Many frequently mutated tumor suppressor genes and oncogenes in HGA are associated with recurrent chromosomal aberrations. Notable examples include loss of heterozygosity (LOH) and inactivating point mutations of the tumor suppressor genes TP53 [5, 6] and PTEN [7, 8], homozygous deletions of CDKN2A [9-11], and amplification with or without activating mutations of the oncogene EGFR [12, 13]. More recently, investigators used genome-wide approaches to catalog the genetic changes in HGA [14, 15]. These groundbreaking studies have led to the current concept that most if not all HGAs are characterized by dysregulation of three core pathways: the receptor tyrosine kinase (RTK)/phosphatidylinositol 3'-kinase (PI3K)/AKT axis, p53 signaling and RB-mediated control of cell cycle progression (Figure ​(Figure1).1). These studies also validated the importance of NF1 inactivation in non-syndromic HGA and identified novel players in gliomagenesis, such as PIK3R1 and IDH1 whose mechanism of action is the subject of intense investigation [16].

The RTK/PI3K/AKT pathway regulates many aspects of cellular physiology including proliferation, protein translation, cell size, cell survival, cell migration and motility, all of which have been implicated in the malignant phenotype (Figure ​(Figure1)1) [17]. The precise outcome of pathway activation is strictly context dependent and it is not entirely clear which of these effects are most important during gliomagenesis. This pathway is targeted for mutational activation at multiple nodes [14, 15]. Amplifications and activating mutations of the RTK genes EGFR, PDGFRA, ERBB2 and MET are seen in 59% of HGAs. The EGFR gene is by far the most commonly activated oncogene in HGA. Another commonly targeted node in this pathway is the production of the secondary messenger molecule phosphatidylinositol (3,4,5)-trisphosphate (PIP3). Three critical genes regulating this process are targeted for mutation: PIK3CA and PIK3R1, which encode the catalytic and regulatory subunits of Class 1A PI3K respectively, and PTEN, the gene encoding the lipid phosphatase that opposes PI3K activity. Together, these genes are mutated in 49% of HGAs. Mutations and homozygous deletions of the NF1 gene are present in 18% of tumors [14, 15]. NF1 contains a GTP-ase activating protein (GAP) domain and is thought to exert tumor suppressive activity primarily by inactivation of RAS proteins. Interestingly, neither study found significant activating mutations of RAS or RAF family members, which are commonly targeted oncogenes in tumors driven by RAS/MAPK signaling such as pilocytic astrocytoma, lung and colon carcinomas and melanoma [4, 18-20]. This suggests that the primary outcome of NF1 inactivation may be to boost PI3K/AKT signaling via crosstalk between pathways, which has been demonstrated in glioma GEMMs (see below). Finally, members of the AKT family are infrequently targeted for gene amplifications. Notably, there is an absence of mutations in downstream effectors of AKT suggesting that multiple effectors of the pathway may play critical roles in gliomagenesis. Alternatively, such mutations may have a net deleterious effect on tumor growth due to the existence of feedback inhibition circuits. These are all potentially critical considerations when selecting appropriate inhibitors of the RTK/PI3K/AKT pathway for HGA therapy.

TP53 is the most commonly mutated tumor suppressor gene in human cancer. The P53 pathway regulates the critical checkpoint that detects oncogenic stress and DNA damage, which if unresolved leads to cellular senescence or apoptosis (Figure ​(Figure1).1). In HGA, the TP53 gene along with genes encoding regulators of p53 stability (MDM2, MDM4 and CDKN2A which encodes p19^ARF) are targeted in at least 87% of cases [14]. Therapeutic strategies aimed at stabilizing p53 via inhibition of MDM2 and/or MDM4 are predicted to be effective in tumors with intact p53 function, therefore characterization of specific pathway mutations is potentially important.

The RB family, and in particular phosphorylation of its primary member RB1, regulates the cell cycle checkpoint at the G₁/S boundary (Figure ​(Figure1).1). The INK4 family of proteins negatively regulates the phosphorylation of RB by suppressing a complex containing D-type Cyclins and the Cyclin-dependent kinases CDK4 or CDK6. While RB1 itself is deleted or mutated in 11% of HGAs, its upstream regulation is more frequently targeted with CDKN2A and CDKN2B, which encode INK4A and INK4B respectively, being homozygously deleted in 47% of tumors [14]. CDK inhibitors under development would not be expected to have an effect on tumors in which RB1 or downstream effectors are mutated.

Using a different approach to target tumor suppressor genes for deletion in post-natal neural progenitor cells resulted in divergent results. Cre-expressing adenovirus was directly injected into the lateral ventricle of mice carrying various combinations of floxed tumor suppressor gene alleles [43]. Only the combination of Trp53 and Pten co-deletion resulted in the development of HGA in mouse brains whereas combined deletion of Trp53 and Rb1 or all three together generated primitive neuro-ectodermal tumors (PNETs). Interestingly, overexpression of Cdk4 was present in the Trp53; Pten deleted HGAs demonstrating that Rb-mediated regulation of cell-cycle progression is a core pathway in these tumors. The selective advantage for the development of PNET in this setting likely reflects a less specific manner in which cells are targeted for Cre-mediated recombination. Indeed, accumulating experimental evidence indicates that different types of brain tumors, and even different subgroups within a histological diagnosis, arise as a consequence of specific driver mutations occurring in specific susceptible cell populations [44, 45]. It follows that for relatively promiscuous driver mutations such as loss of Trp53, Rb1 or Pten, the cell being targeted as well as the developmental context in which mutations occur, are likely to be critical determinants of the resulting tumor type.

The importance of developmental context is apparent in a study in which an intragenic deletion of Trp53 in embryonic neural stem cells was mediated by a GFAP-Cre transgene [46]. HGAs were primarily observed however, a discernible fraction of tumors that developed were medulloblastomas, likely arising from progenitor cells targeted in the developing mouse brain. In agreement with the two previously mentioned studies the HGAs described in these mice also demonstrated dysregulation of the Rb and RTK/Pi3k/Akt pathways.

The contributions of various intermediates of the RTK/Pi3k/Akt pathway have been explored in GEMMs. Consistent with the finding that Pten deletion alone does not lead to gliomagenesis, activation of other pathway intermediates on their own is similarly inefficient at initiating HGAs. Neither virally-induced expression of the activated EGFR variant, EGFRvIII [47], nor PDGF-B in adult neural progenitor cells [48] was sufficient on their own to initiate HGA. However, when coupled with germ-line deletion of Cdkn2a and deletion of Pten in the striatum, EGFRvIII expression resulted in robust tumor formation [47]. Likewise, germ-line deletion of Cdkn2a or Trp53 coupled with expression of PDGF-B efficiently initiated HGAs in adult mice [48]. Underscoring the differences between adult and pediatric tumor biology, when PDGF-B was expressed by targeted viral transduction into neonatal neural progenitors, many mice developed low grade gliomas while a small number harbored HGGs [49]. The proportion and aggressiveness of HGGs were greatly enhanced by germ-line loss of the Cdkn2a gene but not by Trp53 deletion.

The role of the Nf1 tumor suppressor gene in gliomagenesis has also been investigated using GEMMs. Conditional ablation of the gene in the developing brain using a GFAP-Cre driver mouse line was insufficient to initiate gliomagenesis [50] however, when paired with germ-line or conditional deletion of Trp53, malignant gliomas were efficiently produced [51, 52]. As in the other models discussed, tumors in this context demonstrated Rb pathway dysregulation as demonstrated by robust expression of Cdk4 and Cyclin D1 [52]. Interestingly, the Pi3k/Akt pathway was significantly activated in Nf1-null HGAs implying crosstalk between these downstream RTK effectors. Furthermore, loss of the Pten tumor suppressor enhanced gliomagenesis in this model [51]. These investigators subsequently showed that deletion of Nf1 and Trp53 with or without Pten loss in adult neural progenitors gave rise to HGAs [53].

It is particularly interesting to note that in these murine astrocytomas, multiple mutations within core pathways were frequently observed [42]. RTK gene amplifications were noted in both Trp53 and Pten as well as in Trp53, Pten and Rb1 mutated tumors. Importantly, these amplifications appeared to drive maximal signaling through the Pi3k pathway suggesting that ablation of Pten function, the principal negative regulator of Pi3k signaling, is not sufficient to activate the pathway. HGAs with concurrent mutations of PTEN and RTK genes have been noted in several studies [14, 58]. Furthermore, several murine astrocytomas were characterized by co-amplifications of more than one RTK gene [42], a phenomenon that has also been appreciated in human tumors [14, 59, 60]. In similar observations, multiple mutations affecting Rb pathway signaling were observed in both mouse [42] and human HGAs [14, 59]. These observations have important implications for the selection of targeted agents and suggest that pathway inhibition at multiple nodes may be required to achieve maximal therapeutic effect.

A recent study employed a retrovirus encoding PDGF-B and the Cre recombinase which was directly injected into the subcortical white matter of adult mice [61]. HGAs arose efficiently in mice in which Pten or the combination of Pten and Trp53 were conditionally targeted. Gene expression analyses revealed that all tumors expressed a signature strongly associated with the Proneural subgroup which is consistent with PDGFRA amplified tumors in adults [29, 31]. The authors also used the mouse HGA expression profiles to identify two groups of patients with Proneural tumors with significantly different survival outcomes [61].

However, a recent report utilized the same GFAP-Cre driver line in conjunction with an elegant genetic tag to label cells that had undergone homozygous deletion of Trp53 and Nf1 [63]. These investigators demonstrated that while tumor suppressor deletion occurred in the NPC, these cells remained quiescent until after differentiation into oligodendrocyte precursor cells (OPCs) when they acquired a proliferative phenotype associated with cellular transformation. Similarly, the OPC was also implicated as the transforming cell type in experiments where mutant cells were retrovirally labeled [61]. Interestingly, in the latter study, viral injections were targeted to subcortical white matter and not to the subventricular zone. Clearly, GEMMs using both virally delivered Cre and mouse Cre-expressing lines have demonstrated that HGAs can arise outside of the canonical adult neural progenitor cell niches [42, 47, 48]. The nature of the tumor initiating cell has yet to be rigorously investigated in these models.