Using CAR-M cell therapy to treat solid tumors: a detailed review

This review aims to provide a comprehensive and up-to-date overview of CAR-M therapy with a focus on its application in solid tumours. We discuss the current advances in CAR design, engineering strategies, sources of macrophages, and their mechanisms of antitumour activity in detail. Importantly, we highlight emerging strategies to overcome challenges, such as limited persistence, phenotypic instability, and TME-induced immunosuppression. We also explored the translational potential of CAR-Ms and novel delivery platforms by integrating data from preclinical and early clinical studies. This review contributes to the field by outlining the most recent developments in CAR-M immunotherapy, identifying key obstacles to its clinical success, and proposing future directions to unlock its full therapeutic potential.

CARs are synthetic receptors composed of three main domains: an extracellular antibody domain that specifically binds the appropriate antigen, typically an scFv, a transmembrane domain anchoring the CAR in the cell membrane, and an intracellular signalling domain that contains the signal transduction domain and triggers activation of immune effector functions. The modular design of CARs allows for customisable targeting and downstream responses, depending on the effector cell type. Although originally developed for T cells, CARs have been adapted for use in macrophages, with modifications that reflect their distinct biological roles and signalling requirements. Understanding the general structure of CARs provides a foundation for interpreting specific engineering strategies employed in CAR-M development.

CAR-Ms are an emerging class of engineered cellular immunotherapies designed to enhance the natural tumour-infiltrating, phagocytic, and immunomodulatory functions of macrophages. Unlike CAR-T and CAR-NK cells, CAR-Ms have demonstrated superior persistence within the TME and possess the ability to degrade ECM, thereby facilitating immune cell infiltration. The structural design of CAR-Ms has evolved over successive generations to maximise their therapeutic efficacy and adaptability. Early CAR constructs originally developed for T cells also served as templates for CAR-M engineering, with preclinical studies adopting modular architectures tailored to macrophage-specific functions. The first CAR constructs designed for CAR-T cells also provided guidance for CAR-M development (18, 27, 28). In vitro and in vivo preclinical studies have followed the basic principles of CAR-T technology and have progressed rapidly to generate new generations of CARs. Structurally, CAR constructs in CAR-Ms are similar to those anchored on CAR-T cells (29, 30) (Figure 1).

The extracellular antibody domain (scFv) consists of heavy and light chains of variable regions of monoclonal antibodies connected by a flexible linker. This modular domain is responsible for recognising tumour-associated antigens such as CD19, human epidermal growth factor receptor 2 (HER2), and CD47, which are commonly expressed in tumour tissues (31–36). Regarding preclinical advances in CAR construct refinement, the scFv design remains the favoured extracellular domain and is preserved in most next-generation CAR constructs, even when changes occur in the intracellular signalling domain, targeted antigens, or gene delivery methods (37–40). In addition to conventional scFv designs, some researchers have proposed replacing the classical antibody domain with natural receptor-ligand pairs in new CAR-M generations. In a 2021 study by Niu et al., Raw264.7-derived mouse macrophages were modified with a CAR construct incorporating a chemokine (C-C motif) ligand 19 (CCL19) extracellular domain (41). Preclinical data showed the emergence of a highly immunosuppressive subpopulation expressing lipid droplets and CC chemokine receptor 7 (CCR7), migrating from tumour tissues to lymphoid organs, such as the thymus and spleen. CCL19⁺ CAR-Ms were directed against these immunosuppressive cells in vivo by binding to CCR7 antigens. The results demonstrated that CCL19⁺ CAR-Ms prolonged the survival of breast tumour-bearing mice by exerting strong phagocytic activity on tumour cells and inhibiting the migration of the immunosuppressive subpopulation to lymphoid organs (41). The intracellular signalling domain, the most diversified component in CAR structures, triggers downstream pathways that convert the initial antigen recognition signal into an anti-tumour response. CD3ζ and its homologue Fcγ receptor (FcRγ) are the most common intracellular activating modules introduced into the first generation of CAR-Ms to promote effective phagocytic activity (31, 33).

In one of the initial functional models of CAR-Ms, Morrissey et al. explored the phagocytic potential of CAR-Ms targeting CD19 and CD22 surface antigens of B cells. J774A.1 mouse macrophages expressing anti-CD19 or anti-CD22 CAR constructs fused with Megf10 or FcRγ intracellular domains exhibited significant antigen-specific engulfment of CD19- or CD22-coated silica beads compared with wild-type (WT) macrophages. Similar pilot engulfment assays confirmed the capacity of these CAR-Ms to internalize CD19⁺ Raji B human cells, demonstrating both “biting” and complete “eating” of cancer cells (31).

Although CD3ζ and FcRγ can mediate strong opsonisation and phagocytosis, alternative intracellular domains have been investigated to further enhance CAR-M function. Co-stimulatory molecules, such as 4-1BB (CD137), CD28, and OX40, originally used in second-generation CAR-T cells to reinforce CD3ζ signalling, have also been explored in CAR-M engineering (42–44).

In addition to classical co-stimulatory domains, such as CD28 and OX40, recent constructs have incorporated novel intracellular modules, such as CD86 and 4-1BB. For example, the CCR7-targeting CAR-M developed by Niu et al., integrated toll-like receptor (TLR) signalling domains such as TLR2, TLR4, TLR6, MerTK, and 4-1BB-CD3ζ modules. When co-cultured with 4T1 breast tumour cells, macrophages expressing any of these constructs exhibit strong tumour-killing activity (41).

The increasing diversity of intracellular domains beyond the basic generations not only enhances CAR-M phagocytic capacity, but also expands their complementary antitumour functions, as will be discussed later. CAR-M technology has progressed over multiple generations, improving its therapeutic efficacy and adaptability (Figure 2; Table 1).

CAR-M technology has progressed through three successive generations, each incorporating increasingly sophisticated features to enhance the therapeutic efficacy. First-generation CAR-Ms primarily focus on tumour antigen recognition and direct phagocytosis, using intracellular signalling domains such as FcγR and Megf10, with some variants incorporating PI3K signalling to improve whole-cell engulfment (31). Second-generation CAR-Ms integrate additional costimulatory domains, including CD3ζ- and Toll-like receptor (TLR)-based modules, to amplify pro-inflammatory responses and support antigen presentation. These constructs facilitated enhanced cytokine secretion, T cell activation, and macrophage persistence. Notably, adenoviral vectors, such as Ad5f35, and the use of induced Pluripotent Stem Cells (iPSCs)-derived CAR macrophages (CAR-iMACs) enabled scalable production and prolonged M1 polarisation (45). For example, anti-HER2 CAR-M (CT-0508) was engineered for HER2-positive tumours and showed durable antigen presentation and macrophage activity (46). Third-generation CAR-Ms have introduced features such as polycostimulatory domains, cytokine secretion modules, and nanocarrier-based delivery systems to further remodel the TME, boost T-cell recruitment, and enhance in vivo efficacy. A recent example includes DNA nanocomplex-mediated CAR-HER2 macrophages, which selectively target tumours while maintaining potent antitumour activity (47).

Ongoing refinements in CAR-M technology aim to improve its efficacy, persistence, and adaptability, positioning it as a promising tool for advanced cancer immunotherapy.

Despite these advancements, challenges, such as macrophage plasticity, limited expansion in vivo, and potential off-target effects, remain. Continued development has focused on refining tumour antigen specificity, improving persistence, and ensuring clinical safety for broader application.

The idea of TAMs as a therapeutic cell substitute is based on the plasticity of macrophages, meaning that they are reprogrammable at functional and phenotypical levels; thus, they are exemplar cell candidates for use in tumour eradication programs. Furthermore, several studies have shown that macrophages are naturally capable of infiltrating solid tumours and exhibiting sustainable antitumour activity (29, 48). Following the publication of early pioneering reports, CAR-M-based immunotherapy options have increased owing to their superiority in terms of expansion, infiltration, and activity compared to CAR-T cell therapy. Targeting the improvement of CAR technology, CAR-M pipeline projects may evaluate the potential of different macrophage sources. However, phenotype pursuit and stable transformation are not sufficient to evaluate CAR-Ms for clinical application. Many of preclinical studies on murine and human tumor models are nominated as proof-of concept studies, testing the fruitfulness of a wide range of macrophages cell lines, immortalized murine macrophages isolated from the bone marrow, and human THP-1 macrophages (31, 37, 40, 49, 50).

Macrophages play a key role in clearing pathogens and maintaining the immune balance. Mostly known for the phagocytosis process, they are classified as M1 and M2 type macrophages to conduct a pro- or anti-inflammatory response, respectively. As mentioned above, TAMs are tumour tissue-resident macrophages that favour tumour growth and are associated with poor prognosis in most cancers, owing to their predominant M2 phenotype. In addition, TAMs regulate the cytokine and chemokine networks and influence other immune cells. This modulation creates an immunosuppressive TME that favours tumour growth. As TAMs play a pivotal role in cancer progression, they have become prominent targets for antitumour immunotherapy research. In addition to TAM depletion therapies, TAMs reprogramming strategies require special attention. Functional reprogramming and phenotypic repolarization of TAMs can be achieved using TLR agonists, PIK3 inhibitors targeting the phosphatidylinositol 3-kinase/protein kinase-B (PIK3-PKB/Akt) signalling pathway, co-delivery of signal transducers and activators of the transcription 6 (STAT6) inhibitor, and IκB kinase-β (IKKb) small interfering (siRNA) to decrease M2-TAMs frequency and increase M1-TAMs with anti-tumourigenic activity (61–63). Building on TAM reprogramming strategies, researchers have developed CAR-M therapies to further enhance antitumour responses. Parallel to this strategy, the reprogramming of TAMs can consist of functionality modulation, represented by the enhancement of phagocytic activity and minimisation of the immunoregulatory impact of TAMs in the TME, such as the unfavourable activation of regulatory T cells (Tregs). Blocking the CD47- signal regulatory protein alpha (SIRPα) interaction by using blocking antibodies or small drugs to silence the ‘don’t eat me signal’ from solid or liquid tumours is a promising way to prolong the survival (64, 65). Indeed, TAM-targeted immunotherapy is a good option to treat cancer because of the phenotypic plasticity of TAMs and their adaptive capacity to reinforce their killing activity. Inspired by CAR-T cell immunotherapy, a gene-editing strategy can be introduced in TAM-targeted therapy programs to obtain more satisfactory results in solid cancer treatment. CAR-M therapy has emerged as a promising response to the disadvantages of CAR-T and CAR-NK therapies. Experiments confirmed that CAR-M antitumour activity builds on natural macrophage mechanisms. In vitro and in vivo models demonstrated the ability of CAR-Ms to reduce the tumour burden and elucidate their cytotoxic mechanisms Figure 3. In this section, we provide a review of pioneering preclinical studies that thoroughly investigate CAR-M anti-tumoricidal effects.

CAR-Ms exhibit multiple antitumour mechanisms (66, 67).

Several preclinical models have been developed since the demonstration of the anti-cancer activity of biter and eater macrophages by Morrisey et al. to enhance phagocytosis (33, 39, 41). Gill, who was nominated for the design of the first clinical trial on CAR-Ms in 2020, demonstrated the phagocytic activity of anti-CD19, -CD22, -HER2, and -mesothelin CARs using THP-1 human macrophage cell lines on antigen-positive target cells (33, 52). In a recently published study by Chupradit et al., phagocytosis was enhanced in a hepatocellular carcinoma model to target the CD147 tumour-associated antigen using CAR-Ms derived from THP-1 human cells. It has been noticed that such CAR-Ms are homologues of natural macrophages, but are equipped with enhanced phagocytosis activity, exhibiting significant cytotoxicity specific to tumour cells (68). The anti-cancer activity of CAR-M extends beyond mere tumour cell engulfment and killing. CAR-Ms polarized towards M1 pro-inflammatory phenotype can acts as antigen-presenting cells and infiltrate the cytotoxic CD8⁺ T cells, Th17 cells, NKs and neutrophils to create a collaborative immunological niche against malign cells (49, 53, 69).

Creating more complex CAR structures is the next step in CAR programs to strengthen the functional abilities of CAR-Ms and reduce off-target specificity. Eisenberg et al., designed a new generation of CAR constructs using mucin1 (MUC1) tumor-associated antigen-specific scFv domain, CD3ζ domain, and CD28 and OX40 costimulatory domains. Anti-cancer assays on a solid lung tumour or malignant pleural effusion model have shown that anti-MUC-1 CAR-Ms can be differentiated with a stable CD86⁺ TNF-α⁺ IL-8⁺ IL-1β⁺ M1 phenotype, displaying tumour-restricted phagocytosis and antigen-specific lysosomal processing with negligible off-target specificities (70).

To enhance the autonomous activity of CAR-Ms, Zhang et al., engineered HER2-targeted CAR-Ms with integrated shRNA targeting SIRPα. These shSIRPα-CAR-Ms displayed reinforced proinflammatory polarisation, superior phagocytic capacity, and improved cytotoxicity in both patient-derived tumour organoids and in vivo models (71).

Building on these efforts, recent studies have introduced synthetic modules to further improve the functional sophistication of CAR-Ms. For instance, Huang et al., developed dual-specific CAR-M constructs capable of simultaneously engaging tumour antigens and innate immune receptors, thereby enhancing phagocytic activity and TME modulation (72). In another example, CD147-based intracellular signalling domains promote tumour matrix degradation and cell motility, increasing infiltration while preserving cytotoxic function (73).

Zhao et al., used CRISPR-Cas9 to edit metabolic pathways in CAR-Ms, boosting their inflammatory output and oxidative burst, which resulted in more durable M1 polarisation and tumour rejection in vivo (74). Complementary strategies include combining CAR-Ms with immune checkpoint blockade; for example, co-administration of PD-1 inhibitors has been shown to synergise with CAR-M-induced T-cell recruitment and remodelling of the immunosuppressive niche (38).

Together, these findings underscore the emergence of multifunctional CAR-M designs that not only enhance direct cytotoxicity but also shape the tumour microenvironment and coordinate with other immune compartments to sustain robust antitumour responses.

A common challenge in CAR-M research is that the TME often drives macrophages toward an M2 anti-inflammatory phenotype, which impairs their antitumour efficacy. Thus, TME conditioning is a very important aspect to be visualised by CAR researchers to allow CAR-Ms to gain an M1 phenotype and assist in direct anti-cancer activity accompanied by the infiltration of other immune cells to create a tumour-invasive habitat. Zhang et al., tested Raw264.7 mouse macrophages in a murine breast cancer model to consider the CAR-M effect on the ECM, which is a highly rigid, complex acellular network within the TME that acts as a dense barrier that impedes most cell- and drug-based antitumour therapies (37, 75). The team introduced a CAR motif to bind HER2 tumour antigen and aimed to activate CD147 signalling to increase the expression of matrix metalloprotease (MMP), degrade ECM, and allow T cells to infiltrate the tumour. These promising results showed that infusion of CD147 targeting CAR-Ms inhibited the growth of HER2⁺ 4T1 tumour cells and allowed T cell trafficking into tumours via ECM degradation (37). Strong antitumour activity has also been observed in anti-HER2 and anti-CD47 THP-1 human macrophages towards HER2⁺ SKOV3 and CD47⁺ A2780 human ovarian cancer cells (76). CD47 is a critical “don’t eat me signal” biomarker which allows cancer cells to evade the innate immune defense. The high expression level of CD47 and activation of the CD47-SIRPα pathway are correlated with a low survival rate in cancer patients (77). This in vitro CAR-Ms model merits special attention by pointing to a combined antigen-specific anticancer effect manifested by phagocytosis and cytotoxic T lymphocyte (CTL) polarisation. This therapeutic effect was validated in a xenograft tumour model using BALB/c nude mice transplanted with THP-1 and SKOV3 cells. The inhibitory effect was versatile and was expressed by phagocytosis, infiltration by CTLs, and switching of M2 macrophages to M1 macrophages within the TME. A recent study by Chen et al., introduced a novel CAR-M design that co-expresses a chimeric receptor with an FcRγ signalling domain and a soluble SIRPα decoy, effectively neutralising the CD47–SIRPα axis and enhancing phagocytosis, even under low-antigen conditions. This engineered CAR-M not only boosted direct tumour cell engulfment but also reshaped the TME by promoting infiltration of cytotoxic T cells and inflammatory myeloid populations, highlighting a promising approach to overcome immune escape in solid tumours (78).

Combination strategies involving CAR-Ms are increasingly being explored to overcome the key limitations of monotherapies for solid tumours. While CAR-Ms exhibit unique capabilities, such as matrix degradation, tumour infiltration, and pro-inflammatory polarisation, their efficacy can be hampered by immunosuppressive cues within the TME and limited persistence. Combining CAR-Ms with complementary modalities, such as immune checkpoint inhibitors, targeted therapies, and chemotherapeutics, aims to boost antitumour responses through synergistic mechanisms. This section highlights selected preclinical and translational efforts that exemplify the rationale and potential of CAR-M combination strategies.

To address CAR-M exhaustion and improve their persistence in the TME, Lei et al., engineered constructs with increased antigen specificity, resulting in enhanced antitumour activity and reduced off-target effects (45). In this design, the CAR construct included the classical CD3ζ intracellular domain, which was further strengthened by the addition of a tandem TLR4-derived toll/IL-1 receptor (TIR) domain. Locked onto EGFRvIII- and Glypican-3 (GPC3), the new version of CAR-iMACs polarised with LPS stimulation seems to be resistant to the immunosuppressive effect of the TME and are conferred with an ameliorated adhesive nature and displayed a strong anti-cancer effect on EGFRvIII+ U87MG-human glioblastoma cell lines compared to WT or first-generation CAR-iMACs. In line with this functional buildup, the adoptive transfer of a-EGFRvIII CD3ζ-TIR-CAR-iMACs diminished the tumour growth rate and prolonged the survival of animals in an orthotopic glioblastoma model. The dual signalling offered by the CD3ζ-TIR intracellular domain augments the secretion levels of IL-6, IL-12, IL-23, and TNF-α pro-inflammatory cytokines and ensures antigen-dependent antitumour activity, reflected by a robust phagocytic capacity, TNF-α dependent clearance of apoptotic tumour bodies, and emendation of the TME. This study emphasised the potential synergy between CAR-iMACs and CD47 antibodies. The combinatory infusion strategy achieved a stronger anticancer effect in a hepatocellular carcinoma model and resulted in significant remission in tumour-bearing mice. In 2023, Wang et al., applied metabolic gene knockout on CAR-iMACS to neutralise the Aconitate Decarboxylase 1 (ACOD1) gene and obtained M1 polarized CAR-iMACS with a more potent phagocytic and cytotoxic activity against cancer cells. The M1 profile was positively correlated with high expression of CD80/CD86 costimulatory molecules, release of IL-6, IL-1β cytokines and low expression of CD163/CD206 anti-inflammatory proteins. The anti-cancer effect has been validated in vivo by human HO-8910 driven ovarian and human AsPC-1 driven pancreatic cancer models, where tumour-bearing animals had better survival under the control of CAR-iMACs. The tumour suppression effect of CAR-iMACS was even more pronounced by adding anti-CD47 or anti-programmed cell death protein-1 (PD1) immune checkpoint inhibitors in the treatment regimen (79). The PD1-PD-L1 interaction is one of the most studied signalling mechanisms in cancer treatment. PD1- programmed death ligand-1 (PD-L1) pathways keeps the TME in an immunotolerigenic state and favours tumor growth. Unfortunately, this has been clearly demonstrated in most patients with solid tumours and there is no satisfactory response to anti-PD1 monotherapy (80). In a recently published paper by Pierini et al., a clinically relevant fully immunocompetent syngeneic mouse model was developed to consider how CAR-Ms endeavoured to reduce the coverage of the PD1 blockade strategy. Promising results have shown that anti-HER2 CAR-Ms not only have a multifaceted anti-tumour effect by directly killing tumour cells and by remodelling the TME, but also sensitise solid tumours to anti-PD1 immunotherapy (36). Finally, the newly established combined strategy included the synergistic use of CAR-T and CAR-M cells. Liu et al., were the first to demonstrate beneficial cooperation between CAR-T and CAR-Ms (48). Anti-CD19 CAR-Ms carrying the FcRγ intracellular domain and anti-CD19 CAR-T cells showed significant cytotoxicity against Raji B cells. However, the conjugation of the antitumour effect from each cell type exhibited a stronger effect than that obtained from CAR-Ts or CAR-Ms alone. The mechanisms underlying the trademark of the combinatory effect have been described as the differentiation of macrophages into M1 macrophages by CAR-T cell-derived cytokines and the enhancement of cytotoxicity from CAR-Ms. CAR-Ts and CAR-Ms boost each other in a circular cascade because CAR-Ts augment the expression of CD80 and CD86 surface co-activation molecules on CAR-Ms, which in turn ensures T cell priming and activation via the CD80 and CD86-CD28 signaling pathway. Finally, the alliance of CAR-Ts with CAR-Ms remodulates the TME to the detriment of the tumour via a large cytokine network including IL-6, IL-1β, and IFN-γ. In conclusion, preclinical data have shown that CAR-Ms are sufficiently flexible to be introduced into a combined antitumour strategy, and encouraging observations endorse their beneficial assistance and future evaluation in clinical trials.

The recapitulatory information drawn from these proof-of-concept studies indicates that CAR-Ms are highly likely to be modulated to create a better version of the previous generation, and their beneficial effects can be accentuated in combined immunotherapy strategies. One of the brilliant advancements in CAR-M generation methods is that the nanotechnology field and in situ programming have been integrated into CAR engineering to allow the evolution of third-generation CAR-Ms. Despite the widespread use of adenoviruses and lentiviruses in CAR-M generation, concerns over the immunogenicity of viral vectors, safety risks (particularly the potential for insertional mutagenesis), and significant economic burden have driven researchers to explore alternative gene delivery strategies and enhance safety protocols.

A transient delivery method of chemically modified mRNA has been optimised to induce M1 type macrophages in the presence of IFN-γ, and this non-viral delivery system induced persistent expression of CAR constructs and enhanced antitumour activity (81).

Nanomaterial-based carriers are a rising trend in classical viral gene delivery systems, owing to their low cost, large-scale production, and production simplicity (82). Among them, lipid nanoparticle vectors (LNPs) shine amongst others by their efficiency in mRNA vaccine development against COVID-19 (83, 84). The translational potential of the LNPs was integrated into mRNA technology using the CAR-M engineering platform. The team of Zhongfeng Ye et al., successfully engineered functionally active CAR-Ms using LNPs to transfer CD19 mRNA, and in vitro preclinical experiments demonstrated that CAR-Ms are highly cytotoxic against B lymphoma (85). The following year, Yang et al., encapsulated anti-GPC3 CAR into LNPs to specifically edit macrophages in a hepatocellular carcinoma (HCC) model (39). The combination of CAR-M therapy with the blockade of the CD24–Siglec-G axis, achieved through the use of Siglec-G lacking ITIM motifs, reduced tumor burden and prolonged survival in HCC-bearing mice by enhancing CAR-M-mediated phagocytic activity. Kang et al., reported a proof-of-concept study of an anaplastic lymphoma kinase (ALK)-expressing neuroblastoma solid tumour model in 2021 (47). By using macrophage-targeted polymer nanocarriers as gene delivery vectors, the group designed an anti-ALK CAR to transfer into bone marrow-derived macrophages (BMDMs), within M2-BMDM were reprogrammed at the same time as CAR-IFN-γ gene constructs to shift the M2 phenotype to M1. This study has indicated that the switch-on of strong anti-cancer activity is associated with phagocytosis, antitumour immunoregulation, and the reduction of tumour size.

In a 2022-dated preclinical study, Chen and his group, working on glioblastoma multiforme known to have a high rate of relapse due to residual glioma stem cells (GSCs) after surgical intervention, were motivated to generate in vivo CAR-Ms targeting GSCs as part of postsurgical therapy (86). The group used a nanopore loaded with an anti-CD133 gene-hydrogel structure and co-injected the structure with anti-CD47 antibodies in the resection intracavity to edit local macrophages in situ. The results clearly revealed that edited CAR-Ms were either able to deteriorate GSCs via efficient phagocytic activity or mounted an adaptive immune response by gaining antigen-presenting capacity. Building upon advances in nanoparticle-based therapeutics, Zhou’s team, in 2025, integrated in situ CAR-M generation with LNP-mediated delivery to mitigate the challenge of nanoparticle sequestration by tissue-resident macrophages (as detailed previously), thereby harnessing their intrinsic capacity for in vivo uptake. The group developed a 4T1- trophoblast antigen 2 (Trop2) breast cancer model to target the Trop2 tumour-associated antigen. In situ delivery of Trop2-CARs led to the development of CAR-Ms, which selectively destroyed Trop2⁺ cancerous cells and simultaneously boosted the activation of NK cells and CTLs.

Reprogramming macrophages into CAR platforms holds substantial promise for the treatment of solid tumours. Notwithstanding the preclinical findings that have begun to be appraised in the context of clinical trials, in silico immuno-oncology models based on CAR-M therapies are of vital importance in determining the key components of remission and relapse phase-associated immune responses and predicting the long-term benefits and side effects of therapy and therapy resistance mechanisms. In silico modelling of CAR platforms is still in its infancy and mostly dedicated to CAR-T cell-targeted immunotherapy (87, 88). Recently, computational tools have been integrated into CAR-M research. In a TME-targeted in silico study directed by Sugimura R. in 2024, anti-inflammatory transforming growth factor (TGF)-β cytokine is one of the nefarious cytokine keeping TME in a immunosilent atmosphere to benefit the growth of tumor (89). The group processed their computational knowledge using in vitro and in vivo liver cancer models and reprogrammed macrophages to secrete anti-TGFβ scFv (AB-CAR-M). Preliminary results demonstrated that the supernatants from AB-CAR-Ms induced CTL-related anticancer activity, suppressed pro-tumourigenic regulatory T cell activity, and boosted pro-inflammatory M1 macrophages within the TME.

Since the latest encouraging reports on CAR-Ms in solid tumour treatments, the rapid development of CAR-M therapies in solid tumours has prompted researchers, clinicians, and clinical investigators to consider these immunotherapeutic cells as alternative personalised therapy options to fight non-oncologic diseases. CAR-Ms have recently emerged as promising immune therapeutic agents for managing the pathogenesis of various diseases (90). Inspired by the fact that CAR-Ms act in an antigen-dependent manner and modulate the milieu in which they are localised, several preclinical studies have focused on the ability of CAR-Ms to eliminate exogenous pathogens or abnormal cells that present a specific antigen signature. Since 2021, CAR-Ms have been subjected to in vitro and in vivo studies on an extended area of pathogenesis, including viral and bacterial infections, neurodegenerative diseases, and cardiac diseases (35, 91–93). As discussed in previous studies, CAR-Ms have demonstrated potential to clear SARS-CoV-2 infection from the immune system (26). In a murine model of septicaemia caused by methicillin-resistant Staphylococcus aureus (MRSA) infection, LNPs loaded with CAR mRNA targeting SasA surface protein and siRNA targeting the CASP11 evasion biomarker mediated the in situ generation of CAR-Ms, which exhibited potent phagocytic activity against MRSA (91).

Neurodegenerative diseases, such as Alzheimer’s disease (AD), have also been targeted using CAR-Ms. AD is an uncured neurodegenerative disease that affects millions of individuals worldwide. Although the pathogenesis is not yet fully understood, the deposition of β-amyloid plaques in brain tissue and the consequent neurotoxic cascades and destructive neuroinflammation are accepted as key signatures of the disease. In addition to the anti-β-amyloid monoclonal antibody drug approved by the FDA in 2021, antigen-specific CAR-Treg infusion has also been proposed as a potential therapeutic strategy owing to the immunomodulatory character of Tregs, which can dampen excessive neuroinflammation and induce local tolerance (94). More recently, in 2024, Kim et al., explored CAR-M immunotherapy for AD (92). This work aligns with recent reviews on the role of CAR-Ms in neuroinflammatory disorders and the therapeutic potential of targeting the central nervous system (95). They debuted the study with intracranial administration of a first-generation CAR-M targeting β-amyloid plaques differentiated from Hoxb8 murine cells, which showed a beneficial effect on the plaque load. Based on these encouraging results, the group designed an upgraded version that could be maintained by macrophage colony-stimulating factor (M-CSF) secretion in an autocrine manner. As expected, anti-β-amyloid CAR-Ms secreting M-CSF could persist for 30 days in local tissue compared to the first generation of CAR-Ms with only two weeks of lifespan and presented a reinforced phagocytic activity on plaque load. In the cardiovascular field, CAR-T cell-based therapies have recently been explored for cardiac injury, and CAR-M immunotherapeutic approaches have emerged as promising alternatives for cardiac injury treatment (96). CAR-M-based immunotherapeutic assays have also been considered as alternative methods.

CAR-M therapy has been proposed as an alternative to T-cell–based approaches for cardiovascular diseases. Wang et al., working on myocardial fibrosis as a result of myocardial ischemia-reperfusion injury, designed a-FAP (fibroblast activation protein) CAR-Ms (93). Two weeks of treatment indicated that CAR-Ms could infiltrate the cardiac tissue, mediate the phagocytosis of fibroblasts, help recover cardiac function, and thus attenuate the extreme inflammation related to ischaemia-reperfusion injury. By harnessing the superlative immune properties of macrophages, researchers have made exciting breakthroughs that could translate CAR-M programming into non-cancerous diseases. Although CAR-based immuno-oncologic key points serve as a reference for their preclinical applications in non-cancer diseases, very little is known about clinical trials targeting CAR-Ms in cancer; thus, CAR-Ms need time to be validated into non-tumour therapeutic classes. These preclinical studies collectively underscore the broad therapeutic potential of CAR-Ms beyond oncology, establishing a foundation for future clinical translation.

Although early preclinical studies have confirmed the feasibility of CAR-M therapies in animal models, further refinement is necessary to overcome the complex barriers of solid tumours, including immunosuppressive cytokine networks, hypoxia, and stromal fibrosis. Enhancing the in vivo efficacy remains a priority to ensure robust tumour infiltration, sustained antitumour activity, and durable immune system engagement. Several innovative strategies have been developed and validated, specifically in preclinical in vivo models, offering a roadmap for future clinical applications.

These strategies encompass not only improvements in CAR architecture (e.g. dual-costimulatory domains and cytokine expression modules) but also modulation of macrophage metabolic programming to favour M1 polarisation and enhanced persistence within the TME. For instance, metabolic reprogramming through inhibition of immunoregulatory pathways, such as ACOD1 or IDO1, has shown promise in increasing the inflammatory and cytotoxic capacity of CAR-Ms. Additionally, CAR-Ms can be engineered to resist key immunosuppressive signals (for example, TGF-β and IL-10) through dominant-negative receptors or decoy constructs and have demonstrated improved persistence and functionality in vivo.

Beyond genetic strategies, environmental conditioning via pretreatment with cytokines (e.g. IFN-γ) or metabolic adjuvants is being explored to “prime” CAR-Ms for pro-inflammatory activity. Collectively, these approaches illustrate a growing toolbox of engineering and conditioning strategies aimed at maximising CAR-M efficacy in vivo and overcoming the resistance mechanisms prevalent in solid tumours.

Combination strategies have emerged as a promising approach to enhance CAR-M potency within the immunosuppressive TME, where high levels of regulatory cytokines, inhibitory ligands, and suppressive immune cells pose significant challenges to sustained macrophage activation. These strategies aim to synergise CAR-M activity with other therapeutic modalities to achieve deeper tumour regression and broader immune engagement.

Dong et al., demonstrated that combining HER2-targeted CAR-M with oxaliplatin chemotherapy significantly improved tumour regression compared to monotherapies in gastric cancer xenografts (97). This chemoimmunotherapy synergy is thought to be mediated by oxaliplatin-induced immunogenic cell death and increased antigen presentation, which sensitises tumours to macrophage-mediated phagocytosis and antigen cross-priming.

Similarly, Zheng et al., showed that the co-administration of c-MET-targeted CAR-M with gemcitabine or irinotecan amplified antitumour effects in pancreatic cancer models (100), suggesting that certain chemotherapeutics may remodel the TME in ways that enhance macrophage infiltration and persistence. Importantly, chemotherapeutic agents may also reduce the density of stromal barriers and myeloid-derived suppressor cells (MDSCs), further facilitating CAR-M efficacy.

In the immunotherapy domain, Pierini et al., reported that HER2-CAR-M reprogrammed the TME, enhancing CD8⁺ T cell infiltration, and sensitising tumours to PD-1 checkpoint blockade in vivo (36). This positions CAR-Ms not only as direct effectors but also as immunological “primers” that convert immunologically “cold” tumors into “hot” ones, enabling previously refractory tumours to respond to checkpoint inhibitors.

In addition to PD-1 inhibitors, combinatory approaches involving anti-CD47, anti-CSF1R, or anti-TGF-β therapies are under investigation to prolong M1 macrophage phenotypes and block alternative immunosuppressive loops. Altogether, these combinations highlight the versatility of CAR-Ms as both effector and facilitator cells in multiagent therapeutic strategies.

The tumour stroma, predominantly composed of CAFs, represents a major physical and immunological barrier to effective therapy. Thus, stromal targeting using CAR-M cells is a key innovation. Mao et al., engineered a FAP-specific CAR-M capable of selectively eliminating CAFs, leading to reduced fibrosis, enhanced CD8+ T cell infiltration, and significant tumour growth delay in vivo (98). This pioneering study underscores the importance of remodelling the TME to facilitate immune access and enhance responsiveness to therapy.

CAFs not only produce dense ECM components that hinder immune cell infiltration, but also secrete immunosuppressive cytokines and chemokines that promote tumour progression and inhibit effector cell function. Therefore, the selective elimination of CAFs by CAR-Ms can exert a dual benefit: mechanically opening the tumour for immune cell infiltration and biologically reversing immune suppression. Importantly, their study revealed that CAR-M-mediated stromal depletion not only disrupted the physical barriers surrounding tumour nests but also created a more permissive microenvironment for endogenous immune responses.

Building on this strategy, Zhang et al., designed a quadrivalent CAR-M system capable of simultaneously targeting FAP (on CAFs) and GPC3 (on tumour cells) while expressing a dominant-negative TGF-β receptor to resist stromal-mediated immunosuppression (102). Delivered via in vivo lipid nanoparticle (LNP) systems, these quadrivalent CAR-Ms achieved potent stromal remodelling and durable tumour regression in hepatocellular carcinoma models. This approach reflects the growing interest in multifunctional CAR-M designs that can simultaneously address multiple components of the tumour architecture.

Together, these studies demonstrate that disrupting the stromal architecture is essential for enhancing CAR-M efficacy in vivo. By combining physical barrier elimination with immunological reprogramming, stroma-targeted CAR-Ms offer a promising route for amplifying the penetration and antitumour activity of both innate and adaptive immune cells in solid tumour settings.

As described in more detail in the previous section, focused on in vitro improvements, genetic modifications have been strategically explored to reprogram macrophages towards a pro-inflammatory (M1-like) phenotype and enhance their anti-tumour functions. Building on these in vitro findings, several studies have successfully validated the impact of genetic engineering on preclinical in vivo models.

For example, Wang et al., demonstrated that deletion of the ACOD1 gene in iPSC-derived mesothelin-targeting CAR-Ms enhanced pro-inflammatory activation and reactive oxygen species (ROS) production, significantly boosting antitumour activity in ovarian and pancreatic cancer models (79). This metabolic reprogramming led to a higher expression of M1 markers (CD80 and CD86), increased secretion of IL-6 and IL-1β, and reduced expression of M2-associated proteins (CD163 and CD206), confirming a stable inflammatory phenotype. These results emphasise the role of intrinsic macrophage metabolism in the modulation of CAR-M potency.

Similarly, Duan et al., adopted a complementary strategy by incorporating TLR4 or IFN-γ receptor signalling domains into VEGFR2-targeted CARs (103). These synthetic modules promote M1-like activation and enhance tumour infiltration and suppression in breast cancer xenografts. The inclusion of TLR4 domains mimicked microbial danger signals, stimulating the production of TNF-α, IL-12, and other pro-inflammatory cytokines that helped overcome immunosuppressive cues in the TME.

An additional strategy to preserve the antitumour phenotype involves targeting the intrinsic regulators of macrophage polarisation. Ziane-Chaouche et al., demonstrated that Furin knockdown in CAR-Ms prevented their M2 reprogramming in the TME, sustained M1-associated cytokine profiles, and increased their ability to activate T cells and eradicate tumor cells in patient-derived tumoroids and xenograft models (104).

Overall, these studies illustrate that rational engineering of CAR signalling domains or macrophage intracellular programs can substantially increase effector function, persistence, and specificity in vivo. Future designs may benefit from modular “plug-and-play” CAR systems incorporating synergistic co-stimulatory, cytokine, or metabolic modules tailored to the tumour type and TME characteristics.

Effective infiltration of CAR-M into the tumour bed is crucial for therapeutic success, particularly in solid tumours that are characterised by dense stromal components and abnormal vasculature. Several studies have investigated optimised delivery routes to maximise tumour homing and tissue penetration.

Li et al., demonstrated that intraperitoneal administration of mesothelin-targeting CAR-M resulted in superior tumour infiltration and control compared to intravenous infusion in ovarian cancer models (101). This finding underscores the importance of locoregional delivery for enhancing cell persistence and functional activity at tumour sites. Similarly, Dong et al., found that local intratumoural delivery significantly improved HER2-CAR-Ms efficacy in gastric cancer xenografts (97), likely because of enhanced retention within the TME and immediate proximity to tumour cells.

These observations suggest that conventional systemic delivery routes may limit CAR-M accumulation in tumours, particularly in immunologically “cold” or poorly perfused tumour types. Thus, alternative administration strategies, such as intratumoural, intracavitary, or intrapleural injection, are being increasingly considered to overcome these limitations and maximise local cell density and antitumour activity.

Moreover, the route of administration may influence the macrophage phenotype post-infusion. Local exposure to tumour antigens and cytokines may sustain M1 polarisation and enhance the immunomodulatory functions of CAR-Ms. Therefore, tailoring the delivery strategy to the tumour type, anatomical location, and desired immune outcomes is a key consideration in CAR-M therapeutic design.

Although ex vivo engineering approaches have successfully generated potent CAR-Ms for preclinical studies, their clinical translation faces significant logistical challenges, including scalability, manufacturing complexity, and patient-specific variability. Therefore, innovative strategies that enable efficient in vivo reprogramming of endogenous macrophages are gaining increasing attention as a promising approach.

Traditional ex vivo manufacturing involves leukapheresis, cell isolation, differentiation, viral transduction, and reinfusion, which require time, specialised facilities, and significant financial resources. These complexities are particularly problematic in the context of autologous therapies, in which patient-to-patient variability can compromise product consistency and delay treatment.

In contrast, the in vivo reprogramming of macrophages represents a paradigm shift, enabling on-site genetic modification without the need for ex vivo manipulation. Nonviral delivery systems, particularly lipid nanoparticles (LNPs), have emerged as promising platforms for this purpose. As previously described, several studies have shown that the in situ delivery of CAR constructs via LNPs or polymer-based carriers can effectively program macrophages within the TME, leading to tumour-selective CAR-M generation and robust antitumour activity.

This approach offers multiple advantages: it circumvents the need for individualised cell manufacturing, potentially enables repeated dosing, and leverages the natural tropism of macrophages at tumour sites. Furthermore, the use of tissue-specific promoters or localised administration routes (e.g. intratumoural injection) can help to restrict CAR expression to the desired myeloid subsets, thereby improving safety. As in vivo delivery technologies mature, they are likely to play a central role in scalable implementation of CAR-M therapy.

As previously discussed in the section focused on in vitro CAR-M generation, classical ex vivo strategies involve isolation of peripheral blood monocytes or bone marrow-derived cells, differentiation into macrophages using cytokines (M-CSF or GM-CSF), and genetic modification via viral vectors, most notably lentiviruses or adenoviruses. Building on these established methodologies, several preclinical studies have demonstrated that ex vivo engineered CAR-M can exert potent antitumour activity in vivo.

Dong et al., utilised an Ad5f35 vector system to transduce HER2-specific CARs into human peritoneal macrophages, achieving significant tumour control in gastric cancer models (97). This study validated the antitumour potential of CAR-Ms and highlighted their capacity to infiltrate tumour tissues and modulate the immune microenvironment. Similarly, Pierini et al. demonstrated that ex vivo engineered HER2-CAR-M could remodel the tumour microenvironment and sensitise tumours to PD-1 checkpoint blockade (36), indicating a synergistic role for CAR-Ms in combination immunotherapies.

Ex vivo approaches also offer a controlled environment for optimising the transduction efficiency, CAR construct expression, and functional polarisation of macrophages prior to infusion. These features are particularly advantageous for the development of “next-generation” CAR-Ms with enhanced cytokine production, antigen presentation, or resistance to TME-mediated suppression.

Nevertheless, the complexity, cost, and individualised nature of ex vivo production remain significant hurdles for widespread clinical application. For clinical-grade manufacturing, scalable good manufacturing practice (GMP)-compliant protocols and standardised cell sources, such as iPSC-derived macrophages, are being explored to bridge the gap between the bench and bedside. In this regard, ex vivo CAR-M engineering continues to serve as a crucial platform for proof-of-concept studies and the development of lead candidates for early phase clinical trials.

Recent preclinical studies have developed novel in vivo CAR-M programming platforms to overcome the challenges associated with ex vivo macrophage expansion and engineering. These strategies enhance accessibility, reduce manufacturing complexity, and open new avenues for the treatment of solid tumours by directly reprogramming endogenous macrophages. Nonviral delivery systems, particularly LNPs, have emerged as promising platforms for this purpose. Kang et al., employed mannose-conjugated polyethylenimine (MPEI) nanocomplexes to deliver plasmids encoding anti-ALK CAR and IFN-γ into endogenous macrophages, using a PiggyBac transposon system for stable genomic integration in vivo (47). CAR was designed to promote tumour cell phagocytosis, whereas IFN-γ secretion enhanced macrophage polarisation toward an M1-like, pro-inflammatory phenotype. The in vivo experiments were conducted using a murine neuroblastoma model (Neuro-2a, ALK⁺). Notably, direct intratumoural injection of the nanocomplexes resulted in a higher CAR-M programming efficiency and better tumour control than systemic intravenous administration. Extending this approach, Zhang et al., engineered quadrivalent LNPs encoding two distinct CAR constructs to simultaneously target hepatocellular carcinoma (HCC) and tumour stroma. One CAR recognised GPC3-expressing tumour cells and incorporated a Super IL-2 cytokine module designed to stimulate T-cell activation locally. The second CAR targeted FAP-expressing CAFs and was fused to a dominant-negative TGF-β receptor to resist stromal immunosuppression. In preclinical HCC models, this dual-targeting strategy achieved potent stromal remodelling, enhanced T cell infiltration, and durable tumour control (102). Systemic intravenous administration preferentially reprogrammed hepatic macrophages, leading to effective TME remodelling, enhanced CD8⁺ T cell infiltration, and durable immune memory formation. Although additional early-stage efforts are emerging, these studies represent the most advanced and translationally relevant examples of in vivo CAR-M reprogramming to date. Together, they highlight the feasibility, flexibility, and clinical promise of this approach, where vector design, CAR architecture, and delivery route critically shape the therapeutic efficacy and immune activation.

Accurate monitoring of CAR-M in vivo is essential to evaluate its biodistribution, persistence, tumour infiltration, functional reprogramming, therapeutic efficacy, and potential toxicity (105). In addition, advanced imaging strategies provide crucial insights into CAR-M dynamics within the TME, helping to optimise therapeutic protocols and predict clinical outcomes (106). Monitoring approaches can be broadly categorised into two groups: experimental modalities widely used in preclinical CAR-M studies, and clinically validated imaging technologies initially developed for macrophage tracking, which hold promise for future translation to human applications (Table 2).

Fluorescence-based techniques, including genetic expression of Green Fluorescent Protein (GFP) and membrane dye labelling with DiR or DiD, have been extensively used for short-term CAR-M tracking in small animal models (47). These approaches allow the visualisation of cell migration and tumour infiltration via confocal or intravital microscopy. For example, Klichinsky et al., employed VivoTrack680, a near-infrared lipophilic membrane dye, to label HER2-targeted CAR-Ms and monitor their biodistribution and tumour infiltration in vivo, demonstrating robust accumulation within xenograft tumours (33). However, their limited tissue penetration and lack of clinical translatability have restricted their use in preclinical studies.

Bioluminescence imaging (BLI), utilising luciferase-expressing CAR-M or tumour cells in combination with luciferin substrate injection, provides sensitive, non-invasive monitoring of CAR-M persistence and antitumour activity over time (33). BLI is a standard tool in numerous preclinical CAR-M studies, including HER2-targeted (97), c-MET-targeted (99, 100), and stroma/tumour dual-targeted quadrivalent CAR-M models (102). This modality enables longitudinal assessment of therapeutic responses and survival outcomes in vivo, providing critical validation prior to clinical translation.

Emerging high-resolution ex vivo technologies such as RNA in situ hybridisation (RNAscope) and multiplex ion-beam imaging (MIBI) have further enhanced the ability to characterise CAR-M localisation and interactions with the TME at single-cell resolution. For example, Reiss et al., successfully applied RNAscope technology to detect CAR-M in tumour biopsies of patients treated with CT-0508, providing spatial confirmation of CAR-M infiltration within the solid tumour microenvironment in a first-in-human clinical trial (46).

Together, these techniques enable comprehensive spatiotemporal tracking of CAR-M behaviour in vivo and offer valuable platforms for correlating phenotypic persistence with therapeutic efficacy and safety. As clinical development advances, continued innovation in imaging and molecular tracking will be crucial to inform dosing, scheduling, and patient-stratification strategies.

Several imaging platforms originally developed for macrophage tracking have demonstrated clinical feasibility and can be adapted for CAR-M monitoring in patients. Magnetic Resonance Imaging (MRI) with superparamagnetic iron oxide nanoparticles (SPIONs), such as Ferumoxytol (an FDA-approved agent), enables non-invasive visualisation of macrophage infiltration into tumours (107). Sharkey et al., demonstrated the ability to track SPION-labelled macrophages using MRI, offering a clinically compatible approach for future CAR-M studies. Positron Emission Tomography (PET) is another highly sensitive imaging method. Radiolabelling of macrophages ex vivo with isotopes such as ⁸⁹Zr-oxine or using probes targeting macrophage markers such as CD206 (mannose receptor) enables the quantitative assessment of cell distribution in vivo (106). Single-photon emission computed tomography (SPECT) has also been employed for macrophage imaging using radiolabelled antibodies against myeloid markers (108). Photoacoustic Imaging (PAI), a novel hybrid optical ultrasound modality, offers high-resolution, real-time imaging without the need for ionizing radiation (108). Although still largely experimental, PAI holds promise for future clinical applications in immune-cell monitoring. The integration of clinically established modalities, combined with novel probe development, could significantly enhance the real-time, non-invasive evaluation of CAR-M therapies during human trials.

CAR-M represents a promising new cell therapy platform designed to overcome several key challenges faced by T cell-based therapies, particularly in the treatment of solid tumors. Their unique ability to penetrate the TME and stimulate innate and adaptive immune responses has sparked considerable interest. CAR-M is entering the early phase of clinical trials following excellent results observed in vitro and in vivo (33, 97, 100), with a few studies currently registered on ClinicalTrials.gov. The details of these studies can be found in Table 3 and are currently evaluating the safety and efficacy of CAR-M therapies in solid tumors, particularly those overexpressing HER2. A recent review by Li et al., comprehensively summarized the preclinical and clinical progress of CAR-Ms, emphasizing translational hurdles and the therapeutic potential of combining CAR-Ms with chemotherapy or checkpoint blockade to enhance efficacy in solid tumors (109).

One of the most advanced is NCT04660929↗, a first-in-human phase 1 trial investigating CT-0508 (46) (Figure 4), an autologous anti-HER2 CAR-M, in patients with advanced HER2-overexpressing solid tumors. In this open-label study, CT-0508 was administered without lymphodepleting chemotherapy, highlighting its favourable safety profile. No dose-limiting toxicities or high-grade cytokine release syndromes (CRS) were observed. Preliminary efficacy data indicated stable disease in a subset of patients, particularly those with high HER2⁺ expression, along with evidence of TME remodeling and T-cell recruitment. Building on this foundation, the ongoing NCT06224738↗ trial explored the combination of CT-0508 and pembrolizumab, a PD-1 checkpoint inhibitor, for HER2⁺ tumors. This phase 1 dose escalation study aimed to assess the safety and potential synergistic activity of this combination. This is based on preclinical findings suggesting that CAR-M-induced tumor inflammation may sensitize tumors to immune checkpoint blockade, potentially overcoming T-cell exhaustion and enhancing antitumor immunity.

Complementing these clinical studies, the NCT05007379↗ trial focuses on a translational approach by testing CAR-M therapies in patient-derived breast cancer organoids. This cohort study aimed to evaluate the antitumor activity of CAR-Ms in ex vivo models that reflect clinical heterogeneity and treatment resistance to help identify responsive tumor subtypes and refine future therapeutic strategies. Although no clinical results from this trial have yet been published, the research group has contributed valuable publications and protocols demonstrating the feasibility and relevance of using ex vivo organoid models to optimize CAR-M development (110, 111).

NCT06562647↗ is the first-in-human, single-center, single-arm, dose-escalation exploratory clinical trial that evaluates the safety, tolerability, pharmacokinetics, and preliminary efficacy of SY001, a mesothelin-targeted CAR-M therapy, in patients with advanced solid tumors (101). The trial commenced on April 12, 2023, and is currently recruiting participants. Initial clinical data from the first two patients indicated that SY001 was well tolerated, with no dose-limiting toxicities observed. Only two grade 3 adverse events occurred, and common grade 1–2 adverse events included fever and elevated C-reactive protein levels. Notably, no CRS or neurotoxicity related to cell infusion has been reported previously. The NCT03608618↗ trial is a phase 1, first-in-human, dose-escalation study assessing MCY-M11, a mesothelin-targeting CAR-M therapy, in patients with advanced ovarian cancer and malignant peritoneal mesothelioma (Figure 5). MCY-M11 was generated using a rapid, non-viral, mRNA-based manufacturing platform to produce autologous fresh peripheral blood mononuclear cells (PBMCs) transfected with anti-mesothelin CAR mRNA. The therapy was administered intraperitoneally without preconditioning chemotherapy, following a 3 + 3 dose-escalation design across four planned dose levels, with one cycle consisting of three weekly infusions.

Despite promising early safety signals, challenges such as cell persistence, manufacturing scalability, and tumor infiltration efficiency remain critical areas for further investigation.

In the first in-human phase 1 trial of CT-0508 (46), an anti-HER2 CAR-M therapy, the overall safety profile was favourable. No dose-limiting toxicities (DLTs), high-grade CRS (≥ grade 3), or immune effector cell-associated neurotoxicity syndrome (ICANS) were observed. The most common adverse event was mild CRS (grades 1–2), occurring in 64.3% of patients, which resolved spontaneously or with supportive care; only one patient required tocilizumab. Infusion-related reactions, gastrointestinal symptoms (e.g., nausea and vomiting), and transient cytopenia were also reported, all of which were low-grade. Importantly, the therapy was administered without prior lymphodepletion, thereby reducing the risk of prolonged immunosuppression or hematologic toxicity. These findings suggest that CAR-M therapy can be administered safely in heavily pretreated patients with solid tumors, with a distinct toxicity profile compared with CAR-T cell therapies. Table 4 summarizes the toxicities observed with CT-0508.

In the data reported in ASCO for the first-in-human phase 1 trial of MCY-M11, a mesothelin-targeting CAR PBMCs (112) (including CAR-M) therapy delivered intraperitoneally, the overall safety profile was favourable. No dose-limiting toxicities (DLTs), high-grade CRS (≥ grade 3), or immune effector cell-associated neurotoxicity syndrome (ICANS) were observed. The most common treatment-related adverse events were mild (grades 1–2), and included transient fever, abdominal pain, nausea, and fatigue. A single case of grade 2 pericarditis associated with fever and transient neutropenia was reported, which was attributed to on-target off-tumor effects and possible low-grade CRS. It resolved fully with supportive care without the need for corticosteroids or cytokine blockade. To date, no infusion-related reactions or treatment-related deaths have been reported. Importantly, MCY-M11 was administered without preconditioning chemotherapy, thus minimizing the risk of prolonged immunosuppression or hematologic toxicity. These findings indicate that intraperitoneal CAR-M therapy can be safely administered to patients with advanced solid tumors, with a manageable toxicity profile distinct from that observed with CAR-T cell therapies. Table 5 summarizes the treatment-related toxicities observed in MCY-M11.

Initial clinical data from early phase trials suggest that CAR-M therapy may induce antitumor activity in heavily pretreated patients with solid tumors. In the NCT03608618↗ trial evaluating MCY-M11, intraperitoneal administration of anti-mesothelin CAR-Ms led to disease stabilization in a subset of patients with advanced ovarian cancer and malignant peritoneal mesothelioma, with stable disease observed in several patients after a single treatment cycle. Similarly, in a phase 1 trial of CT-0508 (NCT04660929↗), preliminary efficacy signals included stable disease in patients with HER2-overexpressing tumors, particularly in those with HER2 3+ expression. In addition to clinical stabilization, evidence of TME remodeling and T-cell recruitment has been reported in CT-0508-treated patients, supporting the hypothesis that CAR-M therapies may modulate the immune landscape to enhance antitumor responses. Although objective responses remain limited at this stage, these findings are encouraging and justify the continued investigation of combination strategies and expanded dosing regimens. These early findings suggest that CAR-M therapies may achieve durable tumor control by reshaping the TME, even in the absence of strong cytoreductive responses.

The current clinical development of CAR-M therapies has focused primarily on solid tumours, particularly those characterised by the high expression of tumour-associated antigens and an immunosuppressive microenvironment. Ongoing trials, such as NCT03608618↗ and NCT04660929↗, are evaluating CAR-Ms that target mesothelin and HER2, respectively, in patients with advanced ovarian cancer, malignant peritoneal mesothelioma, breast cancer, and gastroesophageal cancer. These indications were selected based on the biological rationale that macrophages can penetrate dense stromal barriers and modulate the TME, thereby overcoming the limitations of T cell-based therapies. Beyond the current targets, CAR-M therapies have the potential to expand to a broader range of solid malignancies, including pancreatic cancer, non-small cell lung cancer, and glioblastoma, in which effective immune infiltration remains a major therapeutic challenge. Preclinical models also suggest opportunities for targeting antigens, such as EGFR, diasilogangloside (GD2), and prostate specific membrane antigen (PSMA). Additionally, CAR-Ms can play a role in haematologic malignancies by exploiting their innate phagocytic capabilities and antigen-presenting functions; however, clinical exploration in this area remains limited. Future directions may include the development of armoured CAR-Ms engineered to secrete pro-inflammatory cytokines or express checkpoint blockade molecules, further enhancing their antitumour efficacy across diverse tumour types.

Despite encouraging early phase results, several challenges remain in the clinical translation of CAR-M therapy. One of the most significant limitations is the relatively short persistence of CAR-Ms in vivo, particularly when using transient mRNA-based platforms, which may limit the durability of the antitumour responses. Strategies to enhance macrophage survival and functionality post-infusion are under active investigation, including the use of autocrine cytokine loops, genetic modifications to resist apoptosis, and incorporation of long-lived memory-like traits. Efficient homing and infiltration into solid tumour sites are also critical hurdles. Although macrophages naturally migrate to sites of inflammation, tumour-associated immunosuppressive signals such as IL-10, TGF-β, and hypoxiacan impair their recruitment or reprogram them toward an M2-like, tumour-supporting phenotype. Optimising the route of administration (e.g. intratumoural or intraperitoneal delivery), engineering enhanced chemokine responsiveness, and shielding CAR-Ms from immunosuppressive cues are active areas of research. Macrophage phenotypic plasticity, which is advantageous for reprogramming, poses a major risk. CAR-Ms may revert to an M2-like, anti-inflammatory, and pro-tumour phenotype under TME pressure. Sustaining M1 polarisation in vivo remains a significant challenge, particularly in solid tumours with a strong immunosuppressive architecture. This has prompted interest in “armored” CAR-Ms expressing pro-inflammatory cytokines or dominant-negative receptors that counteract key TME-derived inhibitory signals. Another complicating factor is the absence of lymphodepleting preconditioning in current CAR-M protocols. While this avoids the toxicities associated with lymphodepletion, it may also limit the therapeutic “niche” for infused cells and reduce engraftment. Future studies should explore tailored, nontoxic conditioning regimens that enhance macrophage persistence without compromising safety.

From a logistical and manufacturing perspective, current CAR-M therapies typically rely on fresh autologous cell products and rapid turnaround, making large-scale deployment challenging. The development of allogeneic or off-the-shelf CAR-M platforms, such as iPSC-derived or -immortalised macrophages, may improve scalability, although they are still in the early phase of development. Furthermore, regulatory frameworks for macrophage-based therapies remain less defined than those for CAR-T cells, adding complexity to their clinical translation.

Therefore, safety remains a concern. Given the widespread expression of many solid tumour-associated antigens, the risk of on-target off-tumour toxicity must be carefully assessed. Engineering strategies to improve antigen specificity, the use of dual antigen recognition, and incorporation of synthetic logic gates or suicide switches are currently being evaluated to mitigate these risks. Finally, the lack of predictive biomarkers for patient stratification and absence of robust immunocompetent translational models hinder both clinical trial design and therapeutic refinement. Addressing these challenges is essential to fully unlock the therapeutic potential of CAR-M therapies in the clinical setting.

CAR-M therapies represent a novel and versatile approach to cancer immunotherapy, particularly for the treatment of solid tumours, for which conventional T cell-based strategies have shown limited efficacy. As preclinical data accumulate and clinical trials progress, several key developments are expected to shape the future integration of CAR-Ms into clinical oncology. First, enhancements in cell engineering to improve in vivo persistence, tumour homing, and resistance to immunosuppressive signals are essential for achieving durable and reproducible clinical responses. Emerging strategies include the use of synthetic signalling domains, epigenetic reprogramming, and the incorporation of metabolic or cytokine modules that promote long-lasting M1-like activity within the TME. Secondly, combination therapies are likely to play a central role in maximising the clinical impact of CAR-Ms. The pairing of CAR-Ms with immune checkpoint inhibitors, chemotherapy, radiotherapy, or oncolytic viruses may overcome the mechanisms of adaptive resistance and synergistically enhance antitumour immunity. In particular, the use of CAR-Ms as TME modulators to “prime” tumours for subsequent T-cell engagement is an attractive concept that has been actively explored. Third, innovation in manufacturing platforms is crucial to increasing accessibility. The development of standardised GMP-compliant protocols for rapid autologous production, as well as scalable allogeneic macrophage banks derived from iPSCs or universal donor sources, could significantly reduce the production costs and turnaround time. These advances are particularly relevant in the context of rapidly progressing cancers where treatment delays are critical.

Additionally, future CAR-M designs will likely integrate safety switches, logic gates, and tunable activation systems to enhance the control and minimise on-target off-tumour toxicity. Refining antigen selection based on tumour specificity and heterogeneity as well as incorporating dual-antigen targeting approaches may also improve selectivity and broaden clinical applicability. Beyond technical and biological improvements, the successful translation of CAR-Ms into routine clinical practice will require well-designed randomised trials to demonstrate clear therapeutic benefits over standard-of-care treatments. The identification of predictive biomarkers for patient stratification and response monitoring will be critical for personalised therapy and outcome optimisation. If these scientific, logistical, and regulatory milestones are achieved, CAR-M therapies have the potential to become a valuable addition to the immunotherapeutic arsenal, initially for solid tumours, and potentially expand into haematologic malignancies or non-oncologic indications such as fibrotic or infectious diseases.

Looking ahead, several key areas will be pivotal for advancing CAR-M therapies toward clinical maturity. One critical objective is to enhance CAR-M persistence and maintain a stable M1 pro-inflammatory phenotype in vivo, as these features are essential for sustained anti-tumour activity. Achieving this may require genetic modifications that promote survival signals, resistance to immunosuppressive cues from the tumor microenvironment (TME), and incorporation of cytokine support systems. Equally important is the optimisation of antigen selection and CAR construct design to maximise efficacy, while minimising off-target effects. Dual-antigen targeting and tunable activation systems can improve the specificity and safety profiles, particularly in solid tumours with heterogeneous antigen expression. Combination strategies are emerging as a promising direction. Pairing CAR-Ms with immune checkpoint inhibitors, chemotherapy, or oncolytic viruses may amplify therapeutic responses by addressing multiple layers of tumour immune evasion. CAR-Ms may serve not only as direct cytotoxic agents, but also as TME modulators that facilitate the recruitment and activation of other immune cell types. Refining manufacturing processes on the translational front remains a major challenge. Developing consistent, scalable protocols—whether for rapid autologous production or allogeneic “off-the-shelf” platforms—is essential for broad clinical accessibility.

Finally, the identification of predictive biomarkers for patient selection and response monitoring is vital for personalised application of CAR-M therapies. These biomarkers could guide treatment decisions and help stratify patients most likely to benefit from this approach. Beyond oncology, CAR-M is also being explored for other indications such as infectious diseases, neurodegenerative disorders, and cardiovascular conditions. The unique properties of macrophages may enable diverse therapeutic applications. Although still in the early stages, CAR-M therapy shows promise as a versatile new tool for cancer immunotherapy. Continued optimisation of CAR designs, manufacturing processes, and combination strategies is critical for realising the full potential of this approach. Larger randomised trials are needed to definitively establish the clinical efficacy and role of CAR-M in cancer treatment.