Macrophages are abundant in almost all tissues of all multicellular organisms, estimated to total 1010 cells in adult humans. Macrophage function is essential for health, as strikingly demonstrated by the severe pathologies of macrophage-deficient animal models. Therefore, understanding the activation and differentiation of macrophages is essential for understanding immunology.
In this Pillars of Immunology article, Charles Mills and colleagues (1) first identified the fundamental M1/M2 polarization axis of macrophages. Mills, who died on May 28, 2017, identified the most important dichotomy in macrophage function: he observed that M1 macrophages kill (infectious organisms, virus-infected cells, or tumor cells) and M2 macrophages heal (sterile wounds and, with less success, cancer).
Mills’ observation that macrophages could promote or inhibit cancer growth dates back even farther, and this outcome was found to correlate with the way macrophages metabolize arginine (2). Arginine is uniquely depleted at sites of inflammation and thus is a limiting metabolite (3–5). Both M1 and M2 macrophages avidly consume this amino acid: M2 macrophages metabolize arginine to ornithine and urea through the arginase pathway (5, 6). Macrophage production of ornithine is required for many repair processes, including cell proliferation and collagen biosynthesis (5). The other side of arginine metabolism is active in M1 macrophages and produces toxic NO (4, 7). M1 macrophages induce inducible NO synthase (iNOS), the enzyme that produces large amounts of NO and citrulline. Although iNOS expression is low in human blood monocyte-derived macrophages, iNOS is expressed at substantial levels in inflamed human tissue macrophages and infiltrating monocyte-derived macrophages (8–10). NO is not only cytotoxic, but it spawns many downstream toxic metabolites that together constitute the M1 killing machinery. Rapid killing is important, because pathogens proliferate out of control within hours or days unless killed by innate responses (11). The adaptive immune system is too slow to be a primary host defense (12), especially on first encounter of a pathogen.
In his seminal paper (1), Mills discovered that Th1-oriented mouse strains such as C57BL/6 preferentially induced M1 macrophages and that Th2-oriented mouse strains such as BALB/c preferentially induced M2 macrophages in response to the same stimulus (Listeria infection). He later called the M1/M2 dichotomy a fork in the road to health and disease (13). That the M1/M2 dichotomy is found in nonvertebrate animals lacking T or B cells (14) and in mice lacking an adaptive immune system (1) underlines the primacy of macrophage polarization over T cell polarization. Immunology had it backward (15). Today, M1 is often used for classically activated macrophages (16) and M2 for alternatively activated macrophages (17). Although there are similarities in phenotype, this use of the M1/M2 nomenclature is incorrect: unlike classically activated macrophages, M1 polarization does not require IFN-γ, and unlike alternatively activated macrophages, M2 polarization does not require IL-4.
Mills was always interested in functions, not in cell subsets for the sake of subsets. He summarized this in two reviews (18, 19). In this Pillars of Immunology commentary, I can only touch on a few key functional aspects. Macrophages must sample their environment and routinely encounter senescent and dead cells and debris. In their default M2 mode, they ingest these materials and help restore lost cells and intercellular matrices through the production of various growth factors (12, 20). However, not all they encounter is benign, and macrophages have hundreds of receptors they use to detect any kind of danger, be it from infectious organisms or dangerously modified endogenous proteins and other biomolecules (21). Unlike T cells, macrophages recognize pathogens directly (21–24). Thus, macrophages support tissue integrity (20, 25) by providing growth factors and healing powers (M2), or by becoming aggressive and killing the invaders (M1).
Killing comes at a price: M1-mediated killing is associated with collateral tissue damage that M2 macrophages must clean up and heal after the dangerous invader has been eliminated. Interestingly, although tissue macrophages are seeded by progenitor cells from the fetal yolk sac or liver (26, 27), can locally self-renew (28), and are also replenished by bone marrow–derived monocytes (29), they all show strong M2 polarization in healthy tissues under homeostatic conditions (30).
Although there are similarities between M2 macrophages and alternatively activated (by IL-4) macrophages (31), they are not the same. No IL-4 is found in healing wounds (32). Similarly, it is well known that LPS is sufficient to induce an M1 phenotype, and T cell–derived IFN-γ, the hallmark of classical macrophage activation (16), is not required. Thus, classically activated macrophages are an amplified (by IFN-γ) version of M1 macrophages, but in the primary immune response, macrophages instruct T cells. Mills gleaned this from the observation that resistance (C57BL/6) or susceptibility (BALB/c) to Leishmania infection was still observed in the absence of adaptive immunity (1). M1 macrophages promote Th1 cells, which in turn produce IFN-γ that strengthens and enhances M1 polarization in a feed-forward loop (33–35). When the cytokine(s) used to drive macrophage phenotypes are known, it is useful to employ a recently proposed nomenclature in which M is followed by the cytokine or other stimulus in parentheses, such as M(IL-4) (36).
Many investigators have found that macrophage phenotypes other than M1/M2 exist. A particularly interesting phenotype is that of regulatory macrophages (Mregs) (37). Although Mregs are controversial (36) and may be a heterogeneous population, the functionality of Mregs was tested in vivo, suggesting that they have a specialized function (38). Future work using higher resolution methods such as mass cytometry and single-cell transcriptomics is necessary to more fully define Mregs. It is worth mentioning that the M1/M2 nomenclature was later adopted by Mantovani et al. (39), who studied macrophages in vitro (40, 41). The transcriptomes observed in macrophages treated with cytokines, immune complexes, and other stimuli in vitro (42) are quite different from M1 and M2 macrophage transcriptomes in vivo (43, 44). For cytokine-induced macrophage polarization in vitro, experimental guidelines and a nomenclature have been proposed: M(cytokine) (36), where (cytokine) delineates the stimulus. Thus, M4 macrophages would be called M(PF4) because they are induced by platelet factor 4 (45).
This nomenclature works well for macrophages induced by defined mediators, but not for macrophages in vivo. In vivo, macrophage phenotypes are not shaped by individual cytokines. Instead, the strongest factor is the infection and danger status that would induce M1, or absence thereof that keeps macrophages in M2 (healing) mode as proposed by Mills et al. (1). A recent analysis of the genome-wide peritoneal macrophage transcriptomes of 83 recombinant inbred mouse strains confirmed that the M1/M2 axis is indeed the main macrophage polarization axis in vivo (30).
In summary, Charles Mills’ discovery of M1 and M2 macrophages proved fundamental to immunology and beyond, because the M1/M2 paradigm is directly applicable to health and disease. As an example, Mills discovered that M1 macrophages can kill cancer cells and M2 macrophages promote cancer (12, 46). Indeed, enrichment for M1 gene expression signatures correlates with favorable outcomes, and enrichment for M2 signatures correlates with poor outcomes in many cancers (30). Recent advances in pharmacology show that it is possible to shift tumor macrophages from M2 to M1 and cure cancer in mice (47). Such interventions may become important enhancers of checkpoint inhibitors (48), or they may become effective cancer therapies in their own right. Although the dust is still settling, Mills’ paper has been the impetus for much subsequent work that has enhanced our still evolving understanding of macrophage activation and its downstream consequences. Mills’ pioneering work has its shortcomings and has required later modifications, many of which are discussed in this commentary. The need for continued refinement is a hallmark of progress in scientific concepts and should not be viewed as detracting from Mills’ achievements.
The author has no financial conflicts of interest.