Immunotherapy is one of the promising methods for the effective treatment of many types of cancer. There are several strategies for cancer immunotherapy: oncolytic viral therapy, adoptive cell therapy, therapy with immune checkpoint inhibitors (ICIs), and therapy based on key cytokines and chemokines of the tumor microenvironment (TME) [1]. For decades, immunotherapy was a concept primarily based on the enhancement of cytotoxic immunity, where T cells and NK cells are considered the major anti-cancer immune cell subsets [2–7]. However, a number of promising and efficient approaches in animal studies failed in clinical trials. Only a few therapeutic strategies that came to clinical trials demonstrate effectiveness in a particular group of patients. Currently, the most challenging issue is to identify criteria discriminating responders from non-responders in order to efficiently apply expensive immunotherapy only to those patients who will benefit with high probability.

Limitations associated with the lack of efficiency of immunotherapy include components of the TME, the microbiome, intratumor heterogeneity, and tumor mutational load [8]. TAMs are the most abundant immune cell population in the TME and represent a promising therapeutic target for improving tumor resistance to immunotherapy [9]. The role of tumor-associated macrophages (TAMs) at each stage of solid tumor progression was recognized and demonstrated in multiple animal models. In cohorts of cancer patients, the correlation of TAM phenotypes with poor prognosis and poor responses to conventional anti-cancer therapies was found [10]. Surprisingly, in contrast to other cancer types, the total number of TAMs in colorectal cancer (CRC) correlates with a favourable prognosis. However, M2-like macrophages were indicative of poor prognosis in multiple CRC cohorts [11]. Low-grade glycolysis was found to provide the metabolic conditions for TAMs to support tumor progression [12, 13].

Macrophages are conventionally classified into two states of polarization: M1-like macrophages and M2-like macrophages [14]. M1-like macrophages, or classically activated macrophages, possess antimicrobial, pro-inflammatory, and antigen-presenting properties, while M2-like macrophages, also known as alternatively activated macrophages, exhibit anti-inflammatory, pro-tumor, and antimicrobial properties [15]. Macrophages can be repolarized under diverse factors secreted by immune cells, tumor cells, microorganisms, and other components of the tissue microenvironment [15].

Macrophages in the TME are primarily recruited from circulating bone marrow-derived monocytes and differentiate into TAMs [16, 17]. For a long time it was believed that TAMs have an M2-like phenotype expressing major pro-tumor molecules, e.g., mannose receptors (CD206), scavenger receptors (CD163), vascular endothelial growth factor (VEGF), interleukin (IL)-10, C-C motif chemokine receptor 2 (CCR2), CD204, and C-C motif chemokine ligand 22 (CCL22) [18, 19] However, recent single cell and spatial transcriptomic data allowed identifying that pro-tumoral TAMs have a mixed M1/M2-like signature, and the origin of TAMs is critical for their pro-tumoral activation [10]. Functional activity of TAMs is formed by transcriptional, epigenetic, and metabolic reprogramming [20]. The activation state of TAMs becomes an effective predictor for the immunotherapy used. Moreover, TAMs are an attractive target for the development of new immunotherapy. However, multiple clinical trials with agents targeting viability, differentiation, recruitment, or activation state of TAMs failed [10]. Despite this, more and more TAM-based approaches are under development to overcome resistance to conventional anti-cancer therapy.

In our review, we combined recent data on the state-of-the-art in the progress and challenges in immunotherapy of CRC, highlighting novel TAM-based approaches and potential clinical significance of TAMs for the efficiency of existing immunotherapy with ICIs. For this purpose, we focused on the relevant articles containing major research terms including “colorectal cancer”, “tumor-associated macrophages”, “immunotherapy”, “immune checkpoint inhibitors”, and “targeted therapy”. The review elucidates the mechanisms of the synergistic effects of combined therapeutic schemas where TAMs are used to enhance the effects of ICIs.

CRC is the third most common cancer and the second most deadly cancer among all cancer types [21]. The number of new cases of CRC is projected to increase from 1.93 million cases per year to 3.2 million by 2040 [22]. CRC usually begins as a benign tumor, which develops into a malignant one within 10–20 years [23].

CRC is classified according to the CpG island methylation phenotype (CIMP) and microsatellite instability (MSI) status, and is divided into four molecular subtypes: CIMP⁺/MSI⁺, CIMP⁺/MSI^–, CIMP^–/MSI⁺, and CIMP^–/MSI^– [24]. It was shown that the number of intraepithelial TAMs detected by the pan-macrophage marker CD68, as well as the number of the intraepithelial marker of M2-like macrophages detected by CD163, was elevated in patients with high MSI (MSI-H) and high CIMP levels [25].

A classification of CRC based on genomic and transcriptional profiling data was proposed in 2016, and includes four consensus molecular subtypes (CMSs): CMS1 (MSI immune), CMS2 (canonical), CMS3 (metabolic), and CMS4 (mesenchymal) [26]. CMS1 tumors are hypermutated and characterized by MSI, high levels of CIMP cluster, the presence of BRAF mutations, and diffuse infiltration of immune cells (primarily T-helper 1 and cytotoxic T cells). The CMS2 tumors are characterized by the activation of downstream targets of the WNT and MYC pathways, as well as a loss of tumor suppressor genes and an increase in the number of oncogene copies. KRAS mutations and metabolic deregulation predominate in CMS3 tumors. CMS4 tumors have increased expression of proteins involved in various oncogenesis pathways: epithelial-mesenchymal transition (EMT), activation of the transforming growth factor β (TGF-β) signaling pathway, angiogenesis, matrix remodeling pathways, and the inflammatory complement system [26]. Among all CRC subtypes, CMS4 tumors are highly infiltrated with CD68⁺ macrophages, as well as myeloid cells, stem cells, and stromal clusters, enriched with the M2-like CD163⁺ subset of macrophages [27, 28]. Transcriptional profiling revealed that CMS4 CRC tumors were highly enriched in signatures associated with altered macrophage activation and downregulation of phagocytosis [28]. CRC metastases to the lung and liver are usually classified as CMS2. However, peritoneal metastases are characterized by CMS1, CMS2, and CMS4 signatures. CMS4 peritoneal metastases are enriched in macrophages (CD163⁺ and CD206⁺), cancer-associated fibroblasts (CAFs), and endothelial cells, as well as have increased expression of the immune checkpoint molecules (TIGIT and PD-L2) [29].

The term CRC encompasses two types: colon cancer (CC) and rectal cancer (RC). However, accumulated data suggest that they should be considered as two separate diseases [30, 31]. CC and RC differ in anatomy and topography, epidemiology, risk of carcinogenesis, histology, as well as symptoms, primary prevention, and treatment [30, 31]. Immune-associated differences were found for these two histological locations. Compared to RC, in CC, the neutrophil-to-lymphocyte ratio and platelet-to-lymphocyte ratio, as well as the systemic inflammatory response index, systemic immune inflammation index, and total systemic inflammation index, were significantly increased. At the same time, blood monocyte levels did not differ significantly between patients with CC and RC [32]. CD3⁺ and CD8⁺ T lymphocyte infiltration in the tumor nest and CD3⁺ lymphocytes in the stroma of CC, but not RC tumors, correlated with better overall survival [33]. High tumor infiltration with NK cells, macrophages, and CD4⁺ T cells correlated with a low probability of local recurrence in RC [34]. High numbers of both peritumor and intratumor macrophages and T cells (CD3⁺, CD8⁺) correlated with a low probability of distant recurrence in RC [34]. Our recent study demonstrated an increased number of peripheral blood monocytes and an accumulation of intratumor TAMs expressing glycolysis activator PFKFB3 in patients with CC compared with those with RC [12]. We demonstrated that PFKFB3 may induce changes in lipid and amino acid metabolism in macrophages, but it had a dual effect in the context of pro- and anti-tumor effects [13]. Differences in gene expression (such as S100P, Reg family, ACTN1, CAMK2G, ACAT1) and signaling pathways (metabolic pathway, cell cycle pathway, p53 pathway) between CC and RC have also been demonstrated [35, 36]. However, non-hypermutated CC and RC tumors cannot be distinguished at the genomic level [37]. Right-sided and left-sided СС are distinct in genomics. Right/ascending colon is more often hypermethylated and has elevated mutation rates [37]. Right-sided CC had MSI-H and gene mutations in BRAF, KRAS, SMAD4, TGF-β, PIK3CA, PTEN, and AKT1 compared to left-sided CC. No significant differences were observed in HRAS, NRAS, APC, TP53, and FBXW7 between the right-sided CC and left-sided RC groups [38]. Tumors with a high mutational load and overexpression of tumor neoantigens are more sensitive to immunotherapy. Approximately 80–85% of patients with CRC are unresponsive to ICIs because they have “cold” tumors with microsatellite stable (MSS) or low MSI (MSI-L) [39]. “Cold” tumors often harbor immunosuppressive cell populations, such as TAMs, regulatory T cells (Tregs), and myeloid-derived suppressor cells (MDSCs), which can potentially be targets for immunotherapy to induce tumor cold-to-hot transition [40]. Both deficient mismatch repair (dMMR) and MSI-H tumors can provide de novo tumor antigens caused by accumulating mutations, leading to tumor immunogenicity. In contrast, mismatch repair proficient (pMMR) CRC cells exhibit weak immunogenicity and are resistant to immunotherapy [41]. Tumors with MSI-H typically respond well to ICIs, including antibodies against programmed cell death-1 (PD-1), programmed cell death ligand-1 (PD-L1), and cytotoxic T lymphocyte-associated antigen-4 (CTLA-4). However, in MSI tumors, immunotherapy may also be ineffective due to the reprogramming of key metabolic pathways in CRC. This results in the formation of by-products that affect the immune response, reducing the anti-tumor activity of immune cells and contributing to resistance to ICIs [42, 43].

The development of CRC is a multi-stage process. Focal changes in the intestinal mucosa lead to the formation of benign polyps, followed by the active proliferation of cells that can eventually acquire malignant properties [44]. Chronic inflammation is one of the causes of CRC. Chronic inflammation may be associated with changes in the levels of inflammatory markers in immune and tumor cells [45]. Furthermore, chronic inflammation facilitates DNA damage, resulting in the activation of pro-oncogenes and the inactivation of tumor suppressor genes [46].

The inflammatory process involves various immune cells, such as macrophages, lymphocytes, as well as proinflammatory factors—tumor necrosis factor (TNF)-α, IL-6, and IL-1β, and reactive oxygen species [47, 48]. Immune cells play a key role in facilitating the transition from inflammation to carcinogenesis [47]. In the early stages of CRC development, M1-like macrophages play a key role, creating an inflammatory environment that promotes mutational changes [49, 50]. The transcription factor (TF) NF-κB and the inflammatory cytokine TNF-α are key regulators of the inflammatory process in the intestinal mucosa. Specifically, activation of the NF-κB pathway promotes the polarization of pro-inflammatory M1-like macrophages [51]. Moreover, epithelial MUC1 induces CCL2 expression and mediates macrophage recruitment and activation [49]. Macrophages secrete the pro-inflammatory cytokine IL-1β, TNF-α, and IL-6, and may promote the transformation of cells into a stem cell state, leading to dysplasia and colitis-associated tumorigenesis [52]. After cell transformation, peripheral blood monocytes are attracted to the tumor site, secreting growth factors and chemokines (CCL2, CCL5, VEGF, and TGF-β) [50]. Moreover, CRC cells secrete EGF to alter macrophage polarization toward M2-like, which, at the late stages of tumor development, promotes tumor immune evasion and contributes to cancer progression [49, 50].

TAMs create a favorable environment for tumor cell growth, proliferation, invasion, and metastasis [11]. Accumulated experimental data demonstrated contradictory pro- and anti-tumor activities of TAMs in CRC [11]. The number of macrophages determined by the general marker CD68 positively correlated with a favorable prognosis in CRC [53–56]. Macrophages with M1-like and M2-like phenotypes were simultaneously present at the front of tumor invasion, and an increase in the total number of macrophages was associated with a better prognosis [57]. However, the predominance of M2-like TAMs (CD163⁺, CD206⁺, SPP1⁺, and Stabilin-1⁺) in CRC tumors, on the contrary, correlated with an unfavorable prognosis and disease progression [58–62]. In vitro studies showed that tumor factors derived from CRC cell lines did not affect the shift in the polarization of M1-like macrophages towards M2-like, but, on the contrary, increased the expression of the M1-like marker CD86 and decreased the expression of the M2-like marker CD163 [63]. Macrophages stimulated with conditioned media from CRC cell lines (HT-29 or HCT116) had higher levels of expression of both M2-like markers CD163 and IL-10, and M1-like markers IL-1β, interferon (IFN)-γ, and TNF-α [58]. Exosomes secreted by tumor cells can promote mixed polarization of macrophages. Exosomal vesicles from the SW620 cell line induced a mixed pattern of secretion of M1-like [C-X-C motif chemokine ligand 10 (CXCL10), IL-6, IL-23] and M2-like (IL-10) cytokines in inactive M0-like macrophages [64].

Thus, the high plasticity of macrophages and their complex role in tumor progression make them an ideal target for the development of immunotherapeutic approaches to increase the effectiveness of the existing CRC therapies [65].

The standard of care for CRC includes surgery, chemotherapy, radiotherapy, targeted therapy, and immunotherapy (Figure 1). Surgical intervention is the primary method of radical treatment for CRC; in early stages, it is used alone, while in locally advanced tumors it is combined with neoadjuvant or adjuvant therapy [66, 67]. In patients with RC, treatment tactics depend on the extent and location of the tumor process: if the upper third of the rectum is affected, radical surgery or neoadjuvant chemotherapy (FOLFOX/XELOX) followed by surgery is performed; for cancer of the mid and lower third of the rectum, radiation therapy (RT) or chemoradiotherapy (CRT) with total neoadjuvant therapy and surgical treatment are used. In patients with CC and locally advanced resectable tumors, neoadjuvant FOLFOX/XELOX chemotherapy and radical surgery are performed. For stages 2 and 3 CRC, immunotherapy with ICIs as neoadjuvant therapy is usually used for patients with dMMR or MSI-H tumors. For metastatic CRC (mCRC), various CRT, targeted therapy, or immunotherapy regimens are used depending on the patient`s condition and the effectiveness of previous treatment [66, 67].

Immunotherapy has revolutionized cancer treatment by introducing novel mechanisms that engage the host’s immune system against tumor cells [68, 69]. Immunotherapy efficacy largely depends on specific molecular features of the tumor, in particular on MMR status.

ICIs represent the most clinically established form of immunotherapy in CRC [70, 71]. Monoclonal antibodies targeting the PD-1/PD-L1 axis (e.g., pembrolizumab or nivolumab), as well as those targeting CTLA-4 (ipilimumab), are approved for dMMR/MSI-H mCRC [68, 69]. The high neoantigen burden in these tumors, comprised of mutant proteins unique to cancer cells, enhances their immunogenicity, making them more susceptible to ICIs [72]. For such tumors, ICIs can yield durable responses and are considered standard therapeutic options [69].

In 2014, the first anti-PD-1/PD-L1 drug pembrolizumab, was approved by the FDA for the second-line treatment of advanced malignant melanoma. Subsequently, it was approved for mCRC second-line treatment in 2017 [41]. Ipilimumab plus nivolumab was approved by the FDA for first-line treatment of unresectable or dMMR/MSI-H mCRC [73].

The insufficient effectiveness of existing immunotherapy (ICIs) for treating CRC has prompted a search for novel therapeutic approaches based on either combination therapy or monotherapy. A number of studies are currently underway in preclinical in vitro and in vivo models (Table 1, Figure 2), as well as in clinical trials.

The overwhelming majority of combination therapies presented in Table 1 demonstrated acceptable safety profiles in animal tumor models. No toxicity [100, 101], weight loss [93, 98, 101–105], and no histological, morphological, or functional organ changes [93, 101] were observed.

Currently, nivolumab, pembrolizumab, dostarlimab, and ipilimumab are approved for the treatment of CRC [68, 69]. Other immunotherapy drugs, such as AMP-224, atezolizumab, avelumab, camrelizumab, durvalumab, envafolimab, sintilimab, spartalizumab, tislelizumab, and toripalimab, did not show significant anti-tumor effects for CRC treatment [68].

Ongoing clinical trials demonstrate that TAMs can be an indicator of response to immunotherapeutic treatment in CRC (Table 2).

The Phase Ib REGONIVO and Phase Ib TASNIVO clinical trials for MSS or pMMR CRC evaluated the clinical efficacy of regorafenib plus nivolumab and pimitespib (HSP90 inhibitor TAS-116) plus nivolumab, respectively [117]. Objective tumor response was observed in 36% of cases in the REGONIVO group and 16% of cases in the TASNIVO group [117]. Analysis of pre-treatment tumor immune infiltrates showed that in the REGONIVO group, the density of CD206⁺CD11b⁺ M2-like macrophages was significantly higher in responders compared to non-responders. In the TASNIVO group, the opposite results were observed: the density of M2-like macrophages was significantly lower in responders than in non-responders [117]. Another study comprehensively evaluated tumor tissue from patients with hypermutated CRC treated with pembrolizumab (KEYNOTE 177 clinical trial) or nivolumab (non-clinical trial patients) to identify tumor cell and immune cell interactions in response to immunotherapy [118]. Higher numbers of CD68⁺CD74⁺ cells were observed in CRC tumors with durable response (DB-CRC) than in CRC tumors without durable response (nDB-CRC). Interestingly, 80% of CD68⁺CD74⁺ cells expressed markers of antigen presentation (HLA-ABC, HLA-DR, CD40, CD16, and CD163) [118]. Responsive tumors contained abundant infiltration of PD-L1⁺ macrophages in close proximity to PD-1⁺ cytotoxic T cells. CD68⁺CD74⁺ macrophages with antigen-presenting ability served as a key predictor of treatment response, highlighting their critical role in mediating pembrolizumab’s therapeutic effect in the TME [118]. In a phase I clinical trial evaluating combination therapy with oncolytic virotherapy and anti-PD-1 antibodies, only one out of four patients demonstrated a durable response [99]. A comprehensive multi-omics analysis revealed that the patient who responded to the combination therapy had low infiltration of T cells and NK cell subsets, but a higher number of macrophages compared to non-responders. Interestingly, the responder had elevated levels of both M1-like and M2-like macrophages [99].

In a Phase Ib/II clinical trial evaluating the efficacy of pembrolizumab in combination with XL888 (an Hsp90 inhibitor) in patients with CRC, metastatic liver biopsies were analyzed [119]. Using multiplex immunohistochemistry, the authors found that patients receiving the combination therapy, compared with those receiving pembrolizumab alone, had a trend toward decreased numbers of CD68⁺ and CD68⁺IL6⁺ macrophages. However, no ORR was observed in the group treated with combination therapy. The best result was achieved in 25% of patients, who had stable disease [119]. In advanced MSS/pMMR CRC (N = 110) after first-line treatment failure, treatment with sintilimab (a PD-1 inhibitor) plus bevacizumab (the study was not registered) increased CD8⁺ T cell infiltration and reduced the numbers of TAMs and CAFs. At the same time, PFS rates in the sintilimab plus bevacizumab treatment group were significantly higher than in the FOLFIRI plus bevacizumab treatment group. The experimental group also had statistically significantly higher rates of partial responses [90, 120]. The Phase 1 PICCASSO study investigated the efficacy of a combination of pembrolizumab and maraviroc (a CCR5 inhibitor) in refractory pMMR CRC [121]. CCR5 regulates macrophage polarization toward the pro-tumor M2-like phenotype, and CCR5 inhibition led to anti-tumor macrophage activation. Only one patient achieved a partial response; in the majority of patients (94.7%), the best response was disease progression during the treatment, but after the treatment, patients demonstrated superior response rates [121]. The phase II CAPability-01 trial in unresectable locally advanced or MSS/pMMR mCRC demonstrated that chidamide combined with sintilimab (anti-PD-1) with or without bevacizumab (anti-VEGF) converted immunosuppressive microenvironments into immunoactive states [122]. Patients receiving triple combination therapy (chidamide, sintilimab, and bevacizumab) compared with patients receiving double combination therapy (chidamide, sintilimab) showed a higher 18-week PFS (64.0% vs. 21.7%) and ORR (44.0% vs. 13.0%). At the same time, in patients who responded to triple therapy, an increase in the number of monocytic lineage cells was observed in their tumors [122].

All types of therapy demonstrated an acceptable safety profile. The most common adverse events were grade 3, and in studies with pembrolizumab [123], pembrolizumab plus XL888 [119], and pembrolizumab plus maraviroc [121]—grades 4 and 5. The incidence of some adverse event grade ≥ 3 did not exceed 28% when regorafenib plus nivolumab [124], pimitespib plus nivolumab [125], pembrolizumab [123], pembrolizumab plus maraviroc [121], pembrolizumab in combination with XL888 [119], and sintilimab plus chidamide with or without bevacizumab [122] were used.

Accumulating evidence from immunopathological studies, pre-clinical models, and ongoing clinical trials highlights the key role of TAMs in shaping the response to immunotherapy in CRC. However, the multifaceted phenotype and plasticity of TAMs across patients and disease stages hinder the identification of effective immunotherapeutic targets and the optimization of current treatment regimens. In this context, advanced omics technologies, combined with state-of-the-art bioinformatics approaches, provide a powerful framework for dissecting functional programs in TAMs and prioritizing novel immunotherapeutic targets.

In particular, advances in the omics era have enabled whole-transcriptome profiling of TAMs. Differential gene expression analysis of bulk RNA-seq data allows identification of dozens of differentially expressed genes (DEGs) [126, 127]. Gene set enrichment analysis (GSEA) facilitates functional dissection of DEGs based on their known involvement in distinct biological pathways [128]. GSEA helps prioritize TAM-associated genes involved in immunoregulatory processes using functional annotations from public databases such as Gene Ontology [129], KEGG [130], Reactome [131], and CellMarker 2.0 [132]. Gene network analysis provides a framework for further prioritization of DEGs. The WGCNA method [133] constructs gene co-expression networks using correlation patterns among a list of variable genes. Further analysis of the resulting co-expression networks identifies gene modules—clusters of co-expressed genes—and their corresponding hub genes, defined as the most highly interconnected genes within each module. The relevance of WGCNA is highlighted by its ability to identify disease-associated gene modules and hub genes related to clinical outcomes, immune states, and metabolic dysfunction across diverse pathological contexts, including cancer [134–136]. Importantly, functional perturbation of WGCNA-identified hub genes represents a promising strategy for target discovery, as silencing of such central regulatory nodes enables assessment of their causal contribution to disease-associated programs and highlights candidates with potential therapeutic relevance [137, 138].

Several studies have identified gene modules associated with immunosuppression in TAMs, demonstrating that WGCNA-derived networks capture functionally relevant immune programs. For example, immunosuppressive modules enriched for hypoxia- and angiogenesis-related genes were linked to poor prognosis in glioblastoma and characterized by elevated expression of TAM-associated regulators such as TREM1, whose functional perturbation attenuated tumor-promoting properties of TAMs [139]. Similarly, module-based immune scoring approaches in glioma and hepatocellular carcinoma revealed coordinated expression of immune checkpoint molecules, inflammatory mediators, and M2-like macrophage markers, e.g., CD163, S100A9, and SPP1, highlighting modular immune states associated with tumor progression, immune evasion, and unfavorable patient outcome [140, 141]. Furthermore, several independent studies have reported immune-related modules in circulating monocytes. In triple-negative breast cancer, an immunosuppressive module composed of genes such as CD163, S100A9, TREM1, and FCN1 was found to be correlated with an unfavorable response to neoadjuvant chemotherapy [142]. We previously identified diverse co-expression modules in high-grade ovarian cancer following neoadjuvant chemotherapy, including circulating monocyte modules characterized by activated antigen processing and presentation, as well as suppression-related gene programs involving RUNX1 and TREM1 [143]. Altogether, WGCNA-derived co-expression modules indicate a significant skewing of monocytes toward a suppressive state already at the circulating stage after chemotherapy.

Single-cell omics approaches greatly expand our understanding of the complex phenotype of TAMs [144]. Single-cell multi-omics technologies have enabled reaching an unprecedented level of profiling, extending the reconstruction of gene co-expression networks to gene regulatory networks. The SCENIC method integrates transcriptome and chromatin accessibility to construct gene regulatory networks that facilitate the detection of key TFs [145]. SCENIC has been applied across diverse disease contexts to infer TF-driven macrophage programs. For example, SCENIC-based analysis revealed regulatory similarities between TAMs and cirrhosis-associated macrophages, enabling construction of a macrophage-naive CD4⁺ T cell-related transcriptional score derived from target genes such as BOD1, SEC61A1, RHEB, CFL1, PTMA, C1orf109, and E2F5 [146]. In oral squamous cell carcinoma, SCENIC further resolved heterogeneous CD163⁺ macrophage subpopulations with distinct immunoregulatory programs, identifying CMKLR1⁺ macrophages as a key subpopulation that inhibits tumor progression [147]. Further coupling of gene regulatory networks with AI-based approaches has enabled in silico TF perturbation analysis. The CellOracle method enables the identification of transcriptional shifts and the prediction of alterations in differentiation and activation trajectories of cells in response to in silico inactivation of key TFs [148]. In silico perturbation analysis using CellOracle has shown consistency with subsequent experimental silencing, as illustrated by predictions that disruption of FOSL1 attenuates migratory transcriptional programs, which were experimentally validated by reduced keratinocyte motility upon FOSL1 siRNA knockdown [149]. Moreover, in macrophages, CellOracle in silico perturbation was supported by a preprint demonstrating that RUNX1 silencing reprograms cells toward a reparative phenotype and promotes cardiac recovery [150]. Collectively, the above-mentioned findings highlight that the profound plasticity of TAMs necessitates integrative single-cell multi-omics and computational bioinformatics frameworks to reliably identify and functionally validate new immunoregulatory targets.

Accumulated data demonstrate the undoubted role of TAMs in tumor progression and have established TAMs as important determinants of the response to immunotherapy in CRC. However, most studies to date have focused on correlative associations rather than mechanisms, leaving a gap in our understanding of how specific TAM subsets directly modulate ICI responses in CRC.

Key findings obtained using omics technologies and state-of-the-art molecular genetic methods have expanded our understanding of the functional diversity of TAMs in human tumors. The most recent studies identified between 6 and 23 macrophage subsets of TAMs, each with its own functional activity: immunoregulatory TAMs, lipid-associated TAMs, pro-angiogenic TAMs, inflammatory TAMs, immunostimulatory macrophages, and others [144, 151–153]. Nonetheless, the functional relevance of many of these subsets remains largely inferred from transcriptomic signatures, with limited validation at the protein level or in functional models, raising concerns about whether they represent sustainable or transient activation states.

Moreover, current research is still limited by the high heterogeneity of TAMs in tumors, necessitating the search for at least a unified method for macrophage typing in tumors. Critically, most biomarker studies rely on single-time point biopsies, ignoring the dynamic plasticity of macrophages under therapeutic pressure. This static view may explain why candidate biomarkers such as CD163 or CD206 have failed to consistently predict immunotherapy outcomes in CRC. In this context, many earlier studies examined the functional activity of TAMs using a limited number of markers, such as F4/80, CD206, CD163, MHCII, IFN-γ, and iNOS, underscoring the need for more comprehensive approaches. What is more important, a novel insight emerging from recent studies is that TAM heterogeneity does not simply exist between diverse tumor types, but is also intratumoral and spatially organized. It means that the location of a TAM subset relative to tumor-immune interfaces may be as critical as its polarization state. Future biomarker search must therefore consider spatial resolution and dynamic changes in TAM functional states to capture clinically relevant TAM phenotypes.

The complexity of the mechanisms regulating the formation of functional phenotypes of macrophages requires the development of a multi-target approach aimed at TAMs. That is why the clinical translation of TAM-targeting strategies (e.g., CSF-1R inhibitors, CCR2 antagonists) has demonstrated modest results in CRC, largely due to compensatory recruitment of macrophages in the TME, off-target effects, and the absence of predictive biomarkers for patient selection. Moreover, the high plasticity of macrophages in response to various stimuli contributes to the formation of heterogeneity between patients [154], which requires patient stratification based on macrophage biomarkers, coupled with the development of an individualized macrophage-based immunotherapy approach. Combination therapy is the future direction, but clinical translation still requires more precise stratification. A critical inference is that combination therapies should move beyond simple “reprogramming” or “depletion” paradigms. Instead, they must be tailored to the dominant TAM subset present in each patient, guided by real-time immune monitoring. Without such stratification, even rationally designed combinations are likely to fail.

To achieve such a level of precision, appropriate preclinical models are essential. The use of mouse models for research purposes has a number of advantages: small size, rapid reproduction, low cost, ease of carrying out genetic modifications, and, as a result, the possibility of conducting numerous studies in a relatively short period [155]. However, differences in size, physiology, and target homology between humans and mice lead to limitations in the use and interpretation of specific therapies [155]. When using mouse models, a number of factors must be considered to effectively evaluate therapy. Pre-clinical models involve subcutaneous, orthotopic, and xenograft mouse tumors. The choice of mouse model is important for assessing marker expression and the effectiveness of drug therapy [156–159]. Tumor cell morphology, drug uptake, and gene expression may differ depending on the model chosen [156–159]. Orthotopic human tumor xenografts are considered more suitable for predicting drug response. In contrast, genetically modified mouse strains are more suitable for studying the role of genes in tumor development and progression [160]. Furthermore, differences in the outcomes of pre-clinical versus clinical trials of immunotargeted drugs may stem from differences in the TME, particularly immune cells. Today, the optimal approach is to use humanized mouse models, which are physiologically and pathologically the best to mimic human tumors [161]. Such models most accurately reflect the interactions between immune and tumor cells. However, even humanized models often fail to recapitulate the full spectrum of TAM heterogeneity seen in patients, as they are typically engrafted with a limited set of immune lineages. Future models should integrate patient-derived tumor fragments with autologous immune cells to preserve the native TAM diversity and spatial architecture. Taken together, integrating multi-omics analyses with functionally annotated preclinical models, including humanized mice, will be essential to develop effective strategies for TAM programming at the transcriptional, epigenetic, and metabolic levels.

In summary, macrophage plasticity, the choice of study model, the choice of markers of therapy effectiveness, differences in TME, and differences in physiology between mice and humans—these factors may be the reason why immunotherapy shows effective results in preclinical in vitro and in vivo models, but the results in clinical trials are still unsatisfactory. This translational gap emphasizes the need for a paradigm shift: from viewing TAMs as a homogeneous myeloid population to targeting them as a dynamic ecosystem that adapts to both therapy and the evolving TME.