Brain Tumor Microenvironment and Host State: Implications for Immunotherapy

Glioblastoma (GBM) is the most lethal and common primary brain tumor in adults. Despite an aggressive standard of care treatment regimen of surgical resection, radiochemotherapy, adjuvant chemotherapy, and tumor-treating fields prognosis remains poor with a 2-year survival rate of only 43% (1). Although the implementation of immunotherapy has proven extremely successful in more immunogenic cancers, no survival benefit has been observed in patients with GBM thus far. (2). The GBM TME shares components with these more treatable cancers but is also made unique by the brain tissue-resident cell types. In addition to these unique cellular components it is also insulated by the blood–brain barrier (BBB), which contributes to the brain being widely considered a relatively immune privileged organ. Immune privileged organs have tightly regulated immune responses, which leads to a naturally more immunosuppressive environment.

In addition to tumor cell–extrinsic components of the TME that lead to poor treatment response, there are several tumor-intrinsic properties that lead to poor immunogenicity and immunosuppression. GBM has recently been characterized into several subtypes based on the dominant aberrant transcriptional program. These are termed proneural (PN), mesenchymal (MES), and classical (CL; ref. 3). There is a great degree of heterogeneity in these subtypes between patients, as well as within an individual tumor (4). In addition to the molecular subtypes based on global transcriptional programs, gliomas, including GBM, have also been stratified according to specific genomic aberrations: mutations in the Telomerase Reverse Transcriptase (TERT) promoter, alterations in the isocitrate dehydrogenase-1 (IDH) gene, and co-deletion of chromosome arms 1p and 19q (5). Regardless of GBM stratification method, mutations found in GBM are rarely homogenous and few of these mutations result in a surface protein modification that is unique to the brain tumor (6). As such, implementation of antigen-specific therapies is proving difficult and often results in immune escape (7).

In this review, we discuss the components of the brain TME and how they may contribute to treatment response. We will also briefly review therapies that aim to directly target the TME to lower the barrier of immunosuppression in the hopes of making antigen-specific therapies more effective.

The TME of GBM is unique in its cellular composition and accessibility to immune cells. The factors that make the TME unique are also what contribute to its highly immunosuppressive and “cold” TME phenotype. Unlike the consistently mutating tumor cells, the stroma of the TME is a genetically stable therapeutic target. Reducing the immunosuppression caused by these stromal cells has the potential to promote functional effector T-cell infiltration and create new opportunities for treatment. Here we discuss how these nonimmune (Fig. 1) and immune (Fig. 2) stromal elements contribute to immunosuppression and the “cold” TME phenotype.

The BBB is functionally distinct organ structure of the brain, which has poor permeability to many frontline therapeutics and impedes the migration of some immune effectors under certain conditions. In the naïve state, the BBB provides a significant restriction to the permeability of many therapeutics especially large or hydrophilic molecules. However, the BBB loses its integrity in many pathologic situations including primary and malignant brain tumors and becomes increasingly permeable to therapies it would have otherwise blocked. Even within these pathologic situations there is a difference in permeability between gross tumor and infiltrated brain (8, 9). A high degree of vascularity with atypical organization and reduced structural integrity is a common characteristic of GBM. This leakiness results in high interstitial fluid pressure, a great degree of hypoxia and necrosis, as well as edema (10). Vascularization of brain tumors, or angiogenesis, is considered unique due to the biological homology between vascular and neural networks (11). It has recently been suggested that neural stem cells and glioma stem cells can differentiate into endothelial cells within the glioma vasculature (12, 13). This ability of glioma stem cells allows them to form glioma stem cell reservoirs in the perivascular niche (PVN), where they are insulated and can safely proliferate (14, 15).

Glioma stem cells (GSC) are associated with the endothelial cells of the PVN (14, 15). The number of GSCs associated with vessels in the TME strongly correlates with increasing tumor grade (16, 17). A feedback loop exists between the endothelium and the GSCs as the endothelium releases factors that drive tumor sphere formation and GSCs release factors that accelerate angiogenesis (16). Nitric oxide (NO) is one such endothelium derived factor that reinforces stem-cell like characteristics in GSCs (18). GSCs have been shown to additionally recruit monocytes to the TME and polarize them to a protumor phenotype via secretion of CCL2 and CSF-1 (19). GSCs also directly inhibit T-cell activation, proliferation, and induce T-cell apoptosis (20). Finally, it has been suggested that GSCs induce functionally active regulatory T (T_Reg) cells (20). These effects combined result in GSCs mediating a great degree of immunosuppression while remaining difficult to target.

Astrocytes provide structural support in the brain by maintaining homeostasis. They are typically localized to the PVN and play an important role in maintenance of the BBB (21). Astrocytes are thought to have protumor functions via secretion of neurotrophic factors which support proliferation of glioma cells (15). In the naïve brain-activated astrocytes supply growth factors and cytokines to enable the repair of brain tissue during different forms of injury. This process is referred to as reactive gliosis, which is one mechanism of wound healing in the brain (22). In the TME these growth factors have been shown to support tumor growth and mediate resistance to therapy (23). In addition to supplying growth factors, they also secrete metalloproteinases which create a favorable environment for tumor invasion (24).

Neurons are a brain-specific cell type, like astrocytes, which are thought to contribute to the creation and outgrowth of tumors. Neurons provide mitogenic signals within the brain to drive neural stem cell growth (25). Recent studies show that neuron-derived neuroligin-3 (NLGN3) increases proliferation of tumor cells via tumor-intrinsic PI3K signaling. It was also shown that in human GBM NLGN3 expression inversely correlates with survival (26). In breast cancer brain metastases, it has also been shown that increased neurotransmitters released by neurons serve as an oncometabolite (27). Whether this process occurs in GBM as well remains to be determined, but it serves as an example where neuron-derived products serve protumor roles in the brain TME.

Tumor-infiltrating lymphocytes (TIL) have the potential to exert both pro- and antitumor functions in the TME. T cells are the primary lymphoid component of the TME but compose less than 0.25% of cells isolated from human GBM biopsies (28). CD8⁺ cytotoxic T lymphocytes (CTL) are considered critical for tumor clearance, but account for less than a quarter of the already sparse TIL population of the TME (28). Functional characterization of the CTLs found in the TME has shown that these cells have impaired effector functions and an exhausted phenotype, rendering them ineffective in their role as cytotoxic lymphocytes (29). Similarly, CD4⁺ T helper cells, which typically have antitumor functions, may correlate with poor survival outcomes (30).This is likely explained by a large percentage of the CD4⁺ TIL population being T_Reg cells, and the remainder being functionally exhausted (31). FOXP3⁺ CD25⁺ T_Reg cells, a CD4 subset, are functionally immunosuppressive. Efforts aimed at depleting these protumor lymphocytes have had modest effects at increasing CTL function in GBM (31).

Bone marrow–derived macrophages/monocytes (BMDM) are the primary immune cell in GBM and compose up to 30% of the tumor mass (32). There are two distinct macrophage populations in the GBM TME, BMDMs, and microglia (33, 34). Collectively, these are referred to as tumor-associated macrophage (TAM). Unlike BMDMs, Microglia develop from yolk sac progenitor cells, and are not replenished postnatally by hematopoiesis (35). During tumor progression, monocytes and macrophages can extravasate into the TME through the compromised BBB (36). Conventionally, macrophages have been thought to exist in either the inflammatory (M1) or wound healing (M2) phenotype. Recent works suggest that this is a gross oversimplification due to the staggering functional diversity and plasticity of TAMs across tumor types (37). In the GBM TME TAMs have a distinct protumor role and their accumulation correlates with tumor grade (34). Functionally, TAMs in the GBM TME, are only producing low levels inflammatory cytokines and lack the ability to aid in T-cell responses via costimulation (38). In addition, they have been found to be great contributors to the immunosuppressive TME via release of soluble factors that dampen the immune response (39).

Natural killer (NK) cells are innate lymphoid cells that identify and kill tumor cells by sensing danger or damage signals (40). NK cells are found in GBM and have been shown to be effective at inducing lysis in GSCs (41). Unfortunately, GBM is known to express HLA-G, which acts as an inhibitory ligand for activated NK cells (42). This likely aids in their evasion from NK-cell–mediated cell killing. Macrophages have been shown to mediate NK-cell activation, but lose this priming ability when macrophages adopt a protumor phenotype (43). A definitive link between TAM polarization and NK-cell activation has not yet been proven in GBM however. Like TAMs, NK cells are a component of the GBM TME that possess the ability to lyse tumor cells in an antigen-independent manner but are rendered functionally suppressed.

The interplay between the cancerous cells and the surrounding stroma is critical in developing the TME present with gliomas. The stroma response to tumor-extrinsic factors such as constant cycles of hypoxia, acidosis, necrosis, angiogenesis, and granulation (44, 45). These processes were originally perceived to be the responsible for the immunosuppressive nature of the TME. This constant state of “chronic inflammation” was what originally granted tumors the term “the wound that never heals” (46). However, this original belief fails to explain how tumors of conventional histology and location can have disparate TME between patients (47). Factors such as age, HLA-type, and genetics may explain some differences between patients, however, these factors do not account for the recent observations where different tumor lesions located within the same organ in a patient can have different TME characteristics (48). Therefore, although tumor-extrinsic factors definitely contribute to the formation of the immunosuppressive TME, tumor-intrinsic factors in the form of tumor genetic or epigenetic changes have recently been demonstrated to play a critical role in shaping this milieu (Fig. 3).

The results obtained from the clinical trials using checkpoint blockade strategies have highlighted the relevance of the role of tumor-intrinsic factors in the establishment of the TME and response to therapy. This field has divided tumors into “cold tumors” versus “hot tumors,” where “cold tumors” are characterized by the lack of T cell infiltrate within the TME, whereas “hot tumors” are infiltrated with mostly CD8⁺ T cells, many of which appear locally activated yet are extrinsically suppressed (49). It is currently believed that “cold tumors” are less responsive to immunotherapies whereas “hot tumors” are primed to respond. The majority of GBM samples display a “cold tumor” phenotype with few CD8⁺ TIL.

GBM is a high-grade tumor that arises from astrocytic origin and it is largely confined to the central nervous system (CNS). Extrinsic factors that directly impact the TME and T-cell penetration have been previously characterized. As mentioned above, the presence of a BBB, neurons, microglia, and surrounding astrocytes are critical in shaping the TME. The direct interaction of these components with the tumor appears to be regulated by tumor-intrinsic mutations. Tumor-intrinsic factors resulting from the PI3K, Ras–MAPK, WNT/β-catenin, p53, and IDH pathways have been shown to modulate the TME and facilitate the emergence of “cold tumors” (50). Coincidentally, despite GBM containing an average of 40 mutations per tumor, whole genome sequencing of GBM tumor samples have revealed that the main pathways altered in GBM are PI3K, Ras–MAPK, p53, and IDH (3, 51). The WNT/β-catenin pathway has not been shown to be commonly mutated yet the pathway appears to be active within a proportion of GBM tumors (6). Although most of the studies of these pathways in GBM are focused on how these regulate proliferation, survival, and invasion, more research is required to determine their role in shaping immune resistance.

PI3K is a major pathway active in GBM, which can be activated in multiple ways. Signaling from the EGFR (52), and c-mesenchymal–epithelial transition receptor (c-met) are common pathways responsible for PI3K activation (53). Loss-of-function mutations of the tumor suppression PTEN, which results in constitutively active PI3K signaling, is commonly seen in GBM (54). Active PI3K in GBM has been shown to result in the expression of the immune checkpoint ligand PD-L1 (55). The interaction of PD-L1 with the receptor PD-1 is one of the main pathways to suppress T-cell responses within the tumor context.

The Ras–MAPK pathway is another major pathway in GBM downstream of the EGFR (52) and c-met receptors as well (53). Many GBM samples harbor loss-of-function mutation on the NF1 gene, which results in further Ras activation (56). Active Ras–MAPK pathway has also shown to contribute to modulation of the TME through the induction of IL6 mRNA (57). IL6 is a pleiotropic cytokine, which is known to induce CCL2 expression. CCL2 is an abundant chemokine within GBM tumors and has been shown to mediate recruitment of monocytes into tumors. IL6 also leads to the activation of NF-κB and STAT-3 in infiltrating monocytes and GSCs, which are known to express the IL6 receptor (58). Active STAT-3 promotes the recruitment of monocytes via CXCL1, CXCL2 expression, and the induction of a suppressive TME by inducing macrophages to the protumor phenotype (58). In addition to the induction of IL6 Ras–MAPK signaling process includes the activation of p38 MAP kinase (59), which plays a critical role in the induction of TGFβ mRNA (60). TGFβ is a central pleiotropic cytokine in shaping the immunosuppressive TME, inducing the protumor TAM phenotype, impairing dendritic cell (DC) migration and cytokine sections, inhibiting T-cell responses, and limits the infiltration of leukocytes into the tumor (61). Although it is not well known if GBM mutations have an intrinsic effect in regulating TGFβ expression, posttranscriptional regulation is critically dependent on the Ras–MAP–p38 pathway.

The WNT/β-catenin pathway has recently been explored in more detail in GBM samples. Although this pathway does not appear to be frequently mutated in GBM, there is significant epigenetic regulation which leads to elevated levels of WNT proteins such as Wnt5A (6). WNT proteins bind to Frizzled receptors and signal through the β-catenin protein, which enters the nucleus and binds transcription factors to activate gene transcription (62). The WNT/β-catenin pathway has recently been shown to play a pivotal role in the melanoma TME, leading to the establishment of “cold tumors” by its effect on limiting DC infiltration and blunting both T-cell priming and effector T-cell recruitment (63). The relevance of this pathway in the TME of GBM is currently being investigated, but one of the well-known effects of WNT/β-catenin pathway is its role in the maintenance of the BBB integrity (64). In addition, it is known to further promote the secretion of IL6 and TNFα in infiltrating TAMs. As this pathway is further elucidated, it remains critical to determine if the observations seen in melanoma translates into GBM tumors.

The well-known tumor suppressor p53 is largely mutated in GBM tumors (3). The loss-of-function mutation is believed to occur early during tumorigenesis. Loss-of-function mutations in the p53 gene result in increased proliferation, reduced cell death, and genetic instability (65). Recent studies have also shown that the loss of p53 function reduces the induction of inflammatory cytokines capable of alerting the immune system and activating NK cells (66). Reintroduction of p53 in GBM tumors leads to the induction of TNFα, and resulted in increased TIL populations (67). It is expected that p53 could sense DNA damage in tumors and induce PTEN expression in those tumors which PTEN is functional, and reduce PI3K activity as well (68), resulting in reduced immunosuppression. More studies are required to elucidate the detailed functional role of p53 in GBM in regulating the “cold tumor” phenotype.

Mutations can occur in either IDH1 or IDH2 and because these genes are very similar they will collectively be referred to as IDH throughout. IDH is a ubiquitous enzyme responsible for catalyzing the conversion of Isocitrate into 2-oxoglutarate as part of the tricarboxylic acid cycle (TCA) cycle (69). IDH has been shown to be mutated in ∼10% of primary GBM, but in as many as 90% secondary GBMs (70). It has recently been demonstrated that the IDH mutation IDH R132H results in the production of the oncometabolite (D) 2-hydroxygluterate (2HG), which directly modulates the TME reducing chemo-attractants responsible for the recruitment of TILs and limiting the function of NK cells (71–73). More studies are in necessary to address other potential mechanisms by which IDH R132H and its oncometabolite, 2HG, modulate the TME in GBM.

One pathway which etiology still needs to be defined in GBM is the indolamine 2,3-dioxygenase (IDO). This pathway has recently been demonstrated to be active in GBM tumors as well as in surrounding stroma (74). IDO is generally activated upon IFNγ signaling or B7 signaling in DCs. However, it remains unknown how this is activated in GBM tumor cells (75). Recent studies highlight the possibility of WNT/β-catenin signaling in regulating IDO expression (76). Its expression and enzymatic activity leads to reduce tryptophan levels, an essential amino acid, and increases the synthesis of toxic kynurenine metabolites (77). Decreased levels of tryptophan reduce T-cell activity, and increased kynurenine metabolites lead to T-cell death (77). IDO is also responsible for T_Reg cells recruitment and activation within the GBM tumors (78). T_Reg cells are then capable of shaping the surrounding TME by secreting IL10, and TGFβ, therefore leading to M2 polarization and immunosuppression.

Mutations intrinsic to the tumor, in the previously mentioned pathways, have direct consequences on shaping the TME, and serve to promote the “cold” tumor phenotype; however, the resultant environment is also caused by the interplay between the tumor itself and surrounding stroma. Production of cytokines like IL6 and stabilization of cytokines such as TGFβ signal the surrounding stroma to activated molecules such as STAT-3. This results in the release of chemokines responsible for monocyte and macrophage recruitment. These cells will in turn interact with the TGFβ, as well as respond to the tumor-extrinsic factors such as hypoxia, acidosis, and necrosis to establish the TME characteristic of GBM.

Immunotherapeutic interventions aimed at lowering the barrier of immunosuppression and converting the TME from a “cold” to a “hot” phenotype will be instrumental for improving success of frontline immunotherapies in GBM. Having large amounts of functional CTLs in the TME is known to correlate with treatment outcomes and survival across solid tumors, and CTL function is predicated on sufficiently low immunosuppressive factors in the TME. Here we discuss how different immune interventions may serve to reduce these immunosuppressive factors.

TAMs serve as a genetically stable target for therapeutic intervention. Although they take on a protumor phenotype in the GBM TME, efforts to re-educate TAMs to a more inflammatory state has becoming a topic of great interest (79). Microglia are dependent on colony stimulating factor-1 (CSF-1) and pharmacologic interventions aimed at inhibiting the CSF-1 receptor (CSF-1R) have shown the ability to either deplete or re-program TAMs in some preclinical murine models (80–82). While programming TAMs to an antitumor phenotype is expected to be more effective than depletion, the plasticity of TAMs likely leads to a less durable antitumor phenotype. Finding ways to irreversibly polarize TAMs to an antitumor phenotype without reducing their migration into the TME could greatly reduce the degree of immunosuppression, and potentially open the door to antigen-specific therapies becoming more effective.

Small-molecule inhibitors targeting the PI3K, Ras–MAPK, signaling pathways and the IDH and IDO enzymatic activity have recently been developed and are under current evaluation in combination with active immunotherapy or checkpoint blockade. The original purpose of these agents were to directly limit the protumorigenic activity of these pathways, however, most of these inhibitors failed when utilized as a single agent due to the large degree of tumor genetic heterogeneity present within and across patients with GBM (83, 84). Nonetheless, despite their failure to mediate antitumor efficacy as single agents, the combination of these inhibitors is expected to alter the TME in ways that might allow the immune system to infiltrate mediate tumor rejection, when combined with other immune modulatory agents such as vaccines, adoptive therapy or checkpoint blockade.

Immune checkpoint blockade therapy involves the targeting of mechanisms that exist to maintain self-tolerance via inhibition of T-cell activation. In the GBM TME exhausted TILs are known to have elevated expression of cytotoxic T-lymphocyte–associated antigen-4 (CTLA-4), programmed cell death receptor-1 (PD-1), and T-cell inhibitory receptor 3 (TIM-3; refs. 85–87). mAbs blocking these receptors have shown promise and even gained approval in a variety of cancers (88). Unfortunately, use of these antibodies as monotherapies or combination therapy failed in a phase III clinical trial for use in GBM (89). Despite the inability of immune checkpoint blockade to restore tumoricidal function of T cells in GBM, it remains promising as an addition to other TME-directed treatments as a means of further reducing immunosuppression.

Successful immunotherapy requires the presence of functional T cells within tumors, but as we previously mentioned the immunosuppressive nature of the GBM TME limits T-cell infiltration, and TILs capable of infiltrating the tumor are rendered dysfunctional (29). Thus, methods which improve T-cell infiltration like in situ cytokine infusion aimed to increase migration to the TME, or T-cell modification which enhances migratory capacity into the TME, are highly desirable. Although several strategies to modulate the GBM TME via cytokine infusion have shown to increase T-cell infiltration, function and efficacy (90–92), licensing T cells to migrate into the TME has proven challenging. Previous studies focused on enhancing the expression of the adhesion molecule very late antigen-4 (VLA-4; ref. 93), which resulted in modest enhancements in infiltration and antitumor efficacy. Therefore, GBM TME modulation through direct infusion of immunomodulatory cytokines or forced expression of the adhesion molecules VLA-4 on T cells could be beneficial for improving the selective infiltration of functional T cells within brain TME. The application of these strategies in combination with the above-mentioned strategies aimed at improving T-cell function within the GBM TME could synergize in overcoming the immunosuppressive barrier present within GBM and mediate tumor eradication.

GBM's TME is extremely immunosuppressive due to tumor-intrinsic and tumor-extrinsic components. This leads to unique challenges in treating this cancer. Many therapies that have succeeded in more immunogenic cancers have failed as a result of this immunosuppression. Because of the great degree of heterogeneity and the adaptive nature of tumors, antigen-specific therapies will likely need to be supplemented by therapies which aim to directly reduce immunosuppression via targeting the genetically stable stroma. Considering the abundance of TAMs, their genetic stability and their paramount role in the maintenance of the immunosuppressive TME within GBM, we expect that TAM-directed immunotherapeutic strategies would greatly reduce the degree of immunosuppression present in GBM. TAM-directed immunotherapeutic strategies could promote T-cell effector function and trafficking thereby shifting the GBM TME from “cold” to a “hot” phenotype. Furthermore, combination therapy including TAM-directed therapies and checkpoint inhibitors may synergize in enhancing the antitumor effect of TILs. To determine the effect of these TAM-directed immunotherapeutic strategies, clinical trials should focus on evaluating brain tumor penetration by TILs, on-target effects on TAMs and changes in immunosuppressive markers in the TME. GBM offers the possibility of evaluating these agents in the neo-adjuvant setting, which could then allow for tumor resection and careful examination of the GBM TME. Discovering how different subtypes of GBM respond best to different TAM-directed therapies could make possible the use of individualized antigen-specific treatments, which have a great degree of neo-epitope coverage.

T.F. Gajewski reports receiving commercial research grants from Merck, Bristol-Myers Squibb, Ono, Incyte, Aduro, Evelo, and Bayer, holds ownership interest (including patents) in Jounce and Evelo, and is a consultant/advisory board member for Merck, Aduro, Fog Pharma, Jounce, Adaptimmune, and FivePrime. J.H. Sampson holds ownership interest (including patents) in Annias Immunotherapeutics (which has licensed intellectual property from Duke related to the use of the pepCMV vaccine in the treatment of glioblastoma multiforme) and Istari Oncology (which has licensed intellectual property from Duke related to the use of poliovirus and D2C7 in the treatment of glioblastoma), and is an inventor on patents related to PEP-CMV DC vaccine with tetanus, as well as poliovirus vaccine and D2C7 in the treatment of glioblastoma. No potential conflicts of interest were disclosed by the other authors.

The funders had no role in the preparation of the manuscript or decision to publish.

This work was supported by grants from the NIH: R01-NS099463 (to J.H. Sampson), P50-CA190991 (to J.H. Sampson), R01-CA177476 (to J.H. Sampson), U01-NS090284 (to J.H. Sampson), and R01-NS085412 (to J.H. Sampson).