Macrophages are one of the major immune infiltrates in pediatric glioma (12).Tumor-associated myeloid (TAM) cells are generally separated into bone-marrow derived macrophages (BMDM) or tumor-associated macrophages from a monocytic origin (moTAM) and macrophages from a microglial origin (mgTAM). In general, LGG tend to have more resident microglial TAMs, while the HGG have the highest proportion of infiltrating monocytic TAMs (18).

NK-cells are part of the innate immune system and function as cytotoxic effector cells (35). In PA, CD56+ NK cells are virtually absent compared to GBM (27). In a TCGA study, LGG with high expression of the Killer cell lectin-like receptor (KLR) family (encoding NKG2 NK-cell receptors), was associated with an improved prognosis (36). In the subtypes of pHGG, more NK-cells have been observed in hemispheric WT pHGG compared to brainstem H3 K27M pHGG (29). In another study, however, no significant difference was found in the infiltration of NK-cells between pHGG and normal control tissue (35).

Eosinophils are another understudied cell type in glioma, with only few studies present. In the study of Grabovska et al., eosinophils were found to contribute up to 13% of non-cancer cells in the tumor; and this tends to collide with the types of brain tumors that have high numbers of Tregs and NK-cells. The number of eosinophils present was negatively associated with tumor grade and there were significantly less eosinophils in G34-mutated GBM (14). Regarding the location of the tumors, significant differences were identified between WT hemispheric pHGG and K27M brainstem pHGG, with more eosinophils in the hemispheric tumors (29).

In an early study by Fossati et al., the level of infiltrating neutrophils correlated with tumor grade (37). Other studies confirmed this finding and showed that neutrophils promote the proliferation of glioma stem cells. Additionally, the neutrophils in the tumor possibly promote angiogenesis by inducing a proneural to mesenchymal subtype transition, through the upregulation of S100A4 (38). In pHGG, neutrophil infiltration was shown to be a negative prognostic factor for both hemispheric and brainstem locations (29).

In adult GBM patients, the pretreatment neutrophil-to-lymphocyte ratio (NLR) correlated with OS, with a high pre-treatment NLR correlating with a high neutrophil infiltration and low T-cell infiltration in the tumors, which predicted a poor OS (39).

T-cells are one of the better studied cell types in the TME. In general, the lower grade tumors have a higher T-cell infiltration compared to higher grade tumors (26, 40). But even within the lower grade tumors, the levels of infiltration can be highly variable (40). For instance, the study by Robinson et al. identified higher T-cell infiltration and -density in pleomorphic xanthoastrocytoma (PXA) and ganglioglioma (GG), even though the study population was small. These two groups also had higher proportions of T-cell clusters (defined as T-cells with more than 10 other T-cells close to them in a 30 µm radius) compared with pilocytic astrocytoma (PA) or pHGG, which is an interesting finding, since PXA and GG are generally low in mutational burden. Additionally, the presence of the stem cell marker SOX2 in the tumor cells was inversely correlated with the T-cell infiltration and TCF1+ T-cells were enriched in perivascular areas. Tang et al. also studied the TME of PXA and GG using methylCIBERSORT, and observed that they exhibit lymphocytic infiltration, with significant increases of CD8+ T-cells in PXA, and suggested that this could be used to enhance the effects of immunotherapy (41). While they observed significant increases in CD8+ T-cells in PXA, no significant differences were seen in CD14+ macrophages, CD19+ B-cells, endothelial cells or fibroblasts (41). In a TCGA study of LGG, enrichment of memory CD8+ T-cells was associated with a better prognosis (36).

Considering pHGG, Lin et al. found that the TME in DIPG is rather non-inflammatory, with low levels of infiltrating T-cells, combined with low to absent expression of chemo- and cytokines (30). The location and type of the underlying driver mutation in pHGG also seem to influence the type and number of T-cells present. For instance, more CD8+ T-cells can be found in hemispheric H3 WT pHGG compared to H3 K27M-mutated pHGG located in the brainstem (29). This was also found in the study by Lieberman et al., where increased numbers of CD8+ T-cell infiltration in pLGG and pHGG were observed in comparison to control tissue, but not seen in DIPG (28). Histone H3-mutant glioblastoma are depleted of lymphocytic clusters compared to IDH-mutant and WT glioma. This compares to the immunologically quiet immune subtypes as described for LGG (42).

T-helper cells (CD4+) are important in the mediation of the anti-tumor immunity of the CD8+ cytotoxic T-cells. In the study of Grabovska et al., the infiltration of T-helper cells decreased with increasing WHO grade. Additionally, in H3 G34-mutant pHGG, higher than median concentrations of T-helper cells were associated with poor OS (14).

Additionally, the TME in glioma can have hypoxic regions and is low in nutrients, which can negatively impact the function of the T-cells, resulting in T-cell suppression, reduced memory T-cell survival and inhibition of T-cell proliferation (43).

An exception to the fact that pediatric glioma are generally cold tumors, are the tumors arising in children harboring germline biallelic mismatch repair deficiency (MMRD). In these patients, there is the potential to use immune checkpoint inhibition (ICB), while the efficacy of ICB is lower in the non-MMRD tumors, based on a more limited amount of tumor-infiltrating lymphocytes (44). It was shown that a high CD8+T-cell infiltration in tumors with MMRD + PPD (polymerase proofreading deficiency) predicted response and an improved OS (45). However, a large study investigating adult glioma with acquired MMRD after temozolomide treatment, observed that the hypermutation results in subclonal alterations which do not generate optimal antitumor T-cell responses. Both the MMR-deficient and the MMR-proficient glioma had a significant lack of T-cell infiltrates (in this study compared to MMR-deficient colorectal carcinoma) (46).

Little is known about the spatial organization of the TME in pediatric glioma. Pediatric HGG showed the most positive CD3- and CD8-staining around the perivascular spaces and necrotic areas (13), a phenomenon that was not observed in PA (27). In a study by Schaettler et al., the immunogenic profile of spatially distinct regions was studied from both glioma and brain metastases in adult patients. Glioma were mutationally more heterogeneous and subclonal and the intratumoral TCR repertoire was more heterogeneous in spatially distinct regions with locally expanded T-cell clones. In their study, they did not observe significant intratumoral differences in the immune infiltrate or activation, despite the heterogeneity in the variants and neoantigens within glioma based on bulk RNA-sequencing (47). Another study in adult glioma showed that T-cells infiltrating in GBM had high expression of immune checkpoints (PD-1, LAG-3, TIM3 or TIGIT), which were correlated with a loss of effector function (48). It remains to be determined whether the same principles apply to pediatric tumors.

B-cells are involved in the adaptive immune system and are responsible for antigen presentation and antibody production. Their role in tumor immunity is still somewhat controversial, and can vary among tumor types (53).

In comparison with previously discussed cell types, B-cells have not been studied intensively. As mentioned before, differences were observed concerning the number of T-cells present in PXA and GG, but in this study, no significant differences in the number of B-cells were observed (41). In pHGG, differences were observed, based on the underlying molecular alterations of the tumor. In the hemispheric WT pHGG, memory B-cells, compared with CD4+ regulatory T-cells and activated dendritic cells had a positive prognostic impact (29). G34-mutated tumors had an increased B-cell gene expression prevalence (42).

Another study in adult glioma identified B-cells to influence the other cells of the TME. These B-cells differentiate into regulatory B-cells (Bregs) under the influence of PlGF (placental growth factor)-carrying exosomes, that in turn suppress the activity of CD8+ T-cells (54).

Dexamethasone, a commonly used corticosteroid, is frequently used in patients with brain tumors for its anti-inflammatory function and to treat brain swelling due to the tumor or treatment. However, it is known for its toxicity on immune cells, which can in turn have effects on the composition and function of the TME. Furthermore, it influences the blood-brain barrier (BBB), closing the latter and possibly resulting in a lower tissue penetration of systemic therapies (62).

Temozolomide (TMZ), a commonly used alkylating agent which excellent brain penetration, can cause lymphopenia and is also used for lymphodepletion prior to the use of T-cell-based immunotherapies (43). TMZ has an immunomodulatory effect, such as the enhancement of antigen-specific T-cell proliferation (63). It induces reactive astrogliosis, similar to RT and is discussed later on (59).

Regarding other types of chemotherapy, little is known on their effects on the TME. Several large studies, such as the HIT-GBM studies (64, 65), have been conducted, however these only investigated responses and toxicities and not the effects on the TME. Preclinical studies using co-cultures of glioma cells and astrocytes treated with TMZ and vincristine, show evidence of a chemoprotective effect based on direct contact between astrocytes and glioma cells (66). Other preclinical work showed that there was an increased chemoresistance in adult GBM for cisplatin, TMZ and etoposide in hypoxic conditions (67).

Anti-VEGF therapy (bevacizumab), was initially designed to affect brain tumor vascularization and has been described to reduce brain edema by improving the structure of the blood-brain-barrier (68), and as such can also affect the influx of inflammatory cells towards the tumor (69). Bevacizumab did not improve event-free survival (EFS) in pediatric patients with newly diagnosed HGG when added to standard radiotherapy and temozolomide treatment (HERBY trial) (70). When further analyzing these cases on a molecular and pathological level, however, the presence of high levels of CD8+ T-cells in the central tumor area was significantly associated with a better OS when receiving bevacizumab in addition to the standard RT/TMZ treatment (71). Some tumors acquire resistance to the treatment, which can be due to an increased recruitment of neutrophils during this treatment, promoting glioma progression. This could be due to S100A4, and upon depletion of S100A4, anti-VEGF therapy in mice bearing glioma was more efficacious (38). Upon anti-VEGF treatment, a correlation was seen between the decrease in tumor vascularity and reduction in myeloid cell infiltration in orthotopic glioma (72). Another study in adult GBM patients also indicated that only patients with a high peripheral neutrophil count benefited from treatment with bevacizumab (73).

Radiotherapy (RT) is frequently used in the treatment of pediatric brain tumors, but has various side effects. Due to the possible neurocognitive side effects, RT is generally avoided in patients younger than 5 years. Furthermore, in patients with underlying neurofibromatosis type 1 (NF1), RT is generally avoided since they have an increased risk for secondary malignancies or cerebral vasculopathy (74, 75).

RT induces lethal DNA damage to the cells of the tumor and its surroundings. It results in the release of tumor-associated antigens, which then activate antigen-presenting cells (APC). These migrate to the draining lymph node where cytotoxic T-lymphocytes are primed, resulting in an adaptive immune response (76). Furthermore, radiotherapy can induce immunogenic cell death at the lesions borders, caused by the danger-associated molecular patterns (DAMP) that recruit immune cells and an increase in MHC1 expression (77). Therefore, RT can increase the tumor immunogenicity, turning a “cold” TME into a “hot” TME and thus sensitizing them for immunotherapy (77, 78). However, one has to consider that it is sometimes necessary to associate dexamethasone to treat the side effects of RT, which could theoretically counteract the tumor immunogenicity.

RT also induces changes in astrocytes, inducing reactive astrocytes, which promote the stemness in GBM, which could possibly contribute to recurrence in a later stage. Astrocyte-derived transglutaminase 2 (TGM2) has been identified as a potential radiation-induced modifier of the TME in adult GBM, protecting against cell death through radiation (79).

In other cancer types, the combined use of radiation therapy and inhibition of TGFβ (using neutralizing antibodies), resulted in a T-cell mediated rejection of the irradiated tumor and non-irradiated metastases in mice models (i.e. generating an ‘in situ’ tumor vaccine). Additionally, an improved outcome was seen when this was combined with PD-1 blockade (80).

Furthermore, radiotherapy could also induce an immunosuppressive state in the TME on the long term. Fibroblast attracted to the tumor (cancer-associated fibroblasts, CAF), can influence the immune cells and tumor progression through the release of cytokines and growth factors (81).

The classically used treatment options for pediatric glioma thus influence the TME in several ways and can have opposite effects. Additionally, the ‘side’ effects of these treatments provide other therapeutic options.

Besides the classic treatment options, a lot of research is ongoing to improve the PFS and OS of pediatric glioma patients. Of course, these novel treatments can also influence the TME. Keane et al. showed that the pHGG cells can polarize microglia towards a tumor-supportive phenotype, a phenomenon that could be reverted by reducing Ezh2 gene expression in the microglia using EZH-inhibitors, which caused a decrease in the global H3K27me3 expression levels, and promoted an activation state in the microglial cells, resulting in antitumoral properties (82). Yin et al. had similar findings using co-culturing models where EZH2 was suppressed, resulting in an increase of M1-markers and an increase in the phagocytic capacities of microglia (83).

The mTOR pathway is another promising treatment option. In the study of Hsu et al., the combined use of rapamycin and hydroxychloroquine influenced the polarization of the TAMs from an M2 to an M1-phenotype. In such context, ICB could be more beneficial (84).

HDAC8-inhibition using the specific inhibitor PCI-34051 causes an enhanced activity of NK-cell in vitro, and increases in the infiltration of NKG2D+ and CD69+ NK-cells in vivo. Additionally, HDAC8-inhibition inhibits chemotaxis of microglia in the tumor in vitro. Further in vivo data shows that HDAC8-inhibition stimulates the expression of pro-inflammatory genes in TAM, indicating that HDAC8 on itself drives the TAMs to an anti-inflammatory phenotype (85).

Furthermore, research is ongoing on the use of nanoparticles in the treatment of pediatric glioma. These might offer a better drug delivery to the tumor and lower possible side-effects. Additionally, nanoparticles also stimulate the immune system and thus influence the TME in these tumors, however, their specific effects have not been studied in pediatric glioma (86).

Both PD-1 and PD-L1 are frequent targets in ICB. Programmed death ligand 1 (PD-L1; B7-H3; CD274) is the ligand of programmed death receptor 1 (PD-1) and in normal situations, plays a role in the maintenance of immune cell tolerance (89).

The efficacy of ICB has been studied in children, however with unfavorable success. This has been associated with the low mutational burden in these tumors, the low expression of major histocompatibility complex and the high prevalence of macrophages in the TME (44, 45).

The exception here seems to be hypermutant tumors, as already discussed before, driven by mutations leading to mismatch repair deficiency (MMRD) (44). Children with constitutional MMRD develop several types of tumors and have a poor prognosis, frequently not reaching adulthood (90). Recently, a trial was conducted where pediatric patients with MMRD and/or polymerase-proofreading deficiency (PPD) tumors were treated with PD-1 inhibitors (pembrolizumab/nivolumab). Here, an OS of 39.3% and a progression-free survival of 26.9% was reported for recurrent/progressive CNS tumors, which is a major improvement compared to their standard (unsuccessful) treatment. An improved survival upon ICB could be seen in tumors with high levels of single nucleotide variants (SNVs) and total MS-indels. They also discovered that clonal mutations were strong predictors of survival and response. Furthermore, a high CD8-T cell infiltration was predictive for response and improvement of survival, while no association between CD68 with response or survival could be identified. After all, the pediatric MMRD+PPD gliomas responded better to ICB compared to their adult counterparts (45).

Bockmayr et al. identified within the mesenchymal subgroups of HGG, two immunologically infiltrated clusters, where an increased gene expression indicative of a higher infiltration of CTL in one of the subgroups could be used for ICB. Additionally, they showed that in glioma which were K27-mutated, PD-L1 and CTLA4 correlated with a worse prognosis. This was the opposite for G34-mutated tumors, which also seem to rely more on TGFB1 and HAVCR2 for their immune escape. Therefore, ICB targeting the PD-1/PD-L1 axis could be beneficial for patients with DMG with H3 K27M mutations (42), while targeting HAVCR2/TIM3 could be an alternative approach (91).

TGFB has been suggested to play a role in the resistance to PD-L1 therapy, since it would be contributing to the exclusion of T-cells in the tumor. TGFB would have an inhibitory effect on the cytotoxic activity of T-cells and would promote the generation of regulatory T-cells (92).

A recent clinical study investigated the effect of MDV9300 (pidilizumab) in patients with DIPG (NCT01952769), which is an immune-modulating IgG1 monoclonal antibody targeting the NOTCH ligand DLL1 and to a lesser extent PD-1 (93, 94), which seems to enhance endogenous antitumor hematologic malignancies. Although a small study, the preliminary results implicated a tolerable safety profile and suggests some clinical activity (94).

Vaccine therapy comprises several types of therapy, such as dendritic cell vaccination, vaccination with tumor lysate or peptide vaccination with glioma-associated antigens. These vaccines could possible synergize with CAR T-cell therapy (43).

In DC vaccination therapies in DMG, the lack of immune cells may hamper its efficacy, suggesting that strategies to enhance immune cell trafficking should potentially be combined (95).

The heterogeneity in tumors can cause difficulties for the use of vaccines since not all tumor cells express the same tumor antigens, resulting in the escape of some subclones of the tumor to treatment. This might be overcome by the use of vaccines that target multiple neoantigens.

Pollack et al. investigated the use of glioma-associated antigen (GAA) epitopes in a peptide vaccination study in children with HGG and recurrent LGG, showing preliminary evidence of immunological and clinical response. However, no assessments on the changes in the TME itself were performed (96, 97).

The H3 27M in pHGG has been identified as a possible target in vaccine strategies. It was shown that that the H3 K27M can act as an immunogenic neoepitope, and is targetable by a specific peptide vaccine. In a humanized mouse model, it resulted in IFNγ immune responses, induced by both CTLs and Th1-cells (98). Others developed a high-affinity TCR that recognizes an H3.3K27M epitope which is expressed by H3.3K27M+HLA-A*0201+ glioma cells. This finding has been used to start a vaccine trial with the H3.3K27M26-35 epitope peptide in children with H3.3K27M+ DIPG or HGG (PNOC-007, NCT02960230) (99). However, the study is still enrolling patients and no results have been described so far.

Chimeric antigen receptor (CAR) T-cell therapy is based on the immune system of the patient itself, where the immune cells of the patient are engineered to target a tumor-selective epitope. Theoretically, this could result in more specific targeting of the tumoral cells, while the normal tissue (which does not express the neo-antigen) could be spared. Additionally, targeting tumors with CARs is not dependent on MHC-antigen presentation. This could be beneficial in those tumors with a low tumor mutation burden or with defects in antigen presentation (43). CAR T-cell therapy has been used before in B-cell malignancies, with good clinical activity (100). However, in adult brain tumors, this therapy had variable results, where a lot of the determinants are still incompletely understood (43). For pediatric brain tumors, several targets are being evaluated, with several still in clinical trials (101).

Haydar et al. investigated 49 patient-derived orthotopic xenograft models of pediatric brain tumors and established a rank order of expression for 5 potential targets (B7-H3, GD2, EphA2, IL-13Ra2 and HER2), and found that B7-H3 and GD2 were highest in expression (102). Trials with CAR T-cells against B7-H3 (CD276) are still ongoing, but some preclinical data shows possible benefit from this therapy (103). Anti-GD2 CAR T-cells with an incorporated 4-1BBz costimulatory domain, demonstrated a robust antigen-dependent cytokine generation and DMG cell killing in vitro (104). In some of the first phase I clinical trials in a limited patient group (n=4), which only recently published its first results, the anti-GD2 CAR T-cells showed promising results on the tumor volume, although not without therapy-related side effects (105).

Importantly, it was observed that the expression of these target antigens is heterogeneous, illustrating the need to target multiple antigens, to avoid an immune escape (102). Additionally, several reports found that there can be a loss or decrease in the target antigen expression, implying that in the further development of this therapy for brain tumors, additional strategies will be necessary to minimize or avoid antigen escape (43).

The TME of the tumors can influence the effect of CAR T-cell therapy, where suppressive monocyte-derived TAMs could influence the effector functions and microglia-derived TAMs are needed to promote the effector functions of the T-cells (101). Tregs could also potentially influence the efficacy of CAR T-cells, where they can augment the growth of the tumor and suppress the T-cell effector molecules in the TME of glioma. This can result in possible negative effects on the function of the CAR T-cell (101). The CAR T-cells on the other hand can also induce changes in the TME. For example, an increased production of INFγ caused by the CAR T’s, can induce immunostimulatory effects and inhibit the macrophages to differentiate to a M2-phenotype (78).

The use of combination therapies was also proposed, such as the addition of PD-L1 inhibition, together with CAR T-cell therapy. This might overcome the immunosuppressive TME, and therefore also improve the persistence and functioning of the CAR T-cells. Clinical phase I studies (in adults, NCT03726515) are currently being performed but no results have been reported yet (43).

Oncolytic virotherapy delivers a live but modified virus that specifically targets tumor cells. The advantage of this type of therapy is that the tumor cells are directly targeted or that the host immune system is activated to attack the tumor cells by stimulating the effector function and antitumor activity of T-cells in the TME. Oncolytic viruses have also been suggested to be used in combination with CAR T-cells, since these viruses can be engineered to express transgenes which can target the suppressive TME (43).

A recent pilot study showed an increased influx of CTLs, combined with a greater proportion of macrophages and microglia after treatment with oncolytic HSV G207 therapy in a pediatric patient with pGBM. This would suggest that oHSV therapy would increase or amplify the anti-tumor immune response (106). Additionally, these T-cells also expressed higher levels of checkpoint proteins such as CTLA-4, PD-1, PD-L1 and IDO after treatment (106).

Tejada et al. describe a clinical phase I trial in which the oncolytic virus DNX-2401 (formerly Delta-24-RGD) is administered intratumorally to children with newly diagnosed DIPG. Preliminary results indicate that this therapy is well tolerated, but longer follow-up and enrolling additional patients to further assess efficacy and safety is still needed (107).

Mendez et al. generated an endogenous mouse model of ACVR1-mutant brainstem glioma and tested the effects of immune stimulatory gene therapy using an adenovirus that expressed thymidine kinase and fms-like tyrosine kinase 3 ligand (Flt3L). The adenoviral delivery of the TK/Flt3L caused an anti-tumor immunity with the recruitment of anti-tumor specific T cells, which even resulted in an increased median survival. The treatment with TK/Flt3L resulted in an increase in tumor-specific CD8+ T-cells and an increased toxicity (enhanced IFNγ production) (108). Before, research of the same group had already shown the efficacy of an immune stimulatory gene treatment in rat and mouse models of adult GBM, leading to tumor regression and long term survival, and went further on to a clinical trial in adult GBM patients [NCT01811992 (109)] (110, 111). Other types of gene therapy for glioma fall beyond the scope of this review and are described elsewhere (112).

Recently, an engineered human NK-cell therapy for adult glioblastoma was developed, with NK-cells having multi-CARs to target GD2 and NKG2D. In addition, an antibody fragment that blocks the activity of CD73 is released via the cleavage of a tumor-specific linker, dependent on proteases in the TME. However, this is still preclinical data, based on studies in a patient-derived xenograft model of adult GBM (113). Shortly, the NCT05887882 trial (phase 1) will start in pediatric glioma patients, concerning the intratumoral injection of ex vivo expanded NK cells (see Table 2).