TLR-driven early glycolytic reprogramming via the kinases TBK1-IKKε supports the anabolic demands of dendritic cell activation

TLR agonists induce a rapid increase in glycolysis by DCs

We analyzed GM-DCs for real-time changes in the rate of extracellular acidification (ECAR), as a measure of lactate production (a surrogate for the glycolytic rate), and the mitochondrial rate of oxygen consumption (OCR), directly following stimulation with lipopolysaccharide (LPS). While the OCR remained stable after stimulation, there was a rapid increase in the ECAR (Fig. 1a); this was independent of iNOS (Supplementary Fig. 1). However, consistent with published work⁶, the long-term commitment to glycolysis was dependent on NO, since in the presence of a general inhibitor of NOS, the TLR-induced increase in the glycolytic rate returned toward baseline by 9 h after stimulation, whereas in the absence of the inhibitor, the ECAR remained elevated beyond that time (Supplementary Fig. 1). We confirmed the rapid induction of glycolysis by LPS by measuring increases in extracellular lactate concentrations (Fig. 1b) and glucose consumption (Fig. 1c). Those results were further supported by analysis of 1,2-¹³C-glucose tracing by gas chromatography–mass spectrometry, with which we observed more rapid incorporation of glucose-derived carbon into pyruvate and lactate after stimulation with LPS for 1 h than in unstimulated (control) conditions (Fig. 1d). Finally, global profiling of metabolites in GM-DCs that had been stimulated for 1 and 3 h with LPS revealed selective LPS-induced changes in the abundance of metabolic intermediates of glycolysis and associated pathways, such as the pentose phosphate pathway (PPP) and tricarboxylic acid (TCA) cycle (Fig. 1e and Supplementary Table 1). Other TLR agonists induced a similar response (Fig. 1f). In in vitro cultures of conventional SIRPα⁺ and CD24⁺ DCs differentiated from bone marrow in the presence of the growth factor Flt3L, the ECAR was also rapidly increased after stimulation of TLRs (Fig. 1g). Thus, rapidly increased glycolysis was a cardinal feature of the TLR-induced DC activation.

DC activation depends on an early increase in glycolysis

We sought to determine whether the early increase in glycolysis was important for DC activation. We stimulated GM-DCs with LPS in the presence of 2-deoxyglucose (2-DG), which inhibits hexokinase activity⁸; 2-DG effectively blocked glycolysis, as shown by the lack of an increase in the ECAR after stimulation with LPS (Fig. 2a). The LPS-induced surface expression of the costimulatory receptor CD40 and costimulatory molecule CD86 and MHC class I and class II and production of interleukin 6 (IL-6), IL-12p70 and tumor-necrosis factor (TNF) at 6 h were substantially impaired by 2-DG (Fig. 2b,c and Supplementary Fig. 2a). Notably, that regulation seemed to be entirely at the translational level, as mRNA encoding each of those proteins was induced equivalently in GM-DCs stimulated with LPS in the presence and absence of 2-DG (Fig. 2d). The production of effector cytokines in T cells is controlled post-transcriptionaIly by aerobic glycolysis through a mechanism that involves the binding of GAPDH to AU-rich elements in the 3′ untranslated region of mRNA encoding interferon-γ (IFN-γ), which results in impaired translation of IFN-γ ⁹. Given the presence of AU-rich elements in the 3′ untranslated region of Il6, Il12b and Tnf mRNA¹⁰, we assessed whether glycolysis controls cytokine expression in DCs in a similar manner. However, the binding of Il6, Il12b and Tnf transcripts to GAPDH was very low in DCs after stimulation with LPS and did not increase much in the presence of 2-DG (Supplementary Fig. 2b), which suggested that the observed post-transcriptionaI regulation of cytokine expression was controlled by other processes. iNOS is not yet expressed at that time point after LPS stimulation, and mitochondria are active⁶, and consistent with the presence of still-functional mitochondria, cell viability was unaffected by the presence of 2-DG (Supplementary Fig. 2c). We observed a similar impairment in the TLR-induced expression of surface markers in 2-DG-treated DCs generated from bone marrow in the presence of Flt3L (Supplementary Fig. 2d). Furthermore, blocking glycolysis in iNOS-deficient (Nos2^−/−) GM-DCs suppressed the LPS-induced expression of the chemokine CCL19 receptor CCR7 (Fig. 2e) and consequently impaired migration toward CCL19 (Fig. 2f). In addition, inhibition of glycolysis compromised the ability of Nos2^−/− GM-DCs to induce T cell proliferation (Fig. 2g) and IFN-γ secretion (Fig. 2h), although notably, under these conditions, the frequency of T cells that produced IL-17A and expressed the transcription factor Foxp3 increased (Fig. 2h). Thus, an increase in aerobic glycolysis early after stimulation with TLR agonists was required for the activation and function of DCs.

Glycolysis feeds the TCA cycle to support DC activation

TLR signaling drives the bioenergetically demanding expression of a large set of genes¹¹. Since we observed that ATP concentrations in GM-DCs dropped by ~40% within 2 h after stimulation with LPS (Supplementary Fig. 3a), we first determined whether that increase in glycolysis was needed to provide extra ATP for DC activation. For cells that depend on glycolysis as their chief source of ATP, the recycling of NADH generated during glycolysis back to NAD⁺ by the conversion of pyruvate into lactate catalyzed by the enzyme lactate dehydrogenase A (LDHA) is essential for maintenance of a high glycolytic flux. However, we found that although inhibition of LDHA by oxamate blocked the early increase in the ECAR (Supplementary Fig. 3b), it had no effect on DC activation (Supplementary Fig. 3c,d), which challenged the proposal of an essential role for glycolytic ATP in DC activation.

On the basis of those findings, we tested the alternative hypothesis that glucose was serving as a key source of carbon for other anabolic programs that underpinned the DC activation. Since metabolite profiling revealed that DC activation was associated with changes in pyruvate metabolism and the TCA cycle (Fig. 1e), we hypothesized that the increase in glycolysis may have served to generate more pyruvate to feed the TCA cycle. In this case, an increase in glycolysis would be expected to be associated with coincident changes in mitochondrial activity. We found that the mitochondrial membrane potential and spare respiratory capacity, a measure of mitochondrial fitness^12,13, increased rapidly after stimulation with LPS (Fig. 3a,b), indicative of increased mitochondrial activity. Consistent with those findings, tracing of 1,2-¹³C-glucose by gas chromatography–mass spectrometry revealed more rapid incorporation of glucose-derived carbon into TCA cycle intermediates in cells stimulated for 1 h with LPS than in control cells left unstimulated for that same period of time (Fig. 3c). In addition, the LPS-induced increase in the spare respiratory capacity was prevented by 2-DG (Fig. 3d), which indicated that the increase in mitochondrial fitness was mediated by that enhanced flux of glycolytic carbon into the TCA cycle. To address whether flux of pyruvate into the TCA cycle is important for DC activation, we suppressed expression of the mitochondrial pyruvate carrier MPC-1 (ref. 14) (Fig. 3e). GM-DCs transduced with short hairpin RNA (shRNA) specific for Mpc1 (which encodes MPC-1) were less activated after stimulation for 6 h with LPS than were GM-DCs transduced with control shRNA (Fig. 3f,g); pharmacological inhibition of MPC-1 with α-cyanocinnimate resulted in a phenocopy of those results (Supplementary Fig. 4a,b). In contrast, when we inhibited fatty acid oxidation or glutaminolysis, two other chief potential sources of carbon for the TCA cycle, we noted no major effects on DC activation (Supplementary Fig. 4a,b). The proposal of the importance of glycolysis for fueling the TCA cycle for DC activation was further supported by data from experiments in which we manipulated the flux of pyruvate into lactate or acetyl-CoA through the use of retroviral transduction to promote or inhibit the expression of LDHA and the pyruvate dehydrogenase kinase PDK1. We found that preventing the flux of pyruvate (and therefore glucose-derived carbon) into the TCA cycle by overexpressing LDHA and PDK1 inhibited activation, whereas reducing expression of those enzymes led to more activation (Supplementary Fig. 4c).

DC activation depends on glycolysis-fueled synthesis of fatty acids

Inhibition of the mitochondrial production of ATP by oligomycin had a minimal effect on the early activation of GM-DCs by LPS (Supplementary Fig. 5a,b). That result indicated that the increase in the flux of glucose-derived carbon into the TCA cycle in response to stimulation via TLRs was not needed to fuel oxidative phosphorylation to generate ATP for DC activation. We therefore reasoned that the increase in glycolysis served to replenish intermediates of the TCA cycle that were being extracted for other biosynthetic processes important for DC activation. Consistent with that, we found that intracellular concentrations of citrate and isocitrate, but not of subsequent intermediates in the TCA cycle, dropped after stimulation with LPS (Fig. 4a). Moreover, the percentage of glucose-derived carbon was significantly lower in intermediates of the TCA cycle downstream of citrate than in citrate itself (Fig. 3c), which indicated ‘preferential’ removal of citrate from the TCA cycle during DC activation. Citrate is one of the main sources of carbon for the de novo synthesis of fatty acids¹⁴, which requires the export of citrate from the mitochondria into the cytosol via the citrate shuttle Slc25a1 (ref. 15). Citrate is subsequently converted to acetyl-CoA, followed by its conversion by acetyl-CoA carboxylase into malonyl-CoA, which is then used for the elongation of fatty acid chains by fatty acid synthase. Our finding of lower citrate concentrations in activated DCs led us to hypothesize that the rapid increase in glycolytic flux after stimulation via TLRs was needed to support the de novo synthesis of fatty acids. Consistent with that, stimulation with LPS resulted in the accumulation of lipids in Nos2^−/− DCs (Fig. 4b) that was diminished when we inhibited either glycolysis or the de novo synthesis of fatty acids by treatment with 2-DG or the fatty-acid-synthase inhibitor C75, respectively (Fig. 4c). Moreover, stimulation of GM-DCs with LPS resulted in increased incorporation of glucose-derived carbon into lipids that was prevented by C75 (Fig. 4d). We addressed the importance of the synthesis of fatty acids from citrate for DC activation by measuring the effect of LPS on GM-DCs in which expression of Slc25a1 had been inhibited by retrovirally introduced shRNA (Fig. 4e) or on GM-DCs were incubated with the acetyl-CoA carboxylase inhibitor TOFA or with C75. We found that each of those three treatments impaired the LPS-induced activation of GM-DCs (Fig. 4f–h), including secretion of the inflammatory lipid mediator prostaglandin E₂ (Supplementary Fig. 5c).

Next we addressed the role of the de novo synthesis of fatty acids in TLR-driven activation of DCs. De novo–synthesized fatty acids may act as ligands for transcription factors, such as peroxisome proliferator-activated receptors, that could regulate TLR-induced gene expression^16,17. However, rosiglitazone, a synthetic ligand for peroxisome proliferator-activated receptor-γ, did not abrogate the defect in LPS-driven activation when fatty acid synthesis was inhibited (Supplementary Fig. 5d). Consistent with that finding, inhibition of fatty acid synthesis had no effect on the abundance of mRNA encoding cytokines or costimulatory molecules (Fig. 4i), which suggested post-transcriptional regulation of LPS-induced gene expression by fatty acid synthesis. Those observations led us to hypothesize that de novo lipogenesis may be important for the synthesis of additional membranes to expand the endoplasmic reticulum (ER) and Golgi to support the increase in the synthesis, transport and secretion of proteins in response to stimulation via TLRs. We addressed this possibility by using transmission electron microscopy to visualize ER and Golgi. We found that both compartments were expanded considerably in GM-DCs at 3 h after stimulation with LPS and that this process was significantly attenuated by the inhibition of fatty acid synthesis (Fig. 4j,k). These data suggested that TLR-driven glycolysis in DCs, through the generation of citrate, supported the de novo synthesis of fatty acids to promote the expansion of ER and Golgi networks required for DC activation.

Glycolysis-driven PPP is required for DC activation

The PPP, an offshoot of the glycolytic pathway, has a crucial role in the generation of NADPH, which is critical for the synthesis of fatty acids as well as for the maintenance of cellular redox balance by controlling the concentration of reduced glutathione (GSH) and the production of ribose-5-phosphate used for nucleotide synthesis. Metabolite analysis indicated considerable changes in the PPP after DC activation (Fig. 1e). Specifically, consistent with a high demand for products of the PPP, we observed that intracellular concentrations of NADPH (Fig. 5a) and ribose-5-phosphate (Fig. 5b) dropped rapidly in DCs after stimulation with LPS. Indeed when we suppressed expression of glucose-6-phosphate dehydrogenase, the first enzyme in the PPP (Fig. 5c), LPS-driven expression of proinflammatory cytokines (Fig. 5d) and lipid accumulation (Fig. 5e) were diminished. Those results suggested that the flux of glucose into the PPP had a crucial role in DC activation by supporting fatty acid synthesis through the generation of NADPH. Together these data indicated that the increased flux of glucose-derived carbon into the TCA cycle and PPP was essential for TLR agonist–induced DC activation and suggested an important role for the de novo synthesis of fatty acids from citrate, fueled by glycolysis, for LPS-driven activation of DCs.

TBK1-IKKε controls the early increase in glycolysis

Signaling via growth factors and insulin receptors leads to Akt-dependent increases in glycolysis¹⁸. Moreover, Akt is essential for the sustained metabolic commitment to glycolysis that is apparent in iNOS-expressing DCs following activation⁴. We reasoned, therefore, that Akt would be essential for the early increase in glycolysis induced by TLR agonists. We found that stimulation of GM-DCs with LPS led to the rapid phosphorylation of Akt at both Thr308 and Ser473, necessary for full Akt activation¹⁹, and phosphorylation of PRAS40, a direct downstream target of Akt (Fig. 6a). Consistent with the proposal of a crucial role for Akt in the metabolic response to stimulation via TLRs, two distinct inhibitors of Akt (triciribine and Akt1/2 VIII) substantially blunted the early increase in glycolysis and concomitant DC activation induced by LPS (Fig. 6b–d and Supplementary Fig. 6a). Unexpectedly, whereas inhibition of the canonical Akt activators phosphatidylinositol-3-OH kinase (PI(3)K) and mTOR diminished the LPS-induced phosphorylation of Akt at Ser473 and of PRAS40, on the basis of both immunoblot analysis (Fig. 6e and Supplementary Fig. 6c) and flow cytometry (Supplementary Fig. 6b), that was not sufficient to impair the increase in glycolysis (Fig. 6f and Supplementary Fig. 6a). Akt has been identified as a substrate for TBK1 and IKKε²⁰, two closely related kinases downstream of various TLRs²¹ that can be activated independently of PI(3)K^22,23 and are known for their importance in the induction of type I interferons. We found that after stimulation with LPS, TBK1 was rapidly phosphorylated in GM-DCs in a PI(3)K-independent manner (Fig. 6g and Supplementary Fig. 6e); TBK1 was also phosphorylated in response to the TLR2 ligand Pam₃CSK₄ (Supplementary Fig. 6d). Moreover, GM-DCs in which TBK1 and IKKε were inhibited by BX795 showed substantial impairment in the phosphorylation of Akt at Ser473 and Thr308 and of PRAS40 and the ribosomal protein S6, and this was more profound than the inhibition observed after inhibition of PI(3)K or mTOR (Fig. 6e and Supplementary Fig. 6b,c). Consistent with that, in contrast to inhibition of PI(3)K or the metabolic checkpoint kinase complex mTORC2, inhibition of TBK1-IKKε blunted the glycolytic response (Fig. 6f) after stimulation with LPS. Those effects were recapitulated in GM-DCs derived from IKKε-deficient (Ikbke^−/−) bone marrow in which TBK1 was knocked down by shRNA (Fig. 6h,i and Supplementary Fig. 6f) and in GM-DCs treated with amlexanox, a different inhibitor of TBK1-IKKε²⁴ (Supplementary Fig. 6g). Together these data indicated TBK1 and IKKε were the main kinases upstream of Akt that regulated metabolic reprogramming in DCs after stimulation by TLR agonists.

Akt-driven activation of HK-II is critical for early glycolysis

One mechanism by which Akt has been postulated to enhance glycolytic flux is by promoting the association of HK-II with voltage-dependent anion channels located in the outer mitochondrial membrane, where presumably it can directly use mitochondrial ATP^25,26. The stimulation of GM-DCs with LPS resulted in enrichment of phosphorylated HK-II in the mitochondrial fraction and increased overall hexokinase activity, effects that were prevented by the inhibition of Akt and of TBK1-IKKε (Fig. 7a,b and Supplementary Fig. 7a). That increase in hexokinase activity was functionally critical, as treatment with a cell-permeable peptide corresponding to the mitochondrial and voltage-dependent anion channel–binding motif of HK-II, which competitively dissociates HK-II from the mitochondria^27,28 (Fig. 7c), prevented the LPS-induced increase in hexokinase activity (Fig. 7d) and impaired the rapid burst in glycolysis (Fig. 7e) and associated DC activation (Fig. 7f,g). Those effects were comprehensively recapitulated by clotrimazole, a drug that dissociates hexokinases from mitochondria²⁷ (Supplementary Fig. 7b–f).

Akt can also enhance glycolytic flux directly by promoting the translocation of glucose transporters to the cell membrane²⁹ and/or by enhancing the enzymatic activity of the glycolytic enzyme PFK1 by promoting the activity of PFK2, its allosteric activator³⁰. However, after stimulation via TLRs, we observed neither a rapid increase in the surface expression of GLUT1 (Slc2a1), the dominant glucose transporter expressed by these cells, nor a substantial effect of the knockdown of PFK2 on the LPS-induced rapid increase in glycolysis or on DC activation (Supplementary Fig. 8a–d). In addition, Akt can enhance glycolytic flux indirectly via mTORC1 through activation of the transcription factor HIF-1α, which drives the expression of several glycolytic enzymes, including HK-II and LDHA³¹. However, the TLR-driven early glycolytic response was unaffected in GM-DCs that either were treated with the mTORC1 inhibitor rapamycin or were deficient in raptor, a key component of mTORC1 (Supplementary Figs. 6a,b and 8e,f). Moreover, during the rapid TLR-driven increase in glycolysis, the expression of HK-II and LDHA protein remained constant (Supplementary Fig. 8g), and neither global inhibition of transcription with actinomycin D nor knockdown of HIF-1α via shRNA had an effect on TLR-induced early metabolic reprogramming (Supplementary Fig. 8h–k), which suggested that neither mTORC1 nor HIF-1α had a prominent role in the rapid, Akt-driven changes in glycolytic metabolism that occurred in DCs stimulated by TLR agonists. Together these data suggested that TBK1-IKKε and Akt were essential for the increased glycolytic flux after stimulation by TLR agonists by directly promoting the association of HK-II with mitochondria.

Early glycolysis is important for DC function in vivo

To investigate whether DCs rapidly increased glucose uptake in response to stimulation via TLRs in vivo, we gave mice injections of 2-NBDG, a fluorescent analog of 2-DG, after intravenous challenge with LPS. Injection of LPS resulted in a reproducible albeit slight increase in the uptake of 2-NBDG by splenic CD11b⁺ DCs but not by CD8α⁺ DCs (Fig. 8a). However, both CD11b⁺ and CD8α⁺ DCs showed an increase in the ECAR in response to stimulation via TLRs ex vivo (Fig. 8b). The finding that the increase in the ECAR in response to LPS relative to the baseline ECAR was significantly lower in CD8α⁺ DCs than in CD11b⁺ DCs, coupled with the relative insensitivity of 2-NBDG labeling in vivo, may explain why we were unable to detect an increase in glucose uptake by CD8α⁺ DCs in vivo (Fig. 8a). As for signaling, we found that similar to results obtained for GM-DCs, the TLR-induced glycolytic response in ex vivo splenic CD11b⁺ DCs was dependent on Akt (Fig. 8c), TBK1-IKKε (Fig. 8d) and the association of HK-II with mitochondria (Fig. 8e) but was not dependent on mTOR or PI(3)K (Fig. 8d). Furthermore, in vivo activation of both DC subsets by LPS was significantly impaired when we gave mice simultaneous injection of 2-DG to block glycolysis (Fig. 8f). Likewise, ex vivo secretion of IL-12 in response to LPS was diminished in the presence of 2-DG in both DC subsets (Fig. 8g). Moreover, DCs activated by LPS in vivo in the presence of 2-DG were impaired in their ability to prime ovalbumin (OVA)-specific CD4⁺ and CD8⁺ T cells in vivo (Fig. 8h). In addition, DCs in which fatty acid synthase was inhibited during LPS-driven activation were considerably compromised in their ability to prime CD8⁺ T cell responses in vivo (Fig. 8i,j), which demonstrated an essential role for TLR-induced glycolytic metabolism and downstream fatty acid synthesis in the immunological priming function of DCs in vivo.

Mice

CD45.1⁺ and CD45.2⁺ wild-type mice, Nos2^−/− mice, Ikbke^−/− mice, Rptor^flox/flox mice, Itgax^cre mice and C57BL/6 mice with transgenic expression of OVA-specific CD4⁺ T cells (OT-II) or CD8⁺ T cells (OT-I) were from The Jackson Laboratory. Mice were maintained at Washington University under specific pathogen–free conditions under protocols approved by the Institutional Animal Care and Use Committee and were used for experiments between 6 and 12 weeks of age.

In vitro DC differentiation

Bone marrow–derived DCs were generated as described⁴⁵. Bone marrow cells were differentiated for 7 d in the presence of GM-CSF (20 ng/ml) in RPMI-1640 medium containing 10% FCS, 100 U/ml penicillin-streptomycin and 2 mM -glutamine (‘complete DC medium’) or for 8 d with Flt3L (160 ng/ml) in DMEM containing 10% FCS, 100 U/ml penicillin-streptomycin and 2 mM -glutamine. Bone marrow was cultured for 8 d with Flt3L, then CD24⁺SIRP1α⁻ and SIRP1α⁺CD24⁻ DCs were isolated with a FACSAria II (BD). To rule out potential secondary effects of TLR-induced iNOS expression on DC metabolism as a result of interference with oxidative phosphorylation in GM-DCs by iNOS-derived NO⁶, Nos2^−/− GM-DCs were used for all experiments in which DCs were studied more than 6 h after activation, when expression of iNOS begins (i.e., cell death, CCR7 expression, migration, lipid staining and in vitro and in vivo T cell–priming studies).

In vivo studies

For 2-NBDG-labeling experiments, mice received 2-DG (4 g per kg body weight) or vehicle control intravenously 30 min before receiving LPS (5 μg) intravenously; 1 h later, 2-NBDG (50 μg) was injected and 15 min later mice were killed and splenic DCs were analyzed by flow cytometry (in the fluorescein isothiocyanate channel) for the uptake of 2-NBDG. For in vivo activation studies, mice were killed 6 h after of LPS and splenic DCs were analyzed by flow cytometry for the expression of activation markers (antibody identification below). Splenic DCs were isolated as described⁶. To increase yields of splenic DCs for in vivo T cell–priming studies and ex vivo metabolic analyses, mice were given subsequent injection of 4 × 10⁶ Flt3L-expressing B16 tumor cells 12 d before isolation of DCs. For in vivo T cell–priming studies, splenic DCs were purified from C57BL/6 mice pretreated with LPS with or without 2-DG, then the cells were incubated for 45 min at 37 °C ex vivo with 1 μg/ml synthetic OVA peptide (amino acids 323–339 or 257–264). Then those cells were washed and injected into congenic mice that had received 1 × 10⁶ CellTrace Violet–labeled OT-I and OT-II cells 1 d previously. The proliferation of T cells in the spleen was determined by dilution of CTV, as measured by flow cytometry, 60 h later. Endogenous responses of OVA-specific CD8⁺ T cells were analyzed in draining lymph nodes 7 d after footpad injection of GM-DCs that had been washed three times following treatment for 6 h with LPS and OVA (1 mg/ml of endotoxin-free egg white prepared in the laboratory) in the presence or absence of C75.

DC culture

DCs were cultured at a density of 2 × 10⁵ cells per well in 200 μl complete DC medium and, where appropriate, were preincubated for 30 min with peptides MIASHMIACLFTELN(β-Ala)GYGRKKRRQRRG-amide (HKII), and GYGRKKRRQRRG-amide (control) (10 μM) from Selleckchem, α-cyanocinnamate (500 μM), 2-DG (10 mM), oligomycin (1 μM), etomoxir (200 μM), 6-diazo-5-oxo-L-norleucine (DON; 100 μM), clotrimazole (25 μM), amlexanox (200 μM), TOFA (5-(tetradecyloxy)-2-furoic acid; 20 μM), C75 (20 μM), actinomycin D (1 μg/ml) and rosiglitazone (5 μM) from Sigma-Aldrich; triciribine (20 μM), LY294006 (20 μM) and BX795 (5 μM) from Calbiochem, S-ethyl-isothiourea (500 μM) from Cayman Chemical; Akt1/2 VIII (5 μM), wortmannin (1 μM) and eapamycin (100 nM) from Millipore; and KU0063794 (100 μM) from Tocris Biochemicals. Subsequently, DCs were stimulated for 6 h (unless stated otherwise) with LPS (Escherichia coli serotype 0111:B4; 100 ng/ml; Sigma-Aldrich); the TLR1-TLR2 agonist Pam₃CSK₄ (1 μg/ml), the TLR2-TLR6 agonist Pam₂CSK₄ (1 μg/ml), the TLR3 agonist poly(I:C) (10 μg/ml), the TLR7-TLR8 agonist R-848 (1 μg/ml) and the TLR9 agonist CpG-B (5 μg/ml; all InvivoGen). Cells used for intracellular cytokine staining were stimulated in the presence of 10 μg/ml brefeldin A (Sigma). The concentrations of cytokines and prostaglandin E₂ in supernatants of DCs cultured for 6 h with or without LPS and with or without inhibitors were determined by cytometric bead array (BD) or enzyme immunoassay (Cayman), respectively, according to the manufacturer’s instructions.

Retroviral transduction

Retroviral transduction of DCs was accomplished as described⁴. Sequences for shRNA were obtained from Open Biosystems and were cloned into LMP retroviral vectors expressing either human CD8 or green fluorescent protein. For gene-overexpression studies, gene sequences were cloned into MSCV IRES retroviral vectors expressing human CD8 or green fluorescent protein. Transduced cells were gated or sorted on the basis of the expression of human CD8 or green fluorescent protein.

Migration assay

CCL19 (0.2 μg/ml) in complete DC medium was added to the bottom chamber of 24-well Transwell plates (polycarbonate filter with a 5-μm pore; Costar). DCs (4 × 10⁵ cells per Transwell) were added to the upper chamber, followed by incubation for 2 h at 37 °C. Transmigrated cells were collected from the lower chamber and counted on a flow cytometer.

Proliferation and polarization of T cells

DCs were preincubated for 30 min with the appropriate inhibitors and were subsequently treated for 6 h with LPS and OVA (1 mg/ml of endotoxin-free egg white prepared in the laboratory), then were washed three times. DCs were cultured at various ratios with 5 × 10⁴ CD62L^hiCD44^lo naive OT-II or OT-I T cells labeled with CFSE (carboxyfluorescein diacetate succinimidyl ester). On day 3, T cell proliferation was assessed by flow cytometry as dilution of CFSE. Cytokine production by T cells was determined by intracellular staining (antibody identification below) on day 6 after restimulation for 5 h with PMA (phorbol 12-myristate 13-acetate) and ionomycin in the presence of brefeldin A.

Metabolism assays

For real-time analysis of the ECAR and OCR, DCs were analyzed with an XF-96 Extracellular Flux Analyzer (Seahorse Bioscience) as described⁶. ATP concentrations were determined with an ATP Determination Kit according to the manufacturer’s instructions (Invitrogen). The concentration of glucose and lactate in the medium, cellular ratio of NAPD⁺ to NADPH and hexokinase activity were assessed with a Glucose Assay Kit and Lactate Assay Kit (Eton Bioscience), NADPH quantification Kit (Biovision) and hexokinase colorimetric assay (Biovision), respectively, according to each manufacturer’s recommendations. Metabolon did the global metabolite profiling of stimulated DCs, and the Metaboanalyst 2.0 web-based pipeline was used for metabolic-pathway-enrichment analysis. For 1,2-¹³C-glucose–tracing experiments, DCs were left unstimulated or stimulated for 1 h with LPS, in the presence of 1,2-¹³C-glucose (Cambridge Isotope Laboratories) or ¹²C-glucose. Metabolites from cultured cells were extracted and then were analyzed by NMR and gas chromatography–mass spectrometry as described^46,47. Glucose-derived lipid biosynthesis was assessed by culture of DCs (2 × 10⁶) for 6 h in complete DC medium containing 2μM U-¹⁴C-glucose (250 mCu/mmol; PerkinElmer). Lipids were extracted with a solution of water, methanol and chloroform at a ratio of 1:1:1, and the organic layer was isolated, dried with a Speedvac and resuspended in a solution of methanol and chloroform at a ratio of 1:1, and incorporated radioactivity was measured with a MicroBeta Liquid Scintillation Counter (PerkinElmer).

Flow cytometry

Antibodies used for flow cytometry analysis were as follows: anti-CD172a (P84), anti-CD24 (M1/69), anti-CD11c (N418), anti-CD8α (53–6.7), anti-CCR7 (4B12), vTNF (TN3–19; all from eBioscience); anti–MHC class II (I-A/E^b; M5/114.15.2; Biolegend); anti-CD11b (M1/70), anti-CD86 (GL1), anti-CD40 (HM40-3), anti–MHC class I (H-2K^b; AF6-88.5), anti-CD44 (IM7), anti-IL-12p40/70 (C15.6), anti-IL-6 (MP5-20F3), antibody to Akt phosphorylated at Thr308 (J1-223.371) and antibody to Akt phosphorylated at Ser473 (M89-61; all from BD). Ovalbumin-specific CD8⁺ T cells were quantified by being stained with MHC–peptide tetramers of H-2K^b and OVA amino acids 257–264. For intracellular staining, DCs were fixed in 4% ultrapure paraformaldehyde, then were stained for 1 h at 20 °C in 0.2% saponin buffer. Mitochondrial membrane potential and lipid content were assessed with 20 μM DiOC₆ (3,3′-dihexyloxacarbocyanine iodide; Invitrogen) and 0.5 μg/ml BODIPY 493/503 (Invitrogen) in the fluorescein isothiocyanate channel, following 20 min of incubation at 37 °C. Global transcription was assessed by a Click-it RNA HCS assay according to the manufacturers’ instructions (Invitrogen) and was analyzed by flow cytometry. Samples were analyzed on a BD Canto II Flow Cytometer.

Cell fractionation and immunoblot analysis

Mitochondria-enriched extracts were prepared from 5 × 10⁶ cells with a Mitochondria Isolation Kit according to the manufacturer’s instructions (Mitosciences). Cell lysate preparation, SDS-PAGE, electrophoretic transfer, immunoblot analysis and development with enhanced chemiluminescence were done as described⁴⁸. Immunoblot analysis used anti-β–actin (13E5), anti–hexokinase II (C64G5), anti-IKKε (D61F9), anti-TBK1 (D1B4), anti-prohibitin-1 (2426), anti-Akt (C67E7), anti-α-tubulin (11H10), antibody to TBK1 phosphorylated at Ser172 (D52C2), antibody to Akt phosphorylated at Thr308 (D25E6), antibody to Akt phosphorylated at Ser473 (D9E), antibody to S6 phosphorylated at Ser235 and Ser236 (D57.2.2E), anti-S6 (5G10), antibody to PRAS40 phosphorylated at Thr246 (C77D7), anti-PRAS40 (D23C7), anti-raptor (24C12) and anti-LDHA (2012; all from Cell Signaling), at a dilution of 1:1,000 or 1:2,000 (S6), followed by detection with horseradish peroxidase–linked antibody to rabbit IgG (1:2,000 dilution; 7074; Cell Signaling). An antibody to residues specifically phosphorylated by Akt (phospho-Akt substrate; 110B7E; Cell Signaling) was used for analysis of the phosphorylation of hexokinase by Akt.

Quantitative RT-PCR

RNA was isolated with an RNeasy Kit (Qiagen), and single-strand cDNA was synthesized with a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The Taqman method was used for real-time PCR with primers from Applied Biosystems and an Applied Biosystems 7000 sequence-detection system. The expression of mRNA was normalized to that of mRNA encoding β-actin.

Immunoprecipitation

DCs (20 × 10⁶ to 30 × 10⁶) were harvested and then were lysed in lysis buffer (10 mM HEPES, pH 7.0, 100 mM KCl, 5 mM MgCl₂, 0.5% NP-40, 1 mM DTT and 0.2 mg/ml heparin). Lysates were precleared, then were immunoprecipitated overnight with rabbit polyclonal antibody to GAPDH (G9545; Sigma). Immunocomplexes were recovered through the use of protein A agarose (Invitrogen) and were washed once with a low-salt solution, a high-salt solution and LiCl (Sigma) and twice with Tris-EDTA buffer, and were eluted with 1% SDS and 0.1 M NaHCO₃. Crosslinkage of mRNA encoding GAPDH was reversed by incubation of the eluted fraction for 1 h at 42 °C, followed by an additional 2 h at 65 °C in the presence of 200 mM NaCl and 20 μg proteinase K. RNA was then extracted from the eluted fraction with TRIzol, and total RNA was quantified with Nanodrop (Thermo Scientific) or mRNA expression was measured by quantitative RT-PCR as described above.

Transmission electron microscopy

For ultrastructural analysis, cells were fixed for 1 h at 65 °C in 2% paraformaldehyde and 2.5% glutaraldehyde (Polysciences) in 100 mM sodium cacodylate buffer, pH 7.2. Samples were washed in cacodylate buffer and then were post-fixed for 1 h in 1% osmium tetroxide (Polysciences). Samples were then rinsed extensively in distilled H₂O before en bloc staining for 1 h with 1% aqueous uranyl acetate (Ted Pella). After several rinses in distilled H₂O, samples were dehydrated in a graded series of ethanol and were embedded in Eponate 12 resin (Ted Pella). Sections (95 nm in thickness) were cut with an Ultracut UC7 ultramicrotome (Leica Microsystems), then were stained with uranyl acetate and lead citrate and viewed on a 1200 EX transmission electron microscope (JEOL USA) equipped with an AMT eight-megapixel digital camera (Advanced Microscopy Techniques).

Statistical analysis

Data were analyzed with Graphpad Prism software (version 5). Two-group comparisons were assessed using unpaired or, where indicated, paired two-tailed Student’s t-tests. The use of these tests was justified on the basis of assessment of normality and variance of the distribution of the data. Differences were considered significant when P values were below 0.05. The identities of micrographs were removed before analysis. No randomization or exclusion of data points was used. Pilot in vivo studies were used for estimation of the sample size required to ensure adequate power.