Therapeutic Manipulation of Tumor-associated Macrophages: Facts and Hopes from a Clinical and Translational Perspective

Attempts to stimulate effective immune responses against cancer have kept immunologists busy for many years. After decades of modest outcomes, a significant breakthrough was achieved with the use of inhibitors to immune checkpoints (e.g., PD-1 and CTLA-4) that reactivate antitumor immunity, demonstrating the possibility of manipulating a patient's immune system to elicit significant responses (1, 2).

Cooperation between innate and adaptive immunity is desirable for achieving a long-lasting, efficient antitumor response. However, innate immune cells in patients with cancer are not only mostly inefficient against tumor cells, but actually promote tumor progression and hamper treatment efficiency (3–8). Tumor-associated macrophages (TAMs) are the most represented component of innate immunity in the tumor microenvironment (TME). Evidence that TAMs are tumor promoting is accepted overall, although some distinctions should be considered for specific tumors (e.g., colorectal cancer) or conditions (early tumors vs. established tumors). TAMs support cancer cell survival and proliferation, distant spreading, and angiogenesis; they also build an immunosuppressive milieu, which hampers the cytotoxic function of T lymphocytes. Therefore, there is a growing interest in manipulating TAMs for therapeutic purposes.

Some macrophage-directed strategies have shown successful results in experimental settings and are now considered in clinical oncology as promising therapies (3–5, 9–12). Although initial preclinical studies were primarily meant to deplete macrophages in the TME or prevent their arrival at tumor sites, recent evidence has indicated that the reeducation of macrophages into antitumor cytotoxic effectors might be more beneficial in eliciting an effective antitumor immune response (3). This latter strategy relies on the remarkable plasticity of macrophages to reprogram their functions in response to external cues. Macrophages are heterogeneous cells characterized by high phenotypic and functional flexibility (3, 13–15). Although any definition is too tight for the varied macrophage world, the original simplified paradigm categorized macrophages into two extreme polarizations: M1 and M2 types. TAMs were assigned to an M2-like phenotype because they substantially differed from M1 macrophages in terms of surface markers and functional activities and were more similar (but not identical) to M2 macrophages (3). In recent years, single-cell resolution studies of RNA sequencing, mass cytometry by time-of-flight, and multiplexed cytometry have disclosed a picture of TAM heterogeneity that is even more complex (16). In addition to their complex diversity, there is emerging evidence that TAMs originate not only from bone marrow–derived circulating precursors (monocytes), but also from tissue-resident macrophages, with distinct features (17–19).

In this review, we summarize the state-of-the-art of the pharmacologic manipulation of TAMs to counteract their negative effects on tumor progression and improve the treatment of patients with cancer.

The way TAMs have been manipulated for therapeutic purposes can be divided into two main approaches: (i) reduction of their numbers in the tumor tissue, either by in situ elimination or inhibition of their arrival from the circulation and (ii) reprogramming of TAMs into antitumor M1-like macrophages (Fig. 1).

Early attempts to reduce macrophage count in solid tumors employed toxic compounds, such as bisphosphonates, which induce macrophage apoptosis. The depletion of TAMs using this approach, in combination with chemotherapy or hormonal therapy, has shown benefits in the clinic for the treatment of bone metastases, for instance, in breast and prostate cancers (20). Similarly, trabectedin, a marine-derived registered antitumor agent that is specifically cytotoxic to the monocytic cell lineage through TRAIL-dependent apoptosis, also showed antitumor efficacy mediated by the elimination of TAMs (21).

Instead of using cytotoxic drugs, other approaches to reduce macrophages in tumors have employed inhibitors of chemoattractants that modulate the recruitment of circulating monocytes at the tumor site or factors important for their survival and differentiation. Notably, in the context of cancer-related inflammation, tumors produce hematopoietic growth factors, stimulating the so-called emergency myelopoiesis. This expansion of myeloid cells includes TAM progenitors (monocytes) and myeloid-derived suppressor cells (MDSC; ref. 22).

mAbs or kinase inhibitors used to disrupt the chemokine–receptor axis (i.e., CSF-1/CSF-1R, CCL2-CCR2, or CXCR4/CXCL12), controlling monocyte recruitment, have been tested with some success in experimental tumors (3, 12, 23), but with no substantial positive results in the clinic even when combined with chemotherapy or immunotherapy. Among these drugs, as the CSF-1 ligand (M-CSF) is the master regulator of macrophage survival and differentiation, inhibitors of the CSF-1 receptor (small molecules, such as PLX7486 and PLX3397, and the antibody, AMG820) have gained interest, showing excellent antitumor activity in experimental mouse models (24, 25), and are now being tested in numerous clinical trials in combination with a broad variety of antitumoral strategies (Table 1; ref. 26). Among the challenges to the effective application of these approaches in the clinic are the constant release of new monocytes from the bone marrow; the inefficient depletion of TAMs, which resulted in a maximum of 50% reduction; and the large redundancy of the chemokine world, with many different ligands and receptors.

However, it must be considered that the unselective depletion of monocytes/macrophages may be harmful, as their role in normal tissues is crucial for host defense and homeostasis (13). Recent knowledge of the wide heterogeneity of TAMs suggests that it is probably more beneficial to target only specific subsets with high immunosuppressive and protumor activities. As an example, the specific depletion of CD163⁺ TAMs in an experimental model of melanoma attenuated immune suppression and induced the arrival of T cells achieving tumor regression, while the pan-targeting of TAMs did not have therapeutic effects (27).

The intrinsic functional plasticity of macrophages provides the rationale to pursue approaches to reprogram TAMs from M2-like immunosuppressive and tumor-promoting cells into M1-like macrophages with immunostimulatory, antitumor phagocytic and cytotoxic activities (3, 11–14). The reprogramming of TAMs may have a chance of success for three reasons: (i) relief of the oppressed immune compartment, (ii) direct cytotoxicity of macrophages toward cancer cells, and (iii) exploitation of high numbers of TAMs already present at the tumor site.

The activation of innate immunity against cancer cells has been endeavored since the beginning of immunotherapy, with early strategies employing microbial molecules, such as bacillus of Calmette-Guérin (BCG; ref. 28), bacterial muramyl-tripeptide (MTP-PE; ref. 29), or immunostimulatory cytokines (e.g., IFNs and IL2; ref. 3).

In the last decades, our knowledge has greatly increased with the identification of some specific molecules and mechanisms to better modulate macrophage antitumor functions.

A step forward was the discovery that selected chemotherapeutics, such as doxorubicin, oxaliplatin, and bortezomib, induced immunogenic cell death (ICD) and the activation of the immune system against tumors (30, 31). Cancer cells dying via ICD expose calreticulin on the plasma membrane, releasing cytokines and damage-associated molecular patterns (DAMP), such as ATP or HMGB1, and tumor antigens, which ultimately stimulate an antitumor immune response (30, 31). This phenomenon of cytotoxic drug–mediated immune activation was noted many years ago in preclinical models, where tumors with a high density of TAMs responded better to doxorubicin (3). Similar results were found in recent years, when tumors from patients with pancreatic cancer with high TAM numbers responded better to gemcitabine. In vitro experiments demonstrated that macrophages treated with gemcitabine were reprogrammed toward an M1 phenotype with the upregulation of cytotoxicity-related genes (32). Also, in preclinical models, taxanes showed the ability to reprogram M2-like TAMs into M1-like macrophages in a TLR4-dependent manner (33); likewise, in patients with metastatic gastric tumors treated with a novel taxane (cabazitaxel), there was evidence that a high macrophage signature was associated with improved survival (34). In colorectal cancer, high TAM density in tumors was a good marker to identify stage III patients responding better to 5-fluorouracil adjuvant therapy (35).

Other conventional therapeutic approaches may cause the ICD of cancer cells; some studies showed that low-dose radiotherapy, photodynamic therapy (PDT), and oncolytic viruses induced the release of DAMPs, tumor antigens, and immunostimulatory signals, eventually leading to a switch of TAMs into M1 antitumor effectors (36, 37).

The eponymous function of macrophages, their phagocytic activity, can be exploited to eliminate malignant cells. This phagocytic activity of macrophages is inhibited when the molecule, signal regulatory protein α (SIRPα), mainly expressed by myeloid cells, is engaged by the transmembrane protein, CD47, expressed by tumor cells. It has been demonstrated that the pharmacologic inhibition of the SIRPα/CD47 axis, with blocking antibodies, restores phagocytosis by macrophages and the killing of tumor cells. These treatments induced tumor regression in various preclinical cancer models and are now under evaluation in clinical trials (Table 1; ref. 38). Promising results were obtained for anti-CD47 mAbs in patients with lymphoma, in combination with anti-CD20 mAbs (39), and other clinical trials are testing their combination with anti–PD-1 therapy in different solid tumors. Blockade of the antiphagocytic pathway was generally well-tolerated in clinical studies, but one concern is that CD47 is also expressed by normal cells, which can sequester the therapeutic anti-CD47 antibodies and reduce treatment response. Interestingly, other antiphagocytic molecules have been recently discovered, such as beta 2-microglobulin (MHC class I complex) binding to the inhibitory receptor LILRB1 (40) and CD24 binding to SIGLEC-10 on macrophages (41). This redundancy of inhibitory molecules may explain why some patients do not benefit from these treatments.

The activating receptor CD40 on the macrophage surface has been targeted with agonistic anti-CD40 mAbs, which mimic the ligand CD40L, expressed by activated T cells. In preclinical studies, anti-CD40 mAbs were effective in reeducating immunosuppressive TAMs into M1-like macrophages, reestablishing immune surveillance. Mechanistically, agonist CD40 mAbs stimulated TAMs to produce the chemokine CCL5, which increased the influx of CD4⁺ T cells into the TME (42, 43). This preclinical evidence stimulated clinical studies of numerous anti-CD40 agonist mAbs in combination with checkpoint immunotherapy, chemotherapy, or targeted therapies in patients with advanced solid tumors.

Toll-like receptors (TLRs), particularly abundant on innate immunity cells, recognize the pathogen-associated molecular patterns of microbes and signal the cell to activate an immunostimulatory response (44, 45). BCG, stimulating TLR2 and TLR4, is the first TLR-stimulating agent approved by the FDA, and it is still used to treat bladder cancer (28, 46). The receptors, TLR3, TLR7, TRL8, and TLR9, are localized in endosomal compartments and serve as nucleic acid sensors, activating NF-κB and triggering the secretion of immunostimulatory cytokines, including the antitumor cytokine, IFN type I. The potential of TLR ligands to activate and reprogram TAMs has been investigated in preclinical cancer models, showing effective antitumor immune responses (47, 48). Among TLR synthetic agonists, only imiquimod (TLR7) has been approved by the FDA for the topical treatment of squamous and basal cell carcinoma (49). As the systemic administration of TLR agonists may be burdened with toxicity, they have been administered locally or intratumorally in accessible lesions, showing a satisfactory efficacy/toxicity balance. Topical administration or local injection of TLR agonists is under evaluation in phase I–III trials for different tumor types (refs. 50, 51; Table 1). For example, poly(I:C) nanocomplexed with polyethyleneimine (BO-112) is being evaluated in clinical trials after demonstrating significant antitumor responses in several mouse cancer models (52). Its analog, poly-ICLC, complexed with carboxymethylcellulose (Hiltonol), is being investigated in clinical trials, alone or in combination with anti-PD-1 therapy (53). The TLR9 agonists, DV281 and SD-101, are in early-phase studies in patients with advanced non–small cell lung cancer and with metastatic melanoma (54–56). Notably, the intratumoral injection of TLR agonists has shown not only excellent local antitumoral efficacy, but also a reduction in noninjected distant lesions, indicating the systemic activation of the immune system (57). With the rationale to boost immunity and “warm” immunologically cold tumors, TLR agonists are under evaluation in combination with checkpoint inhibitors or adoptive T-cell therapy (47, 52, 58, 59). Other clinical trials have evaluated TLR agonists as adjuvants, combined with tumor antigens, for cancer vaccination purposes (60, 61).

Tumor-induced emergency myelopoiesis leads to the expansion of immunosuppressive TAMs and MDSCs (22).

This emergency myelopoiesis is positively regulated by the transcription factor, C/EBPβ, and by the nuclear receptor, RORC1 (22, 62); thus, to downmodulate immunosuppressive myeloid cells, several specific inhibitors of RORγ are currently being characterized in preclinical models.

Another key immunosuppressive pathway in myeloid cells is PI3Kγ, which negatively regulates NF-κB activation. The pharmacologic inhibition of PI3Kγ has shown the ability to revert immunosuppression, reprogram TAMs, and increase CD8⁺ T-cell recruitment into tumors, resulting in tumor growth inhibition (63). PI3Kγ inhibitors have been combined with vasculature-disrupting agents, other kinase inhibitors, and checkpoint inhibitors and have achieved better results (12, 64, 65). Despite hundreds of ongoing clinical trials using pan-PI3K inhibitors, only a few early-phase studies are employing specific inhibitors to the myeloid-γ isoform in patients with cancer (Table 1).

Macrophage polarization is regulated by genetic, epigenetic, and metabolic programs, which have been pharmacologically intervened for therapeutic purposes.

The genetic reprogramming of TAMs has been successfully demonstrated in different murine models using nanoparticles loaded with mRNAs encoding IFN regulatory factor 5 and its activating kinase, IKK-β. These mRNAs were complexed with cationic poly(β-amino-esters), and the formed nanoparticles coated with PGA-di-mannose to target and reprogram TAMs showed antitumoral efficacy (66).

The epigenetic modulation of TAMs has been demonstrated with histone deacetylase (HDAC) inhibitors. The selective class IIa inhibitor, TMP195, stimulated the macrophage-mediated phagocytosis of tumor cells and TAM reprogramming into antitumor effectors in a mouse mammary model, showing better outcomes in combination with chemotherapy or anti-PD-1 (67). Interestingly, low-dose epigenetic compounds in combination with chemotherapy decreased metastatic spread via the inhibition of myeloid cell recruitment at premetastatic sites (68). Thus, pan-HDAC inhibitors are currently being tested in the clinic in combination with chemotherapy or immunotherapy, and their effect on macrophage polarization is under study.

The altered metabolism in the tumor tissue has important implications for macrophages. For instance, the product lactate, secreted by hypoxic cancer cells and fatty acid oxidation favors the immunosuppressive polarization of macrophages (69, 70), and the targeting of tumor-derived metabolites, such as adenosine, glutamine, and lactate, has been investigated in preclinical models for their effects on tumors and macrophages. Glutamine metabolism is involved in important biological functions, such as nucleotide synthesis, amino acid production, redox balance, glycosylation, and extracellular matrix production. A small prodrug molecule (6-diazo-5-oxo-L-norleucine) was used to block glutamine metabolism in a mouse breast cancer model, showing inhibited tumor growth and metastases not only by acting on tumor cells, but also by enhancing macrophage activation and reducing immunosuppression (71). Similarly, glufosinate, an inhibitor of glutamine synthetase, reduced metastasis formation in highly metastatic mouse models, and the antitumoral effect was related to the downmodulation of angiogenesis and immunosuppression, with the reprogramming of TAMs into antitumor effectors (72).

Tumor-derived lactate promotes the M2 polarization of macrophages via the activation of the ERK/STAT3 signaling pathway or the sensor protein, Gpr132 (73). Accordingly, the pharmacologic inhibition of the ERK/STAT3 axis, with selumetinib or stattic, or the inhibition of Gpr132 reduced lactate-induced M2 polarization and showed significant antitumor effects in preclinical murine models (74, 75). Metformin, an antidiabetic agent displaying multiple mechanisms of action on important metabolic pathways, has shown the ability to increase macrophage-mediated phagocytosis of tumor cells in vitro and inhibition of STAT3 activation in vivo (76). Metformin has been investigated in several clinical trials in oncology patients. For instance, a recent trial reported that low-dose metformin reprogrammed the immune environment of esophageal carcinoma with a decreased number of TAMs and increased infiltration of cytotoxic T lymphocytes.

Retinoic acid, with multiple roles in cell biology, has recently shown the ability to favor immunosuppressive differentiation of macrophages in the context of soft-tissue sarcoma, and the genetic deletion or pharmacologic inhibition of retinoic acid resulted in enhanced antitumor immune responses (77). As many human cancers have high levels of retinoic acid–producing enzymes, the targeting of retinoic acid in tumors could be a strategy to prevent monocytes from differentiating into TAMs.

Similarly, indoleamine 2,3 dioxygenase (IDO), which converts tryptophan into kynurenines, is a well-known immunosuppressive mediator produced by TAMs that inhibits cytotoxic T-cell antitumor functions and favors the expansion of regulatory T cells (78). Although controverted results have been observed, several IDO inhibitors are being tested in clinical trials (Table 1).

Overall, these findings support the concept that metabolically rewiring the TME may have significant positive effects on TAM reeducation, although the potential for therapeutic interventions in patients with cancer remains to be investigated.

Long before the complex macrophage world was fully understood, attempts were made to inhibit tumor growth and metastasis with the transfer of macrophages in mouse models. Various approaches were used: activated macrophages, macrophages as vehicles for drug delivery (including drugs formulated with nanoparticles), and genetically modified macrophages, reviewed in (79). Few human clinical studies were performed, substantially giving modest results, if any. Reasons for failure include the lack of knowledge about the biodistribution and survival of the transferred cells or about the influence of the TME on incoming macrophages. Nevertheless, these pioneering studies paved the way for more recent approaches, such as the transduction of a chimeric antigen receptor (CAR) into macrophages. Inspired by the success of CAR-T cells genetically engineered to express antigen-specific receptors, Klichinsky and colleagues succeeded in transducing anti-HER2 into primary human macrophages [CAR-macrophages (CAR-M)] using a modified replication-incompetent adenovirus. They demonstrated that the adoptive transfer of CAR-M with sustained expression of the transgene efficiently reduced tumor growth in immunodeficient mice with HER2-positive human tumors (80). The success of this technique opens new avenues in the use of engineered macrophages with the aim to exploit their potential antitumor cytotoxicity.

Therapeutic manipulation of TAMs has been investigated in the last years using different therapeutic approaches, such as mAbs, TLR agonists, RNA-based strategies, genome editing, epigenetic, and metabolic modulatory drugs. So far, experimental results have indicated that the reprogramming of TAMs is likely to be more successful than their indiscriminate elimination. Furthermore, preclinical and initial clinical studies have shown that rather than monotherapies, combination therapies that tackle different mechanisms are more efficient for the reeducation of TAMs in vivo.

In this scenario, three main challenges can be recognized: (i) the identification of receptors for more specific targeting of TAMs, (ii) improvement in drug delivery to reach TAMs in the TME, and (iii) the achievement of an optimal and long-lasting reprogramming of TAMs under safe conditions. The specific targeting of TAMs is impractical nowadays, as no receptors are known to be expressed exclusively by TAMs. Further basic research and knowledge of TAM biology are needed to address this challenge. Improved drug delivery might be achieved with the direct injection of accessible tumor lesions or with nanotechnology approaches to achieve better drug biodistribution and penetration at tissular and cellular levels. Effective and long-lasting reprogramming must face the problem that the immunosuppressive TME continuously affects TAMs and their progenitor cells. New discoveries from basic research are incessant and exciting; we expect that forthcoming macrophage-directed therapies will provide better outcomes and improve patients' therapeutic response to oncology treatments.

No potential conflicts of interest were disclosed.

Research activity of P. Allavena, C. Anfray, and F. Torres Andón was supported by the EU Project EURONANOMEDIII (2-INTRATARGET). F. Torres Andón was recipient of a grant by the AECC (Asociación Española Contra el Cáncer, Spain). A. Ummarino is recipient of a fellowship from the Italian Association for Cancer Research (AIRC).