Neurotherapeutics (2017) 14:284–297

DOI 10.1007/s13311-017-0519-x

REVIEW

Glioma Subclassifications and Their Clinical Significance

Ricky Chen 1 & Matthew Smith-Cohn 1 & Adam L. Cohen 2 & Howard Colman 3

Published online: 9 March 2017

# The American Society for Experimental NeuroTherapeutics, Inc. 2017

Abstract The impact of targeted therapies in glioma has been the development and testing of new specific targeted therapies
modest. All the therapies that have demonstrated a significant for particular glioma subtypes. This review aims to summarize
survival benefit for gliomas in Phase III trials, including radi- the current literature regarding glioma subclassifications and
ation, chemotherapy (temozolomide and PCV [procarbazine, their clinical relevance in this evolving field.
lomustine, vincristine]), and tumor-treating fields, are based
on nonspecific targeting of proliferating cells. Recent ad- Keywords Glioma . Ependymoma . Targeted therapy . IDH
vances in the molecular understanding of gliomas suggest mutation . MGMT methylation . TERT promoter . EGFR .
some potential reasons for the failure of more targeted thera- BRAF . 1p/19q co-deletion . 2HG . MR spectroscopy .
pies in gliomas. Specifically, the histologic-based glioma clas- Vaccine therapy
sification is composed of multiple different molecular sub-
types with distinct biology, natural history, and prognosis.
As a result of these insights, the diagnosis and classification Introduction
of gliomas have recently been updated by the World Health
Organization. However, these changes and other novel obser- Gliomas are the most prevalent primary tumors of the brain and
vations regarding glioma biomarkers and subtypes highlight spinal cord. Histologically, they share characteristics of normal
several clinical challenges. First, the field is faced with the glial cells and are generally named according to these similarities.
difficulty of reinterpreting the results of prior studies and ret- However, whether gliomas originate from normal glial cells, glial
rospective data using the new classifications to clarify prog- or neural precursors, stem cells, or other cell types remains a topic
nostic assessments and treatment recommendations for pa- of investigation [1]. Historically, gliomas have been diagnosed
tients. Second, the new classifications and insights require and classified based on histopathology. In the 2007 World
rethinking the design and stratification of future clinical trials. Health Organization (WHO) classification, the main glial tumor
Last, these observations provide the essential framework for groups included astrocytic tumors, oligodendroglial tumors,
oligoastrocytic tumors, ependymal tumors, and neuronal and
mixed neuronal-glial tumors (such as gangliogliomas) [2].
These groups included more circumscribed grade I tumors such
as pilocytic astrocytomas, pleomorphic xanthoastrocytomas, and
* Howard Colman
subependymal giant cell astrocytomas, as well as the more com-
[email protected] mon infiltrating gliomas, including grade II oligodendrogliomas
and astrocytomas, and grade III anaplastic oligodendrogliomas,
1
Department of Neurology, Clinical Neurosciences Center, University anaplastic astrocytomas, anaplastic oligoastrocytomas, anaplastic
of Utah, Salt Lake City, UT, USA ependymomas, and grade IV glioblastomas (GBM). The last sev-
2
Division of Oncology, Department of Internal Medicine, Huntsman eral decades of dedicated research into the biology of gliomas has
Cancer Institute, University of Utah, Salt Lake City, UT, USA resulted in a rapidly accelerating process of discovery that has
3
Department of Neurosurgery, Huntsman Cancer Institute and Clinical uncovered some of the key genetic and molecular underpinnings
Neuroscience Center, University of Utah, Salt Lake City, UT, USA of these tumors. These observations have contributed to a new
Glioma Subclassifications and Their Clinical Significance 285

understanding of glioma biology that has altered the current clas- of this difference, including age, patient performance status,
sification and provided new insights into tumor initiation, ontog- and extent of resection. However, even when corrected for
eny, and tumor progression. Later, we will explore how these clinical and histologic prognostic factors, we are unable to
findings have had a significant impact on the diagnosis and man- account for significant proportions of the differences in sur-
agement of many different subtypes of gliomas. These findings vival. These observations suggest that significant differences
are also opening up fertile avenues of investigation into novel in underlying tumor biology within diagnoses were not
therapeutic targets and approaches for specific subtypes of glio- accounted for in this classification. Recently, more detailed
mas. In this review, we will first describe the rationale for molec- understanding of molecular alterations within gliomas have
ular classification, then we will discuss the classification of the allowed for a refinement of diagnostic criteria, prognostic bio-
most common gliomas (astrocytomas and oligodendrogliomas), markers, and the beginnings of our ability to utilize effective
including rare subtypes of GBMs, next we will discuss the clas- targeted therapies in molecularly defined glioma subtypes.
sification of rare types of gliomas, and we will conclude with a
discussion of possible therapeutic implications of molecular MGMT Promoter Methylation
classification.
An example of a molecular biomarker of glioma subtypes that
has prognostic, predictive, and clinical application is MGMT
Limitations of Histologic Classification of Gliomas (O6-methylguanine-DNA methyltransferase) promoter methyl-
ation. MGMT promoter methylation was initially identified as a
Until recently, the standard criterion used by pathologists and prognostic and predictive marker within the diagnosis of GBM
neuropathologists for the diagnosis and grading of gliomas in patients treated with temozolomide [6]. Temozolomide meth-
were defined in the 2007 version of the WHO classification ylates purine bases, including the O6 and N7 sites on guanine
[2]. In this classification, histologic diagnosis (designated by and N3 site on adenine. O6-Methylguanine induction is believed
the aforementioned subtypes) and grading based on degrees of to be the primary cytotoxic event, leading to the insertion of
malignancy were the Bgold standard^ for diagnosis and treat- thymine instead of cytosine during cellular replication, causing
ment. Tumors are graded ranging from WHO grade I to grade double-strand breaks, failure of DNA replication, and ultimately
IV, generally based on characteristics of increasing malignan- cell death [7]. As a mismatch repair enzyme, MGMT removes
cy including the presence and degree of atypia and mitotic the O6-methylguanine adducts induced by alkylator chemother-
activity, as well as specific hallmarks for some subtypes such apies, thereby abrogating their cytotoxic effects. Repair of DNA
as microvascular proliferation and/or pseudopalisading necro- depletes the MGMT protein, which the cells must replenish.
sis in the case of GBM, which is classified as grade IV [3]. Thus, higher levels of MGMT are thought to lead to temozolo-
Although this histologic-based classification system has mide resistance [7].
evolved from prior versions over the years and has served cli- Methylation of the MGMT promoter in the CpG-rich region
nicians well, there are some limitations that have, in part, led to results in epigenetic silencing (decreased expression) of the
the recent revisions. One problem of the histologic-based diag- MGMT protein. MGMT promoter methylation is seen in ap-
nosis system is that it is subject to significant interobserver proximately 40% of all GBMs [8] and can be measured
variability. Although studies of this question have reached dif- through various methodologies, such as methylation microar-
ferent conclusions, some have shown that concordance among a ray or bisulfate sequencing. Higher levels of MGMT promoter
group of individual neuropathologists reviewing a case can be methylation predict longer survival in GBM, particularly
as low as 52%, with particular disagreements related to differ- when temozolomide is part of the upfront therapy. In the orig-
ences in classification of astrocytic versus oligodendroglioma inal publication from Stupp et al. in 2005 [9], among people
versus oligoastrocytoma tumors, as well as interobserver differ- receiving radiation and temozolomide, MGMT promoter
ences regarding differentiation of grade II from grade III tumors methylation was associated with improved median survival
[4]. Thus, the diagnostic criteria differentiating astrocytoma of 21.7 months versus 12.7 months for patients with
from oligodendroglioma are best suited for prototypic cases unmethylated tumors. Long-term follow-up from that initial
but may be be too practically imprecise for most tumors, which study has substantiated this finding, with survival of 48.9% at
have a degree of mixed features [5]. Accurate classification by 2 years, 27.6% at 3 years, and 13.8% at 5 years for patients
histology can be further complicated by insufficient or nonrep- with temozolomide-treated GBMs with MGMT methylation
resentative tissue sampling [4]. versus 14.8%, 11.1%, and 8.3%, respectively, for patients with
Beyond the diagnostic challenges, the traditional classifi- unmethylated MGMT [9]. However, while MGMT promoter
cation and grading schemes historically lacked precision in methylation is useful as a prognostic and predictive marker, it
prognosis even for patients with the same diagnosis (e.g., does not appear to define distinct diagnostic subtypes of glio-
GBM), where survival may vary from weeks or months to mas per se. As described below, the genetic alterations in
multiple years. A number of clinical factors can explain some gliomas appear to be more suitable diagnostic classifiers, with
286 Chen et al.

MGMT promoter methylation occurring in all of these genetic occur at the substate recognition site and significantly alter
subtypes at varying rates. the enzyme active sites of the enzyme in gliomas, leading to
a neomorphic change in IDH function. Mutant IDH genes
Reclassification of Lower-Grade Gliomas result in proteins that convert α-ketoglutarate to the putative
oncometabolite 2-hydroxyglutarate (2HG) [15]. Not only
As the limitations of the histopathological classifications for have IDH mutations been shown to be one of the earliest
gliomas have become clear in view of accumulating molecu- events in glioma formation, but the resulting production of
lar, genomic, and epigenetic data, our traditional diagnostic 2HG appears to drive extensive epigenetic changes that alter
criteria are being reconfigured, with significant implications cellular differentiation and could contribute to oncogenesis.
for diagnosis and management, as seen in Fig. 1. However, IDH mutation alone was not able to transform nor-
The most recent reclassification of lower-grade gliomas mal or immortalized astrocytes in culture, suggesting there are
(LGGs) is based on work by Cairncross et al. [11] that was likely other contributing mechanisms [16, 17].
initially published starting in the late 1990s with the identifi- The majority of LGGs (65–90%) have mutations in IDH
cation of chromosome 1p/19q loss and accelerated with more [18]. Independent of grade, the presence of IDH mutations in
recent identification of IDH mutation, ATRX mutation, and gliomas confer significantly better progression-free survival
other alterations in these tumors and GBM [12, 13]. than in their IDH wild-type counterparts, irrespective of treat-
ment received [19]. The vast majority of primary GBMs ob-
Significance of IDH Mutations in Gliomas served in older patients without any lower-grade precursor are
IDH wild-type, while virtually all secondary GBMs that arise
The isocitrate dehydrogenase (IDH) family of enzymes cata- from LGGs demonstrate IDH mutation. This finding and the
lyzes the conversion of isocitate to α-ketoglutarate while ability to examine prognosis based on IDH mutation status has
converting nicotinamide adenine dinucleotide phosphate led to the surprising observation that many LGGs that are IDH
(NADP+) to reduced NADP+ (NADPH), both as part of the wild-type can be as aggressive as GBM and have prognoses
Kreb cycle and in the cytoplasm. Somatic mutations in genes that are quite similar to histologic grade IV tumors [20]. Some
encoding 2 of the isoforms of isocitrate dehydrogenase (IDH1 authors refer to these IDH wild-type LGGs as pre-GBM or
and IDH2) are present in a variety of cancers [14]. In gliomas, GBM-like. Conversely, GBMs that are IDH mutant generally
they were initially identified in a high percentage of LGGs and have a significantly better prognosis, not only compared with
secondary GBMs and a lower percentage of primary GBMs IDH wild-type GBM, but also relative to IDH wild-type lower
[12]. The vast majority of mutations are caused by an amino- grade (II–III) tumors. Detailed sequencing of individual tu-
acid substitution at position 132 in IDH1 from arginine to mors and matched initial and recurrent tumors suggests that
histidine (R132H) and substitution at position 172 in IDH2 IDH mutation is a relatively early event (perhaps the initiating
from arginine to lysine (R172K), although other substitutions event) in these tumors [21]. IDH mutation is thought to be
at these or nearby sites are seen. All pathogenic mutations followed by other mutations such as TP53 and ATRX in

Fig. 1 A simplified algorithm for

classification of the diffuse
gliomas based on histological and
genetic features (see text and [10]
for details). A caveat to this
diagram is that the diagnostic
Bflow^ does not necessarily
always proceed from histology
first to molecular genetic features
next, as molecular signatures can
sometimes outweigh histological
characteristics in achieving an
Bintegrated^ diagnosis. A similar
algorithm can be followed for
anaplastic-level diffuse gliomas.
*Characteristic but not required
for diagnosis. Reprinted with
permission from the World Health
Organization [10]
Glioma Subclassifications and Their Clinical Significance 287

astrocytomas and co-deletion of 1p/19q (potentially along 2 trials establish chemotherapy with radiation as the standard
with CIC and FUBP1 mutations) in oligodendrogliomas of care for anaplastic gliomas with 1p/19q co-deletion.
[22]. IDH mutant LGGs may eventually undergo malignant Although not part of the prior (2007) WHO classification
transformation when tertiary alterations such as KRAS, system, these data were strong enough that many clinicians
PIK3CA, PDGFRA, MET, PTEN, and N-Myc are acquired and neuropathologists have been using 1p/19q status
[22, 23]. Interestingly, a 2012 study by Turcan et al. [15] Binformally^ in the diagnosis and management of gliomas
found that IDH1 mutations in glioma lead to genome-wide for some time. Specifically, tumors with a mixed
increases in DNA methylation. This hypermethylator pheno- oligoastrocytoma histology were diagnosed and treated either
type is associated with proneural gliomas and improved sur- as astrocytoma if 1p/19q intact or oligodendroglioma if 1p/
vival. A very small group of IDH-mutated tumors lack DNA 19q co-deleted. As we will explore later, this approach is now
hypermethylation and have poor survival. formalized in the new WHO classification system, and the
The mechanism of tumor initiation by IDH mutation is prior mixed oligoastrocytoma diagnosis is only used when
controversial, but IDH mutation has several consequences that molecular classifiers are not available.
may contribute to tumorigenesis. While the specific mecha-
nism of IDH mutation (and potentially 2HG) that results in the
oncogenic switch in gliomas remains unknown, potential Comprehensive Molecular Profiling of Infiltrative
mechanisms have been identified, including inhibition of Gliomas
hypoxia-related proline hydroxylases, inhibition of DNA
demethylases, inhibition of histone demethylases, and alter- Although the data behind molecular classification of LGGs
ations in glumatate metabolism. Meanwhile, reduction of the has been building over many years and centered around iden-
mitochondrial NADPH pool can reduce sensitivity of cells to tification of new single prognostic or classification markers, 2
redox associated apoptosis [24] and may also contribute to recent landmark publications have combined these observa-
tumor progression while increasing sensitivity to oxidative tions into a more comprehensive view of glioma molecular
damage from radiation treatment. Further work is needed to genetics. These findings have resulted in the recent revision of
elucidate the specific role of IDH mutation and the patholog- the official WHO classification of gliomas and other tumors.
ical consequences that clearly impact tumor evolution and The Cancer Genome Atlas Network (TCGA) effort used
prognosis. retrospective tumor samples from multiple centers in the US
and utilized multiple molecular platforms to perform a com-
prehensive molecular characterization of both GBM and
Chromosome 1p/19q Deletion LGGs. The technologies used for the analysis included next-
generation sequencing, gene expression by microarray and
As previously mentioned, chromosome 1p/19q co-deletion RNA sequencing, whole-genome methylation array, and re-
has been known as a diagnostic and prognostic marker of verse lysate protein arrays. These studies found that it is pos-
oligodendrogliomas since 1998 [11]. More recent studies de- sible to subclassify LGGs into 3 main molecular groups
termined that the co-deletion of these 2 chromosomal arms is (Table 1) [13]. The first group are IDH1/IDH2 mutated,
due to a balanced translocation between chromosomes 1 and DNA hypermethylated, and chromosome 1p/19q co-deleted
19 and subsequent loss of other chromosomal arms [25]. This tumors, which carry other molecular markers that correspond
mechanism explains why gliomas with either 1p or 19q dele- to molecular oligodendrogliomas. These tumors generally
tion alone (nonconcurrent deletion) do not share the survival have a good prognosis (median survival ~7 years in the
benefit of co-deleted tumors [26]. TCGA analysis), regardless of whether they are histologically
The initial retrospective series and subsequent retrospective grade II or III. A second group of IDH1/IDH2 mutated and
analyses of large randomized trials have validated 1p/19q de- DNA hypermethylated tumors have intact 1p/19q, and these
letion as a strong prognostic and predictive marker in grade II instead carry alterations in ATRX and TP53, corresponding to
and III gliomas. In RTOG 9402, the presence of 1p/19q co- molecular astrocytomas. Their prognosis is intermediate be-
deletion in anaplastic oligodendrogliomas and tween the molecular oligodendrogliomas and the molecular
oligoastrocytomas was associated with longer overall survival GBMs (median survival ~5 years). The last group are gliomas
when treated with radiation alone (7.3 vs 2.7 years with intact that are IDH1/IDH2 wild-type, which all have intact chromo-
1p/19q). Addition of PCV (procarbazine, lomustine, vincris- some 1p/19q. These tumors have poor prognosis, regardless of
tine) chemotherapy led to a doubling of overall survival in co- grade. Thus, although some of these tumors demonstrate low-
deleted tumors to 14.7 years but did not change the survival er histologic grade, they behave clinically as Bpre-GBMs,^
for 1p/19q-intact tumors [27]. The EORTC 26951 from with a prognosis of only 1.7 years median survival, which is
Europe presented similar chemosensitivity and favorable just fractionally longer than tumors with similar molecular
prognosis with 1p/19q co-deleted tumors [28]. Together, these features and a histologic grade IV [29].
288 Chen et al.

Table 1 The Cancer Genome

Atlas Network classification of Best prognosis Intermediate prognosis Worse prognosis
low-grade glioma
IDH status Mutated Mutated Wild type
DNA methylation Hypermethylated Hypermethylated NA
1p/19q deletion status Deleted Intact Intact
Other mutations NA ATRX, TP53 NA
Grade Grade 2/3 Grade 2/3 Grade 2/3

NA = not applicable

TERT Promoter Mutations question, and the biological mechanism of this association
remains unknown.
Maintenance of telomeres is essential for cancer cells to avoid
senescence and maintain proliferative potential. One mecha- Histone-Mutant Gliomas
nism for maintaining telomeres is to increase expression of the
telomerase reverse transcriptase, encoded by TERT. Mutations Sequencing of pediatric and young adult gliomas revealed ad-
in the promoter region of TERT increase expression and serve ditional subtypes unique to this patient population. A subset of
as an important biomarker of glioma subtypes. Large-scale pediatric tumors harbor mutations in the histone 3.1 and 3.3
analysis of over 1000 cases from databases derived from pa- proteins, encoded by the HIST1H3B and H3F3A genes, respec-
tients from University of California, San Francisco; the Mayo tively [33, 34]. These proteins play an important role in com-
Clinic; and the TCGA identified 5 principal groups based on 3 plex post-translational epigenetic expression by mediating
molecular markers: IDH mutation, 1p/19q co-deletion, and changes in the heterochromatin structure of DNA and directing
TERT promoter mutation [30]. Most GBMs and GBM-like the interactions of transcriptional activators and repressors, as
LGGs had TERT mutation only with wild-type IDH and 1p/ well as regulation of telomeres. H3F3A mutations leading to a
19q intact. Among LGGs, tumors that carried all 3 alterations K27M substitution occur in 78% of diffuse intrinsic pontine
(triple-positive) typically showed oligodendroglioma histolo- gliomas (DIPGs) and 22% of GBMs outside of the brainstem
gy and favorable prognosis. LGGs with only IDH mutation [34]. DIPGs without H3F3A mutations carry HIST1H3B mu-
were consistent with molecular astrocytomas and had inter- tations also leading to a K27M substitution. Mutations of
mediate prognosis but still had medial survival of many years. H3F3A occur predominantly in childhood gliomas (age 5–23
Intriguingly, the worst prognosis among LGGs is seen in those years) and present as midline gliomas involving thalamus,
with TERT mutation only. Based on these findings, the pres- pons, and spinal cord, particularly DIPGs. A separate amino-
ence of an isolated TERT mutation in a LGG may be an indi- acid substitution of glycine 34 to arginine or valine (G34R/V)
cator that the clinical behavior of a tumor may be more akin to in H3.3 appear mostly in pericallosal GBMs in adolescent and
a grade IV tumor [30]. young adults aged 9–42 years [35]. Studies showing overlap-
Recent evidence suggests that TERT has a dichotomous ping mutations with key established alterations in genes such as
prognostic effect that may somehow enhance sensitivity to TP53, ATRX, and DAXX suggest that these hotspot mutations
temozolomide and improve survival in the context of could be linked to the activation of distinct programs of gene
MGMT methylation, while conversely promoting a more re- expression integral to the pathogenesis of these tumors [33].
sistant and prognostically worse phenotype in the context of K27M mutations of H3F3A in DIPGs have significant prog-
an unmethylated MGMT. In a recent analysis from Japan [31] nostic implications and are associated with worse survival than
comparing patient outcomes and molecular subtypes, TERT wild-type DIPGs, which by necessity carry HIST1H3B muta-
promoter mutation was a poor prognostic factor only in tions [36]. Promising in vitro studies indicate these mutations
MGMT unmethylated GBM, and MGMT methylation had a may be future targets for both peptide-based vaccines and chi-
larger effect on prognosis in those tumors with TERT meric antigen receptor T cells [37, 38].
promotor mutation than in TERT wild-type tumor. A separate
analysis of a cohort of patients from University of California, The New WHO Classification of Infiltrating Gliomas
Los Angeles, showed TERT promoter mutation as a prognos-
tic factor only in tumors with MGMT methylation. Similar to These developments regarding the biology and classification
the Japanese cohort, MGMT methylation had a large prognos- of gliomas are so significant that the WHO elected to issue an
tic effect on tumors with TERT promoter mutation and mini- updated WHO Classification of Tumors of the Nervous
mal prognostic effect on wild-type tumors [32]. Because of the System in 2016, which was earlier than the planned update
correlative nature of these studies, the direction of the causal [10]. One of the key changes in this update is the inclusion of
relationship between TERT and MGMT remains an open molecular diagnostic criteria for the classification of
Glioma Subclassifications and Their Clinical Significance 289

infiltrating gliomas. Instead of the histologic subtypes on the prognoses. Of 183 GBM samples examined, 31% were
old classification (astrocytoma, oligodendroglioma, and proneural, 20% were proliferative, and 49% were mesenchy-
oligoastrocytoma), the new criteria incorporate testing for mal. Interestingly, nearly all WHO grade III tumor specimens
IDH mutation, chromosome 1p/19q deletion, and histone mu- examined (65/73; 89%) were classified as proneural. When
tations. Thus, astrocytomas and GBM are divided into IDH initial and recurrent tumors were examined from the same
mutant and IDH wild-type; oligodendrogliomas are defined as patient, recurrent tumors had a tendency to shift towards the
IDH mutant and 1p/19q co-deleted; and diffuse midline glio- mesenchymal subclass; 8 of 26 pairs of tumors (initial and
mas are defined by histone H3 K27M mutations. If molecular reccurrence) transitioned to mesenchymal, while the remain-
testing is not available, tumors are classified similarly to the der remained the same subclass. The precise etiology of this
prior 2007 classification, based on histology (astrocytoma, transition is unclear, but studies have found that radiation can
oligodendroglioma, and oligoastrocytoma), but the designation convert proneural to mesenchymal subtype gliomas, a process
BNOS^ is added to indicate that the molecular criteria are thought to be mediated by tumor necrosis factor-α. This tran-
missing from the diagnosis. The differences between the sition results in radioresistance of glioma cells, perhaps in a
2007 and 2016 classification systems are summarized in nuclear factor κ-light-chain-enhancer of activated B cells (NF-
Fig. 2. kB)-dependent manner [40].
In the 2010 TCGA classification of GBM [41], 4 gene
Mutational, Gene Expression, and Methylation Subtypes expression subtypes were identified, based on multidimen-
sional gene expression and genomic clustering. The gene ex-
Beyond genetic and epigenetic alterations, gene-expression pression subtypes were called classic, mesenchymal,
studies in GBM have identified several key subtypes, some proneural, and neural. In the classical subtype, 97% of glio-
of which are highly correlated with specific genetic changes. mas harbored high-level EGFR amplification, with EGFR
In 2006, Phillips et al. [39] published their findings of poten- mutation seen in over one-third of cases. The mesenchymal
tially 3 divisions within gliomas based on gene expression subtype was characterized by higher frequency of NF1 muta-
microarray: proneural, proliferative, and mesenchymal/angio- tions and was associated with stimulation of genes upregulat-
genic. The mesenchymal/angiogenic phenotype had the worst ing the tumor necrosis factor superfamily and NF-κB

Fig. 2 Differences between the

2007 and 2016 World Health
Organization classifification
systems. Reprinted with
permission from Elsevier [5]
290 Chen et al.

pathways. The proneural subtype was strongly associated with roughly 2% of all grade IV gliomas and are more prone to
PDGFRA amplifcations and IDH1 mutations, which induce extracranial metastasis than other gliomas [46, 47]. A series of
the global hypermethylation phenotype called glioma-CpG 19 gliosarcomas showed identical mutations in the glial and
island methylator phenotype (G-CIMP) [42]. sarcomatous components, supporting a monoclonal origin for
DNA methylation can also be used as a classifier of mo- both histologic components [48]. The mutations seen are typ-
lecular subtypes in GBM and LGGs, and may be useful as a ical of GBM, including TP53 mutations, PTEN mutations/loss
more general classifier than genetic or gene expression alter- of heterozygosity, CDKN2A deletions, and TERT promoter mu-
ations. In 2012, Sturm et al. [35] identified 6 main biological tations. Defining molecular alterations or targetable molecular
subgroups across pediatric and adult GBM, based on global alterations common in gliosarcoma have not been identified.
DNA methylation clustering, and these methylation clusters Giant cell GBMs constitute 1% to 5% of all GBMs, and are
demonstrated overlap with some of the previously described characterized by a predominance of bizarre, multinucleated gi-
genetic and gene expression subtypes. The main methylation ant cells [2, 49]. The median age at diagnosis is about 44 years
subtypes were IDH (associated with IDH mutation); K27 (as- [50]. TP53 mutations are common (78–90%), while epidermal
sociated with histone 3.3 K27 mutation); G34 (associated with growth factor receptor (EGFR) amplification is rare. Alterations
histone 3.3 G34 mutation); RTK 1 BPDGFRA^ (associated of PTEN can occur via both mutation (33%) or loss of hetero-
with increased frequency of PDGFRA amplification); mesen- zygosity at 10q (50%) [50]. However, despite the young age of
chymal; and RTK II Bclassic^ (associated with higher frequen- diagnosis and high rate of TP53 mutations, giant cell GBMs are
cy of EGFR amplification and chromosome 7 gain, and chro- distinct from typical secondary GBMs based on a short length of
mosome 10 loss). More recently, investigators from the prediagnosis symptoms and lack of IDH and ATRX mutations.
TCGA effort have published a combined multiplatform anal- In a recent study by Erson-Omay et al. [51], 6 of 720 exome-
ysis of GBM along with grade II and III gliomas. This analysis sequenced glioma samples harbored somatic mutations in the
identified the mesenchymal and classical subtypes of GBM, exonuclease domain of the polymerase epsilon gene (POLE).
as well as the subtypes of LGG described above. The previ- This mutation was found to cause ultramutated giant cell high-
ously identified proneural group clustered with IDH mutated, grade glioma subtype with a better prognosis (26.93 vs
1p/18q non-codeleted LGGs [43]. The previously described 6.93 months of progression-free survival). Histologically,
neural subgroup was not identified. Given that neither the POLE mutant tumors harbored multinucleated giant or bizarre
Phillips group nor the Sturm analysis identified a neural-like cells and had a tendency to be infiltrated by immune cells.
subgroup, the existence of this subgroup is controversial. Similarly, a study looking at colorectal cancers with this muta-
One concern with gene expression and gene methylation- tion [52] also found better prognosis and found increased infil-
based subtyping is heterogeneity over time and space. As a tration of immune cells predominantly comprised of CD8+ lym-
given biopsy, and even a good resection specimen, likely re- phocyte effector cytokines, indicating the mutation was associ-
flects only a geographic part of a diffusely infiltrating glioma, ated with increased immunogenicity. Studies have suggested
neither the histological nor the molecular approaches are ca- that tumors with mismatch repair deficiency, which by exten-
pable of dealing with how heterogeneous gliomas truly are. sion include hypermutator phenotype gliomas such as those
For example, recent studies have found that different parts of with POLE mutations, have a marked increase in the number
GBM tumor can have different transcriptional subtypes [44]. of mutation-associated neoantigens, thereby producing more
Moreover, by analyzing gene expression patterns of paired potential targets of endogenous T-cell responses. Furthermore,
primary and recurrent GBMs Wang et al. [45] found that immune checkpoint inhibitors such as pembrolizumab may
two-thirds switched subtype at recurrence. These results raise have increased efficiency in these tumors [53].
questions about the clinical utility of transcriptional subtypes One of the newest additions to the updated 2016 WHO
at our current level of knowledge about the information they tumor classification guidelines, epithelioid GBMs are distin-
provide. guished histologically by epithelioid cells with abundant cy-
toplasm, prominent nucleoli, and rhabdoid cells [10, 54]. They
Rare Glioma Subtypes typically occur in the first 3 decades of life [55], with 1 study
[56] comprised of 6 patients finding a median age at diagnosis
There are several other known subtypes of primary GBM that of 7.6 years and a median survival of 169 days. Although the
have been identified based on histology, including giant cell rhabdoid appearance can lead to confusion with atypical
GBM, gliosarcoma, and the most recently identified variant teratoid rhaboid tumor, epithelioid GBMs maintain expression
epithelioid GBM [10]. Like the majority of other primary of INI1 [57]. Unique to this subgroup are BRAFV600E muta-
GBMs, these subtypes lack IDH mutations. tions (50%) and frequent hemizygous deletions of ODZ3 [58].
Gliosarcoma is a rare form of high-grade glioma that has There have also been case reports with TERT promotor muta-
both sarcomatous and malignant glial components by histolo- tions [59]. Also suggesting a unique molecular basis for
gy and immunohistochemistry. Gliosarcoma comprises epithelioid GBM, this subclass of GBM does not harbor
Glioma Subclassifications and Their Clinical Significance 291

alterations typically seen in primary GBMs such as EGFR Supratentorial location and high grade are independent predic-
amplification and chromosome 10 loss [10]. tors of poor progression-free survival [71].
Other glioma subtypes are also characterized by high rates Histologically, ependymoma grades are determined based
of BRAFV600E mutations. Pleomorphic xanthoastrocytoma on features similar to other malignancies, including mitotic
(PXA), like epithelioid GBM, occurs mainly in children and figures, cellular pleomorphism, and invasion into tissue. The
adults under the age of 30 years [60]. Histologically, PXA is gra de I g roup con sists of su bep endy momas a nd
characterized by pleomorphic giant cells, xanthomatous cyto- myxopapillary ependymomas, grade II of ependymomas,
plasm, and infiltration with reticulin and lymphocytes. Up to and grade III of anaplastic ependymomas. Grade III
two-thirds of PXAs carry BRAFV600E mutations [61, 62]. The ependymomas are associated with higher risk of drop metas-
presence of a BRAFV600E mutation may help distinguish a tasis [66]. The new 2016 WHO classification of tumors of the
PXA from a giant cell GBM, which is important given the central nervous system maintains the majority of the prior
m a r k e d p r o g n o s t i c d i ff e r e n c e b e t w e e n t h e t w o . histologic ependymoma diagnoses, but recognizes a new ge-
Gangliogliomas are mixed glioneuronal tumors and are the netically defined ependymoma subtype known as RELA
most common neoplasm causing chronic focal epilepsy in fusion-positive [10].
young patients [63]. BRAFV600E mutation can be seen in In 2015, Pajtler et al. [72] used 500 ependymomas to clas-
20% to 50% of gangliogliomas of the brain [62, 64]. sify molecularly ependymal tumors. Excluding myxopapillary
Pilocytic astrocytomas make up 20% of pediatric brain tumors ependymomas and subependymomas, they found 5 molecular
and appear to have about a 9% rate of BRAFV600E mutation groups of ependymomas, with 1 subgroup in the spine and 2
[65]. This mutation was found to be associated with each in the supratentorial brain and posterior fossa (Fig. 3).
extracerebellar location. The therapeutic implications of Spine ependymomas form a distinct clinical group with an
BRAF mutations are discussed at the end of this review. age distribution higher than other ependymomas [73]. Most
spine WHO grade II or III ependymoma have chromosomal
instability and mutations in NF2. This finding is consistent
Ependymomas with prior work showing germline and sporatic mutations of
NF2 are associated with spinal ependymomas [73].
Ependymomas make up 3% of all primary central nervous sys- Within the posterior fossa, Pajtler et al. [72] identified 2groups,
tem tumors [66]. Ependymomas are histologically identified which had been previously identified using clinical and molecular
based on perivascular pseudorosettes and true ependymal ro- criteria by Witt et al. [74] and Mack et al. [75]. These groups have
settes [67]. The 10-year overall survival is about 64% in pedi- beenreferredtoasgroupAandGroupB.GroupAischaracterized
atric patients and ranges from 70% to 89% in adult patients [68]. by young age, poor prognosis, and a hypermethylated CIMP phe-
The current standard of care is surgical resection and radiation notype. Group B is characterized by older age, including adults,
therapy [69]; there have been no systemic therapies to date with better prognosis, chromosome instability, and lack of a CIMP
proven survival benefit [70]. Ependymomas have historically phenotype. These differences are not recognized by the 2016
been classified based on location, grade, and histology. WHO classification.

Fig. 3 Subgroups of
ependymoma and their molecular
characteristics by location in the
neuroaxis. Reprinted with
permission from [67]
292 Chen et al.

The supratentorial group of ependymomas can be divided Novel Imaging

based on alterations in RELA and YAP1. In 2014, Parker et al.
[76] found that fusions involving RELA are driver mutations The increased levels of 2HG within IDH-mutated gliomas can
for about two-thirds of supratentorial ependymomas. RELA is be used as a biomarker via noninvasive imaging. Magnetic
the principal effector of canonical NF-κB signalling, and it is resonance spectroscopy has been shown in multiple studies
most commonly fused to an uncharacterized gene, C11orf95. to detect elevated 2HG levels and in the future may be able
The C11orf95–RELA fusions resulted from chromothripsis (an to identify reliably a tumor’s IDH mutation status without
event characterized by a multitude of clustered genome rear- need of a biopsy [78, 79]. This novel technique has potential
rangements of chromosomes [77]) involving chromosome for preoperative determination of IDH mutation status in a
11q13. Supratentorial ependymomas without RELA fusions suspected glioma, which may facilitate neurosurgical plan-
have fusions involving the transcription factor YAP1, often with ning for the desired extent of resection [80]. One recent retro-
FAM118B or MAMLD1. YAP1 fusion is associated with a better spective series indicated that the prognostic associations of
prognosis than RELA fusion-associated ependymomas. Neither extent of resection of enhancing and nonenhancing portions
of these alterations impacts treatment yet. of gliomas had significant differences based on IDH mutation
status. While extent of resection of enhancing disease was
associated with better outcome in IDH wild-type gliomas,
Clinical Applications larger volumetric resection of nonenhancing disease had a
significant impact on survival only in the IDH-mutant tumors
The rapidly evolving understanding of the molecular subtypes [80]. As new therapies develop targeting IDH1 mutations,
of gliomas has a number of clinical implications and applica- 2HG magnetic resonance imaging spectroscopy has the po-
tions (Table 2). These applications include diagnostic imag- tential to act as a pharmocodynamic biomarker, providing a
ing, pathologic testing requirements, clinical trial planning, noninvasive measure of a treatment’s efficacy. Further explo-
and targeted treatment of gliomas. ration and validation is needed to refine the technology for

Table 2 Emerging targeted therapies for glioma subtypes

Subtype Markers Median age (years) Emerging targeted therapies

Mesenchymal Glioblastoma NF1 deletion 58 Trametinib

NF-kappaB activation
Classical glioblastoma EGFR amplification 56 Erlotinib
EGFR vaccines
Anti-EGFR antibody–drug conjugates
Proneural/IDH mutant Glioblastoma IDH mutation 52 IDH vaccine
G-CIMP phenotype IDH inhibitors
Glutaminase inhibitors
Checkpoint inhibitors
Midline gliomas H3F3A K27M mutation 5–11 H3F3A K27M vaccine
Proneural/RTK glioblastomas PDGFRA amplification Sunitinib
Epithelioid glioblastoma BRAFV600E 8 Dabrafenib
ODZ3 deletion Vemurafenib
Pilocytic astrocytoma BRAF fusion 5-14 Sorafenib
Giant cell glioblastoma TP53 mutation 44 Immune checkpoint inhibitors
POLE mutation
Diffuse astrocytoma IDH mutation 36 IDH vaccine
ATRX mutation IDH inhibitors
TP53 mutation
Diffuse oligodendroglioma IDH mutation 35–44 IDH vaccine
1p19q deletion IDH inhibitors
CIC or FUBP1 mutation
Preglioblastoma IDH wild type Imetelstat
TERT promoter mutation
Pleomorphic xanthoastrocytoma BRAFV600E mutation 22 Dabrafenib
Vemurafenib
Gangliogliomas BRAFV600E mutation 9–25 Dabrafenib
Vemurafenib

EGFR = endothelial growth factor receptor; IDH = isocitrate dehydrogenase

Glioma Subclassifications and Their Clinical Significance 293

accurate quantification of the metabolite over time and avoid- prior trials were unknowingly populated by patients whose
ance of false-positives and -negatives [81]. tumors had similar morphology but diverse molecular and
genetic profiles. Given our current understanding of the mo-
Refining Routine Diagnostic Practices to Aid in Prognosis lecular data, the unseen variation in molecular alterations (and
hence varied prognoses and underlying biology) within
The newly revised WHO criteria for classification and grading groups of tumors in prior trials potentially diluted the power
of gliomas reflects a paradigm shift in the diagnosis and man- of the studies and obscured the detection of possible treatment
agement for these tumors by encouraging integration of geno- effects. Presumably, certain therapies may show clinical activ-
mic data into the standard diagnostic workup. Routine testing of ity only in certain molecular subgroups. The phase III
1p/19q co-deletion by fluorescent in situ hybridization and im- AVAglio study of bevacizumab for newly diagnosed GBM
munostaining for loss of ATRX is needed to delineate found no overall survival benefit in the population of all pa-
oligodendrogliomas from astrocytomas, resolving the frequent tients with GBM [85], but a post-hoc analysis did identify a
ambiguity of mixed features on histology. Similarly, IDH muta- potential survival increase (17.1 vs 12.8 months) within the
tion status should be ascertained for every oligodendroglioma proneural IDH1 wild-type GBM subgroup [86], warranting
and astrocytoma, irrespective of grade by immunostaining or further study.
genetic sequencing where immunohistochemical testing is ini- In diffuse glioma, EGFR amplification and MGMT meth-
tially negative. For potential PXAs, pilocytic astrocytomas, ylation define specific subpopulations against which to test
gangliogliomas, epitheliod GBMs, rhabdoid meningiomas, new therapies targeted to those markers. EGFR is a receptor
and astrocytic tumors before the age of 30 years, BRAFV600E tyrosine kinase that activates P13/AKT mitogenic pathways,
testing should be done [82]. These tests allow for accurate diag- resulting in activation of MTOR, an enzyme that is key to
nosis and prognostication. This array of molecular information increase angiogenesis, proliferation, and survival of tumor
can be used to examine and synthesize a case by correlation with cells [87]. In GBM, EGFR is overexpressed in 50% to 60%
the clinical context, imaging features, and histology so as to of patients, and 30% of patients harbor a specific deletion
draw up the best possible representation of a given tumor’s called epidermal growth factor receptor variant III
biology. This arms clinicians with powerful information that (EGFRvIII), which is an independent negative prognostic in-
often greatly increases confidence in the diagnosis, which may dicator [88]. Despite success in lung cancer, EGFR tyrosine
lead to improved management strategies. kinase inhibitors such as gefitinib and erlotinib produced no
In certain situations, unique features in the case can also survival benefit in phase II trials of EGFR-amplified GBM
prompt further testing that may, in fact, alter the diagnosis and [89, 90]. Cetuximab, a high-affinity monoclonal antibody that
carry prognostic implications. One conceivable and illustra- binds EGFR and causes destruction of the receptor, also failed
tive example might be a nonenhancing brainstem tumor to show clinical activity combined with chemotherapy [91].
biopsied in a young adult patient that was given the diagnosis The anti-EGFRvIII vaccine, rindopepimut, showed promise in
of Bgrade II astrocytoma^ but mutation in IDH was lacking— earlier studies but failed to improve survival in the recent
highly atypical in a LGG of younger onset. Further testing phase III ACTIV trial [92]. Thus, unfortunately, despite great
may reveal the presence of H3F3A K27M mutation, which is promise, we have yet to uncover any effective therapies
associated with DIPG. Though more prevalent in children, targeting EGFR in gliomas. It is unclear whether these failures
DIPGs occasionally present later in life and histologically in EGFR-targeted therapy are owing to limits of brain pene-
convey a spectrum of findings that could be consistent with tration, differences in the specific mutations in gliomas com-
different glioma types and grades [83]. This molecular finding pared with other cancers, redundant activating pathways or
would thus portend a worse prognosis than expected from the escape mechanisms, or intratumoral heterogeneity. Ongoing
histology. From a management perspective, such a tumor studies of EGFR-targeting therapies include a promising
might be treated more like a pediatric DIPG, where the stan- monoclonal antibody specifically engineered for the
dard of care is radiotherapy alone, rather than like an adult EGFRvIII mutant receptor that is in phase I trials of GBM
GBM, although further study on optimal treatment of each [93] and an antibody–drug conjugate targeting activated
molecular subtype is needed [84]. EGFR that is in a phase III trial in EGFR-amplified GBM
[94]. Additionally, development of the clinical applications
of humanized anti-EGFR variant III chimeric antigen receptor
Therapeutics and Clinical Trial Planning T cells is currently being explored [95].
Another established marker important to clinical trials is
Testing New Targeted Therapies in Defined Subgroups the methylation of the MGMT promoter. As discussed,
MGMT promoter methylation serves as a predictor of outcome
Prior to molecular classification, clinical trials for patients and enhances sensitivity to temozolomide treatment through
with glioma were stratified by clinical variables, meaning the suppression of DNA repair enzyme, MGMT. In addition,
294 Chen et al.

this epigenetic marker is being used as a basis for targeted Specific Abnormality in a Tumor Gene in People With
therapies. For example, there are current phase II/III clinical Advanced Stage Cancer (TAPUR) trial are phase II umbrella
trials of a poly-adenosine diphosphate ribose polymerase in- trials that enroll patients on the basis of genetic events rather
hibitor, veliparib, that disrupts an alternative pathway of DNA than tumor history. For example, patients with brain tumors
repair (base excision repair) combined with temozolomide in with mutations involving neurofibromin mutations, PI3Kα,
people with newly diagnosed MGMT methylated GBMs [96]. and phosphatase and tensin homologue mammalian target of
However, as the survival benefit from temozolomide for pa- rapamycin mutations and tuberous sclerosis complex muta-
tients with GBMs that are MGMT unmethylated is significant- tions can qualify for trial of targeted therapies for these muta-
ly less than for those whose tumors are methylated, this sub- tions (NCT02465060, NCT02693535). GBM-specific trials
population stands in even greater need of effective matching treatments to particular mutations, such as the
nonalkylating therapies. For this reason, some clinical trials Individualized Screening Trial of Innovative Glioblastoma
enrolling patients with newly diagnosed GBM require MGMT Therapy (INSIGHT) trial (NCT02977780), are expected in
unmethylated status as their main eligibility criteria. the near future.

Treatments Targeting IDH Mutations Molecular Markers Assisting in Understanding

and Overcoming Mechanisms of Treatment Resistance
Treatments targeting IDH-mutated tumors are in clinical trials.
Strategies include directly targeting IDH through small mo- Molecular studies have helped us to investigate the mecha-
lecular IDH inhibitors or IDH-targeted vaccines [97]. Other nisms of tumoral resistance against standard clinical treat-
strategies target metabolic consequences of IDH mutations, ments and illuminate the potential genetic events involved in
such as dependence on glutaminase [98]. developing recurrent, progressive disease. Mutational analysis
by sequencing of tissue samples in patients with low-grade
Use of Existing Targeted Agents in Tumors With Specific astrocytomas at initial resection and subsequent resections
Mutations for recurrent tumor have revealed linear and branching pat-
terns of mutational evolution traceable in most cases by shared
Gliomas that harbor the BRAFV600E may be sensitive to drugs mutations to a single original clone [111]. In such cases, IDH1
targeting this mutation. BRAF is a serine/threonine protein mutation was the only shared mutation identified across all the
kinase that plays an integral role in cell differentiation, growth, patients studied, which not only implicates IDH1 mutation as
and proliferation through its role in the Ras/Raf/MEK/extra- a pivotal inciting event in gliomagenesis, but also supports the
cellular regulated kinase pathway [99, 100]. Studies on potential for targeted therapies against IDH1 [111]. These
BRAFV600E mutant melanomas have found success with the studies have found that treatment with the alkylator temozo-
BRAF inhibitors dabrafenib or vemurafenib [101]. Experience lomide, the most common chemotherapy for gliomas compris-
with these agents in gliomas with BRAFV600E mutations are ing standard of care, was associated with eventual develop-
mostly limited to small case reports. Treatment of recurrent ment of hypermutated, temozolomide-resistant tumors in 6 of
gangliogliomas with the BRAF inhibitor dabrafanib led to a 10 patients studied [111]. Many of these hypermutated tumors
partial response in 1 of 3 patients with progression-free sur- developed mutations in the tumor suppressor gene retinoblas-
vival ranging from 4 to 10 months [102]. Several case reports toma and Akt-mammalian target of rapamycin pathway ap-
hav e demo nstrated r espon ses t o ve mu rafen ib of parently after treatment with temozolomide, implicating these
gangliogliomas with BRAFV600E mutations [103–105]. Two specific driver mutations for malignant transformation into
case reports of treating children with epithelioid GBMs har- secondary GBM. Follow-up studies have also uncovered an
boring BRAFV600E mutations with vemurafenib have shown association between mutations in mismatch repair genes or
responses or long-term progression-free intervals [106, 107]. mismatch repair silencing by MGMT methylation and devel-
Multiple case reports have now been published with responses opment of hypermutator tumors after temozolomide therapy
of PXA to BRAF-targeted treatment, such as vemurafenib [112]. As discussed above, targeting these hypermutated tu-
[107–110]. Thus, BRAFV600E-targeted therapy is very promis- mors with checkpoint inhibitors or other immunotherapy is an
ing in selected subsets of gliomas. area of active study in gliomas as in other solid tumors.
Several ongoing trials are testing treating tumors with tar-
getable mutations, such as BRAFV600E, with targeted drugs
independent of histology. The National Cancer Institute- Conclusion
sponsored Molecular Analysis for Therapy Choice (NCI-
MATCH) trial and the American Society of Clinical The classification of GBMs into subtypes based on genomic,
Oncology (ASCO)-sponsored Testing the Use of Food and epigenomic, and proteomic profiles has come a long way. In
Drug Administration (FDA) Approved Drugs That Target a addition to changing the way gliomas are being diagnosed,
Glioma Subclassifications and Their Clinical Significance 295

molecular analyses have deepened our understanding of their 16. Turcan S, Rohle D, Goenka A, et al. IDH1 mutation is sufficient to
establish the glioma hypermethylator phenotype. Nature
underlying biology. Discoveries of multiple biologically and
2012;483:479–483.
prognostically important markers have led to new categoriza- 17. Lu C, Ward PS, Kapoor GS, et al. IDH mutation impairs histone
tions of gliomas. Moreover, there are many promising thera- demethylation and results in a block to cell differentiation. Nature
pies targeted to specific pathways essential to tumor activity. 2012;483:474–478.
Clinical trials can now utilize molecular stratification to har- 18. Sun H, Yin L, Li S, et al. Prognostic significance of IDH mutation
in adult low-grade gliomas: a meta-analysis. J Neurooncol
ness greater statistical power to detect possible treatment ef- 2013;113:277–284.
fects of the new agents. However, the promise of this new 19. Wick W, Hartmann C, Engel C, et al. NOA-04 Randomized phase
molecular era is still limited by our current understanding III trial of sequential radiochemotherapy of anaplastic glioma with
and the need for more and better therapies to procarbazine, lomustine, and vincristine or temozolomide. J Clin
Oncol 2009;27:5874–5880.
target alterations in gliomas
20. Dunn GP, Andronesi OC, Cahill DP. From genomics to the clinic:
biological and translational insights of mutant IDH 1/2 in glioma.
Neurosurg Focus 2013;34:1–15.
21. Suzuki H, Aoki K, Chiba K, et al. Mutational landscape and clonal
architecture in grade II and III gliomas. Nat Genet 2015;47:458–468.
References 22. Wakimoto H, Tanaka S, Curry WT, et al. Targetable Signaling
Pathway Mutations Are Associated with Malignant Phenotype in
1. Modrek AS, Bayin NS, Placantonakis DG. Brain stem cells as the IDH-mutant gliomas. Clin Cancer Res 2014;20:2898–2909.
cell of origin in glioma. World J Stem Cells 2014;6:43–52. 23. Cohen A, Sato M, Aldape K, et al. DNA copy number analysis of
2. Louis DN, Ohgaki H, Wiestler OD, et al. The 2007 WHO classi- Grade II–III and Grade IV gliomas reveals differences in molecu-
fication of tumours of the central nervous system. Acta lar ontogeny including chromothripsis associated with IDH muta-
Neuropathol 2007;114:547–547. tion status. Acta Neuropathol Commun 2015;3:34.
3. Cahill DP, Sloan AE, Nahed BV, et al. The role of neuropathology 24. Oizel K, Gratas C, Nadaradjane A, Oliver L, Vallette FM,
in the management of patients with diffuse low grade glioma. J Pecqueur C. D-2-Hydroxyglutarate does not mimic all the IDH
Neurooncol 2015;125:531–549. mutation effects, in particular the reduced etoposide- triggered
apoptosis mediated by an alteration in mitochondrial NADH.
4. Coons SW, Johnson PC, Scheithauer BW, Yates AJ, Pearl DK.
Cell Death Dis 2015;6:e1704–e1710.
Improving diagnostic accuracy and interobserver concordance in
25. Jenkins RB, Blair H, Ballman KV, et al. A t(1;19)(q10;p10) me-
the classification and grading of primary gliomas. Cancer
diates the combined deletions of 1p and 19q and predicts a better
1997;79:1381–1393.
prognosis of patients with oligodendroglioma. Cancer Res
5. Perry A, Wesseling P. Histologic classification of gliomas. 1st ed.
2006;66:9852–9861.
Elsevier B.V; 2016.
26. Vogazianou AP, Chan R, Backlund LM, et al. Distinct patterns of
6. Hegi ME, Diserens A-C, Gorlia T, et al. MGMT gene silencing
1p and 19q alterations identify subtypes of human gliomas that
and benefit from temozolomide in glioblastoma. N Engl J Med
have different prognoses. Neuro Oncol 2010;12:664–678.
2005;352:997–1003.
27. Cairncross G, Wang M, Shaw E, et al. Phase III trial of chemora-
7. Zhang J, Stevens MFG, Bradshaw TD. Temozolomide: mecha- diotherapy for anaplastic oligodendroglioma: long-term results of
nisms of action, repair and resistance. Curr Mol Pharmacol RTOG 9402. J Clin Oncol 2013;31:337–343.
2012;5:102–114.
28. van den Bent MJ, Brandes AA, Taphoorn MJB, et al. Adjuvant
8. Esteller M, Garcia-Foncillas J, Andion E, et al. Inactivation of the procarbazine, lomustine, and vincristine chemotherapy in newly
DNA-repair gene MGMT and the clinical response of gliomas to diagnosed anaplastic oligodendroglioma: long-term follow-up of
alkylating agents. N Engl J Med 2000;343:1350–1354. EORTC Brain Tumor Group study 26951. J Clin Oncol 2013;31:
9. Stupp R, Hegi ME, Mason WP, et al. Effects of radiotherapy with 344–350.
concomitant and adjuvant temozolomide versus radiotherapy alone 29. Brennan CW, Verhaak RGW, McKenna A, et al. The somatic
on survival in glioblastoma in a randomised phase III study: 5-year genomic landscape of glioblastoma. Cell 2013;155:462–477.
analysis of the EORTC-NCIC trial. Lancet Oncol 2009;10:459–466. 30. Eckel-Passow JE, Lachance DH, Molinaro AM, et al. Glioma
10. Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health groups based on 1p/19q, IDH, and TERT promoter mutations in
Organization Classification of Tumors of the Central Nervous tumors. N Engl J Med 2015;372:2499–2508.
System: a summary. Acta Neuropathol 2016;131:803–820. 31. Arita H, Yamasaki K, Matsushita Y, et al. A combination of TERT
11. Cairncross JG, Ueki K, Zlatescu MC, et al. Specific genetic pre- promoter mutation and MGMT methylation status predicts clini-
dictors of chemotherapeutic response and survival in patients with cally relevant subgroups of newly diagnosed glioblastomas. Acta
anaplastic oligodendrogliomas. J Natl Cancer Inst 1998;90:1473– Neuropathol Commun 2016;4:1–14.
1479. 32. Nguyen HN, Lie A, Li T, et al. Human TERT promoter
12. Yan H, Friedman A, Reardon D, Herndon J. IDH1 and IDH2 mutation enables survival advantage from MGMT promoter
mutations in gliomas. N Engl J Med 2009;360:765–773. methylation in IDH1wild-type primary glioblastoma treated
13. The Cancer Genome Atlas Research Network. Comprehensive, by standard chemoradiotherapy. Neuro Oncol 2017;19:394–
integrative genomic analysis of diffuse lower-grade gliomas. N 404.
Engl J Med 2015;372:2481–2498. 33. Schwartzentruber J, Korshunov A, Liu X-Y, et al. Driver muta-
14. Yang H, Ye D, Guan KL, Xiong Y. IDH1 and IDH2 mutations in tions in histone H3.3 and chromatin remodelling genes in paedi-
tumorigenesis: mechanistic insights and clinical perspectives. Clin atric glioblastoma. Nature 2012;482:226–231.
Cancer Res 2012;18:5562–5571. 34. Wu G, Broniscer A, McEachron TA, et al. Somatic histone H3
15. Thompson CB. Metabolic enzymes as oncogenes or tumor sup- alterations in pediatric diffuse intrinsic pontine gliomas and non-
pressors. N Engl J Med 2009;360:813–815. brainstem glioblastomas. Nat Genet 2012;44:251–253.
296 Chen et al.

35. Sturm D, Witt H, Hovestadt V, et al. Hotspot mutations in H3F3A 55. Kleinschmidt-DeMasters BK, Aisner DL, Birks DK, Foreman
and IDH1 define distinct epigenetic and biological subgroups of NK. Epithelioid GBMs show a high percentage of BRAF
glioblastoma. Cancer Cell 2012;22:425–437. V600E mutation. Am J Surg Pathol 2013;37:685–698.
36. Khuong-Quang D-A, Buczkowicz P, Rakopoulos P, et al. K27M 56. Broniscer A, Tatevossian RG, Sabin ND, et al. Clinical, radiolog-
mutation in histone H3.3 defines clinically and biologically dis- ical, histological and molecular characteristics of paediatric epi-
tinct subgroups of pediatric diffuse intrinsic pontine gliomas. Acta thelioid glioblastoma. Neuropathol Appl Neurobiol 2014;40:327–
Neuropathol 2012;124:439–447. 336.
37. Hou Y, Kohanbash G, Okada K, et al. Novel and shared 57. Makuria AT, Rushing EJ, McGrail KM, Hartmann D-P, Azumi N,
neoantigen for glioma T cell therapy derived from histone 3 var- Ozdemirli M. Atypical teratoid rhabdoid tumor (AT/RT) in adults:
iant H3.3 K27M mutation. J Immunother Cancer 2015;3:P445. review of four cases. J Neurooncol 2008;88:321–330.
38. Kohanbash G, Okada K, Liu S, et al. HG-81 novel and shared 58. Alexandrescu S, Korshunov A, Lai SH, et al. Epithelioid glioblasto-
neoantigen for glioma T cell therapy derived from histone 3 vari- mas and anaplastic epithelioid pleomorphic xanthoastrocytomas-
ant H3.3 K27M mutation. Neuro Oncol 2016;18:iii67.1. same entity or first cousins? Brain Pathol 2015;26:215–223.
39. Phillips HS, Kharbanda S, Chen R, et al. Molecular subclasses of 59. Matsumura N, Nakajima N, Yamazaki T, et al. Concurrent TERT
high-grade glioma predict prognosis, delineate a pattern of disease promoter and BRAF V600E mutation in epithelioid glioblastoma
progression, and resemble stages in neurogenesis. Cancer Cell and concomitant low-grade astrocytoma. Neuropathology
2006;9:157–173. 2017;37:58–63.
40. Bhat KPL, Balasubramaniyan V, Vaillant B, et al. Mesenchymal 60. Ida CM, Rodriguez FJ, Burger PC, et al. Pleomorphic
differentiation mediated by NF-κB promotes radiation resistance xanthoastrocytoma: natural history and long-term follow-up.
in glioblastoma. Cancer Cell 2013;24:331–346. Brain Pathol 2014;25:575–586.
41. Verhaak RGW, Hoadley KA, Purdom E, et al. Integrated genomic 61. Dias-Santagata D, Lam Q, Vernovsky K, et al. BRAF V600E
analysis identifies clinically relevant subtypes of glioblastoma mutations are common in pleomorphic xanthoastrocytoma: diag-
characterized by abnormalities in PDGFRA, IDH1, EGFR, and nostic and therapeutic implications. PLOS ONE 2011;6:e17948–
NF1. Cancer Cell 2010;17:98–110. e17949.
42. Noushmehr H, Weisenberger DJ, Diefes K,et al. Identification of a 62. Schindler G, Capper D, Meyer J, et al. Analysis of BRAF V600E
CpG island methylator phenotype that defines a distinct subgroup mutation in 1,320 nervous system tumors reveals high mutation
of glioma. Cancer Cell 2010;17:510–522. frequencies in pleomorphic xanthoastrocytoma, ganglioglioma
43. Ceccarelli M, Barthel FP, Malta TM, et al. Molecular profiling and extra-cerebellar pilocytic astrocytoma. Acta Neuropathol
reveals biologically discrete subsets and pathways of progression 2011;121:397–405.
in diffuse glioma. Cell 2016;164:550–563. 63. Blumke I, Wiestler OD. Gangliogliomas: an intriguing tumor en-
44. Gill BJ, Pisapia DJ, Malone HR, et al. MRI-localized biopsies tity associated with focal epilepsies. J Neuropathol Exp Neurol
reveal subtype-specific differences in molecular and cellular com- 2002;61:575–584.
position at the margins of glioblastoma. Proc Natl Acad Sci U S A 64. Gupta K, Orisme W, Harreld JH, et al. Posterior fossa and spinal
2014;111:12550–12555. gangliogliomas form two distinct clinicopathologic and molecular
45. Wang J, Cazzato E, Ladewig E, et al. Clonal evolution of glioblas- subgroups. Acta Neuropathol Commun 2014;2:18.
toma under therapy. Nat Genet 2016;48:768–776. 65. Dougherty MJ, Santi M, Brose MS, et al. Activating mutations in
46. Kozak KR, Mahadevan A, Moody JS. Adult gliosarcoma: epide- BRAF characterize a spectrum of pediatric low-grade gliomas.
miology, natural history, and factors associated with outcome. Neuro Oncol 2010;12:621–630.
Neuro Oncol 2009;11:183–191. 66. Armstrong TS, Vera-Bolanos E, Bekele BN, Aldape K, Gilbert
47. Han SJ, Yang I, Tihan T, Prados MD, Parsa AT. Primary MR. Adult ependymal tumors: prognosis and the M. D.
gliosarcoma: key clinical and pathologic distinctions from glio- Anderson Cancer Center experience. Neuro Oncol 2010;12:862–
blastoma with implications as a unique oncologic entity. J 870.
Neurooncol 2010;96:313–320. 67. Wu J, Armstrong TS, Gilbert MR. Biology and management of
48. Reis RM, Könü-Lebleblicioglu D, Lopes JM, Kleihues P, Ohgaki ependymomas. Neuro Oncol 2016;18:902–913.
H. Genetic profile of gliosarcomas. Am J Pathol 2000;156:425– 68. Ostrom QT, Gittleman H, Liao P, et al. CBTRUS statistical report:
432. primary brain and central nervous system tumors diagnosed in the
49. Kozak KR, Moody JS. Giant cell glioblastoma: a glioblastoma United States in 2007–2011. Neuro Oncol 2014;16(Suppl. 4):iv1–
subtype with distinct epidemiology and superior prognosis. 63.
Neuro Oncol 2009;11:833–841. 69. Merchant PTE, Li C, Xiong PX, Kun PLE, Boop FA, Sanford RA.
50. Oh JE, Ohta T, Nonoguchi N, et al. Genetic alterations in A prospective study of conformal radiation therapy for pediatric
gliosarcoma and giant cell glioblastoma. Brain Pathol 2016;26: ependymoma. Lancet Oncol 2009;10:258–266.
517–522. 70. Bouffet E, Tabori U, Huang A, Bartels U. Ependymoma: lessons
51. Erson-Omay EZ, Çağlayan AO, Schultz N, et al. Somatic from the past, prospects for the future. Childs Nerv Syst 2009;25:
POLEmutations cause an ultramutated giant cell high-grade glio- 1383–1384.
ma subtype with better prognosis. Neuro Oncol 2015;17:1356– 71. Vera-Bolanos E, Aldape K, Yuan Y, et al. Clinical course and
1364. progression-free survival of adult intracranial and spinal
52. Domingo E, Freeman-Mills L, Rayner E, et al. Somatic POLE ependymoma patients. Neuro Oncol 2015;17:440–447.
proofreading domain mutation, immune response, and prognosis 72. Pajtler KW, Witt H, Sill M, et al. Molecular classification of
in colorectal cancer: a retrospective, pooled biomarker study. ependymal tumors across all cns compartments, histopathological
Lancet Gastroenterol Hepatol 2016;0:207–216. grades, and age groups. Cancer Cell 2015;27:728–743.
53. Le DT, Uram JN, Wang H, et al. PD-1 blockade in tumors with 73. Ebert C, Haken von M, Meyer-Puttlitz B, et al. Molecular genetic
mismatch-repair deficiency. N Engl J Med 2015;372:2509–2520. analysis of ependymal tumors. NF2 mutations and chromosome
54. Akimoto J, Namatame H, Haraoka J, Kudo M. Epithelioid glio- 22q loss occur preferentially in intramedullary spinal
blastoma: a case report. Brain Tumor Pathol 2005;22:21–27. ependymomas. Am J Pathol 1999;155:627–632.
Glioma Subclassifications and Their Clinical Significance 297

74. Witt H, Mack SC, Ryzhova M, et al. Delineation of two clinically 92. Malkki H. Trial Watch: glioblastoma vaccine therapy disappoint-
and molecularly distinct subgroups of posterior fossa ment in Phase III trial. Nat Rev Neurol 2016;12:190.
ependymoma. Cancer Cell 2011;20:143–157. 93. Taylor TE, Furnari FB, Cavenee WK. Targeting EGFR for treat-
75. Mack SC, Witt H, Piro RM, et al. Epigenomic alterations define ment of glioblastoma: molecular basis to overcome resistance.
lethal CIMP-positive ependymomas of infancy. Nature 2014;506: Curr Cancer Drug Targets 2012;12:197–209.
445–450. 94. Radiation Therapy Oncology Group. A study of ABT-414 in sub-
76. Parker M, Mohankumar KM, Punchihewa C, et al. C11orf95– jects with newly diagnosed glioblastoma (GBM) with epidermal
RELA fusions drive oncogenic NF-κB signalling in growth factor receptor (EGFR) amplification (intellance 1).
ependymoma. Nature 2014;506:451–455. Available at: https://clinicaltrials.gov/ct2/show/NCT02573324.
77. Forment JV, Kaidi A, Jackson SP. Chromothripsis and cancer: Accessed 24 October 2016.
causes and consequences of chromosome shattering. Nat Rev 95. Johnson LA, Scholler J, Ohkuri T, et al. Rational development and
Cancer 2012;12:663–670. characterization of humanized anti-EGFR variant III chimeric antigen
78. Choi C, Ganji SK, DeBerardinis RJ, et al. 2-hydroxyglutarate receptor T cells for glioblastoma. Sci Transl Med 2015;7:275ra22.
detection by magnetic resonance spectroscopy in IDH-mutated 96. Gupta SK, Kizilbash SH, Carlson BL, et al. Delineation of
patients with gliomas. Nat Med 2012;18:624–629. MGMT hypermethylation as a biomarker for veliparib-mediated
79. Pope WB, Prins RM, Albert Thomas M, et al. Non-invasive de- temozolomide-sensitizing therapy of glioblastoma. J Natl Cancer
tection of 2-hydroxyglutarate and other metabolites in IDH1 mu- Inst 2015;108.
tant glioma patients using magnetic resonance spectroscopy. J 97. Schumacher T, Bunse L, Pusch S, et al. A vaccine targeting mutant
Neurooncol 2012;107:197–205. IDH1 induces antitumour immunity. Nature 2014;512:324–327.
80. Beiko J, Suki D, Hess KR, et al. IDH1 mutant malignant astrocyto- 98. Seltzer MJ, Bennett BD, Joshi AD, et al. Inhibition of glutaminase
mas are more amenable to surgical resection and have a survival preferentially slows growth of glioma cells with mutant IDH1.
benefit associated with maximal surgical resection. Neuro Oncol Cancer Res 2010;70:8981–8987.
2013;16:81–91. 99. Wan PTC, Garnett MJ, Roe SM, et al. Mechanism of activation of
81. Andronesi OC, Rapalino O, Gerstner E, et al. Detection of onco- the RAF-ERK Signaling pathway by oncogenic mutations of B-
genic IDH1 mutations using magnetic resonance spectroscopy of RAF. Cell 2004;116:855–867.
2-hydroxyglutarate. J Clin Invest 2013;123:3659–3663. 100. Davies H, Bignell GR, Cox C, et al. Mutations of the BRAF gene
in human cancer. Nature 2002;417:949–954.
82. Behling F, Barrantes-Freer A, Skardelly M, et al. Frequency of
101. Chapman PB, Hauschild A, Robert C, et al. Improved survival
BRAF V600E mutations in 969 central nervous system neo-
with vemurafenib in melanoma with BRAF V600E mutation. N
plasms. Diagn Pathol 2016;11:55.
Engl J Med 2011;364:2507–2516.
83. Buczkowicz P, Bartels U, Bouffet E, Becher O, Hawkins C.
102. Chamberlain MC. Recurrent ganglioglioma in adults treated with
Histopathological spectrum of paediatric diffuse intrinsic pontine
BRAF inhibitors. CNS Oncol 2016;5:27–29.
glioma: diagnostic and therapeutic implications. Acta
103. Rush S, Foreman N, Liu A. Brainstem ganglioglioma successfully
Neuropathol 2014;128:573–581.
treated with vemurafenib. J Clin Oncol 2013;31:e159–e160.
84. Jones C, Karajannis MA, Jones DTW, et al. Pediatric high-grade 104. del Bufalo F, Carai A, Figà-Talamanca L, et al. Response of re-
glioma: biologically and clinically in need of new thinking. Neuro current BRAFV600E mutated ganglioglioma to Vemurafenib as
Oncol 2016 Jun 9 [Epub ahead of print]. single agent. J Transl Med 2014;12:356.
85. Chinot OL, Wick W, Mason W, et al. Bevacizumab plus radiother- 105. Aguilera D, Janss A, Mazewski C, et al. Successful retreatment of
apy–temozolomide for newly diagnosed glioblastoma. N Engl J a child with a refractory brainstem ganglioglioma with
Med 2014;370:709–722. vemurafenib. Pediatr Blood Cancer 2015;63:541–543.
86. Sandmann T, Bourgon R, Garcia J, et al. Patients with proneural 106. Kleinschmidt-DeMasters BK, Aisner DL, Foreman NK. BRAF
glioblastoma may derive overall survival benefit from the addition VE1 immunoreactivity patterns in epithelioid glioblastomas posi-
of bevacizumab to first-line radiotherapy and temozolomide: ret- tive for BRAF V600E mutation. Am J Surg Pathol 2015;39:528–
rospective analysis of the AVAglio trial. J Clin Oncol 2015;33: 540.
2735–2744. 107. Robinson GW, Orr BA, Gajjar A. Complete clinical regression of
87. Liu P, Cheng H, Roberts TM, Zhao JJ. Targeting the a BRAF V600E-mutant pediatric glioblastoma multiforme after
phosphoinositide 3-kinase pathway in cancer. Nat Rev Drug BRAF inhibitor therapy. BMC Cancer 2014;14:397.
Discov 2009;8:627–644. 108. Skrypek M, Foreman N, Guillaume D, Moertel C. Pilomyxoid
88. Heimberger AB, Hlatky R, Suki D, et al. Prognostic effect astrocytoma treated successfully with vemurafenib. Pediatr
of epidermal growth factor receptor and EGFRvIII in glio- Blood Cancer 2014;61:2099–2100.
blastoma multiforme patients. Clin Cancer Res 2005;11: 109. Lee EQ, Ruland S, LeBoeuf NR, Wen PY, Santagata S. Successful
1462–1466. treatment of a progressive BRAF V600E-mutated anaplastic pleo-
89. Raizer JJ, Abrey LE, Lassman AB, et al. A phase II trial of erlo- morphic xanthoastrocytoma with vemurafenib monotherapy. J
tinib in patients with recurrent malignant gliomas and nonprogres- Clin Oncol 2016;34:e87–e89.
sive glioblastoma multiforme postradiation therapy. Neuro Oncol 110. Chamberlain MC. Salvage therapy with BRAF inhibitors for re-
2010;12:95–103. current pleomorphic xanthoastrocytoma: a retrospective case se-
90. Uhm JH, Ballman KV, Wu W, et al. Phase II evaluation of gefitinib ries. J Neuro Oncol 2013;114:237–240.
in patients with newly diagnosed grade 4 astrocytoma: Mayo/ 111. Johnson BE, Mazor T, Hong C, et al. Mutational analysis reveals
North Central Cancer Treatment Group Study N0074. Int J the origin and therapy-driven evolution of recurrent glioma.
Radiat Oncol Biol Phys 2011;80:347–353. Science 2014;343:189–193.
91. Hasselbalch B, Lassen U, Hansen S, et al. Cetuximab, 112. van Thuijl HF, Mazor T, Johnson BE, et al. Evolution of DNA
bevacizumab, and irinotecan for patients with primary glioblasto- repair defects during malignant progression of low-grade gliomas
ma and progression after radiation therapy and temozolomide: a after temozolomide treatment. Acta Neuropathol 2015;129:597–
phase II trial. Neuro Oncol 2010;12:508–516. 607.