Toll-like Receptor 4–Induced Glycolytic Burst in Human Monocyte-Derived Dendritic Cells Results from p38-Dependent Stabilization of HIF-1α and Increased Hexokinase II Expression

Toll-like receptors are common sensors expressed by dendritic cells (DCs) recognizing pathogen-associated molecular patterns and damage-associated molecular patterns, initiating an appropriate immune response to fight infection (1). TLR signaling triggers human DC maturation, promoting the secretion of proinflammatory cytokines and inducing the expression of costimulatory molecules so that DCs acquire the ability to stimulate naive T cells (2). TLR4 is a receptor for Gram-negative bacteria LPS (3). This receptor can also be activated by other microbial ligands such as viral proteins (4–6) and by damage-associated molecular patterns such as high-mobility group box 1 (HMGB) (7) or oxidized phospholipids (8). TLR4 is also involved in the sensing of nutrients and metabolic stress. Indeed, saturated free fatty acids can activate TLR4 signaling via their binding to the hepatokine Fetuin A, which is a TLR4 ligand (9). Its increased production in steatotic and inflamed liver contributes to the secretion of inflammatory cytokines by monocytes and adipose tissue (10). By regulating the expression of many genes involved in innate immunity and metabolic reprogramming (11), TLR4 is at the crossroads of innate immunity and metabolic inflammation (12) and is involved in the pathogenesis of metabolic diseases such as obesity and type 2 diabetes (13).

Glucose is a major nutrient for cellular bioenergetics production. Glycolysis converts glucose to pyruvate by a series of enzymatic reaction steps, generating 2 mol of ATP per mole of glucose. Intermediary metabolites of glycolysis are precursors also fueling the pentose phosphates, lipids, and amino acid pathways (11). Three rate-limiting enzymes are controlling the glycolytic flux. The first one, the hexokinase (HK), controls the conversion of glucose to glucose 6-phosphate. There are four isoenzymes of HK (HK-I, II, III, IV) encoded by four different genes (HK1, 2, 3, 4). HK1 and HK3 have a large spectrum of tissue expression, whereas HK2 and HK4 have an expression profile that is more restricted to metabolically relevant tissues. HK2 is overexpressed in virtually all cancer cells. Other rate-limiting enzymes of glycolysis are the phosphofructokinase (PFK) and pyruvate kinase. PFK is encoded by three different genes (PFKL, PFKP, and PFKM) whose expression is tissue dependent. Pyruvate kinase is encoded by the PKM gene in immune cells, generating two isoforms (PKM1 and PKM2). The pyruvate produced by glycolysis can be converted to lactate that is excreted by specific monocarboxylate transporters or to oxaloacetate or acetyl-CoA to fuel the tricarboxylic acid cycle, coupling glycolysis to oxidative phosphorylation, for the generation of high amount of ATP in the presence of oxygen. During hypoxia, oxidative phosphorylation is reduced and glycolysis is increased to face energetic needs (14). However, activation of glycolysis can also occur under aerobic conditions, and O. Warburg (15) first discovered that tumor cells have a high rate of glycolysis and that most pyruvate is converted to lactate, even when oxygen is available. Both innate and adaptive immune cells can shift to aerobic glycolysis upon stimulation. Cell metabolism is now appreciated as a key regulator of T cell differentiation, modulating effector and memory functions (16). The transcriptional control of glycolysis in T cells mainly relies on the expression of hypoxia-inducible factor 1 (HIF-1) and the proto-oncogene MYC (17). Accumulation of HIF-1α in normoxic conditions has been observed in several cell types upon LPS stimulation (14, 18, 19). Several molecular mechanisms may result in this accumulation, including activation of the mammalian target of rapamycin complex 1 (mTORC1) (20). In primary human plasmacytoid DCs, increased glycolysis was found to play a key role in the regulation of antiviral functions, including IFN-α production (21). Tolerogenic DCs also show a metabolic signature characterized by high mitochondrial respiration and glycolytic capacity (22). TLR4 stimulation of DCs or macrophages results in increased glycolytic activity, an essential process for their proper activation (23–29). Indeed, a metabolic reprogramming to aerobic glycolysis was described to be associated to a proinflammatory M1 phenotype of murine macrophages, whereas glycolytic activity was unchanged in alternate activated M2 macrophages (11, 30). The molecular mechanisms involved in glycolysis upregulation upon TLR4 stimulation remain to be clarified and may differ according to cell types and functions.

In murine bone marrow–derived DCs (BM-DCs), two different molecular pathways have been depicted for glycolytic reprogramming upon TLR4 stimulation. Activation of TBK1-IKKε and AKT kinases was found to control the early increase of glycolysis, favoring mitochondrial translocation of HK-II, fueling the tricarboxylic acid cycle and fatty acids synthesis (24). In these DCs, the late increase in glycolytic metabolism was proposed to be a survival response to maintain ATP production despite the inhibition of oxidative phosphorylation by NO produced by the inducible NO synthase (25). This pathway is unlikely to be involved in human DCs because this enzyme is not expressed (31), and NO production could not be detected in these cells (27). Studies exploring the molecular mechanisms involved in the regulation of glycolysis in primary human DCs are lacking for a better understanding of the reciprocal interactions between cellular metabolic activity and the functional DC activation state.

Therefore, we explored the pathways involved in the control of glycolytic activity upon TLR4 stimulation of human monocyte-derived DCs (MoDCs). In contrast to murine DCs, increased glycolysis did not rely on the TBK1- or AKT-dependent pathways but on the engagement of a TLR4–p38–MAPK axis stabilizing HIF-1α and upregulating HK2 expression. Interestingly, this pathway does not appear to be involved in the activation of glycolysis triggered by TLR1/2 or TLR2/6 stimulation.

Ultrapure TLR4 ligand LPS from Escherichia coli O111:B4, TLR2/6 ligand peptidoglycan (PGN) from Staphylococcus aureus, and synthetic TLR1/2 ligand Pam₃CSK₄ (Pam), all from InvivoGen (Toulouse, France), were dissolved in sterile PBS. FP7 was synthesized from commercially available d-glucose by multistep organic synthesis according to published procedures (32). The purity of the molecule was assessed by nuclear magnetic resonance and mass spectrometry analysis. A 5-mM stock solution of FP7 was prepared in ethanol/DMSO 1:1 and conserved at −20°C.

The inhibitors SB203580 (SB), SP600125, PD98059, BX795, and LY294002, purchased from InvivoGen, rapamycin, provided by Calbiochem (San Diego, CA), and HIF-1α DNA-binding inhibitor, echinomycin (Echino), provided by Cayman Chemical (Ann Arbor, MI), were dissolved in sterile culture grade DMSO and stored at −20°C. AZD5363 and SC75741 were purchased from Selleckchem (Houston, TX), prepared in DMSO, and stored at −80°C. Pepinh MYD, pepinh TRIF, and pepinh control, purchased from InvivoGen, were dissolved in PBS and stored at −20°C.

Glucose uptake was measured by incorporation of 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-d-glucose (2-NBDG; Sigma-Aldrich, Saint-Quentin Fallavier, France) in culture medium.

Monocytes were purified from human peripheral blood of healthy donors (33), obtained from the Etablissement Français du Sang. All experiments were performed in accordance with the guidelines of the World Medical Association’s Declaration of Helsinki. Experimental procedures were approved by the local institutional review committee. Briefly, PBMCs were isolated by standard density gradient centrifugation on Ficoll–Hypaque (Eurobio, Courtaboeuf, France). Mononuclear cells were separated from PBLs by centrifugation on a 50% Percoll solution (GE Healthcare, Velizy, France). Monocytes were purified by immunomagnetic depletion using Pan Mouse IgG Dynabeads (Thermo Fisher Scientific, Villebon sur Yvette, France) with a mixture of mAbs anti-CD19 (4G7 hybridoma), anti-CD3 (OKT3 hybridoma; American Type Culture Collection, Manassas, VA), and anti-CD56 (NKH1; Beckman Coulter, Villepinte, France). Monocyte purity was >90% as assessed by CD14 labeling without CD3⁺, CD19⁺, and CD56⁺ contaminating cells (data not shown).

DCs were differentiated from monocytes cultured at 37°C under 5% CO₂ atmosphere for 6 d at 10⁶ cells/ml in RPMI 1640 medium with GlutaMAX (Thermo Fisher Scientific) supplemented with 10% FCS (PAN-Biotech, Aidenbach, Germany), 40 μg/ml gentamicin, human recombinant GM-CSF, and IL-4 (PeproTech, Neuilly-Sur-Seine, France). Six days later, DCs were harvested, washed twice, and resuspended for treatment at 10⁶ cells/ml fresh RPMI 1640 medium supplemented with 40 μg/ml gentamicin and 10% FCS (complete medium). DCs were >95% pure as assessed by CD14 and CD1a labeling.

DCs were treated with inhibitors or control solvent for 30 min before addition of TLR ligands or the same volume of PBS as control. Pretreatment of DCs with inhibitory peptides, pepinh MYD, TRIF, or control was carried out 6 h before LPS addition. For quantitative PCR analysis, cells were collected after 6 or 24 h incubation. Supernatants and cells were collected after 24 h for glucose assay, cytokines assay, phenotyping, or cell lysate preparation for Western blot and HK assay.

Monocyte and DC phenotypes were determined after labeling with FITC-labeled anti-CD14, -CD19, –HLA-DR, -CD80, -CD54 and PE-labeled anti-CD3, -CD56, -CD1a, -CD86, -CD83, and -CD40 (Beckman Coulter). Flow cytometry analysis was carried out using a FACSCanto II (BD Biosciences, Le Pont de Claix, France). For each double labeling, at least 5000 events in the live cell gate were acquired. Viability of cells was monitored using propidium iodide staining (BD Biosciences) of dead cells.

Clarified culture supernatants were collected 24 h after treatment and stored at −20°C. IL-8, MIP-1β, IL-6, IL-10, IL-12p40, and TNF-α were assayed using quantitative cytokine-specific Cytometric Bead Array Flex Sets (BD Biosciences).

Metabolites were quantified in freshly clarified cell supernatants and control medium using glucose (HK) and lactate assay kits (Sigma-Aldrich) according to the manufacturers’ instructions. Quantifications were normalized to DC cell number at day 6.

For total RNA isolation, 6- or 24-h treated DCs were collected and washed twice with PBS. RNA extraction was performed using the NucleoSpin RNA Mini kit from Macherey-Nagel according to the manufacturers’ instructions. cDNA were synthetized using RNA-to-cDNA kit (Applied Biosystems, Villebon sur Yvette, France) before quantitative PCR experiment using the following primers: for HK1 DNA forward 5′-GCCTCTTATTTGAAGGGCGG-3′, reverse 5′-GACACAGTCATCATCGGACG-3′; for HK2 DNA forward 5′-TCCCCTGCCACCAGACTA-3′, reverse 5′-TGGACTTGAATCCCTTGGTC-3′; for HK3 DNA forward 5′-CATCGTGGACTTCCAGCAGAAG-3′, reverse 5′-CTTGGTCCAGTTCAGGAGGATG-3′; for HK4 DNA forward 5′-CAGAAGGCTCAGAAGTCGGG-3′, reverse 5′-TGGTGTTTGGTCTTCACGCT-3′; for PKM1 DNA forward 5′-CGAGCCTCAAGTCACTCCAC-3′, reverse 5′-GTGAGCAGACCTGCCAGACT-3′; for PKM2 DNA forward 5′-ATTATTTGAGGAACTCCGCCGCCT-3′, reverse 5′-ATTCCGGGTCACAGCAATGATGG-3′ (34); for PFKL DNA forward 5′-AAGAAGTAGGCTGGCACGACGT-3′, reverse 5′-GCGGATGTTCTCCACAATGGAC-3′; for PFKP DNA forward 5′-AGGCAGTCATCGCCTTGCTAGA-3′, reverse 5′-ATCGCCTTCTGCACATCCTGAG-3′; for PFKM DNA forward 5′-GCTTCTAGCTCATGTCAGACCC-3′, reverse 5′-CCAATCCTCACAGTGGAGCGAA-3′; for TBP DNA forward 5′-CCACGAACCACGGCACTGATTT-3′, reverse 5′-CAGTCTGGACTGTTCTTCACTCTT-3′; and SYBR Green master mix on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Marnes-la-Coquette, France). Levels of mRNA for specific genes are reported as relative gene expression normalized to the housekeeping gene TBP.

The method used for extracting HK from DCs was adapted from Kuang et al. (35), and the assay was a modification of that described by Ramière et al. (36). −80°C frozen pelleted cells were homogenized (100 μl/2 × 10⁶ cells) in precooled homogenization buffer (50 mM Tris–HCl, 250 mM sucrose, 5 mM EDTA [pH 7.4], 5 mM 2-ME, 10 mM glucose, 0.2% Triton X-100). After a 20 min incubation on ice, homogenates were pulse-sonicated 5 s at half power of the device (EpiShear Probe Sonicator; Active Motif, La Hulpe, Belgium). Homogenates were then centrifuged at 500 × g for 20 min at 4°C. Supernatants were immediately used for determination of HK activity. HK activity was measured spectrophotometrically through NADP⁺ reduction in the glucose 6-phosphate dehydrogenase–coupled reaction. HK activity was assayed in medium containing 50 mM triethanolamine (pH = 7.6), 550 mM d-Glucose, 100 mM MgCl₂, 14 mM NADP⁺, and 125 U/ml glucose 6-phosphate dehydrogenase (Saccharomyces cerevisiae), equilibrated to 37°C. The reaction was started by addition of ATP (final concentration 19 mM), and absorbance was continuously recorded for 30 min at 340 nm (TECAN Infinite M200). Dilutions of purified recombinant HK (Sigma-Aldrich) were used as standards, and results were expressed as international unit per milligram of proteins.

Day 6 DCs were pretreated in water bath at 37°C with inhibitors for 30 min before TLR ligand addition for 15 min. TLR stimulation was stopped on ice before two cold PBS washes. Then, cells were fixed with Cytofix solution (BD Biosciences) for 10 min at 37°C before permeabilization with methanol 90% for 30 min on ice. Phosphoprotein staining was performed by incubation 1 h at room temperature with the following anti-human Abs: purified anti–phospho-p38 MAPK (Thr180/Tyr182; Cell Signaling Technology, Leiden, the Netherlands), PE-conjugated anti-phospho–NF-κB (Ser536; Cell Signaling Technology), purified anti–phospho-AKT (Ser473; Cell Signaling Technology), PE-conjugated anti–phospho-SAPK/JNK (Thr183/Tyr185; Cell Signaling Technology). When unlabeled primary Ab was used, cells were further incubated 1 h at room temperature with secondary Alexa Fluor 488–conjugated anti-rabbit Ab (Thermo Fisher Scientific). Cells were analyzed using a FACSCanto II (BD Biosciences).

DCs were lysed in buffer containing 20 mM Tris-HCl, 180 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, and 1× mixture of protease inhibitors (Sigma-Aldrich). Lysate proteins were separated by NuPAGE 4–12% Bis-Tris Gel (Thermo Fisher Scientific) in MOPS running buffer. Gels were transferred to nitrocellulose membranes with the Trans-Blot Turbo Transfer system (Bio-Rad). Membranes were blocked for 1 h with PBS buffer containing 0.1% Tween 20 and 5% nonfat dry skim milk and incubated 1 h with the primary Ab against HIF-1α (Novus Biologicals, Lille, France) or actin (clone AC-40; Sigma-Aldrich). Membranes were washed and incubated 1 h with HRP-coupled goat polyclonal anti-rabbit IgG or HRP-coupled goat polyclonal anti-mouse IgG (Jackson ImmunoResearch Laboratories, Suffolk, U.K.). Western blotting experiments were developed using SuperSignal West Femto or Pico Chemiluminescent Substrate (Thermo Fisher Scientific). Chemiluminescent signals were acquired and quantified with a LAS4000 Imager (GE Healthcare).

Data are expressed as mean ± SEM. The Student t test was used for comparisons of two sample means. Multiple group comparisons were performed by one-way ANOVA followed by Bonferroni multiple comparisons test and was applied to the analysis of all dose-dependent experiments. A p value <0.05 was considered statistically significant.

MoDCs stimulated by LPS for 24 h showed an increased glucose consumption and lactate production compared with nonstimulated MoDCs (Fig. 1A, 1B), indicative of an increased glycolytic activity. The amplitude of this glycolytic burst was dependent on the dose of LPS and was statistically significant from 10 to 1000 ng/ml. We previously characterized a selective TLR4 antagonist, FP7, inhibiting the secretion of proinflammatory cytokines by LPS-stimulated MoDCs with an IC₅₀ below 1 μM (27). FP7 also prevented LPS-induced secretion of IFN-β (Supplemental Fig. 1), suggesting that FP7 blocked both endosomal TRIF–dependent and surface MyD88–dependent TLR4 signaling (37). Because FP7 competes with LPS for binding to the TLR4 MD-2 subunit (32, 38) and prevents ligand-mediated TLR4 internalization (F.A. Facchini, D. Di Fusco, S. Barresi, A. Minotti, F. Granucci, G. Monteleone, F. Peri, and I. Monteleone, manuscript in preparation), these results suggest that FP7 may not directly inhibit endosomal TLR4 signaling but rather inhibits upstream activation of TLR4 receptor by LPS. Treatment of cells with 10 μM FP7 prior to LPS stimulation prevented the increase in glucose consumption and lactate production triggered by LPS from 0.1 to 100 ng/ml (Fig. 1A, 1B), showing that this glycolytic regulation was TLR4 dependent. At the highest LPS concentration (1000 ng/ml), FP7 did not totally block the increase of the glycolytic flux, consistent with previous data (27). Using a fluorescent deoxyglucose as a nonhydrolyzable substrate of HK (2-NBDG), we measured the capacity of glucose uptake by the cells after 6 and 24 h LPS stimulation. No difference in the uptake could be detected between control and LPS-stimulated MoDCs (Fig. 1C), indicating that increased glycolytic activity in TLR4-stimulated MoDCs was not due to an increased availability of intracellular glucose. We thus analyzed the expression of the three rate-limiting glycolytic enzymes, HK, PFK, and PKM, at the mRNA level. Compared with control cells, HK2 was strongly induced in LPS-stimulated MoDCs, whereas HK3 was reduced (Fig. 1D). HK1 remained unchanged, whereas HK4 was not expressed in MoDCs. In contrast to PFKM and PKM1, PFKL, PFKP, and PKM2 were highly expressed in MoDCs but not significantly augmented upon LPS stimulation (Fig. 1E, 1F). The data indicate that the expression profile of HK isoenzymes is modified upon TLR4 stimulation. This new expression profile correlates with a major increase in HK activity in total lysates of LPS-stimulated MoDCs compared with unstimulated cells (Fig. 1G).

To investigate the molecular mechanisms involved in the regulation of glycolysis upon TLR4 stimulation, we screened a panel of inhibitors described to target main actors of the TLR4 signaling pathway. A scheme positioning the inhibitors and their target is presented on a simplified view of the TLR4 signaling pathway (Fig. 2). TLR4 signaling in MoDCs results in the phenotypic and functional maturation of these cells, inducing cytokine secretion and enhancing the expression of surface maturation markers. MoDCs were treated with these drugs prior to LPS stimulation for 24 h, and secretion of proinflammatory cytokines, expression of phenotypic maturation marker, and glucose consumption were measured.

As expected, p38–MAPK inhibitor SB inhibited the secretion of IL-6, IL-8, and MIP-1β (Fig. 3A) (39, 40). Echino strongly reduced IL-8 secretion, and the inhibitor of TBK1, BX795, inhibited MIP-1β secretion induced by LPS. Surface expression of the MHC molecule HLA-DR, costimulation molecules CD86 and CD40, and the CD54 adhesion molecule was analyzed by FACS (Fig. 3B, 3C). SB significantly inhibited the expression of CD86 and CD54 and reduced (although not significantly) CD40 expression compared with LPS-stimulated MoDCs. SC75741 and AZD5363 mainly reduced CD54 expression. The other inhibitors did not induce significant changes in MoDC phenotype.

Among the tested inhibitors, p38–MAPK inhibitor SB was the most potent inhibitor of the glycolytic burst triggered by LPS (Fig. 4A). Echino, which prevents DNA binding of HIF-1α to hypoxia response elements, also significantly reduced the action of LPS on glucose consumption (Fig. 4A). The other drugs did not significantly reduce glucose consumption by LPS-stimulated cells. The inhibitors showed no significant toxicity (Fig. 4B). These results indicate that p38–MAPK and HIF-1α are major actors of the increased glucose consumption triggered by LPS and suggest an unexpected link between them. By blocking dimerization of MyD88 or TRIF, we observed that both TLR adaptors were recruited to trigger the glycolytic burst upon TLR4 stimulation (Supplemental Fig. 2), pointing to a common downstream pathway that may involve TRAF6, a key component upstream of p38–MAPK.

BX795 altered the secretion of cytokines and/or the expression of phenotypic marker without affecting LPS-triggered regulation of glycolysis (Figs. 3A–C, 4A). In contrast, treatment by the HIF-1α antagonist Echino resulted in both reduced LPS-induced glycolytic activity and IL-8 secretion (Figs 3A, 4A), indicating that reduced glycolytic burst may contribute to the reduction of cytokines secretion or maturation markers expression. Accordingly, we previously reported that inhibition of the glycolytic burst by 2-deoxyglucose (2-DG) treatment affected both cytokine secretion and phenotypic marker expression induced by TLR4 stimulation (27).

MoDC stimulation by LPS resulted in a strong increase of HK2 expression that can be detected as soon as 6 h after LPS addition (Fig. 5A). At this early time, HK3 expression was unchanged compared with control cells, whereas it was decreased 24 h poststimulation (Fig. 1D). HK1 remained constant (Fig. 5A). Using the inhibitors that reduced glucose consumption by LPS-treated MoDCs, we found that SB and Echino strongly reduced HK2 transcription induced by LPS (Fig. 5A). SB treatment of MoDCs inhibited the phosphorylation of p38–MAPK without affecting the phosphorylation of NF-κB and JNK, as expected (Fig. 5B). HIF-1α is an important transcription factor for the regulation of HK2 expression (41). It is constitutively expressed, and the regulation of its degradation controls its accumulation and activity (42, 43). In total lysates from MoDCs, we observed that LPS stimulation for 24 h resulted in an important accumulation of HIF-1α that was prevented by SB treatment (Fig 5C, 5D). Therefore, HIF-1α accumulation triggered by LPS appeared to be dependent on p38–MAPK activation.

As other bacterial ligands can induce MoDCs functional activation, we tested the effect of PGN and Pam, the ligands for TLR2/6 and TLR1/2, respectively, on glucose consumption. Both TLR ligands significantly increased glucose consumption of MoDCs, although to a different level (Fig. 6A). PGN, like LPS, induced a strong increase of glucose consumption, whereas Pam was less efficient (50% of the LPS increase). Inhibition of glucose consumption by SB reached 70 and 65% when it was induced by LPS–TLR4 and Pam–TLR1/2, respectively, whereas it was only 35% after PGN–TLR2/6 stimulation. p38–MAPK phosphorylation was detected upon stimulation of all these TLRs and was inhibited by SB (Fig. 6B). The inhibition of this kinase by SB strongly impacted the secretion of cytokines triggered by TLR4. Cytokine secretion elicited by TLR2/6 or TLR1/2 tended to be reduced by SB but to a lesser extent (Fig. 6C, 6D). Therefore, the amplitude of glycolytic burst and of its inhibition by SB indicates that glycolysis regulation by TLRs is unlikely to result from a common mechanism.

Under the same conditions, we analyzed cells content in HIF-1α and observed that TLR1/2 and TLR2/6 stimulation of MoDCs, by Pam or PGN, respectively, resulted in a weak accumulation of HIF-1α compared with TLR4 stimulation by LPS (Fig. 6E). SB treatment did not modify the amount of HIF-1α upon TLR2/6 and TLR1/2, in contrast to TLR4 stimulation. In line with this observation, the inhibition of HIF-1α transcriptional activity by Echino did not significantly impact glucose consumption triggered by PGN and Pam (Fig. 7), suggesting that this pathway does not play a major role in glycolysis regulation by TLR2/6 and TLR1/2. As expected, glycolysis inhibition by 2-DG reduced glucose consumption triggered by PGN and Pam. The screening of other inhibitors showed a weak contribution of TBK1 in the upregulation of glucose consumption by PGN, whereas PI3K, AKT, and mTOR inhibition did not significantly alter the glycolytic burst (Fig. 7). The screening did not allow the identification of a pathway involved in the glycolytic regulation upon TLR1/2 stimulation by Pam.

Therefore, the regulation of HIF-1α stabilization upon PGN or Pam stimulation did not depend on p38–MAPK activation. TLR2/6 and TLR1/2 stimulation resulted in a slight accumulation of p38-independent HIF-1α and increased glucose consumption. In contrast, TLR4 regulation of glycolysis in human MoDCs implicates p38-MAPK–dependent stabilization of HIF-1α (Fig. 8).

The present study aimed at characterizing the molecular mechanisms involved in the metabolic reprogramming of human MoDCs triggered by TLR4 stimulation. We showed that TLR4-dependent glycolysis activation in these cells was correlated to an increased global HK activity and an increased expression of HK2. HIF-1α, a master transcriptional regulator of glycolytic enzymes, accumulated in TLR4-stimulated MoDCs. An inhibitor of the transcriptional activity of HIF-1α inhibited both TLR4-induced HK2 expression and glucose consumption. Data are highlighting the role of p38–MAPK activation in HIF-1α stabilization and glycolysis regulation upon TLR4 stimulation. Our study has analyzed in detail the molecular pathways of glycolysis regulation in this model of immunostimulatory DCs (44, 45). Malinarich et al. (22) compared the metabolic profile of tolerogenic DCs and immunostimulatory MoDCs and found that tolerogenic DCs had the highest glycolytic capacity upon ATP synthase inhibition, whereas they had a similar glycolytic rate to that of LPS-stimulated MoDCs. Similarly to us, they found that LPS stimulation enhanced lactate secretion and glycolytic rate of MoDCs without affecting glucose uptake.

Previous results with mouse DCs showed that activation of TBK1-IKKε and AKT kinases controlled the early increase in glycolysis triggered by TLR4 (24). In our experimental conditions, TBK1 and AKT inhibitors did not impact the glycolytic activity of LPS-stimulated MoDCs (Fig. 4A). Moreover, no phosphorylation of AKT could be detected after LPS stimulation of MoDCs (cf. Supplemental Fig. 3). Pam induced AKT phosphorylation (cf. Supplemental Fig. 3); however, inhibiting PI3K or AKT kinases did not significantly alter the glycolytic reprogramming of MoDCs upon TLR1/2 or TLR2/6 stimulation (Fig. 7). TLR2/6 stimulation by PGN induced an important increase of glucose consumption by MoDCs, which was significantly reduced by TBK1 inhibition (Fig. 7). This is consistent with recent data showing that TBK1 is involved in TLR2 signaling (46). Our results indicate that TBK1 contributed to this metabolic regulation by mechanisms different from those described for murine BM-DCs (24) because AKT was not involved. Thus, glycolysis activation by TLR4, TLR2/6, and TLR1/2 appears to engage distinct signaling pathways in human MoDCs.

Major differences in the regulation of glycolytic activity have been reported between mouse and human DCs. mTOR was shown to be an important regulator of LPS-induced murine DC activation and glucose consumption, but not in human myeloid DCs (47). In agreement with these results, we observed that the mTORC1 inhibitor rapamycin did not impact the upregulation of glycolysis upon LPS stimulation in MoDCs (Fig. 4). Moreover, human MoDCs and macrophages do not produce any NO (27, 31), an inhibitor of the oxidative phosphorylation that was shown to boost the glycolytic pathway in mouse DCs (25). Therefore, results obtained from murine BM-DCs or macrophages cannot be readily extrapolated to human DCs.

Previous works indicated that LPS stimulation of mouse BM-DCs and human MoDCs resulted in HIF-1α accumulation (14, 18, 19). In normoxia, HIF-1α degradation is regulated by hydroxylation of proline and asparagine residues by prolyl-hydroxylase domain enzymes (PHDs). The interaction of the von Hippel–Lindau factor with these hydroxylated residues recruits an E3 ubiquitin–ligase that targets HIF-1α to the proteasome for degradation (48). Hypoxia induces HIF-1α accumulation via PHD inhibition due to the lack of its oxygen cosubstrate (18). However, data describing the molecular mechanisms controlling HIF-1α accumulation in primary cells upon inflammation in normoxic conditions are sparse. In murine macrophages, LPS-induced HIF-1α accumulation seems to require NF-κB– and ERK-dependent transcriptional events (49, 50). In other studies, reactive oxygen species production and succinate accumulation upon LPS stimulation inhibit PHD activity, thus increasing HIF-1α stability (51, 52). In murine BM-DCs, sequestration of iron, a PHD cofactor, stabilizes HIF-1α upon LPS stimulation (18). The AKT–mTOR–HIF-1α pathway was found to upregulate glycolysis in murine monocytes (20). In human MoDCs, our results point to a different molecular mechanism involving a p38-MAPK–dependent HIF-1α stabilization. This is crucial to the glycolytic regulation upon TLR4 activation, but p38 activity is not involved in the stabilization of HIF-1α upon TLR2/6 stimulation. Although p38–MAPK phosphorylation is induced by TLR4, TLR1/2, and TLR2/6 (Fig. 6B), the consequence on HIF-1α accumulation varies. This differential regulation points to an indirect control of HIF-1α stability by p38–MAPK. The molecular mechanism explaining the differential regulation of HIF-1α accumulation in MoDCs according to TLR stimulation remains to be discovered. In tumor cells, previous studies showed that p38–MAPK could contribute to HIF-1α stabilization (53, 54). One possible mechanism proposed by Khurana et al. (55) in mouse embryonic fibroblasts is that the RING finger ubiquitin ligase Siah2 can be phosphorylated by p38–MAPK, increasing the degradation of PHD3, therefore reducing HIF-1α degradation.

In both mouse and human models, HIF-1α stabilization alone was not sufficient to induce DC maturation, but it positively contributes to their functionality upon TLR stimulation (14, 19). In mouse BM-DCs, HIF-1α induction by the combined effects of LPS and hypoxia plays a key role in the regulation of glucose consumption, DC maturation, and their ability to stimulate allogeneic T cells (14). In human MoDCs, HIF-1α stabilization could synergize with TLR stimulation to favor maturation (19).

The HIF-1α DNA-binding inhibitor, Echino, reduced the secretion of IL-8 by LPS-stimulated DCs (Fig. 3A). IL-8 secretion highly depends on enhanced gene transcription, and IL-8 promoter mainly recruits NF-κB, AP-1, and NRF (56) but does not contain HIF response elements. The reduction of IL-8 secretion by Echino treatment may thus result from the limitation of the glycolytic burst of MoDCs. Indeed, previous results indicate that HK inhibition by 2-DG suppressed increased glycolysis and reduced the secretion of many cytokines, and especially IL-8, upon LPS-stimulation (27, 29). By reducing the glycolytic burst triggered by LPS, Echino may impede optimal maturation and especially impact maturation processes that require active gene transcription, such as cytokine neosynthesis. Reduced expression of phenotypic maturation markers of MoDCs was observed with 2-DG, which strongly inhibits glycolysis (27), but not with Echino, which only partially reduces LPS-induced glycolytic burst (Fig. 3B, 3C). Glycolysis inhibition by 2-DG strongly reduced the secretion of inflammatory cytokines and altered the expression of some phenotypic maturation markers triggered by PGN and Pam (Supplemental Fig. 4), reinforcing previous results highlighting the importance of the glycolytic burst for TLR-induced DC maturation (27, 29). Reduction by Echino of both the glycolytic burst (Fig. 7) and the secretion of proinflammatory cytokines (Supplemental Fig. 4) triggered by PGN and Pam was not statistically significant. Although we cannot totally exclude a contribution of HIF-1α, this pathway did not appear to play a major role in glycolysis regulation by TLR2/6 and TLR1/2.

Increasing evidence indicates that metabolic pathways play important roles in the fine tuning of immune cell functions. Detailed studies have been conducted on T cell differentiation and mouse macrophage functional polarization according to their metabolic state (11). The signaling of many receptors can result in the regulation of metabolic pathways by controlling the activation of transcription factors and the expression of metabolic enzymes. Some transcription factors such as HIF-1α regulate the expression of both metabolic enzymes and proinflammatory cytokines such as IL-1β in mouse M1 macrophages (52) and TNF-α in human monocytes (20). The concentration of metabolites varying according to the metabolic state of the cells can also impact their immune function. These metabolic signals can be detected by various sensors and are important regulators of inflammation (57).

Our results also show that the increased HK activity after 24 h stimulation by TLR4 of human MoDCs is correlated to increased HK2 gene expression, whereas HK3 was downregulated, HK1 did not change, and HK4 was not expressed. These results differ from those obtained in murine bone marrow–derived macrophages, in which HK1 but not HK2 was upregulated via mTORC1 activation upon LPS and ATP stimulation (58). This difference, however, is in agreement with the lack of effect of the mTORC1 inhibitor rapamycin on glycolysis upregulation in MoDCs. HKs catalyze the first step of glucose metabolism, producing glucose 6-phosphate, which is a precursor for glycolysis but also for the pentose phosphate and hexosamine biosynthetic pathways. Thus, HK plays a key role in glycolysis and biosynthetic pathways. HK-II is the predominant isoenzyme in insulin-sensitive tissues such as heart, adipose tissue, and skeletal muscle, and it is also upregulated in many types of tumors associated with enhanced aerobic glycolysis (Warburg effect), suggesting that it is a major regulator of the glycolytic rate in these cells (59). HK-II activity is upregulated by its binding to the outer mitochondrial membrane protein voltage-dependent anion channel 1 (VDAC1). Our results indicate that HK2 upregulation also plays a key role in metabolic regulation of human primary DCs. Although the molecular mechanisms triggered by TLR4 to stimulate glycolysis are different in humans MoDCs and mouse BM-DCs, they can both involve HK-II protein modulation (24). In the future, HK-II intracellular localization should be investigated in MoDCs in response to TLR4 stimulation. Moreover, the contribution of other nonenzymatic functions of HK-II should also be analyzed. Indeed, overexpression and mitochondrial association of HK-II confer protection to apoptotic or necrotic stimuli in different cell types by several mechanisms (59). Increased expression of HK2 upon LPS stimulation may contribute to prosurvival effects of LPS stimulation of MoDCs. HK-II, but not other HKs, bind and inhibit mTORC1 in the absence of glucose, facilitating autophagy in response to glucose starvation. Thus, HK-II can protect cells from cellular damage and provide energy by recycling intracellular constituents (60).

Metabolomics of immune cells recently established that TLR4 induces glycolytic reprogramming and a drastic rewiring of tricarboxylic acid cycle to generate inflammatory intermediates that contribute to macrophages and DCs activation. Dysregulated TLR4-triggered inflammatory response can be involved in pathologic conditions such as sepsis (61), neuropathic pain (62), amyotrophic lateral sclerosis (63), some autoimmune diseases such as rheumatoid arthritis (64), or obesity-associated type 2 diabetes (65). Our results unravel a novel molecular mechanism linking TLR4 signaling and glycolysis activation and, together with previous reports, reinforce the idea that a better understanding of mechanism underlying TLR4-induced glycolytic reprogramming should offer innovative therapeutic windows.

We acknowledge the contribution of Mélanie Tonnerre for technical assistance, of the Etablissement Français du Sang Auvergne-Rhône-Alpes, and SFR Biosciences (UMS3444/CNRS, US8/Inserm, ENS de Lyon, Université Claude Bernard Lyon 1) facility AniRA-Cytométrie.

The authors have no financial conflicts of interest.