Microglia derive from hematopoietic precursor cells of the yolk sac during early embryonic development (17, 21), and Runt related transcription factor-1 (Runx-1) and colony stimulating factor-1(CSF-1) play critical roles in their development (21). Due to their powerful phagocytic function, microglia can engulf abnormal entities in the CNS, such as tumor cells, necrotic cell fragments, and pathogens (21). In addition, microglia perform immune regulatory functions, interact with neurons and glial cells, and promote angiogenesis (22). Therefore, microglia not only effectively respond to CNS damage, infection, and mutation, but also play a critical role in the development and homeostatic maintenance of brain (21, 22). Microglia rapidly shift from a ramified resting state to an amoeboid-like activated state and release reactive oxygen species (ROS), proinflammatory cytokines, and chemokines in response to exogenous and endogenous stimuli, such as infection and injury (23). This effect is mainly achieved through pattern recognition receptors (PRRs) expressed in microglia, which recognize pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) (24, 25). Recent studies have also shown that microglia play an important role in synaptic formation in the mature brain (26), serving as intermediates for information exchange through hemichannels and gap junctions with neurons and other glial cells (26). Moreover, during the development of primary tumors, activated microglia can kill tumor cells by secreting proinflammatory cytokines and other factors (9). However, with tumor progression, the phenotype and function of microglia change, causing damage to normal neural structures and promoting tumor growth (9, 27). This topic will be discussed in detail later.

Due to the lack of reliable in vivo and in vitro experimental models of microglia, the investigation of microglia in the context of tumors remains limited. Alongside primary microglia cells extracted from experimental animals or human brains, immobilized murine (BV-2) and human (HMO6) microglia cell lines have been developed (28). Nevertheless, it is crucial to note that there exist differences between the primary brain-derived microglia and the immortalized BV-2 cell line at the transcriptional level (29). Recent advancements in flow sorting technology and transcriptome sequencing, several specific marker molecules of primary microglia have been identified. For example, some transcription factors, including Rhox5, E2f6, Hoxc6, and Ppargc1b are exclusively expressed in microglia (30, 31). Furthermore, certain membrane proteins, Lrp8 and Lpcat3, which are associated with lipid metabolism, and ion transporters, like Slco4a1 and Slc30a5, are unique to microglia compared with macrophages (30, 31). Potential markers that distinguish CNS-derived microglia and BMDMs are systematically summarized in Table 1 .

Brain-infiltrating bone marrow-derived macrophages (BMDMs) originate from hematopoietic stem cells (32) and infiltrate the brain parenchyma in large numbers after the development of GBM due to the destruction of blood-brain barrier and release of multiple chemokines by the tumor tissue (33, 34). They are mainly located in perivascular and necrotic regions to address the ischemic regions of the tumor (33). Although microglia and infiltrating BMDMs have distinct origins, they perform similar immune regulatory functions and express several common markers, such as ionized calcium-binding adapter molecule 1 (IBA1), CD11b, CD68, CX3C chemokine receptor 1 (CX3CR1) (9, 35, 36). Among these markers, IBA-1 and CX3CR1 were thought to be specific to microglia, however, subsequent studies confirmed that they were also expressed by BMDMs (22, 37). In contrast, major histocompatibility complex (MHC) class II and subsequent Sall1 are specific to microglia and can be used to differentiate between microglia and macrophages in the brain parenchyma (22, 38). In addition, transmembrane protein 119 (TMEM119) is a recently discovered marker with high specificity for microglia that can distinguish microglia and macrophages in both human and mouse models (35, 39). CD49D is specifically expressed by BMDMs infiltrating malignant brain tumors in mice and humans, but not by microglia (33). C-C chemokine receptor-2 (CCR2) was previously considered as another specific marker to differentiate macrophages in TME, but subsequent studies found that CCR2 is also expressed by microglia, particularly those activated by interaction with glioma cells (40). Previously, CD45 has been used to distinguish between resident microglia (CD45^low) and infiltrating macrophages (CD45^high) (41), and in combination with CD11b, microglia and macrophages in rodent and human GBM can be defined as CD11b⁺/CD45^low and CD11b⁺/CD45^high populations, respectively (42). However, CD45 expression increases in microglia activated in the TME, which adds complicates the identification of microglia and macrophages in tumors (42). To date, no lineage-specific markers have been developed to accurately distinguish these two cell populations, which has made it challenging to assess the specific role of each cell type in tumor development.

Mills et al. were the first to define macrophages as M1 polarization type for pro-inflammatory (antitumor) and M2 polarization type for anti-inflammatory (protumor), referring to the dichotomy of CD4⁺ T cell function into T helper (Th) 1 and Th2 lineages (43). Ponomarev et al. subsequently verified the polarization of microglia in the rodent CNS (44). Numerous CNS pathological events, such as tumor occurrence, injury, microbial infection, and degenerative diseases, can cause the polarization of microglia/macrophages (33, 45–47). For instance, microglia, which account for the majority of GBM, can be polarized into the M1 phenotype under the stimulation of lipopolysaccharides (LPS) or interferon-γ (IFN-γ) (13). M1 type microglia reactively express co-stimulatory molecules such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), and chemokine (C-X-C motif) ligand 10 (CXCL10), and high-level of MHC class II, which endows them with the function of antigen presentation (13). In the context of glioma, M1 microglia play an anti-tumorigenic role through antigen presentation mediated adaptive immunity response (phagocytosis of tumorigenic cells) and secretion of proinflammatory factors (8, 48). Inflammatory factors such as TNF-α further recruit peripheral macrophages into the tumor (49). M1 microglia/macrophages have the classic activated phenotype, and their activated receptors and functions are relatively clear. The polarization of M2 microglia and macrophages is strictly regulated and presents a highly dynamic state, often followed by M1 polarization (49, 50). Type II inflammatory factors such as IL-4, IL-10, and IL-13 generally mediate M2 polarization, which can prevent further tissue damage by secreting multiple anti-inflammatory factors to downregulate the inflammatory response (51, 52). In the context of glioma, M2 microglia promote tumor development (53). M2-polarized microglia/macrophages are differentiated by the co-expression of surface markers such as CD163, CD204, CD206 as well as arginase (53). We have a systematic list of markers for M1 and M2 microglia/macrophages in Table 1 . M2 microglia/macrophages can be further divided into several subtypes (M2a, M2b, M2c) based on their expression of different transcription factors, effector functions, and secretion of cytokines and chemokines (54). M2a and M2b microglia/macrophages are responsible for Th2 activation and immunoregulation (55, 56), while the M2c subtype attenuates inflammatory response, promotes matrix deposition and tissue remodeling, significantly promoting tumor development in the neoplastic context (57–59).

Recent single-cell sequencing studies on GBM suggest that individual cells can express genes promoting inflammation (M1) and genes promoting immunosuppression (M2) (60, 61). These findings challenge the precision of the conventional M1/M2 typing of GBMs and suggest that the polarization may be a highly continuous process in the context of GBM, where GAMs are highly plastic (8). Furthermore, there are likely unclearly defined GAM subtypes having important specific functions in human and rodent GBM. For example, the nonpolarized M0 subtype may have been overlooked to some extent, even though it represents a weakened M2 subtype (62). Accurate classification of GAMs is critical for guiding the development of therapeutic drugs, but it is challenging to establish a systematic microglia/macrophage classification system. Thus, the classical subtype classification model still holds important reference value.

A large number of previous studies have demonstrated that glioma cells secrete different chemokines, which serve as chemoattractants and mediate the recruitment of GAMs. These chemokines include monocyte chemoattractant protein (MCP)-1 (alternative name: C-C motif ligand 2 (CCL2)) (66), MCP-3 (67), stroma-derived factor (SDF)-1 (alternative name: CXCL12) (68), lysyl oxidase (LOX) (69), macrophage colony-stimulating factor (M-CSF) (70), and glial cell–derived neurotrophic factor (GDNF) (71). The role of MCP-1 in GAMs has been verified by various in vivo and in vitro studies, where the MCP-1 expression level is highly correlated with the grade of glioma (66). A recent study indicated that the cooperation between β- Catenin and MCP-1 may be responsible for the rapid and highly heterogeneous growth of Isocitrate dehydrogenase wildtype GBM (72). M-CSF is another important chemokine, which not only promotes the mobility of GAMs but also mediates the M2 polarization of GAMs (73). Gliomas having a phosphatase and tensin homolog (PTEN) deletion highly express LOX, which activates the β 1 integral/proline-rich tyrosine kinase 2 pathway in GAMs, aiding their recruitment (69).

In recent years, researchers have suggested alternative factors that might impact the recruitment of GAMs and M2 polarization, such as peritumoral hypoxia microenvironment-induced factors and GSCs (74). Hypoxia may be the most critical regulatory factor for the recruitment of GAMs, as a large portion of the glioma is hypoxic and harbors large numbers of M2 polarization-type GAMs. Recent studies suggest that hypoxia may affect the expression pattern of chemokines, especially in perivascular niches. For example, Guo et al. revealed that TGF-α mediated upregulation of periostin (POSTN) in the peritumoral region of the glioma significantly promotes the recruitment of GAMs (75). Another recent study showed that arginine methyltransferases (PRMTs) enhance the recruitment of GAMs by enhancing hypoxia-inducible factor-1 mediated hypoxia, thus promoting glioma progression (76). Specific chemokines secreted by GSCs may recruit certain GAMs subtypes. For example, POSTN secreted by GSCs specifically recruits M2 GAMs through integrin αvβ3 signaling pathway (53). In addition, some cytokines with undefined functions have also been reported to participate in the recruitment of GAMs. For example, dual function cytokine IL-33, secreted at the nucleus of gliomas, was recently demonstrated to recruit and activate circulating and resident innate immune cells (77).

The tumor-promoting effects of GAMs have been demonstrated in organotypic brain tumor-slice cultures as well as several in vivo models (78–80). Co-culture of glioma cells with microglia extracted from mouse brain significantly increased glioma cell migration, while the deprivation of microglia significantly promoted tumor growth and invasion (78, 79). A range of anti-inflammatory and pro-tumoral factors secreted by GAMs have been identified as playing important roles in glioma cell invasion, including transforming growth factor beta (TGF-β), epidermal growth factor (EGF), IL-6, IL-1β, stress-inducible protein-1 (STI-1), matrix metallopeptidase-2 (MMP-2), and MMP-9 (27). Among these, MMP-2 and MMP-9 are important effector molecules that enhance glioma cell invasiveness by breaking down extracellular matrix (ECM) components such as collagen and elastin. MMP-2 expression is positively correlated with glioma invasiveness and poor patient prognosis (81). TGF-β, an inhibitory immune regulatory factor, is the most widely explored factor in the mechanism of invasive growth of glioma. TGF- β released from GAMs has been demonstrated to promote the secretion of MMP-2 and MMP-9, resulting in enhanced GBM invasiveness (82). Recent studies have revealed that MMP-2 may be an important downstream molecule of CCL5 in promoting the migratory and invasive activities of GBM (83). MMP-9 has also been reported to be an effector molecule in the tumorigenic infiltration of GBM mediated by GAMs (84). GBM released CCL2 has been reported to upregulate IL-6 expression, which is responsible for GBM invasiveness in a TLR4-dependent fashion (85). The expression level of IL-6 is highly correlated with the pathologies of GBM patients (86). EGF is considered a promoter of GAM-mediated tumor invasiveness, as it is not detectable in the supernatant or cell lysate of glioma cells cultured separately (70). EGF primarily binds to the surface of GBMs to promote tumor invasion. Amplification of the EGFR gene and its truncation mutant are present in over half of primary GBM and are indicative of highly aggressive tumors (87). It should be noted that the regulatory pathways discussed above do not strictly follow the effect of GAMs on GBM in the context of GBM. Many regulatory factors can also be released by tumor cells and influence GAMs.

Angiogenesis is a critical factor in the growth and progression of glioma. GAMs extracted from GL261 gliomas have been shown to release a plethora of proangiogenic molecules, including vascular endothelial growth factor (VEGF) and CXCL2 (88). Depletion of resident microglia significantly reduces tumor vessel count, and a high spatial correlation between GAMs and tumor neovascularization has been reported (88). For instance, specific GAMs that are absent in normal brain tissue were directly detected in the perivascular niche (merged with CD31⁺ vessels) (89). Furthermore, studies have not only verified the direct contact effect of GAMs with blood vessels around and inside the tumor using allografted mice but have also detected elevated levels of various angiogenesis-inducing factors, such as VEGF, VEGFR1, CCR2, CXCR4, CCL2/5, and CXCL2/10/14 (14, 90–93). Recent research has confirmed the strong correlation between GAMs and tumor blood vessels in GBM patients and observed the restructuring of the blood vessel architecture, indicating the potential of anti-angiogenic therapy in the treatment of GBM (88). Consistent with this, VEGF receptor (VEGFR) blocker Sunitinib and VEGF inhibitor Bevacizumab have been shown to promote survival by reducing tumor angiogenesis in the mouse GBM model (94).

The immune-escape mechanism of GBM relies on two main factors: the intrinsic characteristics of neoplastic cells and the immunosuppressive TME mediated by GAMs. Glioma cells are difficult to identify by the immune due to their downregulation of human leukocyte antigen (HLA) molecules that cover the surface specific tumor antigen molecules (95). However, the immunosuppressive TME has been recognized as a more significant contributor to promoting tumor immune evasion, facilitating its growth, invasion, and recurrence (78, 79). GAMs, especially the type M2 subtype, are considered major contributors to TME due to their secretion of type II immune factors such as TGF-β, IL-4, and IL-10 (4, 6). These cytokines upregulate transcription factors such as signal transducer and activator of transcription 3 (STAT3) in glioma cells, which subsequently triggers tumorigenic immune responses (27). STAT3 plays a vital role in the interaction between tumor cells and GAMs, and its activation significantly enhances tumor-promoting immune regulation while inhibiting tumor-killing immune response (96, 97). Several studies have demonstrated that glioma cells secrete S100 calcium binding protein B (S100B), which promotes M2 polarization of GAMs through the receptor for advanced glycation end products (RAGE)-STAT3 signaling pathway (98). Additionally, other growth factors secreted by type M2 GAMs, such as platelet derived growth factor (PDGF), EGF, and fibroblast growth factor-2 (FGF-2), have been reported to activate STAT3, promoting GBM progression (99–101). In vitro studies have shown that conditioned medium derived from glioma cells can activate the STAT3 pathway of microglia, promoting their polarization towards the M2 subtype and secretion of IL-6 and IL-10 (102).

A recent study found that blocking the breakdown of interferon gamma inducible protein 16 (IFI16) with drugs can activate the STAT3 signaling pathway (103). This pathway is crucial for GBM progression because it promotes cell proliferation, invasion, and regulation of the TME. Moreover, high levels of phosphorylated-STAT3 (pSTAT3) are associated with more severe gliomas and poorer patient outcomes (104–106). This is likely because pSTAT3 can suppress the immune system, leading to greater resistance to standard cancer treatments and an increased risk of tumor recurrence (105, 106).

GCSs are a group of cells that are resistant to chemotherapy and contribute to tumor growth and recurrence (107–109). They share stem cell-like characteristics, which enable them to regenerate and differentiate into different cell types (108, 109). This allows them to continuously generate new tumor cells at the site of the tumor. GSCs are similar to NSCs in terms of their phenotypic characteristics. For example, they can form neurospheres in vitro and express common markers such as nestin, Sox2, and Musashi-1 (110, 111).

GSCs have low mitotic activity, which makes them resistant to therapies directed at actively dividing cells, such as temozolomide (TMZ)-based chemotherapy, and enables them to cause tumor recurrence (112, 113). They also have low expression of molecules that present tumor antigens to CD8⁺ T cells, and immune checkpoint pathways are often activated in glioma cells (114). These mechanisms help tumor cells evade immune surveillance. GSCs interact with various components in TME to maintain their drug resistance and tumorigenicity. They are located in specific niches where they are protected from therapy exposure and niche-specific factors (95, 115, 116). GAMs mainly accumulate in perivascular and perinecrotic hypoxic niches, where they initiate interactions with SGCs to affect glioma prognosis (74, 95). GSCs secrete chemokines (such as VEGF, CCL2, CCL5, and CCL7), to recruit GAMs to tumor mass and induce M2 type polarization of GAMs, aiding their transformation and proliferation by building a tumorigenic TME. Apart from GSCs-secreted chemokines, GSCs also exhibit higher levels of neurotensin than those in non-GSC glioma cells (27, 117). In vitro studies have shown that GSCs recruit GAMs more effectively than glioma cell lines. Moreover, a positive correlation between GAMs and GSCs has also been reported (118). These findings emphasize the importance of GSCs for GAMs recruitment.

Conversely, GAMs have been shown to play a role in promoting GSCs (119–122). Research by Wang et al. demonstrated that IL-6, which is secreted by microglia, can function as a growth factor for GSCs and enhance their biological function (122). It is important to note that only GAMs, not the naïve resident microglia, can promote GSC amplification (123). Interestingly, microglia from healthy individuals can even suppress glioma growth by expressing IL-8 and MCP-1 (123). In general, GAMs and GSCs interact extensively in glioma, jointly regulating tumor progression, recurrence, and drug resistance.