Redox Control of Inflammation in Macrophages

I. Introduction

I nflammation is characterized by the cardinal signs redness, swelling, heat, pain, and loss of function. Rather being a disease, inflammation is the response of the immune system to damage caused to the cells of an organism by diverse pathogens, physical insults, injurious chemicals, or an unbalanced immune cell control. Inflammation usually is a local healing response, but may proceed to a chronic state, if the insult persists and carries the risk to progress to a body-wide response, with the danger of a major organ failure and death. In response to injury, a multifactorial network of chemical signals, comprising inorganic metals, gaseous molecules, radicals, amino acid-derived molecules, peptides and proteins, lipids, as well as sugar-containing moieties, including RNA/DNA, initiate, maintain, and terminate a host response to the afflicted tissue. This scenario involves activation and directed migration of leukocytes to the sites of inflammation. Macrophages that are present in virtually all tissues are important immune effector cells. Macrophages sense not only injury and infection but also other types of noxious conditions such as hypoxia and metabolic stress. Moreover, macrophages fulfill many trophic functions that are essential for tissue maintenance and homeostasis, and presumably for stress adaptation to promote tissue repair and restoration of homeostasis. Macrophages respond to endogenous and exogenous danger signals by changing their functional phenotype to fulfill their roles as sensors, transmitters, and responders of inflammation that are required for host defense, wound healing, and immune regulation.

The term oxidative stress was defined as a disturbance in the prooxidant–antioxidant balance in favor of the former (264). This term describes the conditions when the redox state of main cellular redox systems, for example, ascorbate, glutathione, vitamin E, lipoic acid, NADPH, or NADH, is shifted to the oxidized state. Cells cannot tolerate this condition for extended periods and may respond with pathophysiological events such as protein oxidation, DNA damage, and lipid oxidation. Consequently, cells either repair the damage or die by necrosis or by more subtle ways subsumed by the apoptotic pathways. Prooxidants encompass a wide range of molecules, including superoxide radical anion, hydrogen peroxide, alkoxy- and peroxyradicals, and peroxynitrite, often combined under the term reactive oxygen species (ROS) (79). A redox balance is achieved by the activity of generating enzymes such as the mitochondrial electron transport chain NADPH oxidases, peroxisomes, xanthine oxidase, cytochrome P450, lipoxygenases, and cyclooxygenases, as well as scavenger enzymes such as superoxide dismutase, catalase, glutathione peroxidase, thioredoxin, glutaredoxin, and others (78). Oxidative stress, being associated with an irreversible cell/tissue damage and linked to a variety of diseases, has long been acknowledged as a detrimental consequence of (chronic) inflammation. However, the mediators once thought to cause oxidative stress, often followed by an increase of the intracellular concentration of oxidized glutathione (GSSG) or NAD⁺/NADP⁺ and thus causing toxicity, are now recognized as critical players in physiological signal transmission systems. Changes elicited by ROS are often fully reversible and more subtle, and often not associated with a generally disturbed cellular redox balance. These signaling events are known as redox signaling or redox regulation. Redox regulation is integral to cellular communication and, like other universal networking systems, controls the physiological cellular responses. It is presumed that most, if not all, of the classical transcriptional events within cells are modulated or critically depended on redox signaling [for references, see (73)]. Here we will summarize the essential components of the redox control system that operates in macrophages to guarantee diverse functions in evoking chemical toxicity and immune-regulatory functions. Formation of ROS, expression and function of the natural antioxidant antidote system, targets of redox signaling, operating macrophage responses, and potential therapeutic opportunities will be addressed. Specifically, we discuss the formation of redox species, their primary targets and actions, redox-controlling mechanisms during macrophage development and/or polarization, transcriptional/translational networks that contribute to inflammation, and its antagonization as well as the impact of lipid mediators in shaping the role of macrophages during the onset, propagation, and resolution of metabolic diseases.

II. Macrophage Development

Macrophages are long-lived innate immune cells of the mononuclear phagocyte system that ubiquitously populate distinct organs and tissues and can be considered as a prototypic immune cell (176). Monocytes have the capacity to differentiate into elicited macrophages within tissues. They develop from hematopoietic stem cells in the bone marrow and migrate to the circulation, from where they are recruited to the periphery (86). However, in vertebrates, macrophages are generated through at least two distinct mechanisms of hematopoiesis. For instance, embryonal macrophages develop in the yolk sac of mice from mesenchymal progenitors from embryonic day 8, without passing through a monocytic stage (12, 281). Due to this developmental difference and/or dependent on environmental cues in local environments, macrophages in adult animals exhibit a varying morphology and function.

A reflection of this diversity can be found in the different names, that is, Kupffer cells, microglia, marginal-zone macrophages, red-pulp macrophages, subcapsular sinus and medullary macrophages, metallophilic macrophages, Langerhans cells, alveolar macrophages, or osteoclasts that have been given to specific tissue macrophage populations. It is important to stress that some investigators question the classification of Langerhans cells (skin-resident phagocytes) and microglia (brain-resident phagocytes) as macrophages. These cells differ from some other macrophage populations with regard to their ontogeny and steady-state renewal conditions as outlined below. However, their major outstanding feature is their dependency on interleukin 34 (IL-34) as opposed to colony-stimulating factor 1 (CSF1), which are both ligands for the CSF1 receptor (316). While CSF1 receptor-deficient mice lack virtually all macrophage populations, CSF1-deficient mice are depleted of various macrophage populations, but not Langerhans cells or microglia. Unchallenged CD34-deficient mice lack Langerhans cells and microglia, but do not show alterations in, for example, alveolar macrophages or Kupffer cells (316). Langerhans cells are often associated with a functionally distinct group of cells within the mononuclear phagocyte system, the dendritic cells (DCs). DCs themselves are divided into conventional DCs (cDCs) and plasmacytoid DCs, of which only cDCs will be discussed here. The defining functional property distinguishing cDCs from macrophages is their superior capacity to initiate adaptive immune responses. Therefore, cDCs in an immature stage take up the extracellular antigen, migrate to the draining lymph nodes, while differentiating to a mature state, where they cross-present the acquired antigens to T cells and finally succumb to apoptosis (109, 337). Whether cDCs can be easily distinguished from macrophages is a matter of debate. Macrophages cross-present antigens to T cells and show migratory potential and the initially proposed cell surface markers for discriminating macrophages from cDCs, such as expression of CD11c or major histocompatibility complex II on cDCs and CSF1 receptor or F4/80 on the macrophage side, thus failed to fulfill their promise (109, 123). However, recent progress in the field allows the discrimination of at least certain cDC subsets from macrophages, although being currently limited to the murine system. These cDCs develop from terminally committed cDC precursors (Fig. 1) in an fms-related tyrosine kinase 3-ligand (Flt3-L)-dependent manner as opposed to the CSF1 receptor ligands and show a higher migratory potential, a higher capacity to cross-present antigens, and a higher susceptibility to undergo cell death (109, 337). Apart from this, a core cDC signature was established, which revealed cDC-specific cell surface markers such as Flt3 and CD26. Interestingly, these markers were not sufficiently coexpressed by certain phagocyte subsets (including Langerhans cells) that were previously regarded as cDCs, thus enlarging the macrophage subset family (193). It will be important to investigate whether these findings can be translated to the human system.

OPEN IN VIEWER

FIG. 1.

Developmental sources of adult tissue macrophages. Adult tissue macrophages are derived either from embryonic macrophages that develop early in the embryonal yolk sac and proliferate in situ, or from blood monocytes that develop along a specific line of progenitors in the bone marrow. There is speculation about a lymphoid source of macrophages in the lymphoid organs (dotted arrow). HSC, hematopoietic stem cell; CMP, common myeloid progenitor; CLP, common lymphoid progenitor; MDP, monocyte/dendritic cell progenitor; CDP, common dendritic cell progenitor. For details, see text.

The term macrophage was first introduced by Metchnikoff in 1887 (141). Since then, the origin of tissue macrophages in the adult organism has been an issue of debate. After the introduction of the mononuclear phagocyte system by van Furth, Cohn, and others (303), the notion that macrophages in each tissue develop from monocytes also under steady-state conditions became prevalent, although early conflicting data existed (309). Evidence supporting the mononuclear phagocyte system arose from adoptive transfer studies or administration of clodronate liposomes, showing that monocytes contribute to tissue macrophage pools in the spleen and the intestines (74, 305, 328). Further, mice deficient for factors regulating the development of precursor cells into monocytes and their differentiation into macrophages, such as the transcription factor PU.1 (PU.1^−/−) (187, 252), lack tissue macrophage populations. However, recent studies indicate that the origin of macrophages in adult tissues and during steady state versus inflammation is heterogeneous. Langerhans cells and microglia, as well as Kupffer cells, alveolar macrophages, and peritoneal macrophages, proliferate and sustain their presence in vivo independent of the bone marrow or monocytes (252). Further, humans with a combined deficiency in monocytes, DCs, and lymphocytes do not show defects in Langerhans cells and other macrophage populations (14). Bone marrow transplantation studies and in vivo tracing of yolk sac-derived primitive macrophages strongly suggested that the microglia are generated during early hematopoiesis in the yolk sac (90). Finally, two ontogenetically distinct populations of macrophages were identified, based on differential expression of the specific mouse macrophage marker F4/80. Yolk sac-derived macrophages expressed high levels of F4/80, whereas monocyte-derived macrophages expressed low levels of F4/80 (252). Bone marrow transplantation and fate-mapping strategies demonstrated that in adult mice, yolk sac-derived macrophages contribute significantly to the pools of Kupffer cells, Langerhans cells, alveolar macrophages, and macrophages of the pancreas, the spleen, and the kidney, as well as microglia (252). Thus, at least in the mouse, two systems of macrophage development overlap and two ontogenetically distinct macrophage populations persist in adult animals. It will be exciting to determine whether distinct functional properties can be assigned to yolk sac-derived versus bone marrow-derived macrophages and whether targeting one population over the other might be of benefit for therapeutic approaches. There is evidence that monocyte-derived macrophages in the peritoneal cavity are more potent regarding antigen presentation and in generating NO upon activation (89), whereas yolk sac-derived macrophages may show an altered response to hypoxia, evident by high basal expression of the transcription factor hypoxia-inducible factor (HIF)-3α (252). Unfortunately, we might not be yet at the end of the road of macrophage developmental diversity, as even lymphoid progenitor cells retain potential to differentiate into macrophages (142). However, the contribution of lymphocyte progenitors to macrophage pools in vivo remains to be determined. The complexity of macrophage differentiation is summarized in Figure 1.

III. The Distinguished Redox Species: Nitric Oxide and Superoxide Radical Anion

A. Formation of NO and associated species

In the following paragraph, we recapitulate the essentials on the formation of the classical radical products that are produced upon macrophage activation, that is, nitric oxide and superoxide radical anion that account not only for toxicity but also for redox signaling.

The notion that a structurally simple free radical is produced in cells and functions as a messenger molecule initiated a major paradigm shift in discriminating oxidative stress versus redox signaling. Table 1 lists the nitrogen oxides that arise during redox biology of NO by causing its either oxidation or reduction and provides some of the targets that interact with those species.

OPEN IN VIEWER

Table 1.

Nitrogen Species That Occur from the Oxidation or Reduction of NO

Molecular species	Common abbreviation	Description	Established targets
Nitric oxide	NO	Fee radical, uncharged, product of NO synthase reaction	sGC, COX, non-heme iron, thiyl radicals
Nitroxyl	HNO or NO−	Reduced state of NO, product of Angeli's salt, electrophile	Transition metals, nucleophiles
Nitrosonium ion	NO+	Oxidized state of NO, (unstable at neutral pH)	Thiolate anion
Nitrite	NO2 −	Reaction product of NO with O2 − (can be reduced to NO in the presence of iron)	Heme
Nitrate	NO3 −	Oxidation product of NO
Dinitrogen trioxide	N2O3	Nitrosating agent (formed by reacting NO with O2 −)	Thiols, amines
Nitrogen dioxide	NO2	Strong oxidizing agent, free radical, nitrating agent	Thiols, phenolics (tyrosine)
Peroxynitrite	ONOO−	Formed by the reaction of NO with O2 −, nitrating agent	Thiols, transition metals
Peroxynitrate	O2NOO−	Formed by the reaction of NO2 with O2 −
Nitrosoperoxocarbonate	ONOOCO2	Reaction product of ONOO− with CO2 (short-lived)	Mostly tyrosine
Hydroxylamine	NH2OH	Product of nitroxyl reaction
Ammonia	NH3	Metabolic waste product

For more information of potential target interactions, please see the text.

sGC, soluble guanylyl cyclase; COX, cytochrome-c oxidase.

It was Hibbs et al. (115), who first observed NO production by immune cells, that is, macrophages, and reported that arginine was required for the antipathogen and antitumor activity of these cells. The same group connected cytokine-inducible NO production from l-arginine to tumor eradication in mice, while others proved that macrophages could make nitrite and nitrate (272). Stuehr and Nathan went on to show that macrophages produced NO for cytostasis and respiratory inhibition in tumor cells (273), whereas Lancaster and Hibbs demonstrated the formation of iron–NO complexes (162). These early studies connected NO to innate immunity and paved the ground for understanding redox biology of NO and its derived species [for references, see (116, 329)].

The major source of NO is its enzymatic production in a reaction using l-arginine, NADPH, and molecular oxygen to generate NO and citrulline. Three distinct NO synthase (NOS) isoforms are known. NOS3, endothelial type NOS (eNOS), and NOS1, neuronal type NOS (nNOS), are constitutively expressed, release continuously/pulsatile low amounts of NO, originally linked to reactions in the cardiovascular and neuronal system (60). NOS2, inducible-type NOS (iNOS), is cytokine inducible in immune cells, preferentially macrophages, delivers NO for prolonged periods and generates concentrations ranging from 10 nM to low μM amounts. However, macrophages may also express nNOS to increase their phagocytic capacity (122). Complexity increases because iNOS may also generate N-hydroxyarginine, an inhibitor of arginase, as well as O₂ ⁻. More recently, it has been proposed that nitrite (NO₂ ⁻) can be reduced back to NO by certain heme proteins. Data suggest that deoxyhemoglobin- and deoxymyoglobin-dependent nitrite reduction may contribute to hypoxic vasodilatation (91). In contrast to the NOS reaction, this pathway of NO production may be oxygen independent, may occur in a different cellular microcompartment, that is, mitochondria, and may account for NO formation under hypoxia [for references, see (116)]. In addition, multiple indirect evidence for the existence of an mitochondrial NO synthase (mt-NOS) are published, despite no gene sequence for such an enzyme exists, and the protein nature of this enzyme remains obscure (88). Although the mt-NOS awaits rigorous characterization, it may directly interact with cytochrome c oxidase (COX) to block its activity, a situation of particular importance under hypoxia (222).

The chemical biology of NO can be divided into direct and indirect effects as shown in Figure 2. Direct effects are the reactions that allow NO to directly react with biological target molecules and are reported at concentrations of NO at a low nM range [for references see, (290)].

OPEN IN VIEWER

FIG. 2.

Direct and indirect signaling of NO. Direct effects of NO are discriminated from indirect ones by the reactions with oxygen or superoxide radical anion, provoking oxidative or nitrosative stress. For details, see the text.

At low NO, with very limited O₂ ⁻ being around, nitrosylation and thus activation of soluble guanylyl cyclase (sGC) occur (Fig. 3). sGC is considered the natural NO receptor and the most sensitive direct target, responding to roughly 1–30 nM NO. Next in the sensitivity are other heme and nonheme iron targets such as COX or prolyl hydroxylases (PHD), which, when inhibited, account for the interference with the respiratory chain or stabilization of HIF-1α or HIF-2α. This is achieved at concentrations of roughly 30–100 nM.

OPEN IN VIEWER

FIG. 3.

NO actions at distinct concentrations. Formation of NO by NO synthase (NOS) or derived from nitrate under hypoxia with NOS activity being antagonized by arginase, which consumes l-arginine as well. Exemplified are relative levels of NO required for distinct target interactions and accompanying biological outcomes ranging from redox regulation to oxidative stress. For details, see the text.

Indirect reactions occur at NO concentrations above roughly 400 nM and require that NO reacts with oxygen or a superoxide radical anion to generate reactive species, which subsequently target biological molecules. Indirect effects are further divided into oxidative and nitrosative stress. During oxidative stress, the oxidation state of a target molecule increases, whereas nitrosative stress implies the formal addition of NO⁺ (nitrosonium) to a thiol, secondary amine, or hydroxyl group. The formation of N₂O₃ in the reaction of NO with O₂ is a mild oxidant and prefers to nitrosate nucleophiles such as thiols and amines. Peroxynitrite and nitrogen dioxide, which can be formed in the reaction of NO with O₂ ⁻, are potent oxidants and may provoke oxidative stress associated with the oxidation and nitration of proteins, lipids, or DNA. The balance between oxidative and nitrosative reactions was found to largely depend on the flux rate of NO formation, as autoxidation of NO first generates NO₂, which in the presence of increasing NO concentrations is converted to N₂O₃ (290). Thus, when NO levels are above those of O₂ ⁻, the CO₂-OONO⁻ (nitrosoperoxocarbonate) intermediate is converted to NO₂ and N₂O₃, which is accompanied by changes in the redox profile from oxidative to nitrosative reactions (134). Formation of relatively large amounts of NO and O₂ ⁻ would favor nitration via peroxynitrite. Thus, O₂ ⁻ generation during NO signaling can regulate the concentration of NO, thereby ablating many of the NO-target interactions. During the interaction of ROS with NO, the dominant outcome of the reaction is scavenging NO by its simple conversion to nitrite/nitrate. When more NO than O₂ ⁻ is produced, NO₂ and N₂O₃ are formed (Fig. 2). Conversely, when O₂ ⁻ is in excess, oxidative and nitrosative stress is quenched, suggesting that the direct ONOO⁻ chemistry has a limited role, although N₂O₃ formed under these conditions can oxidize, nitrate, and nitrosate proteins. Apparently, the reaction of NO with O₂ ⁻ consumes most of the NO and produces the intermediates associated with indirect signaling effects. This scenario lowers NO concentrations, while creating potentially new pathways of signal transmission. During the course of macrophage activation, the flux of these radicals constantly changes, which makes prediction on a redox-regulated reaction extremely difficult and indeed limits more generalized conclusions. During in vitro macrophage activation with lipopolysaccharide (LPS)/cytokine, the early O₂ ⁻ formation normally is separated from the later NO formation (226, 248). The temporal changes in radical formation in microcompartments where these radical meet will then affect the resulting signal quality, which at present is not fully accessible due to methodological limitations. It should also be mentioned that NO can ablate the oxidation chemistry mediated by H₂O₂ or ROS (330). The attenuation of metal/peroxide oxidative chemistry, as well as lipid peroxidation, appears as a rational explanation by which NO may limit oxidative injury to cells.

With NO increasing to ∼400 nM or higher, the tumor suppressor p53 is stabilized, and concentrations are reached, where NO can induce apoptosis, become cytostatic as well as cytotoxic (Fig. 3).

Clearly, this marks the transition to indirect effects, but can be causatively linked to iNOS activation. NO concentrations at which S-nitrosation occur are not well defined, but overlap with the iNOS activity as well. S-nitrosation has been shown to regulate a broad spectrum of cell functions during the transport, storage, and delivery of NO in cell signaling and inflammation (265).

Another important regulatory feature of NO formation is the availability of oxygen. The concentration of oxygen at which the activity of the NOS is half-maximal, that is, the K_M, is considerably different for the three isoforms. With oxygen around the K_M or lower, NO formation is a direct result of the oxygen concentration. The K_M for iNOS is 135 μM, making NO production from iNOS under hypoxia extremely inefficient (274). Importantly, iNOS can become a source of superoxide radical anion/H₂O₂ in the absence of arginine, thus contributing to the NO/O₂ ⁻ balance. Moreover, l-arginine is the substrate not only for NOS enzymes but also for arginase (Fig. 3) (201). Arginase 1 is localized in the cytosol and produces ornithine, which is the starting point of polyamine synthesis (putrescine, spermidine, and spermine). Polyamines contribute to DNA stabilization, ion channel transport, or cell proliferation. Arginase 2 is found in the mitochondria, where ornithine production is channeled into collagen biosynthesis. Arginase isoenzymes are inducible and thus compete with NOS for their substrates. Another important aspect for NO production is species specificity. Although human macrophages express iNOS mRNA and protein, the expression levels and NO quantities are significantly lower than those of rodent macrophages (320). In humans, the induction of iNOS requires synergistic combinations of immune signals with the notion that a simple cytokine mix of tumor necrosis factor-α (TNFα)/interferon-γ (IFNγ) is insufficient. The explanation probably resides in the significant mismatch of the human versus murine promoter with respect to potential transcription factor-binding sites.

IV. Macrophage Plasticity in the Light of Redox Biology

Metchnikoff described phagocytes as cells that eat (phagocytose), digest, and feed other cells, coupled by a remarkable migratory capacity. He designated these functional properties as a prerequisite to cope with physiological inflammation, supporting development and harmonizing the integrity of the organism. The term pathological inflammation comprised the recognition and phagocytosis of nonself and unwanted self to protect against injurious agents (288). The defense function has been the focus in macrophage research for a long time, while the metabolic function (e.g., secretion of trophic factors), which is closely coupled to regulating tissue development and repair, became a focus of intense research recently. Details concerning the roles macrophages play under physiological and pathological inflammation have been reviewed elsewhere (204, 228, 267). Briefly, macrophages sense cellular debris and pathogens, phagocytose them, and subsequently may present the acquired antigens to adaptive immunity, although with a limited efficiency as compared to professional antigen-presenting cells such as DCs. After contact with pathogens, macrophages produce a wide array of inflammatory mediators that determine immune responses.

The local tissue environment dictates specific tasks performed by macrophages. However, although macrophages are well adapted to the needs of their present locality, they retain a remarkable functional plasticity, enabling them to respond to the disturbances in tissue homeostasis. Macrophage activation patterns in a skin wound may serve as an initial illustrating example. Macrophages are rapidly recruited to the sites of tissue damage, as was recognized by Metchnikoff in one of his key experiments, when he introduced a splinter into a starfish larva. Interestingly, a key molecule for rapid macrophage recruitment to wounds might be H₂O₂, which is produced by the wound epithelium in zebrafish larvae through DUOX (212). In a similar model system, H₂O₂ was required to recruit phagocytes to early-transformed cells (68). The connection between an evolutionary-conserved function (phagocyte recruitment) and an evolutionary stable simple molecule is certainly charming, but the relative impact of H₂O₂ compared to other phagocyte chemoattractants as well as its function in higher vertebrates [where recruitment of monocytes as compared to resident macrophages prevails (204)] remains to be proven. Once macrophages reach the site of injury, they face pathogen-associated molecular patters (PAMPs) or intracellular constituents released by dead cells known as alarmins or damage-associated molecular patterns (DAMPs) (13), to which they respond utilizing similar sensing mechanisms by inducing an inflammatory reaction. This inflammatory reaction, which again involves the generation of NO and/or ROS, serves to clear the site of tissue injury from harmful agents. In a next step, macrophages switch their phenotype to promote repair and resolve the inflammatory response (5, 167). Accordingly, macrophage depletion shortly after introducing skin wounds in mice resulted in delayed re-epithelization and angiogenesis, as well as prolonged inflammation (177). Resolution of inflammation likely requires macrophage interaction with apoptotic resident cells (5, 128), leading to secretion of anti-inflammatory mediators such as TGFβ or IL-10 and modulation of ROS and NO production, as outlined in more detail below. Finally, to restore the tissue to its original state, newly recruited macrophages leave the wounded site mainly by emigrating to the lymphatic system, again underscoring the high motility of these cells (256).

It has to be noted that there is still uncertainty in the literature whether macrophages readily switch phenotypes in vivo or whether inflammatory versus healing or regulatory macrophages are derived from different monocytic precursors (86). At least two monocyte subpopulations are known in mouse and human blood, which are distinguished based on the expression of different cell surface markers and migratory patterns (85, 342). In the murine system, inflammatory monocytes are Ly6C⁺ CCR2⁺ and CX3CR1⁻. These cells are short-lived and upon infection are rapidly recruited from the bone marrow to the site of inflammation in response to the CCR2 ligand CCL2, where they differentiate to inflammatory macrophages or monocyte-derived DCs (85, 86, 94). Resident Ly6C⁻ CCR2⁻ and CX3CR1⁺ monocytes are long-lived cells that initially may not migrate to the sites of infection due to the absence of CCR2 expression. However, they may be recruited at later stages [e.g., when the CX3CR1 ligand fractalkine is produced from resident macrophages due to interaction with dying cells (99)], when tissue regeneration is required, and when they differentiate preferentially to macrophages with an anti-inflammatory phenotype. Under steady-state conditions, these monocytes are thought to replenish the tissue-resident macrophage populations (86, 94). The notion that Ly6C⁺ and Ly6C⁻ monocytes develop preferentially to inflammatory or anti-inflammatory macrophages, respectively, may simply reflect the impact of the current microenvironment at the time of their recruitment to the inflammatory site, as both subsets can be activated to support inflammation as well as its resolution and repair. Nevertheless, selective depletion of Ly6C⁻ monocytes due to genetic ablation of NR4A1 enhanced the development of atherosclerosis, which was attributed to enhanced generation of inflammatory macrophages (104). Thus, the intrinsic potential of the monocyte subpopulations may influence the resulting macrophage phenotype, although further studies are required to support this conclusion and to evaluate the model in the human system.

Despite the evidence summarized above, macrophages retain plasticity after differentiation from distinct monocyte subsets and switch the phenotypes readily at least in vitro, which provoked a considerable effort to classify macrophage activation states systematically, as summarized in Figure 4. Acquisition of different phenotypes is referred to as macrophage polarization (182).

OPEN IN VIEWER

FIG. 4.

Macrophage subtypes. Macrophages can be polarized into three functional activation states. Stimulation with lipopolysaccharide (LPS) and interferon (IFN)γ generates classically activated macrophages that are characterized by expression of nuclear factor kappa light-chain enhancer of activated B-cells (NF-κB), signal transducer and activator of transcription 1 (STAT1), and interferon regulatory factor 5 (IRF5). Downstream of the interleukin-4 (IL-4) receptor, activation of STAT6, peroxisome proliferator-activated receptor (PPAR)γ, and IRF4 generate wound-healing macrophages. Regulatory macrophages are characterized by expression of STAT3, CCAAT/enhancer-binding protein (CEBP)-β, and cAMP-responsive element-binding protein (CREB) downstream of stimulation with IL-10, G-protein-coupled receptor (GPCR) ligands, and others. Arrows between the macrophage subtypes indicate convertibility as well as the occurrence of mixed or hybrid phenotypes. Markers of the distinct macrophage populations as well as their functional profiles are indicated. For details, see text. On top of polarization of mature macrophages, inflammatory versus resident monocytes may exhibit the potential to differentiate to classically activated or wound-healing macrophages, respectively. During Th1 inflammation, classically activated macrophages are initially recruited to the site of infection, whereas during Th2 inflammation, enhanced macrophage numbers may result from local proliferation.

Initially, proinflammatory macrophages (now termed classically activated macrophages) were described as the main responders to infection (178). Classical macrophage activation requires priming by IFNγ (50), followed by an appropriate stimulus indicating infection or danger, such as the microbial cell wall component LPS, certain IL-1 family proteins, or TNFα. Downstream of these effectors, transcription factors such as nuclear factor-kappa light-chain-enhancer of activated-B cells (NF-κB), signal transducer and activator of transcription 1 (STAT1), as well as interferon regulatory factor 5 (IRF 5) are activated to induce inflammatory gene transcription (163). Classically activated macrophages produce a wide range of proinflammatory cytokines such as TNFα, IL-1β, IL-12, IL-23, and IL-6, as well as inflammatory chemokines, and are believed to show a higher capacity to present antigens (199). The latter might indeed reflect their origin from inflammatory monocytes, as discussed above. Overall, classical activation endows macrophages with the potential to recruit inflammatory cells and to support Th1/Th17 immunity (83). Thus, their critical function is defense against invading microbes, for which they are further endowed with a high bactericidal capacity through their production of high levels of NO and ROS (302). Similar mechanisms might be responsible for the antitumor potential of classically activated macrophages.

Macrophages exhibiting tissue repair properties, also termed alternatively activated macrophages, were first described in the 1990s (268). These cells arise from IL-4 receptor activation by IL-4 or IL-13 and show attenuated production of proinflammatory cytokines, ROS, and NO, the latter due to upregulation of arginase. Further, alternatively activated macrophages secrete immunosuppressive cytokines such as IL-10 or IL-1ra (IL-1 receptor antagonist) and chemokines such as CCL17, CCL18, and CCL22 to attract anti-inflammatory leukocytes, upregulate phagocytic receptors, and produce extracellular matrix (ECM) components and growth factors such as fibroblast growth factors, vascular endothelial growth factor (VEGF), and others (83, 93, 199). These mediator profiles enable alternatively activated macrophages to limit inflammation, to produce the ECM, to induce angiogenesis, but also to combat extracellular parasite infection (83, 93, 199). Transcription factors involved in shaping the wound-healing macrophage phenotype are STAT6, IRF4, and peroxisome proliferator-activated receptor (PPAR)-γ (163). An interesting feature of the wound-healing macrophages is their potential to proliferate in vivo upon induction of Th2 inflammation (129). This might be required to enhance their numbers for effective killing of extracellular parasites in the absence of inflammatory monocyte recruitment. Mechanistic explanations how IL-4 triggers macrophage proliferation await their identification.

In an analogy to the T-helper-cell nomenclature, where Th1 cells are associated with the response against bacteria or viruses and Th2 cells are associated with the response to parasitic infection and with limiting Th1-associated immunity, classically activated opposed to alternatively activated macrophages were denoted as M1 and M2 macrophages, respectively (182, 195). However, after the identification of alternatively activated macrophages, several other functional macrophage phenotypes were described, mostly generated in vitro by stimulation with immunomodulatory agents such as glucocorticoids, immune complexes, immunosuppressive cytokines such as IL-10, as well as various G protein-coupled receptor ligands or dying cells (93, 199). These stimuli generate macrophages that regulate immunity and tissue homeostasis in discrete, but also sometimes overlapping, ways. Common markers and transcription factors involved in shaping the phenotype of regulatory macrophage populations are therefore naturally diverse. However, CCAAT/enhancer-binding protein (CEBP)-β, cAMP-responsive element-binding protein (CREB), and STAT3 emerge as critical transcriptional regulators of these cells (81, 163), and the production of high levels of IL-10 as well as upregulation of sphingosine kinase 1 and TNF superfamily member 14 are associated with regulatory macrophage polarization (199).

To account for the resulting complexity, especially with regard to the notion that different stimuli provoke similar macrophage activation patterns, a classification focusing on the physiological outcome, that is, the functional properties of activated macrophages, has been proposed (199). Macrophages were separated into three functional states analogous to the three primary colors in a color wheel, allowing the occurrence of hybrid-type macrophages, analogous to secondary colors in the color spectrum, and thus an infinite number of functional macrophage subtypes. The defined functional states comprise classically activated (M1), wound-healing (alternative-activated or M2), and regulatory macrophages (Fig. 4) (199). We will follow this model in the present review, as macrophage subtypes will be described that do not fit into the M1/M2 nomenclature. A prominent example of hybrid-type macrophages are tumor-associated macrophages (TAMs) that exhibit healing and tissue remodeling properties that produce inflammatory mediators (IL-6 and IL-23), but also negatively regulate other tumor-associated immune cell populations through IL-10 (230), thus combining the functional properties of all three primary macrophage populations. It is therefore not surprising that transcription factors such as NF-κB and STAT1 (classically activated macrophages), PPARγ (wound-healing macrophages), as well as STAT3 and CREB (regulatory macrophages) are connected to the TAM phenotype (319).

There are still important open tasks/questions with regard to macrophage plasticity. A major drawback is the lack of definite markers to identify functional macrophage subpopulations (markers, which are unambiguously associated with a specific function) in tissues. This would allow predicting individual macrophage function, for example, in tissue sections. Recent transcriptome studies resulted in the identification of gene signatures for murine macrophage subpopulations (204). However, mRNA expression at a certain timepoint (mostly in vitro) does not necessarily correspond to stable expression of the gene product in a tissue. Further, gene profiles derived from mouse macrophages are difficult to transfer to human cells, as, for instance, classically activated human macrophages may not express iNOS, whereas alternatively activated human macrophages do not express arginase, which are put forward as prototypic markers for the respective macrophage populations in mice. Techniques such as membrane proteomics of macrophages sorted from tissues in a discrete pathological state might result in the identification of the elusive phenotype markers.

Importantly, the impact of redox modification of signaling or structural proteins on macrophage function is largely unexplored, although data with regard to classical activation are available. The role of ROS (H₂O₂, HOCl, O₂ ⁻, and reactive nitrogen species) in NF-κB activation in different cell types is controversially discussed, owing primarily to the unspecific effects of antioxidants (butylated hydroxyanisole, glutathione (GSH), and N-acetyl-cysteine) used in respective studies (92). However, it seems clear that ROS are involved in the activation process of classically activated macrophages. Downstream of signaling through the LPS receptor complex consisting of Toll-like receptor (TLR)4, CD14, and NOX4-dependent ROS production was involved in the activation of NF-κB as well as p38 mitogen-activated protein kinase (MAPK), which is required for LPS-induced activator protein 1 (AP-1)-dependent transcription (92). Further, mitochondrial ROS seem to foster inflammasome activation as outlined below. Various redox-sensitive proteins are involved in the signaling cascades triggered by inflammatory mediators (75). It will therefore be a challenging task to elucidate in detail how redox reactions regulate polarization of macrophages into different functional subpopulations. Of specific interest will be to determine whether ROS generated upon classical activation interfere with priming macrophages to become tolerant or activated toward an anti-inflammatory or regulatory macrophage phenotype. For instance, the induction of LPS tolerance in macrophages requires chromatin remodeling (76), which is a redox-sensitive process (232). Further, human individuals with chronic granulomatous disease (CGD), having a genetic defect in ROS production, show hyperinflammatory responses (26). At this point, it is interesting to note that classically activated versus wound-healing macrophages were also denoted as reductive or oxidative, respectively, due to differences in glutathione levels (203).

V. Redox Systems Modulate and Determine the Macrophage Phagocyte Function

In the following paragraph, we will concentrate on the interaction of macrophages with pathogens and dying cells and summarize, as schematically outlined in Figure 5, the mechanisms how this interaction is controlled by and how it controls redox signals.

OPEN IN VIEWER

FIG. 5.

Redox regulation of macrophage interaction with dying cells. Release of alarmins such as high-mobility group box 1 (HMGB1) and adenosine triphosphate (ATP) from necrotic cells (NC) induces macrophage activation concomitant with upregulation of inducible NO synthase (iNOS) and activation of the NOX2-containing NADPH oxidase complex. In apoptotic cells (AC), reactive oxygen species (ROS) generated at the mitochondria deactivate HMGB1 by oxidation. Cytochrome c (Cyt c) released from mitochondria during apoptosis in an oxidized form in turn oxidizes phosphatidylserine (PS) at the inner leaflet of the plasma membrane, leading to its externalization, which requires inhibition of aminophospholipid translocase. PS, oxPS, as well as oxidized phosphatidylcholine (oxPC) are recognized by macrophage scavenger receptors through bridging molecules, which may themselves undergo oxidative modifications. Downstream of AC recognition, unknown mechanisms (dotted arrows) activate PPARγ, which inhibits the NOX2 complex in macrophages. Mechanisms of arginase induction and induction or repression of iNOS upon AC–macrophage interactions are controversially discussed and therefore are not clearly indicated. For details, please see the text. FcR: Fc (antibody) receptor; nAb: natural (IgM) antibody; CRP: C-reactive protein.

VI. Redox Systems and Regulatory Macrophage Function

As discussed in the previous chapter, apoptotic cell interactions limit the production of ROS and NO by macrophages. Apoptotic cell-stimulated macrophages can be regarded as regulatory macrophages, and there is more information in the literature, indicating that the functional properties of regulatory macrophages are determined by their limited, but not absent, potential to produce ROS and NO. The clearance of apoptotic neutrophils as well as the stimulation with LPS and immune complexes generated regulatory macrophages that produced NO and primed Th2 responses (63, 69). The glucocorticoid dexamethasone increased ROS production in alternatively activated macrophages, thereby inducing immunosuppressive regulatory T-cells (157). Along this line, macrophages from patients with CGD, lacking functional NOX2, showed hampered regulatory T-cell induction and thus reduced T-cell suppression (156). Interestingly, tumors contain a population of immunoregulatory myeloid cells, myeloid-derived suppressor cells (MDSCs), which express iNOS and arginase simultaneously, thereby producing low NO levels, and also generate ROS. These macrophage-related cells, which can differentiate into TAMs, inhibit antitumor T-cells and/or generate regulatory T-cells (80). It is tempting to speculate that apoptotic cells may induce a similar response in macrophages at sites of tissue injury and inflammation, such as in tumors, to limit adaptive immune responses. Besides NOS and NOX, another enzyme completes the redox arsenal of regulatory macrophages that aims to limit T-cell responses. Indoleamine 2,3-dioxygenase, an intracellular heme enzyme that is involved in tryptophan catabolism, is expressed by apoptotic cell-stimulated (233, 322) and other regulatory macrophages (21). Depletion of tryptophan, the least-abundant essential amino acid, blocks effector T-cell proliferation and enhances the suppressive activity of regulatory T-cells (7). These data indicate that an essential feature of regulatory macrophages, the modulation of T cell responses, is shaped by redox processes. As studying the regulatory macrophage biology progresses, new redox-dependent functions of this therapeutically highly interesting cell population are likely to emerge (21, 72).

VII. Targets of the NO/O₂ ⁻ Redox Biology in Macrophage Function

The NO/ROS redox biology not only helps to eliminate pathogens or to kill tumor cells but also shapes other aspects of macrophage biology. Here we describe some of the established NO targets in macrophages and discuss potential functional consequences.

VIII. Redox Signals and Hypoxia-Inducible Macrophage Responses

Redox-sensitive transcription factors with a relevance to classical and alternative macrophage activation are the HIF-1 and HIF-2. Macrophages accumulate at sites of inflammation such as wounded areas or tumors where hypoxia prevails, activating HIF-1 and HIF-2. Macrophages and in particular the HIF-system within these cells play decision-making roles of how tissues respond to environmental and cell autonomous signals to deliver cell-adaptive responses. Here we will briefly characterize how redox signals impinge on the HIF system, and how this system in turn affects the macrophage biology.

A. NO and the hypoxia-responsive system in macrophages

HIF proteins are composed of HIFα subunits that are often kept at low levels due to continuous 26S proteasomal degradation. Mechanistically, both HIF-1α and HIF-2α are continuously hydroxylated at proline residues by PHDs, whose activity is regulated by oxygen availability, thus serving as oxygen sensors [reviewed in (138)]. Hypoxia attenuates PHD activity, stabilizes the HIFα subunits, and allows their dimerization with HIF-1β and gene regulation. In some analogy to PHDs, factor inhibiting HIF (FIH) hydroxylates a C-terminal asparagyl residue in the HIFα subunits, which impairs the transcriptional activity of HIF. Readers are referred to excellent reviews for further details of HIF regulation [for review, see (98, 138)].

It emerges that the PHD activity is affected not only by oxygen but also by NO and/or ROS (Fig. 6).

OPEN IN VIEWER

FIG. 6.

Redox biology of prolyl hydroxylases (PHD). ROS generated by NOX, an altered antioxidant status or mitochondrial respiration, as well as hypoxia or NO block PHD activity. This stabilizes hypoxia-inducible factor (HIF)-1α, allows the interaction with HIF-1β, and subsequent target gene activation. iNOS as one HIF-1-regulated target may start a positive amplification loop to stabilize HIF-1α. iNOS not only is an HIF-1 target but also a classical NF-κB-responsive gene, activated in LPS-stimulated macrophages. o- - -o; refers to the PHD and HIF-α interaction.

NO-releasing drugs or NO derived from iNOS stabilizes HIF-1α and activates HIF-1 [for review, see (25)]. Recently, it has been shown that NO reacts with a truncated PHD2 catalytic domain (PHD2_181–426), both by coordinating the active Fe(II), as shown by electron paramagnetic resonance analysis, and/or via S-nitrosation of cysteine residues, as shown by mass spectrometry studies (36). Although the biological significance of S-nitrosation is not fully elaborated, it is possible that liberation of NO from S-nitrosated PHD2 can result in inhibition via reaction at the active site Fe(II). Stabilization of HIF-1α by NO might be relevant in certain forms of human cancer, where inhibition of iNOS interferes with the accumulation of HIF-1α (231), with the notion that lowering the HIF-1 activity reduces tumor progression (317). Considering the similarity in reactions catalyzed by PHDs and FIH, it is not surprising that the FIH activity is also blocked by NO [reviewed in (24)]. In addition, stabilization of HIF-1α through S-nitrosation at Cys533 within the oxygen-dependent degradation domain, with NO being generated from neighboring macrophages, was shown in tumor tissue after ionizing radiation (170).

The NO-stabilizing effect changes to a destabilizing one, when NO is supplied under hypoxia (25, 217). This paradox can be resolved, considering that NO competes with O₂ for the binding to mitochondrial COX, which consumes most of the oxygen within a cell (Fig. 7). At a low pO₂, when NO competes with oxygen at COX, oxygen will be spared by mitochondria and instead can support the PHD activity [reviewed in (24)]. In analogy, also mitochondrial respiratory chain inhibitors or eNOS-generated NO lowers accumulation of HIF-1α under hypoxia (59, 102). This intriguing argument in sparing oxygen is not without limitations, as NO binding to COX may trigger O₂ ⁻ formation, which leads to not only the O₂ consumption but also the trapping of NO. However, experiments looking for cellular hypoxia by pimonidazole staining showed that at intermediate oxygen concentrations (3% O₂), hypoxia staining and HIF-1α expression were detectable, but strongly reduced after inhibition of the electron transfer chain (59), clearly indicating that more oxygen is available for the PHD reactivation, thus allowing the cells to return to normoxic conditions. Formally, pimonidazole staining experiments have not been carried out with NO under hypoxia, but as the PHD activity is reactivated and hypoxia staining correlated with PHD activity in other experiments, the potential inference by O₂ ⁻ is of no or minor importance.

OPEN IN VIEWER

FIG. 7.

PHD activity and mitochondrial oxygen consumption. In the presence of oxygen, the PHD activity continuously hydroxylates HIF-1α, which allows its subsequent proteasomal degradation. Oxygen is also consumed during mitochondrial respiration, with cytochrome-c oxidase (COX) reducing oxygen to water. Inhibition of COX by NO shunts the mitochondrial-used oxygen toward PHD activity. For details, see the text.

A more direct connection between hypoxia and the NO/iNOS system exists, as the iNOS promoter contains classical hypoxia-responsive element sites. However, despite the potential HIF-1-binding site, hypoxia alone only poorly (if at all) induces iNOS expression (58). Considering that the K_M value of iNOS for oxygen is 135 μM, it is not surprising that NO formation under hypoxia (1% O₂) is also attenuated (189), irrespective to iNOS expression. This becomes relevant during bacterial infections (225). Under normoxia, the bacterial cell wall components contribute to the expression of HIF-1α, likely via NF-κB activation, as discussed below. HIF-1, presumably with the help of supportive transcriptional regulators such as NF-κB, causes iNOS expression and NO formation, whereas interfering with NO formation lowers HIF-1α expression. Apparently, HIF-1 is central to an NO-dependent amplification loop during the response of murine macrophages to bacterial infection. The relevance of the HIF-1/NO-signaling axis might be restricted to the conditions that allow iNOS to be active rather than being dampened by hypoxia. The fact that iNOS expression in the human system is much more restricted compared to the murine system may further limit the relevance of the HIF-1/NO axis for human diseases.

D. HIF in classically activated macrophages

Classical macrophage activation at the sites of inflammation increases the demand of ATP and O₂ in an area facing a shortage in O₂, thus being or becoming hypoxic. Under resting physiological conditions, oxygen delivery and consumption are balanced, whereas during inflammation, the oxygen demand exceeds oxygen delivery, and consequently, HIFα subunits are stabilized (119, 326). Besides oxygen, additional cofactors such as iron, 2-oxoglutarate, and ascorbate (138) regulate PHD activity. Iron chelation, but also oxidation of the iron, either directly or indirectly by reducing cellular antioxidants attenuates PHD activity (Fig. 8). Bacterial- or LPS-activated macrophages produce oxidants via activation of NOX or iNOS and promote HIF-1α protein accumulation even under an ambient oxygen level [for references, see (53)]. Macrophage responses to eukaryotic microorganisms also affect HIF, as Leishmania-infected macrophages accumulate HIF-1α and HIF-2α (52). Oxidants such as oxLDL provoke HIF-1 activation as well (227, 260), and proinflammatory cytokines such as TNFα or IL-1β induce HIF-1α protein accumulation by increasing either ROS production or phosphoinositide-3-kinase (PI3K)-dependent translation [Fig. 8, (53)].

OPEN IN VIEWER

FIG. 8.

Transcriptional, translational, and post-translational regulation of HIF-1α. In the presence of oxygen, PHDs continuously hydroxylate HIF-1α protein and thereby mark it for proteasomal degradation. Hypoxia, NO, or ROS inhibit PHD activity to stabilize HIF-1α, which dimerizes with HIF-1β to promote target gene induction. IL-1β, tumor necrosis factor-α (TNFα), or LPS increases HIF-1α translation or use ROS to block PHD activity, while inflammatory-associated transcription factors such as NF-κB, STAT3, and/or nuclear factor of activated T-cells (NFAT) increase HIF-1α mRNA expression. For details, see the text.

Other small-molecule mediators of the inflammatory microenvironment, such as adenosine, may also induce PI3K-dependent HIF-1α translation. Although the steady state of HIF-1α protein expression is predominantly determined by PHD-mediated degradation, increased protein translation may shift the balance and overwhelm the degradation system, adding another level of HIF-1 regulation.

At the transcriptional level, classically activated macrophages show, among others, NF-κB activation, which is facilitated via the classical/canonical or alternative/noncanonical pathway. Canonical NF-κB activation requires the assembly of an I-kappa-B kinase (IKK) complex consisting of the catalytic kinases IKKα and IKKβ, and the regulatory component IKKγ (NEMO) (Fig. 9). This complex phosphorylates IκB and targets it for proteasomal degradation, releasing the p50/relA (nuclear factor of kappa light polypeptide gene enhancer in B-cells 1/v-rel reticuloendotheliosis viral oncogene homolog A) heterodimer for nuclear translocation (221). IKK complex formation is usually induced after ligand binding to TLR, TNFα, or cytokine receptors. Noncanonical NF-κB activation is stimulated by specific TNF receptor family members, which releases p52 (nuclear factor of kappa light polypeptide gene enhancer in B-cells 2) (221). Like HIF, NF-κB is activated by ROS [for overview, see (197)] and hypoxia. Hypoxic activation of NF-κB is weaker compared to proinflammatory activation, but still is detectable in vivo (71). Mechanistically, several options have been proposed, for example, tyrosine phosphorylation of IκB, hypoxic-induced transactivation via p42/44 (MAPK 1/3), and PI3K (155). Recently, NF-κB activation under hypoxia was linked to the HIF-system. FIH hydroxylates p105 and IκBα at conserved ankyrin repeat domains, and IKKβ is hydroxylated by PHDs, which causes their inactivation (39, 48) (Fig. 9).

OPEN IN VIEWER

FIG. 9.

Regulation of NF-κB and HIF-1α by hypoxia. Proinflammatory macrophage activation by LPS causes formation of the I-kappa-B kinase (IKK) complex composed of inhibitor of NF-κB (IκB) kinase g (NEMO), IκB kinase α (IKKa), and IκB kinase β (IKKb), which then phosphorylates IκB to release the NF-κB dimer p50 and RelA. PHDs hydroxylate and thereby inactivate IKKβ. Under hypoxia, or by blocking PHD activity with dimethyl oxalylglycine (DMOG), NF-κB is activated, and HIF-1α is stabilized. Both transcription factors translocate to the nucleus, bind to the DNA, and recruit cofactors such as E1A-binding protein p300 (p300/CBP) to induce, at least in part, the overlapping target genes, that is, iNOS. ROS in activated macrophages attenuates PHD activity, but additionally increase binding of p300/CBP to HIF-1. For details, see the text.

Analyzing HIF-1α expression in classically activated macrophages shows an increase in HIF-1α mRNA [see (53)]. An NF-κB binding site was identified within the HIF-1α promoter, and binding of several NF-κB members was confirmed (19, 304). Studies in mice lacking IKKβ pointed to NF-κB as a mandatory transcriptional activator of HIF-1α. Apparently, a basal NF-κB activity is required for HIF-1α protein accumulation under hypoxia (239). Transcription factors such as Krüppel-like factor 2 that are able to modulate NF-κB activity also alter HIF-1α accumulation in macrophages (179). Therefore, HIF-1α accumulation/HIF-1 activation in classically activated macrophages integrates the signals of PHD inactivation, increased PI3K-mediated HIF-1α translation, and/or an NF-κB-induced HIF-1α mRNA increase (Fig. 8). Hypoxia-induced NF-κB activation may cooperate with activated HIF-1, which can support/enhance macrophage function in the hypoxic inflammatory microenvironment.

The interaction of NF-κB and HIF-1 continues at the level of transcriptional activation. Copper metabolism (Murr1) domain-containing 1 is transported out of the nucleus during hypoxia, and this transport is required for activation of NF-κB and HIF-1 (200). After DNA binding, HIF-1 recruits cofactors such as E1A-binding protein p300 (p300/CBP) to initiate target gene transcription. Binding of cofactors is inhibited by FIH-dependent asparagyl hydroxylation. In macrophages, FIH activity is suppressed by Mint3, which sequesters FIH to the perinuclear region and consequently enhances target gene expression (246). The interaction of HIF-1 with p300 is also disturbed by SUMOylation of p300. Micromolar concentrations of H₂O₂ enhance stability and abundance of SENP3 (SUMO1/sentrin/SMT3-specific peptidase 3) in the nucleoplasm to initiate the removal of the SUMOylation sites in p300, thereby activating the HIF-1 transcriptional activity, which might be of particular relevance in the pro-oxidant milieu created by classically activated macrophages (121). On the other side, in classically activated macrophages, NF-κB and HIF-1 share the coactivator p300, and NF-κB competes for coactivator binding, to repress HIF-1 target gene expression (190). In contrast, genes such as iNOS are induced by both HIF-1 and NF-κB and are synergistically induced under hypoxic and proinflammatory conditions (189). Therefore, it seems to be a gene-specific effect whether the interaction of HIF-1 with NF-κB acts synergistically or inhibitory. To generalize, in classically activated macrophages, NF-κB and HIF-1 are activated in parallel and share transactivation mechanisms to induce common target genes, which implies a concerted biological action in these cells (for summary see Fig. 9).

This is reflected in an in vivo colitis model, where administration of dimethyloxallyl glycine (DMOG) activated HIF-1 as well as NF-κB and conferred protection (49). Although HIF is considered to mediate prosurvival signals under hypoxia, Shah et al. demonstrated that activation of HIF-1 increases damage in this dextran–sodium sulfate-induced colitis model (258), implying that NF-κB accounts for the protective effect of DMOG. Protection by DMOG was also confirmed in TNFα-induced loss of an ileal barrier function by reducing apoptosis in enterocytes (117). In vitro experiments suggest an HIF-dependent inhibition of Fas-associated death domain protein expression that is required for TNFα-induced apoptosis, showing that HIF is at least partly involved in the protective effects of DMOG in the intestine. On the other hand, HIF is involved in the induction of several cytokines in response to LPS, and lethality of LPS-induced endotoxic shock is largely abolished in myeloid-specific HIF-1α and HIF-2α knockout mice (46, 125, 224). The phenotype is reversed by knockout of PHD3 with enhanced lethality in endotoxic shock or sepsis, accompanied by the enhanced production of cytokines, migration of macrophages into organs, and enhanced organ damage (149). The increased production of cytokines resulted from an enhanced activation of HIF-1 and NF-κB, but not of HIF-2, in PHD3 knockout cells, again showing a close relationship of HIF-1 and NF-κB in classically activated macrophages. Surprisingly, the knockout of PHD1 did not alter the LPS-induced lethality at all, while a haplodeficiency of PHD2 slightly decreased the survival of animals, leaving the possibility that a complete knockout of PHD2 might also enhance lethality in septic mouse models. In most cells, PHD2 was established as the most important PHD to regulate HIFα protein expression as reflected by the embryonic lethality of a complete PHD2 knockout in mice (282). Nevertheless, in classically activated immune cells, PHD3 seems to have a prominent role.

Parallel to bacterial infection, eukaryotes such as Leishmania induce classical activation of macrophages (302). The resulting production of oxidants such as O₂ ⁻, H₂O₂, or NO is designated to destroy the phagocytosed parasite. Although some parasite stages are sensitive toward oxidants, parasites may survive within the macrophage by eliminating ROS via their own antioxidant defense system. Hypoxia allows macrophages to control Leishmania, an effect likely involving ROS, because treatment with antioxidants such as N-acetylcysteine or ebselen inhibits these activities (52). Although accumulation of both HIFα subunits in infected macrophages was demonstrated, the details of this process are unclear.

E. HIF in alternatively activated macrophages

Besides fighting invaded pathogens, wound healing is a major function of macrophages. Wound-healing macrophages are characterized by a lower ability to present antigens and to produce proinflammatory cytokines, but they secret components of the ECM and polyamines. Administration IL-4 to a wound in vivo decreased the wound size and increased collagen expression (29). HIF-1α is expressed in wounds (336), and a knockout of HIF-1α in myeloid cells enhanced early wound closure, generating an increased number of activated keratinocytes (218).

Macrophages are able to accumulate HIF-1α and HIF-2α. While HIF-1α accumulates in classically activated macrophages, HIF-2α is barely detectable (283). In contrast, macrophages polarized by IL-4 toward a wound-healing phenotype show increased HIF-2α, but reduced HIF-1α, accumulation, implying opposing functions of macrophage HIF-1 and HIF-2. This is in contrast to the studies indicating impaired acute inflammatory responses in myeloid-specific HIF-2α knockout mice that parallel the findings in myeloid-specific HIF-1α knockout mice (46, 125, 224). An explanation might be that during inflammation, macrophages change their phenotype resulting in a mixed macrophage population in vivo, allowing the simultaneous accumulation of HIF-1α and HIF-2α. This is particularly well established for tumor tissue, where TAMs express both HIF isoforms and acquire a mixed phenotype with features of all three macrophage phenotypes (199, 285). In tumor cells and TAMs, STAT3 is required for HIF-1α expression and necessary for the proangiogenic function of these cells (96, 213). In addition, STAT3 directly interacts and stabilizes HIF-1α and increases its hypoxic accumulation (136). Apoptotic tumor cells, which are abundant in tumors, provoked transcriptional nuclear factor of activated T-cell (NFAT)-dependent upregulation of HIF-1α mRNA, subsequent protein expression, and HIF-1 activity in macrophages (113). Mechanistically, sphingosine-1-phosphate and transforming growth factor-β were identified as the activators of NFAT to cause transcription of HIF-1α mRNA. In contrast to HIF-1α, no information is available whether HIF-2α can also be transcriptionally regulated.

Growing evidence suggests that we urgently need to understand the role of tumor-infiltrating cells such as macrophages to entirely explain tumor formation and progression. The tumor microenvironment impairs the antitumor activity of macrophages and other immune cells, but enhances their tumor-supporting properties (146). This is reflected by the phenotype of associated macrophages, characterized by low arginase1 and high iNOS expression, observed in tissues surrounding various tumors as early as 1 week after carcinogen treatment (236). Similar activation is also observed in bone marrow-derived macrophages of these mice, but not in tissue-resident macrophages of intact organs, limiting the influence of the tumor microenvironment to the tumor site and the bone marrow. Hypoxia is characteristic for tumors, and exploring tumor models using either myeloid-specific HIF-1α or HIF-2α knockout mice revealed that each knockout reduced tumor mass and/or tumor grade by different mechanisms (58, 125). HIF-1α knockout macrophages show a limited ability to reduce T-cell proliferation, which was likewise reduced in wild-type macrophages by inhibiting iNOS. Using an in vitro spheroid model, Werno et al. demonstrated that HIF-1α knockout macrophages invading tumor spheroid showed enhanced signs of alternative activation (325). In contrast, HIF-2α-deficient macrophages are impaired in their ability to invade the tumor sites (125). Apparently, in TAMs, both HIF isoforms are important for a macrophage tumor-promoting function.

An important, but unresolved, question is whether hypoxia or HIF contributes to macrophage polarization. Their inflammatory products characterize classically activated macrophages. As NOX2 and iNOS use O₂ to produce ROS or NO, their activities decrease in a hypoxic atmosphere, limiting macrophage proinflammatory functions (186, 240). On the other side, HIF-1 increases the expression of markers for classical activation such as iNOS. Haplodeficiency of PHD2 skews macrophages toward an alternative and/or proangiogenic phenotype (284). Analysis of HIF-1/-2 and NF-κB activation showed that HIF activity was not significantly enhanced in haplodeficient cells, compared to complete knockouts, while NF-κB was activated and seemed to convey the phenotype shift. Macrophages bearing a genetic alteration toward enhanced ROS formation and HIF-1 activation showed increased classical, but decreased alternative, activation markers (314). An HIF-1α knockdown indicated that indeed HIF-1 may enhance classical activation, but still autonomous cell activation was necessary to unleash the HIF-1-promoting effect. In line, the expression of alternative activation markers such as scavenger receptor 1/CD204 was suppressed by HIF-1 (263). As summarized in Figure 10, one may hypothesize that HIF-1 promotes classical activation, while NF-κB more likely causes alternative polarization under hypoxia.

OPEN IN VIEWER

FIG. 10.

Impact of HIF-1α and HIF-2α on macrophage polarization. HIF-1 and HIF-2 share, in part, similar targets, but the individual knockout of the HIF-α subunits in macrophages provokes different phenotypes. HIF-1 enhances glycolysis, the antibacterial capacity, including iNOS, but decreases expression of CD206. Although antioxidant enzymes are HIF-2 regulated, sestrin2 is an HIF-1 target in macrophages. Cytokine production is attenuated by the knockdown of both isoforms. While TNFα predominantly is under the influence of HIF-1, IL-6 is HIF-2 dominated. Angiogenesis is affected by HIF-1 (via vascular endothelial growth factor [VEGF]), while HIF-2 antagonizes angiogenesis (VEGF receptor1 [VEGFR1]). For details, see the text.

IX. Nrf2 at the Transition from Classical to Alternative Macrophage Activation

Macrophages recognize and fight pathogens in part by creating a toxic environment. Therefore, self-protective mechanisms are required to guarantee their survival. These should be turned on tightly coupled to redox changes occurring during macrophage activation. The most prominent redox-sensitive transcription factor accounting for protective mechanisms is Nrf2. Nrf2 is a CNC-bZIP (Cap‘n'collar-basic region leucine zipper) transcription factor (127). As shown in Figure 11, under resting conditions, Nrf2 binds to the cytosolic inhibitor Kelch-like ECH-associated protein 1 (Keap1)/INrf2, an adaptor protein to the cullin-3 E3 ubiquitin ligase complex (308), which targets Nrf2 for proteasomal degradation to keep the cytosolic level of Nrf2 low. Keap1 is a sensor of electrophiles and oxidants (333), containing 27, in part highly redox-active cysteine residues (Fig. 12).

OPEN IN VIEWER

FIG. 11.

Mechanisms of nuclear erythroid-2-p45-related factor 2 (Nrf2) accumulation. Under basal conditions, (A) Nrf2 binds to its cytosolic inhibitor Kelch-like ECH-associated protein 1 (Keap1)/INrf2, an adaptor protein to the cullin-3 (Cul3) E3 ubiquitin ligase complex, which targets Nrf2 for proteasomal degradation. After oxidative, nitrosative, or electrophilic stress, (B) Keap1 releases Cul3 and Nrf2, which stabilizes Nrf2, with the subsequent binding of DnaJ homolog 1 (DJ1). This complex mediates the nuclear import of Nrf2. Protein kinase C (PKC) conferred Nrf2 phosphorylation at serine 40 (S40) to dissociate Keap1 and Cul3 from Nrf2, followed by its nuclear import with DJ-1 (C). In the nucleus, Nrf2 binds to antioxidant element-binding sites (D) and induces the expression of defensive genes after coassociation to small musculoaponeurotic fibrosarcoma oncogene homologues, activating transcription factor (ATF)-4, or JunD, as well as to the transcriptional coactivator p300/CBP. Nuclear Nrf2 is exported from the nucleus after GSK-3β-dependent Fyn phosphorylation (E), which itself translocates to the nucleus to phosphorylate Nrf2 at tyrosine568 (Y568). In the cytosol, Nrf2 is again degraded via Keap1 (A).

OPEN IN VIEWER

FIG. 12.

Hinge-and-latch model of Keap1 binding to Nrf2. (A) Under basal conditions, Keap1 forms a homodimer via their broad complex, tram–track and bric-a-brac domains (BTB). This domain contains one out of three cysteine residues (Cys151), which are important for the redox modification of Keap1. In the intervening region (IVR), two other relevant cysteine residues are located (Cys273 and Cys288). The sulfhydryl groups (–SH) of these cysteine residues remain unaltered and partially coordinate Zn²⁺. One Keap1 molecule of the homodimer binds via its double glycine repeat and the C-terminal region domain (DC) to the conserved amino acid motif aspartic acid-leucine-glycine (DLG) of Nrf2 with low affinity (latch). The other Keap molecule binds with its DC domain to the conserved amino acid motif glutamic acid-threonine-glycine-glutamic acid (ETGE) of Nrf2 with high affinity (hinge). (B) After oxidative, carcinogenic, or electrophilic stress, the three –SH groups (Cys151, Cys 273, and Cys288) are oxidatively modified (-S-E). This releases Keap1 from the Nrf2 DLG motif, whereas binding to the Nrf2 ETGE motif remains intact. However, Cul3 association is attenuated, thus stabilizing Nrf2.

Apart from ROS-dependent modifications of Keap1 cysteine residues, electrophilic-dependent stress responses acting via inhibition of phosphatases can provoke Nrf2 phosphorylation and subsequent stabilization. Moreover, the protein human DnaJ homolog 1 inhibits Nrf2 proteasomal degradation by blocking binding of Keap1 to Nrf2 (38). Stable Nrf2 translocates to the nucleus and binds to the antioxidant-responsive element sites in the promoters of target genes with other bZIP transcription factors such as small Maf proteins (173), activating transcription factor 4 (ATF4) (111), c-Fos/Fra (148, 306), or JunD (294). Moreover, binding of the coactivator p300/CBP (275) enhances transcription of target genes such as glutathione S-transferase, NADP(H) quinone oxidoreductase 1, heme oxygenase-1 (HO-1), uridine 5′-diphosphate-glucoronyl transferase 1a6, epoxide hydrolase, glutathione peroxidase-2, peroxiredoxins, glutathione reductase, or ATF3 (34, 118). These are members of the phase II antioxidant and detoxifying systems, which are important for macrophage survival. Moreover, ATF3 lowers cytokine expression to protect against endotoxic shock (118). Considering the protective role of Nrf2, it is not surprising that a conditional LysM-Cre-driven Nrf2 knockout in the myeloid lineage increases inflammatory response after LPS challenge (153). LPS-treated Nrf2-deficient macrophages enhance ROS formation (154), which augments the TLR4 transport from the trans-Golgi network to the cell surface, consequently amplifying TLR4 downstream signaling (153). This effect was absent in Nrf2^−/− NOX2^−/− double-knockout mice (154). Likely, Brutons tyrosine kinase mediates TLR4-dependent activation of Nrf2 (307). The importance of Nrf2 in macrophages is evident for several diseases, often associated with macrophage malfunction and frequently characterized by chronic inflammation. For example, chronic obstructive pulmonary disease (COPD) improves when Nrf2 is activated or constitutively overexpressed in mice (108, 278). Enhancing Nrf2 activity stimulates the phagocytic capacity of alveolar macrophages, which allows the efficient removal of pathogens and probably apoptotic cells. Pulmonary Nrf2 dominates in the epithelium and in alveolar macrophages, but the mechanisms why Nrf2 is lost during COPD progression remain unclear (100, 181). However, when Nrf2 is exogenously activated in a COPD mouse model by sulforaphane or an imidazole derivative of 2-cyano-3,12-dioxooleana-1,9(11)-dien-28-oic acid, lung emphysema is clearly reduced (181, 277).

Another chronic inflammatory disease affected by Nrf2 is atherosclerosis. Disruption of Nrf2 attenuates apoplipoprotein E (ApoE)-mediated atherosclerosis in mice (276). During atherosclerosis, a specific macrophage phenotype develops in response to atherogenic lipids via Nrf2-dependent gene expression (1). These Mox macrophages are, compared to classically activated or proresolving wound-healing macrophages, characterized by a unique expression pattern of antioxidative and detoxifying genes. Among others, decreased phagocytosis and reduced migratory properties are two features, although as of yet, it has not been clarified whether these macrophages are pro- or antiatherogenic. Understanding the role of Nrf2 in deactivating oxLDL-generated foam cells, we noticed attenuated ROS as well as proinflammatory cytokine production (159). In foam macrophages from Nrf2^−/− mice, LPS-induced proinflammatory responses were restored. Mechanistically, oxLDL induced Nrf2-dependent antioxidative responses, which in turn attenuated the ROS-dependent activation of CCAAT/enhancer-binding protein (C/EBP) family members to reduce proinflammatory cytokine production. In experimental animal models of atherosclerosis, deletion of Nrf2 is protective. Therefore, Nrf2 seems to have a proinflammatory function as well. In line, cholesterol crystals activate Nrf2 and the NOD-like receptor family, pyrin domain containing 3 (NLRP3) inflammasome and thus IL-1β production (77).

During acute inflammation, like sepsis, macrophage Nrf2 appears in the focus as well. Sepsis is characterized by a hyperinflammatory-phase with a cytokine storm of classically activated macrophages (266), contributing to a systemic inflammatory response. During this phase, NOX2 generates ROS (154), which stabilize Nrf2 to initiate self-protection of macrophages (148, 154). Normally, a second, hypoinflammatory phase, also known as immune paralysis, follows, because macrophages acquire a proresolving, tolerant phenotype (151). In these macrophages, NOX2 is inhibited (132, 311), making Nrf2 stabilization by ROS unlikely.

Often, oxidative signals activate more than one redox-sensitive transcription factor (335) with the option that they may substitute each other (95). One example is HIF-1α, which contributes to HO-1 expression during alcohol-induced inflammation and liver injury in Kupffer cells. The knockdown of HIF-1α and Nrf2 completely blocked HO-1 expression, suggesting a mutual effect of both transcriptional regulators (335). Interestingly, in smokers with a primary spontaneous pneumothorax, the Nrf2 target HO-1 is exclusively induced by HIF-1 (95). Moreover, there is also a link between Nrf2 and the second transcription factor family described here, the PPARs.

X. Alternative Macrophage Activation and PPARs

Three members of the PPAR nuclear receptor family have been described. PPARα is predominantly expressed in the liver, controlling fatty acid oxidation (56). PPARβ/δ is ubiquitously expressed, regulating fatty acid metabolism, mitochondrial respiration, thermogenesis, muscle lipid metabolism, and macrophage phenotype switching (56). PPARγ is also expressed in adipocytes, macrophages, and DCs (205, 293). All three members are activated by nonesterified fatty acids, polyunsaturated fatty acids, eicosanoids, or prostanoids, provoking binding to an identical PPAR-responsive element (PPRE) sequence motif. Predominantly, PPARγ has been established as an anti-inflammatory macrophage modulator. Therefore, we focus on this family member. In macrophages, PPARγ causes a transrepression of proinflammatory genes. As shown in Figure 13, ligand-activated PPARγ binds as a heterodimer with retinoid x receptor to the PPRE of target genes (ABCA1, CD206, and CD36) (20, 210), which mainly affects lipid uptake and export.

OPEN IN VIEWER

FIG. 13.

PPARγ-dependent transactivation of target genes. PPARγ binds as a heterodimer with retinoid x receptor (RXR) to PPAR-responsive elements (PPRE) in the promoter regions of target genes. After coactivator binding, expression of target genes is induced. Biological outcomes are indicated. For details, see the text.

However, as identified in thioglycolate-elicited macrophages, PPARγ initiates a potent anti-inflammatory response (238), with several potential explanations. PPARγ not only transactivates the genes by binding to their PPRE, it also acts in a non-DNA-bound manner. This involves protein–protein interactions between PPARγ and proinflammatory transcription factors such as AP-1, NF-AT, NF-κB, and STAT-1 (237), as shown in Figure 14A. By scavenging these transcription factors, expression of proinflammatory genes is blocked. Binding to transcriptional coactivators, for example, steroid receptor co-activator 1 or p300/CBP, PPARγ inhibits AP-1, and reduces NF-κB-dependent gene induction (Fig. 14B).

OPEN IN VIEWER

FIG. 14.

Anti-inflammatory roles of PPARγ not requiring DNA binding. PPARγ binds a heterodimer with RXR to proinflammatory transcription factors (A) and transcriptional coactivators (B). This binding to PPARγ/RXR attenuates an anti-inflammatory macrophage phenotype. Moreover, PPARγ prevents NF-κB expression by inhibiting nuclear receptor corepressor (NCoR) degradation. This repressor remains bound to DNA by SUMOylated PPARγ (C), blocking proinflammatory gene expression. By a so-far unknown mechanism, the PPARγ/RXR heterodimer attenuates mitogen-activated protein kinase (MAPK) activation (D), to block the downstream transcription factors. Cytosolic PPARγ binds to PKCα to inhibit its activation and translocation to the cell membrane, subsequently blocking NOX activation and ROS formation (E).

These coactivator complexes possess histone acetyltransferase activity, allowing the transcription machinery to gain access to the promoter region (51). As shown in Figure 14C, SUMOylated PPARγ transrepresses promoter activation of the transcription factor NF-κB by inhibiting the removal of the corepressor complex nuclear receptor corepressor/silencing mediator of retinoid and thyroid hormone receptors (130). In addition, PPARγ can block MAPK by a so-far unknown mechanism (Figure 14D). MAPK inhibition prevents phosphorylation and thus activation of downstream transcription factors that are necessary for MAPK-dependent proinflammatory gene expression (55). Casein kinase-II phosphorylation of PPARγ makes PPARγ accessible for chromosomal region maintenance 1/exportin1 nuclear export machinery, to localize PPARγ to the cytosol (312). There, PPARγ can directly bind to protein kinase C (PKC)α to interfere with the PKCα translocation to the cell membrane and thus its activation (Fig. 14E). Blocking the PKCα activity attenuates NOX-dependent ROS formation, which contributes to monocyte/macrophage desensitization because of apoptotic cell phagocytosis (132, 311). It is assumed that this inhibitory function of PPARγ reduces redox stress of these desensitized, compared to classically activated, macrophages. However, a detailed description of the macrophage redox status related to different macrophage phenotypes remains elusive.

The mechanism how apoptotic cells activate PPARγ in macrophages is unknown. One explanation might be redox regulation of PPARγ. PPARγ contains nine cysteine residues, eight located in the two zinc-finger motifs of the DNA-binding domain and one is present at the 3′-terminus of the hinge domain. Hypothetically, redox-modified cysteine residues in the zinc finger can block DNA binding and favor protein–protein interactions. A redox-modified cysteine residue located in the hinge domain might affect ligand binding, consequently altering downstream signaling. Interestingly, two studies, although not performed in macrophages, observed activation or deactivation of PPARγ-dependent gene expression in response to lower (291) or higher (16) ROS, thus supporting the redox hypothesis. Future work is necessary to analyze to which extent PPARγ redox modification alters its function in distinct macrophage subtypes.

As outlined for Nrf2, PPARγ is linked to atherosclerosis as well. During the initiation phase of atherosclerosis PPARγ activation may reduce disease progression by attenuating the expression of proinflammatory genes (131). At later stages of the disease, PPARγ activation induces the expression of CD36, a scavenger receptor that facilitates the uptake of oxLDL into macrophages already located in the intima, consequently supporting their differentiation into foam cells. Recent data support the concept that CD36 is an Nrf2 target under proinflammatory conditions, whereas PPARγ agonists seem to be ineffective (216), perhaps because PPARγ DNA-binding is blocked due to redox modified cysteine residues. However, PPARγ indirectly stimulates, via liver x receptor, the expression of ABCA1, a reverse cholesterol transporter (33). Moreover, caveolin-1, another player in reverse cholesterol trafficking, is also a PPARγ target gene in macrophages (120). In vivo studies support the assumption of a protective role of PPARγ (205) in various mouse models of atherosclerosis (6, 33, 41, 168). PPARγ deletion in macrophages provokes disease progression, whereas PPARγ activation reduces lesion formation. From these results, an anti-inflammatory role of PPARγ is obvious. In line, experiments using a PPARγ knockout model identified PPARγ to control alternative macrophage polarization. Mice with a functional deletion of PPARγ in the myeloid lineage failed to generate a wound-healing macrophage phenotype (215). However, in a different mouse strain, PPARγ seems to be dispensable for alternative macrophage polarization (183). These apparent contradictions may rather suggest a modulator function of PPARγ in alternative macrophage polarization. A summary of PPARγ and Nrf2 in macrophage polarization is shown in Figure 15. Interfering with the PPARγ activity in macrophages will be a challenging task for the future.

OPEN IN VIEWER

FIG. 15.

Role of Nrf2 and PPARγ in macrophage polarization. Classical activation of macrophages activates NOX2 via PKCα. ROS may kill bacteria or modify Keap1 to release Nrf2 and to activate the protective genes. Under conditions of high ROS production, PPARγ likely is redox modified (RM). This blocks transcription of PPARγ target genes, but does not interfere with its DNA-independent actions (removal of transcription factors and/or coactivators). In alternatively activated macrophages, ROS formation is attenuated by PPARγ-dependent PKCα inhibition, and Nrf2 is degraded. Redox modifications of PPARγ are reversed, allowing to regain transcription of genes such as CD36 or CD206.

Although ligands such as the thiazolidinediones are pharmacologically available and already in use, the development and characterization of drugs belonging to the new class of selective PPARγ modulators, which should have a cell-type-specific effect, will allow to bias PPARγ dependency of macrophage polarization.

XI. ROS Support Classical Activation in Response to Modified Lipoproteins

B. Modified lipoproteins and endoplasmic reticulum stress

Endoplasmic reticulum (ER) stress is a major mechanism underlying macrophage cell death during atherosclerosis development (254). ER stress may be triggered by disturbed ER functions, for example, an increase of misfolded proteins, inhibited glycosylation, altered ER calcium homeostasis, or perturbed ER membrane structures (Fig. 16). Three signaling proteins sense ER stress: inositol-requiring protein-1α, protein kinase RNA–like ER kinase (PERK), and ATF-6. Activation of these proteins triggers a homeostatic unfolded protein response (UPR), aimed at relieving ER stress by reducing general protein synthesis and increasing folding chaperones.

OPEN IN VIEWER

FIG. 16.

ROS in endoplasmic reticulum (ER) stress-induced cell death. ER stress induces an unfolded protein response (UPR), enhancing transcriptional activation of C/EBP-homologous protein (CHOP) and its target ER oxidation 1α (ERO1α). ERO1α activates IP₃ receptor 1 (IP3R1) and calcium release, which activates the calcium-/calmodulin-dependent protein kinase IIγ (CAMKIIγ)-JNK cascade and transcriptional upregulation of NOX2. NOX2 is also activated in an acidic compartment by signals derived from activation of pattern recognition receptors such as Toll-like receptor (TLR)2, heterodimerizing with scavenger receptor CD36, and activation of extracellular signal regulated kinase(s). NOX2-derived ROS further activate CAMKIIγ-JNK, to enhance the UPR signal, thus contributing to the mitochondrial membrane permeabilization and induction of apoptosis. The CAMKIIγ-JNK cascade also induces Fas transcription, activating the extrinsic apoptotic pathway.

A prolonged and excessive UPR can stimulate apoptosis, where C/EBP-homologous protein (CHOP), a downstream target of PERK, plays a major role. Macrophage ER stress by atherogenic stimuli linked to cell death has extensively been characterized. This integrates ER signals and CD36/Toll-like receptor 2 signaling in response to oxidized phospholipids or saturated fatty acids (255). A signal emanating from the ER due to accumulation of unesterified cholesterol upregulates intracellular calcium, via the CHOP-ER oxidation 1α (ERO1α) pathway (171). This signal in turn activates calcium-/calmodulin-dependent protein kinase IIγ (CAMKIIγ), which triggers apoptosis (292). A recent study highlighted the key role of ROS, generated by NOX2, in transmitting detrimental ER signals (172). ER stress activators, such as free cholesterol or 7-ketocholesterol, induced the expression of NOX2 via a CHOP-ERO1α-CAMKIIγ-JNK-pathway, and NOX2-deficient macrophages showed reduced apoptosis upon ER stress induction. Further, NOX2 was necessary for ER stress-induced IP₃ receptor activation and mitochondrial calcium loading as well as depolarization, preceding mitochondrial cell death induction. NOX2 was also involved in scavenger receptor/TLR-activated proapoptotic signaling, provoking the sustained clustering and activation of NOX2 in acidic cellular compartments (255). Although the role of NOX4 in macrophage ER responses is unknown, it induces ER stress and apoptosis in vascular smooth muscle cells (220). In contrast, recent data suggest a cytoprotective role of NOX4 in endothelial cell via activation of Nrf2 (250). Collectively, current evidence supports an important role of NOX2-derived ROS in ER stress-induced macrophage apoptosis. This likely contributes to the formation of the necrotic core and plaque destabilization in advanced atherosclerotic lesions.

XII. ROS: Linking Classical Macrophage and Inflammasome Activation

Inflammasome activation is a part of the innate immune response to pathogens, but also accompanies the development of autoimmune and chronic inflammatory diseases (271, 324). As presented in Figure 17, inflammasomes are multicomponent platforms that contain caspase-1 to process IL-1β and IL-18. Inflammasomes sense a variety of danger signals, including bacteria, viruses, pathogenic crystals, and aggregates, through a family of Nod-like receptors (NLR). Among these, NLRP3 is the most studied NLR, responding to pathogens, exogenous and endogenous danger signals such as extracellular ATP, uric acid, and cholesterol crystals, alum, or silica. Current concepts propose a two-step NLRP3 inflammasome activation. A priming signal, such as LPS, induces transcription of IL-1β and NLRP3 itself, whereas an activating signal causes caspase-1 and IL-1β proteolytic processing.

OPEN IN VIEWER

FIG. 17.

ROS and NOD-like receptor family, pyrin domain containing 3 (NLRP3) inflammasome activation. TLR receptor agonists deliver a priming signal to increase IL-1β and NLRP3 transcription and pro-IL-1β production. Activating stimuli, for example, ATP, stimulate P2X7 receptor-mediated potassium efflux, which promotes mitochondrial ROS generation and oxidation of mitochondrial DNA (mtDNA). Oxidized mtDNA binds NLRP3 and drives the inflammasome assembly, consisting of oligomers of NLRP3, apoptosis-associated speck-like protein containing a CARD (ASC) and pro-caspase-1, culminating in caspase-1 cleavage, and activation and IL-1β processing. Lysosomal rupture and cathepsin release induced by cholesterol or monosodium urate crystals also activate the inflammasome. Damaged, ROS-producing mitochondria are removed by mitophagy, while inhibition of mitophagy by saturated fatty acids (SFA) increases mtROS and enhances inflammasome activation. ROS also support a priming signal.

The nature of the activating signal is subjected to intensive investigations, with three potential events being implicated: K⁺ efflux, lysosomal rupture, and ROS production. The involvement of ROS in NLRP3 inflammasome activation was first suggested by ROS scavenger interventions (47, 223). A later study proposed mitochondrial ROS to activate the NLRP3 inflammasome (340). In unstimulated cells, NLRP3 colocalizes with ER structures, while another inflammasome component, apoptosis-associated speck-like protein containing a CARD, is cytosolic. Inflammasome-activating stimuli relocalized both proteins to the perinuclear ER and mitochondrion-containing structures, where they may sense ROS generation. This sensing may involve thioredoxin-interacting protein (TXNIP) (339), which dissociates from its binding partner thioredoxin in an ROS-dependent manner and translocates to mitochondria/ER structures, where it directly binds NLRP3. However, this pathway was questioned by experiments in TXNIP-deficient macrophages (185).

Redox regulation of inflammasome activation may be even more complex as outlined recently (244). Several studies suggested a role of the antioxidant defense system in regulating the inflammasome activity. A negative effect of ROS on IL-1β processing was reported for peritoneal macrophages in a study of SOD1-deficient mice (188). This was attributed to glutathionylation and inhibition of caspase-1. Another study suggested a biphasic, redox control mechanism of IL-1β processing in human monocytes (286), whereby initial ROS generation triggers thioredoxin upregulation and cystine/cysteine import/export, which then enhances IL-1β secretion after TLR activation. Differences in the basal redox state among primary monocytes, macrophages, and myeloid cell lines were suggested to explain the discrepancies in IL-1β secretion in response to TLR activation in these cells (28). Further evidence that antioxidant responses contribute to inflammasome activation came from Nrf2 knockout macrophages, which display reduced IL-1β secretion in response to cholesterol crystals, which was proposed to underlie an antiatherosclerotic phenotype of ApoE Nrf2 double knockouts (77). It is worth mentioning that also an excessive reductive environment causes mitochondrial stress and ROS release (338).

It has recently been suggested that ROS may participate in inflammasome priming rather than activation, based on diphenylene iodonium (DPI)- and N-acetylcysteine-sensitive induction of IL1β and Nlrp3 mRNA by LPS, whereas DPI addition after priming failed to modulate inflammasome activation (11). Although the use of the unspecific ROS-scavenging compound N-acetylcysteine points to ROS as an inflammasome activator, NOX-dependent ROS generation can be excluded, taking into consideration that DPI as a competitive inhibitor of flavin-containing cofactors was without effect. A crosstalk between ER stress and the inflammasome activation pathways may exist, as ER stress activators also stimulated IL-1β processing in human macrophages (191). Molecular details of this crosstalk remain elusive, but may involve induction of apoptotic mechanisms as discussed above.

In conclusion, most of the experimental evidence thus far supports a critical role of mitochondrial ROS for NLRP3 inflammasome activation. However, further research is needed to provide mechanistic details on how NLRP3 senses ROS signals and whether pathways involving lysosomal rupture crosstalk to mitochondria. Spatiotemporal changes in ROS generation or the activity of antioxidant defense systems may also modulate the components of the inflammasome-signaling cascade, such as redox regulation of caspase-1.

XIII. Autophagy Dampens ROS-Dependent Inflammasome Activation

Autophagy is a process whereby cellular components, including whole organelles, are degraded in specialized phagolysosomal compartments known as autophagosomes. Autophagy emerges as a fundamental process in many aspects of immune cell physiology (54). Autophagosome formation requires the autophagy-related gene (Atg) protein family members, is followed by autophagosome maturation, and fusion with the endosomal and lysosomal organelles for cargo degradation. Autophagy contributes to cell homeostasis, providing nutrients during starvation, and is critical for cell quality control and disposal of dysfunctional or damaged organelles. In immune cells, autophagy participates in pathogen elimination, either directly or downstream of TLR activation (54). Recent data also showed a critical role of autophagy in regulating the inflammasome activity. Thus, a study of Zhou et al. revealed a role of autophagy-mediated disposal of damaged ROS-producing mitochondria to suppress IL-1β secretion (340). When autophagy was inhibited by knockdown of beclin 1 or Atg5, inflammasome activation by monosodium urate crystals or nigericin was enhanced. Mitochondrial ROS were also critical to explain the effects of autophagy inhibition on inflammasome activation (206). It was shown that in the absence of autophagic proteins LC3B or beclin 1, defective mitochondria accumulated upon treatment with LPS and ATP, which was associated with increased mitochondrial ROS generation. The involvement of mitochondrial ROS in inflammasome activation was further confirmed by the lower IL-1β secretion in cells devoid of mitochondrial DNA or in macrophages treated with the mitochondrion-specific ROS scavenger Mito-TEMPO. Defective mitophagy may also contribute to inflammasome activation by saturated fatty acids (Fig. 17). Mitophagy decreased during combined LPS/palmitate treatment of macrophages, which was associated with increased ROS formation and inflammasome activation (323). In contrast, increasing autophagy by adenosine monophosphate (AMP)-dependent protein kinase (AMPK) activation may rescue cells and prevent ROS formation and inflammasome activation. Another role for autophagy may be the sequestration and degradation of pro-IL-1β (107). Thus, inhibition of autophagy would sustain pro-IL1β for processing by the inflammasome. It has been shown recently that several NLRP3-activating stimuli trigger autophagy of whole inflammasomes, thereby limiting IL-1β production (262).

In the context of atherosclerosis, macrophages deficient in autophagy promoted atherosclerosis in ApoE knockout mice by activating the inflammasome and increasing IL-1β production (234). Inflammasome activation during atherosclerosis proceeds through accumulation of cholesterol crystals and subsequent lysosomal destabilization (61). In another atherosclerotic mouse model, autophagy deficiency increased macrophage apoptosis, while decreasing efferocytosis of dead cells (174). In both cases, the proatherosclerotic effects of autophagy deficiency were linked to increased ROS generation, although the source of ROS may differ (mtROS in the first model and NOX in the second).

XIV. The Redox Sensor AMPK Favors Regulatory Macrophage Activation

A critical component of homeostatic cell responses toward a variety of stress signals is AMPK (105). AMPK senses energy depletion by means of regulation through the AMP/ATP and adenosine diphosphate/ATP ratios. AMPK activation switches off anabolic processes, such as protein and fatty acid synthesis, while promoting catabolic processes, for example, glycolysis, fatty acid oxidation, and autophagy. In macrophages, AMPK supports regulatory macrophage polarization induced by IL-10 and TGFβ (245), probably by activating Akt. Further, AMPK counteracts fatty acid-induced inflammatory responses of macrophages either by increasing β-oxidation (82) or by activating NAD⁺-dependent deacetylase SIRT1 (334), thus promoting deacetylation and inactivation of NFκB. These two mechanisms may actually be linked, since AMPK-driven changes of the NAD⁺/NADH ratio and SIRT1 activation were shown to demand β-oxidation (27). In addition, an AMPK-dependent mitophagy may play an important role in ROS-mediated inflammasome activation by saturated fatty acids as discussed above (323). Studies from other cellular systems have shown several levels of interplay between ROS and AMPK. AMPK can be activated by ROS directly through the redox modification of cysteines (345) as well as by redox regulation of its upstream activator liver kinase B1 homolog (2, 331) and redox-induced changes of intracellular calcium, activating the AMPK upstream kinase CAMKKβ (202). In turn, AMPK activity may affect ROS formation by upregulating MnSOD, thioredoxin (42, 160), or UCP2 (332). AMPK may also interfere with NOX activation by attenuating NOX2 expression in response to angiotensin (251) or attenuating NFκB-dependent expression of several NOX components by inhibiting IκB proteasomal degradation (315). Whether AMPK interferes with macrophage NOX activation is still unknown.

XV. Iron Availability Dictates Macrophage Polarization

Iron not only plays a central role in different metabolic pathways as an essential cofactor of proteins/enzymes, but also is of crucial importance for the regulation of immune cell function. Recently, a role of iron in immunity, specifically in macrophage polarization, emerged (Fig. 18). Macrophages have evolved multiple pathways to acquire, recycle, process, store, and transport iron, including expression of the receptor for the main iron carrier protein transferrin, the iron storage protein ferritin, and lactoferrin receptors to take up the antimicrobial iron transport protein lactoferrin.

OPEN IN VIEWER

FIG. 18.

Role of iron in macrophage polarization. Role of iron in controlling macrophage immune responses associated with changes in iron transport, storage, recycling, and acquisition. For details, see the text.

Iron can also be ingested by transmembrane acquisition, but the major iron uptake route in macrophages is phagocytosis of aging erythrocytes. In addition, macrophages express the surface receptor CD91, the hemopexin receptor, which takes up heme bound to the heme-sequestering protein hemopexin. The uptake of heme iron induces HO-1 in macrophages. Iron is needed for cytotoxic immune defense mechanisms as a central catalytic compound for the production of toxic hydroxyl radicals (243). Cytokines and radicals produced and released by immune cells are in turn crucial to intracellular iron homeostasis. Estimated 10%–20% of ingested iron remains in macrophages as a labile iron pool, which is necessary to maintain cellular iron homeostasis and to fulfill their effector functions. These are often regulated at the post-transcriptional level by iron-binding proteins (IRP), which bind to the RNA stem–loop structures, the iron-responsive elements (IREs) in the untranslated region (UTRs) of their target mRNAs. IREs have been identified in the 5′-UTR of mRNAs coding for proteins related to iron storage (ferritin), iron incorporation, for example, erythroid aminolevulinic acid synthase (e-ALAS), which catalyzes the synthesis of d-aminolevulinic acid (ALA), the first common precursor in the biosynthesis of tetrapyrroles, and iron export (ferroportin [FPN]). For iron uptake, protein (transferrin receptor [TfR]) IREs are found in the 3′-UTR. A functional IRE has also been noticed in the mRNA for HIF-2α, suggesting a negative feedback control of the HIF-mediated response under conditions of limited iron availability (343).

In response to inflammation, iron is sequestered specifically in macrophages, which leads to a state of systemic inflammation-associated anemia. Proinflammatory cytokines and acute-phase proteins provoke iron sequestration in macrophages. LPS and proinflammatory cytokines such as IL-6 are described to induce hepcidin, the main iron regulator for whole-body iron homeostasis (289). Hepcidin is an acute-phase protein, mainly produced in the liver, which regulates iron transport across the gut mucosa. Hepcidin is also implicated in iron retention in macrophages, as it directly binds to FPN, the only known cellular iron exporter, to provoke its internalization and degradation. It also seems that iron retention in macrophages plays a crucial role in combating bacterial infections (208). The availability of iron influences the production of cytokines and thus significantly modulates macrophage polarization (see Fig. 18). In this sense, iron chelation results in marked suppression of mononuclear cell activation by inhibiting TNFα production (327). Polarization of macrophages plays a crucial role for their function in iron metabolism. Wound-healing macrophages are characterized by a rapid mobilization of their intracellular iron pools releasing it to the iron-lacking environment (235). During inflammation, proinflammatory cytokines downregulate the expression of the iron exporter FPN (43), whereby iron remains sequestered in macrophages. This seems to be a defense mechanism to withdraw iron from invading pathogens. Polarization of macrophages dictates the expression of genes involved in iron metabolism. Classically activated macrophages express high ferritin, natural resistance-associated protein 1 (Nramp-1), HIF-1α, SOD1, and low TfR. In contrast, wound-healing macrophages upregulate TfRs, divalent metal transporter-1, IRP1, IRP2, FPN, and HO-1. Thereby, the macrophage iron status influences its immune effector functions by regulating cytokine formation and the induction of the antimicrobial machinery (Fig. 18). Since iron is an essential compound for bacterial growth, the control over iron homeostasis is central. Phagolysosomal proteins such as Nramp-1 that are associated to resistance toward infection can act as iron transporters. Iron-laden macrophages lose their ability to kill intracellular pathogens by IFNγ-mediated pathways due to reduced formation of NO. In the presence of iron, iNOS transcription is blocked because iron inhibits binding of C/EBPβ and HIF-1 to the iNOS promoter (3). Consequently, macrophages show a lower inducibility in response to cytokines. The induction of lipocalin-2 (Lcn-2) by LPS could represent an alternative pathway of sequestering iron. Lcn-2 acts as a potent bacteriostatic agent due to its high affinity to iron, thereby sequestering iron from bacterial siderophores (198).

Macrophages also deliver iron to their neighborhood (81, 235). It was demonstrated that these macrophages express higher levels of HO-1 and FPN to increase the iron pool within the local microenvironment. Iron release from wound-healing macrophages contributes to their roles in tissue repair and tumor growth (Fig. 18). Higher iron availability could influence the growth of adjacent fibroblasts, since iron represents an essential factor for cellular proliferation and DNA synthesis. Iron overload also enhances ECM deposition due to the induction of PHD, which is important for collagen biosynthesis, thereby promoting repair, but possibly also fibrosis. Additionally, we recently described a role of Lcn-2, produced by wound-healing macrophages, during regeneration of an ischemic kidney (137). High FPN iron release may also occur in TAMs. This causes higher iron availability in the tumor microenvironment, which might represent a novel mechanism of protumorigenic macrophage characteristics. Previous data from our group showed that TAMs express and release Lcn-2 to induce proliferation and growth of tumor cells. Lcn-2 is currently discussed as an alternative iron shuttle protein by binding to its specific receptor (NGALR) on the cell surface. In addition, it may function as a possible iron export system because Lcn-2 binds iron with a high affinity inside macrophages and exports iron via exocytosis (Fig. 19). Iron-laden Lcn-2 shuttles to the tumor cell, where the uptake of iron–Lcn-2 complexes ensures their survival and growth. Conversely, the internalization of iron-free Lcn-2 would induce apoptosis, which is driven by iron deficiency and increased expression of the proapoptotic protein Bim (57). It is therefore tempting to speculate that tumor cells specifically take up iron-laden Lcn-2 to ensure their own survival. However, a defined mechanism of the tumorigenic function of Lcn-2 is still missing.

OPEN IN VIEWER

FIG. 19.

Role of lipocalin-2 (Lcn-2) as an alternative iron transporter. Signals generated by tumor cells initiate Lcn-2 expression in macrophages, its loading with iron, and release. Together with iron release via ferroportin, macrophages provide iron to tumor cells, which is used by them to stimulate their survival and growth. For details, see the text.

It can be speculated that alternative iron transport mechanisms mainly occur under pathophysiological conditions, while transferrin-bound iron transport is coupled to physiology. A possible nontransferrin-dependent mechanism may be linked to Lcn-2, since it binds iron with a high affinity and can act independently of transferrin. Additionally, high Lcn-2 concentrations are found in chronic inflammatory diseases such as cancer. Therefore, it may be speculated that Lcn-2 can provide iron to cells under pathophysiological conditions, associated with bacterial infections, inflammation (45), ER stress (261), or anemia (18). Biological functions of Lcn-2 are determined by binding to iron. In contrast to iron-deficient Lcn-2, iron-laden Lcn-2 increases intracellular iron, thereby inhibiting apoptosis by inducing the protective, antiapoptotic protein Bcl-2. It can further be postulated that Lcn-2 may serve as a new iron-handling protein in macrophages, adopting the macrophage to different stimuli.

XVI. Conclusions

The unique feature of macrophages is their high functional diversity, which positions them at the center of pathways that promote inflammation and control its resolution. Sensing environmental and internal signals allows their polarization, which coins macrophage functional properties. Redox signaling contributes to shape the role of macrophages during the onset, propagation, and termination of inflammatory responses. A large number of redox-active components operate simultaneously, being produced by a class of redox-generating systems that comprise NOX enzymes, mitochondria, or NO synthases and are balanced by the antioxidative capacity of a cell. Redox biology adds to macrophage polarization and thus their role in health and, disease. Multiple redox targets, among them several transcription factors such as Nrf2, PPARγ, HIF-1, or HIF-2, transform redox signals into the gene and protein signatures and the associated characteristic functional consequences. Table 2 summarizes information on how distinct macrophage phenotypes are related to health and disease in man or mice (and zebrafish) and how distinct macrophage functional characteristics are linked to defined redox signals.

It is evident that macrophages not only add to control the onset of disease conditions but also determine the course of disease as well as the process of disease resolution. Formation of ROS and/or NO during classical macrophage activation may favor phagocytosis, while little is known whether these distinguished redox signals determine alternative macrophage activation, although NOX4 might contribute to desensitize cells. In the context of lipid overload, macrophages respond with the characteristics of classical activation in conjunction with ER stress and oxidation of mitochondrial DNA. Signatures of classical macrophage activation are connected with activation of HIF-1 and NF-κB, which may operate in concert to boost inflammatory reactions, with the notion that HIF-1 is under redox control. How HIF-2 affects macrophage polarization is less understood. Although a few alternative activation markers are regulated by HIF-2, a generalized concept did not emerge so far. Despite HIF-1 and HIF-2 are associated with macrophage polarization, their transcriptional capacity apparently does not directly drive the polarization program. The redox-sensitive Nrf2 may add to a macrophage safeguard mechanism. When stabilized in classically activated macrophages, it signals back to further limit inflammation and cell self-destruction. Opposed is the role of PPARγ. Initially, this transcriptional machinery is operating at a low level, while it gets active during alternative activation to directly interfere with proinflammatory signaling. Iron sequestration and its release may further add to the macrophage phenotype. Sequestering iron in macrophage depletes the microenvironment, to limit pathogen or tumor cell replication and fostering the production of inflammatory components. In contrast, iron release may be beneficial for cell replication, to promote wound healing with the same mechanism hijacked by tumor cells to stimulate their own survival and growth. In the future, it will be challenging to use this knowledge of redox-regulated processes to shift macrophage polarization and thereby use distinct phenotypes either to limit inflammation or to boost tissue repair and regeneration. In contrast to oxidative stress, redox regulation emerges as a unique navigating mechanism to shape the multiple facets of the continuum of the macrophage subtypes in our body.