New Directions in the Treatment of Glioblastoma

Zachary J. Reitman, MD, PhD; Frank Winkler, MD, PhD; Andrew E. H. Elia, MD, PhD


Semin Neurol. 2018;38(1):50-61. 

In This Article

Molecular Features

Integrated Genomic Analyses

Integrated genomic analyses in the past decade have uncovered mutational, copy number, gene expression, and epigenetic alterations in GBM.[5–8] The most frequent somatic alterations in GBMs include point mutations in the promoter of the telomerase catalytic subunit TERT (in 60–80% of GBMs), deletion of CDKN2A/B (~50%), amplification or mutation of the receptor tyrosine kinase EGFR (~40%), inactivation of the tumor suppressors TP53 (~30%) and PTEN (~30%), alterations in PIK3CA or PIK3R1 (~20%), CDK4 amplification (~15%), RB1 inactivation (~10%), NF1 mutation (~10%), and IDH1/2 mutation (5–10%).[5,8,43] Uncommon activating mutations of the kinase BRAF (1–2%) occur in a new 2016 WHO-defined entity called "epithelioid glioblastoma"[44–46] and are notable given the activity of BRAF-targeted therapies in other cancer types.[47] Most EGFR-amplified GBMs contain additional activating mutations or rearrangements of EGFR, such as the EGFR vIII deletion which is present in ~30% of GBMs.[8] Pathway analysis has revealed that most GBMs contain alterations in the TP53, RB, and RTK/PI3K/RAS pathways, and that genes in each pathway tend to be altered in a mutually exclusive manner. For instance, most tumors contain an alteration of at least one RB pathway gene (CDK4, CDK6, CCND2, CDKN2A/B, or RB1).[5,8] The IDH and TERT mutations have helped to refine the classification of WHO grade II to IV adult diffuse gliomas, as most primary GBMs are IDH wild-type with TERT mutations, while grade II to III astrocytomas and secondary GBMs are IDH mutant without TERT mutations. Oligodendrogliomas contain both IDH mutations and TERT mutations, in addition to characteristic loss of heterozygosity of 1p/19q (Figure 2).

Figure 2.

IDH, TERT, and ATRX alterations refine the classification of adult diffuse gliomas. The combination of IDH and TERT genotyping aids in the classification of diffuse gliomas. Most primary glioblastomas (GBMs) contain TERT mutations but not IDH mutations; grade II to III astrocytomas and secondary GBMs contain IDH mutations but not TERT mutations; and grade II to III oligodendrogliomas contain both TERT and IDH mutations. Also shown are TP53 and ATRX mutations commonly found in astrocytic tumors and 1p/19q loss of heterozygosity found in grade II to III oligodendrogliomas.

The Cancer Genome Atlas and other groups have proposed classification systems for GBM based on bulk tumor mRNA expression patterns. This has been used to classify tumors as proneural, neural, mesenchymal, or classical subtypes, which associate with specific genetic events and clinical profiles.[7] For example, the proneural subgroup contains the IDH-mutant GBMs; EGFR alterations are enriched in the classical subgroup, and NF1 alterations are enriched in the mesenchymal subgroup. Furthermore, the mesenchymal subtype has been associated with radioresistance,[48] while the proneural subtype was linked to an overall survival benefit from bevacizumab in the AVAglio trial.[49] However, single-cell RNA sequencing has revealed that most GBMs comprise tumor cell subpopulations exhibiting most or all of these expression subtypes,[50] calling into question their clinical utility. DNA methylation signatures have also been proposed to classify GBM, and a glioma CpG island methylator phenotype (G-CIMP) was identified that tightly associates with IDH-mutant GBM.[51] The complexity of these classification schemes and the problem with tumor heterogeneity has limited their translation to the clinic.


The DNA repair enzyme MGMT protects against the cytotoxic effect of temozolomide by removing methyl adducts deposited on guanine bases by the drug. Epigenetic silencing of MGMT by methylation of its promoter remains an important GBM prognostic marker and predictor of temozolomide response. In EORTC26981/22981 NCIC CE.3, MGMT promoter methylation was an independent favorable prognostic factor for overall survival.[52] Among patients with MGMT promoter methylation, a survival benefit was observed for those who received temozolomide compared with those who did not (21.7 versus 15.3 months). In the elderly NOA-08 and Nordic trials, temozolomide benefit was similarly observed in patients with MGMT promoter methylation but not in patients without it. In the elderly EORTC 26062/NCIC CE.6 trial, however, a strong trend toward temozolomide benefit was observed in patients whose tumors lacked MGMT promoter methylation. MGMT status may influence the decision to withhold temozolomide or the duration of temozolomide treatment in the elderly (Figure 1), although further research is needed given results from the EORTC 26062/NCIC CE.6 trial. Notably, the G-CIMP phenotype partially overlaps with MGMT promoter methylation, which occurs in 79% of G-CIMP tumors yet only 46% of non-G-CIMP tumors.[8] Lastly, in contrast to MGMT promoter methylation which increases temozolomide sensitivity, mutation of mismatch repair (MMR) genes confers temozolomide resistance in cellular models and is associated with GBM recurrence after temozolomide treatment in patients.[6,53,54]

IDH Mutations

Perhaps the most important molecular marker for grade II to IV gliomas, mutations in isocitrate dehydrogenases 1 and 2 (IDH1 and IDH2, collectively referred to as IDH) have made their way into widespread clinical practice since their discovery in a genome-wide analysis in 2008.[5] Mutations in IDH1 and less commonly IDH2 are found in almost all grade II to III astrocytomas, grade II to III oligodendrogliomas, and secondary GBMs that have transformed from lower grade gliomas, while 90% of primary GBMs are wild-type for IDH[55] (Figure 2). The current WHO classification now categorizes IDH wild-type and IDH-mutant GBM as distinct entities.[56] Among GBM patients, IDH mutation is associated with a younger median age at diagnosis, a longer median overall survival, a predilection for the frontal lobes, and less necrosis and contrast-enhancement on MRI[55,57] (Figure 3). IDH mutational status can be routinely assessed by immunohistochemistry for the IDH1-R132H mutation, which accounts for ~90% of all IDH mutations, or by sequencing-based analyses. Mechanistic studies have demonstrated that IDH mutations result in a gain of function that produces the oncometabolite 2-hydroxyglutarate (2HG).[58] This metabolite promotes epigenetic silencing through inhibition of α-KG-dependent dioxygenases, which include Jumonji histone lysine demethylases and TET DNA hydroxylases that promote histone and DNA demethylation, respectively[59,60] (Figure 4). In addition, 2HG can activate EGLN prolyl hydroxylases to promote HIF1 degradation and cellular transformation.[60] Notably, IDH mutation alone is not sufficient for tumorigenesis but requires additional genetic alterations.[59,61] These studies have led to the development of small molecule IDH inhibitors, which are in early clinical trials, and synthetic lethal approaches to target IDH-mutant tumors.[59]

Figure 3.

Features of GBM classified by different driver alterations. Typical age at diagnosis, associated genetic alterations, anatomic location, tumor features, and prognosis for GBMs classified by histone 3 status, IDH mutation status, and MGMT status. ALT, alternative lengthening of telomeres; DIPG, diffuse intrinsic pontine glioma; GBM, glioblastoma; G-CIMP, glioma CpG island methylator phenotype; IDH, isocitrate dehydrogenase; MGMT, O6-methylguanine methyltransferase.

Figure 4.

Emerging genetic alterations in GBM. The mutational spectrum and molecular mechanisms thought to promote tumorigenesis for IDH1 and IDH2,60 TERT,62 ATRX,64 H3F3A, and HIST1H3B.73,101 ALT, alternative lengthening of telomeres; ETS, E-twenty-six; GBM, glioblastoma; GABP, GA-binding protein; HR, homologous recombination; 2HG, 2-hydroxyglutarate; H3K27me3, histone 3 lysine 27 trimethylation; H3K36me3, histone 3 lysine 36 trimethylation; IDH, isocitrate dehydrogenase; PRC2, polycomb repressive complex 2; TERT, telomerase reverse transcriptase; TF, transcription factor.

Telomere Maintenance—TERT and ATRX

Telomeres are incompletely replicated during the cell cycle and thus shorten at each cell division, eventually causing cellular senescence. To prevent senescence, most GBMs upregulate the enzyme telomerase (TERT), which can extend chromosome ends. In up to 83% of primary GBMs, upregulation is accomplished through TERT promoter mutations,[43] which can aberrantly recruit the ETS transcription factor GABP[62] (Figure 4). A subset of GBMs maintains telomere length through another pathway called alternative lengthening of telomeres (ALT) that is associated with better overall survival[63] and transformation from lower grade astrocytomas. Inactivation of the helicase and histone chaperone ATRX is associated with the establishment of ALT.[64] Recently, ATRX mutations have been discovered in 85% of grade II to III astrocytomas,[65] in the majority of secondary GBMs,[66] and in 30% of pediatric GBMs[67] and correlate with the ALT phenotype in these tumors. ATRX may inhibit ALT by protecting against replication fork stalling at telomeres and thereby minimizing telomeric homologous recombination which is necessary for ALT[64] (Figure 4). ATRX may reduce fork stalling by unwinding quadruplex DNA at telomeres and by promoting telomeric deposition of the histone H3.3.[64] Notably, TERT and ATRX mutations are mutually exclusive in GBM[43] and therefore define GBM subsets with distinct pathways for maintaining telomere length (Figure 2). Interestingly, ALT cells are sensitive to inhibition of the replication stress kinase ATR,[68] making this kinase a potential target for the treatment of gliomas with inactivation of ATRX.

Histone H3 Mutations—K27M and G34R/V

Recently, genome-wide sequencing studies have discovered mutations in histone H3 genes in GBM. Specifically, the mutation K27M occurring in the histone variants H3.3 and H3.1 has been identified in 20% of pediatric GBMs[67,69] and nearly 80% of pediatric diffuse intrinsic pontine gliomas (DIPGs).[69] This mutation is present in not only pediatric tumors but can also occur in GBMs in young adults.[70,71] Importantly, K27M tumors tend to be centered in midline structures (such as the pons, midbrain, and thalamus) (Figure 3), which has led to the definition of "diffuse midline glioma, H3 K27M-mutant" as a new grade IV entity in the 2016 WHO classification.[56] Mechanistically, K27M mutation leads to global reduction in histone H3 K27 trimethylation through impaired recruitment of PRC2 and inhibition of the K27 methylase EZH2 within the PRC2 complex[72,73] (Figure 4). Consistent with this mechanism of pathogenesis, targeted increase in global H3 K27M trimethylation levels using demethylase or deacetylase inhibitors inhibits growth of K27M tumors in xenograft studies.[74,75] Additional therapeutic targets for K27M tumors include the activin receptor ACVR1, which harbors stimulating mutations in ~20% of K27M DIPGs,[76] and the p53 phosphatase PPM1D, which is mutated in nearly 40% of K27M brainstem gliomas in children and adults.[77]

In addition to K27M, recurrent mutations in Gly34 of histone H3.3 (G34R/V) have been identified in GBMs. While the mutations are just a few amino acids apart in histone H3.3, tumors with G34R/V mutation are clinically and biologically distinct from those with K27M mutation. G34R/V tumors tend to occur in the cerebral hemispheres, follow a more favorable clinical course, and have a different epigenetic signature compared with K27M tumors[71] (Figure 3). Like IDH-mutated GBMs, the G34 GBMs almost invariably contain ATRX mutations.[67] The G34 tumors also appear to comprise a more heterogeneous group, with histopathologic features typical of either GBM or PNETs (primitive neuroectodermal tumors), which has led to the proposal that they be considered a single clinical entity.[46] Proposed mechanisms of oncogenesis for the histone 3 mutations and for other emerging molecular alterations detailed above are shown in [Figure 4].

Intratumoral Heterogeneity

Molecular intratumoral heterogeneity occurs commonly in GBMs and can manifest as the presence of different subclones in the same tumor.[78] For example, subclones with mutually exclusive amplifications of different receptor tyrosine kinases (EGFR, PDGFR, or MET) have been found intermingled within the same tumor.[79,80] Intratumoral heterogeneity has also been observed for PTEN deletion[57] and for gene expression signatures identified by single cell RNA-sequencing.[50] Tyrosine kinase inhibitors targeting EGFR and a vaccine against the constitutively active EGFR vIII mutant have demonstrated limited success in clinical trials,[81–83] possibly because EGFR alterations affect only a subclonal portion of tumor cells and are late events.[57,78,79] In contrast, IDH and histone H3 mutations harbor characteristics that suggest they are initiating events. For instance, IDH mutations in secondary GBMs appear at the earliest biopsy of the disease,[54,55] are generally stable throughout tumor development,[54,55] and have a mutational signature consistent with appearance before or concurrent with p53 mutation.[57] Similarly, histone H3 K27M mutations in DIPGs are homogeneous throughout tumor tissue.[84] These potentially early driver alterations, which define GBMs with distinct clinical features including prognosis, patient age, tumor location, and other genetic features (Figure 3), may represent promising therapeutic targets. Notably, founder events in IDH-WT GBMs, which account for the majority of adult GBMs and which can arise within months, remain to be more clearly identified.[78,85] Ultimately, given the diversity of pathways involved in GBM pathogenesis, combination of molecular therapies will likely be necessary.

Tumor Microtubes as a Novel Mechanism for GBM Progression and Resistance

The recent discovery of ultra-long membranous protrusions that extend from astrocytoma cells, called tumor microtubes (TMs), has changed our understanding of GBM[86,87] (Figure 5). TMs interconnect individual GBM cells in a single communicating syncytium with extensions that invade and colonize the brain by sending nuclear material to distant locations.[86] Multiple characteristics of TMs have translational importance. First, approximately half of the cells in a malignant astrocytoma are connected through the TM network. Importantly, these cells are resistant to the cytotoxic effects of both radiotherapy[86] and temozolomide.[88] Second, TMs extend to surgical lesions in a mouse model, which may contribute to local recurrence after resection.[88] Third, GBM cells hijack CNS developmental programs, involving factors such as GAP-43[86] and Ttyh1,[89] to promote TM formation (Figure 5), making these pathways promising therapeutic targets despite their attenuated function in the adult CNS. Fourth, development of functional TMs depends on intact 1p/19q status, which may be due to the presence of multiple neurotrophic factors, such as Ttyh1, on the 1p and 19q arms[86,89] (Figure 5). This point is interesting because it may explain why patients with 1p/19q codeleted gliomas have better survival independent of IDH mutation.[90] Finally, communication via the TM network, which is enabled by Cx43 gap junctions between cells, can be blocked by gap junction inhibitors and other means to impair TM function.[87] In summary, gliomas appear to be organized as a single functional unit with syncytial connections between the glioma core and extensions that invade adjacent brain tissue. Research in this area is likely to shed further light on pathways regulating TM formation and on oncogenic consequences of TM function, which may lead to novel therapies that inhibit this important mechanism of GBM resistance.

Figure 5.

Tumor microtube (TM) functions during glioma progression. Overview of the four known molecular functions of TMs, whose development depends on intact 1p/19q status. While GAP-43 is necessary for all four functions (invasion, proliferation, interconnection, and network formation), TTYH1 appears to promote only invasion and tumor cell proliferation. Cx43, connexin 43 gap junctions; ER, endoplasmic reticulum; MT, microtubules; Mito, mitochondria; ICW, intercellular calcium waves; TM, tumor microtube. Adapted from prior reference.86