Glioma-Associated Microglia Characterization in the Glioblastoma Microenvironment through a ‘Seed-and Soil’ Approach: A Systematic Review

Background and aim: Ever since the discovery of tumor-associated immune cells, there has been growing interest in the understanding of the mechanisms underlying the crosstalk between these cells and tumor cells. A “seed and soil” approach has been recently introduced to describe the glioblastoma (GBM) landscape: tumor microenvironments act as fertile “soil” and interact with the “seed” (glial and stem cells compartment). In the following article, we provide a systematic review of the current evidence pertaining to the characterization of glioma-associated macrophages and microglia (GAMs) and microglia and macrophage cells in the glioma tumor microenvironment (TME). Methods: An online literature search was launched on PubMed Medline and Scopus using the following research string: “((Glioma associated macrophages OR GAM OR Microglia) AND (glioblastoma tumor microenvironment OR TME))”. The last search for articles pertinent to the topic was conducted in February 2022. Results: The search of the literature yielded a total of 349 results. A total of 235 studies were found to be relevant to our research question and were assessed for eligibility. Upon a full-text review, 58 articles were included in the review. The reviewed papers were further divided into three categories based on their focus: (1) Microglia maintenance of immunological homeostasis and protection against autoimmunity; (2) Microglia crosstalk with dedifferentiated and stem-like glioblastoma cells; (3) Microglia migratory behavior and its activation pattern. Conclusions: Aggressive growth, inevitable recurrence, and scarce response to immunotherapies are driving the necessity to focus on the GBM TME from a different perspective to possibly disentangle its role as a fertile ‘soil’ for tumor progression and identify within it feasible therapeutic targets. Against this background, our systematic review confirmed microglia to play a paramount role in promoting GBM progression and relapse after treatments. The correct and extensive understanding of microglia–glioma crosstalk could help in understanding the physiopathology of this complex disease, possibly opening scenarios for improvement of treatments.


Introduction
Ever since the discovery of tumor-associated immune cells, there has been growing interest in the understanding of the mechanisms underlying the crosstalk between these cells and tumor cells, which can range from conferring a growth advantage to the latter to enabling the cancer cells to escape autoimmunity. To achieve this, a significant effort has been made to define the types and characteristics of the cells constituting the tumor 2 of 35 microenvironment (TME) and to characterize how these cells assist the tumor cells by producing cytokines, chemokines, growth factors, and triggering the release of inhibitory immune checkpoint proteins from T cells.
Gliomas, with emphasis on glioblastomas (GBMs), are no exception and they present a wide range of glioma-associated macrophage and microglia (GAMs). Extensive literature in this field has been published to gain a better understanding of the composition of the TME and of the role that resident microglia and immune cells from peripheral blood play in ensuring an immunosuppressive environment in which the tumor can thrive undisturbed.
The nature of the TME shapes both therapeutic responses and resistance, justifying the recent impetus to target its components. Despite the growing body of evidence on the topic, many aspects are yet to be explored in the field. Open questions are related to microglia distribution in the peritumoral area, its spatial and molecular interaction with GBM infiltrative margins and stem cell compartment, and the real extension of the so-called TME with respect to the tumoral core. A correct and extensive understanding of microglia-glioma crosstalk could help in understanding the physiopathology of this complex disease, possibly opening scenarios for improvement of treatments. An immunosuppressive pressure acts on the microglia and macrophage towards a polarization to an M2 protumor-immunosuppressive cellular phenotype. However, it is not ultimately clear to what extent an M2 macrophage distribution is capable of maintaining an immunologically downregulated environment.
A "seed and soil" approach has been recently introduced to describe the GBM landscape: the tumor microenvironment acts as fertile "soil" interacting with the "seed" (glial and stem cells compartment). Microenvironmental contribution seems critical to gaining a better understanding of the unique challenges GBM poses and could be pivotal in: (a) developing new treatments targeting GAMs, thus rendering the tumor once again targetable by host immunity, and slowing its progression and aggressiveness by inhibiting the microglia and tumor cell crosstalk (b) defining the hystotypes of the tumor and degree of response to treatment based on the GAMs composition and infiltration pattern.
In the following article, we provide a systematic review of the current evidence pertaining to the characterization of GAMs and microglia, and macrophage cells in the glioma TME.

Methods
This study was conducted in accordance with the PRISMA-P (Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols) guidelines. An online literature search was launched on PubMed Medline and Scopus using the following research string: "((Glioma associated macrophages OR GAM OR Microglia) AND (glioblastoma tumor microenvironment OR TME))". The last search for articles pertinent to the topic was conducted in February 2022. Two authors, C.M.D. and G.M., independently conducted the abstract screening for eligibility. Any discordance was solved by consensus with a third, senior author, G.D.P. No restrictions on the date of publication were made. Exclusion criteria were as follows: no comparative study design, studies published in languages other than English, and metaanalysis. A systematic abstract screening of the references (forward search) was performed to identify additional records.

Results
The search of the literature yielded a total of 349 results. Duplicate records were then removed (n = 10), via title and abstract screening; 235 studies were found to be relevant to our research question and were assessed for eligibility (Figure 1). Upon a full-text review, 58 articles were included in the review. The search of the literature yielded a total of 349 results. Duplicate records w removed (n = 10), via title and abstract screening; 235 studies were found to be rel our research question and were assessed for eligibility (Figure 1). Upon a full-text 58 articles were included in the review. The reviewed papers were further divided into four categories based on the (1) Microglia maintenance of immunological homeostasis and protection autoimmunity (Table 1); (2) Microglia crosstalk with dedifferentiated and stem-like glioblastoma cells (T (3) Microglia migratory behavior and its activation pattern (Table 3). The reviewed papers were further divided into four categories based on their focus: (1) Microglia maintenance of immunological homeostasis and protection against autoimmunity (Table 1); (2) Microglia crosstalk with dedifferentiated and stem-like glioblastoma cells (Table 2); (3) Microglia migratory behavior and its activation pattern (Table 3).  Although relb coordinates anti-inflammatory feedback in astrocytes, this mechanism does not function in GBM cells due to both the limited activity of SIRT1 and the presence of YY1 in the nuclei.
These data strongly support the idea that p52-independent relb signaling is critical in GBM development or progression.   An immune gene signature was constructed and showed a favorable efficiency in predicting the prognosis of patients with glioma. A high-risk score of this signature predicted enrichment of macrophages and less response to ICB therapy in glioma.
The higher complexity of immune cells was shown in high-risk patients, and the risk score was significantly positively correlated with the abundance of macrophages. Genes in our signature are highly expressed in macrophages by an open single-cell sequencing dataset of GBM.
Patients with a high-risk score showed enrichment of macrophage in a tumor-supportive state, but not an antitumor state. The comprehensive characterization of immune cells from a total of nine GBM tissues revealed a unique immune landscape in GBM at the single-cell level and identified SPI1 as a potential immunotherapeutic target against TAMs in GBM.
The study's results confirm that TAMs may occupy over one-third of the GBM tumor, with the ratio of the bone marrow-derived macrophages and the brain-resident microglia being 2:1, but infiltrating T lymphocytes comprising less than 2% of the tumor mass. SPI1 was a vital regulon in all states of TAMs Shen et al. [17] 15 October 2021 Transcriptomics 3810 Monocytes and T cells Datasets from NCBI, ICGC, CGGA This reveals two molecular subtypes (i.e., C1/2) of brain tumors featured by distinct immune infiltration signatures and prognosis.
The C1/2 subtypes can distinguish glioma patients with different prognoses stratified by histology, tumor grade, and genomic alteration. In addition, the C1/2 subtypes can also reflect differences in microvascular infiltration, distant metastasis, and the radio-chemotherapy response of patients. The data argue against the idea that CMV merely plays a bystander role in glioblastoma pathology but is contributing to oncogenesis.
A tropism for CMV antigen expression was found, specifically pp65, in the GCSCs and MS microglia. Finding of production of CMV IL-10 by the GCSCs and its subsequent effect on the MS microglia precursor, the monocyte.
CMV IL-10 activates IE1 in monocytes and there is conversion to the immunosuppressive phenotype M2.

T cells Not specified
This study showed the beneficial role of immune cell infiltration into the tumor in GBM patients, despite multiple mechanisms of tumor immune escape.
Significant positive correlation of increased cd3+ and cd8+ cellular infiltration into the tumor with improved patient survival. This is an immunosuppressive, anti-inflammatory milieu that led to alterations in gene expression of glutamate receptors and transporters as well as GS (glutamine synthase) that were found to differ between TAMs and mdms. The results show that WNT-5A is the only representative of the WNT family upregulated in human glioma compared to nonmalignant control brain tissue.
There is an approximately 4-fold increase in WNT-5A expression between nonmalignant control brain tissue and GBM patients' brain tissue. The highest levels of WNT-5A were found in mesenchymal GBM. Demonstration of CECR1-mediated crosstalk mechanism between macrophages and glioma cells.
The influence of CECR1 on immune cells, and on macrophage polarization especially, seems to predominantly impact the systemic and cerebral vasculature and could affect the tumor vasculature in GBM.
Expression analysis of typical M1 and m2a, b, and c subtype markers validated the presence of these TAM subtypes in human gliomas.

Monocytes and T cells Tumor
This study defines a strategy to determine transcriptional subtypes and associates expression subtypes to the tumor-associated immuno-environment.
The transcriptional glioma subtypes, defined through clustering based on tumor-intrinsic genes, strongly over lapped with the PN, CL, and MES subtypes, but identified the NE subtype as normal NE lineage contamination. Examination of the expression pattern of SERPINA3 at the various stages of glioma progression, including normal brain, pilocytic astrocytomas, diffuse astrocytomas, oligodendrocytes astrocytoma, anaplastic astrocytomas, and glioblastomas multiforme.
Caponegro et al. [30] 2 November 2018 Genomics 667 from TCGA Macrophages RNA-seq data NRP1 expression is correlated with poor prognosis and glioma grade, and associates with the mesenchymal GB subtype.
In human GB, NRP1 expression is highly correlated with markers of monocytes and macrophages, as well as genes that contribute to the protumorigenic phenotype of these cells.
Functional gene analysis suggests that NRP1 is associated with markers of monocytic infiltration and protumorigenic GAMs in human GB. The results demonstrate that both LGG and GB patients with high NRP1 expression have enriched monocytic, macrophage, and M2 macrophage populations. AIF1 and ITGAM (Iba1 and CD11b, respectively) are pan markers of monocytes, macrophages, and microglia, and are highly upregulated across human GB subtypes. Significantly correlated with genes that characterize the M2 pro-tumorigenic GAM signature, such as Adm and Mrc1. Also, significantly correlated with TMEM119 and TMEM173. Negative correlation between the expression of PDGFRA and PAXL in GSCS and GBM tumors. High AXL and PROS1 expression is associated with a poor prognosis in patients with GBM. The study demonstrates a role for PROS1-mediated AXL signaling in cancer, which is negatively regulated by pdgfra.
Couto et al. [32] 6 March 2019 Cellular culture Not specified GAMs Cell lines from biobank The data unravel a new oncogenic role for IL-6 in GBM through direct effects in microvascular ECs and regulation of endothelial permeability with a negative impact on brain barrier functions.
Data demonstrated significant alterations in the barrier properties of human ECs (endothelial cells) when exposed to cm-cc (conditioned medium of coculture), namely an increase in the permeability to smaller (4 kda) and larger (70 kda) molecular size dextrans and a decrease in the teer (transendothelial electric resistance) values.
Gjorgjevski et al. [33] 20 June 2019 Qpcr 20 Macrophages Tumor This study establishes M1and M2-like markers CXCL10 and CCL13 for informative and reliable detection of GBM-associated microglia and macrophage polarization in conjunction with a defined protease profile as molecular determinants for GBM progression.
It was found that MMP9 and MMP14 are negatively correlated with GBM patient survival and associated with the markers to define a more M2-like microglia and macrophage phenotype. CHME3 cells were shown to be responsive to classical proinflammatory signals through the Toll-like 4 receptor (TLR4) stimulation by LPS and IFN-γ, leading to activation of STAT1 and the nuclear factor-kb (NF-kb). When challenged with cytotoxic agents, GBM cells showed an increase in proliferation when in contact with MG. Even a low amount of MG (10%-20%) is able to confer resistance of GBM to cytotoxics.
Chiavari et al. [35] 3 November 2020 Cellular culture and histopathology 18 GAMs Tumor and normalappearing brain The data showed the expression of PDIA3 not only in tumor cells but also in GAMs, supporting its potential role in cellular and molecular processes related to GB.
This study confirmed that GAMs, as the dominant infiltrating immune cell population, exhibit substantial inter-and intratumoral heterogeneity in the GBM immune microenvironment, and increased proportions of exhausted T cell subpopulations and Tregs substantially contribute to local immune suppressive characteristics.
Chen et al. [26] 18 November 2020 Cellular culture 2 (GBM12 and GBM39) Not specified Not specified In the study, the combination of increased proliferation but decreased invasion aligns with the go-or-grow hypothesis but more importantly demonstrates that crosstalk between MG and GBM cells in the tumor microenvironment may have powerful effects on GBM activities tied directly to tumor progression and patient survival.
GO analyses revealed GBM-MG co-culture upregulated genes in a patient-derived GBM specimen associated with cell cycle, RNA and DNA division, and metabolic activity. However, genes involved in cell adhesion and migration showed significant downregulation as a result of GBM-MG co-culture. Il-1β induces the expression of ICAM-1 and VCAM-1 in GBM by binding to the IL-1 receptor, which may play key roles in interacting with GBM and tumor-associated immune cells. were identified as the primary activators driving PYK2 and FAK activation in glioma.
Microglia treated with glioma-conditioned medium (GCM) from primary cell lines increased the expression of genes encoding PDGFβ, SDF-1α, IL-6, IL-8, and EGF.  The enhanced directional migration of macrophages toward hypoxic areas has been attributed to the hypoxia-inducible expression of POSTN in glioma cells which may partially explain the mechanism by which macrophages become trapped in hypoxic regions after they were initially attracted to them. Define the distribution and polarization of TAMs in GBM Immunohistochemical evaluation of macrophage phenotype showed an interesting scenario; the number of cd163+ macrophages was significantly higher (p < 0.001) than those of cd68+ macrophages in the perinecrotic area, parenchyma, and, in particular, in perivascular areas.
In confocal analysis, the cd163+ population consisted of both cd163+/cd68+ macrophages and single stained cd163+ cells. Interestingly, this finding was more evident in perivascular areas.

L. Lisi et al. [48] 2 March 2017
Hisopathology and immunohistochemistry 41 Macrophages Tumor core and healthy brain Results suggest that cd163 expression is higher within the tumor than in the surrounding periphery in both male and female patients.
In the present work, we show that cd163 expression is higher within GBM specimens than in the surrounding periphery in both male and female patients. We report that both inos (m1 marker) and arg-1 (m2 markers) are present both within the tumor and in peripheral parenchyma, albeit unevenly distributed.
Annovazzi et al. [49] 12 September 2017 Histopathological 108 Macrophages Tumor and 10 healthy controls Characterization of different patterns of microglia composition based on the degree of infiltration.

The transition from HIA (High infiltration area) to ST (solid tumor)
was marked by an almost disappearance of RM (reactive microglia) forms, which could be reduced, destroyed, or switched to serve a different function. During tumor growth, opcs (oligodendrocyte precursor cells) and macrophages and microglia migrate and proliferate rapidly in the border region, where they secrete growth factors and cytokines, causing GBM cells to acquire stem cell profiles and chemo-radio resistance.
The rapid reaction potential of opcs and macrophages and microglia provides advantages for GBM cells in the formation of the GSC niche in the border because GBM cells possess higher proliferation and migration potential, thereby necessitating rapid adaptation of supportive cells. Newly diagnosed cases with a high CCL5 level were associated with increased tumor volume. GAMs of GBM were involved in the production of CCL. It was observed that high-grade glioma contained a higher level of GAM infiltration. Glioma cells elicited a more robust homing response toward GM-CSF-activated GAM-CM.
Landry et al. [52] 11 November 2020 Genomics 14 tumors Macrophages Tumor core and tumor periphery Analyze geographical differences in macrophage recruitment and activation through sequential activation states. Find distinct activation and maturation processes between tumor core and periphery.
It was found that TAMs in the tumor core mostly originate from the bone marrow-derived pool whereas those in the tumor periphery are largely derived from microglial cells, supporting prior research. Cells in the tumor core evolve from a "pre-activation" state toward a proinflammatory state. Tumor periphery, by contrast, it was found that cells transition from a preactivation state towards pro-oncogenic activation (M2 or "alternative" activation state).

Discussion
GBM is a complex solid tumor with a highly inflammatory tumor environment. Approximately 30-50% of the brain tumor mass is constituted by microglia, monocytes, and macrophages, defined as GAMs [55]. These cellular populations are the major cerebral immune components [17], responsible for maintaining brain homeostasis, producing cytokines, chemokines, and growth factors (which altogether constitute the TME), and regulating tumor progression [56] (Figure 2).

Discussion
GBM is a complex solid tumor with a highly inflammatory tumor environment. Approximately 30-50% of the brain tumor mass is constituted by microglia, monocytes, and macrophages, defined as GAMs [55]. These cellular populations are the major cerebral immune components [17], responsible for maintaining brain homeostasis, producing cytokines, chemokines, and growth factors (which altogether constitute the TME), and regulating tumor progression [56] (Figure 2).
All the relevant findings regarding microglia maintenance of immunological homeostasis and protections against autoimmunity are summarized in Table 1.
All the relevant findings regarding microglia maintenance of immunological homeostasis and protections against autoimmunity are summarized in Table 1.
GAMs constitute a crucial component of the TME, together with extracellular matrix (ECM) and other nonmalignant cells, such as endothelial cells, fibroblasts, and T lymphocytes which play a pivotal role in the development, growth, and malignant characteristics of gliomas [60]. Within this complex cellular multiplicity, different GAM components can be identified [2]: a first group originating from the embryonic yolk sac and considered central nervous system (CNS) immune tissue residents, and a second group consisting of the CNS-infiltrating leucocytes (e.g., monocytes, T, B, and natural killer cells), normally circulating within the blood vessels except in the case of a blood-brain barrier disruption, as described in inflammatory and neoplastic CNS pathologies [5]. Although monocytes, macrophages, and microglia have several common markers, such as an ionized calciumbinding adapter molecule 1 (Iba1) and CX3CR1, some studies regarding transcriptome analyses comparing microglia and other myeloid immune cells, have identified a multitude of genes dependent on the TGF-β signaling, codifying for Sall1, TGF-βr1, P2ry12, Fcrls, and Gpr34 that can be considered specific markers for microglia [61]. GAMs can exhibit a different spectrum of phenotypes. As generally accepted, the activated microglia and macrophages can present an antitumoral immune phenotype, defined as M1, through the secretion of proinflammatory cytokines, such as tumor necrosis factor-alpha, IL-1beta, and inducible nitric oxide synthase [1]. Alternatively, GAMs have also been reported to play a crucial role in some GBM features, such as growth, invasion, proliferation, and immunosuppression [62,63]. Particularly, GAMs are forced to switch to the M2 phenotypes in the GBM microenvironment, secreting factors such as IL-10, IL-4, IL-6, macrophage colony-stimulating factor, TGF-β, macrophage inhibitory factor, and prostaglandin E2, which, due to their anti-inflammatory action, facilitate tumoral immune-escape and increase tumor invasiveness, angiogenesis and growth, contributing to the creation of an immunosuppressive tumoral microenvironment [38]. However, in recent years different studies reported that M1 and M2 phenotypes would represent only extremes of broader and more articulated cell heterogeneity [64,65]. To reinforce this concept, Landry et al. report significant differences in the GAM populations present in the GBM core, compared to those of the GBM periphery: on one side, core GAMs manifest mainly a proinflammatory phenotype correlated with Programmed cell Death-1 (PD-1) signaling; on the other, peripheral GAMs exhibit an anti-inflammatory phenotype and a strong association with NFkB signaling [52].
Among the mechanisms that seem to support the proinflammatory activity of the GBM core, deacetylase Sirtuin 1 (SIRT1), whose gene is deleted in 80% of GBM tumors, seems to play a role. As a result, GBM cells continuously produce cytokines and factors attracting and activating glioma-associated microglia and macrophages, promoting a proinflammatory loop [7]. Another potential role in modulation and interaction between GBM cells and GAMs might be played by the ERp57/PDIA3 (protein disulfide-isomerase A3), an endoplasmic reticulum protein present both in GAMs and GBM cells, whose expression and activity were found to be directly proportional to the polarization capacity towards the protumor M2 phenotype of microglia [35]. This complex tumoral microenvironment, which mainly expresses immunosuppressive characteristics, especially in the tumoral periphery, determines a limited migration of T lymphocytes which represent less than 2% of the neoplastic cell mass [16]. This aspect has a pivotal role in therapeutic strategies, being immunotherapies targeting T cells, such as monoclonal antibodies against programmed cell death 1 (PD-1) or cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), which are not suitable for treating GBM [66].
A further significant role in tumor metastasis and invasiveness is played by cell adhesion molecules (CAMs), which are cell surface-proteins that mediate cell-cell adhesion, particularly between immune cells and target tissues. Intercellular CAM (ICAM-1, also known as CD54) and vascular CAM (VCAM-1, also known as CD106), are upregulated by several proinflammatory cytokines, improving the immune-mediated response. Furthermore, ICAM-1 and VCAM-1 play important roles in the adhesion of cancer cells to the endothelium, especially in the context of the inflammatory microenvironment, supported by the high concentration of IL-1b, correlated with higher grade gliomas [57,67]. Through binding to IL-1R present on the surface of GBM cells, IL-1 can promote the cascade of the MAPK and p65 signaling pathways, resulting in the production of proteins and soluble forms of VCAM-1 and ICAM-1, which enhance monocytes adhesion and secretion of CCL2 and IL-6 through activated macrophages. This would seem to modulate the immunosuppressive activity, increasing the survival of myeloid monocytes recruited to the TME and polarizing their differentiation toward M2-type macrophages. Furthermore, the immunosuppressive microenvironment is not exclusive to GBMs: it has also been demonstrated in IDH mutant anaplastic astrocytomas and in IDH mutant/1p-19q codeleted anaplastic oligodendrogliomas [13].

Microglia Crosstalk with Dedifferentiated and Stem-like Glioblastoma Cells
Glioma stem cells (GCS) are a chemo-resistant population that can drive tumor growth and relapse. The traditional theory of cancer stem cells defines them as a minor subpopulation of self-renewing malignant cells that maintain a low but steady level of unlimited proliferation [57,68]. The latter maintains the tumor, and these cells' low mitotic activity protects them from treatment approaches that are directed against actively dividing cells [29,69]. Therefore, these cells can survive treatment and give rise to recurrences [18][19][20][21][22][23][24]. In addition, stem cells interact with the TME, and their interplay is mandatory to develop biological resistance and sustain the tumorigenic process [31,39]. General features of GSCs are treatment resistance and association with tumor recurrence. They reside within specific anatomic niches, which can be seen as specialized microenvironments ensuring their stemness, proliferation, and apoptosis resistance, analogous to tissue stem cell niches. These niches shield GSC functionally by providing prosurvival cues and anatomically by blocking them from therapy exposure [18,30,33,41].
All the relevant findings regarding microglia crosstalk with dedifferentiated and stem-like glioblastoma cells are summarized in Table 2.
Both clinical outcomes and mathematical modeling show that GSCs are a key mechanism in determining the resistance of the whole tumor to therapy. Isolated cell lines of glioblastoma do not show such a marked resistance as observed in GBM patients. In addition to this, in the brain-mimetic biomaterial platform for the 3D culturing of patientderived GBM cells, the modulation of hyaluronic acid content and mechanical properties of biomaterials were required to recreate the known resistance to epidermal growth factors receptor (EGFR).
As previously discussed, TME niches play a multifaceted role in regulating GSCs and this motivates further investigation. Additionally, it has been demonstrated that GSCs immune evasion is critical to sustaining the tumorigenic process. GSCs express low levels of molecules involved in the processing and presenting tumor antigens to TCRs, a crucial stimulatory signal to the T-cell response. Consequently, they escape from recognition by antitumor immunity and possibly actively suppress T-cell activation. GSCs express various molecules that deliver either stimulatory or inhibitory signals during direct physical contact with tumor-infiltrating lymphocytes (TILs). The balance of these opposing signals regulates the amplitude and quality of TIL response and the aberrant activation of the inhibitory signals, also known as immune checkpoints, is a mechanism utilized by cancer cells to evade immune attacks [43,68].
Given the therapeutic potential and the promising therapies targeting the PD-1/PDL1 axis in GBM, the association between GSCs and the PDL1 axis in GBM deserves further analysis and investigation.
Hsu et al. recently demonstrated that epithelial-mesenchymal transition (EMT) enriches PD-L1 in GSCs by the EMT/β-catenin/STT3/PD-L1 signaling axis, in which EMT transcriptionally induces Nglycosyltransferase STT3 through β-catenin, and subsequent STT3-dependent PD-L1 Nglycosylation stabilizes and upregulates PD-L1. The axis is also utilized by the general cancer cell population, but it has a much more profound effect on GASCs as EMT induces more STT3 in CSCs than in nonGASCs. They further identified a noncanonical mesenchymal-epithelial transition (MET) activity of etoposide, which suppresses the EMT/β-catenin/STT3/PD-L1, leading to PD-L1 downregulation and sensitization of cancer cells to anti-Tim-3 therapy. On the one hand, this gives hope, on the other hand, it must be said that the expression of PD-L1 in GASC has never been described and further studies would be necessary to understand the possible potential for research and therapeutics [74].
This confirms GSCs and their interplay with TMEs to be mandatory for biological resistance.
Against this background, our group previously isolated a subpopulation of stem cells, called glioma-associated stem cells (GASCs). Indeed, GASCs are devoid of tumor-initiating properties, but show stem cell properties and the ability to support, in vitro, the biological aggressiveness of tumor cells [75].
In the infiltrating front of the tumor, transcriptomic suggested the presence of a GASC population possibly responsible for tumor recurrence. As already mentioned, even when GBM resection is performed beyond the tumor edge, there is no assurance that all tumor cells can be located and resected: infiltrating tumor cells are enriched with GASC, which in turn interacts with the TME, thus promoting tumor growth; in turn, the TME, by favoring hypoxic conditions, contributes to GASC generation.

Microglia Migratory Behavior and Its Activation Pattern
Despite a growing body of evidence, many aspects are yet to be explored regarding the migratory behavior of microglia with respect to tumoral margins. Open questions are related to microglia distribution in the peritumoral area, its spatial and molecular interaction with glioma infiltrative margins, and stem cell compartment [76]. Seminal studies showed how microglia move in a rather random way, whereas glioma cells exhibit a "committed" migratory behavior with significantly increased directionality compared to microglia [77]. However, it is unclear if glioma cells and microglia are responding to different migratory cues or are responding to the same cues but in different ways. GAMs polarization is influenced by both macrophage localization and tumor microenvironment signaling, resulting in a more complex scenario than the simple M1 and M2 activation status. Macrophage polarization in GBM has not yet been fully elucidated, and most results have been obtained in experimental nonhuman settings [46][47][48][49][50]78].
All the relevant findings regarding microglia migratory behavior and its activation pattern are summarized in Table 3.
Glioma cell-GAM crosstalking is fundamental to understanding microglia migratory behavior and its activation pattern in the center of the tumor and the surrounding periphery. Several studies have confirmed the central role of this crosstalk, highlighting a strong correlation between the local density of glioma cells and the rapidity of GAM migration and polarization [77,78]. Glioma cells stimulate the motility of microglial cells at the peritumoral infiltrative margins and after activation in a short period, microglia may enable more contact with cells via this random migration, resembling a surveillance function. However, microglia accumulation and polarization should not be regarded as a mere nonspecific reaction to tissue injury with consequent cytokine gradients, as it ultimately reflects their active participation in supporting and promoting the invasive phenotype of astrocytoma cells [44,[51][52][53]79].
Recent evidence highlights that the migration of tumoral and microglia compartments is 'driven' by, (a) a condition of peritumoral hypoxia with associated hypoxia-induced factors and, (b) inflammatory mediators produced by immune cells in the TME.
Hypoxia probably is the strongest determinant in shaping glioma and GAM migration. Indeed, it is well known that gliomas contain large hypoxic areas, and that a correlation between the density of M2-polarized GAMs and hypoxic areas exists. This suggests hypoxia plays a supportive role during GAM recruitment and M2 induction. Recent literature demonstrated perivascular niches in hypoxic areas and that hypoxia can affect chemotactic factors expression in such niches [46,49,80,81]. This mechanism underlies hypoxia-induced GAM recruitment and polarization, even if a clear description of M1-and M2-polarized cells distribution in a "topographic fashion" is still lacking in literature [47].
To date, in normal brains, the macrophage population is mainly located in perivascular (Virchow-Robin) spaces of cerebral microvessels. These are continuously repopulated by blood-derived monocytes and macrophages, and more rarely by resident brain microglia [47]. A considerable amount of M2 macrophages can be identified in perinecrotic and perivascular areas, which are indicators of an advanced stage of the tumor. The accumulation of GAMs in avascular and necrotic areas strongly accounts for their exposure to hypoxia, recognized as a key stimulus for alternative macrophage activation. Altogether, these data further support the notion that the macrophage phenotype could be a result of glioma progression owing to an interactive network involving glioma stem cells, proinflammatory-activated glioma cells, GAMs, and other components of the perivascular niche [76]. Recent studies confirmed a high prevalence of M2-polarized macrophages mainly disposed around the clusters of proliferating vessels. However, it must be acknowledged that whereas 'quantitative' information on GAM has been quite widely investigated, little is known about GAM spatial distribution within and around tumor mass and if a possible 'gradient of activation status' at the tumor periphery exists.
Furthermore, Guo et al. showed how hypoxia-increased periostin (POSTN) expression in glioma cells actively promotes the recruitment of macrophages and that hypoxiainducible POSTN expression was increased by TGF-α via the RTK/PI3K pathway [46].
Other studies have shed light on the mechanisms that underlie GAM recruitment and M2 polarization under hypoxic conditions in gliomas. Specific chemokine or metalloproteinase upregulation has shown a distinct 'homing effect' with regards to GAMs. As shown in the seminal study by Yu-Ju et al., chemokine C-C ligand 5 (CCL5) modulates the migratory and invasive activities of glioma cells in association with metalloproteinase 2 (MMP2) expression. In response to CCL5, glioma cells undergo a synchronized increase in intracellular calcium levels and glioma cells tend to migrate toward GAM-conditioned media activated by a granulocyte-macrophage colony-stimulating factor (GM-CSF) in which CCL5 is abundant. Moreover, an association between CCL5 and GAM activation has been demonstrated [51]. Authors suggest that modulation of glioma calcium levels may restrict the effect of CCL5 on glioma invasion and could be a potential therapeutic target for alleviating glioma growth. Indeed, GBM expresses a plethora of macrophage chemoattractants, such as IL-10, macrophage migration inhibitory factor (MIF), cytokines of the CSF family (macrophage colony-stimulating factor1, M-CSF, and granulocyte-macrophage colony-stimulating factor, GM-CSF), monocyte chemotactic protein 1 (MCP-1), alternative macrophage activation-associated CC chemokine-1 (AMAC-1), thymus and activation regulated chemokine (TARC) belonging to the CC chemokine family, chemokine (C-X-C motif) ligand 4 (CXCL4), chemokine (C-X3-C motif) ligand 1 (CX3CL1/fractalkine), and stromal cell-derived factor 1 (SDF-1).
Regarding inflammatory mediators, it has been explained in the previous sections how inflammatory cytokines are determinants in GAM polarization and activation [2]. CC motif chemokine ligand 2 (CCL-2), also known as monocyte chemoattractant protein 1 (MCP-1), is a major chemokine that acts as a glioma cell-derived monocyte chemotactic factor. Indeed, overexpression of CCL-2 results in increased migration. Conversely, stromal cells such as macrophages secrete more CCL-2 into the TME than cancer cells. In a positive feedback loop, GBM cells secrete CCL-2 and attract macrophages and microglia; these cells also secrete CCL-2, and the number of activated cells is increased. Thus, GBM cells, macrophages, and microglia exhibit a robust reciprocal network in proliferation and migration. Other authors showed how chemokine receptor pairs CXCL12/CXCR4/CXCR7, CXCL16/CXCR6, and CX3CL1/CX3CR1 are involved in tumor progression and that GBM-associated macrophages and microglia are also characterized by expression of these chemokine receptor pairs indicating a pivotal role of this expression profile in GAM biology in gliomas [44,82].

Targeting the Tumor Microenvironment: The Disillusionment with Current Immunotherapeutic Treatment
Targeting therapeutics to the TME offers promise to improve patient survival and quality of life. Unfortunately, most of the clinically tested GBM-targeted therapies have shown little efficacy so far, such as erlotinib targeting the often overexpressed EGFR. At present, the anti-VEGF bevacizumab is the only drug targeting GBM TMEs that is approved by the US Food and Drug Administration (FDA). In addition, traditional treatments, radiotherapy, and chemotherapy with TMZ have shown to promote TME remodeling [83][84][85][86][87]: -Radiotherapy improves the blood-brain barrier (BBB) permeability to chemotherapy; triggers TME remodeling via increased GAM infiltrates and improved GSC radiation resistance by activating DNA damage checkpoints to repair DNA damage. -Temozolomide (TMZ) triggers a proinvasive TME phenotype by altering proteoglycans and glycosaminoglycans (GAGs) content.
As already stressed, TME niches play a multifaceted role in regulating GSCs and GSC immune evasion is critical to sustaining the tumorigenic process. GSCs express low levels of molecules involved in the processing and presenting of tumor antigens to TCRs, a crucial stimulatory signal to the T-cell response. Consequently, they escape from recognition by antitumor immunity and possibly actively suppress T-cell activation. GSCs express various molecules that deliver either stimulatory or inhibitory signals during direct physical contact with TILs. The balance of these opposing signals regulates the amplitude and quality of TIL response, and aberrant activation of the inhibitory signals, also known as immune checkpoints, is a mechanism utilized by cancer cells to evade immune attacks.
PD-1 belongs to the family of immunoglobulins and is expressed predominantly by activated T lymphocytes. It is often activated by PD-L1, one of the ligands known to be expressed by antigen-presenting cells (APCs), B lymphocytes, and parenchymal cells. Importantly, the expression of PD-L1 has been detected in glioma. In normal conditions, PD-1/PD-L1 engagement occurs controlling a prolonged activation of the immune system, often avoiding autoimmunity processes. Therefore, the PD-1/PD-L1 pathway has been appropriated by tumor cells to resist antitumor responses and facilitate tumor survival. Early immunotherapeutic attempts were focused on targeting PD-1 expression in the general cancer cell population. However, failure of anti-PD1 therapy has been seen in Checkmate 143, 498, and 548 clinical trials [95][96][97].
Thus, the focus has been shifted to PD-L1. In the TME, PD-L1 is regulated mainly by cytokine, while receptor antigen signaling is influenced by hypoxia, cytokines, and oncogenes. GBM cells express PD-L1, which engages with the PD-1 receptor primarily on T cells and attenuates its functions, effectively reducing the antitumor activity of these cells. Nevertheless, studies have shown heterogeneity of PD-L1 expression in tumor mass; a greater expression was observed at the edges of the tumor than in the core. Several phase I and II trials are focusing on PD-L1 in gliomas. Current evidence demonstrates that: (1) PD-L1 quantitative expression has an impact on survival, independently of gender and age [98]. (2) PD-L1 overexpression is significantly associated with poor OS for patients from Asia and America, while no significant association for the survival of patients from Europe. This "ethnic bias" of PD-L1 has been observed in several clinical studies for patients with other solid tumors, such as KEYNOTE-161 in esophageal squamous cell carcinoma and KEYNOTE-063 in advanced gastric or gastro-esophageal junction cancer [98]. (3) IDH1-wildtype status in glioblastoma was PD-L1 expression positive, suggesting PD-L1/IDH1-wildtype association. From a molecular point of view, it could be that IDH1 mutation results in PD-L1 promoter hypermethylation, thus downregulating the expression of PD-L1. Therefore, PD-L1 immune checkpoint inhibitors analysis might not be advisable because of the globally low PD-L1 expression in patients with IDH1-mutant glioblastomas.
To sum up, higher expression of PD-L1 (both at protein and mRNA levels) is linked to a worse outcome. PD-L1 may expand and maintain immunosuppressive Tregs, which are associated with decreased survival in glioma patients. Beyond the direct impact on effector cells, the blockade of the PD-L1/PD-1 axis may reduce Treg expansion and further improve T cell function [98].

Conclusions
Microenvironmental contribution seems therefore critical to gaining a better understanding of the unique challenges GBM poses. Aggressive growth, inevitable recurrence, and scarce response to immunotherapies are driving the necessity to focus on GBM TMEs from a different perspective to possibly disentangle its role as a fertile 'soil' for tumor progression and identify in it feasible therapeutic targets. Against this background, our systematic review confirmed: (1) Microglia play a paramount role in the maintenance of immunological homeostasis and protection against autoimmunity and its activation pattern at the TME level, polarized toward an M2 phenotype as selected by environmental pressure. This suggests that further investigation of microglia phenotypic characterization at the microenvironment level (M1 vs. M2 phenotype) is needed. (2) Microglia crosstalk with dedifferentiated and stem-like glioblastoma cells in perivascular and perinecrotic hypoxic niches, where they start crosstalk with the staminal compartment ultimately promoting disease progression and relapse after treatments. (3) Microglia demonstrate migratory behavior with respect to infiltrative margins of tumor cells. However, there are still many issues to be investigated. While the classification of macrophages or microglial cells into the M1 or M2 polarized state is a well-established approach in most preclinical models, the same is not true in the clinical research setting, because of the high degree of diversity and plasticity shown by these cell types. Therefore, dichotomizing GAMS into M1 and M2 activation status might be over simplistic as, indeed, a clear distinction between these phenotypes cannot be clearly distinguished. The resulting definitions of transcriptomic-based functional phenotypes of GAMs from human and experimental rodent gliomas are conflicting and indicate a mixture of M1 and M2 phenotypes. Cells within the tumor often display a complex pattern of phenotypes, upregulating both M1 and M2 molecular markers, and the prevalence of one phenotype on the other might also depend on the stage of disease [99].
In addition, the danger of 'oversimplification' goes along with the lack of universally recognized markers of the functional phenotype of GAMs. Although the association of GAM subtypes and patient overall survival has been observed in several papers still no consensus exists on reliable gene expression-based markers [100].
To further confound the research scenario on the topic it should also take into account that it is hard to define realistic research models as significant differences between human and mice models exists in terms of microglia polarization and inconsistency between rodent and human GAMs, regarding markers, has been reported. Despite some similarities, the mouse and rat models represent different pathways of GAM activation; comparative analysis of GAM transcriptomics across different in vivo models of human, mouse, and rat, failed to reproduce consistent microglia phenotypes that could be classified according to previously reported gene signatures and showed remarkably low similarity between models [99].
Despite the growing body of evidence on the topic reported in this paper, many aspects are yet to be explored in the field [4,26,45,54,98,101,102]. The correct and extensive understanding of microglia-glioma crosstalk could help in understanding the physiopathology of this complex disease, possibly opening scenarios for improvement of surgical strategies and medical treatments. Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.

APCs
Antigen presenting cells BBB Blood brain barrier CAMs cell adhesion molecules CNS