Research Article

mTOR- and HIF-1α–mediated aerobic glycolysis as metabolic basis for trained immunity

See allHide authors and affiliations

Science  26 Sep 2014:
Vol. 345, Issue 6204, 1250684
DOI: 10.1126/science.1250684

Structured Abstract

Introduction

Trained immunity refers to the memory characteristics of the innate immune system. Memory traits of innate immunity have been reported in plants and invertebrates, as well as in mice lacking functional T and B cells that are protected against secondary infections after exposure to certain infections or vaccinations. The underlying mechanism of trained immunity is represented by epigenetic programming through histone modifications, leading to stronger gene transcription upon restimulation. However, the specific cellular processes that mediate trained immunity in monocytes or macrophages are poorly understood.

Graphic

Aerobic glycolysis as metabolic basis for trained immunity. In naïve macrophages during aerobic conditions, glucose metabolism is mainly geared toward oxidative phosphorylation providing adenosine triphosphate (ATP) as the energy source. In contrast, long-term functional reprogramming during trained immunity requires a metabolic shift toward aerobic glycolysis and is induced through a dec tin-1–Akt–mTOR–HIF-1α pathway.

Methods

We studied a model of trained immunity, induced by the β-glucan component of Candida albicans, that was previously shown to induce nonspecific protection against both infections and malignancies. Genome-wide transcriptome and histone modification profiles were performed and pathway analysis was applied to identify the cellular processes induced during monocyte training. Biological validations were performed in human primary monocytes and in two experimental models in vivo.

Results

In addition to immune signaling pathways, glycolysis genes were strongly upregulated in terms of histone modification profiling, and this was validated by RNA sequencing of cells from β-glucan–treated mice. The biochemical characterizations of the β-glucan–trained monocytes revealed elevated aerobic glycolysis with reduced basal respiration rate, increased glucose consumption and lactate production, and higher intracellular ratio of nicotinamide adenine dinucleotide (NAD+) to its reduced form (NADH). The dectin-1–Akt–mTOR–HIF-1α pathway (mTOR, mammalian target of rapamycin; HIF-1α, hypoxia-inducible factor–1α) was responsible for the metabolic shift induced by β-glucan. Trained immunity was completely abrogated in monocytes from dectin-1–deficient patients. Blocking of the mTOR–HIF-1α pathway by chemical inhibitors inhibited trained immunity. Mice receiving metformin, an adenosine monophosphate–activated protein kinase (AMPK) activator that subsequently inhibits mTOR, lost the trained immunity–induced protection against lethal C. albicans infection. The role of the mTOR–HIF-1α pathway for β-glucan–induced innate immune memory was further validated in myeloid-specific HIF-1α knockout (mHIF-1α KO) mice that, unlike wild-type mice, were not protected against Staphylococcus aureus sepsis.

Discussion

The shift of central glucose metabolism from oxidative phosphorylation to aerobic glycolysis (the “Warburg effect”) meets the spiked need for energy and biological building blocks for rapid proliferation during carcinogenesis or clonal expansion in activated lymphocytes. We found that an elevated glycolysis is the metabolic basis for trained immunity as well, providing the energy and metabolic substrates for the increased activation of trained immune cells. The identification of glycolysis as a fundamental process in trained immunity further highlights a key regulatory role for metabolism in innate host defense and defines a potential therapeutic target in both infectious and inflammatory diseases.

A BLUEPRINT of immune cell development

To determine the epigenetic mechanisms that direct blood cells to develop into the many components of our immune system, the BLUEPRINT consortium examined the regulation of DNA and RNA transcription to dissect the molecular traits that govern blood cell differentiation. By inducing immune responses, Saeed et al. document the epigenetic changes in the genome that underlie immune cell differentiation. Cheng et al. demonstrate that trained monocytes are highly dependent on the breakdown of sugars in the presence of oxygen, which allows cells to produce the energy needed to mount an immune response. Chen et al. examine RNA transcripts and find that specific cell lineages use RNA transcripts of different length and composition (isoforms) to form proteins. Together, the studies reveal how epigenetic effects can drive the development of blood cells involved in the immune system.

Science, this issue 10.1126/science.1251086, 10.1126/science.1250684, 10.1126/science.1251033

Abstract

Epigenetic reprogramming of myeloid cells, also known as trained immunity, confers nonspecific protection from secondary infections. Using histone modification profiles of human monocytes trained with the Candida albicans cell wall constituent β-glucan, together with a genome-wide transcriptome, we identified the induced expression of genes involved in glucose metabolism. Trained monocytes display high glucose consumption, high lactate production, and a high ratio of nicotinamide adenine dinucleotide (NAD+) to its reduced form (NADH), reflecting a shift in metabolism with an increase in glycolysis dependent on the activation of mammalian target of rapamycin (mTOR) through a dectin-1–Akt–HIF-1α (hypoxia-inducible factor–1α) pathway. Inhibition of Akt, mTOR, or HIF-1α blocked monocyte induction of trained immunity, whereas the adenosine monophosphate–activated protein kinase activator metformin inhibited the innate immune response to fungal infection. Mice with a myeloid cell–specific defect in HIF-1α were unable to mount trained immunity against bacterial sepsis. Our results indicate that induction of aerobic glycolysis through an Akt–mTOR–HIF-1α pathway represents the metabolic basis of trained immunity.

In classical descriptions of host defense mechanisms, innate immune responses that are rapid, are nonspecific, and lack memory are distinguished from specific T and B cell–dependent immune responses, which are highly specific and have the capacity to build immunological memory. The hypothesis that the innate immune system is incapable of mounting adaptive responses (1) is contradicted by studies showing that organisms lacking a specific immune system, such as plants or insects, are able to respond adaptively to infection (2, 3) and that innate immune cells, such as macrophages, have adaptive characteristics (4). In line with the proposal that there are nonspecific adaptive responses in the innate immune system, T and B cell–independent protective effects of monocytes and natural killer (NK) cells have been demonstrated in models of bacterial and viral infections, respectively (5, 6). Furthermore, epigenetic reprogramming at the level of histone H3 methylation has been proposed as the molecular mechanism responsible for long-term memory of innate immunity (5, 7), and this process has been termed trained immunity.

Initiation of innate immune memory through trained immunity is likely to be responsible for the nonspecific protective effects of certain vaccines (8). Furthermore, the increased inflammatory responsiveness of monocytes and macrophages due to trained immunity appears to play a central role in inflammatory diseases (9). From this perspective, the capacity of innate immunity to mount adaptive responses both redefines the function of innate immunity and identifies a potential therapeutic target in human diseases. It is thus essential to understand the cellular and molecular mechanisms that mediate trained immunity, in hopes of harnessing their therapeutic potential. Although epigenetic modifications are known to underlie information storage during innate immune memory in both plants (10) and mammals (7), less is known regarding the molecular pathways and downstream mechanisms that lead to trained immunity.

Transcriptome and epigenetics of monocytes

Candida albicans and its main cell wall constituent, β-glucan, induce trained innate immune memory both in vitro and in vivo (7). We performed an unbiased assessment of whole-genome mRNA expression, histone methylation, and acetylation patterns after training human primary monocytes with β-glucan, the major Candida cell wall structure that mediates trained immunity, which induces nonspecific protection against both infections and malignancies (11). An in vitro experimental model of β-glucan–induced trained immunity was established in monocytes (Fig. 1A). β-Glucan training of cells induced a potentiated cytokine production upon restimulation with lipopolysaccharide (LPS) 7 days later (Fig. 1B). An enhanced response was also observed after stimulation with the TLR2 ligand Pam3Cys or with nonrelated Gram-negative and Gram-positive bacteria (fig. S1). Assessment of histone 3 Lys4 trimethylation (H3K4me3) and histone 3 Lys27 acetylation (H3K27Ac) identified promoters that were specifically induced by β-glucan training (Fig. 1C). Pathway analysis of the promoters potentiated by β-glucan identified innate immune and signaling pathways up-regulated in trained cells that are responsible for the induction of trained immunity (7, 12).

Fig. 1 Trained immunity in monocytes.

(A) Schematic of in vitro trained immunity experimental setup. (B) TNF-α levels after 7 days in β-glucan–treated cells. Data are means ± SEM (n = 8, *P < 0.05, Wilcoxon signed-rank test). (C) Genome-wide H3K4me3 (red) and H3K27Ac (blue) epigenetic modifications 7 days after β-glucan treatment. Ratios of β-glucan/RPMI for both H3K4me3 and H3K27Ac modification were calculated for each promoter. The promoters that display significantly higher or lower ratio (P ≤ 0.05, t test) relative to median values are called β-glucan–induced promoters and β-glucan–repressed promoters, respectively. Box plots show distributions of the sequence read density (reads per kilobase) for all promoters, β-glucan–induced promoters, and β-glucan–repressed promoters in each data set. In each box plot, the band inside each box (midpoint) represents the median value, and upper and lower borders of the box represent the Q3 (third quartile) and Q1 (first quartile) values, respectively. The upper line represents the maximum value within the upper bound [Q3 + 1.5 × (Q3 – Q1)]; the lower line represents the minimum value within the lower bound [Q1 – 1.5 × (Q3 – Q1)]. Dots represent observed points outside the upper and lower bound. (D) Epigenetic modifications in the promoter regions of the genes involved in glycolysis and mTOR pathways. The box plots were analyzed as in Fig. 1C. (E) Schematic representation of the up-regulated enzymes (red) in the glycolysis pathway. (F) Representative screen shots of H3K4me3 (red) and H3K27Ac (blue) modifications in the promoter region of pyruvate kinase (PKM) and hexokinase. (G) Differential gene expression analysis between the β-glucan–treated group and the control group. Genes in the glycolysis pathway that are up-regulated by the β-glucan training are highlighted in the box at right. The colors in the heat map represent the normalized RNA levels of identified differential expressed genes (false discovery rate = 0.01, relative change ≥ 1.5) in three mice per group.

In addition to immune signaling pathways, epigenetic profiling of trained monocytes on the basis of both methylation and acetylation patterns identified a signature associated with central metabolism (fig. S2) and an increase in the promoters of genes encoding enzymes involved in glycolysis and its master regulator mTOR (mammalian target of rapamycin) (Fig. 1, D and E). Furthermore, after priming of monocytes with β-glucan, genes involved in glycolysis, such as hexokinase and pyruvate kinase, were epigenetically up-regulated 1 week later (Fig. 1F and fig. S3). The gene expressing mTOR and the glycolytic genes that are targets of the transcription factor HIF1α were also enhanced by β-glucan (fig. S4). In line with this, HIF-1α activation was increased in β-glucan–trained monocytes (fig. S5). In addition, glycolysis genes were also up-regulated in vivo in mice challenged with β-glucan, as revealed by total RNA sequencing analysis in splenocytes of these mice (Fig. 1G and fig. S6).

Glycolysis and monocytes

Monocytes from peritoneal exudates rely on glycolysis as a main energy source (13). The role of glucose as an energy substrate for monocytes is demonstrated by the blockade of monocyte stimulation and trained immunity by incubation of cells with 2-deoxy-d-glucose, a glucose analog that cannot be metabolized by the cells and inhibits glycolysis (fig. S7). This is in line with observations that activated macrophages, dendritic cells, and TH1 and TH17 lymphocytes undergo a switch from oxidative phosphorylation to aerobic glycolysis (14).

Consistent with these findings, monocytes trained with β-glucan showed a reduced baseline oxygen consumption on day 7 relative to naïve cells; this finding is compatible with the hypothesis that these cells underwent a shift from oxidative metabolism toward glycolysis. Moreover, trained cells showed a decreased maximal rate of oxygen consumption after complete uncoupling with carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP), a chemical substrate that permeabilizes mitochondrial membranes and uncouples electron transport systems from the oxidative phosphorylation systems (Fig. 2, A and B), whereas the rate of proton leak–dependent oxygen consumption was not altered (table S1). The latter result indicates a reduction of the capacity of the mitochondrial electron transport chain (ETC) as observed after a period of hypoxia (15). Hypoxia decreases the activity of the ETC complexes I and IV through HIF-1α (16). This hypothesis was reinforced by observations of increased glucose consumption (Fig. 2C), lactate production (Fig. 2D), and ratio of nicotinamide adenine dinucleotide (NAD+) to its reduced form (NADH) (Fig. 2E) in trained monocytes.

Fig. 2 Physiology after β-glucan treatment.

(A) Representative oxygen consumption rate of untreated (RPMI, black) and β-glucan–trained (red) monocytes as determined by high-resolution respirometry (Oxygraph; OROBOROS Instruments, Innsbruck). (B) Baseline (basal oxygen consumption before oligomycin treatment, upper panel) and maximum oxygen consumption rate (maximum oxygen consumption upon FCCP treatment, lower panel) of untrained (open bar) and β-glucan–trained (solid bar) monocytes determined by respirometry and normalized to the leak oxygen consumption. (C and D) Kinetic changes of glucose consumption (C) and lactate production (D) from days 1, 3, and 7 of untreated and β-glucan–trained monocytes. (E) Kinetics of NAD+/NADH ratio determined at days 1, 3, and 7. In (B) to (E), data are means ± SEM (n = 5 to 8, *P < 0.05, Wilcoxon signed-rank test).

Differences in glucose consumption did not offset the high glucose concentrations in the RPMI medium, which suggests that glucose availability is not the limiting factor for the observed training phenotype (fig. S8). In addition, the training effect induced by β-glucan was likely not due to the presence of pyruvate in the culture medium, an intermediate metabolite in glycolysis, because training occurred even when medium devoid of pyruvate was used during the training process (fig. S9).

Earlier studies have shown that a high cellular NAD+/NADH ratio acts through sirtuin-1 to decrease the mitochondrial content (17). This mechanism may explain the observed β-glucan–induced reduction in ETC capacity. In contrast, LPS stimulation leads to a strong but transient increase in the glycolytic process in monocytes (fig. S10); this finding supports the suggestion that, although the acute response of monocytes to LPS is characterized by glycolysis (18), at later time points this response switches to oxidative phosphorylation—a process that subsequently induces immune tolerance by activation of sirtuin-1 and sirtuin-6 histone deacetylases (19). In contrast to LPS-induced tolerance, β-glucan training inhibited the expression of Sirtuin1 (fig. S11). Moreover, the addition of resveratrol, a sirtuin-1 activator, during the first 24 hours of β-glucan training partially inhibited the enhanced interleukin-6 (IL-6) production (fig. S11). This suggests that sirtuin deacetylases play a role in the modulated monocyte functional phenotype and highlights the complex interaction between the intermediate metabolites and subsequent immune responses through chromatin-modifying enzymes (20).

mTOR acts as a sensor of the metabolic environment (21) and functions as a master regulator of glucose metabolism in activated lymphocytes (22). Epigenetic signals at promoters of genes in the mTOR pathway were significantly higher in β-glucan–trained monocytes (paired t test, P < 0.001) than in cells exposed to culture medium (Fig. 3A). Target genes of mTOR, such as EIF4EBP1, displayed a similar pattern (Fig. 3B). In line with this finding, mTOR phosphorylation was up-regulated in trained monocytes as assessed by Western blot (Fig. 3C). Monocytes isolated from patients with a complete deficiency in dectin-1 (23) failed to activate mTOR upon stimulation with β-glucan (Fig. 3D) and failed to enhance tumor necrosis factor (TNF) production upon LPS restimulation (fig. S12), supporting the hypothesis that mTOR phosphorylation is dependent on the dectin-1 C-type lectin receptor.

Fig. 3 mTOR signaling in β-glucan–treated monocytes.

(A) Schematic representation of up-regulated enzymes (red) in mTOR signaling pathway in β-glucan–trained monocytes. (B) Screen shot of H3K4me3 (red) and H3K27Ac (blue) modification in the promoter region of EIF4EBP1 (coding region denoted at bottom), the main target of mTOR, in both RPMI- and β-glucan–treated monocytes. (C) Western blot from cell lysate harvested at day 7 after RPMI or β-glucan treatment. Antibodies specific for endogenous phospho-mTOR (p-mTOR), total mTOR, phospho-AMPK, AMPK, and actin were used to blot the total and phospho proteins, respectively. Representative blots of five independent experiments are shown. The p-mTOR/mTOR ratio is shown as a bar chart (n = 5, P = 0.0625, Wilcoxon signed-rank test). (D to F) The endogenous p-mTOR status of dectin-1–deficient patients was determined by Western blot (D) from cell lysate harvest at day 7 after RPMI of β-glucan treatment and probed with antibodies to p-mTOR and total mTOR, respectively. The p-mTOR/mTOR ratio is shown as a bar chart. Relative cytokine production was determined from cells incubated with rapamycin (mTOR inhibitor) (E) and with AICAR (AMPK inhibitor) and ascorbate (HIF-1α inhibitor) (F) in a dose-dependent manner. In (E) and (F), data are means ± SEM (n = 6, *P < 0.05, Wilcoxon signed-rank test).

Glycolysis in trained immunity

As the data presented above demonstrate activation of mTOR and glycolysis in trained monocytes, we next investigated the causality between these two processes by blocking glycolysis during β-glucan training. Inhibition of mTOR with rapamycin during the first day of stimulation resulted in a dose-dependent inhibition of the training effect induced by β-glucan (Fig. 3E). Indirect inhibition of mTOR with AICAR, an adenosine monophosphate–activated protein kinase (AMPK) activator, had similar effects (Fig. 3F). On the basis of observations that mTOR induction of glycolysis is mediated through activation of HIF-1α and stimulation of glycolytic enzymes (24) and that rapamycin inhibits HIF-1α expression (25), we assessed the effect of a HIF-1α inhibitor on monocyte training. We found that the HIF-1α inhibitor ascorbate also blocked trained immunity in a dose-dependent manner (Fig. 3F).

We further investigated the link between metabolic effects and epigenetic changes by assessing the effects of the epigenetic inhibitors MTA (methylthioadenosine, a methyltransferase inhibitor) and ITF (ITF2357, a histone deacetylase inhibitor) during the training setup on the lactate measurements. As expected, the epigenetic inhibitors had no effect on lactate production in the acute phase (24 hours after β-glucan stimulation; fig. S13). However, lactate production was significantly reduced in the trained monocytes on day 7 when MTA was added to monocytes with β-glucan during the first 24 hours in the incubation period (fig. S13), which suggests that histone methylation also partially modifies the induction of glycolysis in the trained monocytes.

Monocyte mTOR activation

Activation of mTOR by insulin or colony-stimulating factors such as GM-CSF (granulocyte-macrophage colony-stimulating factor) is mediated by intermediary activation of the Akt-PI3K (phosphatidylinositol 3-kinase) pathway (26). A similar signal route is induced in monocytes by β-glucan, as stimulation with β-glucan induced a strong phosphorylation of Akt (Fig. 4A). This effect was again dectin-1–dependent, being absent in monocytes isolated from dectin-1–deficient patients (Fig. 4B). Inhibition of Akt phosphorylation also resulted in down-regulation of mTOR activation (Fig. 4C), demonstrating the relationship between Akt and mTOR activation. Finally, the Akt inhibitor wortmannin inhibited monocyte training by β-glucan in a dose-dependent manner (Fig. 4D).

Fig. 4 Akt–mTOR–HIF-1α pathway downstream of β-glucan stimulation.

(A) Monocytes were treated with either RPMI or β-glucan in the presence or absence of wortmannin, a PI3K inhibitor. GM-CSF stimulation was included as a positive control. Cell lysates were harvested at 5, 15, 30, 60, and 120 min. Akt phosphorylation and actin level were blotted with specific antibodies to p-Akt and actin. Representative blots from three independent experiments are shown. (B) Akt phosphorylation and p-Akt/Akt ratio induced by β-glucan from dectin-1 were determined by Western blot by specific antibodies to p-Akt, total Akt, and actin. (C) Effects of PI3K inhibitors on Akt and mTOR phosphorylation in β-glucan–treated monocytes were determined by Western blot by probing with specific antibodies to p-AKT and p-mTOR. (D) Monocytes were treated with β-glucan in the presence of wortmannin in a dose-dependent manner. Cytokine levels after 7 days were determined by enzyme-linked immunosorbent assay. (E) Relative cytokine production was determined from cells incubated with metformin (AMPK inhibitor) in a dose-dependent manner. (F) Survival of wild-type C57BL/6J mice infected with live C. albicans after training with PBS or β-glucan. Metformin or PBS was given from 1 day before the first nonlethal dose of live C. albicans challenge until 3 days after challenge on a daily basis. (G) Wild-type (WT) and HIF-KO alveolar macrophages at a concentration of 8 × 104 were incubated with PBS or curdlan (100 μg/ml) for 1 hour. Resazurin was added and absorbance was recorded every 30 min for 24 hours. Inset (*) shows absorbance values at the 20-hour time point. Data are representative of three biological replicates. (H) Survival curve of wild-type or mHIF-1α KO mice primed with β-glucan and challenged with a lethal dose of S. aureus infection. In (D) and (E), data are means ± SEM (n = 6, *P < 0.05, Wilcoxon signed-rank test). In (F) and (H), a log-rank test was used to assess significance of the survival curves (*P < 0.05).

Epigenetic reprogramming of monocytes by trained immunity has been reported as a mechanism of nonspecific protection in different models. Mice were protected from lethal disseminated candidiasis after an initial nonlethal Candida albicans infection (7). Similarly, β-glucan also induced protection against infection with a lethal Staphylococcus aureus inoculum (27), while Bacillus Calmette-Guérin (BCG) vaccination protected mice from systemic candidiasis (5). We first assessed whether metformin—which acts through AMPK activation and subsequently mTOR inhibition (28) and is commonly used for the treatment of type 2 diabetes (29)—abrogates the protective effects in these experimental models. In vitro, metformin suppressed trained immunity induced by β-glucan (Fig. 4E), and administration of metformin to mice during and after primary infection with a low-inoculum C. albicans inhibited the protective effects induced by it against secondary disseminated candidiasis (Fig. 4F), demonstrating that mTOR-mediated effects mount a protective trained immunity in vivo.

We assessed whether the effects of mTOR were mediated at the level of innate immunity but not at the level of adaptive T and B cell immunity elicited during vaccination. An experimental model of β-glucan–induced protection against S. aureus sepsis can be observed in myeloid cell–specific HIF-1α conditional knockout mice (mHIF-1α KO) (30). These mice are unable to mount glycolysis specifically in cells of the myeloid lineage. We assessed the metabolic activity of wild-type and mHIF-1α KO macrophages when stimulated with β-glucan. mHIF-1α KO macrophages showed increased chemical reduction of the metabolic indicator resazurin (Fig. 4G), consistent with the hypothesis that HIF-1α induces the switch to aerobic glycolysis in response to β-glucan. In this model, the cells do not undergo the switch in the absence of HIF-1α and are “metabolically” dysregulated. Whereas β-glucan increased the survival of wild-type mice infected with S. aureus from 40% to 90%, the induction of trained immunity was completely abrogated in mHIF-1α KO mice (Fig. 4H). To further dissect which pathways are modulated in the mHIF-1α KO mice, we performed RNA sequencing and compared the differential RNA expression profiles of wild-type and mHIF-1α KO mice. Several interesting genes were specifically up-regulated in wild-type but not in mHIF-1α KO mice (fig. S14 and table S2), including those encoding beclin-1 (an autophagy-related protein), STK11 (an AMPK-related serine-threonine kinase), JHDM1D (jumonji C domain containing histone demethylase), and the FOXO4 transcription factor involved in Akt-PI3K stimulation. Thus, these results demonstrate that stimulation of HIF-1α–mediated glycolysis in myeloid cells is crucial for mounting trained immunity in vivo.

Discussion

The role of histone methylation as a mediator of short-term innate immunological memory in macrophages has been described (7) and has been referred to as a latent enhancer for the epigenetic elements that mediate this phenomenon (31). In this study, whole-genome epigenetic profiling of histone modifications and RNA sequencing analysis have identified both immunologic and metabolic pathways stimulated during trained immunity. A cyclic adenosine monophosphate–dependent pathway mediating trained immunity in monocytes has also been described in an accompanying manuscript (12). In the present study, we identified the metabolic pathways induced in trained monocytes, demonstrating a metabolic switch toward aerobic glycolysis, which is in turn crucial for the maintenance of trained immunity (Fig. 5).

Fig. 5 Model of metabolic activation of trained monocytes, characterized by a shift toward increased aerobic glycolysis and decreased oxidative phosphorylation.

Upon β-glucan/dectin-1 recognition, the AKT–mTOR–HIF-1α pathway is activated and shifts the glucose metabolism from oxidative phosphorylation to aerobic glycolysis. The activated glycolysis state prepares β-glucan–trained monocytes to respond to stimulation in a robust manner. The potential role of rapamycin and metformin in inhibition of trained immunity is also depicted. The metabolic differences between a trained monocyte and naïve monocytes are summarized at the right.

A metabolic switch toward aerobic glycolysis was earlier reported to be a feature of cell activation and proliferation [such an effect was first described in neoplastic cells and termed the Warburg effect (32)] while also playing a role in effector T helper lymphocytes (33) and activated macrophages (34). The elevated glycolysis metabolism observed in trained monocytes might be necessary to equip and prepare cells to respond to the intruding pathogens in a robust and rapid manner through proinflammatory cytokine production and possibly also through enhanced phagocytosis capacity (35).

Although we observed trained immunity in monocytes, this response should not be restricted to cells in the monocyte lineage. Recently, adaptive features of NK cells have been demonstrated to be involved in resistance to reinfection with viruses (6, 36). The specific NK memory cells, like T cells, rapidly proliferate, degranulate, and produce cytokine upon activation. However, it remains to be determined whether metabolic rewiring also plays a role in NK or in other innate immune cells, such as dendritic cells. In addition, it is of interest to determine whether training is contact-dependent or could also be induced by soluble mediators. This is an important question in the field of autoinflammatory and autoimmune diseases, because these diseases are worsened by the unregulated cytokine production. Our results suggest that proinflammatory cytokines such as IL-1β could also induce trained immunity in monocytes in vitro (fig. S15). This hypothesis is further supported by nonspecific protective effects induced by IL-1β, even when injected several days before an experimental infection is induced (37).

One important aspect to note is that the molecular mechanisms investigated in the present study focused on trained immunity in the first 7 days after the initial stimulus. This is the crucial period during which trained immunity offers protection in newborn children against perinatal sepsis (38) and thus is relevant from a biological and clinical point of view. Longer-lasting in vivo effects of trained immunity have been demonstrated in humans (5), and it is important to assess whether these later effects are mediated through similar mechanisms. However, any such later effects are also likely to be exerted at the level of bone marrow myeloid cell progenitors, as recently demonstrated in the case of Toll-like receptor (TLR)–induced tolerance (39).

Our study introduces an interesting preliminary step in understanding the glycolytic process in trained immunity. Hypoxia and glycolysis enhance the proliferative response of macrophages to CSF-1 (40) and sustain the survival of activated dendritic cells (41). Soluble β-glucan from Grifola frondosa induces macrophage proliferation (42), although we were not able to observe these effects with trained monocytes by Candida β-glucan (7). However, epigenetic profiling has identified a cell cycle activation signal in β-glucan–trained cells (12), and it is tempting to speculate that trained monocytes are not only capable of increased cytokine production but also primed to respond to proliferative signals, although this remains to be demonstrated. Finally, the identification of glycolysis as a fundamental process in trained immunity further highlights a key regulatory role for metabolism in innate host defense and also defines a novel therapeutic target in both infectious and inflammatory diseases (9).

Materials and methods

Isolation of primary human monocytes

Blood was collected from human healthy volunteers and two dectin-1–deficient patients after written informed consent (Ethical Committee Nijmegen-Arnhem, approval no. NL32357.091.10). Peripheral blood mononuclear cells (PBMCs) were isolated by differential centrifugation using Ficoll-Paque (GE Healthcare, Diegem, Belgium) from buffy coats obtained from Sanquin Bloodbank, Nijmegen, Netherlands. Monocytes were purified by MACS depletion of CD3-, CD19-, and CD56-positive cells from the PBMCs; CD3 MicroBeads (130-050-101), CD19 MicroBeads (130-050-301), and CD56 were purchased from Miltenyi Biotec (Leiden, Netherlands) and used according to the manufacturer’s protocol. Efficacy of depletion was controlled by flow cytometry (FC500; Beckman-Coulter, Woerden, Netherlands) and was higher than 98%.

Genome-wide sequence analysis

For chromatin immunoprecipitation (ChIP) analysis or RNA sequencing, 10 × 106 CD3CD19CD56 pure monocytes were plated on 100-mm dishes. Monocytes were preincubated with cell culture medium (RPMI) or β-glucan (5 μg/ml) for 24 hours in a total volume of 10 ml. After 24 hours, cells were washed to remove the stimulus and were resuspended in RPMI supplemented with 10% human pool serum. Monocytes were collected before and 6 days after the incubation for ChIP or RNA sequencing. For RNA sequencing, monocytes were collected in TRIzol reagent (Invitrogen, Bleiswijk, Netherlands). The purified materials were then processed to generate genomic DNA for WGBS, RNA (Trizol extraction according to manufacturer instructions; Agilent BioAnalyser RIN >8), and chromatin by fixing the cells in 1% formaldehyde.

Reagents

Candida albicans β-1,3-(d)-glucan (β-glucan) was kindly provided by D. Williams (East Tennessee State University). Reagents used were as follows: LPS (E. coli 0B5/B5, Sigma, Diegem, Belgium), rapamycin (Sigma, R0395), metformin (R&D, AF 1730, Abingdon, UK), AICAR (Sigma, A9978), ascorbate (Sigma, A4034), wortmannin (InvivoGen, tlrl-wtm, Toulouse, France). C. albicans ATCC MYA-3573 (UC 820) cells were heat-inactivated for 30 min at 95°C.

Stimulation experiments

For training, monocytes were preincubated with β-glucan (10 μg/ml) for 24 hours. After 7 days, cells were restimulated with various microbial ligands: LPS (10 ng/ml), Pam3Cys (10 g/ml), and heat-killed S. aureus or heat-killed E. coli (both at 106 microorganisms/ml). After 24 hours, supernatants were collected and stored at –20°C until cytokine measurement. All the cytokine measurements presented were from at least six donors.

To address the HIF-1α–AMPK–mTOR pathway in trained immunity, we added the specific inhibitors together with β-glucan for the first 24 hours in different doses as follows: rapamycin from 1 to 100 nM, metformin from 0.3 to 30 mM, AICAR from 5 to 500 nM, and ascorbate at 5 and 50 μM.

ChIP-seq data analysis

H3K4me3 and H3K27ac ChIP, sequencing, and processing of the data were performed as described (7). The detailed data have been deposited in the GEO database with accession number GSE57206. Sequenced reads of 42-bp length were mapped to human genome (NCBI hg19) using bwa-alignment package mapper (43). ChIP-seq data sets were normalized as described (44), and the sequenced reads were directionally extended to 300 bp, corresponding to the original length of sequenced DNA fragments. For each base pair in the genome, the number of overlapping sequence reads was determined, averaged over a 10-bp window, and visualized in UCSC browser (http://genome.ucsc.edu). These normalized tracks were used to generate the genome browser screen shots. Putative H3K4me3- and H3K27ac-enriched regions in the genome were identified by using MACS (45) with P < 10−8. All the transcription start sites (±1 kb) of genes with significant H3K4me3 signal were regarded as active promoters. H3K4me3 and H3K27ac signals at all active promoters were estimated, and log2 ratios of ChIP-seq signal between treatment and control samples were calculated. Promoters that showed an absolute deviation of 2 times the median (median ± 2 × MAD) of the ratio of ChIP-seq signal (treatment/control) were regarded as regulated promoters (induced or repressed). Sequence reads counted from the normalized ChIP-seq data sets were used to generate the box plots.

Metabolite measurements

Culture medium was collected at days 1, 3, and 7. The glucose and lactate concentrations within the medium were determined by Glucose Colorimetric Assay Kit (K686-100; Biovision, Milpitas, CA) and Lactate Colorimetric Assay Kit (K627-100, Biovision), respectively. NAD+ and NADH concentration were determined by NAD/NADH Quantification Colorimetric Kit (Biovision, K337-100) from the cell lysate according to manufacturer’s protocol. All the metabolite measurement data presented were from at least six donors.

Oxygen consumption measurement

Culture medium was collected from 1 million cells treated with either RPMI or β-glucan. After stimulation, the cells were trypsinized, washed, and resuspended in 60 μl of the collected culture medium. The cell suspensions were then used for cellular O2 consumption analysis. Oxygen consumption was measured at 37°C using polarographic oxygen sensors in a two-chamber Oxygraph (OROBOROS Instruments, Innsbruck, Austria). First, basal respiration (baseline oxygen consumption) was measured. Next, leak respiration was determined by addition of the specific complex V inhibitor oligomycin A (OLI). Then, maximal electron transport chain complex (ETC) capacity (maximum oxygen consumption) was quantified by applying increasing concentrations of the mitochondrial uncoupler FCCP (1 to 14 μM final maximal concentration). Finally, minimal respiration was assessed by adding a maximal (0.5 μM) concentration of the specific complex I inhibitor rotenone (ROT; 0.5 μM) and the complex III inhibitor antimycin A (AA; 0.5 μM).

After establishment of the baseline oxygen consumption rate, cells were treated with the ATP synthase inhibitor oligomycin to determine the rate of proton leak–dependent oxygen consumption, after which the baseline rate value was normalized to the value of the leak rate. Next, the cells were treated with FCCP to determine the maximum oxygen consumption rate. For normalization, the maximum FCCP value was ratioed to the leak value. The oxygen consumption measurement was repeated in monocytes isolated from five healthy individuals.

Western blot

For Western blotting of AMPK, mTOR, Akt (total and phosphorylated), and actin, training was performed as described in stimulation experiments. Adherent monocytes were trained in 24-well plates. After training and the resting period, cells were lysed in 150 μl of lysis buffer. Equal amounts of protein were subjected to SDS-PAGE electrophoresis using 7.5% polyacrylamide gels. Primary antibodies [1:500 and 1:50 000 (actin)] in 5% (w/v) BSA/TBST (5% bovine serum albumin/TBST) were incubated overnight at 4°C. HRP-conjugated anti-rabbit antibody or HRP-conjugated anti-mouse antibody at a dilution of 1:5000 in 5% (w/v) BSA/TBST was used for 1 hour at room temperature. Quantitative assessment of band intensity was performed by Image Lab statistical software (Bio-Rad, CA, USA). The following antibodies were used: actin antibody (Sigma, A5441), mTOR antibody (Cell Signaling, #2972, Leiden, Netherlands), phospho-mTOR antibody (Ser2448) (Cell Signaling, #2971), AMPKα antibody (Cell Signaling, #2532), phospho-AMPKα (Thr172) (Cell Signaling, #2531), Akt antibody (Cell Signaling, #9272), phosphor-Akt (Ser473) (Cell Signaling, #9271). At least four different individual experiments were repeated for each Western blot experiment.

Analysis of RNA sequencing data

Sequencing reads were mapped to the mouse genome (mm10 assembly) using STAR (version 2.3.0). The aligner was provided with a file containing junctions from Ensembl GRCm38.74. In total, there were 507.5 million reads from 12 samples. Htseq-count of the Python package HTSeq (version 0.5.4p3) was used to quantify the read counts per gene based on annotation version GRCm38.74, using the default union-counting mode (The HTSeq package, www-huber.embl.de/users/anders/HTSeq/doc/overview.html).

Differentially expressed genes were identified by statistics analysis using the edgeR package from bioconductor. The statistically significant threshold [false discovery rate (FDR) = 0.05] was applied. For visualization, relative changes larger than 1.5 and FDR of 0.01 were used to plot the expression level of protein-coding genes.

Animal experimental models

The metformin experiment was done at the University of Athens with the approval of the Ethics Committee on Animal Experiments of the University of Athens (approval no. 2550). C57BL/6J female mice (8 to 12 weeks) were used (Jackson Laboratories, Bar Harbor, ME, USA). Mice were injected with live C. albicans blastoconidia (2 × 104 CFU per mouse) or pyrogen-free phosphate-buffered saline (PBS) alone. Seven days later, mice were infected intravenously with a lethal dose of live C. albicans (2 × 106 CFU per mouse). Survival was monitored daily. To assess the involvement of the AMPK-mTOR pathway in the training, metformin (250 mg/kg) or PBS was given via intravenous injection from 1 day before the first nonlethal dose of live C. albicans challenge until 3 days after challenge on a daily basis.

Wild-type (Cre +/+, HIF flox/flox) and HIF-KO mice 8 to 10 weeks old were trained with 200 μl intraperitoneally (i.p.) of either 1 mg of β-glucan particles or sterile PBS on days –7 and –4 prior to tail vein inoculation with 200 μl of 5 × 106 S. aureus strain RN4220 on day 0. Mice were monitored three times daily for survival for 14 days. Data presented are the combined survival data (Kaplan-Meier) from two independent experiments. There were five mice per group in the first survival experiment and seven mice per group in the second survival experiment. A log-rank test was used to assess the statistical significance between the groups. For the RNA sequence analysis, both wild-type and mHIF1α-KO mice were injected with PBS or β-glucan i.p. and the total RNA was extracted from splenocytes at day 4. The total gene expression profiles were accessed by RNA sequencing. This study was carried out in accordance with the recommendations of the National Research Council (46). The protocol was approved by the Dartmouth IACUC (approval no. cram.ra.2).

Metabolic activity assay

Alveolar macrophages were isolated from 6- to 10-week-old HIF-KO and wild-type mice by flushing lungs 10 times with 1 ml of PBS containing 0.5 mM EDTA. Alveolar macrophages were added and allowed to adhere for 1 hour to a 96-well plate at a concentration of 8 × 104 in 200 μl of CO2-independent media (Leibovitz’s L-15, Life Technologies) supplemented with 10% FCS, 5 mM HEPES buffer, 1.1 mM l-glutamine, penicillin (0.5 U/ml), and streptomycin (50 mg/ml). To the media, 10% Resazurin dye (Sigma) was added and the plate was incubated at 37°C for 24 hours, with readings recorded every 30 min at 600 nm. A 690-nm reference wavelength was subtracted from the 600-nm wavelengths and the data were normalized to wells without cells. Curdlan (100 μg/ml, Sigma) was used as a stimulator of metabolic activity.

Statistical analysis

The differences between groups were analyzed using the Wilcoxon signed-rank test (unless otherwise stated). Statistical significance of the survival experiment was calculated using the product limit method of Kaplan and Meier. The level of significance was defined as a P value of <0.05. Cytokine production as well as the band intensity ratio for Western blot were plotted as a bar chart with mean ± SEM. Replicate numbers of the experiments performed are reported in the figure legends.

Correction (9 October 2014): The name and affiliation for Brian M. J. W. van der Veer have been corrected.

Supplementary Materials

References and Notes

  1. Acknowledgments: S.-C.C., J.Q., and M.G.N. were supported by a Vici grant of the Netherlands Organization of Scientific Research and ERC Consolidator grant 310372 (both to M.G.N). C.W. is supported by funding from the European Research Council under the European Union’s Seventh Framework Programme (FP/2007-2013)/ERC grant agreement 2012-322698). Y.L. is supported by Veni grant 863.13.011 of the Netherlands Organization for Scientific Research. R.A.C., and K.M.S. were supported by National Institute of General Medical Sciences grant 5P30GM103415-03 (William Green, PI) and 1P30GM106394-01 (Bruce Stanton, PI), and National Institute of Allergy and Infectious Diseases grant R01AI81838 (R.A.C. PI). R.A.C./K.M.S. thank B. Berwin for the S. aureus. R.J.X funded by DK43351, DK097485, Helmsley Trust, and JDRF. The detailed data have been deposited in the GEO database with accession number GSE57206.
View Abstract

Navigate This Article