The Molecular and Microenvironmental Landscape of Glioblastomas: Implications for the Novel Treatment Choices

Glioblastoma (GBM) is the most frequent and aggressive primary central nervous system tumor. Surgery followed by radiotherapy and chemotherapy with alkylating agents constitutes standard first-line treatment of GBM. Complete resection of the GBM tumors is generally not possible given its high invasive features. Although this combination therapy can prolong survival, the prognosis is still poor due to several factors including chemoresistance. In recent years, a comprehensive characterization of the GBM-associated molecular signature has been performed. This has allowed the possibility to introduce a more personalized therapeutic approach for GBM, in which novel targeted therapies, including those employing tyrosine kinase inhibitors (TKIs), could be employed. The GBM tumor microenvironment (TME) exerts a key role in GBM tumor progression, in particular by providing an immunosuppressive state with low numbers of tumor-infiltrating lymphocytes (TILs) and other immune effector cell types that contributes to tumor proliferation and growth. The use of immune checkpoint inhibitors (ICIs) has been successfully introduced in numerous advanced cancers as well as promising results have been shown for the use of these antibodies in untreated brain metastases from melanoma and from non-small cell lung carcinoma (NSCLC). Consequently, the use of PD-1/PD-L1 inhibitors has also been proposed in several clinical trials for the treatment of GBM. In the present review, we will outline the main GBM molecular and TME aspects providing also the grounds for novel targeted therapies and immunotherapies using ICIs for GBM.


INTRODUCTION
Glioblastoma (GBM) is the most common and aggressive primary CNS tumor (Stupp et al., 2009;Louis et al., 2016;Mendes et al., 2018;Altmann et al., 2019;Ostrom et al., 2019) and it has been included in the group of diffuse astrocytic and oligodendroglial tumors by the 2016 CNS WHO (Louis et al., 2016). It is thought that genetic alterations affecting neuroglial stem or progenitors cells give origin to GBM. The incidence of this tumor seems to increase with age; in fact, 62 years is the median age at diagnosis. Males are affected by GBM tumors 1.7 fold more often than females. According to the presence of mutations in the isocitrate dehydrogenase (IDH) 1 and IDH2 genes GBM is subdivided by the WHO into two major types. More than 90% of GBM cases is constituted by GBM with wild type IDH (Louis et al., 2016). Clinically, grade IV lesions (namely primary GBM) are presented de novo by the majority of patients, while progression from a less aggressive form of WHO grade II diffuse astrocytomas and WHO grade III anaplastic astrocytomas (i.e., secondary GBM) is shown by a small fraction of patients (5-10%; Ohgaki and Kleihues, 2013;Louis et al., 2016). Primary GBM and secondary GBM differ in prognosis and age of onset. As for overall survival (OS) (Doetsch et al., 1999;Louis et al., 2016), primary GBM is typically diagnosed at older age and has a worse prognosis while secondary GBM are less common and affect people under the age of 45; also they develop into low-grade astrocytoma and are associated with better prognosis (Doetsch et al., 1999;Brennan et al., 2013;Ohgaki and Kleihues, 2013;Louis et al., 2016).
Standard of care first-line treatment is constituted by maximal surgical resection (complete resection is performed quite rarely because of the presence of diffuse infiltrations), followed by radiotherapy with concomitant and adjuvant chemotherapy such as the oral alkylating agent, temozolomide (TMZ). Upon this treatment combination GBM show a median OS of about 15 months (Canoll and Goldman, 2008;Stupp et al., 2009;Ohgaki and Kleihues, 2013;Levine et al., 2015).
The increase of patient survival is small and tumors invariably recur after TMZ (Canoll and Goldman, 2008;Stupp et al., 2009;Ohgaki and Kleihues, 2013;Levine et al., 2015). Following the first recurrence, treatment choices can be represented by further surgical resection when possible, or conventional chemotherapy, e.g., TMZ (with different dosing schedules) or nitrosoureas, or treatment with the anti-vascular endothelial growth factor (VEGF) agent, bevacizumab, or the use of the low-intensity alternating electric fields (TTFields). However, these treatments have not achieved significant improvements in survival (Canoll and Goldman, 2008;Stupp et al., 2009;Chamberlain and Johnston, 2010;Stupp et al., 2012;Ohgaki and Kleihues, 2013;Stupp and Hegi, 2013;Chamberlain, 2015). Moreover, the tyrosine kinase inhibitor (TKI) regorafenib has been introduced in the treatment of recurrent GBM .
In the present review, we will outline the principal GBM molecular and tumor microenvironment (TME) aspects providing also the grounds for novel targeted therapies and immunotherapy approaches using ICIs for the treatment of GBM affected patients.
Recent studies have defined the evolution of tumor cells in GBM cases undergoing therapy as a process of clonal replacement where a fraction of tumor cells is eliminated by the treatment while clones of resistant cells are positively selected (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). Specifically, intratumoral heterogeneity, with the presence of resistant subclones, both in low grade and high grade glioma is frequently associated with treatment failure. Although a clearly defined pattern of tumor evolution has not yet been described in GBM, TP53 gene mutations have been recently proposed as a marker of subclonal heterogeneity (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). Glioblastoma evolution is highly branched, specific alterations and evolutionary patterns frequently occurring depending on the treatment. There is no linear link between the dominant clone at diagnosis and the dominant clone at relapse (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). Remarkably, genes such as TP53, EGFR, PDGFRA are frequently subjected to a process of mutational switching where a mutated version of a gene, found at diagnosis, is replaced by another mutated version of the same gene at relapse (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). Hypermutated tumors, which are highly enriched for mutations at CpG dinucleotides, generally harbor mutations in MMR pathway genes, most commonly in MSH6. These MMR alterations have been thought to be associated with putative mutagenic mechanisms of TMZ treatment (Hunter et al., 2006;Cahill et al., 2007;Tcga, 2008;Yip et al., 2009;Brennan et al., 2013;Wang et al., 2016).
A gene expression-based molecular classification has been proposed to integrate somatic mutation and DNA copy number data (Verhaak et al., 2010;Behnan et al., 2019). According to this classification, GBM cases were subdivided in proneural, neural, classical and mesenchymal subtypes. These different subtypes have been associated with gene signatures of normal brain cell types of different neural lineages. Moreover, GBM cases included in the different subtypes have also been associated with a different pathogenesis with GBM clones developing as the result of different causes and/or from different cell type of origin. However, further studies, also investigating glioma stem cells, have been able to identify three subtypes: proneural, mesenchymal and classical subtypes (Verhaak et al., 2010;Behnan et al., 2019).
According to the first proposed classification, GBM cases belonging to the classical subtype show in about the 100% of cases the chromosome 7 amplification paired with chromosome 10 loss. This event is also very frequent in the totality of GBM cases. High-level of EGFR amplification has been observed in 97% of cases belonging to the classical subtype, whereas this alteration has been infrequently found in the other GBM subtypes. Moreover, in association with frequent EGFR alteration, a lack of TP53 mutations has been found in a subset of the classical subtype (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016). The focal 9p21.3 homozygous deletion, targeting the CDKN2A gene, has been also frequently found in the classical subtype, in the 94% of cases belonging to this subtype found associated with EGFR amplification. The homozygous 9p21.3 deletion has been also found mutually exclusive with aberrations in genes belonging to the RB1 pathway such as RB1, CDK4 and cyclin-D2 (CCDN2), thus suggesting that in the cases with focal EGFR amplification the CDKN2A deletion is the sole alteration affecting the RB1 pathway. GBM cases belonging to the classical subtype are also characterized by the high expression of genes belonging to the notch homolog 1, translocation-associated (NOTCH) pathway such as neurogenic locus notch homolog-3 (NOTCH3), jagged-1 (JAG1) and lunatic fringe (LFNG), sonic hedgehog pathway such as smoothened (SMO), growth arrestspecific protein 1 (GAS1) and zinc finger protein GLI1 (GLI1) and the neural precursor and stem cell marker nestin (NES) pathway (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016).
Glioblastoma cases belonging to the proneural subtype were found to be mainly characterized by alterations of PDGFRA and point mutations of IDH1. The focal amplification at the locus 4q12 harboring PDGFRA was associated with high levels of PDGFRA gene expression and the frequent presence of mutations in the PDGFRA gene. The great majority of IDH1 mutations has been found in GBM cases belonging to the proneural subtype. Of note, they have been found to be generally mutually exclusive to PDGFRA alterations. Loss of heterozygosity and mutations of the TP53 gene have been found to be frequent events in the proneural subtype. PIK3CA/PIK3R1 mutations have also been found in the proneural subtypes in cases without PDGFRA abnormalities. The proneural group has been found to be characterized also by the high expression of genes other than PDGFRA that characterize the oligodendrocytic development such as oligodendrocyte transcription factor (OLIG2) and homeobox protein nkx-2.2 (NKX2-2). This group has been found also characterized by the expression of proneural development genes such as SOX family genes and achaete-scute homolog 1 (ASCL1), doublecortin (DCX), delta-like 3 (DLL3), transcription factor 4 (TCF4; Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016).
Glioblastoma cases belonging to the neural subtype were characterized by the expression of genes well-known as neuron markers such as GABRA1, neurofilament light chain (NEFL), synaptotagmin-1 (SYT1) and solute carrier family 12 member 5 (SLC12A5). GBM cases belonging to the neural subtype show an enrichment in genes involved in neuron protection and in axon and synaptic transmission (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016).
Glioblastoma cases belonging to the mesenchymal subtype are frequently characterized by the presence of focal hemizygous deletions at 17q11.2 region encompassing the NF1 gene. This has been frequently associated with low NF1 expression levels. Moreover, mutations at the NF1 gene have been found in GBM cases belonging to the mesenchymal subgroup. Concomitant PTEN mutations have also been found in mesenchymal subgroup cases carrying NF1 mutations (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016). GBM cases belonging to the mesenchymal subtype are also characterized by the expression of mesenchymal markers such as chitinase-3-like protein 1 (CHI3L1) and MET. It has been thought that the higher activity of mesenchymal and astrocytic markers such as CD44 and MERKT is linked to an epithelial-to-mesenchymal transition proper of dedifferentiated and transdifferentiated tumors. Finally, GBM cases belonging to the mesenchymal subtype are also characterized by the high expression of genes belonging to the TNF superfamily pathway and NF-kB pathway such as tumor necrosis factor receptor type 1-associated death domain (TRADD), RELB and TNF receptor superfamily member 1A (TNFRSF1A) (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016).
Several clinical features have been associated with the four subtypes. In particular, an association between proneural subtype and age as well as between this subtype and a trend for a longer survival. However, GBM belonging to the proneural subtype have not shown a survival advantage from aggressive treatment protocols. On the other hand, a clear treatment effect has been observed among GBM cases belonging to the classical and mesenchymal subtypes (Tcga, 2008;Verhaak et al., 2010;Brennan et al., 2013;Wang et al., 2016).
The proneural-to-mesenchymal transition upon tumor recurrence has been proposed as a mechanism of treatment resistance for GBM to radiotherapy and/or chemotherapy. GBM patients belonging to the mesenchymal subtype have been associated with survival shorter than the other subtypes, particularly when cases with low transcriptional heterogeneity are considered. Although in the context of poor prognosis patients, GBM cases belonging to the mesenchymal subtype have been found to show favorable response to immunotherapy and intensive radiotherapy and chemotherapy (Verhaak et al., 2010;Behnan et al., 2019).
Long non-coding RNAs (LncRNAs) are RNA transcripts longer than 200 base pairs which do not code for proteins. Although the human genome harbors more than 50,000 LncRNA genes, they are still poor characterized. However, LncRNAs have been found to play key roles in various cell activities related to regulation of gene expression, protein synthesis, stemness, immunity (Schlackow et al., 2017). Moreover, LncRNAs have been found to exert relevant roles in pathogenesis and progression of various cancers including GBM. In particular, a large number of LncRNAs has been found associated with deregulated gene expression and imbalanced biological processes in GBM (Zeng et al., 2018). In this context, the expression of the LncRNA P73 antisense RNA 1T (TP73-AS1) has been associated with poor outcome in GBM patients. GBM patients belonging to less aggressive subgroups have been found to be characterized by hypermethylation and low expression of TP73-AS1. Moreover, it has been found that TP73-AS1 downregulation is associated with the loss of aldehyde dehydrogenase 1 family member A1 (ALDH1A1) expression and the re-sensitivity of the GBM stem cell (GSC) population to TMZ treatment (Mazor et al., 2019). Expression of the LncRNA HOX transcript antisense intergenic RNA (HOTAIR) in GBM has been demonstrated to be significantly higher than in normal tissues and low grade gliomas. Moreover, HOTAIR has been demonstrated to be an independent prognostic factor in GBM associated with proliferation and tumorigenic potential of GBM cells (Zhang et al., 2015;Zhang et al., 2018). LncRNA colorectal neoplasia differentially expressed (CRNDE) has been found highly expressed in GBM and other brain cancers such as astrocytomas. It has also been explained that its overexpression is associated with promotion of tumor cell growth and migration (Ellis et al., 2014;Wang et al., 2016). LncRNA nuclear enriched abundant transcript 1 (NEAT1) has been shown to be a key regulator of nuclear domains implicated in mRNA nuclear retention and splicing. NEAT1 has been found upregulated in human GBM tissues and GBM cell line models and a high NEAT1 expression has been associated with larger tumor size, higher WHO grade, higher recurrence rate and unfavorable overall survival (He et al., 2016). The LncRNA X-inactive specific transcript (XIST) has been found highly expressed in glioma tissues and GSCc. The knockdown of XIST has been shown to suppress proliferation, migration, invasion and tumorigenic potential of GSCs by upregulating miR152 (Yao et al., 2015). The LncRNA SOX2 overlapping transcript (SOX2OT) is characterized by the fact that its transcription genomic region includes the SOX2 gene; a SOX2OT involvement in the transcriptional regulation of SOX2 has also been observed. SOX2OT has been shown to be involved in the proliferation, migration, invasion of GSCs . The LncRNA H19 has been shown to be upregulated in glioma tissues and associated with poor outcome. Moreover, invasion, angiogenesis, stemness and tumorigenicity of GBM cells have been found enhanced when H19 is overexpressed . The LncRNA LOC441204 has been found highly expressed in glioma tumor specimens and cell lines. Tumor cell proliferation has been found suppressed by knockdown of LOC441204 in glioma. On the other hand, LOC441204-induced tumor cell growth has been shown to be modulated by the stabilization of the βcatenin pathway (Lin et al., 2017). Regarding the role of other LncRNAs, evidence has also been reported about the fact that the high expression of other LncRNAs such as maternally expressed gene 3 (MEG3), metastasis associated lung adenocarcinoma (MALAT1), cancer susceptibility candidate 2 (CASC2), taurineupregulated gene 1 (TUG1), DBH antisense RNA 1 (DBH-AS1), AC005035.1, AC010336.2, AC108134.2, AC116351.2, Clorf132, C10orf91, LINC00475, MIR210HG could be associated with poor outcome in GBM cases (Zeng et al., 2018).

ROLE OF THE GBM TUMOR MICROENVIRONMENT
The brain is distinguished from the other organs by the presence of the blood-brain-barrier (BBB). The BBB provides a selective barrier between the systemic circulation and the brain, thus representing a limit for the delivery of many therapeutic agents Frontiers in Neuroscience | www.frontiersin.org FIGURE 1 | Immune checkpoint inhibitors (ICIs) targets in lymphnodes and in glioblastoma (GBM) tumor microenvironment (TME). Cytotoxic T lymphocyte protein 4 (CTLA-4) blockade mainly acts by targeting Tregs expressing CTLA-4 in lymphnodes. In the context of the GBM TME, programmed cell death protein receptor 1 (PD-1)/programmed death ligand (PD-L1) blockade can overcome the T cell exhaustion and reverse immunosuppression. Miura et al., 2013). However, a loss of BBB integrity could be displayed in the presence of cancer, in particular during the cancer progression. This seems to be the reason why several agents, including ICIs, that are known to be not capable of penetrating the BBB, have however shown in some extent a clinical efficacy (de Vries et al., 2006;van Tellingen et al., 2015). Specialized endothelial cells, pericytes, and astrocytic foot processes, dictating junctional integrity, are the elements that FIGURE 2 | Targeted therapies in glioblastoma (GBM). The introduction of novel targeted therapies has been allowed by the comprehensive characterization of the molecular landscape of somatic genomic alterations identifying a series of mutated genes and abnormal rearrangements potentially utilizable as therapeutic targets.
Frontiers in Neuroscience | www.frontiersin.org constitute the BBB. Moreover, BBB integrity can be also regulated by microglia, being these cells capable of repairing the BBB in a purinergic receptor P2RY12-dependent manner in case of injury (de Vries et al., 2006;van Tellingen et al., 2015).
The complex crosstalk of TME components is involved in the regulation of tumor progression Joyce, 2013, 2017;Gritsenko et al., 2012;Nakasone et al., 2012;Joyce, 2013, 2017). The composition of ECM of normal brain is distinctive, with specific tissue-resident cell types such as neurons, astrocytes and microglia. Moreover, the BBB physically protects the ECM from inflammation (Novak and Kaye, 2000;Mahesparan et al., 2003). The most common component of the brain ECM is hyaluronic acid which is localized in the intraparenchymal region (Kim et al., 2018). The haptotactic cues from the vascular basement membrane, the enrichment of vascular derived chemoctatic cues, as well as interconnected axon tracts can determine the therapeutic resistance of GBM cells in the perivascular space further providing haptotactic cues for cellular invasion (Giese and Westphal, 1996;Nimsky et al., 2005;Gritsenko et al., 2012).
A diffuse invasion pattern characterizes GBM (Young et al., 2015). Healthy tissue beyond the tumor margin is infiltrated by the tumor cells, generally enriched in the GSC stem cell fraction, that either migrate individually or collectively practically impeding complete surgical resection (Eyler and Rich, 2008;Sherriff et al., 2013). On the other hand, GBM tumors rarely intravasate and metastasize from the brain to distant organs Joyce, 2013, 2017).
Glioblastoma frequently develop in a hypoxic microenvironment which can modify the metabolic pathways of GBM cells. The brain has a high metabolism level in which the glucose is the major energy substrate and lactate, ketone bodies, fatty acids and aminoacids can also be employed. The metabolic homeostasis of the brain is maintained by the interaction among its various constituent cells such as astrocytes, neurons and microglia (Gritsenko et al., 2012;Nakasone et al., 2012;Joyce, 2013, 2017). This equilibrium can be altered by genomic aberrations and biochemical variations in GBM cells that often metabolize glucose into lactate even when oxygen is present in a process called Warburg effect. GBM cells can also increase intracellular lipid, aminoacid and nucleotide levels. These metabolic adaptations can favor GBM tumor growth (Gritsenko et al., 2012;Nakasone et al., 2012;Joyce, 2013, 2017).
Hypervascularity is a characteristic of GBM tumors with an increment in angiogenesis compared to healthy brain tissue. This tumor-associated vasculature is not completely formed, with leaky vessels, and associated with an increase in interstitial fluid pressure. A necrotic core softer than the surrounding tissue characterizes the TME of GBM (Brat and Van Meir, 2004;Persano et al., 2011;Hambardzumyan and Bergers, 2015;Chen and Hambardzumyan, 2018). High density regions called pseudopalisades are formed when cells migrate away from the hypoxic regions. Increased matrix production with respect to both necrotic regions and healthy tissues characterizes these regions (Brat and Van Meir, 2004;Persano et al., 2011;Hambardzumyan and Bergers, 2015;Chen and Hambardzumyan, 2018). GBM cells are capable of rapidly invading vasculature (Akiyama et al., 2001;Ponta et al., 2003;Zimmermann and Dours-Zimmermann, 2008;Dicker et al., 2014;Schiffer et al., 2018).
Circulatory and immune systems are connected by the lymphatic system that is involved, together with blood vessels, in the exchange of various elements including fluid, waste, debris as well as immune cells (Engelhardt et al., 2017). Together with the absence of a classic lymphatic drainage system, the CNS exhibits several other peculiar features, such as the presence of tight junctions in the BBB, as well as the limited rejection of foreign tissues within the CNS (Louveau et al., 2015;Schiffer et al., 2018).
There are functional lymphatic vessels in the CNS with the presence of different types of antigen-presenting cells (APCs), including microglia, macrophages, astrocytes and canonical APC such as dendritic cells (DCs; Figure 1; Louveau et al., 2015;Schiffer et al., 2018). In the brain, microglia are the predominant APCs whereas DCs carry out a less relevant role (Lowe et al., 1989;Ulvestad et al., 1994;Weiss et al., 2009;Goldmann et al., 2016).
Activated T cells can invade the CNS. These activated T cells can cover these compartments in an unrestricted manner. On the other hand, antigens can be presented locally or in the draining cervical lymph nodes. Moreover, the BBB can be compromised, thus allowing a relevant infiltration of multiple immune cell types from the peripheral circulation (Weiss et al., 2009). However, GBM tumors present low numbers of tumorinfiltrating lymphocytes (TILs) and other immune effector cell types compared to other tumor types (Schiffer et al., 2018). The interaction of TILs with the TME can cause their re-education. In particular, the local TME can alter T cell effector function in the process related to anti-tumor immunity even in the CNS, where T cell-mediated inflammatory responses are considered poor under normal physiological contexts. The number of antigen-specific TILs can remain relatively low besides frequently displaying an exhausted phenotype. The peculiar immune environment of the brain can be responsible for this reduced quantity and limited activity of T cells in GBM. In particular, there is a specific need of avoiding unrestrained inflammation in the brain given its solid enclosure and the potential for damage from increased intracranial pressure Joyce, 2013, 2017;Gajewski et al., 2017;Keskin et al., 2019). This need is not present with the same extent in peripheral organs. In fact, this environment in which both inflammatory and adaptive immune responses are tightly regulated is specific of the brain; besides there is a variety of immunosuppressive mechanisms at both the molecular and cellular levels (Perng and Lim, 2015). In particular, stromal cells of the brain produce high levels of the classic immunosuppressive cytokines transforming growth factor β (TGFβ), interleukin-10 (IL-10) in response to inflammatory stimuli, including those derived from GBM tumors, in order to maintain homeostasis (Vitkovic et al., 2001;Gong et al., 2012). Furthermore, the accumulation of regulatory T cells (Tregs) is stimulated by IDO which can suppress T cell activity by depleting tryptophan from the microenvironment. Microglia and tumor-infiltrating myeloid cells can also inhibit T cell proliferation and function through the production of high levels of arginase that causes the       Table 1 are the ongoing clinical trials present on ClinicalTrials.gov searching the keywords "glioblastoma multiforme" and "kinase inhibitor" The research has been done adding the following filters: "Completed"; "The research has been performed on October 21st, 2020." depletion of tissue arginine levels (Uyttenhove et al., 2003;Fecci et al., 2006a,b;Wainwright et al., 2012). Immune checkpoints exert a key role in central and peripheral tolerance by counteracting activating signaling (Xu et al., 2018). Under physiological conditions, immune checkpoint molecules represent a negative feedback to regulate inflammatory responses following T cell activation (Krummel and Allison, 1996;Chambers et al., 2001;Collins et al., 2002;Stone et al., 2009;Inarrairaegui et al., 2018). A mechanism used by tumors, including GBM, to inhibit and escape the anti-tumor immune response is represented by the expression of checkpoint molecules, such as cytotoxic T-lymphocyte antigen 4 (CTLA-4) and PD1 (Stone et al., 2009;Francisco et al., 2010;Cheng et al., 2013;Bhandaru and Rotte, 2017;Hui et al., 2017;Wei et al., 2017Wei et al., , 2018Wei et al., , 2019aRotte et al., 2018;Kalbasi and Ribas, 2020;Sharma and Allison, 2020).

SMALL MOLECULES FOR TARGETED THERAPIES IN GBM
The progresses in the molecular classification of GBM have allowed the identification of dysregulated pathways that could represent potential targets for new treatment strategies (Figure 2).
Glioblastoma is a vascularized tumor which is histologically characterized by the expression of VEGF and other proangiogenic cytokines involved in the stimulation of endothelial cell proliferation, migration and survival (Schiffer et al., 2018). In patients with a relapsed GBM, the TKI regorafenib has received approval in the GBM treatment .
Other TKIs targeting VEGF family components have been proposed for the treatment of GBM besides regorafenib. Of note, vascular normalization has been proposed as an alternative strategy for the employment of antiangiogenic therapies in which the objective is to modulate the tumor vasculature in order to reduce hypoxia, and to support physiological angiogenesis. This process could ultimately improve perfusion and drug delivery. In this context, promising results in reducing angiogenesis and normalizing vascularization have been shown by cediranib and sutinib (Batchelor et al., 2013;Grisanti et al., 2019).
The PI3K/mammalian target of rapamycin (mTOR) pathway is a targetable pathway in GBM. In this context, the mTOR inhibitor temsirolimus did not show a treatment efficacy as single agent in recurrent GBM (Chang et al., 2005). Similarly, the pan-PI3K inhibitor buparlisib did not demonstrate a treatment efficacy (Wen et al., 2019). Also treatment combinations of mTOR pathway inhibitors with radiotherapy and TMZ or in combination with radiotherapy only did not show efficacy Wick et al., 2016).
Targeting MDM2 and mouse double minute 4 homolog (MDM4) activity has been suggested for GBM cases carrying MDM2 or MDM4 gene amplification (Wick et al., 2019).
The use of TKIs targeting EGFR as single agents did not demonstrate significant activity for GBM treatment (Lassman et al., 2005;Hegi et al., 2011). It has not yet been agreed on the potential use of MET as target for GBM treatment. The use of the TKIs crizotinib and cabozantinib in recurrent GBM has achieved modest efficacy after several attempts (International Cancer Genome Consortium PedBrain Tumor Project, 2016;Wen et al., 2018). Tests have been carried out for larotrectinib and entrectinib in neutrophic tyrosine receptor kinase (NTRK) fusion-positive GBM without any confirmation on their efficacy (Ferguson et al., 2018). Notwithstanding the frequent expression in GBM of fibroblast growth factor receptors (FGFRs), a relevance as potential therapy target seems to be restricted to GBM exhibiting FGFR-transforming acidic coiledcoil containing protein TACC fusions (Singh et al., 2012), as shown by using the pan-FGFR kinase inhibitor erdafitinib (Di Stefano et al., 2015). A modest treatment efficacy has been obtained for the possible targeting of BRAFV600E mutation in GBM (Kaley et al., 2018). Finally, eribulin has been proposed to inhibit TERT activity in GBM (Takahashi et al., 2019). A list of the current clinical trials employing TKIs for GBM treatment is reported in Table 1.

USE OF ICIs FOR GBM TREATMENT
Following the results of ICIs use in other cancers, the use of PD-1/PD-L1 inhibitors has been proposed for GBM ( Table 2). Clinical trial results have shown that GBM patients with unresectable tumors do not benefit from monotherapy with nivolumab in terms of survival improvement when compared    Table are the ongoing clinical trials present on ClinicalTrials.gov searching the keywords "glioblastoma multiforme" and "Checkpoint." The research has been done adding the following filters: "Not yet recruiting"; "Recruiting"; "Enrolling by invitation"; "Active, not recruiting"; "Suspended"; "Terminate"; "Completed"; "Withdrawn"; "Unknown status"; "The research has been performed on October 21st, 2020." to bevacizumab (Reiss et al., 2017). Moreover, pembrolizumab showed limited activity for GBM (Reardon et al., 2014(Reardon et al., , 2016Schwartz et al., 2016;Reardon et al., 2017;Reiss et al., 2017;Wen et al., 2018). Recent tests have been carried out involving patients with newly diagnosed or relapsed GBMs for the use of ICIs (e.g., nivolumab or pembrolizumab) in neoadjuvant and/or adjuvant administration, although no straightforward results have been obtained (Cloughesy et al., 2019;Schalper et al., 2019).
Glioblastoma tumors of cases non-responsive to ICIs have shown an enrichment in mutations of the PTEN gene (Zhao et al., 2019) that has been associated with an immunosuppressive TME characterized by the presence of GBM cells expressing CD44. PTEN mutant tumors were characterized by highly clustering tumor cells with a lack of T cell infiltration (Peng et al., 2016;George et al., 2017). Furthermore, the poor responsiveness to ICIs of GBM cases carrying PTEN mutations has been related to a low PD-L1 expression for the involvement of the PI3K-mTOR pathway that is downstream to PTEN (Lastwika et al., 2016).
Responsiveness to ICI was associated with the presence of mutations of BRAF/protein tyrosine phosphatase non-receptor type 11 (PTPN11). In this subset of BRAF/PTPN11 GBM patients, treatment combinations of ICIs and MAPK inhibitors could be introduced (Toso et al., 2014;Toso et al., 2014;Ebert et al., 2016;Wang et al., 2016).
The heterogeneous response rate to ICIs highlights the need of identifying the subgroups of patients who could benefit the most from the use of this immunotherapy treatment. PD-L1 expression was the first marker evaluated as predictor of a clinical response to ICIs (Ansell et al., 2015). PD-L1 expression in gliomas was associated with IDH status (Berghoff and Preusser, 2016;Garber et al., 2016;Berghoff et al., 2017). Importantly, mesenchymal GBM has been found having high levels of PD-L1 expression that may suggest that the expression of immune checkpoint proteins and aggressiveness of GBM tumors may be correlated (Garber et al., 2016). More recently, the tumor mutational burden has been proposed as a predictive marker of responsiveness to ICIs. However, it has not generally been demonstrated that the tumor mutational burden is capable of sufficiently predicting long term clinical benefits (Champiat et al., 2014;Rizvi et al., 2015;Schumacher et al., 2015;Le et al., 2017). Moreover, recent studies have shown that higher somatic mutation and neoepitope loads have not been found in GBM cases responsive to ICIs (Zhao et al., 2019). The infiltration of mutation-reactive class I and class II T cells into the tumor seems not to be precluded by a low mutational load in GBM (Cloughesy et al., 2019;Schalper et al., 2019;Zhao et al., 2019). The presence of alterations in the MMR genes is another proposed biomarker (Cloughesy et al., 2019;Schalper et al., 2019). The expression of MHC class I molecules has been associated to responsiveness to ICIs since it is involved in the presentation of antigens and characterized by highly heterogeneous expression levels in GBM (Indraccolo et al., 2019).
Systemic delivery uses existing vessels to deliver anti-tumor drugs to the tumor. To overcome the impediment of the   Table are the ongoing clinical trials present on ClinicalTrials.gov searching the keywords "glioblastoma multiforme" and "CAR-T." The research has been done adding the following filters: "Not yet recruiting"; "Recruiting"; "Enrolling by invitation"; "Active, not recruiting"; "Suspended"; "Terminate"; "Completed"; "Withdrawn"; "Unknown status"; "The research has been performed on October 21st, 2020." FIGURE 3 | Modified chimeric antigen receptor-T (CAR-T) cells to ameliorate treatment efficacy by counteracting the immunosuppressive glioblastoma (GBM) tumor microenvironment (TME). The co-expression of an activating chimeric switch receptor (CSR), that combines the extracellular ligand-binding domain of an inhibitory receptor (PD-1 or CTLA-4) fused through a transmembrane domain with the cytoplasmic co-stimulatory signaling domain of CD28, could improve CAR-T cell efficacy in GBM.
BBB several strategies have been proposed including chemical modification of the drugs, high dose chemotherapy capable of inducing a transient BBB disruption, nanoparticle-based drug delivery and peptide-based drug delivery. Nevertheless, no straightforward results have still been reached (Siegal, 2013). Glioblastoma stem cell cell population has been shown to induce a certain degree of radio-chemoresistance given their high expression of anti-apoptotic proteins, ATP-binding cassette pumps, their increased capability of DNA damage repair, as well as their high capacity of migration and invasion (Bao et al., 2006;Calabrese et al., 2007;Eyler and Rich, 2008;Diehn et al., 2009;Pietras et al., 2014). GSCs have been found capable of secreting angiogenic factors which in turn are responsible for an enhancement in the formation of tumor blood vessels, this has been frequently associated with high tumor aggressiveness. Moreover, the TME cell components can promote GSC survival by VEGF secretion (Wada et al., 2006). The interaction of TME with GSCs can facilitate tumor progression and consequently therapeutic resistance (Chen and Liu, 2012;Miura et al., 2013).
Over the past 10 years, the knowledge regarding genomic features of GBM has been greatly increased by comprehensive multiplatform genome-wide analyses. As a result of these analyses, it has emerged that GBM comprises a group of highly heterogeneous tumor types, each with peculiar molecular/genetic features (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016).
In GBM the phase I/II and III trials investigating the use of therapies molecularly targeting oncogenic alterations did not generally show straightforward results and, consequently, their clinical utilization is still limited. However, although limited activity or no therapeutic efficacy has so far been produced by the use of TKIs, improvement in understanding the mechanisms of action of these compounds could help to determine how to better incorporate their use in the existing treatment modalities. Redundancies are frequently present in the molecular pathways that can be targeted which makes the inhibition of any pathway largely ineffective (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). The failure of targeted therapies can also be ascribed to another possible reason such as the fact that several genomic alterations are important only for the initial stages of tumor progression whereas other molecular mechanisms outweigh their role in the later stages. On the other hand, several genomic alterations in GBM can interfere with GBM cell metabolism. In particular, alterations in the growth factor signaling pathways that can control metabolic flux have been found in high frequency as well as recurrent mutations in IDH1 and IDH2 genes, whose encoded proteins are part of the tricarboxylic acid (TCA) cycle. Alterations of the cellular metabolism, which is controlled also by the biochemical microenvironment, could contribute to the failure of the proposed targeted therapies (Tcga, 2008;Brennan et al., 2013;Wang et al., 2016). A better understanding of the interactions constituting this interplay between altered genome and biochemical microenvironment could contribute to finding more effective treatment strategies in the reverting of altered cellular metabolism of GBM cells.
The treatment of different cancers has markedly been revolutionized by immunotherapy. Nevertheless, the data obtained so far concerning the use of ICIs for the treatment of GBM patients seem to be not sufficient to propose this type of immunotherapy as a standard treatment for GBM (Reardon et al., 2014(Reardon et al., , 2016Motzer et al., 2015;Goldberg et al., 2016;Reck et al., 2016;Schwartz et al., 2016;Reardon et al., 2017;Reiss et al., 2017;Wen et al., 2018;Cloughesy et al., 2019;Schalper et al., 2019).
Another immunotherapy approach that can be used also in combinations with ICIs is the chimeric antigen receptor-T (CAR-T) cell therapy targeting specific tumor associated antigens. The introduction of CAR-T cell therapy approaches also in solid tumors including GBM has been favored by the success of this therapy in hematological malignancies (Neelapu et al., 2017;Maude et al., 2018). Concerning GBM treatment, several clinical trials have been proposed showing that there are still substantial obstacles including TME immune suppression ( Table 3; Morgan et al., 2010;Brown et al., 2015Brown et al., , 2016Zah et al., 2016;Ahmed et al., 2017;Walseng et al., 2017;Richman et al., 2018). To increase CAR-T treatment efficacy several CAR-T modifications have been proposed such as the knocking out of genes encoding T cell inhibitory receptors or signaling molecules (e.g., PD-1 or CTLA-4) or the co-expression of activating chimeric switch receptor (CSR; Figure 3; Prosser et al., 2012;Shin et al., 2012;Ankri et al., 2013;Kobold et al., 2015;Liu et al., 2016).
Understanding the molecular and immunological complexity of GBM more and more could provide the grounds for the introduction of other immunotherapeutic approaches such as the use of CAR-T cell therapy, in combination with ICIs or TKIs, in the treatment paradigm of GBM.

AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.