|Home | About | Journals | Submit | Contact Us | Français|
Macrophages are strategically located throughout the body tissues, where they ingest and process foreign materials, dead cells and debris and recruit additional macrophages in response to inflammatory signals. They are highly heterogeneous cells that can rapidly change their function in response to local microenvironmental signals. In this Review, we discuss the four stages of orderly inflammation mediated by macrophages: recruitment to tissues; differentiation and activation in situ; conversion to suppressive cells; and restoration of tissue homeostasis. We also discuss the protective and pathogenic functions of the various macrophage subsets in antimicrobial defence, antitumour immune responses, metabolism and obesity, allergy and asthma, tumorigenesis, autoimmunity, atherosclerosis, fibrosis and wound healing. Finally, we briefly discuss the characterization of macrophage heterogeneity in humans.
The mononuclear phagocytic system is generated from committed haematopoietic stem cells located in the bone marrow. Macrophage precursors are released into the circulation as monocytes, and within a few days they seed tissues throughout the body, including the spleen, which serves as a storage reservoir for immature monocytes1. When monocytes migrate from the circulation and extravasate through the endothelium, they differentiate into macrophages or dendritic cells (DCs). Thus, the primary role of monocytes is to replenish the pool of tissue-resident macrophages and DCs in steady state and in response to inflammation. Monocytes, DCs and macrophages, along with neutrophils and mast cells, are ‘professional’ phagocytic cells. Professional phagocytes are distinguished from ‘non-professional’ phagocytes according to how effective they are at phagocytosis2. A major factor that differentiates professional and non-professional phagocytes is that professional phagocytes express a multitude of receptors on their surfaces that detect signals that are not normally found in healthy tissues. For example, scavenger receptors are responsible for binding apoptotic and necrotic cells, opsonized pathogens and cell debris. Moreover, professional phagocytes express Toll-like receptors (TLRs), but the interplay between phagocytic receptors (which initiate and assist in the mechanics of phagocytosis) and pattern recognition receptors (PRRs, such as TLRs, which detect ‘non-self ’ or ‘damage’) is complex. The interplay between these receptors is likely to involve synergistic and antagonistic interactions, including downstream signalling mechanisms within the phagocytic cell that remain largely unknown3,4.
Within the mononuclear phagocyte pool, macrophages are often distinguished from DCs by differential expression of surface makers such as F4/80 (which is encoded by EGF-like module containing, mucin-like, hormone receptor-like sequence 1 (Emr1) and is a useful marker of some but not all macrophages in the mouse), CD11b and CD18 (also known as MAC1), CD68 and Fc receptors (TABLE 1). However, few, if any, known marker combinations can definitively segregate macrophages from myeloid DCs at present because these populations exist on a continuum of development from common myeloid progenitors (BOX 1; TABLE 1).
A major problem in defining macrophages and myeloid dendritic cells (DCs) lies in the fact that they express common cell surface markers, as they both arise from common myeloid precursors. A widespread experimental method to separate DCs from macrophages is based on CD11c expression. However, most, if not all, macrophages express low (or intermediate) amounts of CD11c, and this complicates the interpretation of experiments with CD11c-based cell enrichment or depletion165,166. Moreover, although F4/80 is commonly used as a macrophage marker in the mouse, there are probably some cells classified as DCs that also express F4/80, as well as some macrophages that lack F4/80 expression. Thus, the cell surface marker-based separation strategies are only an enrichment for mononuclear phagocytes that have functional properties relative to DCs or macrophages (for example, antigen presentation capacity is relative to DCs rather than macrophages). Even then, ‘DCs’ and ‘macrophages’ isolated from the same organ can have identical stimulatory effects on naive T cells167, and these issues have been discussed at length168.
Another problem in characterizing myeloid cell populations stems from the existing nomenclature for DC and macrophage subsets. For example, TIP-DCs (tumour-necrosis factor/inducible nitric oxide synthase-producing DCs), which are an inflammatory population of newly recruited myeloid cells, can be identified by a surface marker combination of CD11c+CD11b+MHC class-IIhi. However, rather than ‘DCs’, these cells might be considered to be inflammatory macrophages that have been exposed to Toll-like receptor ligands and cytokines in situ, as macrophages express CD11c, and expression of MHC class-II is likely to be induced by local interferon-γ. Moreover, CD169+ subcapsular lymph node phagocytes are essential for tumour-derived antigen presentation in draining lymph nodes169, and their function (that is, good antigen presentation) is most closely associated with conventional DCs; however, they are called macrophages.
Despite these issues, macrophage and DC subset definition can be substantially refined. Lineages can best be defined by lineage-specific genes, as identified by conditional genetic deletion approaches. For example, ablation of basic leucine zipper transcriptional factor ATF-like 3 (BATF3) causes a complete deficiency in CD103+ DCs in the gut, whereas CD8+ DCs are ablated in the absence of interferon regulatory factor 8 (IRF8), nuclear factor interleukin-3-regulated protein (NFIL3) and at least six other transcription factors, while other mononuclear phagocytes remain intact1,53. Observations about the specificity of gene expression of transcription factors and cell surface proteins can be used as a platform for lineage tracing experiments: the success of CX3C-chemokine receptor 1–green fluorescent protein (CX3CR1–GFP) mice for detection of the circulating monocytes is an example of successful lineage tracing in myeloid cells, whereas notable advances have been made in dissecting the fine details of distinct origins and functional properties of the gut mononuclear phagocytes53,54,134.
In this Review, we provide an overview of the homeostatic, protective and pathogenic functions of the various macrophage subsets in health and disease, and discuss the current obstacles to the complete characterization of macrophage heterogeneity and effector function.
Macrophages are divided into subpopulations based on their anatomical location and functional phenotype5 (FIG. 1). Specialized tissue-resident macrophages include osteoclasts (bone), alveolar macrophages (lung), histiocytes (interstitial connective tissue) and Kupffer cells (liver). The gut is populated with multiple types of macrophages and DCs, which have distinct phenotypes and functions, but work together to maintain tolerance to the gut flora and food (BOX 1). Secondary lymphoid organs also have distinct populations of macrophages that perform unique functions, including marginal zone macrophages in the spleen, which suppress innate and adaptive immunity to apoptotic cells6, and subcapsular sinus macrophages of lymph nodes (LNs), which clear viruses from the lymph and initiate antiviral humoral immune responses7,8. Distinct macrophage subpopulations also patrol so-called immune-privileged sites — such as the brain (microglia), eye and testes — where they are assumed to have central functions in tissue remodelling and homeostasis. These tissue-specific macrophage subpopulations ingest foreign materials and recruit additional macrophages from circulation during an infection or following injury.
Because there is great overlap in surface marker expression between the different macrophage subsets9, a useful characterization approach has been to quantify specific gene expression profiles after cytokine or microbial stimulation10 (TABLE 2). Several macrophage subsets with distinct functions have been described. Classically activated macrophages (M1 macrophages) mediate defence of the host from a variety of bacteria, protozoa and viruses, and have roles in antitumour immunity. Alternatively activated macrophages (M2 macrophages) have anti-inflammatory function and regulate wound healing. ‘Regulatory’ macrophages can secrete large amounts of interleukin-10 (IL-10) in response to Fc receptor-γ ligation11,12. Tumour-associated macrophages (TAMs) suppress antitumour immunity, and myeloid-derived suppressor cells (MDSCs) are linked to TAMs and may be their precursors13. Although there are obvious differences among the M2 macrophage, regulatory macrophage, TAM and MDSC subsets, they all exhibit immune suppressive activity14. Consequently, when stimulated, macrophages suppressive activity14. Consequently, when stimulated, macrophages adopt context-dependent phenotypes that either promote or inhibit host antimicrobial defence, antitumour immunity and inflammatory responses. It is generally believed that macrophages represent a spectrum of activated phenotypes rather than discrete stable subpopulations13. Indeed, numerous studies have documented flexibility in their programming, with macrophages switching from one functional phenotype to another in response to the variable microenvironmental signals of the local milieu15–20.
A conventional approach for studying macrophage activation in vitro is the stimulation of cells (plated on plastic) with microbial agonists or cytokines and the measurement of effector cytokine production and changes in gene expression. However, macrophage responsiveness in vivo is different. Should the vast numbers of macrophages that inhabit the colon, liver and lungs respond so readily to external stimulation, then systemic cytokine production would be continuous. Therefore, tissue macrophages, as well as newly recruited monocytes, are subject to a hierarchy of activation states that ensure baseline tissue homeostasis is the ‘default’ and prevent constant inflammation, which is the underlying cause of numerous chronic diseases.
At steady state, tissue macrophages have intrinsic anti-inflammatory functions. For example, colonic macrophages spend their existence bathed in IL-10 and mute any inflammatory response to the gut flora and their products21,22. Disruption of the normal sources or quantities of IL-10 or IL-10 signalling in immune cells leads to massive inflammation in the gut23. Another specialized macrophage type that suppresses immune responses is the marginal zone macrophages of the spleen, which are required to reduce self-reactivity to apoptotic cells6. Depletion of marginal zone macrophages leads to the formation of DNA-specific antibodies and a systemic lupus erythematosus-like autoimmune syndrome.
An initial level of macrophage activation occurs when early warning signals trigger monocyte recruitment and in situ activation or when IL-4 induces in situ macrophage proliferation24. Tissue damage sensing is probably crucial at the second level of macrophage response, regardless of whether the damage is of a microbial nature. The mechanisms of tissue damage sensing have been discussed in recent reviews25,26. Beyond the initial activation and stimulation of macrophages, cooperative actions of multiple sensors, feedforward cytokine networks and inter-organ communication increase the output of monocytes and neutrophils driving inflammatory responses. Macrophage effectors work together in cell-intrinsic and cell-extrinsic networks27. For example, the production of interferon-γ (IFNγ) by T helper 1 (TH1) cells requires IL-12 production from activated mononuclear phagocytes; IFNγ then stimulates macrophages to activate the antimicrobial arsenal28.
A key component of the next layer of the macrophage response is the production of anti-inflammatory feedback mechanisms that encompass cell-intrinsic signalling feedback loops and cell-extrinsic mechanisms, such as the production of IL-10, which is an essential and non-redundant anti-inflammatory cytokine.
The final layer of macrophage response is the least clear and involves the final decision between chronic inflammation and re-establishment of homeostasis. The understanding of the underlying mechanisms that restore homeostasis after an inflammatory reaction underpins all research efforts related to chronic inflammatory diseases.
Mature macrophages are strategically located throughout the body and perform an important immune surveillance function. They constantly survey their immediate surroundings for signs of tissue damage or invading organisms and are poised to stimulate lymphocytes and other immune cells to respond when danger signals are phagocytosed and/or detected by cell surface receptors. For example, when a macrophage ingests a pathogen, the pathogen becomes trapped in a phagosome, which then fuses with a lysosome unless prevented from doing so by pathogen-specific mechanisms. Within the fused phagolysosome, enzymes and toxic free radicals digest and destroy the pathogen. In addition to fighting infections, resident tissue macrophages are involved in maintaining healthy tissues by removing dead and dying cells and toxic materials. For example, alveolar macrophages facilitate the removal of allergens from the lung, whereas Kupffer cells in liver participate in the clearance of pathogens and toxins from the circulation. Tissue macrophages also suppress inflammation mediated by inflammatory monocytes, thereby ensuring that tissue homeostasis is restored following infection or injury. Indeed, important homeostatic functions have been assigned to the mononuclear phagocytes in almost every tissue of the body (FIG. 1).
Because normal cells of the body must not be mistakenly removed or compromised, macrophages are selective of the material that they phagocytose. During and following phagocytosis, PRRs (including TLRs, C-type lectin receptors (CLRs), scavenger receptors, retinoic acid-inducible gene 1 (RIG1)-like helicase receptors (RLRs) and NOD-like receptors (NLRs)) recognize signals associated with invading pathogens, foreign substances (for example, silica or asbestos) and dead or dying cells1,5. Some PRRs (such as the mannose receptor, DC-specific ICAM3-grabbing non-integrin (DC-SIGN) and macrophage receptor with collagenous structure (MARCO)) function in pathogen binding and phagocytosis, whereas signalling PRRs (which include the TLRs, NLRs and RLRs) sense microbial products and aberrant self on the cell surface or in the cytoplasm of cells and activate transcriptional mechanisms that lead to phagocytosis, cellular activation and the release of cytokines, chemokines and growth factors29–31. Macrophages also express numerous secreted molecules, including complement and Fc receptors that bind opsonin molecules, C3b and antibodies, which activate the complement cascade and enhance the process of phagocytosis by tagging the pathogen surface. Thus, macrophages use various surface receptors and secreted molecules to monitor and respond to changes in their environment.
An unanswered question in macrophage biology is whether resident mononuclear phagocyte populations of a given organ sufficiently respond to tissue stress and infection, or whether there is always a requirement for recruitment of new inflammatory cells. In many infections and tissue stress situations, the resident macrophage populations of organs such as the liver, lungs and gut are insufficient to mediate microbial control and subsequent tissue repair. Instead, monocytes enter the damaged organs and differentiate into a spectrum of mononuclear phagocytes. These newly recruited cells are pro-inflammatory, and therefore damaged tissues exist on an inflammatory tightrope where excessive production of inflammatory mediators must be balanced with the need to protect tissue integrity: this process can be considered as ‘orderly’ inflammation32. It is only recently that molecular links between bone marrow mobilization of effector monocytes and specific inflammatory reactions have been elucidated. Therefore, in this section we focus on a series of specific inflammatory responses that we consider to be informative of the general principles of orderly inflammation.
‘Emergency myelopoiesis’ is the process of generating large pools of monocytes and neutrophils from cells in the bone marrow beyond the normal requirements of a healthy person. Tissue stress, including acute and chronic infection, as well as sterile inflammation, drives the production of monocytes and neutrophils in a process that is dependent on cytokines such as granulocyte colony stimulating factor (G-CSF) and chemokines including CC-chemokine ligand 2 (CCL2) and CCL5 (REF. 33) (FIG. 2). The increased production of monocytes and neutrophils is found in many different types of stress and can therefore be considered a common, conserved pathway. Moreover, the production of circulating MDSCs increases in cancer, but also in Crohn’s disease34, autoimmune disease35, transplantation tolerance36 and smouldering sepsis induced by caecal ligation and puncture (CLP)37. CLP-mediated induction of MDSCs is dependent on myeloid differentiation primary response protein 88 (MYD88)37 and therefore we might expect that the TLR and IL-1 receptor (IL-1R) common pathway, via MYD88, induces haematopoietic cytokines, such as G-CSF and granulocyte/macrophage colony stimulating factor (GM-CSF), that act on bone marrow precursors to increase the output of neutrophils and monocytes38,39. The MDSC pool that exits the bone marrow comprises mature and immature mixtures of monocytic and granulocytic cells, suggesting that either the capacity of the bone marrow to mature the cells is compromised or the bone marrow receives signals to expel the haematopoietic cells at an increased rate. This pathway is an example of long-range communication between the damaged site and the bone marrow to generate increased numbers of tissue macrophages.
A widely accepted view is that monocytes adopt two distinct fates after bone marrow exit1. One type of monocyte — which is defined by high expression of CX3C-chemokine receptor 1 (CX3CR1) and low expression of the myeloid marker lymphocyte antigen 6C (LY6C) (TABLE 1) — has a ‘patrolling’ function in and around the vascular endothelium1. Importantly, patrolling monocytes lack the expression of the chemokine receptor CC-chemokine receptor 2 (CCR2) and cannot respond to CCL2. A recent study has shown that the transcription factor NUR77 (encoded by nuclear receptor subfamily 4, group A, member 1 (Nr4a1)) is required for the development of patrolling monocytes40. By contrast, the LY6Chi monocyte pool is linked to inflammation, expresses CCR2 and can be rapidly mobilized1. The spleen harbours large numbers of LY6Chi monocytes in the subcapsular red pulp that rapidly emigrate to inflammatory sites41.
Multiple types of acute infections cause monocyte mobilization, including infection with influenza, Listeria monocytogenes, Toxoplasma gondii and fungi42–45. Recent results have revealed a surprising complexity to chemokine-induced monocyte recruitment. For example, in acute Citrobacter rodentium infection in the gut (a mouse model of severe Escherichia coli infection), the NLR protein nucleotide-binding oligomerization domain protein 2 (NOD2) in non-haematopoietic cells of the gut lamina propria is responsible for CCL2 production and the subsequent recruitment of large numbers of monocytes that flood the colon and become inflammatory macrophages46. This process is essential for bacterial clearance and for the restoration of tissue homeostasis because NOD2-deficient mice cannot clear the bacteria efficiently and thereby have increased bacterial loads and tissue damage.
CCL2 also drives monocyte recruitment in other settings. For example, when the protozoan parasite Leishmania major infects macrophages, it does not induce a strong inflammatory response and few, if any, chemokines and cytokines are made47. Nevertheless, L. major induces a strong inflammatory response at the infection site; it was shown that complement deposition on parasites induces platelets to accumulate at the infection site and release platelet-derived growth factor (PDGF), which stimulates local CCL2 production and thus creates a chemokine gradient to induce monocyte recruitment48 (FIG. 2). Moreover, monocyte recruitment can be initiated by low circulating amounts of TLR agonists that induce CCL2 production in bone marrow mesenchymal cells and drive inflammatory monocytes into circulation49. This mechanism presumably bypasses the splenic reservoir and is thus an example of the diverse mechanisms the body uses to produce sufficient monocytes and get them into circulation, and ultimately into tissues where they terminally differentiate into macrophages.
The fate of the recruited monocytes and their subsequent differentiation into macrophages is a key issue because inflammatory monocytes have the potential to cause tissue damage or even promote metastasis50. Monocytes quickly differentiate into macrophages and DCs at the site, and it remains unclear how their inflammatory activity is constrained, although IL-10 is likely to have an irreplaceable effect in suppressing activated macrophages at the damage site51.
We cannot assume that the circulating LY6Chi monocyte population is uniform. It is possible that the LY6Chi monocyte population consists of both inflammatory and regulatory populations that counter-balance each other, or the LY6Chi monocytes might convert into regulatory macrophages upon exposure to the non-inflammatory tissue mononuclear phagocytes. In this regard, it was recently shown that pre-emptive CSF1 treatment reduced graft-versus-host disease by expanding suppressive or regulatory macrophages52.
The gut has been fertile ground for research into the fate of recruited monocytes. Several groups have established that a population of gut macrophages is exclusively derived from the circulating monocyte pool, whereas another gut mononuclear phagocyte population, which is characterized by the expression of CD103, is a distinct population of resident gut DCs that have their own functional specializations in terms of promoting immune responses53,54.
During C. rodentium and T. gondii infection, the recruited monocyte population is essential to resolve acute inflammation, but must rapidly convert to an anti-inflammatory phenotype following interaction with the gut-resident macrophages in order to restrain excessive responses to the gut flora. Moreover, in the brain, it was recently shown that recruited pro-inflammatory immature LY6Chi myeloid cells convert in situ to regulatory populations that suppress T cell response55. Be it the gut or any other organ system, it remains unclear if and how monocytes differentiate at the damage site and how the overall number of mononuclear cells in an organ are controlled after homeostasis is re-established. It seems likely the underlying plasticity in myeloid lineages and conversion between pro- and anti-inflammatory activities will be a paradigm uncovered in numerous pathological scenarios.
The textbook picture of macrophage differentiation from recruited monocytes was recently challenged by a study that demonstrated that tissue macrophages undergo massive proliferation in TH2-mediated inflammation24. In this scenario, IL-4 produced by TH2 cells is sufficient to cause local macrophage proliferation during helminth infections, resulting in increased numbers of M2 effector macrophages, which expel worms (FIG. 3). Furthermore, recruited M1 macrophages were induced to proliferate as long as sufficient IL-4 was present24. The signalling mechanism regulated by IL-4 to push macrophages into the cell cycle remains unclear, but may be related to the expression of macrophage-activating factor (MAF; also known as c-MAF) and MAFB transcription factors that suppress macrophage proliferation56.
Self-renewal of tissue macrophages is an appealing concept because it would bypass the requirement for bone marrow-generated monocytes and thus allow local sites to develop an anti-inflammatory milieu that allows for wound repair. Presumably, the expanded population of M2 macrophages would also restrain excessive T cell responses by l-arginine depletion57. However, many questions about tissue macrophage self-renewal remain unanswered. It is unclear whether tissue macrophage self-renewal occurs generally in TH2-dominated inflammation. For example, do alveolar macrophages proliferate in asthma and allergic lung inflammation? Similarly, do tissue-resident macrophages proliferate at the sites of deep tissue TH2 responses, such as at sites of schistosome egg deposition in the liver and Trichinella spiralis worm invasion in muscle? Finally, as the gut harbours the largest population of macrophages in the body, do these cells self-renew to perpetuate the necessary numbers of anti-inflammatory macrophages, or do most originate from bone marrow-derived monocytes?
Following tissue injury or infection, the first-responder macrophages usually exhibit an inflammatory phenotype and secrete pro-inflammatory mediators such as tumour necrosis factor (TNF), nitric oxide (NO) and IL-1, which participate in the activation of various antimicrobial mechanisms, including oxidative processes that contribute to the killing of invading organisms51,58. Other mediators produced by activated macrophages include IL-12 and IL-23, which are decisive in influencing the polarization of TH1 and TH17 cells, which further drive inflammatory responses forward. Activated macrophages produce reactive oxygen and nitrogen intermediates, including NO and super-oxide, that are highly toxic for microorganisms but can also be highly damaging to neighbouring tissues and lead to aberrant inflammation32. Indeed, M1 macrophages are believed to participate in various chronic inflammatory and autoimmune diseases59 (FIG. 4). Therefore, pro-inflammatory and antimicrobial M1 macrophage responses must be controlled to prevent extensive collateral tissue damage to the host.
In addition to their innate phagocytic activity and role in antimicrobial immunity, macrophages are intimately involved in wound repair60,61 (FIG. 4). In contrast to pro-inflammatory and antimicrobial M1 macrophage responses, M2 macrophages exhibit potent anti-inflammatory activity and have important roles in wound healing and fibrosis62,63. They also antagonize M1 macrophage responses, which may be crucial for the activation of the wound healing response and for tissue homeostasis to be restored59. Recent studies have also shown that M1 macrophages can themselves ‘convert’ into anti-inflammatory macrophages with an M2 wound-healing phenotype64,65.
M2 macrophages produce growth factors that stimulate epithelial cells and fibroblasts, including transforming growth factor-β1 (TGFβ1) and PDGF66. Macrophage-derived TGFβ1 contributes to tissue regeneration and wound repair by promoting fibroblast differentiation into myofibroblasts, by enhancing expression of tissue inhibitors of metalloproteinases (TIMPs) that block the degradation of extracellular matrix (ECM) and by directly stimulating the synthesis of interstitial fibrillar collagens in myofibroblasts67,68. Macrophage-derived PDGF also stimulates the proliferation of activated ECM-producing myofibroblasts69.
M2 macrophages can also regulate wound healing independently of their interactions with myofibroblasts. Indeed, they produce matrix metalloproteinases (MMPs) and TIMPs that control ECM turnover70, they engulf and digest dead cells, debris and various ECM components that would promote tissue-damaging M1 macrophage responses66,71, and they secrete specific chemokines that recruit fibroblasts, TH2 cells and regulatory T (TReg) cells72,73. Moreover, M2 macrophages produce factors that induce myofibroblast apoptosis74, serve as antigen-presenting cells (APCs) that propagate antigen-specific TH2 and TReg cell responses (which promote wound healing while limiting the development of fibrosis75,76) and express immunoregulatory proteins (such as IL-10, resistin-like molecule-α (RELMα; also known as RETNLα or FIZZ1), chitinase-like proteins and arginase 1 (ARG1)) that have been shown to decrease the magnitude and duration of inflammatory responses and promote wound healing57,77–81 (FIG. 4).
M2 macrophages have been found to regulate important metabolic functions82. These macrophages are induced by peroxisome proliferator activated receptor-γ (PPARγ) signalling and maintain adipocyte function, insulin sensitivity and glucose tolerance, which can prevent the development of diet-induced obesity and type 2 diabetes83,84. A recent paper suggested that IL-4-producing eosinophils are required to maintain M2 macrophages in healthy non-obese mice176. These studies suggest that as obesity progresses, adipose tissue-associated macrophages switch from an M2-like phenotype to a classically activated M1-like phenotype with potent pro-inflammatory activity82, with the NLRP3 inflammasome serving as the molecular switch by sensing obesity-associated danger signals85 (see REF. 86 for a review).
M2 macrophages were originally described as suppressive cells because they inhibit the production of a wide variety of pro-inflammatory mediators87,88. However, the definition and function of M2 macrophages has been expanded, particularly in regards to their role in regulating TH2-type inflammatory responses, as in addition to downregulating pro-inflammatory responses, M2 macrophages are involved in the development of TH2-dependent immunity to some extracellular parasites and fungi89,90.
Numerous studies have identified roles for M2 macrophages in allergic responses driven by IL-4 and IL-13 (REF. 91). However, their function in allergy and asthma remains controversial, with some studies suggesting that M2 macrophages promote allergic inflammation and others indicating a suppressive role for these cells. A recent study suggested that M2 macrophages are required for the development of airway disease following infection with Sendai virus, which is a mouse parainfluenza virus92. The authors found that M2 macrophages secrete IL-13 and that their depletion significantly attenuated TH2-driven inflammation in the lung. M2 macrophages induced during rhinovirus infection have also been shown to exacerbate eosinophilic airway inflammation by producing the chemokine CCL11 (also known as eotaxin 1), which recruits eosinophils93. The epithelial-derived cytokine IL-33 has also been hypothesized to function as a major driver of eosinophilic airway inflammation because it promotes the differentiation of airway macrophages towards an M2 phenotype94,95.
Nevertheless, other studies have questioned the importance of macrophages in the development of allergic airway disease and instead support a role for another type of mononuclear phagocyte, CD11c+ DCs, in the development of eosinophilic inflammation and TH2-associated cytokine production in the lung96. Additional reports have also identified a suppressive role for M2 macrophages in allergy and asthma. Indeed, by facilitating the uptake and removal of fungal conidia, M2 macrophages have been shown to inhibit asthma symptoms associated with chronic fungal infections90. In contrast with M2 macrophages in mice infected with Sendai virus, M2 macrophages producing IL-13 mediated the resolution of respiratory syncitial virus-induced lung injury by reducing inflammation and epithelial damage97. Chitinase proteins expressed by M2 macrophages have also been proposed to suppress allergic inflammation by degrading or sequestering chitin, a potent and highly abundant allergen in the airway80. RELMα, which is expressed by M2 macrophages, eosinophils and epithelial cells, inhibits TH2-driven inflammation in the lung79,98. However, the specific contribution of M2 macrophages and the proteins they express to airway inflammation remains unclear, as the expression of many of these proteins is not exclusive to TH2 cytokine-stimulated macrophages. These studies emphasize the need to elucidate the functions of molecules expressed specifically by M2 macrophages (TABLE 2).
Distinct macrophage subsets have been linked with either protective or pathogenic roles in cancer99. A protective role in tumorigenesis has been described for M1 macrophages, which activate tumour-killing mechanisms and antagonize the suppressive activities of TAMs, MDSCs, M2 macrophages, regulatory macrophages and immature myeloid cells (which have all been shown to suppress adaptive tumour-specific immune responses and promote tumour growth, invasion, metastasis, stroma remodelling and angiogenesis100–105). M1 macrophages also amplify TH1 responses, providing a positive feedback loop in the antitumour response64.
By contrast, TAMs isolated from solid and metastatic tumours have a suppressive M2-like phenotype. Furthermore, accumulating evidence from many tumour models suggests that macrophages contribute to tumour progression, with increasing numbers of TAMs, MDSCs and immature monocytes correlating with poor outcomes106–108. These observations are also consistent with the tumour-promoting activities of IL-4 and IL-13, which also promote M2 macrophage differentiation109– 112. A novel population of forkhead box P3 (FOXP3)-expressing macrophages was also shown to display immunosuppressive properties and promote tumour growth113.
Importantly, IFNγ was recently shown to reverse the immunosuppressive and pro-tumoural properties of TAMs. So, IFNγ could potentially be administered locally to combat the generation and maintenance of immunosuppressive TAMs and thus boost protective M1 macrophage and T cell responses within the tumour microenvironment114. Moreover, blocking nuclear factor-κB (NF-κB) signalling can switch TAMs to an M1-like phenotype that is cytotoxic against tumour cells115. Natural killer T cells can also kill TAMs directly, providing an additional approach for targeting TAMs and promoting tumour-specific immunity116.
A major problem in the analysis of TAMs concerns how the cells are phenotyped and thus categorized. Diverse phenotypes have been attributed to TAMs, and this stems partly from differences in tumour types, donors and isolation techniques. Therefore, TAM phenotyping should rely on defining gene and protein expression profiles in vivo and ex vivo and on comparison of these profiles with the gene expression profiles of conventional macrophage subsets. Moreover, TAMs should be expected to exhibit the same plasticity as other macrophages following cytokine stimulation ex vivo. Undoubtedly, comprehensive profiling of TAMs from both mouse cancer models and human samples will be a key part of understanding the tumorigenesis process, as cancer researchers have increasingly recognized ‘inflammation’ as being inseparable from cancer itself117.
M1-like macrophage-derived TNF, IL-18, IL-12 and IL-23 have been identified as important mediators in several chronic inflammatory and autoimmune diseases, including Crohn’s disease, rheumatoid arthritis, multiple sclerosis and autoimmune hepatitis118–120. For example, during experimental colitis, a subset of CX3CR1int LY6Chi GR1+ (glutathione reductase 1+) macrophages expressing TLR2, CCR2 and TNF was shown to promote inflammation in the colon121. Similarly, in patients with Crohn’s disease, researchers identified a population of CD14+ macrophages that are distinct from the normal intestinal macrophage pool and produce large amounts of pro-inflammatory cytokines, including IL-23 and TNF122,123. Because IL-23 and TNF mediate pathology in Crohn’s disease, these inflammatory macrophages have been hypothesized to contribute to pathogenesis of the disease. Nevertheless, other studies have shown that impaired pro-inflammatory cytokine production by macrophages can also contribute to Crohn’s disease by diminishing the capacity of macrophages to clear potentially pathogenic commensal bacteria from the lining of the bowel119.
Resident tissue macrophages also maintain homeostasis in the intestine by clearing apoptotic cells and debris, by promoting epithelial repair and by producing IL-10, which has been shown to maintain expression of FOXP3 in colonic TReg cells124,125. In a pathology as complex as Crohn’s disease, it is important to bear in mind that the principles of macrophage heterogeneity and plasticity also apply, and thus multiple macrophage populations are likely to have flexible pro- and anti-inflammatory (or homeostatic) effects in the gut and are subject to both temporal and anatomical effects.
Contrasting roles for different macrophage subsets have also been described in the pathogenesis of rheumatoid arthritis. For example, TNF produced by M1-like macrophages was shown to trigger cytokine production by synovial cells, leading to the development of chronic polyarthritis120. By contrast, macrophages producing reactive oxygen species were found to protect mice from arthritis by inhibiting T cell activation126.
Macrophages have also been identified as key regulators in demyelinating diseases of the central nervous system (CNS). Indeed, infiltrating M1-like macrophages are thought to contribute to axonal loss in multiple sclerosis and in experimental autoimmune encephalomyelitis (EAE), an animal model of multiple sclerosis127. Macrophages recruited to the CNS prime T cells to execute a TH1 effector programme in EAE128, whereas recruited myeloid cells producing IL-23 stimulate the production of GM-CSF by helper T cells, which regulates disease development and severity129,130. These observations suggest that macrophages could be targeted to prevent or reduce axonal loss in multiple sclerosis131. However, macrophages also have protective roles in multiple sclerosis by promoting T cell apoptosis and by expressing anti-inflammatory cytokines such as TGFβ1 and IL-10, which contribute to the termination of inflammation132. Moreover, a subset of macrophages expressing the inhibitory receptor CD200 (also known as OX2) has also been shown to prevent the onset of EAE in mice133. Finally, a population of monocyte-derived macrophages was shown to inhibit inflammation in a model of spinal cord injury, providing further evidence for a protective role for macrophages in the CNS134. Thus, macrophages have both protective and pathogenic roles in a wide variety of autoimmune and inflammatory diseases.
It has been appreciated for quite some time that atherosclerosis is both a lipid disorder and inflammatory disease, with macrophages having a central role135. In atherosclerosis, it is thought that macrophages lodge in the intima and subintima of arteries, eventually leading to the formation of obstructive atherosclerotic plaques that are prone to rupture, leading to thrombosis, myocardial infarction or stroke. Studies have suggested that TH1 cells contribute to the development of atherosclerosis by producing IFNγ136, which stimulates the differentiation of highly activated macrophages, termed foam cells, that promote the formation of unstable lesions137. These pathogenic macrophages also express higher levels of scavenger receptors and CD36, which augments the uptake of modified forms of low-density lipoprotein138–140.
By contrast, TH2-associated cytokines, particularly IL-10, seem to have a protective role, as they block the formation of pathogenic M1-like macrophages in atherosclerotic plaques141. Although hypercholesterolaemia was initially hypothesized to be the primary stimulus for the recruitment of macrophages into the arterial wall, immunological and mechanical injuries, as well as bacterial and viral infections, are likely to contribute to the pathogenesis of atherosclerosis137. Toxic blood lipids, such as oxidized low-density lipoproteins (cholesterol) are removed by macrophages as part of their general homeostatic scavenging function139. Therefore, because macrophages facilitate the clearance of cholesterol, they could be viewed as having a protective role in atherosclerosis and lipid homeostasis.
However, hypercholesterolaemic mice that are deficient in macrophages were found to be highly resistant to developing atherosclerosis, suggesting that macrophages primarily have a pathogenic role in the disease142. Depletion of CD11b+ macrophages after plaque formation is, by contrast, less protective, suggesting that monocytes and macrophages are involved in the genesis but not maintenance of atherosclerosis143. Nevertheless, some reports have suggested that decreases in plaque size and regression of atherosclerosis correlates with macrophages emigrating from the plaque135,144. Thus, devising strategies that facilitate the depletion or inactivation of pathogenic M1-like macrophages from actively growing plaques could emerge as a useful therapy for atherosclerosis145,146.
Studies have suggested that progressive fibrotic diseases, such as idiopathic pulmonary fibrosis (IPF), hepatic fibrosis and systemic sclerosis, are tightly regulated by macrophages61. ‘Pro-fibrotic’ macrophages produce various mediators, including TGFβ1, PDGF and insulin-like growth factor 1, that directly activate fibroblasts, and therefore these cells are intimately involved in wound healing (FIG. 4). These secreted proteins regulate the proliferation, survival and activation status of myofibroblasts, which control ECM deposition147–149. Pro-fibrotic macrophages also produce their own MMPs and TIMPs, which regulate inflammatory cell recruitment and ECM turnover70. In addition, they secrete various pro-fibrotic cytokines and chemokines, including IL-1β, which was identified as a potent pro-fibrotic mediator in the lung150,151. IL-1β stimulates TH17 cells to produce IL-17, which was identified as an important inducer of bleomycin-induced pulmonary fibrosis, a fibrotic disorder with characteristics that are similar to those of IPF152. Furthermore, macrophages function as APCs and promote TH2 responses153, which have been shown to induce and activate the pro-fibrotic cytokine TGFβ1 in macrophages through an IL-13- and MMP9-dependent mechanism62,154.
Nevertheless, although macrophages are clearly required for the initiation and maintenance of fibrosis, other studies have suggested that they are also involved in the suppression, resolution and reversal of fibrosis155. Indeed, macrophages phagocytose dead cells and cellular debris, which can help to reduce the danger signals that contribute to the production of pro-inflammatory and pro-fibrotic mediators. Moreover, they engulf and digest ECM components and stimulate the production of collagen-degrading MMPs in other inflammatory cells, including myofibroblasts and neutrophils70. The production of IL-10, RELMα and ARG1 by M2-like macrophages has been shown to suppress fibrosis57,79,156. Thus, with their potential to both induce and inhibit fibrosis, macrophages and the factors they express are integrated into all stages of the fibrotic process (FIG. 4). To better understand the pathogenesis of fibrosis, we therefore need to identify the specific macrophage subsets that promote, inhibit and reverse fibrosis and elucidate the contributions of the unique mediators that are expressed by each population.
Together, these examples illustrate how inflammatory and suppressive macrophages are crucially involved in the progression and resolution of disease. They also demonstrate the complex and often opposing roles of different macrophage subsets in health and disease. A more detailed understanding of the mechanisms that regulate the activation and deactivation of human macrophages is likely to lead to the development of more effective strategies for treating various important inflammatory diseases157.
An important question in understanding the evolution of immune systems concerns the functions of macrophages after the advent of the lymphocyte-based non-self discrimination system. As we have stressed here, immunosuppression is a common trait of all tissue-resident macrophages, and so it seems plausible that control of T cell proliferation and interaction with TReg cells is a recently acquired function that is necessary for tissue homeostasis. All of these properties of macrophages can be readily dissected in mouse models, which leads us to consider the role of macrophages in humans. Here, differences to rodents are apparent in both the types of pathogens that infect humans and the effector molecules that are deployed by macrophages to control infections. Homotropic pathogens, including Mycobacterium tuberculosis, Mycobacterium leprae, Shigella flexneri, Plasmodium falciparum and numerous viruses such as measles and dengue virus, are predominantly or only found in humans. The long lifespan of humans compared to rodents is likely to be a driver of types of immune responses that are needed to control pathogens; the time lag until sexual maturity means that humans need to survive for decades to ensure their children are self-sufficient. For example, prevention of collateral tissue damage and oncogenic somatic mutations may be a factor in human evolutionary fitness compared to shorter-lived animals that quickly produce the next generation. The extrapolation of rodent models in order to understand homotropic pathogens has, however, not kept pace with the need for relevant systems for dissecting human macrophage-based immunity14.
Although murine M1- and M2-polarized macrophage subsets are relatively easy to distinguish based on combinatorial gene expression profiles (TABLE 2), the identification of equivalent subsets in humans has been more challenging. The basic problem is that panels of markers for in vitro-generated human macrophage subsets do not exist or cannot be agreed upon (TABLE 2). One approach to solve this problem is to ablate transcription factors that establish bias in macrophage phenotypes. For example, interferon regulatory factor 5 (IRF5) seems to be crucial for human M1 macrophage gene expression158. Therefore systematic gene expression profiling in IRF5-deficient human macrophages (or in other macrophage populations in which polarization is genetically fixed or biased) stimulated with different cytokines and TLR agonists might reveal panels of genes that associate with polarized subsets.
Moreover, neither ARG1 nor inducible nitric oxide synthase (iNOS) is expressed by in vitro polarized human macrophages stimulated with IL-4 or IFNγ, respectively, in amounts comparable with those expressed by mouse macrophages. So, the discrepancies in arginine-metabolizing enzyme expression are at the centre of an intense debate on similarities between the human and mouse macrophage subsets and their expected functions14. In addition, other effector pathways have undergone major evolutionary changes compared to rodents. For example, the p47 immunity-related GTPase (IRG) family has 20 members in mice but only two in humans (IRGM and IRGC)159,160. It has been shown that IRGM is involved in the protective anti-mycobacterial autophagy response, and variants of IRGM are strongly associated with Crohn’s disease pathogenesis and anti-bacteria autophagy responses161. It is reasonable to postulate that the pool of effector molecules will be more diverse from species to species as pathogens seek to exploit new niches. This controversial area has been extensively discussed162–164, but remains an area ripe for new discoveries, as evolutionary comparisons can be made between model organism and human macrophages to uncover the underlying effector mechanisms of pathogen control and elimination.
Macrophage research undergoes periods of intense activity and continuously provides informative insights for immunologists. Although much current research has focused on the signalling pathways that regulate inflammatory mediator production and subset development, new issues have arisen that need to be resolved within the contexts of normal homeostasis and acute or chronic disease. We identify three areas of research as paramount for further work.
First, the regulation of macrophages in the tissues remains unclear. For instance, it is only in the past few months that M2 macrophage proliferation in situ has been discovered. We also do not understand how homeostasis is restored after infection, how the response to damaged tissues is resolved and what mechanisms are involved in the layered hierarchy of macrophage activation in situ. Indeed, the number and diversity of signals and the magnitude of the response required to switch macrophages into a pro-inflammatory state remains unclear. How is the fate of recruited monocytes regulated? And what happens to excess macrophages in the tissues following deposition of vast numbers of newly recruited monocytes?
The second area of research that requires development is the underlying mechanisms that regulate the plasticity and stability of macrophage populations. As we have described here, most investigators agree that macrophages are highly plastic, yet the assays used to assign phenotypes require further development and standardization. In our view, new work on the transcription factors and epigenetic changes responsible for macrophage plasticity combined with better marker systems will advance the field. This type of work will help to better define macrophage subsets at a molecular level and provide the foundation that is needed to generate new genetic tools, which will finally allow us to interrogate the function of macrophage subsets in vivo.
Finally, the third area concerns the relationship between human macrophages and their cognate animal-derived model systems. This is perhaps the area of work with the biggest potential, as the chasm between understanding mouse and human macrophages is wide.
Work in P.J.M.’s laboratory is supported by The Hartwell Foundation, US National Institutes of Health (NIH) CORE grant P30 CA21765 and the American Lebanese Syrian Associated Charities. T.A.W. is supported by the Intramural Program of the US National Institute of Allergy and Infectious Diseases, NIH.
Competing interests statement
The authors declare no competing financial interests.
Peter J. Murray’s homepage: http://www.stjude.org/murray.
Thomas A. Wynn’s homepage: http://www.niaid.nih.gov/LabsAndResources/labs/aboutlabs/lpd/immunopathogenesissection/pages/wynn.aspx
ALL LINKS ARE ACTIVE IN THE ONLINE PDF