Immune cells are finely attuned to extracellular cues and are able to quickly respond to microenvironmental changes. The rapid rewiring of metabolic pathways produces alterations in the bioenergetic profile of the cell and promotes the synthesis of signaling metabolites that ultimately affect cellular functions. Innate immune memory has been shown to be supported by different metabolic pathways, such as glycolysis, oxidative phosphorylation (OXPHOS), the tricarboxylic acid (TCA) cycle, amino acid, and lipid metabolism (Fig. 2).

Metabolic pathways that support innate immune memory and their interaction with epigenetic regulators. The upregulation of multiple metabolic pathways supports the establishment of trained immunity in distinct cell types by providing energy, deferential building blocks and by modulating protein activity. Metabolites enriched upon trained immunity may activate transcription factors and modulate the activity of epigenetic enzymes. These promote an epigenetic signature that enables the increased transcription of pro-inflammatory and metabolic genes, which in turn amplify the existing metabolic shift and ultimately confer a sustained increase in effector functions. (S1P sphingosine-1-phosphate; LOX lipoxygenase; αKG α-ketoglutarate; TCA tricarboxylic acid; OxPHOS oxidative phosphorylation; AKT protein kinase B; mTOR mammalian target of rapamycin; HIF1α hypoxia inducible factor 1α; LXR liver X receptor; TF transcription factor; Ac acetyl; me methyl). Figure created with Biorender.com

Glycolysis involves the rapid degradation of glucose with production of pyruvate, energy and reducing power in the form of adenosine triphosphate (ATP) and nicotinamide adenine dinucleotide (NADH) respectively. Pyruvate can then fuel the TCA cycle or be oxidized into lactate (Fig. 3A). The upregulation of glycolysis, with increased glucose consumption and release of lactate, has been described in distinct cell types and in response to different inducers of innate immune memory. For example, glycolysis was upregulated in neutrophils isolated from BCG vaccinated individuals, revealing increased amounts of lactate secretion. Enhanced lactate production was correlated with enhanced fungicidal activity, possibly due to the associated increase in the release of reactive oxygen species (ROS) [23]. ROS were also increased in neutrophils of mice exposed to β-glucan [24]. BCG, β-glucan and oxLDL have also been shown to induce glycolysis in monocytes [25,26,27]. Indeed, the concomitant pharmacological inhibition of glucose uptake and exposure to the BCG or oxLDL led to the long-term decrease of responsiveness of human trained macrophages, while the inhibition of glycolysis of control cells did not affect their long-term cytokine production capacity [26, 27]. In addition, glycolysis may also be relevant in the establishment of central innate immune memory. Namely, hematopoietic stem cells of mice treated with β-glucan presented an increased glycolytic rate, and administration of β-glucan together with a glucose uptake inhibitor decreased the myeloid bias of hematopoietic stem cells characteristic of trained mice [17]. This increase in glycolysis was associated with changes in the epigenome, namely the increased deposition of the transcriptional permissive mark H3K4me3 at regions associated with the promoters of glycolytic enzymes [25,26,27]. The upregulation of glycolytic genes might be regulated by the protein kinase B (AKT) – mammalian target of rapamycin (mTOR)—hypoxia inducible factor (HIF)-1α pathway, ultimately contributing for the increased responsiveness characteristic of a trained cell.

Schematic overview of the major metabolic pathways involved in innate immune memory, such as (a) glycolysis, (b) oxidative phosphorylation (OxPHOS), (c) tricarboxylic acid (TCA), glutaminolysis and (d) fatty acid, cholesterol, sphingolipid and oxylipin synthesis. Figure created with Biorender.com

Indeed, inhibition of mTOR with rapamycin or of HIF-1α with ascorbate decreased the pro-inflammatory cytokine production of β-glucan or BCG treated monocytes [25, 26]. Also, the myeloid specific deletion of HIF1A compromised the survival of β-glucan trained mice to a S. aureus infection [25]. In humans, oral metformin treatment decreased the ex vivo induction of BCG- and oxLDL-induced trained immunity in monocytes [26, 27]. Metformin is an antihyperglycemic drug known to inhibit mTOR through activation of AMP-activated protein kinase (AMPK). In addition, it also dampens oxidative phosphorylation (OXPHOS), by inhibition of complex I of the electron transport chain. It is important to highlight that mTOR and HIF-1α not only have a role in the activation of glycolysis but can also modulate other pathways. For example, chromatin immunoprecipitation followed by sequencing analysis of HIF-binding sites has identified approximately 500 sites across the genome [28]. At the metabolic level, HIF-1α has been shown to decrease the transcription of the TCA cycle enzyme succinate dehydrogenase (SDH) [29]. HIF-1α may also enhance the expression of glutamine transporters, thus increasing the intracellular concentration of glutamate [30]. Glutamate in turn may be converted to α-ketoglutarate, a metabolite of the TCA cycle. Of note, the consumption of glutamine to feed the TCA cycle, also known as glutaminolysis, has also been shown to play a role in innate immune memory, as discussed in the later sections of this review.

OXPHOS includes the oxidation of NADH, the creation of a proton gradient which is used by F0F1-ATP synthase to synthesize ATP, and the reduction of oxygen (Fig. 3B). Innate immune memory rely on a highly energetic metabolism, with the concomitantly increase of glycolysis and oxygen consumption, as observed for BCG, oxLDL and β-glucan [26] [27] [31]. Indeed, monocytes treated with the ATP synthase inhibitor oligomycin, followed by exposure to memory inducing stimuli such as β-glucan or oxLDL, presented a decreased long-term cytokine producing capacity when compared to trained macrophages that were not exposed to oligomycin [31, 32]. The potential role of OXPHOS in innate immune memory was also highlighted by suggestive associations between common single nucleotide polymorphisms (SNPs) in OXPHOS-related genes and the variability in the responsiveness of β-glucan- or oxLDL- trained monocytes [31, 32].

The increased mitochondrial metabolism of innate memory cells contributes to the increase in energy stores but also potentiates the production of mitochondrial ROS. Indeed, human monocytes exposed to oxLDL exhibited higher levels of mitochondrial ROS [33]. Similarly, neutrophils of zebrafish previously exposed to the bacteria Shigella respond to stimulation with higher levels of mitochondrial ROS, when compared to naïve neutrophils [34]. A sublethal dose of Salmonella enterica also generated neutrophils that respond to stimulation by producing increased amounts of mitochondrial ROS [35]. In this study, S. enterica infection rewired HSC towards the generation of neutrophils with enhanced bactericidal activity, which was phenocopied by overexpression of the transcription factor CCAAT/enhancer binding protein beta (C/EBPβ). C/EBPβ was also involved in LPS-induced memory of mice HSC, by targeting myeloid enhancers and potentially promoting myelopoiesis [15].

In addition to monocytes and neutrophils, lymphoid cell populations such as the natural killer (NK) cells have also shown memory characteristics. NK cells isolated from BCG-vaccinated individuals have also shown increased proinflammatory cytokine production following ex vivo heterologous stimulation [36]. Curiously, NK cells of human cytomegalovirus (HCMV) seropositive individuals exhibit heightened glycolysis, mitochondrial oxidative metabolism, and mitochondrial ROS, relative to NK cells of HCMV-seronegative individuals. This metabolic rewiring and increase in the capacity to produce IFN-γ was dependent on a chromatin modifying transcriptional regulator [37]. Thus, NK cell memory is potentially regulated by metabolic and epigenetic changes, similarly to monocytes [38].

Together, these findings suggest a relevant role for glycolysis, OXPHOS and also mitochondria metabolism and function in innate immune memory.

The increase of oxidative phosphorylation and TCA cycle activity is characteristic of different trained immunity programs. The TCA cycle is a fundamental mitochondrial process that, through a series of oxidative reactions, reduces NADH and FADH2 coenzymes, contributing directly to the electron transport chain and ATP production (Fig. 3C). The TCA cycle also integrates several anabolic and catabolic pathways and produces metabolic intermediates [39]. Trained immunity is supported by the accumulation of TCA cycle intermediate metabolites. Upon induction and maintenance of trained immunity, two main carbon sources are responsible for the supplementation of TCA cycle metabolites: (i) pyruvate-derived acetyl-CoA and (ii) glutamine-derived α-ketoglutarate.

The former derives from the upregulation of glycolysis, which produces increased levels of pyruvate. Once imported into the mitochondria, pyruvate is oxidized into acetyl-CoA. β-oxidation of fatty acids also feeds the TCA cycle by producing acetyl-CoA and NADH and FADH2. Acetyl-CoA not only fuels the TCA cycle but also acts as a source of acetyl groups for histone acetyltransferases. Indeed, the increase in the intracellular concentration of acetyl-CoA promotes histone 3 lysine 27 acetylation (H3K27ac) and leads to the upregulation of glycolysis-related genes, such as hexokinase 2, phosphofructokinase and lactate dehydrogenase [40]. Particularly, a short exposure to exogenous acetyl-CoA increased the capacity of human monocytes to produce IL-6 upon TLR2 stimulation [41]. Acetyl-CoA can be metabolised and condensed with oxaloacetate to form the TCA cycle intermediate citrate. Citrate levels in stimulated macrophages are central for the induction of proinflammatory factors such as ROS, nitric oxide (NO), and prostaglandins [42]. In the cytosol, citrate can also be cleaved to regenerate Acetyl-CoA and used as a precursor of lipid biosynthesis.

In addition to glycolysis, glutaminolysis fuels the TCA cycle and is necessary for the establishment of trained immunity [43]. Glutamine is converted to glutamate, which in turn is synthesised into alpha-ketoglutarate (α-KG) (Fig. 3C). α-KG is a cofactor necessary for the activity of α-KG dependent dioxygenases (α-KGDD). However, trained monocytes are also characterized by the accumulation of the TCA cycle metabolites succinate and fumarate. These metabolites compete with α-KG and inhibit the activity of multiple α-KGDD [44]. Among the α-KGDD family are the prolyl hydroxylase domain (PHD) enzymes, which hydroxylate the transcription factor HIF-1α, promoting its constitutive degradation. Stabilization of HIF-1α is crucial for the induction of trained immunity, promoting the induction of IL-1β transcription, glycolysis, and expression of histone demethylases [25, 45, 46].

The α-KGDD family also includes histone and DNA demethylases. Particularly, fumarate accumulation in human monocytes was shown to increase the production of pro-inflammatory cytokines after stimulation, while also increasing the deposition of H3K4me3 at regions associated with the promoters of the pro-inflammatory genes TNF and IL6. The enrichment of this transcriptional-permissive H3K4me3 mark was associated with a fumarate-induced decrease in KDM5 activity, a histone H3K4 demethylase [43]. Other α-KGDD have also been implicated in the establishment of innate immune memory. The pharmacological inhibition of the H3K9 histone demethylase KDM4 decreased glycolysis and the capacity of β-glucan trained human monocytes to secrete IL-6 and TNF upon TLR4 stimulation. Accordingly, inhibition of KDM4 increased the deposition of the repressor mark H3K9me3 at regions associated with promoters of IL6 and TNF of β-glucan trained cells [47]. KDM4 expression was also increased in stimulated myeloid progenitor cells of mice trained with a western-type diet [12]. This interplay between metabolic intermediators and the activity of epigenetic enzymes illustrates the cross-regulation of metabolism and epigenetics that underlines innate immune memory [14, 48].

The upregulation of the TCA cycle promotes an accumulation of succinate in trained monocytes [49]. Succinate can be oxidised to fumarate by succinate dehydrogenase (SDH), also known as complex II of the electron transport chain. SDH is a bridge enzyme between the TCA cycle and the OXPHOS, since the oxidation of succinate into fumarate promotes the transference of an electron along the electron transport chain. Genetic human studies have revealed the presence of individual SNPs in SDH genes, which were associated with changes in the capacity of pro-inflammatory cytokine production [49]. Furthermore, the transcription of SDHB and SDHD were increased in human macrophages trained with β-glucan [31]. The increase in SDH transcription was accompanied by the enrichment of TCA cycle metabolites, which were ablated by the pharmacological inhibition of the lysine methyltransferase Set7. Similarly, bone marrow cells of mice exposed to β-glucan exhibited a long-term increase in SDHB that was not observed in β-glucan-trained SET7 deficient animals. The transcription of SDHB appears to be regulated by the acquisition of H3K4me1 at SDHB enhancer regions which is deposited by the activity of Set7 induced by β-glucan [31]. These data highlight that the enzymes that participate in the TCA cycle are also regulated at the epigenetic level, particularly through the action of Set7.

SDH activity is not only regulated at the transcriptional but also at the protein level, namely by Itaconate, a competitive inhibitor of SDH [50]. Itaconate is produced by aconitate decarboxylase 1 (ACOD1) from cis-aconitate, an intermediate of the TCA cycle. It is synthesised in pro-inflammatory macrophages [51] and triggers anti-inflammatory and antioxidant responses [52,53,54]. Itaconate not only inhibits SDH but may also affects other metabolic pathways by reacting with thiol groups [55]. Itaconate and its derivatives have been reported to alkylate the ROS scavenger molecule glutathione [53], and different glycolytic enzymes [52, 55, 56]. In line with the anti-inflammatory activity of itaconate, β-glucan restored the cytokine production capacity of LPS-tolerized monocytes, possibly by inhibiting itaconate production and re-establishing SDH function [49]. Also, supplementation of monocytes with the derivative dimethyl itaconate led to the inhibition of β-glucan-induced trained immunity [49]. In contrast, monocytes exposed to dimethyl itaconate in the absence of other stimuli exhibited features of trained immunity, with metabolic, transcriptional, epigenetic and pro-inflammatory changes similar to those induced by the prototypical inducer of trained immunity β-glucan [57]. DMI administration also increased mice survival to S. aureus infection. Dimethyl itaconate-induced trained immunity possibly involved glutathione metabolism. Glutathione was shown to also be involved in vitro β-glucan-induced trained immunity [58] and its pharmacological inhibition reduces the cytokine production of both DMI- and β-glucan-trained macrophages [57, 58]. Itaconate may also modulate the epigenetic landscape of innate immune cells. Of note, itaconate was shown to inhibit TET DNA dioxygenases upon LPS stimulation, and may also suppress the activity of other α-KGDD [59]. However, its role in the epigenetic control of trained immunity has not been addressed. Overall, itaconate and its derivatives have a dual anti- and pro-inflammatory role in trained immunity that is context dependent.

Amino acids have a central role in cellular metabolism and innate immunity activation. Amino acids are not only the building blocks of proteins but also act as precursors of metabolites involved in the trained immunity phenotype. As discussed above, glutaminolysis is essential for trained immunity induction since it replenishes the TCA cycle via α-ketoglutarate. Indeed, pharmacological inhibition of glutaminase blunted the stimulated production of IL-6 and TNF by β-glucan-trained human monocytes. In line with this, glutaminolysis inhibition also decreased the deposition of H3K4me3 at regions associated with the promoters of the IL6 and TNF genes [43]. In addition to glutamine, human monocytes exposed to β-glucan also exhibited increased consumption of aspartate and methionine [43], however, the role of these amino acids in the establishment of innate immune memory remains to be explored further. Methionine, in particular, is the precursor of S-adenosylmethionine (SAM), a methyl donor critical for DNA and histone methylation. LPS-stimulated macrophages upregulate the methionine cycle to fuel the production of SAM which in turn contributes to histone methylation that promotes IL-1β production [60]. Different amino acids have been shown to influence the activation of innate immune cells, however their contribution to the establishment of memory is underexplored. More studies are warranted to obtain a clearer understanding of their specific role in both immunometabolism rewiring and epigenetic modifications driving innate immune memory.

Lipid metabolism has also been implicated in different programs of innate immune memory. Cells acquire fatty acids through both de novo lipogenesis and by uptake from the environment (Fig. 3D). Dietary fatty acids can act as inflammatory stimuli and their involvement in trained immunity has been reported [61]. A diet rich in saturated fatty acids altered the relative populations of hematopoietic stem cells in mice and increased the responsiveness of bone marrow and circulating monocytes to ex vivo TLR4 stimulation. Particularly, the saturated fatty acid palmitate contributed to the increased response of macrophages, and palmitate exposure was sufficient to enhance the clearance of C. albicans from infected mice that lack mature T and B cells [61]. Similarly, a genetic mouse model of atherosclerosis fed a high-fat diet, which induces transient hypercholesterolemia, exhibited a shift of the hematopoietic compartment towards an enrichment of granulocyte-monocyte progenitors [12]. This myeloid bias was sustained, even after the mice returned to a standard chow diet and the cholesterol levels normalized. These results point to the capacity of dietary compounds to modulate innate immune responses. Of note, a high-salt diet was also recently shown to induce a persistent immune memory imprinted in the hematopoietic compartment, with a transcriptional increase of genes involved in glycolysis and fatty acid metabolism [62].

In addition to dietary enrichment of cholesterol, the intracellular upregulation of cholesterol pathways has also been implicated in trained immunity. Namely, various species of cholesterol esters were enriched in hematopoietic stem cells isolated from β-glucan-trained mice. In this model, the pharmacological inhibition of the cholesterol pathway decreased the hematopoietic myeloid bias induced by β-glucan [17]. Mevalonate, synthetized from acetyl-CoA, may be used for the production of cholesterol and it is thought to be the bioactive metabolite of this pathway in the induction of innate immune memory. Accordingly, the inhibition of mevalonate production decreased lactate secretion, H3K4me3 deposition at regions associated with the promoter of TNF and ultimately the production of TNF, which were enhanced in BCG- or oxLDL- trained monocytes [63]. In line with this, a constitutive trained immunity phenotype was observed in hyper immunoglobulin D syndrome patients, who present with mevalonate accumulation [63]. Thus, mevalonate has been shown to induce maladaptive trained immunity, that leads to the sterile inflammatory attacks observed in these patients, while also contributing to the protection conferred by different stimuli such as β-glucan, BCG and oxLDL [