Molecular Pathways Regulating Macrophage Polarization: Implications for Atherosclerosis

Over the past two decades, the inflammatory hypothesis of atherothrombosis has gained an increasingly strong footing through multiple lines of supportive evidence. Overall, an increased systemic burden of inflammation prompts a higher CVD incidence, as is the case in chronic inflammatory conditions such as rheumatic arthritis [8] and systemic lupus erythematosus [9]. Various soluble mediators of the inflammatory response have been found to predict future cardiovascular risk in atherosclerotic patients (well-described in [10]). High-sensitivity C-reactive protein (hsCRP) has formed a focus point in this respect, as systemic concentrations of this acute-phase protein compared favorably with LDL cholesterol and blood pressure as CVD risk factors [11], and were specifically associated to plaque vulnerability [12, 13]. Building on post hoc analyses from several other large-scale studies (eg, CARE, PROVE-IT TIMI 22, AFCAPS/TexCAPS trials [14‐16]), the JUPITER trial prospectively consolidated the correlation of hsCRP and cardiovascular outcome in a primary prevention setting [17]. Researchers observed that the clinical benefits of statin therapy were greatest when both LDL and hsCRP values were reduced, thus connecting both dyslipidemia and inflammation at the interface of CVD pathogenesis. Intriguingly, even with pre-existent LDL levels below the clinical cut-off point for treatment, persistent inflammation as measured by increased hsCRP levels puts patients at a higher than anticipated risk of CVD. In the AFCAPS/TexCAPS trial, these subjects responded strongly to treatment [16], indicating LDL burden is not a prerequisite to successful therapy. Apart from providing clinicians with valuable information for risk assessment, this finding proposes that an enhanced inflammatory state might in itself justify targeted therapy. Indeed, US and Canadian prevention guidelines have since embraced hsCRP measurements in the considerations for patients at intermediate risk. Moreover, a number of new trials, using either low-dose methotrexate (CIRT) or anti-IL-1β monoclonal antibodies (CANTOS) as anti-inflammatory treatment, are underway to address and possibly validate the hypothesis of inflammatory causality [18•, 19•]. These translational efforts could provide a major argument towards a more systematic implementation of anti-inflammatory therapy in our continuing battle to diminish residual cardiovascular risk.

Substantial experimental evidence complements the broad clinical involvement of inflammation in CVD outlined above. Now most agree that systemic risk factors interact with many cell types (both those intrinsic to the vasculature and immune cells attracted from the circulation) to drive plaque development. Particularly, monocyte-derived macrophages are considered critical participants in the atherogenic process, as they secrete pro-inflammatory cytokines and other mediators that affect lesion progression and stability. Consequently, many experimental studies have successfully targeted the abundance of monocytes/macrophages and their soluble repertoire in atherosclerosis as a means of prevention. For instance, atherosclerotic plaque formation was virtually abolished in hyperlipidemic mice lacking the macrophage-colony stimulating factor (M-CSF) gene, which exhibit impaired monocyte development and subsequent differentiation to macrophages [20, 21]. Other scientific efforts involved the abrogation of chemokine-dependent monocyte recruitment to the plaque [22], in addition to a wealth of studies addressing the various cytokines produced by macrophages and other cells (reviewed in [23]). Although not cell-specific, these data still offer valuable insight into how macrophages contribute to nascent lesions. Macrophage apoptosis is another important feature seen during atherosclerosis development. In early lesions, macrophage apoptosis and plaque size exist in an inverse relationship [24], whereas in later stages this process contributes to the plaque’s lipid core [25]. This ambiguity appears to be mediated by a process termed “efferocytosis” [26]. Combined with proof linking plaque macrophages to matrix metalloproteinase (MMP)-dependent collagen breakdown [27, 28], it is evident that these cells modulate inflammatory mechanisms to determine plaque susceptibility to rupture and clinical atherothrombosis. Associative imaging studies lend weight to this notion by demonstrating that vascular uptake of fluorodeoxyglucose (¹⁸F-FDG) correlates to plaque macrophage load and general inflammatory burden and can thereby assist in the prediction of cardiovascular events [29, 30].

Cultured and tissue macrophages both exhibit pronounced heterogeneity, as was recognized early on [31]. Skewing of macrophages toward distinct polarization programs occurs in response to various environmental cues and has inspired extensive research into their significance in pathophysiology. Reflecting the Th1 and Th2 nomenclature in T-cells, polarized macrophage subsets were originally referred to as classically (M1) or alternatively activated (M2) [32]. The latter group was subsequently divided into M2a, M2b, M2c and tumor-associated macrophages (TAMs) to distinguish between inducing stimuli [33]. Later on, Mosser and Edwards [34] reassigned these macrophage subsets to three different classes: classically activated macrophages (CAMs, corresponding to M1), alternative activated macrophages (AAMs, also referred to as wound healing macrophages and analogous to M2a), and regulatory macrophages (RMs, consistent with M2b/c). These classes are best considered a continuum of functional states that encompasses a broad range of macrophage phenotypes with interchangeable characteristics.

CAMs are typically induced by the Th1-cytokine interferon-γ (IFNγ), possibly followed by activation with a Toll-like Receptor (TLR) ligand, such as lipopolysaccharides (LPS). IFNγ (originally termed macrophage-activating factor) prepares macrophages for a pro-inflammatory environment, giving rise to a potent inflammatory response upon microbial challenge. The resulting phenotype characteristically displays high interleukin-12 (IL-12) and low IL-10 production, combined with enhanced microbicidal effector functions through the induction of NADPH-oxidases and inducible nitric oxide synthase (iNOS). Therefore, these macrophages become very efficient in killing bacteria, viruses, parasites, and fungi. Their continuous induction and sustained activation, however, will cause tissue damage [35].

The Th2 cytokines IL-4 and IL-13 are produced by granulocytes, mast cells, and Th2 cells during injury and infection and generate AAMs [36, 37]. Opposite to CAMs, AAMs dampen inflammatory responses through an IL-12^low and IL-10^high expression profile. Furthermore, AAMs promote tissue repair and fibrosis through increased arginase-1-dependent production of the collagen precursors ornithine and proline [38]. Besides arginase-1, other markers for AAMs include Ym1, Fizz1, and the mannose receptor (MR).

Finally, regulatory macrophages (RMs) are induced in response to a wide range of stimuli, including immune complexes, prostaglandins, G-protein coupled receptor ligands, glucocorticoids, and uptake of apoptotic cells. However, a key cytokine in the induction of RMs is the anti-inflammatory, atheroprotective cytokine IL-10. The main task of RMs is to suppress and control immune responses by producing high levels of IL-10 and thereby contribute to the resolution of inflammatory responses [34, 39].

The current framework of macrophage subsets is very well characterized in vitro but less so in in vivo settings. Here, a greater variety of external challenges elicit macrophage phenotypes that are considerably less adherent to the constraints of the existing paradigm. During monocyte-to-macrophage differentiation and subsequent macrophage activation, the collective imprint of these environmental factors shapes the macrophage phenotype. An intuitive overview of these sequences was recently described by Gordon and Martinez [40].

First, monocyte recruitment from the circulation will be followed by differentiation to mature tissue macrophages. This transition is primarily mediated by the growth factors M-CSF and granulocyte-macrophage colony-stimulating factor (GM-CSF) that influence the inflammatory potential of the resulting macrophage. Maturation with M-CSF was observed to lead to a more anti-inflammatory (IL-10^highIL-12^low) phenotype, whereas GM-CSF-induced differentiation gave rise to a macrophage population with pro-inflammatory (IL-10^lowIL-12^high) characteristics [41, 42].

The phenotype that ensues from the subsequent priming stage hinges on the local balance of cytokines and chemokines in the newly recruited macrophage’s environment. Priming with cytokine stimuli will affect the inflammatory potential of macrophages and their response to other stimuli. In addition to the previously described cytokine stimuli (IFNγ, IL-4/IL-13, and IL-10 for CAMs, AAMs, and RM, respectively), chemokines (eg, CXCL4) [43] and other plaque constituents (oxidized lipoproteins) [44] were also shown to induce unique macrophage phenotypes with distinctive characteristics.

Upon activation by Toll-like or analogous receptor stimuli, the macrophage will undergo functional maturation that results in a rapid induction of anti-microbial pathways for fast killing and clearance of pathogens. Finally, when the macrophage has fulfilled and survived its inflammatory task, it will undergo deactivation. In this final phase, the RM phenotype and its mediators transforming growth factor-β (TGF-β), IL-10, and lipoxins, are key contributors to the resolution of inflammation and tissue repair [40].

Macrophage subtypes have the capacity to switch from one phenotype to another as stimuli from the micro-environment change. Recently, such plasticity was elegantly demonstrated in a murine kidney ischemia-reperfusion model [45•]. Whereas initially pro-inflammatory CAMs were recruited to the kidney, later time points saw previously attracted macrophages switch to an MR⁺-AAM phenotype, signaling the resolution of inflammation and repair. Similar reports were made with regard to other disease states, as for instance in metabolic disease. Adipose tissue macrophages in lean mice resemble an AAM phenotype that supports adipocyte function and insulin sensitivity [46]. Obesity, however, switches the macrophage balance to a CAM phenotype as the result of inflammasome activation by various danger signals from adipocytes, thereby promoting inflammation and insulin resistance [47, 48]. In analogy to this dichotomy, differential macrophage function has been reported to influence the course of various other conditions, such as cancer and infectious disease [34, 49].

Likewise, in atherosclerosis, evidence implies that the balance of macrophage polarization is crucial in determining plaque outcome [49‐51]. Following the progression of atherosclerosis over time, it was shown that plaque macrophages in early lesions express arginase-1 (indicative of AAMs), whereas at later time points the expression of arginase-2 (referred to as a CAM marker) predominated in the plaque [52]. More recently, phenotypic switching in atherosclerosis was described upon induction of disease regression [53••]. Here, plaque macrophages restrained pro-inflammatory marker expression (eg, monocyte chemotactic protein-1 and tumor necrosis factor) and enhanced that of AAM markers such as arginase-1, MR, and CD163 upon normalization of the plasma lipid profile. However, the order of events in this model still requires clarification (ie, does phenotypic switching actively contribute to plaque regression or is it merely a consequence thereof). In human subjects too, it was shown that symptomatic atherosclerotic disease of the carotid arteries is associated with a CAM profile, characterized by high pro-inflammatory cytokine expression [54]. Moreover, enrichment of CD11c⁺-CAMs over MR⁺-AAMs characterizes the epicardial adipose tissue from patients with manifest coronary artery disease [55]. These reports are consistent with the current hypothesis that CAMs propagate plaque vulnerability through their inflammatory potential, whereas their AAM counterparts are likely to benefit plaque stability and regression through a repertoire of repair and fibrotic functions. From this perspective, Fig. 1 provides a schematic overview of the inducers, transducers, and effector functions of these macrophage populations in atherosclerosis development.

Although murine macrophage subtypes are well characterized, the translation to human pathophysiology is not straightforward. Where polarized macrophage subsets in mice can be easily distinguished based on their arginine-metabolizing enzyme expression (i.e. iNOS and arginase-1 for CAMs and AAMs respectively), in vitro polarized human macrophages do not express arginase-1 and fail to produce NO in amounts comparable to that of mouse macrophages [57]. In addition, markers like Ym1 and Fizz1, which are excellent for AAM detection in mice, are not expressed in humans [35]. Circumventing these issues through identification of generally more favorable markers of human tissue macrophages will prove an important step in translational atherosclerosis research.

To achieve expression of the aforementioned surface marker and cytokine profiles, different macrophage subsets are known to employ distinct signaling pathways (Fig. 1). These cascades usually activate certain transcription factors that effectively induce expression of archetypal genes, thereby accounting for the translation of external polarizing cues into a specific subtype [58]. Clearly, this designates transcription factors as important players in macrophage polarization that may represent an attractive alternative to surface markers for discerning macrophage populations in vivo. Moreover, these pathways may eventually yield suitable targets for the inhibition, depletion, or stimulation of a particular phenotype. An overview of transcription factors associated to macrophage heterogeneity and their functional relevance in experimental or human atherosclerotic lesions, as described below, can be found in Table 1.

In addition to transcriptional control, epigenetic regulation is essential for a properly directed expression of target genes. Without modifying the actual genetic code, epigenetic mechanisms affect DNA accessibility to the transcriptional apparatus to alter gene expression and even mRNA degradation by microRNAs. Remarkably, the resulting gene expression patterns can be passed down to daughter cells upon cell division or even trans-generationally. Epigenetic regulation can be achieved through DNA methylation (generally considered a relatively static process), as well as through applying certain modification to histones, such as acetylation or methylation. These marks can be changed very dynamically in response to many environmental stimuli.

In macrophages, several epigenetic mediators now are identified as key players in macrophage function and polarization [93]. For instance, Jmjd3, a JmjC family histone demethylase that erases lysine 27 trimethylation marks on histone 3, has been shown to specifically bind transcription start sites of promoters of NF-κB dependent genes. This would suggest that Jmjd3 provides additional control of NF-κB-dependent gene expression, although most pro-inflammatory genes were expressed independently of Jmjd3 [94, 95]. In addition, it has been described that Jmjd3 plays an essential role in polarization. In response to AAM polarization, Jmjd3 can induce the expression of IRF4, resulting in the transcription of key AAM marker genes [62, 69]. Recently, the histone deacetylase 3 (HDAC3) was also found to mediate macrophage polarization, as HDAC3-deficient macrophages were shown to be hyperresponsive to polarization with IL-4 [96]. Thus, alternative activation of macrophages can be controlled at an epigenetic level by targeting HDAC3 and Jmjd3, suggesting that epigenetic mediators might evolve to become promising immunomodulatory targets. This notion was recently substantiated by the finding that the bromodomain and extra terminal domain family of proteins-inhibitor (I-BET) is able to disrupt chromatin complexes responsible for the expression of key inflammatory genes during macrophage activation and could thereby protect against endotoxic shock and sepsis [97•]. This indeed very nicely demonstrates that synthetic compounds specifically targeting epigenetic mechanisms can act as excellent therapeutic agents for immunomodulation. As future studies further unravel the epigenomic landscape of macrophage subtypes, we will definitely see the development of other immunomodulatory compounds that target epigenetic regulation of macrophages. Moreover, in experimental atherosclerosis, it was shown that cell-specific histone methylation modifications and expression of accompanying lysine methyltransferases occur in carotid arteries. More specifically, differences in histone methylation modifications in vascular endothelial and smooth muscle cells were described between the offspring of hypercholesterolemic and normocholesterolemic mothers [98]. As histone modifications are readily altered in response to environmental stimuli, they can provide an attractive explanation how diet and lifestyle may contribute to atherosclerosis susceptibility [99].