Functional plasticity of macrophages: in situ reprogramming of tumor-associated macrophages

Abstract

The extent to which the functional heterogeneity of Mϕs is dependent on the differentiation of functional sublineages remains unresolved. One alternative hypothesis proposes that Mϕs are functionally plastic cells, which are capable of altering their functional activities progressively in response to progressively changing signaling molecules generated in their microenvironment. This “functional plasticity” hypothesis predicts that the functionally polarized Mϕs in chronic pathologies do not represent Mϕ sublineages but rather, are mutable phenotypes sustained by chronic signaling from the pathological environment. Solid TAMϕs are chronically polarized to provide activities that support tumor growth and metastasis and suppress adaptive immune responses. In support of the functional plasticity hypothesis, administration of slow-release microsphere-encapsulated IL-12 successfully reprogrammed TAMϕs in situ, reducing Mϕ support of tumor growth and metastasis and enhancing Mϕ proimmunogenic activities. Increased knowledge of how Mϕ function is regulated and how polarized Mϕs can be reprogrammed in situ will increase our ability to control Mϕ function in a variety of pathological states, including cancer and chronic inflammatory disease.

Introduction

Cells of the monocyte/Mϕ lineage are notoriously diverse in their functional phenotype (i.e., the selective array of functional activities that they are expressing) [1, 2]. A substantial amount of research has been invested in attempts to resolve the heterogeneity by categorizing Mϕs into subsets defined in parallel with the Th1 and Th2 T cell subsets [3, 4], in parallel with inflammatory versus anti-inflammatory activity [4, 5], or in parallel with the phases of a classical acute inflammatory response, namely, the initial tissue-destructive phase versus the wound-resolution phase [4]. The functional heterogeneity of Mϕs appears to extend well beyond a simple bimodel model such as type 1 and type 2 responses [1, 2, 4]. Inflammatory responses are functionally diverse (e.g., necrotizing, fibrotic, granulomatous, and other), and the signaling cascades involved in different types of inflammation are, to a degree, counter-regulating, thus complicating the concepts of “inflammatory” and “anti-inflammatory.” Broadly dividing Mϕs on the basis of tissue-destructive versus wound-resolving activities appears to be the least conflicted of approaches (Fig. 1) but does not include the many nonpathological activities of tissue Mϕs (e.g., Kupffer cells, microglia, osteoclasts, or alveolar Mϕs). This focus on bimodel models for defining Mϕ subsets does not address the question of how the many unique arrays of functional activities displayed by Mϕs are regulated. Are there multiple sublineages and subsets of Mϕs that display these different functional patterns? Is there interchange among subsets, and/or do Mϕs display a progression of different functional phenotypes in response to a complex array of signals generated by the tissue microenvironment? Resolution of the mechanisms by which Mϕs attain and sustain their polarized functional phenotype(s) in chronic pathologies is critical to the development of means of targeting Mϕ function for therapeutic intervention in those pathologies.

Figure 1.

Functional categories of Mϕ function based on destructive or constructive activities. These two broad categories of Mϕs function less conflicted than the categories of inflammatory versus anti-inflammatory activities or “Th1 versus Th2” categories. However, they do not include many of the physiological activities displayed by tissue Mϕs.

Mϕ FUNCTIONAL HETEROGENEITY

Monocytes taking residence in different tissues rapidly display a functional phenotype characteristic of Mϕs in that tissue (e.g., prefusion osteoclasts, Kupffer cells, microglia, and others) [6, 7]. This ability of blood monocytes to “differentiate” into different functional phenotypes within tissues led to the concept of differentiative plasticity (i.e., that Mϕs displaying different functional phenotypes represent distinct myeloid sublineages). However, the resident Mϕs within a given tissue display substantial heterogeneity, which is associated with their location within the tissue [6, 7]. For example, the functional profile of liver Mϕs is dependent on the proximity to the portal vein. The periportal Kupffer cells display a functional pattern distinct from that of Kupffer cells distal to the portal vessel [6]. These studies suggested that the functional phenotype of tissue Mϕs is controlled by microenvironmental signals within the tissue [1, 8]. The role of microenvironment in regulation of tissue Mϕ function has been demonstrated for mucosal Mϕs, whose functional responses are restrained by the epithelial cells that are in direct contact with the Mϕs [8]. The demonstration of these homeostatic circuits regulating tissue Mϕs, in conjunction with previous demonstrations that the functional responses of tissue Mϕs to stimuli change when the Mϕs are cultured in vitro [1, 9, 10], has led to the opinion that tissue Mϕ phenotype is not an immutable end-stage phenotype but rather, is maintained by a reversible homeostatic mechanism exerted through organ-specific microenvironments.

Upon injury, tissues release a variety of proinflammatory signals whose purpose is to increase vascular permeability and alter the chemokine/adhesion molecule expression by proximal vascular endothelial cells to facilitate diapedesis of leukocytes. Several of these cytokines, notably M-CSF, TNF-α, IL-1β, and IL-6, contribute to the mobilization of myeloid cells from the bone marrow. Monocytes immigrating into the inflammatory site develop an inflammatory phenotype dependent on the signals being generated in the tissue [11, 12]. In an acute inflammatory response, this usually entails production of inflammatory cytokines, antimicrobial oxidative radicals, and tissue-debriding proteinases and expression of elevated phagocytic activity. When the wound is cleared of inflammatory debris, Mϕs contribute to the process of wound resolution, promoting angiogenesis, matrix regeneration, and cell proliferation. This change in functional activity appears to be initiated predominantly by the phagocytosis of apoptotic cells and to be regulated by a variety of tissue-derived cytokines, hormones, and metabolites [11,12,13]. A point that should be noted is that it has not been established formally whether the monocytes entering an inflammatory site progressively change their functional phenotype from inflammatory/debriding activities to tissue-regenerative activities in response to progressive changes in microenvironmental signaling within the tissue [1, 14].

FUNCTIONAL PLASTICITY OF Mϕs: THE BASIS OF Mϕ HETEROGENEITY

The role of cytokines in orchestrating the various synergizing and antagonizing functions of Mϕs first became apparent with the observation that IFN-γ enhanced the cytotoxic activity of activated Mϕs [15]. Nearly 10 years later, Simeon Gordon’s group [16] reported that IL-4 enhanced expression of the mannose receptor, a function distinct from any effect of IFN-γ. Subsequently, they demonstrated that IL-4 induced a set of genes in Mϕs that was unique and distinct from many of the genes known to be induced by IFN-γ [16]. As studies progressed about these two subsets, it was suggested that classically activated Mϕs were proinflammatory and cytotoxic, and alternatively activated Mϕs were anti-inflammatory [3]. Mills et al. [17] examined Mϕ function in four strains of mice, two of which preferentially generated Th1 responses to selected parasites, and two of which preferentially generated Th2 responses. The Mϕs in the “type 1” strains appeared to produce more iNOS and less TGF-β than the Mϕs in the “type 2” strains. Mills et al. [17] suggested the Mϕs be referred to as “M1” and “M2” to correspond with the Th1/Th2 polarity. Many immunologists embraced this nomenclature, which fostered the perception that M1 and M2 Mϕs were separate lineages like “Th1” and “Th2”. Through use, the term M1 was applied to IFN-γ-generated inflammatory and cytotoxic Mϕs, and the term M2 was applied to any Th2 cytokine or glucocorticoid-driven Mϕ, which was by virtue of nomenclature considered to be anti-inflammatory and protissue resolution [3, 4].

Several concerns have been raised about the M1/M2 nomenclature. Treatment of Mϕs with glucocorticoids, IL-10, or TGF-β does not induce the same phenotype as treatment with IL-4 or IL-13 [18,19,20,21,22]. To counter this concern, proponents of the M1/M2 nomenclature have proposed “subsets” be designated (e.g., M2a, M2b, etc.) to indicate the distinct but overlapping phenotypes induced by different type 2 cytokines [4]. However, physiological responses involve multiple cytokines and frequently, multiple-activating molecules. The functional phenotype of the Mϕs is the result of the cross-talk between those multiple signal cascades [8, 23,24,25,26]. The extensive cross-talk between receptors and signaling cascades is the focus of a recent issue of Nature Immunology [27,28,29], which emphasizes the “network” view of system biologists rather than a single pathway view. The network is complex, including transcription-dependent and transcription-independent regulation of receptor expression, synergistic and antagonistic interactions between receptor adaptor proteins, and synergistic and antagonistic interactions between components of different signaling cascades. Given that Mϕs are capable of responding to the complex mixtures of cytokines, chemokines, hormones, and/or pathogen-associated molecular patterns, generated during pathological assaults on a tissue, it is unlikely that just one signaling pathway would be operative in a responding Mϕ. A classic example of progressive changes in Mϕ functional phenotype caused by signal cross-talk is provided by TLR4 ligation, which activates NF-κB via TNFR-associated factor 6, resulting in production of inflammatory cytokines such as TNF-α and IL-6. TLR4 ligation also activates Akt via PI3K, enhancing CREB phosphorylation, which in turn, initiates IL-10 production. IL-10R ligation activates AMPK, which enhances IL-10 production by inactivating the CREB inhibitor, glycogen synthase kinase 3 [30, 31]. IL-10R ligation also inhibits NF-κB activation via activation of AMPK [30] and activates STAT-3, which induces suppressor of cytokine synthesis-3, an inhibitor of gp130 and/or JAK-STAT signaling pathways [27, 32]. Thus, the TNF-α-producing Mϕ (inflammatory), appearing early in the response to TLR4 ligation, is converted to an IL-10-producing Mϕ (anti-inflammatory) by a maze of interactions between signaling pathways [30]. Interaction with other cytokines, chemokines, or receptor complexes may modify the kinetics, intensity, and/or quality (types of effector molecules expressed) of the response by initiating synergizing, complemetary, or antagonizing signaling pathways [27, 28]. For example, IFN-γ prolongs the inflammatory phase of the Mϕ response to TLR4 ligation by inhibiting Akt /CREB activation, thus delaying the induction of IL-10 production and the resultant feedback inhibition of NF-κB activity [27, 33]. As a result of signal cross-talk, IL-4 can promote an “inflammatory M1-like” phenotype (up-regulate IL-12, TNF-α, and iNOS) if present during LPS stimulation of Mϕs but also can promote an “anti-inflammatory M2-like” phenotype if Mϕs are treated with IL-4 overnight prior to activation with LPS [34, 35]. TNF-α and IFN-γ, type 1 inflammatory cytokines, can enhance anti-inflammatory activity in Mϕs primed with TGF-β, a type 2 cytokine [36, 37], and treatment of Mϕs simultaneously with TNF-α and TGF-β induces tartrate-resistant acid phosphatase, an enzyme involved in pathological decalcification by Mϕs [38]. Treatment of Mϕs with IL-10, the prototypic M2 cytokine, can enhance iNOS activity, a prototypic M1 characteristic, under select infectious or hypoxic conditions [39, 40]. Combinations of IL-4 with GM-CSF, M-CSF, glucocorticoids, or TGF-β each induce unique functional phenotypes in Mϕs that are distinct from the phenotypes induced by any of the cytokines alone [21, 37, 41, 42]. It is therefore not surprising that many investigators have described Mϕ phenotypes that are neither M1 nor M2 and/or that are a mixture of inflammatory and anti-inflammatory activities [17, 34, 37, 43]. The term “inflammation” is, itself, somewhat vague. There are many types of inflammation (debriding/necrotizing, fibrotic, granulomatous, etc.), and as demonstrated by the studies discussed above, it is difficult to categorize a cytokine as inflammatory or anti-inflammatory without defining the context (microenvironmental signaling milieu) in which it is acting. We clearly need to define our terminology more precisely if we hope to unravel the mysteries of Mϕ function. The most direct means of clarification at this juncture is to refer to Mϕs by the function(s) they express (e.g., cytotoxic, fibrotic, scavenging) rather than by clustering multiple phenotypes under vaguely defined acronyms.

It has been known for several decades that cytokines can selectively modulate Mϕ functional responses to a given stimulus [44, 45]. In an effort to bring together the concepts of signaling cross-talk and microenvironmental context in a coherent context compatible with the extensive, functional heterogeneity of Mϕs, we presented our perspective on Mϕ functional plasticity, the key premise, which was that the Mϕ functional phenotype could change progressively as the signaling milieu of its microenvironment changed progressively [1, 9]. This hypothesis emphasized the dynamic changes in the signaling environment that influence Mϕ function progressively when tissue homeostasis is disrupted. This perspective is pertinent to Mϕ functional transitions during tissue homeostasis [6,7,8] and inflammatory responses [1, 46], to polarization of Mϕ function in pathological conditions [47,48,49,50,51], and to the potential for therapeutic targeting of Mϕ functional plasticity in those pathologies [1, 9, 34, 52].

TUMOR-Mϕ SYNERGY AND SUBVERSION OF THE IMMUNE SYSTEM

It has become apparent in recent years that most solid tumors actively recruit Mϕs, with which they synergize to establish the proangiogenic, prometastatic, and anti-immunogenic environment that enhances tumor growth and metastasis and subverts the innate and adaptive immune responses [49,50,51]. Mϕs within the tumor mass do not all display the same phenotype. Similar to Mϕs in the lung and other tissues [6, 7], distinct functional phenotypes are displayed by Mϕs in discrete microenvironments of the tumor. Support of angiogenesis and supportive matrix are the dominant activities of Mϕs in hypoxic areas of the tumor. Mϕs proximal to blood vessels secrete epidermal growth factor to stimulate tumor cell migration toward the vessel, secrete matrix metalloproteinase-9 to break down the basement membrane of the blood vessel, and secrete TNF-α to activate endothelial cells, all of which facilitate the intravasation of the tumor cells [49,50,51].

Peritumor Mϕs and Mϕs distal to the tumor produce PGE₂, TGF-β, and IL-10 [53], thus reinforcing the immunosuppressive mandate of the tumor systemically. A degree of this immunosuppressive phenotype can be observed in Mϕs systemically, in every tissue examined, whether or not metastases are evident [52, 54]. How Mϕs in tissues distal to the primary tumor are activated (as opposed to functionally modulated) by the tumor has not been formally established. However, it has been demonstrated that tumor cells release microvesicles that display an array of ligands and cytokines expressed by the tumor cells, which are hypothesized to down-regulate NK cell activity and to stimulate a suppressive functional profile in Mϕs, disrupting the maturation of myeloid dendritic cells and subverting the generation of an adaptive cytotoxic immune response against the tumor [55].

EXPERIMENTAL IMMUNOTHERAPIES IMPACTING Mϕs

It is clear from the discussion above that Mϕs in tumor-bearing hosts are functionally polarized toward immunosuppression and promotion of tumor growth and metastasis. It is now widely accepted that the primary obstacle to immunotherapy of cancer is the strong immunosuppressive environment that is established through TAMϕs [56]. Tumor-bearing hosts thus provide a model to test the hypothesis that Mϕs are functionally plastic and that functionally polarized Mϕs retain that functional plasticity. It has been demonstrated by several research groups that TAMϕs can be induced to display cytotoxic activity against the tumor [57,58,59,60]. A broader goal, which we have pursued, is to reprogram Mϕs to reduce or eliminate Mϕ support for tumor angiogenesis and metastasis and/or Mϕ-mediated immunosuppression and to enhance Mϕ support for reactivation of NK cells and initiation of cytotoxic adaptive immune responses against the tumor.

Of the many cytokines that have been evaluated for cancer therapeutic efficacy, IL-12, which impacts innate and adaptive immune systems, has proven the most interesting [61]. Injection of soluble or lipid-encapsulated IL-12 into tumor-bearing mice resulted in a strong cytotoxic anti-tumor response that is dependent on NK, NKT, Th1, and cytotoxic T cells [62]. We demonstrated recently that IL-12 treatment alters the functional activities of TAMϕs and tumor-infiltrating Mϕs dramatically and rapidly from a tumor-dominated profile (elevated TGF-β, IL-10, MIF, and MCP-1) to a proimmunogenic/inflammatory profile (elevated TNF-α, IL-6, IL-15, and IL-18 and diminished or abrogated expression of TGF-β, MIF, and IL-10) (Fig. 2) [52]. Proangiogenic and prometastatic activities are depressed [63, 64]. Therefore, in mouse models of metastatic cancer, administration of IL-12 appears to undermine Mϕ support for tumor growth and metastasis and also appears to promote an active, cytotoxic response against the tumor. These studies provide strong support for the functional plasticity hypothesis, the mutability of chronically polarized Mϕs, and the therapeutic potential of in situ reprogramming of chronically polarized Mϕ function in cancer and inflammatory disease.

Figure 2.

In situ reprogramming of TAMϕs by IL-12. Treatment of tumor-bearing mice alters the functional phenotype of Mϕs in peritoneum, spleen, and tumor-bearing organs, reducing production of tumor-supportive and/or immunosuppressive cytokines (A) and elevating the production of several potent inflammatory and proimmunogenic cytokines, such as IL-15 (B).

ACKNOWLEDGMENTS

The authors’ research about macrophage biology and its application to tumor immunology was supported by grants from the National Institutes of Health (AI048850 to J. S.), the Susan G. Komen Race for the Cure (R. D. S.), the Kentucky Lung Cancer Research Fund (individually to J. S. and to R. D. S.), and The American Lung Association (S. K. W.) and in part by the Commonwealth of Kentucky Research Challenge Trust Fund (individually to J. S. and to R. D. S.).