentration of 5 Amol/L for 16 h.Drugs were added in duplicate wells using a BioMek-FX and incubated for 24 h, then 10 AL of AlamarBlue (Biosource) were added, and after a further 36 h at 37jC, plates were read using the Gemini XPS fluorometer (Molecular Devices) at 544 nm excitation/590 nm emission.The CellTiterGlo Luminescent Cell Viability Assay kit and the AlamarBlue assay produced similar results.Each plate contained eight positive (no cell treatment, PC) and eight negative (signal background, NC) controls.Data were normalized using the formula: (FI sample À mean FI NC ) / (mean FI PC À mean FI NC ) Â 100.''Hits'' were defined as compounds associated with a signal that shifted by at least 3Â SD from the mean of the entire sample population.The Z factors were 0.58 and 0.53 for the Mgat +/+ and Mgat À/À screen, respectively, whereas the Z' factors were above 0.7 in both screen.Primary ''hits'' were subjected to confirmatory tests done in triplicate and by titration to compare IC 50 values for Mgat5 +/+ and Mgat5 À/À cells.The tyrosine kinase inhibitors AG835, PP2, and genistein were purchased from Calbiochem.

Results
Growth control in PyMT tumor cells is dependent on Mgat5.Serum-starved PyMT Mgat5 À/À tumor cells show low levels of p-Erk nuclear translocation on addition of serum when compared with PyMT Mgat5 +/+ and Mgat5-rescued mutant cells (Fig. 1A).The Mgat5 À/À tumor cells were smaller in size than Mgat5 +/ Mgat5 +/+ tumor cells and rescued mutant cells, suggesting that Mgat5 and its N-glycan products may play a role in growth and cell cycle regulation (Fig. 1B).When injected s.c.into mice, Mgat5 +/+ and Mgat5 À/À tumor cells grew at similar rates, indicating that reduced sensitivity to growth factors in Mgat5 À/À cells does not restrict solid tumor cell growth (data not shown).Growth factor signal to PI3K/Akt/mTOR regulates protein synthesis and cell size, whereas growth factor signal to Erk and cyclin D/cyclin-dependent kinase 4/6 promotes G 1 -S entry (26).Mgat5 À/À cells in asynchronous cultures displayed an abnormal cell cycle profile with Figure 1.Altered growth signaling in Mgat5-deficient PyMT tumor cells.A, Mgat5 +/+ (2.6) and Mgat5 À/À (22.9) tumor cells were serum starved for 24 h and stimulated with FBS, and p-Erk nuclear translocation was measured by ArrayScan.The ''rescued'' are Mgat5 À/À (22.9) cells infected with a retrovial vector for expression of Mgat5 or the Mgat5 L188R mutant that does not localize to Golgi.B, tumor cells were grown in DMEM plus 10% FBS and cell size was measured by FSC.C, cell cycle distribution by FACS analyses following 0, 24, and 48 h of serum starvation.D, Mgat5 +/+ and Mgat5 À/À tumor cells growing in DMEM plus 10% FBS were serum starved for the indicated times.Cell lysates were applied to SDS-PAGE and probed for p-Akt, p-Erk, UCP-1, and g-tubulin.
increased apoptosis and a higher fraction of S and G 2 -M phase cells compared with Mgat5 +/+ cells (Fig. 1C).More cells were observed in S-G 2 -M after 48 h in serum-free medium (SFM) compared with Mgat5 +/+ , reflecting a delay or failure of mutant cells to arrest in G 1 (Fig. 1C).p-Erk and p-Akt levels declined in Mgat5 +/+ cells 24 to 48 h after serum withdrawal but remained elevated in Mgat5 À/À cells (Fig. 1D).Nuclear p53(Ser 15 ) levels were lower in Mgat5 À/À tumor cells under normal growth conditions, consistent with a weakened checkpoint (Supplementary Fig. S1A).The M-phase blocker colcemid reduced growth of Mgat5-rescued mutant cells but not that of Mgat5 À/À tumor cells (Supplementary Fig. S1B).Similarly, the S-phase blocker hydroxyurea slowed Mgat5-rescued cells but was less effective on Mgat5 À/À tumor cells (Supplementary Fig. S1C).These experiments indicate that PyMT Mgat5 À/À cells are functionally deficient in G 1 -S and G 2 -M checkpoints and suggest that shorter pauses in interphase result in reduced cell size.
Mgat5-and PyMT-dependent control of metabolism.Although PI3K/Akt activation is known to stimulate glucose uptake (27), we explored the converse relationship: whether hyperglucose metabolism in Mgat5 À/À cells contributes to PI3K/Akt signaling.Glucose uptake rates were similar in Mgat5 +/+ , Mgat5 À/À , and Mgat5-rescued cell lines cultured in serum-containing medium, but after 24 h in SFM, Mgat5 +/+ and Mgat5-rescued cells showed strong down-regulation, whereas glucose uptake remained unchanged in Mgat5 À/À cells at 24 h and showed an increase by 72 h (Fig. 2A).GLUT1 gene expression is increased by oncogenic transformation of cultured cells and correlates with poor survival of cancer patients (28).Glut1 protein levels were 2-to 3-fold higher in Mgat5 À/À cells compared with Mgat5 +/+ cells (Fig. 2B).Glucose transport V max was 6-fold higher in Mgat5 À/À compared with Mgat5 +/+ cells, whereas K D values were similar indicating that Mgat5-dependent N-glycosylation did not alter the kinetics of transport (Fig. 2C).After 24 h in SFM, ROS levels decreased in Mgat5 +/+ tumor cells but increased 2-fold in Mgat5 À/À tumor cells (Fig. 2D).We might expect the Mgat5 deficiency in surface RTKs to reduce basal growth signaling and suppress glucose uptake in the absence of PyMT oncoprotein.Indeed, Mgat5 À/À mouse embryonic fibroblasts (MEF) proliferate more slowly and display reduced sensitivity to EGF, PDGF, FGF, and IGF (29).Glucose uptake and ROS levels were reduced in Mgat5 À/À MEFs compared with Mgat5 +/+ MEFs (Supplementary Fig. S2A and B; ref. 29).Serum withdrawal further reduced glucose uptake and ROS in both mutant and wild-type (WT) MEFs.Taken together, this suggests that Mgat5 promotes growth signaling and glucose

Cancer Research
Cancer Res 2007; 67: (20).October 15, 2007 uptake in nontransformed cells, but this requirement is removed by the PyMT oncoprotein, which drives a glucose-dependent positive feedback loop.However, Mgat5 expression seems to be required in PyMT-transformed cells to sense serum withdrawal and suppress growth.
Translational initiation and elongation are positively regulated by Akt/mTOR and inhibited by AMP-activated protein kinase (AMPK; refs.30,31).ATP levels were 2-fold higher in Mgat5 À/À cells and increased further upon transfer to SFM, whereas ATP was unchanged in Mgat5 +/+ cells (Fig. 3A).Mgat5 À/À cells displayed a higher rate of [ 35 S]methionine incorporation into proteins, which increased after serum withdrawal at a faster rate than that observed in Mgat5 +/+ cells (Fig. 3B).Pdk1 activates mTOR, and in turn S6 kinase, which phosphorylates eEF2 at Ser 366 , thus activating protein translation.Serum withdrawal leads to inactivation of eEF2 by dephosphorylation at Ser 366 and phosphorylation at Thr 56 (32,33).A comparison of Mgat5 À/À and Mgat5 +/+ cells growing in serumcontaining medium revealed that the basal rate of protein synthesis was higher and Thr 56 p-eEF2 levels lower in mutant cells (Fig. 3B  and C).After 24 h of SFM, Mgat5 +/+ cells showed a marked increase in Thr 56 p-eEF2, accompanied by a small decline in protein synthesis.Although Thr 56 p-eEF2 also increased in Mgat5 À/À cells at 24 h of SFM, it did not reach levels observed in Mgat5 +/+ cells, presumably remaining below a threshold for protein synthesis inhibition.Thr 56 p-eEF2 decreased between 24 and 72 h in SFM for both Mgat5 À/À and Mgat5 +/+ cells, accompanied by increased rates of protein synthesis.Thus, PyMT Mgat5 À/À tumor cells in normal and SFM growth conditions maintain a higher energy charge and protein synthesis rate.
ROS-dependent Erk/Akt signaling in Mgat5 À/À cells.Next, we examined the relationship between elevated metabolism in Mgat5 À/À cells and growth signaling.The excess ROS in Mgat5 À/À tumor cells growing in SFM was suppressed by treating cells for 30 min with carbonyl cyanide FCCP, an uncoupler of oxidative phosphorylation (Fig. 4A).NAC, a thiol reductant, also suppressed ROS in Mgat5 À/À cells.FCCP treatment reduced tonic p-Akt and p-Erk in Mgat5 À/À cells to levels comparable with that of untreated Mgat5 +/+ cells in SFM conditions (Fig. 4B).Mgat5 À/À cells treated with rapamycin, an mTOR inhibitor, displayed a decrease in ROS production, whereas glucose uptake was reduced by treatment with rapamycin or the PI3K inhibitor LY294002 (Fig. 4C and D).ROS levels were unaffected in Mgat5 +/+ cells, whereas glucose uptake was reduced by LY294002.This indicates that high ROS production in Mgat5 À/À cells is dependent in part on PI3K/mTOR activation of metabolism.Mitochondrial ROS is suppressed by the uncoupling proteins (UCP), which catalyze a controlled leakage of protons across the inner membrane.UCP-1 levels were lower in Mgat5 À/À tumor cells than Mgat5 +/+ cells, and following serum withdrawal, UCP-1 increased in Mgat5 À/À tumor cells and decreased in Mgat5 +/+ cells (Fig. 1C).UCP-1 expression is stimulated by superoxides (22), which may provide negative feedback to suppress mitochondrial ROS production.ROS stimulates antioxidant pathways that limit damage (34), suggesting the possibility that high ROS levels in Mgat5 À/À cells may be coupled with higher turnover rates for signaling intermediates.To examine this possibility, we compared the recovery rates of p-Erk, p-Akt, and oxidized PTEN following a pulse of exogenous H 2 O 2 .Akt became highly phosphorylated in both Mgat5 +/+ and Mgat5 À/À within 20 min of H 2 O 2 addition, and the return to baseline was marginally more rapid in Mgat5 À/À cells, as was the recovery of reduced PTEN in Mgat5 +/+ and Mgat5 À/À (Fig. 5A).However, Erk phosphorylation recovered more rapidly in Mgat5 À/À compared with Mgat5 +/+ tumor cells in SFM conditions (Fig. 5B).This suggests that anti-oxidant pathways are also increased in Mgat5 À/À cells and contribute to increased steady-state and turnover rates of p-Akt and p-Erk.The activity of ROS as a PTP inhibitor may render Mgat5 À/À cells less dependent on specific membrane RTKs.Indeed, Mgat5 À/À cells were 5 times more resistant to PP2, an inhibitor of Src family kinases, 10 times more resistant to genistein, a broad range tyrosine kinase inhibitor, and 3-fold more resistant to AG835, an inhibitor of EGFR (Fig. 5C).
To further explore the critical determinant of growth in an unbiased manner, we screened the LOPAC chemical library for compounds with differential effects on growth of Mgat5 À/À and Mgat5 +/+ tumor cells.Only 3 of 1,200 compounds met this criterion.Mgat5 À/À cells were more sensitive to 2,3-dimethoxy-1,4-naphthoquinone (DMNQ), a ROS inducer (Fig. 5D).DMNQ induced f2-fold higher levels of H 2 O 2 in Mgat5 À/À cells than in Mgat5 +/+ cells consistent with differential toxicity (data not shown).Mgat5 +/+ cells were also more sensitive to brefeldin A, which disrupts Golgi, and to camptothecin, a topoisomerase inhibitor.These results are consistent with the defect observed in Mgat5 À/À cells, notably increased oxidative stress, reduced dependency on complex N-glycan, and reduced checkpoint enforcement.Greater resistance to kinase inhibitors suggests that growth and survival in Mgat5 À/À cells are more dependent on redox turnover than extracellular cytokines.
Rescue of metabolic control in Mgat5 À/À cells by hexosamine supplementation.Our results suggest that growth signaling in PyMT Mgat5 À/À tumor cells is less dependent on cytokines and more dependent on positive feedback from glucose metabolism.Mgat5 À/À tumor cells fail to down-regulate growth in SFM, suggesting that h1,6GlcNAc-branched N-glycans may also be required to suppress growth signaling.N-glycan branching is increased with GlcNAc supplementation to the culture medium, and this enhances galectin-dependent binding and retention of receptor at the cell surface (8).GlcNAc is salvaged by the Mitochondrial ROS production and growth signaling in Mgat5 À/À tumor cells.A, ROS levels in Mgat5 +/+ , Mgat5 À/À , and Mgat5-rescued cells.Cells were either untreated or pretreated with 10 Amol/L FCCP for 30 min or with 5 mmol/L NAC for 2 h.P < 0.05 by ANOVA.B, Mgat5 +/+ and Mgat5 À/À cells growing in DMEM with and without FBS for 48 h were either untreated or treated with 10 Amol/L FCCP for 30 min.Western blot probed for p-Akt, p-Erk, and g-tubulin.C and D, Mgat5 +/+ and Mgat5 À/À cells growing in DMEM without FBS for 48 h and treated with the LY294002 or rapamycin for 3 h before measuring ROS and glucose uptake.

Cancer Research
Cancer Res 2007; 67: (20).October 15, 2007 hexosamine pathway and increases UDP-GlcNAc supply to Golgi N-acetylglucosaminyltransferases (Mgat1, Mgat2, Mgat4a/b, and Mgat5; Supplementary Fig. S3A).GlcNAc supplementation does not restore h1,6GlcNAc-branched N-glycans in Mgat5 À/À cells but rather increases the fraction of the next most branched N-glycan (tri-antennary) from 15% to 30% of the total N-glycan pool (8).The tri-antennary and tetra-antennary N-glycans in Mgat5 +/+ cells increase only marginally with GlcNAc supplementation, suggesting that the medial Golgi pathway in Mgat5 +/+ tumor cells is operating closer to saturation for UDP-GlcNAc (8).Activation of PI3K and Erk Figure 5. Redox turnover of PTEN and Erk/Akt signaling in tumor cells.A, cells were serum starved for 24 h and then exposed to H 2 O 2 for 20 min, and lysates were prepared for Western blotting with antibodies to p-Akt and PTEN.Red., reduced; Ox., oxidized.B, time course of p-Erk nuclear-cytoplasmic difference in Mgat5 +/+ and Mgat5 À/À tumor cells following a 30-min pulse of 1 mmol/L H 2 O 2 .C, sensitivity of Mgat5 À/À and Mgat5-rescued Mgat5 À/À cells to PP2, genistein, and AG835.Cells were cultured in the presence of the drug for 4 d in DMEM plus 1% FBS.D, differential sensitivity of Mgat5 +/+ and Mgat5 À/À cells to three compounds identified in a screen of the Sigma LOPAC(1280) chemical library conducted in DMEM plus 1% FBS.Points, mean of three experiments; bars, SD.Cap, camptothecin; Bre, brefeldin A.
Surface levels of ThR in Mgat5 À/À were lower than Mgat5 +/+ cells, and GlcNAc supplementation increased ThR to comparable levels for both cell lines (Supplementary Fig. S3B).Surface levels of EGFR were partially rescued in Mgat5 À/À cells by GlcNAc supplementation (8), and acute responsiveness to both EGF and TGF-h was largely restored (Supplementary Fig. S3C and D).Thus, an increase in the less branched tri-antennary N-glycans is sufficient to restore the surface residency of receptors in Mgat5 À/À cells.Importantly, GlcNAc supplementation normalized glucose and ROS transport in Mgat5 À/À cells under low serum conditions (Fig. 6A).
GlcNAc suppressed the growth of Mgat5 À/À tumor cells to a greater degree than Mgat5 +/+ tumor cells (Fig. 6B; Supplementary Fig. S4A-C).Mgat5 À/À tumor cells displayed a higher proportion of S-phase cells, and GlcNAc supplementation at 20 to 30 mmol/L rescued the cell cycle profile.However, higher concentrations of GlcNAc selectively suppressed proliferation of Mgat5 À/À cells compared with Mgat5 +/+ cells (Fig. 6B; Supplementary Fig. S4A-C).The cell cycle profile in Mgat5 +/+ cells was less severely affected by GlcNAc, with only a small increase in the S-phase population.GlcNAc supplementation restored basal nuclear phospho-Smad2/3 (p-Smad2/3) in Mgat5 À/À cells to levels comparable with that of Mgat5 +/+ cells, presumably reflecting enhanced sensitivity to autocrine/paracrine TGF-h.Similarly, A, [ 3 H]2-deoxy-D-glucose uptake and ROS levels for Mgat5 +/+ and Mgat5 À/À tumor cells cultured in DMEM plus 10% FBS for 48 h either with or without 40 mmol/L GlcNAc, followed by 24 h in SFM.B, S-phase fraction by FACS analysis for Mgat5 +/+ and Mgat5 À/À cells cultured in DMEM plus 10% FBS supplemented with GlcNAc for 48 h (cell cycle distribution is in Supplementary Fig. S4).C, tonic levels of nuclear p-Erk and p-Smad2 expressed as a ratio in Mgat5 +/+ and Mgat5 À/À cells following 48 h of GlcNAc supplementation in normal growth conditions.D, a model of N -glycan-dependent regulation of cytokine receptors and metabolism.The PyMT oncoprotein directly activates PI3K/Akt, promoting positive feedback to glucose uptake.Glucose metabolism increases ROS, thus oxidation of signaling intermediates that stimulate growth.Metabolic flux to the hexosamine and N -glycan branching pathway and/or Mgat5 expression increases affinities for galectin and surface residency of cytokine receptors in a lattice microdomain (see ref. 8).RTKs that promote growth have relatively high numbers of N -glycans with 11.4 F 5.1 N-X-S/T consensus sequences (n = 30 receptor) and estimated to be f70% occupancy.Receptor kinases with low multiplicity function in organogenesis and differentiation/arrest include Tie1, Musk, Ltk, ROR1/2, DDR1, the Eph receptors, and TGF-h/BMP receptors (27 receptors 2.48 F 1.28 N-X-S/T sites/ receptor).In Mgat5 À/À cells, insufficient positive feedback from the cell surface (1) is partially alleviated by ROS-dependent activation of growth signaling intermediates (green ) in the endosomes.However, positive feedback via the hexosamine/N-glycan processing is also insufficient to maintain (2 ) ThR and other low receptors with few N -glycans at levels sufficient to retrain the cell cycle.In PyMT Mgat5 +/+ cells, RTKs attain higher affinities for galectins and are retained at the cell surface in greater numbers and dominate over ThR and other low receptors.

Cancer Research
Cancer Res 2007; 67: (20).October 15, 2007 GlcNAc partially restored basal nuclear p-Erk in serum-depleted conditions (Supplementary Fig. S3C and D).In culture conditions with serum, the ratio of p-Erk to p-Smad2/3 in Mgat5 À/À cells decreased with GlcNAc supplementation but much less in Mgat5 +/+ cells (Fig. 6C).TGF-h/Smad and PI3K/Erk pathways are antagonistic at multiple levels (5,38) and therefore likely to play a role in growth control as has been observed in Caenorhabditis elegans (39).Our results show that Mgat5 products, the h1,6GlcNAc-branched N-glycans, are required to balance growth and arrest signaling in PyMT tumor cells (Fig. 6D).

Discussion
Tumor progression in PyMT transgenic mice is slower in Mgat5 À/À mice than in WT mice, and Mgat5 À/À mammary tumor cells established in tissue culture are deficient in responsiveness to multiple ligands (7,12).Tonic levels of activated p-Erk, p-Akt, ATP, and protein synthesis under normal culture conditions were comparable in Mgat5 À/À and Mgat5 +/+ tumor cells, but surprisingly, removal of serum failed to suppress glucose metabolism, protein synthesis, and ROS production in Mgat5 À/À tumor cells.Elevated ROS and Akt activation in Mgat5 À/À tumor cells is dependent on glucose metabolism because chemical uncoupling of oxidative respiration normalized both.Mutant cells displayed an accelerated turnover of p-Erk and p-Akt, possibly due to parallel increases in ROS production and anti-oxidant activities.Rapamycin reduced ROS and glucose uptake in Mgat5 À/À tumor cells, indicating a requirement for positive feedback through mTOR.Thus, sustained glucose uptake and mitochondrial ROS production in PyMT Mgat5 À/À tumor cells promotes Akt and Erk signaling.The elevated energy charge in Mgat5 À/À tumor is expected to suppress negative regulation of mTOR/S6K by AMPK (40).Indeed, depletion of S6K1 or mutation of the S6K phosphorylation sites in S6 decreases cell size and increases protein synthesis rates, similar in this regard to the phenotype of the PyMT Mgat5 À/À tumor cells (41).Interestingly, nutrient restriction in Drosophila AMPKadeficient epithelial cells results in loss of polarity and overproliferation, suggesting that changes in basic metabolism are an important aspect of the cancerous phenotype (42).ROS has been shown to activate mTOR in cultured human cells, whereas S6K1 phosphorylation becomes insensitive to nutrient deprivation (43).We suggest that ROS-dependent growth signaling and changes in metabolite flux to N-glycan biosynthesis may be critical determinants of cell size and growth autonomy.
Activation of EGFR at the plasma membrane by EGF stimulates PI3K/Akt signaling and H 2 O 2 production via Rac1/NOX in a transient and localized manner that promotes polarization of microfilaments and cell motility (13).In contrast, ROS generated by the electron transport chain is likely to be persistent, and nonpolarized, as mitochondria are generally distributed in the cell.Moreover, activation of EGFR by H 2 O 2 or EGF induces different patterns of receptor phosphorylation, which can affect signaling dynamics.The c-Cbl binding site at Tyr 1045 is under phosphorylation by H 2 O 2 stimulation, whereas four other sites common to EGFR stimulation are phosphorylated.As such, H 2 O 2 prolongs EGFR activation by slowing ubiquitination and transit to the endosomes where inactivation occurs by PTPs (44).Near saturating levels of phosphatase activities are required to maintain the sensitivity of cytokine receptors to ligands.For example, EGFR is under tonic phosphatase suppression in endomembrane compart-ments, but redox-dependent suppression of PTPs in the endosomes promotes ligand-independent activation (13,14).Elevated ROS in Mgat5 À/À tumor cells may favor tonic activation of RTKs present at higher levels in endosomes (7).Mgat5 À/À tumor cells are more resistant to Src and EGFR kinase inhibitors and to brefeldin A, an inhibitor of guanine nucleotide-exchange proteins (BIG) for ADPribosylation factors, which are required in Golgi vesicular transport (45).The relative resistance of Mgat5 À/À tumor cells to these drugs is consistent with a reduced requirement for surface expression of growth factor receptors.PyMT Mgat5 À/À cells are checkpoint deficient and display an abnormal cell cycle distribution.PI3K/Akt activity normally increases at G 1 -S, then declines with the G 2 pause, and increases again at G 2 -M (26).Apparently, high levels of metabolic ROS in PyMT Mgat5 À/À cells blocks the down-regulation of Akt in G 2 and hastens M phase.Consistent with loss of checkpoints and high ROS, PyMT Mgat5 À/À tumor cells are relatively resistant to genotoxic agents and sensitive to DMNQ, a ROS inducer.However, oxidative stress also stimulates acetylation and activation of FOXO and p53 (46,47), which drives expression of checkpoint, DNA repair, and anti-oxidant proteins, including GADD45, cyclin B, plk1, mSOD, and p27 Kip1 (48).Of these competing effects, glucose metabolism and checkpoint dysfunction prevail in PyMT Mgat5 À/À tumor cells, suggesting that Mgat5-modifed N-glycans are required for growth arrest.On serum withdrawal, PyMT Mgat5 À/À tumor cells fail to down-regulate glucose metabolism and ROS, but both are down-regulated in primary MEFs, and to a greater extent in the Mgat5 À/À MEFs.This argues that growth stimulation by PyMT oncoprotein promotes glucose transport, metabolic ROS, and positive feedback to growth signaling, compensating in part for loss of growth factor sensitivity in Mgat5 À/À cells.
Both GlcNAc supplementation and Mgat5 expression promote receptor binding to galectins and shift receptor residency from the endosomes (ROS-dependent RTKs) to the cell surface (ligand dependent), but with qualitative differences.GlcNAc supplementation fully rescued surface ThRII receptor, which has only two N-glycans, whereas EGFR with eight N-glycans remains partially dependent on Mgat5 for optimal hexosamine-dependent surface expression (8).
In keeping with this relationship, increasing GlcNAc supplementation reduced the nuclear p-Erk/p-Smad2/3 ratio in Mgat5 À/À tumor cells, as well as glucose uptake and proliferation, whereas GlcNAc had a more modest effect on Mgat5 +/+ cells (Fig. 6D).Mgat5 expression commonly increases in oncogene-transformed cells, and predictions based on computational modeling indicate that Mgat5 products preferentially increase the surface levels of growth receptors (high number of N-glycans), relative to ThRII and other glycoproteins with few N-glycans (8).Our results suggest that nutrient flux to complex N-glycan biosynthesis coordinates the cellular response to multiple extracellular cues in tumor cells that determine growth, invasion, and sensitivity to drugs.

Figure 3 .
Figure 3. Mgat5-dependent regulation of growth signaling, oxidative phosphorylation, and protein synthesis.A, ATP levels in cell lysates.Representative of three experiments.Columns, mean; bars, SE.P < 0.05 by paired t test.B, [ 35 S]methionine incorporation into protein following a 30-min pulse labeling.Representative of three experiments.Columns, mean; bars, SD.P < 0.05 by paired t test.C, Western blot of cell lysates probed for Thr 56 p-eEF2 and g-tubulin.

Figure 4 .
Figure 4. Mitochondrial ROS production and growth signaling in Mgat5 À/À tumor cells.A, ROS levels in Mgat5 +/+ , Mgat5 À/À , and Mgat5-rescued cells.Cells were either untreated or pretreated with 10 Amol/L FCCP for 30 min or with 5 mmol/L NAC for 2 h.P < 0.05 by ANOVA.B, Mgat5 +/+ and Mgat5 À/À cells growing in DMEM with and without FBS for 48 h were either untreated or treated with 10 Amol/L FCCP for 30 min.Western blot probed for p-Akt, p-Erk, and g-tubulin.C and D, Mgat5 +/+ and Mgat5 À/À cells growing in DMEM without FBS for 48 h and treated with the LY294002 or rapamycin for 3 h before measuring ROS and glucose uptake.

Figure 6 .
Figure 6.Mgat5 balances hexosamine-dependent regulation of metabolism and tumor cell proliferation.A, [ 3 H]2-deoxy-D-glucose uptake and ROS levels for Mgat5 +/+ and Mgat5 À/À tumor cells cultured in DMEM plus 10% FBS for 48 h either with or without 40 mmol/L GlcNAc, followed by 24 h in SFM.B, S-phase fraction by FACS analysis for Mgat5 +/+ and Mgat5 À/À cells cultured in DMEM plus 10% FBS supplemented with GlcNAc for 48 h (cell cycle distribution is in Supplementary Fig.S4).C, tonic levels of nuclear p-Erk and p-Smad2 expressed as a ratio in Mgat5 +/+ and Mgat5 À/À cells following 48 h of GlcNAc supplementation in normal growth conditions.D, a model of N -glycan-dependent regulation of cytokine receptors and metabolism.The PyMT oncoprotein directly activates PI3K/Akt, promoting positive feedback to glucose uptake.Glucose metabolism increases ROS, thus oxidation of signaling intermediates that stimulate growth.Metabolic flux to the hexosamine and N -glycan branching pathway and/or Mgat5 expression increases affinities for galectin and surface residency of cytokine receptors in a lattice microdomain (see ref.8).RTKs that promote growth have relatively high numbers of N -glycans with 11.4 F 5.1 N-X-S/T consensus sequences (n = 30 receptor) and estimated to be f70% occupancy.Receptor kinases with low multiplicity function in organogenesis and differentiation/arrest include Tie1, Musk, Ltk, ROR1/2, DDR1, the Eph receptors, and TGF-h/BMP receptors (27 receptors 2.48 F 1.28 N-X-S/T sites/ receptor).In Mgat5 À/À cells, insufficient positive feedback from the cell surface (1) is partially alleviated by ROS-dependent activation of growth signaling intermediates (green ) in the endosomes.However, positive feedback via the hexosamine/N-glycan processing is also insufficient to maintain (2 ) ThR and other low receptors with few N -glycans at levels sufficient to retrain the cell cycle.In PyMT Mgat5 +/+ cells, RTKs attain higher affinities for galectins and are retained at the cell surface in greater numbers and dominate over ThR and other low receptors.
À/À tumor cells.Leakage of electrons directly to O 2 from complexes I, III, and IV in the electron-transport chain generates superoxide (O 2 ÁÀ ), which is converted to H 2 O 2 by superoxide dismutase.H 2 O 2 titrates redox-sensitive proteins and, notably, oxidizes an essential thiol group found in the active site of PTPs (15).Examples include H 2 O 2 -mediated inhibition of PTEN, LAR, PTP1B, and LMW-PTP, which enhance positive downstream signaling