Introduction

Tumour-associated macrophages (TAMs) make an alluring cancer immunotherapy target. For one, their abundance in tumours allows TAMs to greatly influence the interplay between cancer cells and the surrounding tumour microenvironment (TME), resulting in tumour-promoting milieu essential for cancer progression [

TAMs as cancer therapy targets

A plethora of research has highlighted the various ways in which TAMs promote cancer development and progression, providing ample therapeutic opportunities. In short, TAMs act in every stage of the metastatic cascade by promoting primary tumour growth [24], angiogenesis [25], immune evasion [26], invasion [15, 27, 28] and metastatic spread [15, 29,30,31]. These TAM functions develop gradually when TAMs co-evolve with cancer and are based on homoeostasis-promoting functions of healthy macrophages [4]. Further emphasising their therapeutic potential, TAMs promote cancer immunotherapy resistance by limiting T-cell entry into tumours and suppressing anti-tumour T-cell activation [26, 32, 33]. However, TAMs can fight cancer by phagocytosis [34], killing with reactive radicals [35, 36] and activating anti-tumoural immunity via antigen presentation and cytokine secretion [37]. The long co-evolution with cancer cells ultimately dampens these anti-tumoural TAM properties [4], unless prevented with therapeutics.

Considering the tumour-promoting roles of TAMs, it is no surprise that a high abundance of TAMs is associated with poor prognosis in most solid cancers [38]. Colorectal and prostate cancer are exceptions to this, but when immunosuppressive M2 marker-expressing TAMs are quantified instead, a negative prognosis is evident across cancer types [38]. A high abundance of immunosuppressive TAMs is characteristic of so-called non-inflamed tumours that lack T cells and are often resistant to immune checkpoint inhibitor (ICI) therapy, further contributing to poorer prognosis [39]. However, these overall trends give a simplistic view, as the intratumoural localisation of TAMs and selected treatment regimen also affect their prognostic value. For instance, tumour islet TAMs are often associated with a more favourable prognosis than stromal TAMs [47,48,49]. Once monocytes enter the cancer tissue, their phenotype is modified by the cancer type, affected organ and intratumoural localisation, as these determine the characteristics of local tissue niches. Factors such as neighbouring cells, matrix composition, pH and cytokine environment will then further fine-tune TAM phenotypes and functions [10, 26]. Furthermore, monocyte differentiation into macrophages is altered in cancer, resulting in additional accumulation of immature phenotypes, such as MDSCs [50].

Investigating TAM phenotypes in different tumour areas and cancer types is possible using single-cell and spatial analysis technologies. For instance, Ma et al. used existing datasets of single-cell RNA-sequenced TAMs and identified seven TAM phenotypes common to multiple cancers [10]. Interestingly, these identified TAM phenotypes have both pro- and anti-tumoural properties within the same subset, such as interferon (IFN)-primed TAMs that express IFN-regulated mediators contributing to both T-cell activation and exhaustion, and regulatory TAMs that resemble immunosuppressive M2-like macrophages with their PD-L1, IL-10 and MRC1 expression but also express co-stimulatory and major histocompatibility complex (MHC) molecules [10]. As expected, these TAM phenotypes strongly connect TAM functions with their localisation within the tumour. For instance, pro-angiogenic TAMs reside in hypoxic tumour areas, where they promote metastatic spread and angiogenesis [10], whereas immunosuppressive lipid-associated TAMs are located in areas of invasion, where they suppress T-cell responses [10, 51], and IFN-primed TAMs co-localise with CXCL13-expressing T cells to regulate their responses [10].

Clearly, these different and opposing functions of TAM subsets require tailored therapeutic targeting. It is yet to be determined which of these TAM subsets are susceptible to functional reprogramming and which are the most crucial to be reprogrammed. If these TAM subsets possess both anti- and pro-tumoural capacities, successful manipulation would promote their anti-tumoural functions without activating the pro-tumoural ones. Of note, depending on cancer type, the markers for these subsets may differ, and some of the widely investigated TAM molecules are associated with several subsets, such as TREM2 [10]. Nevertheless, uncovering TAM functions at the subset level will provide opportunities for develo** even more sophisticated TAM-reprogramming therapeutic approaches.

Clinical landscape analysis

To obtain a comprehensive understanding of the clinical landscape of TAM-reprogramming strategies, we gathered data from past and present clinical trials investigating TAM-reprogramming agents in cancer. For this, we first searched PubMed for potential TAM targets using the following search conditions: macrophage and (cancer or neoplasm or tumour or tumor or malignancies) and (phase or clinical or trial or target), where ‘macrophage’ and the cancer-related term had to be in the title or abstract. The obtained 17,457 unique articles published before 1.2.2024 were screened by Rayyan [52] using the following exclusion criteria: (1) the paper does not discuss or identify therapeutic targets, specifically for cancer and in macrophages; (2) the therapeutic strategy primarily depletes TAMs or inhibits their recruitment and (3) the therapeutic is re-purposed. Screening was performed mostly manually, with Rayyan-created ratings used to exclude the lowest matching articles only after screening one-third of all articles and confirming the irrelevance of the lowliest rated articles. The identified TAM targets and therapeutics from the obtained 550 review and research articles or the word ‘macrophage’ were then used as the search condition in clinicaltrials.gov to find clinically investigated TAM-reprogramming therapeutics (condition/disease = ‘Cancer’).

During the search, we excluded withdrawn trials and therapeutics with first trial start date before 1.1.2000, and expanded our original TAM target list with additional TAM-targeting therapeutics from the company pipelines encountered. In this way, 194 clinically evaluated TAM-reprogramming therapeutics were gathered with information on the clinical phase, number of trials, combinatory regimens, first clinical trial date, investigated cancer types and actual number of treated patients. Clinical development was classified as discontinued if the therapeutic had been removed from the company’s pipeline without information on selling the asset, most recent trials had been terminated, or trials had remained inactive for several years without new trials being launched. Additional information was collected from publicly available sources, including published conference abstracts, press releases, web-based company pipelines and quarterly reports. While we call these therapeutics ‘TAM reprogramming’ for clarity, many of them are not macrophage-specific and may have other functions, as described in the discussion of individual targets. We also acknowledge the presence of other crucial pathways controlling TAM phenotypes not discussed in this review. For instance, cytokines, prostaglandins and other inflammatory mediators have additional broad effects on various immune and non-immune cell types and represent potential strategies for modulating cancer-associated inflammation rather than specifically controlling TAMs.

Past and present clinical trials

Overall, the past decade has seen a steep increase in the number of TAM-reprogramming therapeutics that have entered clinical trials (Fig. 1a). Therapeutics targeting the CD47–SIRPα axis have particularly contributed to this rapid development. Additionally, several novel macrophage targets have entered the clinics, including members of the LILRB and scavenger receptor families (Fig. 1b).

Fig. 1: Timeline of clinically evaluated macrophage-reprogramming therapeutics.
figure 1

a Number of macrophage-reprogramming therapeutics entering clinical trials each year since 2000, coloured by target group. b A timeline showing when each macrophage-reprogramming therapeutic target was first clinically investigated in cancer. Bubble colours indicate macrophage target groups.

Such development is encouraging because currently only a few macrophage-modifying therapies have been approved for clinical use, the number of therapeutics against the same target is high, and not many programmes have proceeded beyond phase 2 (Fig. 2a). The approved therapeutics are duvelisib, a PI3Kγ and PI3Kδ inhibitor for haematological malignancies [53], and imiquimod, a toll-like receptor 7 (TLR7) agonist for topical treatment of basal cell carcinoma [54]. Additionally, approved early activators of pattern recognition receptors include Bacillus Calmette-Guerin, an attenuated bacteria that activates TLR2 and TLR4 receptors in non-invasive bladder cancer [54], and mifamurtide, a synthetic analogue of bacterial cell wall that activates TLR4 and NOD2 receptors in osteosarcoma [55].

Fig. 2: Clinical landscape of macrophage-reprogramming therapeutics in cancer.
figure 2

a Number of therapeutic agents that have been investigated in clinical trials. Therapeutics are shown by target group and coloured by clinical development phase as of January 2024. b Number of treated patients by target group, coloured by the development status. c Number of therapeutic agents by target group, coloured by development status and highest clinical development phase at the time of discontinuation. Percentages indicate proportions of discontinued therapeutics. d Bar plots of macrophage-reprogramming therapeutics with ongoing clinical development as of January 2024, grouped by molecule type and coloured by clinical development phase.

Phase 3-investigated therapeutics comprise additional TLR agonists, CD47–SIRPα axis blockers, IDO1 inhibitors and STAT inhibitors (Fig. 2a), and we report staggeringly high numbers of patients treated with these agents (Fig. 2b). Unfortunately, these advanced targets show a high proportion of discontinued therapeutics, such as 100% for IDO1, 61% for TLRs and 50% for STAT, indicating non-favourable therapeutic properties (Fig. 2c). Surprisingly, some targets with known safety- or efficacy-related issues are still highly investigated, such as CD47–SIRPα axis, STING and CD40 (Fig. 2c), reflecting novel approaches taken to develop next-generation agents with a more favourable therapeutic profile. We will review the properties of each target separately further below, but commonly poor efficacy is caused by on-target side effects that limit dosing, or activation of compensatory/counteracting pathways that enable tumour immune escape.

Emerging efficacy- and safety-related challenges have motivated a vast amount of translational research. Novel treatment strategies have already been clinically investigated alongside conventional small molecules and monoclonal antibodies (Fig. 2d), and various approaches have been exploited to circumvent the original challenges. To direct treatment effects, alternative delivery routes (intratumoural vs. systemic) are utilised [56,57,58]. To target systemically administered therapeutics to the TME, antibody–drug-conjugates [59, 60], antibodies with pH-dependent target binding [61] and bi-specific antibodies and fusion proteins have been developed [56, 62, 63]. Although most TAM-targeted bi-specifics bind a second target on cancer cells or the TME, some use the second arm to enhance immune activation by engaging PD-1, PD-L1 or 4-1BB instead (Supplementary Table 1). The efficacy of monoclonal antibodies can be modified by re-engineering the Fc region when the efficacy depends on antibody-induced effector functions or cross-linking [56, 64]. Indeed, we observed that most of the clinical candidates carry an IgG1 Fc region, which has been occasionally modified to decrease or enhance effector functions (data not shown). Finally, more complex delivery systems using exosomes [65, 66], bacteria [67] or viral vectors [68, 69] have also been tested in clinics (e.g. NCT04592484, NCT05375604, NCT04167137, NCT03852511, NCT02654938). Overall, some of the above approaches have demonstrated superior efficacy or safety in preclinical studies [63, 70, 71], supporting the advancement of their clinical development.

Because TAM-targeted therapeutics are predicted to support the efficacy of other treatments rather than eradicate cancer on their own [4, 5], we also evaluated the prevalence of combination treatment regimens. Expectedly, less than a quarter of clinically investigated TAM-reprogramming agents have been studied as monotherapy above phase 1, whereas more than half have been studied in combination with ICIs (Fig. 3a).

Fig. 3: TAM-reprogramming treatment targets by combinatory treatment regimen and subcellular localisation.
figure 3

a Dot plot showing how commonly macrophage-reprogramming therapeutics have been investigated in combination with other treatment types. Dot size indicates number and dot colour proportion of macrophage-reprogramming therapeutics investigated with the indicated treatment combinations, separately for each target. ‘Monotherapy phase 1’ indicates therapeutics at phase 1 clinical development not yet investigated in combination with other treatments. Percentages were calculated from all TAM-reprogramming therapeutics (n = 194). b Illustration depicting subcellular localisation of macrophage-reprogramming therapeutic targets. Targets are coloured by clinical development status with coloured areas depicting proportions among therapeutics against the same target.

To summarise, a surge of macrophage-reprogramming therapeutics has proceeded to clinical development during the past decade, and while several older targets have faced challenges, novel strategies have been undertaken to improve their therapeutic profiles.

TAM-reprogramming therapeutic targets

Next, we will discuss each TAM-reprogramming target separately, focussing on the mode-of-action, types of clinically investigated therapeutics and clinical results thus far. Firstly, we describe strategies that alter specific TAM functions, such as phagocytosis, scavenging, pattern recognition or interactions with other immune cells, and then proceed to therapeutic strategies that alter the overall TAM phenotype towards pro-inflammatory direction. Subcellular localisation of each target is illustrated in Fig. 3b. As the majority of these therapeutics have been evaluated in (advanced) solid tumours and haematological malignancies, we will only occasionally describe the investigated cancer types and provide a supplementary trial ID table (Supplementary Table 1) for further reference.

Phagocytosis checkpoints

CD47–SIRPα

To escape phagocytosis, healthy and cancerous cells express CD47, which binds macrophage SIRPα to inhibit cytoskeletal rearrangements necessary for phagocytosis [72, 73]. Since the first-in-class monoclonal CD47-targeting antibody magrolimab [74], ~50 therapeutics blocking the CD47–SIRPα interaction have been clinically investigated. A notable portion of these therapeutics are bi-specific antibodies and fusion proteins that mostly recognise another molecule on cancer cells or inhibit PD-L1/PD-1. Overall, blocking CD47 in haematological cancers and SIRPα in solid tumours yields better efficacy [75], but two phase 3 magrolimab trials (NCT4313881 and NCT04778397) were recently terminated because of poor efficacy in acute myeloid leukaemia and myelodysplastic syndrome [76, 77]. CD47 blockade on red blood cells prevents the therapeutic from reaching cancer cells and causes anaemia by inducing red blood cell phagocytosis and agglutination. To circumvent this, newer strategies use antibodies that preferentially bind to cancer cells or bi-specific agents to target cancer cells [63, 75]. Furthermore, enhanced phagocytosis can promote either pro- or anti-inflammatory reprogramming of TAMs, depending on signals from the phagocytosed cells and type of phagocytosis. Anti-inflammatory responses would limit monotherapy efficacy by promoting immunosuppression and tumour growth [78].

CD24–Siglec-10

Analogous to CD47–SIRPα interaction, CD24 on cancer cells inhibits phagocytosis by binding to macrophage Siglec-10 [79]. CD24-targeting antibodies of the IgG1 subclass (IMM47, ATG-031) are investigated in phase 1 trials, and at least preclinical IMM47 efficacy depends on its Fc-induced effector functions [80]. Further development should consider other immune- and cancer-related functions of CD24 and additional CD24 and Siglec-10 ligands binding to different glycosylated protein forms [79, 81, 82].

Scavenger receptors

Macrophages clear various endogenous and pathogen-related ligands with scavenger receptors that also regulate subsequent immune responses to these ligands [83, 84]. Apart from clinically investigated CD163 and Clever-1, preclinical studies have identified MARCO [85, 86] and CD206 [87] as potential therapeutic targets. Inhibiting MARCO with the monoclonal antibody PY265 supports anti-tumour immunity via pro-inflammatory conversion of TAMs and MDSCs [88], and specific CD206-targeted peptides can likewise support M1-like macrophage phenotype or deliver therapeutics to CD206-expressing TAMs [87, 89].

CD163

Under homoeostasis, CD163 scavenges haptoglobin–haemoglobin complexes [90], and in cancer, CD163 is a widely known marker for tumour-promoting M2-like macrophages, generally associated with poor prognosis [83, 91, 92]. CD163 in TAMs is associated with STAT3 activation and anti-inflammatory IL-10 and TGF-β secretion [93]. A single CD163-targeting antibody (OR2805, IgG1 subclass), which activates T-cell responses in preclinical models [94], is being clinically evaluated in a phase 1/2 trial (NCT05094804).

Clever-1

Clever-1 (Stabilin-1) is a scavenger and adhesion molecule expressed by monocytes, macrophages and endothelial cells [95]. It regulates macrophage lysosomal acidification to halt antigen presentation and T-cell activation [96]. A Clever-1-blocking antibody, bexmarilimab (IgG4), activates T-cells and IFN responses in advanced solid cancers and elicits cancer type-dependent disease control alongside few objective responses in a phase 1/2 trial [96, 97]. Promising objective responses have been observed in combination with azacytidine in acute myeloid leukaemia and myelodysplastic syndrome in a phase 1/2 trial (NCT05428969) that is also recruiting azacytidine-refractory patients [98].

Pattern recognition

Pattern recognition receptors alert innate immunity to pathogens and tissue damage, but their systemic administration is often limited by the associated side effects of overt immune activation [99].

Toll-like receptors (TLRs)

TLRs are widely expressed, but their stimulation on antigen-presenting cells activates pro-inflammatory cytokine secretion, expression of co-stimulatory molecules and antigen presentation to support T-cell activation [100,101,102,103]. Depending on the TLR type, the activating therapeutics are either small molecules or larger lipid or nucleic acid derivatives that can be administered systemically or locally [58, 100]. Several therapeutics have faced discontinuation after showing weak monotherapy efficacy [101], possibly due to the induction of tolerance and simultaneous activation of pro-tumoural pathways that support regulatory T cells (Tregs), MDSCs and cancer cell proliferation [104,105,106]. Mostly, these agents are now deemed as boosters for other therapeutics, especially cancer vaccines and ICIs [102]. As an exception, the TLR7 agonist imiquimod and the attenuated TLR2- and TLR4-activating Bacillus Calmette-Guerin bacteria have been approved for skin carcinoma and non-muscle invasive bladder carcinoma, respectively [54].

STING

Upon recognising cytosolic pathogen-derived or damaged endogenous DNA, cGAS produces cyclic GMP–AMP, which activates STING to elicit type I IFN production and NF-κB activation. These pathways promote cancer cell death, anti-inflammatory macrophage polarisation, antigen presentation, T-cell priming and recruitment [107, 108]. However, chronic STING activation may have undesired opposite effects, such as enhanced cancer cell survival and immunosuppressive IDO1 induction [108]. Nevertheless, at least twenty STING-activating therapeutics have been clinically evaluated in phase 1 and 2 trials. Early STING agonists were intratumourally administered synthetic cyclic dinucleotides (e.g. ADU-S100, MK-1454) characterised by insufficient monotherapy efficacy, limited penetration inside the cells and susceptibility to enzymatic degradation [57, 109, 110]. Therefore, alternative delivery methods, such as liposomal formulations, nanoparticles, antibody–drug-conjugates (TAK-500, XMT-2056), exosomes (exoSTING: NCT04592484) and a bacterial vector (SYNB1891: NCT4167137) are being developed [109]. Additionally, compounds that enable intravenous administration (SB11285: NCT04096638) and a polymer that prolongs STING activation (ONM-501: NCT06022029) are under clinical investigation.

HMGB1

As a DNA-binding protein, HMGB1 supports chromatin organisation and transcription, but extracellular HMGB1 released from dying cells and activated macrophages stimulates inflammatory responses via TLRs and RAGE [111, 112]. In cancer, some HMGB1-activated pathways aggravate prognosis by supporting invasion and metastasis [112, 113]. The HMGB1-binding prodrug SB17170 modulates myeloid cell cytokine secretion to increase T-cell infiltration [114], and it is being evaluated in a phase 1 trial (NCT05522868).

Dectin-2

Dectin-2 (CLEC6A) defends against fungi and mycobacteria [115, 116]. Its activation stimulates the secretion of pro-inflammatory chemokines and cytokines, including TNFα and IL-12 [116, 106, 152, 192], tendency to alter TAMs rather than healthy tissue macrophages [193], lack of compensatory pathways in the TME [149] and access for therapeutic manipulation within solid tumours [57]. To direct therapeutic effects away from macrophages in healthy tissues, therapeutic efficacy could depend on tumour-specific expression, ongoing chronic inflammation at tumour sites or monocyte-derived macrophages. Enhancing specificity with such TME-specific targets or TME-directed delivery mechanisms could potentially widen the therapeutic window that is currently limited by side effects [22, 56, 63, 194]. Alternatively, targeting circulating monocytes can affect their differentiation into TAMs and therefore provide access within solid tumours without the need for direct tissue penetration.

Fundamental understanding of TAM biology will pave the way to more effective treatment approaches [4, 5, 195]. Unfortunately, TAM subset complexity revealed by single-cell RNA sequencing [Full size table