Obstacles and opportunities for understanding macrophage polarization

Review on new issues concerning macrophage polarization and the impact of polarization and categorization of macrophage subsets on immunology.

Abstract

Macrophages are now routinely categorized into phenotypic subtypes based on gene expression induced in response to cytokine and pathogen-derived stimulation. In the broadest division, macrophages are described as being CAMs (M1 macrophages) or AAMs (M2 macrophages) based on their exposure to TLR and IFN signals or Th2 cytokines, respectively. Despite the prolific use of this simple classification scheme, little is known about the precise functions of effector molecules produced by AAMs, especially how representative the CAM and AAM subtypes are of tissue macrophages in homeostasis, infection, or tissue repair and how plasticity in gene expression regulates macrophage function in vivo. Furthermore, correlations between mouse and human tissue macrophages and their representative subtypes are lacking and are a major barrier to understanding human immunity. Here, we briefly summarize current features of macrophage polarization and discuss the roles of various macrophage subpopulations and macrophage-associated genes in health and disease.

Introduction

Researchers have long known that macrophages residing in or migrating to different tissues or sites of infection and damage have distinct appearances and cell surface phenotypes; for example, Kupffer cells (liver resident macrophages) appear microscopically different than splenic red pulp macrophages. Until recently, phenotyping macrophages and other related mononuclear phagocytes, including the many DC subtypes, with cell surface markers such as CD11b, CD68, macrophage antigen-2, and F4/80, has been the mainstay of macrophage characterization. However, the last decade has provided new ways of phenotyping macrophages based on their gene-expression profile in response to specific stimuli. By far, the most often-used terms in gene expression-based macrophage phenotyping are CAMs (also called M1) and AAMs (M2), which are thought to have characteristic gene-expression profiles defined by markers linked to the stimulation conditions used to generate the subtype—TLR stimulation, bacterial infection, and IFN-γ stimulation for CAMs and IL-4/IL-13 for AAMs. It is not surprising that given tendencies of immunologists for cell categorization, CAMs and AAMs have been atomized into smaller tranches such as M1a and M2a, M2b, and so on [1]. A major question, therefore, concerns the function of the different macrophage types in different homeostatic, infection, and tissue-repair scenarios. Surprisingly, little is known about the functions of individual AAM-associated genes in comparison with CAM-associated macrophage-inflammatory and tissue-remodeling products. However, the gap in knowledge concerning AAM effector functions is closing rapidly with recent publications investigating the effects of deletion of two AAM-associated effector genes, Arg1 and Retnla. The purpose of this short review is not to provide a comprehensive summary of macrophage subtypes or the CAM-AAM paradigm; this work has been reviewed by other researchers extensively [2–8]. We will also not discuss tumor-infiltrating macrophages and their characteristics, as recent reviews have addressed this growing area of research [5, 9–14]. Instead, we will focus on the barriers to macrophage phenotyping and how investigation into the underlying functions and expression patterns of macrophage effector genes offers the best way to understand macrophage biology.

MACROPHAGE GENE EXPRESSION IS PLASTIC

Unlike lymphocytes where phenotypic changes can be largely “fixed” by chromatin modifications after exposure to polarizing cytokines, macrophages have a plastic gene expression phenotype that changes depending on the type, concentration, and longevity of exposure to the stimulating agents, as appraised extensively by others [8, 15–20]. Macrophages exposed to strong AAM-polarizing environments can be readily induced to express CAM-associated genes by subsequent exposure to TLR ligands or IFN-γ [18, 21–23]. Indeed, early functional dissection of macrophage phenotypes noted that gene-expression plasticity was the rule rather than the exception when experimental polarization was performed [18, 21–23]. Plasticity in gene expression makes sense from the perspective of macrophages that have to respond to different microenvironmental stimuli, migrate, phagocytose dead cells and debris, and be instructed by different T cell subsets. As plasticity seems to be a general property of macrophages, the rigid extension of AAM or CAM subset′s assignment to in vivo inflammation and homeostasis seems problematic unless the overall immune response is dominated by one type of T cell response. A further complication concerns the fact that the generation of macrophages from different mouse backgrounds results in overlapping but distinct gene-expression profiles [24, 25]. Although the mouse background could have substantial effects on macrophage-effector functions, it is practically and financially difficult to escape the use of the most common background strains such as C57BL/6 and Balb/c, especially when considering immunologic experiments that require corresponding MHC and TCR matches.

THE TYPE OF MACROPHAGE STUDIED IS A MAJOR FACTOR IN ASSIGNING PHENOTYPE

In the mouse system, BMDMs grown in CSF-1-containing media (usually made from L cells) and inflammatory macrophages isolated from the peritoneal cavity after thioglycollate injection remain the workhorses for macrophage experimentation. In humans, practical and ethical considerations limit work predominantly to macrophages generated from blood monocytes in the presence of CSF-1 or from BM stem cells differentiated into macrophages using cytokine cocktails [26]. It is hardly surprising that we are dependent on these macrophage types to study polarization: they provide large numbers of cells cheaply and with high reproducibility. Indeed, the majority of insights into macrophage gene expression, including our own work, has come from isolating and stimulating BMDMs with different cytokines and pathogens [27–29]. However, there are two major limitations with the current approaches. First, little work has been done to compare gene expression profiles from the common experimental macrophage types with unmanipulated tissue macrophages freshly isolated from organs in homeostasis or disease, as there is no simple way to capture the entire macrophage population free of contaminating cells such as neutrophils. Other commonly used markers are not expressed on some tissue macrophages (e.g., F4/80) or perform poorly in magnetic bead isolation experiments such as anti-CD68 (unpublished results). An alternative would be to use lineage-marked macrophages (i.e., with a marker such as GFP). Indeed, Hume and colleagues [30] generated a strain of mice where GFP is driven by the Csf1r promoter (MacGreen mice), and most if not all macrophages are labeled in this system. However, the usefulness of these mice has limitations for genetic-based investigation of specific macrophage effector functions, as the Csf1r-gfp allele would have to be crossed into the relevant knockout or transgenic backgrounds. In short, we do not have a sound understanding of tissue macrophage gene expression as it relates to the commonly used macrophages, especially in humans. Second, studying specific tissue macrophage subtypes, such as Kupffer cells in the liver, osteoclasts in bone, or microglia in the brain, is a far more daunting proposition for the microarray approaches because of the difficulty in obtaining sufficient, highly purified cells for analysis. The latter can be overcome by new generation microarray platforms that require ∼50 ng total unamplified RNA; however, the ability to reproducibly isolate and fully characterize cells from diseased tissues remains a challenging obstacle and may explain why findings with mouse macrophage subsets have been difficult to translate to humans.

HOW DIFFERENT ARE MOUSE AND HUMAN MACROPHAGES? AN EXAMPLE OF AN UNRESOLVED MOUSE—HUMAN DIFFERENCE WITH IMPLICATIONS FOR ANTIMICROBIAL IMMUNITY

A controversial aspect of macrophage biology concerns perceived differences between rodent and human macrophages. Some researchers have even suggested that human macrophages are “fundamentally” distinct from rodent macrophages [31, 32]. At the center of this controversy is the use of arginine for NO production. NO is a central component of the macrophage arsenal against intracellular pathogens, in addition to its myriad roles in signaling, apoptosis, vascular tension, and free radical-induced damage [33]. In macrophages, NO is made by iNOS, an oxidoreductase that catalyzes the conversion of arginine and oxygen to NO and citrulline. The expression of iNOS and NO production is most closely linked with CAMs (see below). The NOS2 gene (encoding iNOS) is present in rodents and humans. In mouse models of intracellular infection, NO production by iNOS has been established—by detailed genetic and pharmacological approaches—as essential for effective immunity to Mycobacterium tuberculosis, Leishmania sp, Toxoplasma gondii, and Trypansoma cruzi, amongst others [33]. However, despite the acceptance of the central antimicrobial effects of NO, the role of NO in human macrophages is remarkably controversial and elicits vigorous exchanges in the literature [31, 32, 34, 35]. What is the source of the controversy? Since the discovery of iNOS and NO production in macrophages, many of the “rules” governing the induction of iNOS expression and NO production have been established in the mouse. For example, the most potent inducer of iNOS expression is a combination of a TLR agonist and IFN-γ, hallmarks of the CAM phenotype: we even know most of the signaling events downstream of the receptors, as well as the sites in Nos2 that respond to TLR + IFN-γ signaling. However, similar experiments performed using human MDMs yield no iNOS and no NO [31, 32]. Similarly, human monocyte-macrophage cell lines cannot be provoked readily to make iNOS or NO. To confuse matters further and in direct opposition to the in vitro data, human tissues, such as TB biopsy samples, express abundant iNOS by IHC, and dozens of publications have reported robust iNOS expression in human tissue samples [33, 34, 36]. These discrepancies have profound importance for human health, as if iNOS and NO are not expressed in human tissue macrophages in vitro, and the IHC data are unfortunate artifacts, then what are the host requirements for microbial control in human macrophages? Should we even be studying mouse models of infection where NO plays an obligate role? Experiments in human macrophages to measure iNOS and NO are performed almost universally in cells derived from donor monocytes and are “differentiated” in vitro in media such as RPMI with CSF-1. There is no guarantee that such a system is permissive to iNOS expression. Maybe there is something missing from the RPMI, obligate epigenetic changes are required, or even factors that inhibit NOS2 expression are coexpressed; many possibilities exist. By contrast, BMDMs or inflammatory peritoneal macrophages from the mouse readily express iNOS and make huge amounts of NO. The simplest hypothesis is that the culture conditions and/or the cell type are lacking in one or more factors that allow iNOS expression in human macrophages in vitro, as articulated before by Nathan and Bogdan, and others [33, 35]. The case against iNOS expression to date in human macrophages is based on the interpretation of negative data and is thus uninterpretable.

A similar scenario exists for the arginases. There are two arginase enzymes: cytosolic Arg1 and the mitochondrial Arg2. The expression patterns of the two enzymes in myeloid lineages are very different and have been studied in a variety of physiological contexts. However, in humans, Arg1 appears to be expressed only in neutrophils [37–39], and mouse Arg1 is expressed in macrophages and neutrophils. Mouse Arg2 is also expressed in macrophages and neutrophils, but in tissue-infiltrating myeloid cells, Arg2 is expressed exclusively in neutrophils (unpublished results). As is the case for iNOS, species and cell-type distinctions in the expression of the arginases have given rise to debate [40]. Accordingly, these diverse and species-specific expression patterns need to be assessed continuously within the frame of a given experimental system.

The apparent species-specificity issues with iNOS, arginases, and NO production encapsulate the human versus mouse issue and also speak to a central problem in macrophage biology: if we cannot understand the differences and similarities between species, then preclinical development of therapies based on mouse models is difficult. It seems likely that the human immune response retains a core set of pathways that are identical to rodents and probably all mammals. By contrast, humans have also evolved pathways that are substantially distinct from rodents and were driven by the evolutionary pressures of new pathogens (diseases such as leprosy, dengue fever, shigellosis, and TB, for example, have no natural counterpart in the mouse), new environments, and the introduction or loss of genes within the human gene pool; a remarkable illustration of gene expansion (or contraction) is the IFN-regulated IRG proteins, where there are two members in humans but ∼18 in the mouse [41]. The first step to unraveling the human-mouse macrophage controversy is first, to perform comparative gene expression studies on various tissue macrophage populations (including cytokine-stimulated and infected populations) from different species and organ systems and establish where the similarities and differences lie.

CAN CAMs BE EASILY DEFINED?

CAMs are generally considered to arise in microenvionments rich in TLR ligands and IFNs. What gene products are expressed under these conditions, and how does the profile of CAM gene expression change over time? Here, the definition of a CAM is set by the way we do experiments involving in vitro stimulation. For example, when experimental macrophages are stimulated with LPS and gene expression measured over time, expression of the early primary and late secondary TLR genes increases, subject to the production of autocrine-paracrine factors (e.g., type I IFNs, IL-6, IL-10) and secondary transcription factors (and myriad other cell intrinsic factors, such as mRNA stabilization and destabilization), in accordance with predictable patterns that have been elucidated recently [42–44]. Gene expression patterns in LPS and IFN-stimulated macrophages can be grouped based on multiple criteria. For example, early response genes include Tnfaip3, Tnf, and Cxcl1, and secondary response genes include secreted factors such as Il10, Il23a, and Il6. Another group of TLR-regulated genes are “housekeeping” in that their expression is constitutively high in resting BMDMs but can be induced further by exogenous TLR activation (e.g., Nfkbia α encoding IκBα). It seems fair to conclude that many of these patterns of gene expression are linked to CAMs and not AAMs, although specific exceptions occur, such as IL-10, which is produced to various degrees by all macrophages as an autocrine-paracrine anti-inflammatory mediator [45]. Many of the TLR-induced gene classes are associated with specific functions that seem distinct from AAMs; for example, a variety of chemokines, including IL-8/CXCL8, MIP-1α/CCL3, MIP-1β/CCL4, IFN-inducible protein 10/CXCL10, and RANTES/CCL5, are released and attract neutrophils, DCs, NK cells, and activated T cells to sites of inflammation. Cytokines including IL-1β, IL-6, and TNF-α regulated myriad factors in the inflammatory microenvironment. CAMs also release proteolytic enzymes including MMP-1, -2, -7, -9, and -12, which degrade collagen, elastin, fibronectin, and other matrix components. Although production of many of these factors is important for host defense, in some circumstances, when CAMs persist in tissues and remain activated, they often induce significant collateral tissue damage as a result of production of proinflammatory, proapoptotic, and tissue-remodeling MMPs in the local environment. Indeed, tissue destruction is a hallmark of the proinflammatory CAM. A good example is the TB granuloma that can persist for years and contains infected and uninfected macrophages that are activated by mycobacteria and autocrine-paracrine cytokines in the local milieu. In this case, we have little idea of how a TB granuloma CAM is related to the in vitro classifications described above.

CANONICAL GENE EXPRESSION LINKED TO AAMs

The characteristic genes associated with AAMs have been assembled previously from gene expression studies in IL-4-treated macrophages, murine asthma models, or macrophages isolated from mice infected with nematodes [2, 3, 8, 46–52]. The majority of these genes shares the property of being regulated directly or indirectly by Stat6 [2, 3]. However, little is known about the function of the products of these genes other than Arg1 and Retnla (discussed below). Three points are worth noting about the AAM signature. First, it is uncertain how tightly associated any of these genes are with AAMs. It is entirely possible that AAM signature genes are expressed in multiple biological contexts and not just a Th2 cytokine-rich environment. Second, it is clear that many of the characteristic genes are not expressed exclusively by macrophages. Third, it is unclear why the AAM signature is stereotypically induced by the Th2 cytokines. One possibility is that AAM gene expression arises as a component of the antihelminth response as a unit. Rather than expending resources to tailor a specific AAM profile to a specific worm, it might be more advantageous and economical to trigger the full response to cover all possible avenues to effective immunity. The canonical AAM response seems to arise in many if not all Th2-dominated immune responses, but the overall role of the response is far from certain. For example, deletion of the IL-4R α chain in macrophages, which will disable the entire Stat6-dependent response in AAMs, has no discernable effect on Nippostrongylus infection, which requires a highly polarized Th2 response for worm expulsion [53]. In contrast, an absence of AAMs has been reported to lead to the rapid demise of mice infected with Schistosoma mansoni, as a result of the development of a stronger proinflammatory response in the intestine [53].

Arg1 IS A USEFUL MARKER OF AAMs ONLY WHEN AAMs ARE THE DOMINANT MACROPHAGE POPULATION

Arg1 is perhaps the most well-known marker of AAMs, and its expression is often used to provide evidence for AAM polarization. Indeed, we know a great deal about the mechanistic basis of Arg1 expression following Stat6 activation by the IL-4 and IL-13 pathways [23, 28, 54–56]. Arg1 expression in BMDMs is almost undetectable but can increase five to six orders of magnitude after IL-4 stimulation [23, 57, 58]. IL-10 synergistically enhances IL-4-induced Arg1 expression, as IL-10, via Stat3, increases the expression of Il4ra, encoding the IL-4R α chain [28]. Most importantly, Arg1 expression in AAM-rich lesions of schistosome egg-induced granulomas is robust, arguing that tissue AAM macrophages accurately reflect in vitro stimulation conditions [53, 59]. However, the notion that Arg1 is an absolute marker of AAMs is untrue. The first clues that Arg1 might be expressed in CAMs, albeit at much lower levels, came from experiments using LPS stimulation of BMDMs or peritoneal-dervied macrophages; Arg1 was expressed but not to the amounts observed with IL-4 [60, 61]. However, when BMDMs are infected with mycobacteria, Arg1 expression increases to the equivalent of an AAM [55]. As bacillus Calmette-Guerin-infected macrophages are certainly not AAMs and as neither Stat6 nor the IL-4Rα was required for Arg1 expression in the infected cells [55], what is the mechanism of Arg1 expression in mycobacterial infection? Recently, we have discovered a pathway that requires Stat3 to induce Arg1 expression in mycobacterial infection [62]. First, Arg1 expression in mycobacteria-infected macrophages has an absolute requirement for MyD88 but not the IL-1R, suggesting that the TLR pathway was responsible. However, MyD88-deficient macrophages can express high amounts of Arg1 mRNA and protein when exposed to sterile supernatants from control-infected macrophages; thus, a soluble factor(s) was involved. Subsequently, we found that mycobacteria-mediated activation of the TLR-MyD88 pathway induces the expression of IL-6, G-CSF, and IL-10 (as expected) and that these three cytokines control, via a Stat3-mediated autocrine loop, the expression of Arg1 [62]. These data have important implications for understanding mycobacterial infection because of the link between Arg1 and NO suppression (discussed below). A second implication is that Arg1 as an AAM marker is only useful in microenvironments dominated by IL-4 and IL-13. In other situations, such as tumor-associated macrophages, the effects of the Stat3 pathway need to be accounted for, as the presence of combinations of IL-6, G-CSF, and IL-10 will complicate the assignment of whether a macrophage is thought to be an AAM based on robust Arg1 expression.

DELETION OF Arg1 CAUSES TWO VERY DIFFERENT OUTCOMES DEPENDING ON THE UNDERLYING DISEASE

Noting that Arg1 expression is observed in AAMs and mycobacteria-infected macrophages, the next challenge was to test the in vivo function of Arg1 in these different situations. When mice lacking macrophage Arg1 (constructed using the Tie2-Cre deleter strain) were infected with M. tuberculosis, increased local tissue NO and a reduction in bacterial numbers relative to controls were observed [55, 62]. These data and corroborating studies with T. gondii infection [55] argue that Arg1 controls NO production by suppressing the total amount of arginine available to iNOS. These results link back to the earliest demonstrations of macrophage polarization, where pretreatment of macrophages with IL-4 to induce Arg1 and subsequent stimulation with LPS and IFN-γ to induce iNOS cause suppression of NO production by arginine limitation [21, 23]. The effects of Arg1 in the in vitro setting can be overcome by adding excess arginine back to the cultures [21, 23]. In vivo, the effect of Arg1 on NO production seems to be rheostat-like in nature: coexpression of Arg1 and iNOS in mycobacteria-infected cells alters the arginine balance such that NO production cannot be “maximal”. Although it seems counterintuitive that the host would not try to maximize NO production to kill pathogens, limitation of NO may be beneficial to the host, as collateral-cellular and tissue damage from free radical damage would be reduced. Of course, intracellular pathogens can also take advantage of the host Arg1-mediated NO suppression pathway, as mycobacteria induce IL-6, G-CSF, and IL-10 production and could feasibly hijack local Arg1 expression, and T. gondii directly phosphorylates Stat6 via rhoptry kinases, which then drives Arg1 expression, bypassing the requirement for host cytokines to induce Arg1 [63]. It is important to note that other examples of pathogen-mediated manipulation of host arginine metabolism have been discovered, suggesting that host-pathogen evolution has selected a variety of ways to take advantage of reductions in NO or the products of arginine metabolism [33].

The role of Arg1 in AAMs is, however, a remarkable contrast from the examples of intracellular pathogenesis. Our initial hypothesis was that Arg1 would promote AAM-mediated immunopathogenesis. As such, when macrophage Arg1 was ablated, we should have observed a reduction in excessive Th2-mediated tissue damage. To dissect macrophage Arg1 function in a model of Th2-mediated inflammation and fibrosis, we began with the S. mansoni egg-induced liver inflammation model, where Arg1-positive macrophages surround worm eggs in the liver and are thought to drive a profibrotic response via the production of proline, one of the downstream products of arginine catabolism. In contrast to our expectations, we found that the inflammatory and profibrotic responses and hepatomegaly were all greatly increased in the absence of macrophage Arg1, leading to early death of the infected mice [64]. The underlying mechanism involves AAM Arg1-mediated suppression of T cell proliferation. In the absence of macrophage Arg1, increased Th2 cell proliferation and numbers were observed in the livers of infected mice, driving the inflammatory response forward and preventing the normal tissue repair and resolution response [64]. Recent data from Herbert and colleagues [59], using a bone reconstitution approach, support our data from the Arg1 conditional knockouts. The excessive inflammatory phenotypes in S. mansoni-infected macrophage Arg1 knockouts were independent of any contributions of NO [64] and suggest that instead of the arginine competition mechanism observed in macrophages coexpressing Arg1 and iNOS, AAM Arg1 locally depletes arginine from the microenvironment, and this is the trigger for T cells to stop proliferating. Whether this mechanism occurs exclusively in defined tissue microenvironments, such as granulomas and tumors, awaits further study.

DELETION OF RETNLA, AN AAM-ASSOCIATED MARKER, IS ALSO REQUIRED FOR Th2 IMMUNOSUPPRESSION

Retnla is a member of a family of cysteine-rich, secreted proteins referred to as resistin-like molecules or found in inflammatory zone, originally identified in the lung. The resistin-like family consists of four members, although only Retnla and Retnlb are induced by Th2 cytokines, and Retnla, predominantly but not exclusively, associates with the AAM gene expression signature. Retnla increases during allergic responses in the lung, as well as in the lung, liver, and gut during most helminth infections, largely as a result of IL-4-, IL-13-, and Stat6-mediated signaling [46, 49]. Studies with the bleomycin model of pulmonary fibrosis revealed that Retnla is highly induced in the lung during fibrogenesis [65]. Retnla-expressing cells were found to activate α-smooth muscle actin and type I collagen-expressing fibroblasts via a notch-1-dependent but TGF-β1-independent mechanism in vivo [65]. Retnla also exerts a proliferative effect on mouse lung fibroblasts [66]. Together, these findings suggested that macrophage-derived Retnla might be involved in fibrogenesis by promoting the differentiation and survival of myofibroblasts.

In addition to the injured lung, Retnla is found in abundance following infection with a variety of metazoan parasites. In schistosomiasis, liver fibrosis and portal hypertension are the primary causes of chronic morbidity and mortality, and Th2 cytokines are essential to the development of fibrosis. Retnla is markedly induced in the granulomatous tissues of S. mansoni-infected mice [51, 52, 67]. Experiments with Retnla-deficient mice revealed that Retnla is not required for the development of helminth-induced CD4⁺ Th2 responses in the lung, liver, or gut. By contrast, deleting Retnla facilitated the development of a much stronger Th2 response to S. mansoni eggs, leading to exacerbated liver fibrosis mediated by IL-13 [67, 68]. Retnla similarly reduced inflammation in the lungs of egg-challenged mice [67, 68]. Therefore, instead of inducing fibrosis, as might have been expected based on prior in vitro studies with Retnla protein, in vivo studies with Retnla-deficient mice suggested that Retnla, like Arg1, has protective, anti-inflammatory activity in schistosomiasis-induced fibrosis by functioning as a negative regulator of the Th2 response. Whether the effects of Retnla are linked exclusively to AAMs, however, awaits further study, as Retnla was also produced at quite high levels by eosinophils and epithelial cells. Development of macrophage-specific knockouts for Retnla, as well as for the other AAM-associated genes will help better clarify their individual roles in host immunity.

ABSENCE OF MACROPHAGE EFFECTOR-SPECIFIC CRE DELETER MICE IS A BARRIER TO UNDERSTANDING MACROPHAGE FUNCTION

Cre deleter mice and loxP modification of genes are one of the greatest advances in basic research, and yet in the macrophage field, we have a paucity of useful mice to dissect macrophage subtype and effector function and leave macrophage populations intact (transient depletion of entire macrophage populations can also be accomplished with Fas-mediated or diphtheria toxin-mediated cell killing [69, 70], through use of mice with complete or partial ablation of the Csf1r gene [71], or by antibody-mediated depletion targeting CSF1R-expressing cells [72]). The most often-used Cre deleter mouse for macrophages is the LysM-Cre stain developed by Imrgard Förster and colleagues [73] over a decade ago. Although the LysM-Cre strain is widely used, there are two major problems with these mice. First, the efficiency of Cre deletion is locus-dependent. For example, the floxed allele of Arg1 and Stat3 requires complete deletion of the other allele to get “good” (i.e., >90%) deletion [55, 74]. A second problem is that lysM (encoded by Lyz2) expression is not observed in all mononuclear phagocytes, seems highly variable from mouse to mouse, and has mosaic-type expression in early hematopoietic precursors [75–77]. The latter complicates the interpretation of experiments where the target allele needs to be deleted uniformly in the myeloid lineages. Another useful strain is the Tie2-Cre deleter (Tie2 is encoded by Tek), of which, there are several variations. In our hands, Yanagisawa and co-workers′ Tie2-Cre deleter [78] is the most efficient deleter in macrophages, providing >99% deletion of the floxed Arg1 allele [55]. However, Tie2-Cre is expressed in all hematopoietic lineages and in the case of the Yanagisawa strain, endothelia. In the example of tissue-specific Arg1 deletion, all myeloid lineages and hepatocytes express Arg1, and so, the effects of Cre in other hematopoietic lineages were irrelevant. Use of such an easy “trick” to knock out Arg1 in macrophages will likely not be so straightforward for other macrophage effector genes whose expression is found in other lineages.

The absence of Cre deleters, which are highly specific for mature tissue macrophages or groups of macrophages, in addition to Cre deleters, which are active in activated macrophages and can definitively mark AAMs, CAMs, or other activated subtypes, is a major barrier to research advances in macrophage biology and immunity in general but by the same token, difficult and potentially expensive to develop and test. For example, an Arg1-Cre strain would theoretically delete in the liver, AAMs, and CAMs; should the gene of interest be essential for liver function, then the deleter mouse would be of limited usefulness. A more useful strain would be a Cre deleter, which was specific for only AAMs or CAMs. However, the development of such strains would require the identification of genes specific only for these cell types, a problem that brings us back to the original issue raised in this review—that we do not know enough about the gene-expression patterns of activated macrophages and how they can be classified. Possibilities include chitinase 3-like 3 for AAMs and Irgm1 (LRG-47) for CAMs, although far more work is needed in this area to define the specificity of gene expression for particular macrophages.

CONCLUSION

Although the use of the CAM and AAM definitions, terminology, and practical experimental approaches has become widespread, much remains to be learned about macrophage polarization and its effects on immune responses. The primary issues concern plasticity of macrophage phenotypes in vivo, correlations between in vitro polarization experiments and in vivo infections, the degree to which tissue macrophages are representative of in vitro primary macrophage models, how closely related human and mouse tissue macrophages are, robust markers for macrophage polarization, and finally, translation of the basic observations of macrophage polarization to human disease and how polarization can be manipulated to alter the outcome of acute and chronic diseases. Because of their inherent plasticity, manipulating macrophages rather than T cells might offer a more attractive approach to treat a variety of important human diseases.