The Impact of Glucose Intermediates, Lactate and Amino Acids on Macrophage Metabolism and Function

Abstract

Macrophages (Mφs) are long-lived innate immune cells present in almost all tissues. In addition to phagocytic properties, Mφs are characterized by their plasticity. They are involved in tissue homeostasis, anti-infectious, pro- and anti-inflammatory responses depending on the needs of the tissue. Mφ functional phenotypes are tightly tied to their metabolic pathways. Glucose-related metabolic pathways including glycolysis, pentose phosphate pathway and glycogen metabolism have been associated with the control of inflammatory response. On the contrary, Krebs cycle activity fueled with glutamine or lactate has been associated with Mφs harboring repair properties. For some metabolites, their fate is directly dependent on Mφ phenotype as exemplified with arginase in murine Mφs: proinflammatory Mφs express nitric oxide synthase synthesizing NO while trophic Mφs express arginase-producing L-ornithine and urea. In this chapter, we propose an overview of the metabolic requirement for Mφs function with an emphasis on the differences between human and murine Mφs.

1. Introduction

Macrophages (Mφ) are myeloid cells of innate immunity present in most tissues which control crucial physiological processes such as antimicrobial and inflammatory responses as well as tissue homeostasis and repair. Tissue-resident Mφs (TRMs) exhibit tissue-specific functions and maintain tissue homeostasis [1, 2]. For example, alveolar Mφs are located in alveoli and eliminate debris, surfactant and apoptotic cells while osteoclasts are located in the bone where they orchestrate bone remodeling. During infection or inflammation, when TRMs are overwhelmed, blood monocytes are recruited and differentiate locally into Mφ (hereafter called recruited Mφ) [3] (Figure 1a). Murine embryos with defects in Mφs development are not viable in both human and mice, genetic defects in Mφ lead to several pathologies including neurodevelopmental defects, bone deformities, multiorgan dysfunctions [4] which emphasize the importance of these immune cells.

Figure 1.

A simplified view of Mφ subsets. (a) In situ Mφ includes TRM and newly recruited Mφ. (b) Human and murine Mφ subsets commonly used for in vitro and ex vivo assays.

Mφs are long-lived cells that continuously sense their environment (pH, O2, cytokines, metabolites) and adapt their functions to tissue demands, a process called functional plasticity [3, 5, 6]. Mφs acquire proinflammatory, trophic or immunoregulatory properties. Based on the classification of CD4+ T lymphocytes as Th1 vs. Th2, a classification of Mφs differentiated in vitro from myeloid precursors has been proposed. It opposed proinflammatory Mφs called M1 with reparative and immunoregulatory Mφs called M2 [7].

The M1/M2 classification is imperfect because it is reductive since these two states represent the ends of a functional polarization continuum [8] exemplified by the multiple intermediary’s phenotypes that have been observed [9, 10, 11, 12, 13, 14, 15, 16, 17]. Although imperfect, the classification based on M1 and M2 Mφs remains commonly used (Figure 1b) [18]. Since the phenotypic and metabolic characteristics of human and murine macrophages (Mφs) differ on many points, a parallel analysis of the two models is required, and results obtained from one model should not be transposed to the other.

In mice, both tissue-resident Mφs and bone marrow-derived Mφs (BMDM) are studied [19]. In human, monocytes-derived Mφs are easily generated in vitro while obtaining access to tissue-resident Mφs is obviously limited [20]. Macrophages (Mφs) rely on three growth factors for their differentiation and survival: M-CSF and IL-34 [21]which signal through their receptor CD115, and GM-CSF, an inflammatory cytokine produced at sites of inflammation that signals through CD116 [22]. It should be noted that in mice, bone marrow precursors exposed to GM-CSF differentiate into both dendritic cells and Mφs in vitro [23] while M-CSF induces the generation of a more homogeneous population of Mφs after a 7-days differentiation protocol [24]. Multiple factors polarize Mφs: Interferon-γ (IFN-γ), interleukin-1β (IL-1β), lipopolysaccharide (LPS) trigger the generation of murine and human M1 Mφs that produce IL-12 and inflammatory cytokines (including TNFα, IL-1β, IL-6) but are weak producer of IL-10 [8]. M2 Mφs are generated in the presence of the polarizing cytokines IL-10, IL-4 and IL-13 and they produce growth factors, immunosuppressive cytokines including IL-10, thereby promoting resolution of inflammation and tissue repair [25]. Of note, human Mφs present in inflammatory and tumor sites usually exhibit a mixed M1/M2 phenotype [9, 10, 11, 12, 13, 14, 15, 16, 17, 26]. An important mechanism by which Mφs polarize toward pro-(M1) and anti-inflammatory (M2) phenotypes is via changes in metabolism, also called metabolic reprogramming, to sustain the cell’s function [27]. Distinct phenotypes of Mφs are associated with the pathophysiology of different disorders: M1 cells are involved in the inflammatory processes associated with diabetes and autoimmune diseases while M2 cells contribute to tumor growth and fibrosis [28].

This issue is complicated by variations between species. Murine and human Mφs differ in the way they are generated in vitro and in their phenotypes [8, 19, 29]. Murine Mφs are usually generated from bone marrow cells while human Mφs from monocytes [19]. Importantly, human inflammatory Mφ can be generated with GM-CSF while GM-CSF switches human myeloid precursor to murine DC [23]. Moreover, human Mφs need LPS activation to reveal their phenotype while for murine Mφs, LPS is a polarizing factor. Finally, in the murine in vitro model, Mφs from distinct strains do not respond equivalently to PRR activation [29] as some murine strains (e.g., C57BL/6) are more prone to inflammation compared to others (BALB/c). Differences between metabolism have been also highlighted [30]. Murine M1 cells polarization enhanced glycolysis, pentose phosphate pathway and displays a broken tricarboxylic acid cycle (TCA), and this permits M1 polarized Mφs to rapidly trigger microbicide activity and to cope with a hypoxic tissue microenvironment, thus meeting their energy needs [31, 32]. In human Mφs, M1 glycolytic Mφs do not harbor a broken Krebs cycle (personal data). M2 cells uptake fatty acids to fuel OXPHOS over glycolysis, contributing to tissue remodeling, repair and wound healing [25, 32, 33]. Another example of the metabolic differences between human and mice is observed regarding nitric oxide. Murine proinflammatory Mφs activated by PRR secrete nitric oxide which is toxic for bacteria and intracellular parasites. They express an inducible form of nitric oxide synthase named iNOS that, once induced, is stable for several hours, thus enabling the production of NO at the micromolar range [34]. In contrast, human Mφs do not secrete NO as the transcription of Nos2, the gene encoding iNOS, is methylated and thus limited [35].

2. Overview of metabolic pathways and technical analysis

The aim of cellular metabolism is to generate ATP, NADPH and FADH2 molecules and, in several immune cells including Mφs, metabolic pathways also support the acquisition of functional phenotypes [36]. In addition to their involvement in energy production and redox regulation, some metabolites are involved in chromatin remodeling and thus directly contribute to the regulation of gene expression (Figure 2) [37, 38]. α-Ketoglutarate (α-KG), succinate and fumarate regulate the activities of enzymes involved in histone methylation and DNA methylation [37].

Figure 2.

Interconnection between metabolic pathways and macrophage’s phenotype and function. Macrophages rely on multiple metabolic pathways to produce energy and cofactors in order to survive and to perform their key functions (efferocytosis, cytokine production). As a central metabolite, G6P can be channeled in different directions: being oxidized to pyruvate along glycolysis, to ribose-5 phosphate (R5P) pr xululose-5P (X5P) via pentose phosphate pathway (PPP), or to glycogen via glucose-1P (G1P). Pyruvate has two fates in macrophages: it can be metabolized by lactate dehydrogenase (LDH) generating lactate which is exported out of the cell with a proton. Pyruvate can be transported into the mitochondria via the mitochondrial pyruvate carrier and then be metabolized into acetyl-CoA. This initiates the tricarboxylic acid cycle or Krebs cycle. A complete cycle generates three NADH and one FADH2 cofactors, which transfer their electrons to the respiratory chain that is localized in the intermembrane space of the mitochondria. In the respiratory chain, electrons are transferred between the four multi-protein complexes (complexes I, II, III and IV), and the complexes I, III and IV pump protons from the matrix across the inner mitochondrial membrane. A proton gradient is then established on which depends ATP synthase (also named complex V) to produce ATP from ADP, a process designated as oxidative phosphorylation. PPP generates abundant NADPH and R5P, respectively, to ensure high levels of reduced glutathione for inflammatory macrophage survival as well as DNA synthesis and protein glycosylation. Glycogen metabolism also activates UDPG/P2Y14 signaling pathway to upregulate the inflammatory gene expression via STAT1activation. Multiple metabolites of the TCA (citrate, acetyl-CoA, succinate) and lactate participate in the regulation of gene expression. HK, hexokinase; GPI, phosphoglucose isomerase; PFK, phosphofructokinase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; PGK, phosphoglycerate kinase; PGAM, phosphoglycerate mutase; PKM, pyruvate kinase; LDH, lactate dehydrogenase; G6PD, glucose-6-phosphate dehydrogenase; 6PGL, 6-phosphogluconolactonase; 6PGD, 6-phosphogluconate dehydrogenase; PGM1, phosphoglucomutase; GYS, glycogen synthase; UGP, UDP-glucose pyrophosphorylase; UDPG, UDP-glucose; PDH, pyruvate dehydrogenase; CIC, mitochondrial citrate carrier; SCOT, succinyl-CoA:3-ketoacid coenzyme A transferase; PDH, pyruvate dehydrogenase; PDK, pyruvate dehydrogenase kinase; CS, citrate synthase; ACO2, aconitase 2; IDH3, isocitrate dehydrogenase 3; OGHD, 2-oxoglutarate dehydrogenase; SCS, succinyl-coenzyme A synthetase; SDH, succinate dehydrogenase; FH, fumarate hydratase; MDH2, malate dehydrogenase 2; ACLY, ATP-citrate lyase; HAT, histone acetyl transferase; STAT, signal transducers and activators of transcription; P2Y14, purinergic 14 receptor; Cyt C, cytochrome c; PE, ribulose-phosphate 3-epimerase; RPI, ribose-5-phosphate isomerase; FBPase, fructose 1,6-bisphophatase; ALDO, fructose-bisphosphate aldolase.

Glycolysis is the metabolic pathway that relies on glucose utilization, leading to the formation of pyruvate, a 3-carbon entity and ATP. At the organism level, glucose homeostasis is tightly regulated and for most tissues is the major pathway in producing ATP [39]. Glycolysis generates ATP faster than mitochondrial oxidative phosphorylation and is thus often a primary metabolic fuel. Glucose metabolism involves multiple interconnected pathways including glycolysis, the pentose phosphate pathway (PPP), glycogenesis, gluconeogenesis and glycogenolysis [40]. Within the cells, free glucose is immediately phosphorylated by hexokinase producing glucose-6-phosphate (G6P), thereby preventing its diffusion out of the cell [41]. G6P is at a crossroads of several metabolic pathways (Figure 2). G6P can be processed in glycolysis that drives the generation of 2 molecules of pyruvate, 2 molecules of ATP and antioxidant power via NADPH. In Mφs, NADPH is required for NADPH oxidase (NOX), a key player in pathogen defense [40]. Pyruvate can be metabolized by lactate dehydrogenase (LDH), leading to the formation of lactate, which is exported out of the cell via monocarboxylate transporters (MCT) along with a proton, enabling the reoxidation of NADH to NAD+. If G6P is oxidized by glucose 6-phosphate dehydrogenase (G6PD), it enters the PPP. The conversion of G6P to glucose 1-phosphate (G1P) by phosphoglucomutase (PGM) supports glycogen synthesis (glycogenesis), a critical intracellular reservoir of carbons. In the liver, glycogen can also be synthesized by non-carbohydrate precursors such as lactate (transported from peripheral tissues) and glycerol (released from lipolysis), a metabolic process called gluconeogenesis, which is responsible for the generation of glucose as a fuel for other tissues. We have recently observed that some human M1 cells are capable of gluconeogenesis (unpublished results). Synthesized glycogen is stored in the cytoplasm or enters glycogenolysis for degradation (glycogenolysis), eliciting G6P (Figure 3) [42].

Figure 3.

Simplified view of glycogen metabolism. Glycogen can be synthesized from Glc (glycogenesis) or from non-carbohydrate substrates (glyconeogenesis). Glycogen degradation (glycogenolysis) gives rise to glucose-6-phosphate (G6P) which supplies three different pathways: glucose release, glycolysis or the pentose phosphate pathway. The enzymes mentioned in the proposal are in purple. Ala, alanine; Gln, glutamine; G6P, glucose-6 phosphate; GYS, glycogen synthase; PYG, glycogen phosphorylase; G6PC3, glucose-6 phosphatase catalytic subunit 3; GPI, phosphoglucose isomerase; G6PD, glucose 6-phosphate dehydrogenase.

The pyruvate produced from glycolysis can also be imported into the mitochondria matrix, converted to acetyl-CoA by pyruvate dehydrogenase complex, and incorporated into the TCA cycle (also named Krebs cycle) in conjunction with oxaloacetate. This cycle that mediates the catabolism of acetyl-CoA from the oxidative decarboxylation of pyruvate is linked to the electron transport chain (ETC), a set of multi-protein complexes (complexes 1–4) located in the mitochondrial inner membrane. The respiratory chain creates a proton gradient across the mitochondrial membrane that enables ATP synthase (complex V) to function, resulting in the generation of 38 ATP, whereas glycolysis alone produces 2 ATP. O2 is the terminal electron acceptor of the respiratory chain, so its function is impaired under hypoxia. The Krebs cycle can also be supplied by fatty acids and amino acids including glutamine or ketone bodies (Figure 2).

To assess Mφ metabolism, multiple techniques are available that do not overlap as they gave different information regarding the pathways and nutrients used [43]. First of all, the analysis of transcripts, notably by single-cell technologies, assesses the expression of enzymes or transcription factors involved in energy production [43]. Regarding glycolysis, enzymatic assays can be used to assess the activity of an enzyme (e.g., LDH) or to quantify in cell culture supernatant glucose and lactate concentrations. Enzymatic activities have also been realized on tissue, thus highlighting metabolic networks in situ [44]. The gold standard to analyze both glycolysis and OXPHOS required an extracellular flux analyzer that generated data in real time (for example, Seahorse® technology by Agilent). Glycolytic flux is assessed by the acidification of cell culture supernatant as lactate export is accompanied with the export of a proton (ECAR, extracellular acidification rate) and OXPHOS is monitored by the consumption of O2 in cell culture. Flow cytometry can also be used to assess the expression of metabolism-related protein expression or to assess the uptake of probes that reflect nutrient uptake or mitochondrial mass for example [45]. Metabolites from both glycolysis and OXHOS can be measured by liquid or gas chromatography-mass spectrometry [43], eventually with heavy isotope-labeled nutrients (e.g., stable isotope 13C-labeled glucose) to do isotope tracing [46].

3. Glucose metabolism

3.1 Glucose metabolism in proinflammatory Mφs

The first studies demonstrating the need for Mφs to use glycolysis date back to the 1960s. The increased use of the glycolysis pathway by Mφs for phagocytosis or cytokine production in a proinflammatory context has been demonstrated both in vitro and in vivo [47, 48, 49]. The M1 cell polarization is associated with increased glycolytic activity, compared to IL-4-exposed Mφs (M2 cells) [31, 50]. Moreover, GM-CSF (that induces inflammatory Mφs) increases human Mφ glycolytic activity more than M-CSF, while M-CSF increases OXPHOS utilization more than GM-CSF [33]. The cell surface expression of the glucose transporter GLUT1 (encoded by slc2a1) is upregulated in M1 inflammatory cells and is associated with increased glycolysis [51] and ROS production [52, 53]. Regarding Mφ stimulation, LPS stabilizes the transcription factor HIF1a [54] and upregulates the expression of several glycolysis enzymes (Figure 2) [55, 56]. Accordingly, LPS-stimulated monocytes harbor an increased ECAR as monitored by Seahorse analysis, thus indicating an upregulation of glycolytic metabolism [57].

Metabolic shifts are associated with profound transcriptional regulation of gene expression, and several studies have identified different glycolytic enzymes as being crucial to Mφ biology. Hexokinase acts as a glucose sensor and phosphorylates glucose for subsequent utilization (Figure 2). The isoform 2 of hexokinase behaves as a pattern recognition receptor as it binds N-acetylglucosamine, leading to NLRP3 inflammasome activation [58, 59], thus linking innate immunity to glycolysis.

The first rate-limiting enzyme in glycolysis is 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK-2/FBPase, PFK), which catalyzes the conversion of fructose-6-phosphate to fructose-2,6-bisphosphate, a rate-limiting step in glycolysis [60]. Mφ activation induces a switch in the expression of PFKFB2 isoform, from the liver type-PFK2 (L-PFKFB 2) to the more active ubiquitous PFK2 (u-PFKFB2) isoenzymes through HIF-1α independent mechanism [31]. This enzyme is a potent driver of glycolysis because its overwhelming activity leads to the production of fructose-2,6-biphosphate, which activates PFKFB1 [61, 62]. The isozymes encoded by Pfkfb3 and Pfkfb4 are implicated in Mφ glycolysis at transcriptional levels upon stimulation with IFN-γ and LPS, and silencing Pfkfb3 has been shown to lower glycolysis in inflammatory Mφs [63, 64, 65]. Recently, Chen and colleagues [65] have shown that PFKFB3-mediated glycolysis promotes IFN-γ-induced M1 polarization through the JAK2/STAT1 signaling pathway [65].

GAPDH catalyzes the conversion of glyceraldehyde-3-phosphate into D-glycerate 1,3-bisphosphate (Figure 2). When glycolysis activity is low, GAPDH binds to TNF-α mRNA and inhibits its transcription, thus decreasing its translation [66]. This is specially the case in endotoxin tolerant monocytes that have been previously exposed to LPS [66]. The RNA-binding property of GAPDH is decreased by its malonylation [67], which consists in the addition of a malonyl-CoA group which is synthetized in the cytosol from acetyl-CoA by acetyl-CoA carboxylase (ACC). GAPDH capacity to behave as a transcriptional repressor is inversely proportional to its activity in glycolysis.

PRR-mediated Mφ activation drives inflammation that is dependent on the glycolytic pathway. This could be explained by the fact that glycolysis is faster than OXPHOS if substrates are available [68]. The production of lactate from glycolysis-derived pyruvate enables the regeneration of NAD+ in cell cytosol. In addition, pyruvate kinase M2 (PKM2), an enzyme from the glycolysis that converts phosphoenol pyruvate (PEP) into pyruvate, is dimerized in LPS-treated Mφs and gets acetylated [69], thus slowing glycolysis and allowing flux of glycolysis intermediates into biosynthetic pathways [70]. PKM2 is a key molecular determinant in the enhancement of proinflammatory response of Mφs, through the HIF-1α-dependent mechanism, leading to IL-1β induction by directly binding to its promoter [70, 71]. Recently, Dong and colleagues have shown that increased production of IL-1β by glycolysis was mediated through enhanced H3K9 acetylation in activated murine M1 cells [72]. The acetylation level increased resulting in the opening of IL-1β gene binding chromatin structures leading to HIF1α/PKM2 fixation [72]. Two of the most important signaling pathways for M1 cell activation, NF-κB and Akt/mTOR, converge to stabilize HIF-1α, which increases the expression of several glycolytic genes [73]. In addition to the modulation of inflammatory response, PKM2 seems to be responsible for increased expression of PD-L1 in TAM from hepatocellular carcinoma [74]; PKM2 binds to hypoxia response elements in Mφs and dendritic cells and, in combination with HIF1α, upregulate PD-L1 transcription [75]. In addition to a crucial role in proinflammatory activation, glycolysis also supports pyroptosis [76] in activated Mφs.

Another key feature of the metabolic reprogramming of M1 phenotype upon activation is the upregulation of the PPP (Figure 2). G6P from glycolysis supplies the PPP, resulting in the production of two major products: nicotinamide adenine dinucleotide phosphate (NADPH), that is crucial for preventing ROS production, and ribose-5-phosphate, essential for nucleic acid synthesis [77]. The glucose 6 phosphate dehydrogenase (G6PD) is the first and rate-limiting enzyme of the PPP [40, 77, 78]. Overexpression of G6PD in murine Mφ cell line enhanced the activation of NF-κB and p38-MAPK signaling pathways and potentiated the expression of inflammatory cytokines (such as IL-6, IL-1β, MCP-1 and TNF-α) as well as ROS production [79]. Moreover, Haschemi and colleagues have identified by kinase screening, CARKL, as a novel regulator of human and murine Mφ activation [80]. CARKL catalyzes the formation of sedoheptulose (S7P), an intermediate of the PPP, whose expression rapidly decreased upon M1-polarized Mφ activation. Inflammatory cytokines and intracellular superoxide production rates were blunted upon CARKL overexpression, an effect mediated by sustained S7P production, suggesting that CARKL downregulation is critical for proper M1 polarization [80].

In addition to glycolysis and PPP, Mφs utilize other glucose-related metabolic pathways to drive their inflammatory phenotype. Recently, the role of glycogen metabolism on metabolic reprogramming has become a recognized feature of some myeloid cells under stress conditions such as hypoxia and glucose deprivation [61, 81, 82, 83]. Glycogen present an important reservoir form of glucose in cells which is essential for energy supply and glucose homeostasis [84]. Glycogen can be generated from glucose (direct pathway, Figure 3) or by using non-carbohydrate substrates, a process called gluconeogenesis that occurs mainly in the liver (indirect pathway) [42]. Glycogen metabolism is regulated by glycogen synthesis (glycogenesis) and degradation (glycogenolysis) requiring a coordination action of two key enzymes, glycogen synthesis (GYS) and glycogen phosphorylase (PYG), respectively. Glycogenolysis-derived G6P can fuel different metabolic pathways (Figure 3). Two studies have revealed a direct association of glycogen metabolism and inflammation in murine Mφs. Ma et al. [81] have reported that, unlike IL-4 stimulation, IFNy/LPS treatment stimulates glycogenesis and glycogenolysis in murine M1 Mφs, by upregulating the enzymes involved in glycogen metabolism (including GYS1, PGM1, UGP2 and PYG). The study demonstrated that glycogen metabolism has a central function in controlling inflammatory murine Mφs by two related mechanisms: (i) M1 Mφ uses G6P-derived glycogenolysis which is channeled through the PPP to produce large amounts of NADPH required for inflammatory Mφ survival, (ii) UDPG-derived glycogenesis binds to P2Y14 receptors to induce inflammatory Mφs, and cytokines such as TNF, IL-6 Il-1b were subsequently upregulated in an autocrine fashion via UDPG/P2Y14/STAT1 signaling transduction [81]. Recently Qian et al. [85] studied the importance of UDPG/P2Y14 signaling pathway in a model of inflammation. The authors showed that HIF-1α directly regulates glycogen synthase 1 (GYS1), thus promoting glycogen synthesis. Knock-down of HIF-1α gene interfered with GYS1 both at the mRNA and protein level, suggesting that GYS1 is a downstream target gene of HIF1a. Moreover, LPS stimulation of murine and human M1 Mφs increases UGPG and P2Y14 secretion, while HIF-1α stabilizer (MK8617) treatment exerts an anti-inflammatory effect by inhibiting UDPG and P2Y14 production, leading to the release of inflammatory mediators in M1 Mφs. This study proposed a novel specific regulatory mechanism to prevent inflammation by which MK8617 prevents intracellular HIF-1α degradation through GYS1/UDPG/P2Y14 pathway, thereby attenuating M1 Mφ inflammation (Figure 2) [85].

The mechanisms of inflammatory metabolic reprogramming have mostly been studied using LPS treatment in vitro or in vivo [10, 86, 87, 88, 89]. Recently, Murugina et al. [90] have shown that metabolic reprogramming from oxidative phosphorylation to aerobic glycolysis in human monocyte-derived Mφs and in mouse peritoneal Mφs was also induced by agonists of NOD1 and NOD2 receptors similar to TLR4 agonist lipopolysaccharide. But the rewiring of metabolism toward glycolysis is not strictly conserved for all PRR as TLR2 ligation in monocytes activated bot glycolysis and OXHPOS, a requirement for phagocytic activity and cytokine production [57].

3.2 Glucose-related metabolism in M2 Mφs

Prototypic or alternative M2 cells are generated in the presence of M-CSF, IL-4 or IL-13, and harbor a trophic/wound healing and immunoregulatory phenotype [22]. In contrast to murine M1 cells, alternatively activated M2 cells possess a fully intact TCA cycle and obtain much of their energy from fatty acid oxidation (FAO) and oxidative metabolism, to execute cellular functions [91]. Furthermore, aerobic and anaerobic glucose metabolism also fuel phagocytosis, which is an energy-demanding process and a key property of M2-polarized Mφs [25]; of note human M1 cells (GM-CSF + IFNγ) do not exhibit phagocytic properties.

Lactate is a signaling metabolite that affects both the phenotype and metabolism of Mφs. Lactic acid (LA), produced by glycolysis from pyruvate by LDH, accumulates in the microenvironment in case of injury, bacterial infection or tumors [92]. In a landmark study in 2014, Ruslan Medzitov and collaborators identified LA as a local factor driving murine M2 cell polarization [93]. Lactate is imported by MCT of the Scl16a family [94] with a proton. Lactate-mediated stabilization of HIF1α was required to increase several M2-specific genes including Arg1 and Vegf (Figure 2). Lactate was oxidized by murine tumor-associated Mφs (TAM) as highlighted in tracing experiments.

We and others have shown that LA triggers inflammatory gene expression in human monocytes [95] and in monocytes-derived Mφs [17, 96]. More precisely, human Mφs generated in the presence of GM-CSF and lactic acid acquired a mixed profile associating M2 and M1 phenotype. Lactic acid driven M2 phenotype included CD163 membrane expression, growth factors secretion and the expression of several M-CSF dependent genes [17]. We identified HIF1α as a factor that unlocked the consumption of M-CSF, while increasing the proinflammatory secretory profile (IL-1b, IL-6, TNF-a). Lactate internalization by Mφs was necessary for the acquisition of the phenotype and our study contributed to explain how human Mφs can harbor both proinflammatory and trophic properties. It has been recently published that lactate induces histone lactylation [97]. In addition to its role once internalized in the cell, lactate is a ligand for the protein-coupled receptor (GPCR) 132 whose activation triggers chemokine synthesis by TAM [74].

It has been demonstrated several years later in murine Mφs that pyruvate generated from glucose or lactate is taken up in the mitochondria and incorporated in TCA cycle. A portion of citrate is imported outside the mitochondria and cleaved by ACLY into acetyl-coenzyme A (acetyl-CoA), thus regulating histone acetylation on M2-specific promoters [98]. The function of ACLY in the regulation of the M2 phenotype of murine Mφs has also been studied in the context of LPS stimulation by Shi et al. [99]. The authors realized transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) and identified numerous genomic sequences whose accessibility was reduced by LA, including proinflammatory genes coding for IL-1β and TNF-α. However, these results are not fully transposable to human Mφs, whose biology differs from murine Mφs notably regarding the phenotype as discussed previously [19, 96].

Following M2 polarization, the Pfkb1 gene is expressed instead of Pfkb3, resulting in higher levels of the liver isoform of PFKB2 and lower levels of fructose-2,6-biphosphate [31]. The downregulation of the glycolytic pathway is compensated with a massive augmentation of an oxidative metabolic program, ranging from fatty acid uptake and oxidation to oxidative phosphorylation and mitochondrial respiration. The molecular pathway that directly links mitochondrial oxidative metabolism to the anti-inflammatory program is upregulated by STAT6-PGC-1 following IL-4 activation [100]. Active STAT6 induces the coactivator protein peroxisome proliferator-activated receptor (PPAR) gamma-coactivator 1 (PGC-1), which in turn induces mitochondrial function to promote murine M2 cell polarization [101, 102]. Indeed, transgenic expression of PGC-1 primes Mφs for alternative activation and strongly inhibits inflammatory cytokine production, whereas inhibition of oxidative metabolism or RNAi-mediated knock-down of PGC-1 attenuates this immune response [103, 104]. Thus, PGC is considered as the key player responsible for the metabolic switch in M2 Mφs.

Even though they consume glucose modestly, human or/and murine M2 cells also rely on glycolysis to maintain their phenotype [30, 105, 106]. Glucose oxidation, but not that of fatty acids, plays a critical role in the early differentiation of M2 Mφs via pyruvate dehydrogenase kinase 1 (PDK1). PDK1 knock-down remarkably enhanced the expression of M2 markers and augmented mitochondrial oxidative phosphorylation at the early time point in IL-4-stimulated Mφs. This study suggests that PDK1 regulates M2 Mφ differentiation via controlling glucose oxidation during the early differentiation of M2 Mφs [107]. Moreover, it has been shown that glucose uptake increases over time in Mφs activated by IL-4 [105]. The implied mechanistic link between glycolysis and M2 Mφ activation is the generation of pyruvate by glycolysis, which then feeds into the TCA cycle to promote Ac-CoA synthesis and histone acetylation by ACLY enzyme [105] or mitochondrial OXPHOS [106]. Efferocytosis is one of Mφ’s key functions as it prevents leakage of intracellular contents into tissue, its malfunction driving pathologies including atherosclerosis [60]. Interestingly, phagocytosis of dead cells by Mφs induces an early and transient increase in glycolysis, with glycolysis rate returning to basal activity after 24 h [59]. This metabolic activity is therefore quite distinct from the increase in glycolysis as part of the inflammatory response, whose action is prolonged over time. It has been shown that this transient glycolysis allows the expression of membrane receptors and also activates one of the pH-sensitive receptors, GPR132, which augments Myc expression and thus drives murine Mφ proliferation [61]. In murine Mφs, the Akt-mTORC1 pathway is responsible for increasing glucose consumption in M2 Mφs following IL-4 stimulation. Thus, increasing ACLY enzymatic activity and histone acetylation and, subsequently, M2 gene induction (including chemokine production and gene regulating cellular proliferation) [105]. In human Mφs, although ACLY inhibition attenuated IL-4-induced gene expression, it failed to alter cellular Acetyl-CoA levels and histone acetylation, suggesting that IL-4-induced gene expression occurred independently of ACLY [108]. Those studies noticed considerable differences in metabolic requirements of human vs. murine Mφs toward IL-4-induced polarization. Finally, polarization of Mφs in M2 phenotype induced CARKL upregulation, enhancing the non-oxidative steps of PPP, leading to ribose-5P production, necessary for nucleotide and UDP-GlcNAC synthesis (Figure 2) [80]. UDP-GlcNAC is required for N-glycosylation of different cell surface proteins (i.e., CD206) abundantly expressed in M2 Mφs [10].

4. Krebs cycle and respiratory metabolism

4.1 Oxidative metabolism in inflammatory Mφs

As discussed previously, glycolysis is upregulated in M1 cells in murine M1 cells, the TCA cycle is truncated after the generation of citrate and succinate with accumulation of TCA cycle intermediaries, such as α-ketoglutarate and fumarate. However, it is important to note that the data on TCA cycle disruption in M1 Mφs come from murine Mφs, and a body of evidence suggests that the metabolism of human and murine Mφs differs fundamentally [30, 33].

Succinate stabilizes HIF1α, thus increasing Il-1b transcription (Figure 2). Moreover, the oxidation of succinate in inflammatory Mφs by the succinate dehydrogenase enzyme (SDH) produces ROS [109], required for the acquisition of inflammatory phenotype [110]. SDH belongs to the electron transfer chain, whose assembly dictates Mφ’s inflammatory response [111].

In addition to its plethora of regulatory roles in cell metabolism, citrate sustains inflammatory Mφ response. LPS stimulation of murine M1 cells induces a downregulation of isocitrate dehydrogenase (IDH), which catalyzes the conversion of citrate to isocitrate [10], and upregulation of CIC [112], leading to citrate accumulation and its exportation from the mitochondria to cytosol, which is essential for NO, ROS and prostaglandin production [113, 114]. Pharmacological or genetic targeting of CIC in human Mφs decreases these inflammatory mediators [112]. ATP citrate lyase (ACLY) catalyzes the conversion of cytosolic citrate to acetyl-CoA and plays a critical role in supporting the inflammatory response through its regulation of many inflammatory gene expressions [115, 116]. In activated human Mφs, ACLY activates NF-κB acetylation, enhancing the transcription of several inflammatory genes such as IL-1β and PTGS2, as well as the mitochondrial citrate carrier (CIC) (Figure 2) [117]. These studies demonstrate how Mφ metabolism is not simply needed for providing the energy but can play a pivotal role at transcriptional regulation of the immune response.

In TLR-stimulated murine M1 Mφs, glycolysis upregulation also leads to an upregulation of glutamine consumption, and to an increase of succinate levels [56] explained by glutamine-dependent anaplerosis or the GABA shunt. Although some discrepancy exists in the literature [10], glutamine also contributes to (murine) M1 cell polarization.

Strategies aimed at switching TAM (M1/M2) into antitumor inflammatory Mφs are in demand in oncology as TAMs are the major infiltrate in solid tumors and they harbor multiple protumor properties [118]. CD40 is highly expressed by Mφs and TAM and an agonistic anti-CD40-antibody reverts TAM phenotype into antitumor cells, improves antitumor response and mice survival [119]. On the metabolic level, we would expect this reversion to be associated with a transition from an oxidative metabolism into a more glycolytic metabolism. Intriguingly, agonistic CD40-antibody hugely increases mitochondrial activity. In addition, in Cpt1-deficient Mφs, the phenotypic reversion was inhibited showing that FAO was unexpectedly associated with M2 to M1 phenotype switch [120]. Moreover, glutamine degradation enabled the production of pyruvate that was metabolized to lactate and this process was required for the maintenance of NAD+/NADH ratio and OXPHOS activity. Finally, the upregulation of inflammatory mediators was mediated by epigenetic regulation that required FAO-derived acetyl-CoA [120]. Thus, phenotypic reversion of murine TAM into M1 cells appears supported by mitochondrial metabolism, not glycolysis. A similar phenomenon was observed in murine Mφs treated with oxidized lipids and LPS that acquired a hyperinflammatory phenotype [121]. While LPS augmented glycolysis as previously shown, oxidized lipids increased glutamine-mediated oxidative metabolism. This led to the export of citrate generated in the TCA cycle from the mitochondria to the cytoplasm where it was converted into oxaloacetate which stabilized HIF1a, as previously reported [122] and thus increasing IL-1b production.

4.2 Oxidative metabolism in M2-alternative activated Mφs

Even though glycolysis is also functioning in M2 cells, M2 polarization has been associated with mitochondrial oxidative metabolism since 2006 [103]. It was observed that murine Mφs exposed to IL-4 exhibited an upregulation of the mRNA expression of acyl-CoA dehydrogenases and enoyl-CoA hydratases, two enzymes involved in mitochondrial FAO [103]. They also identified PGC-1β as an important coactivator for the M2 phenotype. The knock-down of PGC1β, or the use of an inhibitor of ETC, embedded M2 polarization and led to an inflammatory phenotype, even in the presence of Il-4.

Several years later, triacylglycerol was identified as the main FA supporting the acquisition of M2 properties in IL-4-generated murine Mφs. The role of the scavenger receptor CD36, which enables FA internalization and lipolysis by lysosomal acid lipase (LAL), thus promoting TCA utilization in both murine and human M2 cells, was highlighted [123]. CD36 expression is now considered as an M2 marker for both murine and human Mφs [124]. Of note, inhibition of FAO with etomoxir might exert a confounder effect. Depending on its concentrations, etomoxir might lead to the generation of etomoxiryl-CoA that depletes CoA and thus disturbs cellular CoA metabolism [125]. Moreover, the role of FAO regarding human Mφ polarization is debated [112]. The importance of FAO for murine M2-like polarization has been described by multiple labs [103, 126] and multiple substrates fueling FAO permit the acquisition of the M2 phenotype including glycolysis-derived pyruvate [126]. More generally, mitochondrial fitness appears as a key event for murine M2 polarization [107].

Glutamine, a non-essential amino acid, is produced from glutamate and NH3 by glutamine synthetase and is converted by glutaminase into glutamate, thus liberating NH4. It is a substrate for Mφs and lymphocytes [47]. Glutamine fuels TCA by a branching to 2-oxoglutarate [127]. Glutaminolysis is required for proper efferocytosis in vitro and in vivo, as evidenced by LysM-Cre × Gls1fl/fl knock-out mice that do not phagocytose apoptotic cells correctly [128]. When realizing efferocytosis, IL-4-Mφ’s upregulation of glutamine uptake increases the formation of glutamate through Gls1-mediated transamination. Glutamate is then metabolized into TCA, thus providing energy and redox buffering for M2 murine Mφs. Reinforcing its importance for M2-dependent polarization, the inhibition of Kir2.1, a potassium channel that senses extracellular K+ concentrations, embedded glutamine uptake by TAM and was therefore accompanied by their polarization into antitumor M1 Mφs [129]. The metabolization of glutamine, a requirement for the acquisition of IL-4-dependent markers, permits the increase of intracellular a-KG, a Krebs cycle intermediary that is also a co-stimulator factor for Jmjd3, a demethylase that decreases the trimethylation of histone H3 K27 [130] on the promoters of genes encoding for M2-like markers including Arg1 and Mrc1. Glutamine induces M2 cell polarization through the glutamine-UDP-N-acetylglucosamine pathway [10, 130].

Other mitochondrial substrates can be used by Mφs, for example, the ketone body acetoacetate (AcAc) (Figure 2). Two ketone bodies are potent energetic substrates for cells to produce ATP [131], AcAc and D-β-hydroxybutyrate (D-βOHB) but only AcAc can be metabolized by murine and human Mφs due to the absence of expression of the mitochondrial enzyme D-βOHB dehydrogenase (BDH1) [132, 133]. AcAc is produced by hepatocytes and is shuttled in Kupffer cells, the resident Mφs of the liver. The importance of this metabolic connection is observed in Mφs which do not express succinyl-CoA:3-oxoacid-CoA transferase (SCOT, encoded by oxct1). This deletion resulted in an increase of fibrosis in mice exposed to high-fat diet, AcAc is thus a key metabolite to preserve Kupffer cell function [132] and AcAc metabolization imprints Mφ’s phenotype. We recently demonstrated that the oxidation of AcAc alleviates the capacity of human Mφs to survive to lactic acidosis [133]. Human Mφs exposed to lactic acidosis harbor depolarized mitochondria, decrease transiently their mitochondrial mass through mitophagy and stop consuming nutrients. AcAc constitutes an alternative fuel that prevents mitochondrial integrity and nutrient consumption. We also identified AcAc as a metabolite that increases VEGF secretion by both murine and human Mφs (unpublished personal data).

In addition to their action as substrates, KB are seen as signaling molecules that activate GPCR [134]. Moreover, D-βOHB (and not AcAc) is an inhibitor of NRLP3 inflammasome assembly. NLRP3 is a multi-protein complex whose activation triggers the activation of caspase-1 and the subsequent cleavage of its substrates including the pro-form of IL-1β and IL-18. Once cleaved, these inflammatory cytokines are secreted in their active form. Monocytes treated with D-βOHB secreted less IL-1β and IL-18 [135] in vitro; in vivo the administration of D-βOHB in a model of gout decreased inflammation.

Most studies characterizing the proinflammatory M1 phenotype associated with glycolysis and oxidative phosphorylation (OXPHOS) in contrast to the trophic/M2 phenotype relying on oxidative phosphorylation have been conducted in vitro using bone marrow-derived macrophages (BMDM) or monocyte-derived macrophages (Mφs). However, it appears that resident macrophages exhibit unique metabolic profiles, likely reflecting the distinct availability of metabolites depending on the tissue context [136, 137]. For example, peritoneum is a glutamate-rich environment and peritoneal Mφs rely on its use to fuel oxphos via glutaminolysis [136]. Peritoneal Mφs’s oxidative burst relies on glutamine metabolism, as opposed to neutrophil and this does not depend on TLR engagement, this study shows that the environment dictates Mφ metabolism. Not all resident Mφs are equivalently affected by OXPHOS defect as assessed by Wculek collaborators [138] who showed that alveolar Mφs, Langherans cells, Kupffer cells and peritoneal Mφs are partially depleted upon deletion of TFAM expression, a transcription factor involved in the expression of several ETC proteins. Heterogeneity of Mφ metabolism has also been described at the organ level, more specifically in the lung that contains two subsets of resident Mφs: alveolar Mφs and interstitial Mφs [139]. Alveolar Mφs harbor reduced ECAR rates compared to interstitial Mφs [140]. Transcriptomic studies have also shown that the metabolic discrepancy exists between alveolar and interstitial Mφs [139, 140]. To sustain their activity, alveolar Mφs rely on fatty acid oxidation [139]. In the muscle, glutamine use can restrain the activity of satellite cells and this decreases tissue regeneration. The inhibition of glutamate dehydrogenase 1 (GLUD1) in Mφs increases the production and export of glutamine in the extracellular environment, that became available for satellite cells [141]. Glutamine has also been identified as the mediator of a crosstalk between Mφs and ovarian tumor cells, with the secretion of glutamine by TAM improving tumor cell aggressivity [142]. Finally, recent data from the Rathmell lab using positron emission tomography tracers indicate that TAMs are the main glucose consumers in the tumor microenvironment and that glucose consumption is not tied to its local concentration but rather to Mφs’ propensity toward glutamine metabolism [143]. This landmark study shows that, contrary to what was previously thought, in tumor microenvironment, the consumption of nutrients such as glucose or glutamine is not dictated by a restricted environment in terms of concentration, but by the cells’ preference for a metabolic substrate. This notion could be further challenged by the coexistence of multiple Mφ subsets that vary in terms of phenotype [9] and metabolism [144] eventually over time [145]. Metabolic and phenotypic data obtained over time will certainly be very informative since Mφs survive in tissues for a long time.

5. Conclusion

In severe or chronic diseases, Mφs accumulate in damaged tissues where they exhibit inadequate responses and participate in the chronicity of the pathology. Metabolic reprogramming of Mφs has been proposed as a promising therapeutic target to combat inflammatory disorders [30, 36]. Moreover, tumor-associated Mφs (TAMs) promote tumor progression and metastasis whereas Mφ in atherosclerosis exacerbates inflammation [146]. Different strategies to reprogram or eliminate Mφ are currently being evaluated in clinical trials, including novel approaches targeting their metabolism [26]. However, it must be kept in mind that these approaches must target Mφ present at inflammatory sites that are characterized by a nutrient-poor and lactic acid-enriched environment [92]. To date, the metabolic adaptations that allow Mφ to survive and function under these harsh conditions remain largely unknown [133]. Their identification remains a major challenge and is required to propose strategies specifically targeting the metabolism of inflammatory Mφ.

Acknowledgments

This work was supported by institutional fundings from Angers and Sorbonne Paris Nord universities.

Authors contribution

NJ and LP wrote the manuscript. PJ revised the manuscript. All authors contributed to the article and approved the submitted version.