Macrophages are a heterogeneous population of innate myeloid cells involved in health and disease. Macrophages originate from monocytic precursors in the blood and undergo specific differentiation depending on local cues in the tissue [1]. Roughly two distinct polarization states are recognized; the classically activated type 1 macrophages (M1) and the alternative activated type 2 macrophages (M2). In response to activating danger signals delivered by bacterial products or IFNγ, macrophages adapt to a M1 phenotype which is tailor-made to attract and activate cells of the adaptive immune system. Important features of M1 macrophages are the expression of iNOS, ROS and the production of NK and type 1 T-cell stimulating cytokine IL-12. M1 macrophages can phagocytose and kill target cells. M2 macrophages can develop in response to for instance IL-4 or IL-13[2, 3], express abundant scavenger receptors and are associated with high production of IL-10, IL-1b, VEGF and matrix metalloproteinases (MMP). M2 macrophages play a role in parasite clearance and wound healing where they also polarize T cells to Th2 and dampen immune responses [4]. Furthermore macrophages are antigen presenting cells (APC) that express class I and class II HLA-molecules and co-stimulatory/inhibitory molecules to instruct T cells, albeit with lesser efficiency than dendritic cells. Notably, the terms M1 and M2 are an oversimplification of the macrophage types that can be detected. Macrophages display great plasticity and can adapt to a plethora of activation states ranging between the M1 or M2 phenotype depending on the mix of signals in their direct microenvironment. A number of excellent reviews on this topic were recently published [2, 5–7]. In addition, fully polarized M1 and M2 macrophages can be redirected in vitro towards the opposite functional phenotype by treatment of the cells with cytokines [8]. For the remainder of this review we will refer to M1 and M2 macrophages under the notion that both M1 and M2 will reflect a whole range of functional states of macrophages.

Alternatively, the number of tumor associated macrophages may form an epiphenomenon reflecting an inflamed tumor micro-environment in which many types of immune cells can be found [48, 61]. This may comprise strong infiltration by regulatory T cells that exert a negative impact on clinical outcome. On the other hand it may also include the infiltration of tumors by T cells with cytotoxic function and a type 1 cytokine profile [62, 63]. Moreover, macrophages do not only contribute to the local inflammation but also impair the activation of an effective anti-tumor T-cell response in tumor draining lymph nodes by inducing apoptosis [64].

In contrast to studies on macrophages in situ it is easier to assess macrophage function in vitro. All macrophages can phagocyte particles and apoptotic cells. Upon activation M1 macrophages will respond with an oxidative burst by the production of ROS and NO via iNOS as well as by the release of interleukin-12, known for its type 1 T-cell polarizing capacity. The use of NO to identify human M1 macrophages is somewhat debated as it was not detected in human macrophage cultures but abundantly found in tissues as detected by IHC (reviewed [72]). Therefore, in vitro studies on human macrophages focus mostly on the production of IL-12 and IL-10 to discriminate between M1 and M2 subsets [66, 67, 78, 79].

New markers that may assist to discriminate M1 from M2 macrophages are typical transcription factors expressed by these cells [80]. M1 cells highly express IRF-5, a transcription factor that is involved in stimulating the production of type 1 interferon while repressing the production of IL-10. Forced expression of IRF-5 in M2 macrophages switches them to M1 macrophages. Genetic polymorphisms that induce overexpression of IRF-5 mRNA in human auto-immune pathologies like SLE and multiple sclerosis illustrate the link between high IRF-5 expression and type 1 inflammation in human beings (reviewed in [81, 82]). Secondly, the signal transducer and activator of transcription (STAT) protein family is associated with the function of immune cells. Cytokine induced phosphorylation and nuclear expression of STAT1 is associated with a type 1 pro-inflammatory phenotype whereas STAT3 resembles an anti-inflammatory phenotype. STAT3 expression is often found in tumors and tumor-infiltrating macrophages and is induced by IL-6 and IL-10 [77].

Last but not least, macrophages produce chemokines to attract other immune cells and the type of these chemokines may offer help to distinguish between the macrophage subsets. For example, activated M1 macrophages produce CXCL-10 (IP-10) in order to attract CD8 and Th1 CD4+ T cells. M2 macrophages attract immune cells by the production of CCL-22 and CCL18. Differentiation between human M1 and M2 macrophages on the basis of their chemokine production has so far only been evaluated in vitro[18, 79]. In vivo, most studies focused on the total tumor-derived chemokinome and, therefore, the exact role of macrophage-produced chemokines to attract immune cells is not clear yet.

It has to be said that there are a number of reports suggesting that also tumor cells and monocytes might express CD163, the detection of which depends on the antibody that is used [43, 68]. Furthermore, MMPs are reported to be produced by tumors as well [50]. MMPs are thought to contribute to tumor progression, however, there is accumulating evidence that MMPs may also confer tumor protection [70]. In our opinion, careful assessment of macrophages in the tumor requires preferably the use of 3 markers to identify macrophages and their functional profile.

Tumors require nutrients, oxygen and the ability to discharge metabolic waste and carbon dioxide. These needs are addressed by tumor associated neovascularization [11]. Macrophages are the perfect help in this process, especially the wound-healing subset, since they are equipped to remodel tissue and produce VEGF. How tumors attract these myeloid cells has been excellently reviewed before [83].

CCL-2 produced by tumor cells and tumor associated cells attracts and shapes myeloid cells and interferes with osteoclasts in bone- metastatic disease [84, 85]. Blocking CCL-2 reduced the number of tumor-infiltrating macrophages in pre-clinical animal studies. Secondly, the transcription factor Hypoxia Inducible Factor-1 (HIF-1), highly expressed under hypoxic conditions, regulates the attraction of monocytes and macrophages into the tumor amongst others by induction of Colony Stimulating Factor-1 (CSF-1/M-CSF) [83].

Many studies have been conducted to determine which factors induce macrophage differentiation upon arrival in the tumor. The levels of several cytokines were measured in tumor fluids, in ascites or in blood. Furthermore, the gene expression profiles of tumor cells have been analyzed. Interleukin-6 (IL-6) was often reported to be present in tumor fluids and was a produced by many tumor derived cell lines. IL-6 is a cytokine involved in chronic inflammatory diseases such as rheumatoid arthritis [86]. A number of studies on the in vitro differentiation of M2 macrophages - when monocytes or immature DC are exposed to tumor cell supernatant derived from lung, ovarian or cervical cancer cell lines - showed that this differentiation depended on IL-6. In nearly all cases IL-6 was shown to act in synergy with other factors [66, 78, 79, 84, 87]. Binding of IL-6 to its receptor results in activation of STAT3 in immune cells and consequently in the suppression of immune mediators that activate the immune response [77].

CSF-1 (M-CSF) mediates, besides being a chemo attractant, the differentiation of monocytes into CD14+CD163+CD206high M2 macrophages in vitro[67]. A similar dual role has been suggested for VEGF which acts as an angiogenic factor when secreted by wound healing macrophages but which can also impair APC differentiation when secreted by tumor cells [88].

The inflammation-inducible cyclo-oxygenase-2 (COX-2)-derived prostaglandin E2 (PGE2) is overexpressed by several human tumor-types [89]. The role of COX and PGE2 in oncogenesis has been well investigated in colorectal cancer. Blocking this pathway prevented oncogenesis in the low-intestines [90]. In vitro the addition of PGE2 to IL-4/GM-CSF stimulated monocytes skews their differentiation towards M2 macrophages [66, 78, 91, 92]. Notably, overexpression of COX-2 in human lung cancer cells resulted in the production of IL-6 and the subsequent phosphorylation of STAT3. This rendered the tumor cells more resistant to apoptosis as well as increased their VEGF production [93].

Human tumors express and release high levels of Heat Shock Protein 27 (HSP27) (reviewed by [94]). The addition of soluble HSP27 to monocytes in vitro directed their differentiation to macrophages with an immune tolerizing and proangiogenic phenotype [95].

Thus, inflammatory mediators attract myeloid cells to the tumor and direct their differentiation towards tumor promoting macrophages (Figure 1). The macrophage in the tumor tissue is subject to local levels of many factors that lead to a great variety in myeloid cell subsets even within a single tumor. Interestingly, most described factors are also produced by macrophages suggesting a feed forward loop.

The plasticity of macrophages has been shown in vitro but whether M2 macrophages can revert to M1 and vice verse within the tumor environment is an emerging question. In vitro, exposure of murine macrophages to a type 1 cytokine environment skews them towards M1 macrophages. In contrast, the macrophages adapted a M2 phenotype upon exposure to IL-10 [8, 96]. Adaptation to a particular phenotype was reversible by subsequent changes of the cytokine milieu [96]. Also human macrophages show great plasticity with respect to macrophage phenotype following stimulation with anti- or pro-inflammatory stimuli [97]. IFNγ, which was originally called macrophage activating factor, plays a major role in the differentiation towards M1 macrophages. It's produced by both innate and adaptive immune effector cells, including CD8 T cells and several subsets of CD4 T cells, NK cells as well as γδ T cells. IFNγ induces STAT1 and IRF-5 signaling-mediated type 1 cytokine production. Dulluc et al. comprehensively showed that IFNγ can prevent and reverse the M2-skewing effect of ascites from patients with ovarian cancer. Furthermore, pre-treatment of M2 macrophages with IFNγ results in a macrophage type that produces more IL-12 and less IL-10 upon activation with several innate stimuli [79].

Cognate interaction with CD4+ helper T cells constitutes a powerful activation signal for antigen presenting cells. The signal is delivered via CD40-CD40L interaction following T-cell receptor mediated MHC class II-presented antigen recognition. CD40L and IFNγ can activate in synergy but do not have to come from the same CD4 T cell. Exposure of APC to IFNγ-producing CD8+ T cells and CD40-L+ CD4 T cells allows APC to produce IL-12 and to become strong promoters of Th1 responses [98, 99]. We showed recently that tumor skewed IL10-producing M2 macrophages can fully revert to IL12-producing M1 macrophages following interaction with CD4+ Th1 cells or with CD40-L-expressing cells in the presence of IFNγ [66]. Notably, in this in vitro study CD40-CD40L interaction resulted only in a strong activation signal for macrophages that was required but not sufficient to change the macrophage subset. Interestingly, we have observed that a minor population of tumor-infiltrating human CD8+ T cells expresses CD40-L upon activation [100]. This poses the question whether certain populations of intratumoral CD8 T cells can mediate the same effect on macrophages as CD4+ Th1 cells. The important role of T cells in instructing APC is further supported by studies showing that freshly isolated and activated blood-derived regulatory T cells skewed monocytes to M2 macrophages, which were able to produce high levels of IL-10 when stimulated with LPS [101]. The regulatory T cells, defined as CD4+CD25high cells, exerted their effect via the secretion of IL-13, IL-4 and most importantly IL-10 that was produced upon T-cell activation [101].

Mouse tumor-models support the significance of cognate interactions between CD4 T cells and macrophages. In an aggressive breast tumor-mouse model IL-4-producing CD4+ T lymphocytes indirectly promoted invasion and subsequent metastasis of mammary adenocarcinomas by regulating the phenotype and effector function of tumor-associated macrophages [102]. In another study, myeloma tumors were infiltrated by macrophages within 3 days, followed by tumor-specific CD4+ Th1 cells at day 6 after which the macrophages started to kill tumor cells [103]. This indicates that during early tumor development a Th1-macrophage response controls tumor outgrowth. Thus the type of T cell that interacts with tumor associated APC influences the polarization and function of the APC. Depending on the type of T cell, APC can swap their function allowing them to suppress or to stimulate immunity and vice verse.

Studies that evaluate the type of macrophages in relation to the number and type of T cells in patients are scarce. In ovarian cancer high numbers of B7H4 expressing immunosuppressive macrophages correlated with increased numbers of FoxP3+ T cells [30]. Accordingly, in gastrointestinal stromal tumors the number of CD163+M2 macrophages correlated with CD3+FoxP3+ regulatory cells [48]. A similar relation was reported for hepato-cholangiocarcinomas [24]. In contrast, in Hodgkin lymphoma an inverse correlation with CD68+ macrophages and FoxP3+ T cell numbers was reported. Here macrophages positively correlated with higher numbers of granzymeB and PD-1 expressing T cells instead. Alas, in this last study the type of infiltrating macrophages was not evaluated but the number of macrophages correlated to poor prognosis suggesting a M2 type of infiltration [47]. Finally, tumor-infiltrating macrophages were reported to attract IL-17+ CD4 T cells in liver carcinoma, thereby sustaining inflammation and supporting tumor promotion [104]. A recent study showed that a high number of macrophages and CD4+ T cells was correlated to a bad prognosis in breast carcinoma whereas CD8+ T cell dominance related to good prognosis [105]. Based on the MMTV-PyMT breast tumor mouse model, in which these authors also found a relation between macrophages, CD4 T cells and poor outcome, the CD4 T cells are likely to be Th2 cells [102]. Although these studies show that the number of infiltrating macrophages is related to the number of (regulatory) T cells it is not clear if these cells interact and what the outcome of interaction is with respect to macrophage function. For that the functional properties of all infiltrating T-cell subsets should be evaluated.

The strong infiltration of human tumors by activated CD8+ cytotoxic lymphocytes (CTL) and CD4+ helper T (Th) cells but not immune suppressive cells such as regulatory T cells, M2 macrophages and myeloid derived suppressor cells is a hallmark for improved survival after therapy (reviewed by [106]). The treatment of cancer may take advantage of therapies that interfere with M2 macrophages, if combined with standard or immunotherapeutic regimens. Therapeutic modalities may attack at several levels; the attraction, the differentiation or the activation of macrophages.

One therapeutic option is to interfere at the level of macrophage attraction and differentiation by abrogation of the PGE2, IL-6 and STAT3 activation loop. This may directly affect tumor growth and limit the induction of tolerogenic macrophages as well. Several trials were conducted in which PGE2 production was inhibited by blocking of the COX-2 enzyme, often in combination with standard cytotoxic therapy [90, 107]. Since all studies were initiated to block the paracrine loop on tumor cells not much is documented about the immunological effects let alone macrophage differentiation. One report mentions high immune cell infiltration, including macrophages upon 3 days of treatment with the COX-inhibitor [108].

A second strategy to interfere with immune cells in the tumor micro milieu is the blockade of cytokines secreted by the tumor or immune cells (e.g. M-CSF or IL-6). Recently, a phase II trial in patients with ovarian carcinoma was reported were the anti-IL-6 antibody siltuximab was tested [87]. The goal in this study was to deprive the tumor from growth factors and to prevent the differentiation of M2 macrophages. It might prove difficult to reach high enough antibody titers to block cytokine levels as macrophages tend to accumulate in poorly vascularized tumor areas [83]. Blocking of the cytokine receptors is not advised as membrane bound antibodies may activate type 2 macrophages via cross linking of the FcγReceptors and thereby activate tumor promoting cascades. This was recently demonstrated in vitro for the humanized Fc region optimized antibody cetuximab that binds the EGF-Receptor at the tumor cell surface and which caused activated M2 macrophages to produce IL-10 following the cross-linking of Fc-receptors [65].

All above mentioned strategies aim to inhibit the induction of M2 macrophages but do not reprogram M2 macrophages once they exist. In vitro macrophages can adapt to another phenotype upon a strong polarizing stimulus as discussed before [97]. Therefore, deliverance of a local APC-activating stimulus may be an attractive option. Macrophages highly express PAMP-receptors and several agonists for TLR and NOD-receptors are currently under clinical investigation for the treatment of cancer. For instance, easy accessible (pre-)malignancies have already been treated with TLR7 agonists [109]. We showed that activation of M2 macrophages in vitro by a number of different TLR agonist's results in the loss of expression of typical M2 markers but did not alter their typical M2 function [66]. Following this principle through made us realize that it also bears impact on other studies in this field. The group of Zitvogel extensively studied the effects of chemotherapy and showed that HMGB-1 released from dying tumor cells acted as a TLR4 agonist to activate APC [110]. The specific immune modulating properties of proteins released from dying tumor cells (e.g. HMGB-1) on the activation of M2 macrophages via PAMP-receptors is not yet studied.

Reprogramming of M2 to M1 type macrophages not only requires receptor-mediated activation signals but the presence of polarizing cytokines (e.g. IFNγ) as well. TLR are also expressed by T cells and stimulation modulates T-cell responses upon TCR triggering [111]. At the T-cell population level this may result in the enhanced production of IFNγ by T cells but also the downregulation of regulatory T-cell function [111]. One can envisage that if TLR-agonists not only activate macrophages but also T cells to produce IFNγ this will result in reprogramming of the macrophages. On the other hand, our own studies on the TLR expression by patient derived tumor-antigen-specific CD4+ T-cell clones, including helper T cells and regulatory T-cell clones, revealed that human T cells express a number of different TLRs at the mRNA level and that the TLR expression pattern differed per CD4+ T-cell clone independent of their function. When these CD4+ T-cell clones were tested in various activation (proliferation, cytokine production) and suppression assays no consistent gain (proliferation/IFNγ production) or loss in function was observed due to the presence of TLR-agonists either with or without TCR triggering. This suggests that TLR may not play a major role when CD4+ T cells are fully polarized already (unpublished observations). It is, therefore, questionable whether the application TLR agonists may result in repolarization of macrophages.

A recent study revealed that activation of macrophages by the infusion of antibodies against CD40 may induce macrophage-mediated tumor regression in 30% of cases in both a mouse model for pancreatic cancer and in patients with pancreatic cancer [112]. CD40 therapy resulted in the stimulation of secondary lymph node resident macrophages to migrate into the tumor tissue. Whereas, tumor-infiltrating in non-treated animals predominantly produced IL-10, the tumor-infiltrating macrophages of antibody treated mice produced IL-12 [112]. Notably, macrophages already present in the tumor were not stimulated by this treatment, possibly due to a low tissue penetration of the antibody [112]. In vitro CD40-activation with IFNγ was required to effectively reprogram tumor-induced M2-like macrophages into activated IL-12 producing M1 cells (Figure 1). As CD40 is not only expressed by macrophages but also by monocytes, endothelial cells, epithelial cells, and B cells [113], it is likely that the activation of several cell types throughout the body resulted in the deliverance of the necessary pro-inflammatory cytokines that allowed the circulating CD40-activated macrophages to adapt to a M1 phenotype.

In order to reprogram macrophages directly in the tumor-microenvironment, it is essential that CD4+ Th1 cells are locally present. The capacity of tumor-specific Th1 cells to directly alter the tumor microenvironment has also been recognized in studies on tissue-infiltrating CD8+ T cells in mice models. Th1 cells were essential for successful recruitment, local expansion and full effector function of large numbers of CTL by modulation of the local environment [114, 115], and this may have included the repolarization of macrophages. In order to obtain sufficient numbers of tumor-specific CD4+ Th1 cells one may make use of adoptive T-cell transfer protocols [116, 117] or apply strong vaccines [118, 119].

Macrophages play an important role in tumors. Depending on the mode of activation, they may promote tumor growth and suppress local immunity or attack tumor cells and sustain tumor immunity. Immunotherapeutic strategies to combat cancer should incorporate approaches focused on the attraction and polarization of M1 macrophages as well as on the reprogramming of M2 macrophages to the M1 subset. This requires a well-based understanding of the different subsets of macrophages in human tumors as well as their interaction with other members of the immune system, including T helper cells. To this end, the type and numbers of macrophages in the tumor have to be more carefully analyzed by the simultaneous use of several markers that allow discriminating the different macrophages subsets in one run. New and better subset-specific markers need to be identified. In addition, it will require in-depth studies on the interaction of local T cells and macrophages, a subject which currently is only marginally studied in human cancers.