Glycosylation with O-linked β-N-acetylglucosamine (O-GlcNAcylation) is a reversible posttranslational modification that regulates the activity of intracellular proteins according to glucose availability and its metabolism through the hexosamine biosynthesis pathway. This modification has been involved in the regulation of various immune cell types, including macrophages. However, little is known concerning the mechanisms that regulate the protein O-GlcNAcylation level in these cells. In the present work, we demonstrate that LPS treatment induces a marked increase in protein O-GlcNAcylation in RAW264.7 cells, bone marrow–derived and peritoneal mouse macrophages, as well as human monocyte-derived macrophages. Targeted deletion of OGT in macrophages resulted in an increased effect of LPS on NOS2 expression and cytokine production, suggesting that O-GlcNAcylation may restrain inflammatory processes induced by LPS. The effect of LPS on protein O-GlcNAcylation in macrophages was associated with an increased expression and activity of glutamine fructose 6-phosphate amidotransferase (GFAT), the enzyme that catalyzes the rate-limiting step of the hexosamine biosynthesis pathway. More specifically, we observed that LPS potently stimulated GFAT2 isoform mRNA and protein expression. Genetic or pharmacological inhibition of FoxO1 impaired the LPS effect on GFAT2 expression, suggesting a FoxO1-dependent mechanism. We conclude that GFAT2 should be considered a new LPS-inducible gene involved in regulation of protein O-GlcNAcylation, which permits limited exacerbation of inflammation upon macrophage activation.
Glycosylation with O-linked β-N-acetylglucosamine (O-GlcNAcylation) is a posttranslational modification that regulates the activity of cytosolic, nuclear, and mitochondrial proteins. This modification is controlled by only two enzymes: OGT, which adds N-acetylglucosamine (GlcNAc) on serine or threonine residues, and OGA, which removes it. O-GlcNAcylation, which regulates a wide array of biological processes (1), depends on the availability of glucose and its metabolism through the hexosamine biosynthesis pathway (HBP) (Fig. 1A). This modification has been involved in the modulation of various signaling pathways, and perturbations of O-GlcNAcylation participate in several important human pathologies (2–4). Moreover, a number of studies indicated a link between O-GlcNAcylation and inflammatory conditions, including diabetes, autoimmune diseases, and cancer (5–11). Moreover, several lines of evidence have suggested a role for O-GlcNAcylation in immune cell signaling. Indeed, O-GlcNAcylation was first discovered at the surface of lymphocytes (12), and dynamic changes upon lymphocyte activation were described at the beginning of the 1990s (13). More recently, several studies indicated that protein O-GlcNAcylation plays major roles in the immune system. In T cells, TCR activation results in global elevation of protein O-GlcNAcylation (14, 15), increased O-GlcNAcylation of several transcription factors (16–19), and accumulation of OGT at the immunological synapse (14). Moreover, in B lymphocytes, activation of the BCR induces O-GlcNAcylation of several signaling molecules involved in BCR signaling (20, 21). Overall, these studies strongly support a major role in O-GlcNAcylation in adaptive immunity. However, the molecular mechanisms by which immune cell activation modulates protein O-GlcNAcylation remain elusive.
In macrophages, contradictory results have been obtained concerning the effect of LPS on protein O-GlcNAcylation and its role in inflammatory processes. Whereas some studies indicated that LPS induced a general increase in protein O-GlcNAcylation (22, 23), others found that LPS decreases protein O-GlcNAcylation in macrophages (24, 25). Moreover, whereas several lines of evidence indicate that increased O-GlcNAcylation of signaling proteins in the NF-κB pathway potentiates inflammatory processes (26–29), other studies indicate that increased O-GlcNAcylation may inhibit proinflammatory signals (24, 30). More recently, O-GlcNAcylation of the signaling adaptor MAVS in macrophages has been involved in antiviral signaling response (31), whereas increased RIPK3 O-GlcNAcylation was shown to regulate necroptosis signaling upon LPS stimulation (25). Thus, whereas much evidence suggests that modulation of O-GlcNAcylation signaling constitutes an important facet of innate immune response, the mechanisms by which macrophage activation regulates O-GlcNAc and its consequences on inflammatory processes remain to be firmly established.
In the present work, using different macrophage cell models, we clearly demonstrated that LPS markedly stimulates protein O-GlcNAcylation in macrophages through increased expression and activity of the rate-limiting enzyme of the HBP, glutamine fructose-6-phosphate amidotransferase (GFAT). More specifically, we showed that, whereas resting macrophages express mainly the GFAT1 isoform, upon LPS stimulation a marked increase in GFAT2 expression was observed, revealing GFAT2 as a LPS-inducible gene in macrophages that promotes a general increase in protein O-GlcNAcylation. In macrophages with conditional OGT deletion, lack of LPS-induced O-GlcNAcylation was associated with an increase in the production of proinflammatory molecules, suggesting that GFAT2 induction may be part of a regulatory loop that may limit inflammation and/or permit its resolution after TLR4 activation.
Materials and Methods
Chemicals and Abs
Thiamet G (3aR,5R,6S,7R,7aR)-2-(ethylamino)-3a,6,7,7a-tetrahydro-5-(hydroxymethyl)-5H-pyrano.thiazole-6,7-diol) and LPS from Salmonella enterica serotype typhimurium were from Sigma-Aldrich (Saint Quentin Fallavier, France). FoxO1 inhibitor AS1842856 was from Calbiochem. PUGNAc (O-(2-acetamido-2-deoxy-d-glucopyranosylidene)-amino-N-phenylcarbamate) was from Toronto Research Chemicals (North York, ON, Canada). Anti-GFAT2 Ab (EPR 19095) and anti-FOXO1A Ab (chip grade ab39670) were from Abcam; anti-GFAT1 (H-49), anti-NOS2 (N-20), anti-GAPDH (O411), anti-Tubulin (TU-02), anti-UBF (F9), and HRP-conjugated anti-rabbit Abs were from Santa Cruz Biotechnology; the anti-FOXO1 Ab for Western blotting (L27) was from Cell Signaling Technology; anti–O-GlcNAc transferase Ab (O 6264) was from Sigma-Aldrich; anti–O-GlcNAcase Ab (NBP2-32233) was from Novus Biologicals; anti–O-GlcNAc Ab (RL2) was from Thermo Fisher Scientific; anti-Clathrin Ab (610500) was from BD Transduction Laboratories; and HRP-conjugated anti-mouse Ab was from Jackson ImmunoResearch Laboratories.
The cDNA coding for nuclear, cytosolic, and plasma membrane FRET O-GlcNAc biosensors (32) were generously provided by Prof. L.K. Mahal (University of Texas, Austin). Bioluminescence resonance energy transfer (BRET) biosensors were developed based on these FRET constructs by replacing the CFP by a Rluc8 sequence (33).
Human (−801>>>0) and mouse GFAT2 (−501>>>0) putative promoter sequences were amplified by PCR and cloned in a firefly luciferase plasmid (pGL4-20; Promega). FoxO1 binding sites on GFAT2 promoter were identified using the Regulatory Sequence Analysis Tools web server (http://rsat.sb-roscoff.fr/RSAT_home.cgi). Mutagenesis of Foxo1 binding sites on the mouse GFAT2 reporter gene was performed using a mutagenesis kit (QuikChangeII XL; Agilent Technologies).
Constitutively active FOXO1-TM (mutated on the three Akt phosphorylation sites) has been described previously (34).
Cell culture and transfection
RAW264.7 murine macrophage cells were maintained in media containing RPMI 1640–GlutaMAX medium supplemented with 10% FCS, 50 μM 2-ME, 1 mM sodium pyruvate, 10 mM HEPES, and 2 mM l-Glutamine (Life Technologies).
Plasmid transfections were performed by cell electroporation, using the ElectroBuffer Kit (Cell Projects). For each transfection, cells grown to subconfluence in a 100-mm plate were transfected with 15 μg of plasmid DNA. Cells were electroporated at 250 V, 900 μF in 0.4-cm cuvettes (Bio-Rad Laboratories) using the Gene Pulser Xcell electroporation system (Bio-Rad Laboratories). After electroporation, cells were immediately resuspended in culture medium, transferred into 96-well white OptiPlates previously coated with polylysine (Perkin Elmer), then cultured for 18 h at 37°C and 5% CO2 before treatments.
HEK293T cells were cultured in DMEM and transfected with FuGENE as described previously (34).
RAW264.7 cells (1 × 107 cells) were prepared for chromatin immunoprecipitation (ChIP) assay using HighCell# ChIP kit protein A (Diagenode). After cross-linking with formaldehyde, DNA was sonicated into 200–300-bp fragments, and protein–DNA complexes were immunoprecipitated using either FoxO1-ChIP grade Ab (ab39670; Abcam) or control rabbit IgG. Protein–DNA cross-links were reversed by heating, and precipitated DNA was quantified by real-time PCR using primers (Supplemental Table I) that amplify either Foxo1 binding site 1 or 2 present in the 500 bp upstream of the transcription start site of the mouse GFAT2 gene.
Transfected cells were treated with LPS (100 ng/ml) and/or Thiamet G (10 μM) for 24 h. BRET experiments were then performed exactly as described previously (35, 36), using the Infinite F200 PRO microplate analyzer (Tecan). Briefly, cells were preincubated for 5 min in PBS in the presence of 5 μM coelenterazine. Each measurement corresponded to the signal emitted by the whole population of cells present in a well. The BRET signal was expressed in milliBRET units. The BRET unit has been defined previously as the ratio 530 nm/485 nm, obtained in cells expressing both luciferase and YFP, and corrected by the ratio 530 nm/485 nm, obtained under the same experimental conditions in cells expressing only luciferase (37, 38).
Eight- to twelve-week-old C57BL/6J male mice were used. To study the function of Foxo1 in macrophages, we crossed mice carrying two floxed Foxo1 alleles kindly provided by Prof. S.M. Hedricks, University of California, San Diego) with LysM-Cre transgenic mice (kindly provided by C. Peyssonnaux, Institut Cochin, Paris, France), in which Cre is specifically expressed in the myelomonocytic cell lineage (LysM-Cre-Foxo1/Foxo1 knockout [KO] mice, thereafter referred to as Foxo1-KO mice). To study the role of O-GlcNAcylation on proinflammatory effects of LPS, mice with OGT deletion in the myelomonocytic cell lineage were generated by crossing mice carrying two floxed OGT alleles (obtained from Jackson ImmunoResearch Laboratories) with LysM-Cre transgenic mice (LysM-Cre-OGT KO mice, thereafter referred to as OGT-KO mice). All mice were housed in the Institut Cochin animal facility. All experiments were performed in accordance with accepted standards of animal care, as established in the INSERM and the CNRS guidelines and were approved by the National Ethics Committee (APAFIS N°A751402).
Preparation of bone marrow–derived and peritoneal macrophages from mice
Mice were sacrificed by CO2 asphyxiation. Bone marrow–derived macrophages (BMDM) were prepared from bone marrow cells flushed from femurs and tibias (39). Cells were seeded in sterile Petri dishes at the concentration of 1 × 106 cells/ml in RPMI 1640 medium supplemented with 10% FCS, 100 μg/ml gentamicin, 10 mM HEPES, 1 mM sodium pyruvate, 50 μM 2-ME, and 20 ng/ml mouse M-CSF (Miltenyi Biotec). Cells were differentiated for 6 d, then washed to eliminate nonadherent cells, and macrophages were detached in cold PBS−/− and seeded in six-well plates at 1 × 106 cells/ml in complete media. Cells were treated the day after.
For peritoneal macrophage collection, after mice sacrifice, the peritoneum was infused with 10 ml of cold RPMI 1640. Cells were seeded in six-well plates at 1 × 106 cells/ml in complete media and allowed to adhere overnight. Nonadherent cells were then washed away, and macrophages were treated with LPS.
In some experiments, mice were injected i.p. with LPS (0.6 mg/kg). Six hours after injection, mice were sacrificed, and peritoneal cells were harvested as described above and plated in a six-well plate at 37°C with 5% CO2. After 2 h, nonadherent cells were washed away, and adherent cells were lysed for analysis by Western blot.
To evaluate the effect of LPS on cytokine production in vivo in OGT-KO mice, LPS was injected i.p., mice were sacrificed 6 h after injection, and blood was collected by cardiac puncture. Blood was centrifuged at 1500 rpm for 10 min at 4°C, and the serum was collected and frozen at −80°C for subsequent determination of cytokine concentrations using an MSD kit.
Preparation of human monocyte-derived macrophages
Human primary macrophages were isolated from blood of healthy donors (Etablissement Français du Sang, Ile-de-France, Site Trinité; agreement number INSERM-EFS:18/EFS/030) by density gradient sedimentation on Ficoll (GE Healthcare), followed by negative selection on magnetic beads (StemCells, catalog no. 19059) and adhesion on plastic at 37°C for 2 h. Cells were then cultured in the presence of complete culture medium (RPMI 1640 supplemented with 10% FCS [Eurobio Scientific]), 100 mg/ml streptomycin/penicillin and 2 mM l-glutamine (Life Technologies) containing 10 ng/ml recombinant human M-CSF (R&D Systems) for 4–5 d (40).
RNA extraction, reverse transcription, and quantitative PCR
Macrophages cultured in six-well plates were lysed in TRIzol reagent (Life Technologies). RNA was isolated and reverse transcribed (41). Levels of the cDNA of interest were measured by quantitative PCR (qPCR) (LightCycler FastStart DNA Master SYBR Green 1 Kit; Roche Diagnostics). The absence of genomic DNA contamination was verified by treating RNA samples in parallel without reverse transcriptase and controlling for the absence of amplification by qPCR. Gene expression was normalized over cyclophilin and hypoxanthine-guanine phosphoribosyltransferase RNA levels. The sequences of the primers used for qPCR are given in Supplemental Table I.
OGA enzymatic assay
Control- and LPS-treated RAW264.7 cells were lysed in extraction buffer containing 50 mM Tris, 137 mM NaCl, 1% Triton, 10% glycerol, phosphatase (50 mM NaF, 10 mM disodium β-glycerophosphate, 1 mM Na3VO4), and protease (AEBSF, leupeptine, antipaine aprotinine, and pepstatin, 1 μg/ml each) inhibitors. OGA activity was measured using 4-methylumbellifery-N-acetylß-d-glucosamine (MU-GlcNAc; Sigma-Aldrich), which is converted into fluorescent 4-methylumbelliferon upon hydrolysis by OGA and other hexosaminidases (42). The fluorescence of 4-methylumbelliferon was measured at 448 nm after excitation at 362 nm. Fluorescent measurements were performed after 30 and 60 min of incubation at 37°C to ensure that the determination was performed during the linear phase of the reaction. To specifically determine OGA activity versus other hexosaminidase present in the lysate, all reactions were performed both in absence and presence of 100 μM Thiamet G (a specific OGA inhibitor). The difference between the fluorescent signals obtained in the absence (activity of OGA plus other hexosaminidases) and presence of Thiamet G (activity of the other hexosaminidases) reflected the amount of 4-methylumbelliferon produced by OGA (9).
OGT enzymatic assay
Control- and LPS-treated RAW264.7 cells were lysed in extraction buffer containing 50 mM Tris, 137 mM NaCl, 1% Triton, 10% glycerol, phosphatases (50 mM NaF, 10 mM disodium β-glycerophosphate, 1 mM Na3VO4), and proteases (AEBSF, leupeptine, antipaine aprotinine, and pepstatin, 1 μg/ml each) inhibitors. Immunoprecipitation of OGT was performed as described below. Four hundred micrograms of proteins were incubated with 1.5 μg of anti-OGT Ab (Sigma-Aldrich) for 2 h at 4°C. Precipitation was performed by incubating 25 μl equilibrated protein G-Sepharose beads (GE Healthcare) for 30 min at 4°C. After three washes, the precipitated proteins were submitted to an additional wash in OGT assay buffer containing 50 mM Tris–HCl and 12.5 mM MgCl2, pH7.5. An OGT assay was then performed on protein G-Sepharose–bound OGT (33) using the bioluminescent UDP-Glo glycosyltransferase assay (Promega) exactly as described in the manufacturer instructions (43).
GFAT enzymatic assay
Glutamine fructose-6-phosphate amidotransferase enzymatic activity was measured as described previously (44). Control- and LPS-treated RAW264.7 cells, human monocyte-derived macrophages or mouse BMDM macrophages were lysed in extraction buffer containing 50 mM Tris, 137 mM NaCl, 1% triton, 10% glycerol, phosphatase (50 mM NaF, 10 mM disodium β-glycerophosphate, 1 mM Na3VO4), and protease (AEBSF, leupeptine, antipaine aprotinine, and pepstatin, 1 μg/ml each) inhibitors. Fifteen to thirty micrograms of proteins were incubated in a buffer containing 10 mM fructose 6-phosphate, 6 mM glutamine, 0.3 mM 3-acetylpyridine adenine dinucleotide (APAD), 50 mM KC1, 100 mM KH2PO4 (pH 7.5), 1 mM DTT, and 30 U/ml of glutamate dehydrogenase. After incubation for 30 min at 37°C, the change in absorbance due to reduction of APAD to APADH was monitored spectrophotometrically at 365 nm using Tecan Infinite M1000 PRO microplate reader.
Cells were lysed with buffer containing 50 mM Tris–HCl (pH 8), 137 mM NaCl, 10% (v/v) glycerol, 1% (v/v) Triton, 50 mM NaF, 10 mM disodium β-glycerophosphate, 1 mM Na3VO4, and protease inhibitors (1 μg/ml pepstatin, antipain, leupeptin, aprotinin, and AEBSF) supplemented with 10 μM PUGNAc to preserve the GlcNAcylation state of proteins during the extraction procedure. Proteins were then analyzed by SDS-PAGE, followed by Western blotting, as described previously (45). Clathrin was used as a loading control because its molecular mass (180 kDa) is far away from the 50- to 70-kDa region, where many proteins of interest are found (46, 47).
In some experiments, O-GlcNAcylated proteins were precipitated on wheat germ lectin, as described previously (48, 49). Briefly, cell lysates (400 μg of proteins) were incubated for 2 h at 4°C on a rotating wheel with 20 μl of agarose beads coupled to wheat germ lectin (which binds the GlcNAc pattern). At the end of the incubation, the beads were pelleted by centrifugation at 3000 × g for 2 min. The supernatant was removed, and the pellet was washed three times in extracting buffer. Twenty microliters of Laemmli sample buffer were added to the beads, and the samples were boiled at 95°C for 5 min and then submitted for Western blotting as described previously (48, 49).
Western blots were revealed by chemiluminescence (Thermo Fisher Scientific) and visualized using a FUSION FX7-Vilber Lourmat camera. The signals obtained were then quantified using ImageJ software.
Statistical analyses were performed using PRISM software. Comparison between groups were performed using a Student t test or ANOVA followed by a Dunnett or Tukey posttest, as indicated in the figure legends.
LPS stimulation markedly increased protein O-GlcNAcylation in macrophages
Subcellular relocalization of OGT in different cell compartments has been observed upon stimulation of membrane receptors (50, 51), resulting in compartment-specific changes in O-GlcNAcylation activity. To evaluate whether LPS stimulation affects protein O-GlcNAcylation in different cellular compartments in macrophages, we used BRET-based O-GlcNAc biosensors (33). These biosensors are composed of Rluc8 luciferase fused to a lectin domain (GafD), a known OGT substrate peptide derived from casein kinase II, followed by the Venus variant of the yellow fluorescent protein (Fig. 1B). Upon O-GlcNAcylation, the casein kinase peptide binds to the lectin, resulting in a conformational change detected as an increased BRET signal. These biosensors were fused to addressing sequences for targeting to the internal face of the plasma membrane (using Lyn myristoylation/palmitoylation sequence), the cytosol (using the HIV-1 Rev protein nuclear exclusion sequence), or the nucleus (using the SV40 nuclear localization sequence) (32).
RAW264.7 cells transfected with these biosensors were incubated for 24 h in presence of LPS, thiamet G (an inhibitor of OGA), or both. We observed that LPS treatment increased BRET signal with all three biosensors (Fig. 1B), indicating that LPS stimulation promotes a general rather than compartment-specific increase in O-GlcNAcylation in RAW264.7 macrophages. Interestingly, the effects of LPS and Thiamet G were additive, suggesting that the effect of LPS on O-GlcNAcylation is independent of OGA activity.
LPS-induced O-GlcNAcylation of proteins, both in the absence and presence of Thiamet G, was further demonstrated by Western blotting, using anti–O-GlcNAc Ab in RAW264.7 cells and in mouse BMDM and peritoneal primary macrophages, as well as in human monocyte-derived macrophages (Fig. 1C). Therefore, increased O-GlcNAcylation upon LPS stimulation also appears to occur in primary macrophages from different origins and species, suggesting a general mechanism elicited by TLR4 activation.
GFAT2 is an LPS-inducible gene in macrophages
Using RAW264.7 macrophages, we further explored the mechanism involved in this LPS-induced O-GlcNAcylation. As shown in Fig. 2, LPS-induced increase in protein O-GlcNAcylation (Fig. 2A) was not associated with any detectable change in OGT and OGA mRNA or protein expression (Fig. 2B, 2C), suggesting that it was not mediated by regulation of the expression level of O-GlcNAc–cycling enzymes. GFAT, the enzyme that catalyzes the rate-limiting step of the HBP (Fig. 1A), exists as two isoforms (GFAT1 and GFAT2), encoded by two separate genes (also denominated GFPT1 and GFPT2, respectively) that are differentially expressed in a cell type-specific manner. We observed that in the basal state only GFAT1 protein was highly expressed in macrophages, whereas GFAT2 protein expression was barely detectable (Fig. 2B). GFAT1 protein expression was moderately increased (1.3-fold) by LPS treatment, whereas LPS induced a major increase (7-fold) in GFAT2 protein expression (Fig. 2B). In agreement with these results, Fig. 2C shows that LPS increased the expression of GFAT1 mRNA by 2–3-fold, and markedly induced (by 10–15-fold) the expression of GFAT2 mRNA, suggesting a regulation of GFAT1 and 2 at the transcriptional level. Enzymatic assays indicated that, whereas LPS treatment had no detectable effect on OGT or OGA activities, it significantly increased the activity of GFAT in macrophages (Fig. 2D). These results suggest that the LPS-induced increase in the expression of GFAT, and more specifically GFAT2, translates into an increase in the activity of the rate-limiting step of the HBP. Time course experiments indicated that maximal induction of GFAT2 mRNA was observed 3 h after LPS stimulation, whereas GFAT1 mRNA expression was barely modified at early time points (Supplemental Fig. 1A), revealing that GFAT2 is an early TLR4 target gene. However, despite marked stimulation of GFAT2 mRNA at early time points, GFAT2 protein expression was barely increased after 6 h of LPS treatment, and robust stimulation was only detected after 24 h of treatment with LPS (Supplemental Fig. 1B). Interestingly, GFAT activity was also barely increased after 6 h of treatment and closely followed the increase in GFAT2 protein expression at time 24 h (Supplemental Fig. 1C). In agreement with this, LPS-induced increase in protein O-GlcNAcylation could be detected at 6 h but became statistically significant only at time 24 h (Supplemental Fig. 1D).
Increased GFAT2 protein and mRNA expression was also observed in human monocyte-derived macrophages (Fig. 3A), mouse BMDM (Fig. 3B), and as peritoneal macrophages (Fig. 3C). In agreement with these results, GFAT enzymatic activity was also increased upon LPS stimulation in human and mouse primary macrophages (Supplemental Fig. 2A). Moreover, i.p. injection of LPS in mice also induced a modest increase in GFAT1 and a marked increase in GFAT2 protein expression in peritoneal cells, indicating that this response was also operative in vivo (Supplemental Fig. 2B).
Altogether, our results suggest that LPS induces an increase in protein O-GlcNAcylation through stimulation of the expression of the rate-limiting enzyme of the HBP.
Impaired O-GlcNAcylation promotes proinflammatory response in macrophages
To evaluate the role of LPS-induced O-GlcNAcylation in macrophages, we isolated peritoneal macrophages from mice with conditional deletion of OGT in the myeloid lineage. As expected, OGT expression and LPS-induced O-GlcNAcylation were markedly impaired in OGT-KO cells (Fig. 4A), although residual protein O-GlcNAcylation was sometimes observed, probably reflecting incomplete deletion of OGT by LysM-Cre in resident peritoneal macrophages. We noticed that basal GFAT1 and both basal and LPS-induced GFAT2 expression were increased in OGT-KO macrophages, suggesting a compensatory response to the low level of O-GlcNAcylation in these cells. Interestingly, OGT deletion resulted in an increase in LPS-induced NOS2 expression. Moreover, we observed an increased production of IFN-γ and IL-1β in the culture medium of OGT-KO macrophages. These results suggest that increased O-GlcNAcylation upon LPS stimulation may have a counterregulatory effect that restrains excessive cytokine production by macrophages (Fig. 4B). In agreement with this notion, increases in proinflammatory cytokines IFN-γ, IL-1β, and TNF-α were also observed in vivo in the serum of OGT-KO mice i.p. injected with LPS (Fig. 4C). We also evaluated time dependency of cytokine production by measuring IL-1β and IFN-γ in the culture medium of control and OGT-KO macrophages 6 and 24 h after LPS stimulation (Supplemental Fig. 2C). These experiments indicated that after 6 h of treatment, the LPS-induced increase in IL-1β secretion was higher in culture medium of OGT-KO macrophages than wild-type macrophages. IFN-γ secretion also appeared to be higher at 6 h, although the difference was not statistically significant at this time point (Supplemental Fig. 2C).
These results suggest that increased O-GlcNAcylation upon LPS stimulation participates in the counter-regulation of proinflammatory signaling in macrophages. Induction of GFAT2 expression may therefore constitute a new mechanism to limit excessive inflammation upon LPS stimulation. However, it cannot be excluded that the lack of basal O-GlcNAcylation in OGT-KO macrophages may also participate in the observed alterations on gene expression and cytokine production independently of LPS-induced O-GlcNAcylation, for instance, by perturbing the expression or activity of signaling intermediates or transcription factors that are involved in the regulation LPS proinflammatory effects.
GFAT2 expression is dependent on FoxO1 transcription factor in macrophages
The transcription factor FoxO1 has been shown previously to display dual regulatory effects on inflammatory signals in myeloid cells, with either pro- (52–55) or anti-inflammatory (56–59) effects, depending on the pathophysiological context. Analysis of mouse and human GFAT2 putative promoters revealed canonical FoxO1 recognition sites within the 500- and 800-bp regions upstream of the transcription start site of the mouse and human genes, respectively (Supplemental Fig. 3A). We inserted these upstream sequences in a luciferase reporter gene plasmid. Transfection in HEK293T cells with this plasmid revealed that the activities of both human and mouse reporter genes were markedly increased by cotransfection with a constitutively active form of FOXO1, FOXO1-TM (Fig. 5A).
The mouse 500-bp promoter contains two FoxO1 recognition sites (referred to as site 1 and site 2 on Fig. 5B, Supplemental Fig. 3A). To further demonstrate the contribution of these sites, we mutated either one or both FoxO1 binding sites and measured the activity of the mutated promoters in the luciferase assay.
As shown in Fig. 5B, the FOXO1-TM effect was markedly reduced by mutation of site 1 and totally abolished when both site 1 and site 2 were mutated. This suggests that both sites might be important for regulation of GFAT2 expression by FoxO1. Binding of Foxo1 to site 1 and site 2 was further demonstrated on the endogenous GFAT2 promoter in RAW264.7 cells, in ChIP experiments using a FoxO1-specific Ab (Fig. 5C).
These results suggested that FoxO1 may be involved in the regulation of GFAT2 expression in macrophages. To evaluate the potential involvement of FoxO1 in the regulation of GFAT2 expression, we used a small molecule–specific inhibitor of FoxO1 activity, the AS1842856 drug (60). We observed, both in RAW264.7 cells (Fig. 6A) and human monocyte-derived macrophages (Fig. 6B), that inhibition of FoxO1 by this compound markedly impaired the effect of LPS on induction of GFAT2 mRNA and protein expression. The effect of AS1842856 on GFAT1 mRNA and protein expression was quite variable. In human monocyte-derived macrophages, AS1842856 had no significant effect on GFAT1 mRNA or protein expression. In RAW264.7 cells, whereas AS1842856 appeared to reduce the LPS effect on GFAT1 mRNA expression, it had no significant effect on GFAT1 protein expression. In agreement with a major role of GFAT2 in LPS-induced GFAT activity, we observed that basal GFAT activity was not affected by AS1842856 treatment, whereas inhibition of LPS-induced GFAT2 expression by this compound was associated with a marked inhibition of LPS-induced GFAT activity (Supplemental Fig. 3B).
To further demonstrate the role of Foxo1 in LPS-induced GFAT2 expression, we also evaluated the effect of LPS on peritoneal macrophages from mice with conditional Foxo1 deletion in the myeloid lineage. We observed that LPS-induced GFAT2 protein and mRNA expression was markedly impaired in Foxo1 KO macrophages (Fig. 6C). Again, higher variability was observed concerning GFAT1 expression, but Foxo1 invalidation appeared to have no significant effect on GFAT1 mRNA or protein expression. Inhibition of GFAT2 expression was associated with inhibition of the LPS effect on protein O-GlcNAcylation. These results demonstrate that Foxo1 is necessary for induction of GFAT2 by LPS and support the notion that increased GFAT2 mediates the LPS-induced increase in protein O-GlcNAcylation in macrophages.
Several lines of evidence have suggested a role for O-GlcNAcylation in the regulation of inflammatory processes in macrophages (61). However, contradictory results have been obtained concerning the effect of TLR4 activation on protein O-GlcNAcylation. Indeed, depending on the experimental setting, both increases (22, 23) and decreases (24, 25) in the general O-GlcNAcylation profile were observed upon LPS stimulation.
In the present work, we showed, using two different methodological approaches [BRET-based assay (Fig. 1B) and Western blotting on the crude cell lysate as well as the wheat germ lectin-bound fraction (Figs. 1C, 2A)] that LPS induced a major increase in global protein O-GlcNAcylation in RAW264.7 macrophages. The use of plasma membrane-, cytosol-, or nucleus-targeted BRET O-GlcNAc biosensors indicated that the LPS-induced increase in O-GlcNAcylation was not restricted to a specific cell compartment. Increased protein O-GlcNAcylation upon LPS treatment was also confirmed in primary mice and human macrophages, indicating that our observation was not a cell line-specific effect. The reason underlying the discrepancies between different laboratories are unknown at the present time. However, it is possible that technical differences in the extraction procedure may affect the detection of O-GlcNAc by Western blotting. For instance, although investigators generally add protease and phosphatase inhibitors in their cell lysis buffer, they do not mention the use of any hexosaminidase inhibitor to prevent loss of O-GlcNAc during the extraction procedure. Given that O-GlcNAc is a very dynamic and labile modification that can be rapidly hydrolyzed upon cellular damage or during protein isolation (62), it is quite possible that loss of O-GlcNAc might occur during the sample preparation. In contrast, we always included PUGNAc at a concentration of 10 μM in our extraction buffer to preserve the O-GlcNAcylation state of proteins obtained in cells after LPS treatment. In addition, the confirmation of our Western blotting results by an independent technique based on the use of a BRET biosensor, which monitors O-GlcNAcylation changes in intact living cells without any processing of cellular proteins, strongly argues in favor of an LPS-induced general increase in O-GlcNAcylation in macrophages.
Interestingly, the effects of LPS on protein O-GlcNAcylation were additive to those of a maximally inhibitory concentration of the OGA inhibitor Thiamet G. This suggested that LPS-induced O-GlcNAcylation was not mediated by regulation of OGA activity. In agreement with this notion, using a fluorogenic substrate, we observed that OGA activity was similar in cell extracts from control- and LPS-stimulated RAW264.7 macrophages (Fig. 2D). Moreover, using a luminescent assay (33, 43), we found that LPS treatment had no detectable effect on OGT activity in cell extracts from RAW264.7 cells (Fig. 2D). Therefore, LPS-induced O-GlcNAcylation does not appear to be mediated by regulation of O-GlcNAc–cycling enzymes.
One of the most important findings of our study is that LPS treatment resulted in an increased expression and activity of GFAT (Fig. 2). GFAT is the enzyme that catalyzes the rate-limiting step of the HBP, which eventually leads to the production of UDP-GlcNAc, the substrate used by OGT for protein O-GlcNAcylation. GFAT exists in two isoforms, GFAT1 and GFAT2, encoded by two different genes (63, 64). Although little data are available concerning differential roles of these enzymes, these two isoforms present different tissue distribution, with GFAT1 mRNA being predominantly expressed in the pancreas, placenta, and testis, whereas GFAT2 mRNA were found throughout the CNS (64). Although some differences in the regulation of their catalytic activities by cAMP-induced phosphorylation has been described (65), very little is known about differential regulation of GFAT1 versus GFAT2 expression in different cell types. Most surprisingly, to our knowledge, only one study evaluated the expression of GFAT in a mouse macrophage cell line (ANA-1) in macrophages (66). These authors reported that GFAT1 was constitutively expressed in these cells, and they indicated (as data not shown) that no effect of LPS or IFN-γ on its expression was observed. However, GFAT2 expression was not evaluated in this study. Interestingly, we found that GFAT1 mRNA and protein were indeed expressed at significant levels in resting macrophages, whereas GFAT2 expression was barely detectable. LPS stimulated both GFAT1 and GFAT2 expression, although the stimulatory effect was much higher for GFAT2, whereas the effect of LPS on GFAT1 expression was comparatively much lower and quite variable, depending on the experimental conditions. The LPS-induced increase in O-GlcNAcylation is likely to be, at least in part, mediated by the induction of the expression of the GFAT2 isoform in macrophages. Indeed, several observations argue in favor of a predominant role of GFAT2 in the LPS effect on protein O-GlcNAcylation. First, measurement of GFAT activity 6 and 24 h after LPS stimulation indicated that LPS-induced GFAT activity tightly correlated with the LPS-induced GFAT2 protein expression level (Supplemental Fig. 1B, 1C). Second, the inhibition of FoxO1 in RAW264.7 cells markedly impaired GFAT2 expression, whereas it had no significant effect on GFAT1 expression. Accordingly, basal GFAT activity was not affected by AS1842856, whereas LPS-induced activity was markedly impaired upon inhibition of LPS-induced GFAT2 expression by this compound (Fig. 6A, Supplemental Fig. 3B). Third, in Foxo1 KO macrophages, GFAT1 protein expression was not significantly affected, whereas LPS-induced GFAT2 expression was markedly impaired. This was accompanied by blunted LPS-induced protein O-GlcNAcylation, whereas basal protein O-GlcNAcylation remained essentially unaffected (Fig. 6C). Together, these observations strongly suggest that, whereas GFAT1 is the enzyme that controls HBP under basal conditions and thereby permits protein O-GlcNAcylation in the basal state, GFAT2 induction permits increase in the activity of the pathway and mediates LPS-induced protein O-GlcNAcylation.
In addition to GFAT, it cannot be excluded that other enzymatic steps involved in glucose metabolism in the HBP pathway may be affected by LPS treatment, contributing to an increase in protein O-GlcNAcylation. However, even if LPS stimulated the expression or activity of other enzymes, the resulting increase in the flux through the HBP pathway will eventually be conditioned by the activity of the rate-limiting enzyme GFAT and therefore by the LPS effect on GFAT2 expression.
It is also possible that part of the increase in protein O-GlcNAcylation is mediated by subtle changes in OGT expression that were not detected in our studies. Indeed, we noticed on several occasions a small increase in OGT mRNA or protein level. Although these changes generally did not reach significance, we cannot exclude that modification of OGT expression may also participate in LPS-induced protein O-GlcNAcylation.
Our work clearly demonstrates that GFAT2 is an early LPS-inducible gene in macrophages. Although the mechanism by which LPS stimulates GFAT2 expression remains elusive, we provide strong evidence for a role of FoxO1 in this process. Indeed, using gene reporter as well as ChIP assays, we demonstrated the presence of FoxO1 binding sites on the GFAT2 putative promoter. Moreover, pharmacological inhibition or genetic deletion of FoxO1 in macrophages markedly impaired LPS-induced GFAT2 expression, confirming the involvement of FoxO1 in this regulation. However, the exact mechanism by which LPS stimulates FoxO1 activity in these cells remains to be explored. FoxO1 activity is known to be controlled by regulation of its nucleocytoplasmic localization (67). In cell fractionation experiments, we did not detect any significant change in FoxO1 subcellular localization (Supplemental Fig. 3C), suggesting that other mechanisms must be involved. FoxO1 activity can also be controlled independently of any change in its nuclear localization through various posttranslational modifications (68) as well as interaction with numerous binding partners (69). Clearly, elucidation of the mechanism by which LPS induces FoxO1 activity to stimulate GFAT2 expression deserves further investigations.
Previous studies have shown that O-GlcNAcylation can either promote inflammation or reduce it, according to the cellular context and type of insult. Thus, whereas O-GlcNAcylation has proinflammatory effects in situations of chronic hyperglycemia, it appears to be protective in acute stress conditions such as ischemia–reperfusion injury in the heart (61). We observed that impaired O-GlcNAcylation in OGT-KO macrophages resulted in marked increase in NOS2 expression and proinflammatory cytokine production, suggesting that the O-GlcNAc tone may exert a break on inflammatory processes. Therefore, the rapid induction of GFAT2 expression may constitute a protective mechanism to limit exacerbated inflammation upon LPS stimulation.
In summary, we have shown that LPS stimulation promotes a general increase in protein O-GlcNAcylation in macrophages. This effect is, at least in part, mediated by increased expression and activity of the rate-limiting enzyme of the HBP, GFAT, with the GFAT2 isoform being the most responsive to LPS activation. Indeed, whereas GFAT1 may control the activity of the HBP in the basal state, our work demonstrated that GFAT2 is a TLR4-inducible gene in macrophages, permitting a rapid adaptive response to environmental changes.
We are very grateful to Prof. L. K. Mahal for the cDNA coding for the FRET O-GlcNAc biosensors. We also thank Laura Francese for help in some of the BMDM experiments. This work was performed within the Département Hospitalo-Universitaire Autoimmune and Hormonal Diseases.
This work was supported by the Société Francophone du Diabéte (SFD-2013), the Fondation pour la Recherche Médicale (FRM-DEQ20130326518 and FRM-DEQ20150331744) and by the Programme Inter-Cochin (Institut Cochin, CNRS, Institut National de la Santé et de la Recherche Médicale, Université de Paris). L.B. and H.A.-M. held Ph.D. grants from the CORDDIM-Région Ile-de-France. H.A.-M. was also funded by the Fondation pour la Recherche Médicale. J.-L.S.S. was a recipient of a postdoctoral grant from the Consejo Nacional de Ciencia y Tecnologia, Mexico (CONACYT-CVU 376926).
The online version of this article contains supplemental material.
Abbreviations used in this article:
bone marrow–derived macrophage
bioluminescence resonance energy transfer
hexosamine biosynthesis pathway
glutamine fructose-6-phosphate amidotransferase
glycosylation with O-linked β-N-acetylglucosamine.
The authors have no financial conflicts of interest.