Updates in prognostic markers for gliomas

Abstract

Gliomas are the most common primary malignant brain tumor in adults. The traditional classification of gliomas has been based on histologic features and tumor grade. The advent of sophisticated molecular diagnostic techniques has led to a deeper understanding of genomic drivers implicated in gliomagenesis, some of which have important prognostic implications. These advances have led to an extensive revision of the World Health Organization classification of diffuse gliomas to include molecular markers such as isocitrate dehydrogenase mutation, 1p/19q codeletion, and histone mutations as integral components of brain tumor classification. Here, we report a comprehensive analysis of molecular prognostic factors for patients with gliomas, including those mentioned above, but also extending to others such as telomerase reverse transcriptase promoter mutations, O⁶-methylguanine-DNA methyltransferase promoter methylation, glioma cytosine-phosphate-guanine island methylator phenotype DNA methylation, and epidermal growth factor receptor alterations.

Gliomas have an incidence of approximately 5 cases per 100000 persons.¹ Historically, gliomas have been classified according to the microscopic, immunohistochemical, and ultrastructural similarity of the tumor cells with the presumed normal cells from which they arose.

Grading has been based on features such as mitoses, necrosis, and microvascular proliferation. The combination of histologic features and tumor grade has provided important prognostic information. In adults, diffuse gliomas have been categorized into the diagnoses of astrocytoma (World Health Organization [WHO] grade II or III), oligodendroglioma (WHO grade II or III), and glioblastoma (WHO grade IV). The overall survival (OS) for grades II, III, and IV astrocytomas is approximately 6–8 years, 2 years, and 15 months, respectively. In addition to the association with poorer prognosis, higher tumor grade is correlated with more advanced age. Increasingly, genetic biomarkers have become essential components of integrated pathologic diagnoses, and their use has transformed the paradigm of brain tumor classification.

In 2016, the WHO classification schema for adult gliomas was significantly updated to incorporate important new findings on the genomics of diffuse gliomas. In this update, the approach to brain tumor classification was expanded to include both histopathologic and molecular features,² thus integrating phenotypic and genotypic information. Unlike prior editions of the WHO classification, molecular information is now considered integral to the definition of adult gliomas. In particular, the revised edition now requires integrated pathologic diagnoses that include information about the mutation status of the isocitrate dehydrogenase (IDH) genes and about 1p/19q codeletion. Currently, adult gliomas generally fall into 3 major groups: IDH mutant with 1p/19q codeletion, IDH mutant with 1p/19q intact, and IDH wildtype.

In recent years, additional molecular alterations have been discovered that are likely to have important clinical implications for glioma prognostication. To handle the accelerating pace of scientific discovery and the need for clarification and new guidelines for practicing diagnosticians between WHO updates, a consortium has been established, cIMPACT-NOW³ (the Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy), to facilitate consensus review of new diagnostically relevant information and determining how such information should be used in clinical practice and how it can be incorporated into future updates of the WHO CNS tumor classification.

Molecular Prognostic Markers

IDH Mutation

Recurrent mutations in the metabolic gene IDH1 were discovered initially in the context of a genomic evaluation of glioblastoma (GBM),⁴ in which a heterozygous point mutation changed arginine to histidine at amino acid 132 in approximately 20% of the tumors analyzed. Further work demonstrated that mutations at R132 can be detected in up to 80–90% of grade II and III gliomas, including both astrocytic and oligodendroglial subtypes, and in a small subset of GBMs. While the vast majority of mutations involve the R132H substitution, noncanonical variants including R132C, R132S, R132G, and R132L also occur.⁵ The mutations tend to be found in younger patients.⁶ Much less commonly, in less than 1% of IDH mutant gliomas, a mutation of the related IDH2 gene has been observed at an analogous arginine, amino acid 172.^5,6IDH2 mutations tend to occur in oligodendrogliomas.

IDH1 and IDH2 are NADP+-dependent metabolic enzymes that are critical for the oxidative carboxylation of isocitrate to alpha-ketoglutarate, with IDH1 functioning in the cytoplasm and IDH2 working within the context of the Krebs cycle within the mitochondria. Mutation at R132 (or R172 in IDH2) results in the acquisition of a neoenzymatic activity that promotes conversion of alpha-ketoglutarate to R-2-hydroxyglutarate (2-HG).⁷ This change leads to accumulation of exceedingly high levels of the 2-HG metabolite which can promote tumorigenesis. Tumor evolution studies suggest that mutation of IDH is an early event in gliomagenesis.⁸

While a complete understanding of the mechanism of tumor promotion is still under study, the supra-physiologic levels of 2-HG detected in IDH mutant glioma are implicated in the inhibition of a number of 2-oxoglutarate (2-OG)–dependent enzymes. These proteins include hydroxylases, histone demethylases, and DNA repair enzymes which have pleiotropic effects, including effects on cellular epigenetic state.⁹ Consistent with these effects, IDH mutant gliomas exhibit a typical pattern of hypermethylation of cytosine-phosphate-guanine (CpG) islands, commonly referred to as the CpG island methylator phenotype (CIMP).^10,11 The epigenetic patterns promoted by the presence of the IDH mutation is proposed to lock glioma cells in a less differentiated or stem cell–like state, thereby rendering them prone to acquisition of additional genomic alterations that are known to promote tumorigenesis, including TP53 mutation and loss of chromosome arms 1p and 19q.^9,12,13

Of all of the known glioma-associated molecular alterations discovered to date, the presence or absence of an IDH mutation has the largest prognostic significance. IDH mutations are noted in the vast majority of grade II and grade III gliomas,⁶ which are associated with improved survival compared with GBM. Further, there is evidence from retrospective analyses that IDH mutation status is as strong a predictor of survival as histologic grading that differentiates high-grade astrocytomas. In a retrospective analysis of nearly 400 anaplastic astrocytomas (AAs) and GBMs, the presence of an IDH mutation more strongly predicted OS than did histologic grade. The authors demonstrate in a multivariate model incorporating age, IDH mutation status, extent of surgical resection, histologic diagnosis, and MGMT status that the presence of IDH mutation was the dominant prognostic factor (relative risk, 2.7; 95% CI: 1.6–4.5). Simultaneous evaluation of histology and IDH mutation by Kaplan–Meier analysis revealed improved progression-free survival (PFS) and OS for both IDH mutant AA and GBM compared with IDH wildtype AA and GBM.¹⁴

There is a relationship between the presence of an IDH mutation and improved prognosis in grade II and grade III diffuse glioma as well. When 271 grade II diffuse gliomas were evaluated based on a number of clinical and molecular parameters, mutation of IDH was significantly associated with increased OS on univariate analysis (P = 0.002), as well as in a multivariate model incorporating known clinical factors such as age, performance status and extent of resection, and chromosome 1p/19q status (P = 0.003).¹⁵ Likewise, in an analysis of 552 tumors that consisted of both grade II and grade III gliomas, IDH mutant gliomas exhibited prolonged OS compared with IDH wildtype gliomas, regardless of grade. Indeed, only a modest effect of grade on OS was observed in the IDH mutant tumors analyzed.¹⁶

The favorable prognostic profile of IDH mutations was also demonstrated in a comprehensive genomic analysis of 293 grade II and grade III gliomas performed by The Cancer Genome Atlas (TCGA) Research Network. Patients without an IDH mutation exhibited a significantly shorter OS compared with those with an IDH mutation, with a median of 1.7 years for IDH wildtype glioma patients. This is in contrast to an OS of 6.3 years for patients with IDH mutation and no 1p/19q codeletion (astrocytic gliomas) or 8 years for patients with IDH mutation and 1p/19q codeletion (oligodendroglial gliomas).¹⁷ Additionally, in a recent study of grade III glioma patients treated with radiotherapy and either temozolomide (TMZ) or nitrosourea, IDH mutation status was found to be a significant prognostic factor for both OS (hazard ratio [HR] 0.42) and PFS (HR 0.59).¹⁸

Low-grade gliomas that are wildtype for IDH mutation were found to be genomically heterogeneous and to have independent prognostic factors.¹⁹ Out of 718 grade II and III gliomas that were genotyped for IDH mutations, 166 wildtype cases were found. These tumors were genotyped for epidermal growth factor receptor (EGFR) and myeloblastosis (MYB) amplifications as well as telomerase reverse transcriptase (TERT) promoter, H3 histone family 3A (H3F3A), and proto-oncogene B-Raf (BRAF) mutations. The most favorable prognostic factor was found to be MYB amplification, whereas EGFR amplification and H3F3A mutation conferred an unfavorable prognosis.

Table 1.

Summary of prognostic molecular markers in glioma

Molecular Marker	Functional Significance	Frequency	Prognostic Value	Diagnostic Evaluation
IDH mutations	Lead to accumulation of oncometabolite 2-HG7, which inhibits 2-OG dependent enzymes and alters epigenetic state	80–90% grade II and III gliomas, 12% of GBMs	Favorable prognosis Grades II–III: HR 7.5 for IDH mutant vs IDH wildtype tumors Grades III–IV: RR 2.714	Routinely performed IHC, sequencing for IHC-negative samples, imaging
1p/19q codeletion	Deletion of tumor suppressor genes, candidates include CIC33	Defines tumors of oligodendroglial lineage	Favorable prognosis and predictive of response to chemotherapy Grade II: PFS 62 vs 48 vs 20 months for IDH mutant and 1p/19q codeleted, IDH mutant alone and IDH wildtype respectively38 Grade III: Median OS 14.7 vs 2.6 years for codeleted and non-codeleted tumors treated with RT + chemotherapy39	Routinely performed FISH, aCGH, SNP arrays
MGMT promoter methylation	Silencing of MGMT expression, which mediates resistance to alkylating agents	40% of gliomas, more common in lower grade IDH mutant	Favorable prognosis and predictive of response to temozolomide Grade IV: Median OS 18.2 vs 12.2 months for methylated vs unmethylated tumors irrespective of treatment46 Median OS 21.7 vs 12.7 months for methylated vs unmethylated tumors among patients who received treatment with TMZ46	Routinely performed qMSP, IHC
G-CIMP methylation	Silencing of tumor suppressor genes and mismatch repair proteins	8.8% of GBMs 55% of IDH mutant gliomas	Favorable prognosis IDH mutant tumors: median OS 7.2 vs 2.7 years for G-CIMP high vs low respectively62	Not routinely performed
TERT promoter mutations	Reactivation of telomerase and telomere maintenance	80% of IDH wild type GBMs	Poor prognosis Grades II–III: HR 11.74 for TERT mutant only tumors vs tumors with TERT, IDH mutation and 1p/19q codeletion34	Not routinely performed but detected in sequencing panels
EGFR alterations	Constitutive activation of EGFR pathway, involved in cell proliferation, apoptosis control, cell invasion	EGFR amplification: 40–50% of IDH wildtype GBMs EGFR vIII: 50% of EGFR- amplified tumors	High EGFR expression confers a poor prognosis Grade IV: HR 1.57 for high expressing tumors vs low expressing ones82	Not routinely performed but detected in sequencing panels
BRAF V600E mutations	Constitutive activation of MAPK pathway, which controls cell proliferation, differentiation, apoptosis and migration	10–15% of pilocytic astrocytomas, 5–10% pediatric gliomas, 34% of glioneuronal tumors, <2% adult gliomas	Favorable prognosis in young patients All gliomas: HR 0.51 for pediatric patients and 0.43 for younger adults (age <35)91	Not routinely performed but detected in sequencing panels
Histone mutations, H3K27 mutation can occur in histone H3.1 or H3.3	Regulation of transcription mediated by reduction of H3K27 methylation	Defines diffuse midline gliomas, H3K27 mutant, predominantly pediatric	Poor prognosis Pediatric gliomas: median OS 1.04 vs 6.1 years for H3K27 mutant tumors vs wildtype ones94 Adult gliomas: median OS 19.6 months, comparable to IDH wildtype tumors96	Not routinely performed but detected in sequencing panels

Abbreviations: CIC, protein capicua homolog; BRAF, v-raf murine sarcoma viral oncogene homolog B; MAPK, mitogen-activated protein kinase.

The critical role of IDH mutation in prognostication has led to routine evaluation of this biomarker in all glioma samples, whenever possible. IDH mutations are clinically detected using immunohistochemistry (IHC) with a mutation-specific antibody.^20,21 In a number of studies comparing IHC with sequencing, the concordance rate between these methods was found to be between 88% and 99%. The main reason for this discrepancy is that IHC fails to detect less frequent IDH mutations such as R132C, R132L, R132S, and R132G²² and those in IDH2. This prompts the need to evaluate IHC-negative samples using sequencing methods. The IDH status of a tumor can also be assessed using imaging techniques such as magnetic resonance spectroscopy,^23–25 tissue-based analysis such as Raman spectroscopy²⁶ and Fourier-transform infrared spectroscopy,^26,27 as well as intraoperative testing using mass spectrometry imaging^26,28 or rapid genotyping assays.^26,29

1p/19q Codeletion

The association between heterozygous loss of the short arm of chromosome 1 and the long arm of chromosome 19 (1p/19 codeletion) and improved prognosis of glioma has been appreciated since the 1990s.^30–32 The 1p/19q codeletion is found in tumors of the oligodendroglial lineage. Several candidate genes that may be lost with this alteration have emerged, including the homolog of Drosophila capicua (CIC), found to be mutated in up to 50% of oligodendroglial tumors.³³ CIC is a high mobility group box for transcriptional repression and promoter targeting. Loss of this tumor suppressor is thought in part to promote transcriptional programs that favor tumor growth.

Loss of 1p/19q co-occurs with mutation of IDH.³⁴ Patients with oligodendroglial tumors with this combination of molecular alterations tend to survive the longest of all patients of the glioma subtypes. One of the first studies to demonstrate the importance of 1p/19q codeletion on prognosis was a retrospective analysis of 125 tumors analyzed for the presence of 1p/19q codeletion, in which the median OS of codeleted tumors was 11.9 years (95% CI: 10.4 to not reached), significantly longer than the median survival of 8.1 years for patients with tumors that were 1p/19q intact (95% CI: 4.1‒11.2).³⁵

This association of 1p/19q codeletion and prolonged OS has been observed in many other studies since that time, including in large genomic analyses of gliomas^34,36 and prospective clinical trials. The Radiation Therapy Oncology Group (RTOG) trial 9802 was a large study undertaken to compare the efficacy of radiation treatment combined with a 3-drug chemotherapy regimen (procarbazine, lomustine, and vincristine [PCV]) to radiation treatment alone in patients with grade II gliomas who were deemed to be “high risk” and require adjuvant treatment. Notably, patients with the histologic subtype of oligodendroglioma experienced the longest PFS and OS on subgroup analysis.³⁷ Although 1p/19q status was available for only a small number of tumors in the trial population, it is known from other studies that the majority of tumors with oligodendroglial histology have 1p/19q codeletion. In a large, phase III randomized study comparing TMZ with radiation therapy for treatment of patients with grade II glioma, the authors undertook an exploratory analysis based on the molecular features of IDH mutation and 1p/19q status. Patients with combined 1p/19q codeletion and IDH mutation enjoyed the longest PFS regardless of treatment, at 62 months (95% CI: 41–not reached) compared with 48 months (95% CI: 41–55) for IDH mutant alone and 20 months (95% CI: 21–26) for IDH wildtype.³⁸

The improved prognosis associated with 1p/19q codeletion has further been appreciated in grade III tumors as well. RTOG 9402, comparing PCV plus radiation treatment to radiation treatment alone in patients with grade III oligodendrogliomas and oligoastrocytomas, showed that patients with codeleted tumors lived significantly longer than others regardless of treatment group. In this trial, the median OS for patients with 1p/19q codeleted tumors was 14.7 years in the PCV plus radiation group and 7.3 years for radiation treatment alone, compared with 2.6 and 2.7 years for patients with non-codeleted tumors.³⁹ Altogether, these data support the concept that codeletion of 1p/19q is a prognostically favorable molecular marker associated with longer survival times.

In addition to being a favorable prognostic marker, 1p/19q deletion was found to be predictive of response to chemotherapy in oligodendroglial tumors. In the European Organisation for Research and Treatment of Cancer (EORTC) 26951 study, patients with anaplastic oligodendroglial tumors were assigned to receive radiation alone or in combination with 6 cycles of adjuvant PCV.⁴⁰ There was an overall benefit for combination therapy, which was higher for patients with 1p/19q codeleted tumors. Patients with codeleted tumors did not reach OS in the combination therapy arm, while OS was 112 months in the radiation only arm. For the non-codeleted cohort, OS was 25 months for combination therapy and 21 months for radiation alone. Similar results were seen in the RTOG 9402 trial, where patients with codeleted tumors derived a significantly larger benefit from chemotherapy compared with patients with wildtype tumors.³⁹

In addition to 1p/19q codeletion, other chromosomal copy number changes were found to correlate with prognosis in low-grade gliomas. For example, in a study of 231 low-grade gliomas, 25% of tumors were found to have loss of heterozygosity (LOH) on chromosome 9p and 14% had LOH on 10q.⁴¹ These alterations did not associate with each other. LOH at both of these sites correlated with a poor prognosis (HRs for PFS were 1.46 for 9p LOH and 1.49 for 10q LOH, while HRs for OS were 0.98 for 9p LOH and 2.53 for 10q LOH). Expanding further on the relationship between copy number variation (CNV) and prognosis, a recent retrospective study examining >300 astrocytic IDH mutant tumors (grades II–IV) noted a strong, significant association between high CNV load and prognosis. Notably, when the specific regions exhibiting CNV were analyzed, the strongest association with OS was observed for homozygous deletion of cyclin-dependent kinase inhibitor 2A and 2B (CDKN2A/B), found on chromosome 9p21, with IDH mutant patients with CDKN2A/B deletion living for much shorter times following diagnosis compared with those without the deletion.⁴²

As mentioned above, the 2016 WHO classification for nervous system tumors established molecular diagnostic requirements for some intracranial malignancies. For example, the presence of both an IDH mutation and the 1p/19q codeletion is necessary for a diagnosis of oligodendroglioma. Therefore, analysis of 1p and 19q is routinely done as part of clinical practice using fluorescence in situ hybridization (FISH), array comparative genomic hybridization (aCGH) microarrays, single nucleotide polymorphism (SNP) microarrays, or next-generation sequencing techniques.⁴³

MGMT Promoter Methylation

O⁶-methylguanine-DNA methyltransferase (MGMT) is a DNA repair enzyme that removes alkyl groups from the O⁶ position of guanine, which is the critical site modified by alkylating chemotherapeutics.^44,45 This activity allows MGMT to effectively reverse the damage induced by TMZ, the chemotherapy used as standard of care for GBM and frequently used in treatment of grade III and grade II gliomas. Therefore, high levels of MGMT activity can render tumors resistant to alkylating agents. Interestingly, approximately 40% of gliomas exhibit epigenetic modification of the MGMT gene promoter in the form of methylation, which leads to decreased MGMT expression and enhanced sensitivity to TMZ and other alkylating agents.⁴⁴

MGMT promoter methylation serves as both a predictive and prognostic molecular marker in glioblastoma. From the landmark clinical trial that established temozolomide and radiation as standard of care for GBM, methylation of the MGMT promoter was found to be a clear predictive biomarker for tumors that were most sensitive to treatment with TMZ.^45,46 Interestingly, the investigators also noted that there was a significant difference in OS for patients with tumors exhibiting MGMT methylation compared with those without, regardless of treatment received. The median OS for patients with MGMT methylation was 18.2 months (95% CI: 15.5–22), compared with 12.2 months (95% CI: 11.4–13.5) in patients without methylated MGMT.⁴⁶ Analysis of the MGMT promoter methylation status of patients enrolled in RTOG 0525, which compared dose-dense TMZ administration to standard, monthly temozolomide, corroborated the association with methylation status and prognosis. Patients whose GBMs did not have MGMT methylation exhibited more rapid disease progression following diagnosis and a higher risk of death, particularly within the first 2 years following diagnosis (HR 1.87; 95% CI: 1.46‒2.17).⁴⁷

Despite the clear correlation between TMZ treatment and improved survival in patients with methylated MGMT, it is worth noting that patients with glioblastoma with unmethylated MGMT promoters appear to derive some benefit from TMZ. There was a 31% risk reduction for death in patients with unmethylated MGMT promoters who received both radiation and TMZ, compared with unmethylated patients who received radiation alone, though this difference was not statistically significant.⁴⁶ A similar trend of extended OS in patients with unmethylated MGMT promoters treated with both radiation and TMZ compared with radiation alone was observed in a randomized trial of elderly patients with GBM.⁴⁸ Together, these data suggest the presence of a marginal benefit from TMZ in patients with unmethylated GBM.

However, the prognostic value of MGMT promoter methylation is so strong that trials are currently under way to determine whether alternative therapies could be beneficial for patients with MGMT unmethylated GBM. For example, the EORTC 26082 study compared the standard of radiation plus TMZ with radiation plus temsirolimus, an inhibitor of mammalian target of rapamycin (mTOR), in MGMT unmethylated patients. In this study, however, patients receiving temsirolimus did not have a superior 1-year survival compared with patients receiving TMZ.⁴⁹ Additionally, the phase III study CheckMate 489 is aimed at investigating the efficacy of nivolumab and radiation compared with TMZ and radiation in newly diagnosed, MGMT unmethylated GBMs. Additionally, the Individualized Screening Trial of Innovative Glioblastoma Therapy (INSIGhT) is a biomarker-based study looking at the role of alternative adjuvant therapies to TMZ in these patients. The 3 experimental arms include adjuvant neratinib, abemaciclib, or CC-115 (dual inhibitor of mTOR kinase and DNA-dependent protein kinase).⁵⁰

MGMT promoter methylation is more common in lower-grade, IDH mutant gliomas.³⁸ This is consistent with data showing a correlation between the presence of an IDH mutation and a hypermethylator phenotype (discussed in more detail below). When MGMT status was retrospectively investigated in tumors of patients enrolled in the EORTC 22033–26033 trial comparing radiotherapy with dose-dense TMZ, it was noted that 100% of tested IDH mutant, 1p/19q codeleted tumors had a hypermethylated MGMT promoter compared with 86% of IDH mutant, non-codeleted tumors.³⁸ In the IDH mutant tumors in this cohort, the presence of MGMT promoter methylation was correlated with longer PFS only in the TMZ treatment arm,⁵¹ suggesting that MGMT promoter methylation status can be a useful aid for predicting which patients may respond to TMZ. A retrospective review of MGMT promoter status in low-grade glioma samples from patients treated with radiation therapy plus TMZ in the context of the single-arm phase II NRG/RTOG 0424 trial also noted a higher frequency of MGMT promoter methylation in IDH mutant tumors compared with IDH wildtype tumors. Though the predictive nature of MGMT methylation could not be addressed in this single-arm study, a multivariate analysis highlighted the prognostic importance of MGMT methylation in low-grade gliomas, demonstrating prolonged OS and PFS of patients with methylated tumors, independent of IDH mutation status.⁵²

MGMT promoter methylation status is currently clinically determined using quantitative methylation-specific PCR (qMSP) and pyrosequencing techniques⁵³ and in some cases using IHC.²⁸

TMZ resistance can develop following therapy, leading to a “hypermutator” phenotype. Resistance results from mutational inactivation of mismatch repair proteins such as mutS homolog (MSH) 2, MSH6, mutL homolog (MLH) 1, PMS2 (postmeiotic segregation increased 2), POLE (polymerase epsilon), and POLD1 (polymerase delta 1), leading to an accumulation of G/T mismatches in the presence of an alkylating agent and increased mutational burden at a rate of 31.0–90.9 mutations per megabase, most of which are G:C->A:T transitions.⁵⁴ The hypermutator phenotype has been implicated in the progression from low-grade to high-grade gliomas.⁵⁵ In a sequencing study of 23 low-grade gliomas at diagnosis and recurrence, the hypermutator phenotype was found in 6 out of 10 tumors treated with TMZ, all of which had progressed to GBM and had acquired genetic changes in signaling pathways characteristic of this tumor.⁵⁵ The true risk of TMZ-induced hypermutation in glioma progression and its prognostic significance in high-grade gliomas has yet to be determined in larger-scale studies.⁵⁶ The hypermutator phenotype may also have important implications as a biomarker and predictor of response to therapy, as there have been several case reports of hypermutant tumors exhibiting durable responses to checkpoint blockade agents.^57,58

G-CIMP DNA Methylation

Global changes in DNA methylation frequently occur in cancer as ways of regulating transcription of oncogenes and tumor suppressor genes. CpG islands are regions of the genome that are high in guanine-cytosine content and commonly occur in promoters. These regions can be transcriptionally silenced by methylation, which either blocks access to transcription factors or recruits methyl-binding proteins that initiate structural chromatin changes.⁵⁹ The CpG island methylator phenotype (CIMP) was first described in 1999 in colonic tumors as a state of global hypermethylation.⁶⁰ This was differentiated from age-related methylation and was thought to lead to transcriptional repression of tumor suppressors such as p16 and mismatch repair deficiency through inactivation of MLH proteins. A similar phenotype was described in a subset of glioblastomas.⁶¹ This phenotype is strongly associated with IDH mutations and is frequently found in recurrent tumors. In fact, when introduced into primary astrocytes, the IDH mutation was found to be sufficient to cause hypermethylation and the glioma (G)-CIMP phenotype.¹¹

Tumors harboring the G-CIMP phenotype are known to have a favorable prognosis. It is unclear whether this is due to silencing of specific genes induced by methylation or is related to the presence of the IDH mutation. Despite the strong correlation between IDH mutation and G-CIMP, a subgroup of IDH mutant gliomas with a G-CIMP low phenotype was recently discovered.⁶² In this study, IDH mutant gliomas were divided into G-CIMP high, G-CIMP low, and 1p/19q codeleted tumors. Among these subtypes, G-CIMP high tumors had the best prognosis, comparable to the codeleted groups (median OS 7.2 years and 7.9 years, respectively), while G-CIMP low tumors had a significantly worse outcome, with a median OS of 2.7 years.⁶² G-CIMP low gliomas may arise from G-CIMP high ones. Intratumoral heterogeneity was found to be a poor prognostic factor in G-CIMP high gliomas. Nine patients with initially G-CIMP high tumors at diagnosis exhibited G-CIMP low recurrences that were all grade IV and portrayed epigenetic changes that resembled IDH-wildtype primary GBMs.⁶³ While testing for the G-CIMP phenotype is not routinely performed in the clinical setting, assessing genome-wide methylation of tumor is becoming increasingly common and should become routine in clinical practice.⁶⁴

TERT Promoter Mutations

Telomerase reverse transcriptase is the catalytic component of telomerase, which allows for the elongation and maintenance of telomeres at chromosome ends. While telomeres normally shorten with every cell division and allow for a defined lifespan length of any particular cell, cancer cells exhibit aberrant activation of telomerase, which allows for unlimited proliferative capacity. The most frequent mechanism of telomerase activation is through mutations within the promoter of the TERT gene, which were found to be the third most common genetic alteration in cancer after mutations in KRAS and TP53.⁶⁵ These mutations are thought to lead to telomerase reactivation by creating a novel binding site for a transcription factor of the ETS family; however the full transcriptional regulation at the mutant locus has yet to be elucidated.⁶⁶ These mutations were first discovered in melanomas⁶⁵ but were later found in a large number of other tumors, including non–small cell lung cancer,⁶⁷ bladder cancer,⁶⁸ hepatocellular carcinomas,⁶⁹ and glioblastomas.⁷⁰

TERT promoter mutations are found in approximately 80% of IDH wildtype GBM,^71,72 as well as in the majority of IDH mutant, 1p/19q codeleted oligodendrogliomas.^34,71 Recent phylogenetic analysis of pre- and posttreatment GBMs suggests that TERT promoter mutations are an early event in gliomagenesis.⁷³ In GBM, TERT promoter mutations have been associated with worse prognosis compared with that of patients with IDH wildtype GBM in a number of studies.^34,70,74,75 A recent study of 1087 glioma samples subdivided tumors into molecular groups based on 3 genetic alterations: TERT promoter mutations, IDH mutations, and 1p/19q codeletion.³⁴ Grade II and III gliomas with TERT promoter mutations alone harbored the worse prognosis (HR 11.74, 95% CI 6.15–22.41, compared to tumors with all 3 alterations). A similar effect was also seen in GBMs compared to IDH mutant tumors but was not significant on multivariate analysis.³⁴ The impact of the mutation on prognosis may be influenced by a common polymorphism rs2853669, age at diagnosis, and extent of resection.^74,75 There is currently no standard role for detection of TERT promoter mutations in the diagnosis of GBM, but this alteration is frequently included on many tumor sequencing panels. Additionally, a rapid genotyping assay was recently developed to genotype tumors for IDH and TERT promoter mutations intraoperatively.⁷⁶

In gliomas, TERT promoter mutations are mutually exclusive with mutations in the alpha thalassemia/mental retardation syndrome X-linked (ATRX) gene. This gene codes for a telomere binding protein and confers an alternative lengthening of telomeres phenotype,⁷⁷ characterized by long telomeres which are maintained in a telomerase-independent manner. This suggests that telomere regulation is an important process in the development of gliomas.

ATRX mutations were first identified in 31% of pediatric glioma patients and found to co-occur with histone H3 mutations.⁷⁸ They were then identified in grade II and III gliomas and found to co-occur with p53 mutations in this population.⁷⁹ A recent study subdivided 671 grade II and III gliomas into 3 genomically distinct types: type I tumors were characterized by the presence of IDH mutations and 1p/19q codeletion, type II tumors had ATRX and TP53 mutations, and type III were the remaining samples. In this study, type II tumors had an intermediate prognosis with HR of 2.06 compared with type I tumors (HR was 3.40 for type III tumors).⁷⁹ While ATRX mutation status appeared to influence survival in univariate analysis, this effect was not seen in multivariate analysis in this study. Therefore, the full prognostic effect of ATRX mutations in gliomas has yet to be determined.

EGFR Alterations

Amplification of EGFR is considered a hallmark alteration of GBM and is observed in ~40–50% of primary (IDH wildtype) GBM.⁸⁰ About one half of EGFR-amplified GBM express a constitutively activated variant of EGFR known as EGFRvIII, which further dysregulates the EGFR pathway. This variant contains a deletion in exons 2–7 of the EGFR gene, which leads to expression of a constitutively active protein.⁸¹ A recent meta-analysis of 10 articles involving 1074 patients demonstrated that high EGFR expression is associated with poor prognosis in GBM patients, with the pooled HR for OS of 1.57 (95% CI: 1.15–2.14, P = 0.004).⁸² The presence of the EGFRvIII variant was not found to alter prognosis in patients with EGFR-amplified tumors.⁸¹

Despite the high frequency of EGFR alterations in gliomas, these tumors show little clinical response to tyrosine kinase inhibitors that have proven effective in other cancers.⁸³ This is likely due to the fact that EGFR alterations display a significant amount of intratumoral heterogeneity in GBM.⁷³ Single cell sequencing technology allowed us to understand that different EGFR alterations, such as the vIII variant and carboxy-terminal deletions, were found in different subclonal populations within the same tumor.⁸⁴ Of note, while EGFR amplification status of the tumor tends to remain stable over time, expression of EGFRvIII can change. In 15 primary and recurrent tumor pairs that exhibited EGFRvIII expression within the primary specimen, only 8 (53%) of paired tumors continued to maintain EGFRvIII expression at recurrence.⁸⁵

Recently, novel EGFR-targeted therapeutic agents have started to show promise. The monoclonal antibody–drug conjugate depatuxizumab mafodotin (depatux-m, ABT-414) completed phase I studies in patients with recurrent, EGFR-amplified GBM and in newly diagnosed GBM given with concurrent TMZ with encouraging results.^86–88 A phase III study (RTOG 3508) looking at the efficacy of adding depatuxizumab to both radiation and TMZ and Adjuvant TMZ in newly diagnosed GBMs has completed accrual. Additionally, a phase I study of 10 patients treated with chimeric antigen receptor T cells directed against the EGFRvIII protein shows that this treatment is well tolerated, and 1 patient had stable disease at 18 months post-infusion.⁸⁹ However, rindopepimut, a vaccine targeting the EGFRvIII mutation in patients with EGFRvIII-expressing, newly diagnosed GBM, failed to improve survival when added to standard therapy of radiation and TMZ in a large phase III trial.⁹⁰

BRAFV600E Mutations

BRAF encodes for the B-Raf protein kinase, which is involved in growth-promoting pathways. Mutation of the BRAF gene at the V600E hotspot results in constitutive activation of B-Raf and has been detected in a number of cancer types, including in glial tumors, such as pilocytic astrocytoma and glioneuronal tumors in the pediatric population as well as in diffuse gliomas, glioneuronal tumors,²⁷ and GBM in adults. Although the influence of the BRAF V600E mutation on prognosis in gliomas is not entirely clear, there is a suggestion from the literature that the significance of the alteration is dependent on the age of the patient. In a recent meta-analysis reviewing 11 articles describing ~1300 patients with gliomas, the authors found no prognostic relationship of BRAF V600E in patients over 35 years of age. In pediatric patients and younger adults (<35 y), the presence of the mutation is associated with improved survival, with pooled HR of 0.51 (95% CI: 0.34–0.79) for pediatric cases and 0.43 (95% CI: 0.20–0.93)⁹¹ for younger adults. In GBM, the presence of the BRAF V600E mutation has been associated with epithelioid GBM on histopathologic analysis,⁹² which has been reported to have more aggressive behavior and poorer prognosis.⁹³ Further investigation is required to better understand how BRAF V600E may influence outcome. Detection of the alteration has a therapeutic role, as many cancer types with BRAF V600E have been shown to respond to BRAF inhibitors.

Histone Mutations

Mutation of the histone H3 proteins is found in a subset of high-grade gliomas known as diffuse midline glioma, H3 K27M-mutant, as described in the update to the WHO classification of CNS tumors.² These tumors are generally found in the pons, thalamus, and spinal cord, in both adults and children, where they are generally associated with a poor prognosis.⁹⁴ The presence of the histone mutation is mutually exclusive with IDH mutations but can co-occur with mutations in receptor tyrosine kinase/Ras/phosphatidylinositol-3 kinase pathways.⁹⁵ The median age at diagnosis of histone-mutated tumors in adults is the early 30s,⁹⁶ compared with a median age of 64 for GBM. In pediatric patients, tumors with H3K27 mutations were found to have an overall worse prognosis independent of anatomical location. The median OS was 1.04 years for mutant tumors versus 6.1 years for wildtype ones.⁹⁴ Characterization of a small series of adult patients with histone-mutated gliomas suggests that the H3 K27M-mutation is also associated with poor prognosis in older age groups, with median OS of 19.6 months, similar to an OS of 17 months that was observed in IDH wildtype gliomas in this cohort.⁹⁶

Conclusions

The genomic analysis of adult gliomas has led to insight into the underlying pathways that lead to tumor formation. In addition, some of the molecular alterations discovered have clinical and prognostic relevance. We have discussed the key biomarkers that have emerged over the last decade and how they serve to influence prognosis both positively and negatively. Of note, the influence of many of these biomarkers on disease course is by and large related to individual effects on development and growth of these tumors. Therefore, in addition to the prognostic implications associated with the alterations described herein, these biomarkers are potential therapeutic targets that are actively under investigation.

Funding

The present work was not funded by any sources.

Conflict of interest statement. Dr. Brastianos has received Honoraria from Merck and Genentech, is a consultant for Lilly, Merck and Angiochem and has received research funding (to MGH) from Merck. Dr. Cahill is a consultant for Lilly and Merck. Dr. Santagata is a consultant for RareCyte.

Supplement sponsorship. This supplement was funded through an independent medical educational grant from AbbVie.