It is increasingly appreciated that tumor-associated immune cells are significantly different compared with their normal cellular counterparts. However, it remains less well understood how normal cells transition into disease-associated cell types upon initial contact to tumor cells and within different stages of tumor progression. Interestingly, Schulz et al. observed that the instruction and education of TAMs represents an early event during formation of the BrM-TME and occurs rapidly after recruitment of resident TAM-MG or peripheral-derived TAM-MDMs. Analyses of transcriptional program in TAMs isolated from small- vs. large-stage BrM revealed an almost complete absence of differences in gene expression in each population, suggesting stable transcriptomes during BrM progression (32). However, further dissection of potential transition stages based on single-cell approaches are needed to characterize the acquisition of diseases-associated signatures across a broader range of different stages of tumor progression and to identify the progenitor cells that contribute to tumor-associated myeloid cell pool for more precise comparison. In the following paragraph, we will therefore highlight recent insight on signatures of transition states based on trajectory analyses. Interesting observations on the cellular differentiation route of TAM-MDMs were made in a recent study in which the authors dissected the myeloid cell pool in different stages of murine and human glioma by single-cell RNA-Seq (scRNA-Seq) approaches in combination with lineage-tracing experiments in mouse models (34). By adoptively transferring classical or nonclassical monocytes from Cx3cr1^GFP/+ mice into Ccr2-KO mice harboring orthotopically transplanted gliomas, the authors demonstrated that only classical monocytes were able to differentiate into MDMs within the tumor. If this applies to TAM-MDM in BrM requires further investigation.

Microglia under homeostatic conditions represent a heterogeneous cell population depending on their localization within the brain parenchyma. Moreover, MG heterogeneity is modulated by developmental stages with lower heterogeneity found in adults compared with embryonic stages. Given the inherent MG heterogeneity, it is not surprising that brain pathologies induce an even higher degree of heterogeneity (12, 37), which was demonstrated with single cell RNA-seq for TAM-MG (28, 32). For example, by using t-distributed stochastic neighbor embedding (tSNE), a dimensionality reduction method, of RNA-Seq data from single cells, Schulz et al. reported that the majority of TAM-MG from the H2030-BrM model were contributing to three transcriptionally distinct cell cluster in treatment-naïve BrM. The cluster comprising most of the TAM-MG (cluster 9) was represented by high expression of cytokines (Ccl3, Ccl4, Cxcl13), cathepsin Z (Ctsz), the epidermal growth factor receptor 1 (Egfr1), as well as MG typical marker (C1qa, Hexb) (38, 39). The complement member and MG lineage marker C1qa was found to be upregulated in TAM-MG in several studies (Table 2), whereas other members of the MG-specific “sensome” core signature (Cx3cr1, P2ry12, Tmem119) (11, 38, 39) were consequently downregulated in murine and human TAM-MG (Table 2). While an increased expression of C1q members in MG belongs to their activation profile (40, 41), downregulation of homeostatic markers most likely is a consequence concomitantly occurring with downregulation of the homeostatic regulatory gene Sall1/SALL1 in murine and human TAM-MG. This further mirrors a classical activation response of MG associated to any damage of the brain, as observed under neuroinflammatory conditions (42).

Once activated and educated by tumor cells, TAM-MG upregulate several markers known to be crucial for inflammation in the injured brain, thereby probably contributing to BrM progression, e.g., via exerting immune-suppressive functions. Presumably, this is accompanied with profound metabolic changes as seen in an apparent deregulation of members of solute carrier (Slc) genes (32).

Some of the frequently observed markers in MG associated to neurological damage are apolipoprotein E (Apoe/APOE) and the triggered receptor expressed on myeloid cells 2, Trem2/TREM2, which were highly expressed/upregulated in TAM-MG of murine BrM models, whereas the expression of both members only slightly varied in human TAM-MG. The APOE-TREM2 axis has been described to drive activated MG states in neurodegenerative diseases along with downregulation of homeostatic markers (39, 43). However, especially TAM-MDMs showed elevated expression levels of Trem2/TREM2 in preclinical BrM models or patient material of BrM derived from various primary tumors (Table 2). The contributions of TREM2 and APOE to neurological diseases [e.g., Alzheimer’s disease (AD) or multiple sclerosis (MS)] have been extensively described with regard to MG (44, 45). Importantly, it was shown that targeting TREM2⁺ MG represents an interesting approach to attenuate disease progression. In addition to the broadly studied APOE-TREM2 axis, another key player of MG-mediated neurological dysfunction, and ligand for TREM2, is Galectin-3, encoded by Lgals3/LGALS3. Galectin-3 shows a multitude of functions in MG. Elevated extracellular levels within the BrM-TME might drive inflammatory responses in a Toll-like receptor (TLR) 4 binding-mediated self-sustaining manner (46, 47). In line with this, TLR4 expression was found to be upregulated in TAM-MG of H2030-BrM (32). In a recent study, Siew et al. analyzed the contribution of MG-derived Galectin-3 signaling in a mouse model of Huntington’s disease (48). Elevated cytokine levels were attributed to high Galectin-3 signaling, and strategies targeting its expression have been shown to be sufficient in decreasing levels of inflammatory cytokines, thereby ameliorating MG-mediated pathogenesis (48).

While profiling across different conditions revealed that Lgals3/LGALS3 expression levels are strongly increased in TAM-MG, varying but high levels within TAM-MDMs have been found in murine or human BrM as well (Table 2). Similarly, high expression levels were rather associated to TAM-MDMs of murine (49), or human recurrent glioma (34), as revealed in high-dimensional single-cell profiling studies. Therefore, contribution of elevated Lgals3/LGALS3 levels need to be carefully evaluated in a context-specific manner. Another group of genes (S100 family members) shows an interesting alternating pattern across TAMs in mouse and human BrM. S100 members are small, Ca²⁺-binding proteins, which regulate several cellular functions in an autocrine, or paracrine fashion, and can act as damage-associated molecular pattern (DAMP) (50). For example, S100β has been shown to serve as a noninvasive, astrocytic marker of blood-brain barrier (BBB) integrity and function, also in brain tumors (51), whereas several S100A members have been implicated in neurodegenerative diseases like AD (52). In addition, distinct S100 proteins have been associated to the regulation of inflammatory responses and TAM migration and invasion in tumors (50). Hence, upregulation or high expression especially in TAM-MG of BrM (e.g., S100A6, S100A10) (Table 2) (28, 31), might reflect a central mediator of BrM-associated inflammation. In contrast, in glioma elevated expression level of S100A6 has been implicated in a transitory TAM-MDM state in mouse and humans (34) or was part of a strong “macrophage signature” (49). Since S100A6, which was implicated in tumor progression in several other cancers (53), can either act on MG in a cell-restricted manner or can be secreted, elevated expression levels of S100A6 and other S100 members need further examination. To date, it remains unclear if the regulation of S100 members represents a cause or consequence of progressing BrM.

Direct comparison of both TAM populations on the single cell (27, 28, 31), and bulk population level (29, 30, 32) revealed further cell type-restricted molecular profiles in mouse models and human patient samples. TAM-MG showed higher upregulation of distinct proinflammatory cytokines (e.g., CXCL5, CXCL8, IL6) (29), or genes belonging to cell migration, e.g., Vim/VIM (28, 31) within a changing environment.

Cathepsins (Cts/CTS) encompass a family of proteases known to play several protumorigeneic roles in the tumor context by, e.g., remodeling the extracellular matrix (ECM) (54). In the brain, cathepsin S has been described as BrM-promoting via enhancing transmigration through the BBB (19). High expression levels of different cathepsins within TAMs of established BrM further suggest profound remodeling of the TME in outgrowing tumors. Among them, Ctsd (28) and CTSL (31) exhibited higher expression levels in TAM-MG, while Ctsb was strongly upregulated in both TAM populations of different murine BrM models (30, 32), but in human BrM was, together with CTSW, rather enriched in TAM-MDM (29). Hence, different highly expressed members of this family of ECM modulators further suggest an involvement for generating a BrM-promoting environment, but specific contributions for/with each TAM population in BrM have not been elucidated yet.

Recruitment of monocytes to the CNS has been predominantly attributed to the chemokine axis including the receptors Ccr2/CCR2 and Cx3cr1/CX3CR1 (17), which was shown in the brain tumor context using lineage-tracing approaches in glioma mouse models (26, 55). Interestingly, while Ccr2/CCR2 is dramatically downregulated upon entry into the parenchyma and during the monocyte-to-macrophage transition in glioma (55), and also BrM (27), Cx3cr1/CX3CR1 levels are upregulated in TAM-MDMs (Table 2), whereas the protein was downregulated on the majority of MDM subsets but remained abundant on only a small subset (27). These data suggest that high levels of Cx3cr1/CX3CR1 are partially required to integrate into the CNS parenchyma, since the axis is usually involved in glia-neuronal crosstalk (56). Comparing TAM data for the expression of another potent chemokine receptor known to be involved in cell migration, it became apparent that especially TAM-MDMs possess high levels or strongly upregulated the C3A receptor, C3ar1/C3AR1 in murine and human BrM (Table 2). This suggests the contribution of a conserved mechanism of recruitment via the anaphylatoxin/chemokine axis.

Given the fact that bulk analysis usually masks different expression/antigen density levels, Friebel et al. comprehensively dissected the heterogeneity of TAMs derived from different human brain malignancies, and showed that TAM instruction is not a random process, but rather driven by the underlying tumor, both in primary and secondary brain malignancies (27). By combining in total 38 markers for their myeloid panel, the authors captured a broad range of cellular states as depicted by distinct lineage-specific, but also activation markers. Upon merging all TAM CyTOF data from both, glioma and BrM samples, the authors created a detailed relationship and trajectory analyses focusing on abundance of typical monocyte/macrophage markers in silico. One of the common features they found was downregulation of CCR2 and the SIGLEC family member, SIGLEC3 (CD33), which represents a transmembrane receptor implicated in pattern recognition and regulation of phagocytosis, and in that regard has been described to be a risk factor for AD (57). Moreover, transitioning cells were found to commonly upregulate the innate immune sensor receptor CD163, together with the TAM receptor MERTK (Table 2). Monocytes transition to macrophages through a more common MDM state (termed MDM 1), and finally develop into three distinct MDM subpopulations (MDM 2, 3, 4), and this transition was driven by differential regulation of certain markers, including CD169, CD206 (mannose receptor c-type I, MRC1), CD209 (C-type lectin receptor, CLEC4L), CD38, and PD-L1. Some of these markers were further used in combination with IBA1 to specifically stain for MDMs in the TAM compartment (27). Since not only CD206 but also CD209 usually are associated to alternative macrophage activation, it is not surprising that those markers have been found to be predominantly upregulated in TAM-MDMs of murine and human BrM across different conditions (Table 2). MERTK upregulation was found to be gradual, and highest antigen density was allocated to the MDM 4 population (27), whereas bulk RNA-Seq of murine and human TAM-MDMs revealed in general elevated expression levels (Table 2). MERTK and AXL, which were also found to be significantly upregulated especially in TAM-MDMs (Table 2), represent two of the three TAM (TYRO3, AXL, MERTK) receptor tyrosine kinases, which are involved in phagocytosis and regulation of inflammatory responses (58). Interestingly, one of the ligands for AXL, Gas6/GAS6, was also highly expressed within both TAM populations in murine BrM, but rather upregulated in TAM-MDMs of human BrM samples (Table 2). Although GAS6-AXL signaling is present in the healthy CNS, and is associated to phagocytosing MG (59), deregulation can cause enhanced inflammation in the CNS (60). Moreover, this signaling axis can lead to a protumorigenic TME (61) and has been found to be coexpressed in TAMs with high C1QC levels in various primary tumors (62).

Despite several differences between the same TAM population within both mouse and human, further commonly regulated markers of TAM-MDMs included Nr4a2/NR4A2. This nuclear receptor family is known to control macrophage gene expression during inflammation (63) but is implicated in maintaining normal functions of dopaminergic neurons in the healthy brain. Interestingly, NR4A2 has been found particularly upregulated in a transitory monocyte population in glioma (34), further suggesting regulatory involvement in inducing inflammatory phenotypes during MDM development.

While certain patterns associated to these inflammatory states of TAMs seem to be conserved between species, other families of proteins are rather restricted to either a species, or a cell type. One interesting group encompasses the membrane-spanning (MS) protein family of MS4A members (Table 2). While for some of the family members (e.g., MS4A1 = CD20) their functions are well described, most of them remain poorly understood. In a recent study, Liu et al. generated new lineage-tracing mouse models targeting Ms4a3 (Ms4a3^Cre and Ms4a3^CreERT2) and validated lineage specificity of this marker, which specifically distinguishes monocytes and granulocytes from embryonically derived resident macrophage populations, including MG, under steady state but also inflammatory conditions (64). MG possess strong expression levels of certain Ms4a members (Ms4a6b, Ms4a6c, Ms4a6d, Ms4a7) during early development, while these high expression levels are not found in adult MG or in response to injury. Nevertheless, MG of that specific subpopulation of MG identified during early development cluster also showed overlapping features to BAMs (37). Several genes of the MS4A family appeared in all of the BrM-omics studies among the top regulated genes, and some of them have been described as risk factors for AD (65), including Ms4a4c (mouse)/MS4A4E (human), Ms4a6c/MS4A6E, and Ms4a7/MS4A7. Interestingly, all of them were found at high expression levels or strongly upregulated across both TAM populations of murine BrM with slightly higher levels in TAM-MDM (Table 2). Contrary, in human TAMs, MS4A members were found predominantly upregulated or higher expressed per se in TAM-MDMs. Similarly, high and stable expression of Ms4a7 was reported as part of a core signature for BAMs in steady state and under neuroinflammatory conditions (42). Taken together, this data is in line with the situation in glioma TAMs (34) and further corroborate strong similarities in transcriptional profiles between each TAM population derived from distinct diseases. These findings are strengthened by higher intersect levels of each TAM pool between glioma and BrM (29).

Aiming to describe heterogeneity within certain TAM subsets, all of the recent studies looking into the TME by single-cell approaches confirmed that especially the TAM-MDM pool consists of a more diverse range of cellular states (27, 28, 32). Whereas, Guldner et al. even described BAMs to contribute to the TAM-MDM pool within their model. With respect to identity of cell types within cell populations that were FACS purified prior to transcriptomic analyses, it is important to carefully consider procedures of sample preparation and marker selection. Macrodissection of tumor lesions is critical to reduce the risk of diluting the disease-associated cell pool with normal cell types. Moreover, the use of different marker combinations can lead to different assignment of cell types to certain subpopulations. For example, several commonly used markers to discriminate MG and MDM including CD45, SALL1, and TMEM119 (30, 66) are known to show assimilation of expression levels in both populations in brain tumors. Hence, the choice of marker combinations can lead to differences in population assignment. This determines the classification of subpopulations and consequently significantly affects the respective transcriptional programs. Nevertheless, typical non-MG clusters (TAM-MDMs/BAMs) were shown to be dominated by the expression of genes belonging to antigen processing and presentation particularly associated to MHC class II presentation, including H2-Aa, H2-Ab1, H2-Eb, and CD74 (Table 2) in several independent studies (28, 30, 32). Expression of the MHC class II member HLA-DR in human TAMs was similarly higher in TAM-MDMs (Table 2), and upregulated expression can be attributed to the transition from monocytes to MDMs (27), similarly as in the glioma situation (34). Interestingly, this raises the question to which extent each TAM population contributes to T-cell interaction, hence influencing a cancer-promoting, or immunosuppressive TME. Using multiplexed immunostaining, spatial organization of TAMs with respect to T cells was examined in murine (30) and human BrM (29). Both studies found a close proximity of both TAM populations to CD4⁺ and CD8⁺ T cells, yet Niesel et al. observed closer proximity of TAM-MDM to T cells compared with TAM-MG based on discrimination by TMEM119 (30). Moreover, it was observed that PD-L1 was almost absent in tumor-free brains, whereas BrM induced the recruitment of PD-L1⁺ myeloid cells, and levels of PD-L1 were highest among TAM-MDMs (30). On the gene expression level, several costimulatory and also inhibitory markers were found to be present among both TAM populations, whereas most of them were expressed at higher levels in TAM-MDMs (30). Representative comparison of different T-cell regulatory markers across TAMs revealed that both, activating Icosl/ICOSL, and inhibitory Pd-l1/PD-L1 markers are present in TAMs with slight higher levels in the TAM-MDM compartment (Table 2). Given the spatial organization of TAMs within the BrM-TME, one can assume that TAM-MGs in the BrM periphery are in contact with T cells at the tumor-stroma interface, while immunosuppression within the BrM core is fostered by TAM-MDMs once T cells have entered the tumor mass. Furthermore, T-cell profiles confirm exhaustion states within the TME (27, 30). Additional immunoregulatory mediators within BrMs predominantly derived from TAM-MDMs were the chemokines CCL8, CCL13, CCL17, and CCL18 (29). Together, all of them are attributed to an alternative, rather tumor-promoting phenotype (= M2) of macrophages, which is found in many tumors (35). Despite this fact, both TAM-MG and TAM-MDMs upregulate a broad range of inflammatory mediators (29, 32), and hence cannot be classified into the conservative M1–M2 scheme, but rather exist within a continuum, depending on their current local environment.

Finally, in order to understand putative dichotomous functions of TAMs, and their relevance for the inflammatory TME in BrM (Figure 3), detailed annotation and pathway analyses of differently regulated genes were performed based on results obtained from RNA-Seq experiments (28–30, 32).

To delineate functional changes upon TAM instruction, comparative analyses of significantly differently regulated genes of TAMs vs. their healthy counterparts (i.e., normal MG, or blood monocytes) in breast cancer BrM (30), and lung cancer BrM (32) were performed. Together, both studies showed that altered transcriptional profiles in TAM-MG resulted in upregulation of pathways and signaling cascades associated to inflammation, regulation of cytokine production, type I interferon (IFN) signaling, cell migration and motility, and proliferation. Interestingly, TAM-MG were found to be involved in the recruitment and interaction with neutrophils, in both mouse models of BrM (30, 32), and also in human BrM (29). Concurrently, TAM-MG downregulates genes involved in housekeeping functions, such as synapse organization or regulation of neuronal organization.

TAMs derived from the periphery are more plastic than MG as described above. Thus, their altered transcriptional profiles need to be evaluated more carefully in a context-dependent manner. Compared with their normal cellular counterparts, TAM-MDMs upregulated pathways involved in inflammatory responses, cytokine production and interaction, migration, mitosis and cell cycle, and also organization of the ECM (30, 32). In addition, human BrM TAM-MDMs showed slightly higher relevance of genes related to mitosis and cell division, compared with TAM-MGs; however, staining of human BrM samples for proliferation markers indicated proliferation in both cell types (29). However, if this results from an environment promoting local proliferation, or is caused by other stimuli, e.g., prior treatment remains to be elucidated. Analyses of patient samples stratified based on treatment history will be required to gain further insight whether different intervention strategies modulate recruitment and expansion of TAM-MDMs as previously shown in a lung-to-brain model in response to radiotherapy (32). Both human TAM populations showed an enrichment for genes related to type I IFN, and NF-kB signaling (29). Not surprisingly upon entering the BrM-TME, and transitioning from monocytes, TAM-MDMs downregulated genes related to chromatin organization, and intracellular reorganization of, e.g., the cytoskeleton (30).

In addition to lineage-restricted transcriptional programs in TAM populations, it became clear that both TAM populations also share a significant proportion of similarly regulated markers by Top marker principal component, and overlapping gene analysis in human TAMs (29), or by comparison of all differently regulated genes (DEGs) (32).

In H2030-BrM, approximately 300 DEGs were found to be commonly upregulated, and around 900 DEGs commonly downregulated between both TAM populations (32). Functional annotation of all jointly regulated markers revealed the induction of inflammatory pathways, as well as regulation of cell adhesion and cell migration. Downregulated DEGs were mostly associated to homeostatic functions in the brain, e.g., synapse organization. In contrast to the results obtained from experimental BrM models, the number of shared genes between TAM-MG and TAM-MDM was rather small in the human situation (29). The difference can potentially be explained by the fact that human RNA-seq data in this study were not stratified based on primary tumor type thereby possibly missing important similarly regulated genes. Comparison of typical inflammatory cytokines (e.g., CCL2, CCL3, CCL4, IL1B) across mice and human BrM-TAMs revealed that although most of them exhibited much higher expression levels in the BrM situation in mice, it rather were TAM-MDMs upregulating them in the human scenario (Table 2). RNA velocity analysis of single-cell RNA-Seq data of mouse TAMs further showed that Il1b and also Tgfb were genes similarly regulated at convergence points between TAM-MG and TAM-MDM clusters (28).

Osteopontin which is encoded by Spp1/SPP1 is a marker usually associated to MG of early development or has been described as one key marker of all disease-associated MG (DAM) subcluster (37). In BrM, Spp1/SPP1 was found highly expressed or upregulated in both TAM populations across species (Table 2), however with slightly higher expression levels in TAM-MG, similar as in primary brain tumors (34). Although a broad range of cellular functions has been assigned to osteopontin, under inflammatory conditions it most likely regulates inflammation itself via enhancing the recruitment of not only myeloid but also lymphoid cells (67) to the TME. Aiming to reduce or dampen inflammation within the BrM TME, osteopontin hence might represent an interesting target.

In addition to the high expression levels of the complement cascade-initiating member C1q/C1Q within the BrM-TME, two other inflammatory factors, Il1a/IL1A and Tnf/TNF, showed broad abundance and were highly expressed or upregulated in TAMs of murine or human BrMs (Table 2). Together, the presence of all three molecules strongly raises the possibility of TAM-mediated astrocyte activation towards a neurotoxic phenotype (termed A1), which partially would be sustained due to constant factor availability. This mechanism of astrocyte activation has been described in mouse models of neuroinflammatory conditions, and A1 astrocytes were found in samples of various human diseases. A characteristic marker for A1 astrocytes is the central complement component C3 (68). Interestingly, next to C1q/C1Q, other members of the complement system were apparently deregulated in BrM (32), and cross-conditional comparison revealed upregulation of, for example, C4b/C4B among all TAMs in mice and human (Table 2). While oligodendrocyte-derived C4b in mice has been associated to pathogenicity in an AD model (69), its functions within the TAM pool of BrM remain to be shown.

In summary, recent discoveries and previously unknown molecular insights into macrophages/microglia associated to BrM have dramatically shifted the paradigm of BrM-TAMs representing one homogeneous population. In comparison with data derived from recent “omics” studies involving glioma, it became clear that TAMs in brain malignancies constitute a heterogeneous mixture of resident and recruited mononuclear phagocytes, with multifaceted activation states (Figure 3). Moreover, each major subpopulation contributes to the inflammatory TME in a unique way, and disease-specific manner. The discovery of molecular markers present in both TAM populations or conserved between species opens novel avenues to develop targeted approaches in order to fight this deadly disease.