Cancer Immunotherapies Based on Genetically Engineered Macrophages

In the past 20 years, the development of immunotherapies has contributed substantially to progress made in the treatment of malignancies. These treatments work by stimulating, modifying, or inhibiting key immune regulators and aim to rewire both innate and adaptive antitumor immune responses (1). Cell-based immunotherapies rely on the use of modified immune cells to treat cancer. T cells and natural killer (NK) cells were the first immune-cell populations to be used in this context because their primary function is to directly kill infected or tumor cells. T cells that have been genetically modified to express chimeric receptors that combine antigen-binding and T-cell activation activities in a single receptor are known as chimeric antigen receptor (CAR) T cells. Adoptive transfer of CAR T cells has shown considerable promise in battling hematologic malignancies (2–4).

CAR T-cell treatments, however, have so far lacked efficacy in the treatment of solid tumors. Several factors are likely responsible for these disappointing results. First, identifying tumor-specific antigens remains difficult for solid tumors, raising concerns about potential off-target effects. Second, before reaching the cancer cells within the solid tumor tissue, the CAR T cells may encounter physical barriers in the form of tumor-associated macrophages (TAM) and cancer-associated fibroblasts, which produce vast amounts of extracellular matrix (ECM). Third, CAR T cells may not be able to successfully invade the solid tumor microenvironment (TME) due to a lack of metabolic resources or signals provided by TME cell components (5). Finally, due to persistent antigen stimulation arising from tumor cells, CAR T cells can become “exhausted” or dysfunctional, losing their effector function and failing to evolve into effector memory T cells (6). Alternative approaches are currently being developed using new versions of CARs; using NK cells, mucosal-associated invariant T (MAIT) cells, or γδ T cells (7, 8); or using macrophages. The intrinsic features of macrophages make them an ideal candidate to overcome the limitations of CAR T cells.

Macrophages are innate immune cells with a wide range of functions that vary depending on their tissue localization (9). They become activated when receptors they express that recognize danger- or pathogen-associated molecular patterns, such as Toll-like receptors (TLR) or Nod-like receptors (NLR), are engaged by ligands. This ligand recognition results in the production of cytokines like TNFα, IL12, and IL1β (10). These cytokines can have a marked impact on cells in the TME, including promoting antitumor Th1 and Th17 responses (11, 12).

Macrophages are also antigen-presenting cells that can locally stimulate T cells. Although less potent than dendritic cells (DC), macrophages can prime T cells by presenting antigens (13–15). In addition, macrophages can perform cross presentation of exogenous antigens, again with less efficacy than DCs (16). This activity depends on the activation state of the macrophages and on their polarization. Anti-inflammatory macrophages are relatively inefficient at antigen presentation and T-cell priming, as compared with activated proinflammatory macrophages (14). Strategies to induce adaptive antitumor T-cell responses via macrophage manipulation should, therefore, include consideration of their polarization status and may efficiently overcome the limitations of CAR T cells and achieve durable antitumor immunity.

The TME is a complex, heterogeneous mix of cell populations that interact with each other and with tumor cells. TAMs are a key cellular component of the TME in a variety of cancers (17, 18). Use of high-dimensional technologies, including cytometry by time of flight (CyTOF) and single-cell RNA sequencing (scRNAseq), has revealed an important heterogeneity of phagocyte subpopulations present in solid tumors across various tissues. Nevertheless, recent integration of large scRNAseq datasets from different tissues has revealed that some macrophage subsets exhibiting immunosuppressive activities are conserved across neoplasia such as TREM2⁺, HES1⁺, or IL4I1⁺PD-L1⁺IDO1⁺ macrophages (19).

Macrophages can produce and secrete metalloproteases that are able to dramatically modify the ECM within the tumor mass thereby altering the architecture of the tumor tissue (20–22). TAMs can exert forces on collagen fibers and induce their reorganization within the TME, thus favoring tumor evasion (22). Macrophages are also endowed with antitumor capabilities, such as the ability to phagocytose entire tumor cells or undertake antibody-dependent cell phagocytosis (ADCP; ref. 12). One study revealed that in healthy and malignant breast tissues, a discrete population of FOLR2⁺ macrophages may contribute to antitumor immunity (23). These tissue-resident macrophages are located in the tumor stroma near vessels and interact with CD8⁺ T cells. Importantly, the presence of FOLR2⁺ macrophages correlates with T-cell infiltration and favorable clinical outcomes in patients with breast cancer (23).

The prevailing consensus is that tumor-derived cytokines direct myeloid-cell recruitment, and then the TME influences macrophage polarization (24). For instance, it has been shown that human and mouse early lung cancers have similar myeloid populations, including tissue-resident embryonic macrophages that are recruited at the tumor bed (25). Tissue-resident macrophages are redistributed at the tumor's periphery later in its growth, whereas monocyte-derived macrophages are drawn to the tumor bed, where they accumulate (25). As a result, macrophages can represent a significant portion of the tumor mass (26, 27), for instance, up to 50% in some breast tumors (17). They develop into immunosuppressive macrophages that prevent antitumor CD8⁺ T cells from infiltrating the tumor and attract or induce regulatory T cells (25, 28–30). When cocultured with lung tumor cells, alveolar macrophages drive a transcriptional program in tumor cells typical of an epithelial–mesenchymal transition, potentially encouraging metastasis (25).

TAMs secrete growth factors like VEGF (31) and TGFβ (32), which promote tumor growth and invasive behavior (33). They are generally associated with poor prognosis (34–37), although recent studies indicate that their impact on prognosis can vary, depending on their localization (38) and polarization (39). TAMs have therefore been considered for a long time as targets of interest for anticancer therapies. Several strategies aimed at impeding their development (by CSF1R blockade; refs. 40–42), at blocking their recruitment by tumor cells (with CCR2/5 inhibitors; refs. 43, 44), as well as at lowering their immunosuppressive activity (with CD40 agonists; refs. 45, 46) have been developed. These strategies, when implemented individually, have tended to have limited effects in phase I/Ib clinical trials. Few phase II trials were implemented for these strategies as monotherapies, but they are currently being evaluated using combinatorial approaches (47–50). Boosting macrophage capacity to phagocytose tumor cells by blocking the CD47 signal on tumor cells with antibodies has shown more encouraging results in clinical trials (51, 52)

In this review, we discuss attempts to harness the antitumor potential of macrophages through genetic engineering. We cover the remarkable diversity of approaches that have been used to genetically modify macrophages to favor phagocytosis, induce signaling and polarization of macrophages, and stimulate antitumor immunity. Then, we examine the relevance of these approaches in the context of immunotherapies against tumors. Finally, we discuss the challenges and obstacles that remain to be solved to translate these therapeutic strategies to efficient treatments for solid tumors.

Macrophages are absent from the bloodstream and cannot be expanded in vitro because they are terminally differentiated cells. Human macrophages can currently be produced in three ways from their precursors. The simplest method to obtain macrophage precursors is to magnetically separate the 10% of monocytes contained in peripheral blood mononuclear cells (PBMC; ref. 12). Another approach is to isolate hematopoietic stem cells (HSC) from the bone marrow, where they represent around 2% of all bone marrow cells (53). Finally, all types of macrophages can be potentially differentiated from induced pluripotent stem (iPS) cells by a cocktail of growth factors (54). iPS cells are generated in vitro from somatic cells by inducing the expression of 4 transcription factors. Importantly, unlike monocytes, HSCs and iPS cells can divide in vitro and thus be expanded. Adenoviral or lentiviral transduction, as well as gene editing with CRISPR–Cas9 technology, can all be used to genetically modify the three types of precursors.

Macrophages have been engineered by viral-independent techniques, that is, by classical or reverse transfection (55, 56), reaching usually rather low rates. Primary human macrophages are partially refractory to lentiviral transduction due to the restriction factor SAM domain and HD domain-containing protein 1 (SAMHD1). High transduction rates can be achieved by using HIV-2 or SIV viral protein X (VPX) to deplete SAMHD1 in macrophages (57, 58). Other studies used group B adenovirus to create replication-incompetent adenoviral vectors that can infect primary macrophages via their CD46 receptors (59). Adenovirus exposure causes innate sensing via various DNA sensors (60), resulting in macrophage polarization toward a pro-inflammatory phenotype potentially beneficial for the antitumor response (61, 62).

Successful gene editing in primary macrophages has been achieved using a modified recombinant Cas9 for better association with cationic arginine-coated nanoparticles (63). More recently, nanoparticles encapsulating Cas9 recombinant protein bound to guide RNA have been shown to enter and efficiently edit the macrophage genome (64, 65). Moreover, strategies with Cas9 delivered by ribonucleoproteins have proven effective to edit murine macrophages in vivo (66) and human primary monocytes in vitro (67). The rapid progress paves the way for precise and multiple gene editing of primary immune cells for therapeutic purposes.

One of the biggest challenges for any cell-based therapy is the necessity to provide autologous patient cells that have been manipulated in vitro. This requires complex procedures for collecting, treating, and shipping samples, which have been streamlined in the case of CAR T cells by a few companies. However, this highly affects the cost of the treatments. To reduce these costs and procedures, there are many efforts focused on producing “off-the-shelf CAR T cells” that could be infused to groups of patients affected by the same type of malignant pathology, independently of their HLA type. This strategy depends on preventing the development of graft-versus-host responses by the injected CAR T cells, as well as allowing their long-term persistence so that they can perform the function conferred on them by genetic manipulation (gene editing and/or transgene expression).

The aforementioned issues do not apply to macrophages, because they do not proliferate, cannot react to HLA molecules and are less efficient than cytotoxic T cells at direct target cell killing. Although monocytes are easily purified from PBMC, their use requires processing for each patient, like for CAR T cells. Bone marrow samples are more difficult to obtain than blood and their relatively low HSC content (around 2%; ref. 53) prevents their use on a large scale to treat patients. iPS cells remain very promising because they can proliferate in vitro to some extent; be gene-edited, which opens the possibility to render them immunologically silent for the recipient; and be induced to differentiate into different kinds of tissues macrophages (54). However, optimization of several steps will be required before it will be possible to perform large-scale production of “off-the-shelf engineered macrophages” from iPS cells that could be injected to groups of patients sharing the same malignancy. Thus, currently, monocytes remain the principal source of myeloid cells used for therapeutic purposes.

Macrophages are known for their important ability to phagocytose cellular debris, pathogens, apoptotic cells (in which case the process is called efferocytosis; ref. 68), and even intact living cells like tumor cells (12). Enhancing their ability to phagocytose cancer cells is thus a strategy to strengthen their antitumor capabilities (Table 1).

An attractive strategy is to prevent inhibition of phagocytosis by targeting the CD47–SIRPα pathway. When CD47, which is expressed by many cells, including some tumor cells, binds to signal regulatory protein α (SIRPα) on macrophages, the CD47–SIRPα pathway sends a "don't eat me” signal’ to macrophages via the SIRPα, preventing phagocytosis (ref. 69; Fig. 1). Several CD47-specific antibodies and decoy receptor fusion proteins are currently being tested in preclinical models in vivo (70, 71) and in clinical trials for solid tumors (colorectal cancer, melanoma, breast cancer and hepatocellular carcinoma) as well as hematological cancers and have shown encouraging early results (51, 52, 72). However, CD47 is widely expressed on normal cells, including erythrocytes, platelets, and neutrophils. Accordingly, these treatments induce the depletion of these cell populations in patients (51, 73, 74). Of note, CD47 levels diminish only when erythrocytes become senescent, provoking their rapid clearance by splenic macrophages. Rather than blocking CD47, several attempts have been developed to block its counterreceptor SIRPα with antibodies or with small molecules (75–78). As an alternative to antibody treatment, SIRPα was gene-edited in Raw264.7 macrophages, and this enhanced their capacity to phagocytose the U2OS osteosarcoma cell line in vitro (63). Similar gene editing in primary human macrophages expressing an anti-HER2 CAR also enhanced their ability to phagocytose and kill HER2⁺CD47⁺ SKOV3 tumor cells (79). In vivo validation of these results is pending. These approaches are attractive as they may confer better tumor specificities and thus avoid adverse effects observed with CD47-specific antibodies.

CARs are originally chimeric receptors that combine antigen-binding and T-cell activation activities in a single receptor. Several generations and variations of CARs have been produced for T cells (3). CARs for macrophages are more recent, less numerous and designed to harness innate immunity via cytosolic domains of one or more innate receptors, such as the ones of the FcγR, combined with specificity using an scFv specific for a tumor antigen, mostly to promote phagocytosis of tumor cells (Fig. 2). Various combinations based on cytosolic domains of innate receptors are currently being tested preclinically for their capacity to induce strong innate immunity response upon contact with tumor antigen, with the goal being that this will counter the immunosuppressive TME.

In early work, adenoviral transduction of primary human monocytes with a CAR containing an scFv specific for carcinoembryonic antigen and the cytosolic domain of CD64 (FcγRI), reduced tumor growth in vitro and in vivo, but the mechanism of cytotoxicity involved was not investigated (80). Later, CARs carrying an scFv specific for CD19 or CD22 and intracellular domains of Megf10 or FcγR were created to better harness macrophage phagocytic potential [Fig. 2 and ref. (81)]. The expression of the CAR construct provided directed ingestion of CD19- or CD22-coated beads and killing of CD19-expressing tumor cells by a murine macrophage cell line. Although there was some whole-cell phagocytosis documented, tumor cell death was primarily caused by trogocytosis, which corresponds to the internalization of a piece of the target cell membrane by another cell, in this case a macrophage (82). A further increase in tumor cell phagocytosis by macrophages was obtained by adding to the CAR the 38-amino-acid long portion of CD19 responsible for recruiting the p85 subunit, which activates the PI3K pathway (81). However, this study did not evaluate the ability of CAR macrophages to control tumor growth in mouse models.

Yong and colleagues (83) developed a transgenic mouse strain in which the pan-hematopoietic Vav promoter drives the expression of an anti-HER2 CAR gene with CD28 and CD3ζ as intracellular domains (vav-CAR mice). As a result, multiple subsets of immune cells express the CAR in these mice. Macrophages isolated from these mice expressing the anti-HER2 CAR have increased phagocytic activity and release more IL6 and TNFα when cocultured with HER2-expressing tumor cells than control macrophages. Furthermore, when challenged with HER2-expressing tumors, vav-CAR mice outlive wild-type mice. Treatment with clodrolip, which depletes macrophages in vivo, reduces survival of vav-CAR mice bearing HER2-expressing tumors, implying that CAR macrophages play an important role in controlling tumor growth in vivo (83).

Human primary macrophages have been transduced directly with an anti-HER2 first-generation CAR containing the CD3ζ cytosolic domain (84). In vitro, these CAR macrophages specifically phagocytosed the human ovarian cancer SKOV3 cell line, which expresses HER2. In vivo, the CAR macrophages inhibited SKOV3 tumor growth, increased mouse survival, and reduced lung metastatic burden upon intravenous or intraperitoneal injection in tumor-bearing NSG (NOD/SCID/γchain^null) mice. Independently of the gene transduced, adenoviral vector transduction of macrophages enhanced IFN-associated gene expression, activated critical components of the antigen-presentation pathway, and polarized the macrophages toward an inflammatory phenotype. Furthermore, the CAR macrophages retained their inflammatory phenotype for at least 5 days after reaching the tumor. They even partially remodeled the TME by triggering increased proinflammatory gene expression in bystander immune cells (84). This landmark study was the first to unequivocally demonstrate that CAR macrophages can phagocytose tumor cells in vivo, become pro-inflammatory, and control tumor growth in preclinical models.

Zhang and colleagues (85) used a second-generation anti-CD19 and an anti-mesothelin CAR (containing the 4–1BB costimulatory domain and the CD3ζ chain) to transduce iPS cells. According to surface marker expression and transcriptome analysis, macrophages produced from these engineered iPS cells were similar to monocyte-derived macrophages. Coculturing CAR macrophages with CD19- or mesothelin-expressing cells (ovarian or pancreatic tumor cells, respectively) increased phagocytosis and expression of pro-inflammatory cytokines (IL6, IL1β, and TNFα) in vitro. Furthermore, CAR macrophages became polarized toward an inflammatory phenotype (85). However, in vivo, the CAR macrophages only had a minor effect on tumor growth, indicating that further improvements would be needed to achieve full antitumor potential.

To generate several CAR-expressing myeloid populations, HSC have been transduced with an anti-CD19 second-generation CAR (with a CD28 co-stimulatory and a CD3ζ domain) and differentiated into mature myeloid cells consisting of neutrophils, monocytes, and macrophages. In vitro, myeloid cells demonstrated specific killing activity against CD19-expressing human tumor cell lines, as measured by cytotoxicity assays (86). Furthermore, when challenged with Raji tumor cells, which express CD19, humanized NSG mice engrafted with HSC expressing the anti-CD19 CAR outlived mice engrafted with unmodified HSC (86).

Although promising, genetic changes in HSC can result in leukemia or lymphoma, which means that careful characterization of potential off-target effects is required before infusion to patients. To increase safety, HSC have been transduced with an anti-CD19 CAR (with a CD28 costimulatory and a CD3ζ domain) and a herpes simplex virus hyperactive sr39 mutant (HSVsr39TK), which induces cell death when ganciclovir (GCV) is provided (87). Such engineered HSC increased mouse survival when injected into NSG mice carrying CD19⁺ tumor cells. GCV treatment ablated the injected engineered HSC (87). HSVsr39TK can be substituted with a shortened version of epidermal growth factor, leading to similar results (88). The specific in vitro cytotoxicity in these two experiments was due to myeloid cells, and the greater survival in vivo was due to a broad immunological response. However, the importance of CAR macrophages in these settings in comparison to other myeloid cells, such as DC, has yet to be examined.

Overall, several cytosolic domains can activate phagocytosis in CAR-expressing myeloid cells upon engagement of the antigen for which the CAR is specific. This includes domains known to belong to the phagocytic pathway such as activation domains of FcγRs. ADCP is mediated by FcγRs and known to favor antigen cross-presentation by myeloid cells (89). Interestingly, probably due to their origin, CARs have also been designed for usage in macrophages with activation domains from receptors that are absent from myeloid cells. In particular, CD3ζ appears to be efficient in various settings for this purpose (83–85, 87, 88), suggesting that upon engagement and multimerization of the CAR that carries the CD3ζ domain, initial events of signal transduction such as phosphorylation of CD3ζ by ZAP70, a T-cell specific kinase, can be achieved by other kinases in myeloid cells such as Syk (90). In vitro evidence of cross-presentation of the NY-ESO1 tumor antigen by CAR macrophages to T cells has already been obtained (84). It is therefore probable that tumor cell phagocytosis by CAR myeloid cells would result in vivo in tumor antigen cross-presentation, and stimulation of antitumor T-cell immunity.

The various CARs used in myeloid cells have been designed to harness the phagocytosis function of the cells. However, phagocytosis by myeloid cells can result in profound modifications of their phenotype, including the induction of immune tolerance. Phagocytosis of dying cells by macrophages, a process called efferocytosis, is associated with secretion of immunosuppressive cytokines and resolution of inflammation (91). In addition, LC3-associated phagocytosis (LAP) promotes a tolerogenic phenotype in myeloid cells. Impairment of LAP in macrophages upon efferocytosis leads to activation of the STING pathway and production of type I IFN, acquisition of a proinflammatory phenotype by TAMs and control of tumor growth via tumor-infiltrating T cells (92). ADCP activation in macrophages confers a protumoral phenotype mediated by upregulation of programmed death-ligand 1 (PD-L1) and indoleamine 2,3-dioxygenase, leading to the inhibition of NK and T cells (93). These ADCP-mediated upregulations are dependent on the DNA sensor AIM2, which gets activated by tumor DNA upon ADCP and cleaves cGAS, preventing type I IFN responses (93).

It will be critical in a therapeutic setting to precisely characterize the phenotype of engineered macrophages before and after phagocytosis of dying or alive tumor cells, and to assess their capacity to reverse the immunosuppressive phenotype that can be induced by phagocytosis.

Macrophages are plastic cells that can be polarized by their surroundings. Their polarization state ranges from a proinflammatory, antitumor phenotype to an anti-inflammatory, protumor phenotype, with a range of variations between these two states (94). Depending on their polarization state, macrophages secrete different cytokines and promote either antitumor Th1 and Th17 or protumor Th2 T-cell responses (12). In the TME, macrophages are usually polarized toward an anti-inflammatory phenotype, favoring tumor growth and invasive behavior (33). It is possible to reverse the immunosuppressive phenotype of macrophages within the TME by mimicking activation of pathogen recognition receptors (PRR) such as Toll-like receptors (TLR; Table 1). Velazquez and colleagues created a CAR that included an anti-mesothelin scFv and the TLR4 cytosolic Toll/IL1 receptor (TIR) domain (ref. 95; Fig. 2). Mesothelin is overexpressed in a variety of human tumors, including lung and pancreatic adenocarcinomas, as well as ovarian cancers (96). In vitro, coculturing CAR-expressing macrophages with mesothelin-expressing target cells resulted in TNFα production and targeted cell death. When CAR macrophages were administered intravenously into NSG mice with mesothelin-expressing breast cancers, they reduced tumor burden (95). It remains unclear how the TIR domain contributes to the antitumor action of these CAR macrophages.

In an antigen-independent approach, a chimeric receptor bearing the cytosolic component of TLR4 linked to an F36V domain that can dimerize upon binding to a cell permeable chemical inducer of dimerization (CID) was created (97). TLRs dimerize when they meet their ligands, activating the NF-κβ and MAPK pathways. The addition of CID to Raw264.7 macrophages transduced with the F36V domain–containing chimeric receptor stimulated IL6 production via the NF-κβ and ERK pathways, as expected.

Harnessing PRR signaling pathways in CAR macrophages represents an interesting strategy but will require careful characterization of macrophage phenotype after contact with tumor cells in vitro and in vivo, to ensure that an antitumor phenotype is retained. Unlike T cells, macrophages have a finite lifespan and proliferative capability (98). It is therefore essential to rewire them to relieve local immunosuppression and activate an adaptive response.

Several cytokines have pleiotropic proinflammatory properties as they can activate different cell types and enhance immune responses. Hence, they have been used to modify the TME and evaluated for their antitumor effects when expressed by myeloid cells (Fig. 1 and Table 1).

Because of their potent impact on the immune response, type I IFNα and IFNβ have been used in different tumor models (Table 1). The transduction of HSC with an IFN transgene under the control of TIE2 regulatory elements allowed selective activity during myeloid differentiation (99). NSG mice carrying MDA MB 231 breast tumors and engrafted with such transduced HSC exhibited increased expression of IFN-inducible genes in the tumor. HSC with an IFN transgene also induced tumor regression in an oncogene-driven MMTV-PyMT animal model as well as in mice injected with metastatic cells from this model. Regression was associated with higher infiltrations of CD4⁺ and CD8⁺ T cells in the tumor and a lower metastatic burden (99).

Type I IFN also has been engineered into macrophages generated from iPS cells (100), resulting in IFNα- and IFNβ-transduced macrophages (IFNα-MΦ and IFNβ-MΦ), which exhibited antitumor activity in vitro. However, only IFNβ-MΦ significantly reduced tumor burden in SCID mouse xenograft models of gastric and pancreatic cancer. Both IFNα-MΦ and IFNβ-MΦ reduced tumor growth in vitro and in vivo when tested against a metastatic melanoma (101). Interestingly, IFNα-MΦ reduced the presence of anti-inflammatory macrophages in the tumor infiltrate, suggesting that they influenced the composition of the TME.

Because of its capacity to efficiently stimulate naive T cells and thus initiate an immune response, IL12, an inflammatory cytokine produced by DC and macrophages has been used to arm macrophages in several studies (102–104). Murine macrophages transduced to express a murine Il12 transgene (mIL12-MΦ) were tested in a mouse model of metastatic prostate cancer by intratumoral administration (102). mIL12-MΦ reduced primary tumor growth, inhibited spontaneous metastasis, induced stronger tumor infiltration of T lymphocytes, as well as increased NK cell and cytotoxic T-cell activity among splenocytes. Similarly, intratumoral injection of mIL12-MΦ into mice carrying subcutaneous B16.F10 tumors induced some delay in tumor growth (104).

In another study, murine bone marrow–derived myeloid cells expressing an Il12 transgene were generated by lentiviral transduction (103) and injected intravenously in a mouse orthotopic tumor model of rhabdomyosarcoma that spontaneously spreads to the lungs. The injection caused a surge in IL12 production, followed by IFNγ, both in the original tumor and in the lung premetastatic niche. This treatment also reshaped the lung premetastatic microenvironment at the cellular level, increasing the number of T cells, NK cells and activated DCs. At the genetic level, it increased the expression of genes linked with adaptive immunity, antigen processing and presentation, and decreased the expression of immunosuppressive genes, reversing the core immunosuppression program in the premetastic lung (103). Overall, IL12 myeloid cell injection enhanced survival, reduced metastatic burden, and inhibited tumor growth. Similar results were obtained in an animal model of pancreatic cancer metastasis to the liver (103) and in a mouse model of subcutaneous melanoma (104).

Regarding humans, only one study so far has used human monocyte-derived macrophages expressing IL12 (hIL12-MΦ), which when cocultured with T cells in vitro, promoted T-cell activation, as seen by increased CD69 expression and IFNγ production while maintaining a proinflammatory polarization (104). When injected intravenously into NSG mice bearing a subcutaneous U87 tumor on the flank, one percent of hIL12-MΦ trafficked to the tumor flank where they remain for at least 7 days. hIL12-MΦ also induced tumor cell death and IFNγ production on tumor slices of advanced human gastrointestinal malignancies ex vivo (104).

Taken together, these studies suggest that IL12 production by myeloid cells in TME is sufficient to polarize macrophages toward an inflammatory phenotype despite the immunosuppressive environment, resulting in better immune control of tumor growth. Importantly, despite its potency, IL12 local distribution must be strictly managed because it has been shown that IL12 can rapidly produce severe toxicities in patients (105).

Although cytokines have the potential to reprogram the TME, they have side effects, prompting the development of artificial molecules such as bispecific T-cell engagers (BiTE) to bypass the immunosuppressive TME. BiTEs are constructs that include two scFv, one of that is specific for a tumor-associated antigen and the other that can bind T cells. BiTEs promote T-cell–mediated tumor cell death by stabilizing the contacts between tumor cells and T cells (106).

A lentiviral vector coding for BiTEs targeting CD3 on T cells and EGFR variant III (EGFRvIII) as a tumor-associated antigen, with or without an IL12 transgene (BiTEs-MΦ and BiTEs/IL12-MΦ), was used to transduce human macrophages generated from monocytes (107). In vitro, the resulting macrophages released functional BiTEs that stimulated T-cell activation, proliferation, secretion of IFNγ and TNFα, and killing of tumor cells.

Addition of BiTEs-MΦ and BiTEs/IL12-MΦ increased tumor cell lysis in co-cultures of EGFRvIII-expressing U87 cells, with T cells or PBMCs. Only BiTEs/IL12-MΦ, however, triggered increased expression of several proinflammatory cytokines (IFNγ, IL-6, IL-8, and GM-CSF) and induced high expression of granzyme B in CD4⁺ T cells. When injected subcutaneously in NSG mice carrying U87 tumor cells, only BiTEs/IL-12-MΦ prevented the emergence of a detectable tumor for 36 days in vivo (107), but their influence on already formed tumors remains to be determined.

Instead of targeting tumor cells, macrophages can be designed to recognize the cells within the TME, obviating the need to identify specific tumor antigens. In one study, mouse immune cells that displayed an immunosuppressive phenotype after a contact with 4T1 tumor cells, were found to be CCR7⁺ (108). CAR constructs were thus developed to target CCR7⁺ cells using CCL19, a well-known CCR7 ligand (109), and the cytosolic domain of the myeloid–epithelial-reproductive tyrosine kinase (MerTK), which has been implicated in phagocytosis (refs. 108, 110; Fig. 2). In vitro, CCR7⁺ cells were specifically phagocytosed by murine Raw264.7 cells transduced with this chimeric receptor. Intravenous administration of transduced Raw264.7 cells to mice with subcutaneous 4T1 tumors increased median mouse survival, tumor infiltration by T cells, and serum pro-inflammatory cytokine levels (IL1β, TNFα, and IL6; ref. 108). However, MerTK has been also reported to induce an immunosuppressive phenotype in macrophages (111), meaning that there will be a need for careful characterization of these CAR macrophages.

The TME contains not only immune cells, but also fibroblasts, blood vessels, and the ECM, all of which aid in tumor growth and metastasis. Macrophages have been engineered to express NK4, a hepatocyte growth factor (HGF) internal fragment that antagonizes the HGF/c-met receptor axis (ref. 55; Fig. 1). When engaged, this pathway promotes angiogenic and metastatic characteristics in tumor cells. In vitro and in tumor-bearing mice, NK4-MΦ released a large amount of NK4 and demonstrated antitumor efficacy against the Meth-A⁺ sarcoma cell line (55).

The extensive ECM present in the TME may act as a physical barrier preventing access to tumor-specific cytotoxic T cells, contributing to immunosuppression. Macrophages can secrete members of the matrix metalloproteinase (MMP) family, which can disrupt the ECM. Zhang and colleagues (56) developed a chimeric receptor (CAR–CD147) including an anti-HER2 scFv and the cytosolic domain of CD147, a positive regulator of MMP production (Fig. 2). Raw264.7 macrophages expressing CAR–CD147 increased the production of various MMPs when cocultured with HER2-expressing cells. The CAR–CD147 macrophages successfully controlled tumor growth when administered into mice carrying HER2⁺ tumors. According to an ex vivo investigation on tumor spheroids, this control was associated with a partial disruption of the tumor ECM and increased T-cell infiltration in the tumor (56). However, because the ECM has a well-documented role in restricting tumor metastasis (111), an antitumor strategy relying on matrix modification should be considered with caution.

The use of genetically engineered autologous myeloid cells in anticancer immunotherapies has sparked growing interest in the past few years (Table 1). In patients with HER-2–overexpressing malignancies, a first phase I clinical trial using anti-HER2 CAR macrophages has started (112). Combinations with other immunotherapies, such as immune checkpoint inhibitors, are likely to be rapidly evaluated. Engineered macrophages could also be paired with oncolytic viruses to fight solid tumors, akin to previous experiments using CAR T cells (113).

Engineered myeloid cells still face important challenges before becoming adoptive cell treatment in the clinics. First, CAR macrophage development will require the validation of new tumor-specific antigens for targeting, and thus will indirectly benefit from the results obtained in the CAR T-cell field. Second, it will be crucial to determine whether the altered myeloid cells are able to reach and infiltrate the tumor site. There is room for improvement in this regard by additional genetic modifications, engineering, for instance, different chemokine receptors depending on the secretome of the tumor considered. The route of administration, the frequency of the treatment, and the number of macrophages delivered require careful benchmarking and optimization for each cancer type.

Third, because macrophages do not divide, the injected CAR macrophages will need to be sufficiently numerous and potent to tip the balance toward a proinflammatory TME and an efficient immune response in a therapeutic situation. New techniques to expand myeloid cells, or to increase their lifespan, will have to be implemented. Finally, the manufacturing of the myeloid cells will need to be streamlined to reduce the time and the costs of the treatments. Again, the tremendous work achieved in the CAR T-cell field will certainly speed up the development of myeloid cell–based therapies.

Deciphering the appropriate genetic changes to optimize myeloid antitumor responses, as well as designing combinations with other therapies most likely constitute the next challenges to get CAR macrophages to the clinic. The emergence of iPS cell–derived myeloid products raises many hopes for the development of standardized CAR myeloid allogenic products, related to their potential to proliferate and withstand extensive gene editing. In addition, considering the rapid development of RNA carriers linked to the SARS-CoV-2 pandemics, it is possible to imagine that local delivery of lipid nanoparticles (LNP) loaded with RNA encoding CAR constructs could represent a viable strategy for CAR macrophages generation. Given the high numbers of TAMs in solid tumors and their capacity for phagocytosis, LNPs may be preferentially captured by TAMs in solid tumors. Importantly, such a strategy would bypass the heavy manufacturing required for myeloid cell–based therapies. Although one must consider that it may be difficult to efficiently target TAMs or their precursors, and that the short half-life of RNA might hinder the induction of a long-lasting antitumor immune response, these novel strategies may pave the way for interesting developments and improvements of immunotherapies.

J. Nikolic reports a patent for EP22305473 pending and EP22305474 pending. P. Benaroch reports a patent for EP22305473 pending and EP22305474 pending. No disclosures were reported by the other authors.

We would like to thank Nadia Jeremiah and Enzo Poirier for valuable comments and English editing of the article, and Jean-Nicolas Volff and Sophia Belkhir for their assistance and input on an early draft of the work. Grants were provided by ITMO PCSI, Chercher Trouver foundation, PSL/Qlife, Emergent Curie project, Association pour la Recherche sur le Cancer (ARC), Ligue contre le cancer, and Laboratoire d’Excellence (Labex) DCBIOL (ANR-10-IDEX-0001-02 PSL and ANR-11-LABX-0043; to P. Benaroch).