Molecular Markers in Low Grade Glioma – Toward Tumor Reclassification

Abstract

Low-grade diffuse gliomas are a heterogeneous group of primary glial brain tumors with highly variable survival. Currently, patients with low grade diffuse gliomas are stratified into risk subgroups by subjective histopathologic criteria with significant interobserver variability. Several key molecular signatures have emerged as diagnostic, prognostic, and predictor biomarkers for tumor classification and patient risk stratification. In this review, we will discuss the impact of the most critical molecular alterations described in diffuse (IDH1/2, 1p/19q co-deletion, ATRX, TERT, CIC, FUBP1) and circumscribed (BRAF-KIAA1549, BRAF^V600E, C11orf95–RELA fusion) gliomas. These molecular features reflect tumor heterogeneity and have specific associations with patient outcome that determine appropriate patient management. This has led to an important, fundamental shift towards developing a molecular classification of WHO grade II-III diffuse glioma.

Introduction

In the United States 28% of all primary brain and central nervous system (CNS) tumors are diagnosed as gliomas ¹. Based on their infiltrative behavior, gliomas are subdivided into two main subgroups: circumscribed and diffuse. The circumscribed gliomas are generally amenable to total surgical resection and patients with these tumors have improved outcomes compared to patients with diffuse gliomas. The aggressive phenotype of diffuse gliomas is attributed to the tendency of the malignant glioma cells to infiltrate the neuropil along axons and travel far away from the primary tumor site (Figure 1). A malignant glioma cell may travel to the opposite cerebral hemisphere. For this reason diffuse gliomas cannot be completely surgically resected (Figure 1c). Diffuse gliomas encompass two main histological subtypes (astrocytoma and oligodendroglioma) of which a third subtype is derived (mixed oligoastroytoma). Histological criteria established by the World Health Organization (WHO), further stratify gliomas into four grades of aggressiveness (WHO I-IV) ². WHO grade I is reserved for circumscribed glial and glio-neuronal entities usually with favorable prognosis, while diffuse gliomas comprise the more aggressive WHO grades II-IV (Table 1) ². Patients with astrocytomas generally have worse outcomes than patients with oligodendrogliomas (Table 2) ¹.

Figure 1.

Diffuse gliomas are aggressive neoplasms that tend to infiltrate along dendrites and axons and travel long distances from the primary site of origin. Malignant glioma cells are shown infiltrating the cortex and forming aggregates along a blood vessel (yellow arrows) extending all the way to the leptomeningeal surface (black arrows) (A – H&E, Obj: 100X). A higher power of the cortex highlights neoplastic glioma cells (green arrow) surrounding the neurons (yellow arrow) (B – H&E, Obj: 200X). A special immunohistochemical stain for the mutated IDH-R132H protein highlights (in brown) malignant glioma cells infiltrating the normal (pale) subcortical and cortical tissue – note how the subcortex has an incresed density of glioma cells (left upper corner) compared to the superficial cortex (right lower corner) (C – IDH1-R132H, Obj: 40X, scale not available).

Table 1.

Classification and assigned grading of gliomas after the current WHO system ².

	WHO grade
Circumscribed	Astrocytic
	Subependymal giant cell astrocytoma	I
	Pilocytic astrocytoma	I
	Pilomyxoid astrocytoma	II
	Pleomorphic xanthoastrocytoma	II
	Ependymal
	Subependymoma	I
	Myxopapillary ependymoma	I
	Ependymoma	II
	Anaplastic ependymoma	III
Diffuse	Diffuse astrocytoma	II
	Oligodendroglioma	II
	Mixed oligoastrocytoma	II
	Anaplastic astrocytoma	III
	Anaplastic oligodendroglioma	III
	Anaplastic mixed oligoatrocytoma	III
	Glioblastoma	IV
Other	Angiocentric glioma	I
	Desmoplastic infantile astrocytoma	I
	Chordoid glioma of the third ventricle	II
Mixed glio-neuronal tumors	Ganglioglioma	I
	Anaplastic ganglioglioma	III
	Desmoplastic infantile ganglioglioma	I
	Papillary glioneuronal tumour	I
	Rosette-forming glioneuronal tumour of the fourth ventricle	I
	Dysembryoplastic neuroepithelial tumour	I

Table 2.

Five-year relative survival rates for patients with diffuse glioma ¹.

Diffuse glioma (WHO grade)	5-Year relative survival rate (%)(95% Confidence Interval)
Oligodendroglioma (II)	79.5 (77.9–81)
Mixed oligoastrocytoma (II-III)	61.1 (58.6–63.6)
Anaplastic oligodendroglioma (III)	52.2 (49.1–55.1)
Astocytoma (II)	47.4 (46–48.8)
Anaplastic astrocytoma (III)	27.3 (25.6–28.9)
Glioblastoma (IV)	5 (4.8–5.4)

While histologic criteria for diagnosing the most aggressive diffuse glioma (i.e. glioblastoma) are clear, there is extreme variability in interpretation of current morphological criteria and in diagnostic reproducibility for grade II and III diffuse glioma among pathologists ^3–6. Several molecular signatures have now been identified in gliomas, with important diagnostic, prognostic, and/or predictive roles. These genetic alterations have led to further stratification of gliomas into several distinct subgroups. The addition of genetic markers offer better prognostic patient stratification compared to WHO grading alone. Guidelines for a combined molecular-morphologic approach to glioma diagnosis are under development ⁷.

The treatment of WHO grade II-III diffuse gliomas continues to evolve. Retrospective molecular analysis of tumors from patients enrolled in randomized clinical trials have called into question the use of standard histologic grading and have emphasized the relevance of key molecular alterations as important predictive biomarkers with implications for determining appropriate tumor management.

Histopathological classification of WHO grade II-III diffuse glioma

Since Bailey and Cushing’s initial attempt at brain tumor classification in 1926 ^{8, 9}, histological examination has been the mainstay method for risk class assignment, patient outcome stratification, therapy guidelines, and stratification for clinical trials ^{2, 3, 10}. Several potential issues arising during histological examination are responsible for the significant inter-observer variability in achieving diagnostic reproducibility. For example, the current WHO criterion for grade III designation of astrocytomas (i.e. “brisk mitotic activity” ²) does not clearly specify any mitotic figure cutoffs and is, therefore, ambiguous and subjective and often based on the pathologist’s individual experience and bias. Similarly, the criteria for the “mixed oligoastrocytoma” category (i.e. “recognition of neoplastic glial cells with convincing astrocytic or oligodendroglial phenotypes” ²) are subjective. The pathologist’s expertise and experience in neuropathology also significantly impacts accuracy of diagnosis. In a large oncologic center study investigating the rate of diagnosis disagreement after expert neuropathology review, approximately 40% of case reviews had some type of disagreement with the original diagnosis, of which about 9% were serious with immediate impact on treatment ¹¹. Another potential problem in interpretation can be caused by mistakenly omitting foci of high-grade histology (i.e. missing slides from the resection specimen or alternatively, limited surgical sampling in subtotal resections or biopsies and/or suboptimal specimen sampling of the resected specimen) ^{3–6, 11}. Due to one or several of these issues, the histopathologic consensus among neuropathologists after a single review is reached in only approximately 50% of cases ⁵. Although this number could be improved after several case reviews, ⁵ it is imperative to implement more objective criteria and/or incorporate ancillary modalities alongside histological parameters in order to significantly improve the high rate of inter-observer variability in histopathologic diagnosis and classification of diffuse gliomas.

Molecular features of WHO II-III diffuse gliomas

The current WHO classification does not comprehensively reflect diffuse glioma biology and patient outcome. There is extensive evidence that tumors from different patients that have indistinguishable morphology under the microscope do not necessarily share the same biology and do not necessarily reflect similar patient outcomes ^12–16. Molecular subgroups of diffuse glioma, heterogeneous in WHO grade, with different survival outcomes have been described. The use of molecular stratification was superior in predicting outcome compared to the WHO grade alone ¹⁵.

Isocitrate dehydrogenase (IDH) mutations in diffuse gliomas

Key molecular markers in WHO grade II-III diffuse glioma include IDH 1 and 2 (IDH1/2) mutations. These genes encode the Krebs/citric acid cycle family of metabolic enzymes ¹⁷. IDH mutations in diffuse glioma were initially described in 2008 ¹⁸ but occur at lower frequencies in other malignancies ^19–24. In diffuse glioma IDH1 mutations occur more commonly (>90%) than IDH2 mutations and are mutually exclusive. IDH mutations are common in grade II-III diffuse glioma (~65–80%) and secondary glioblastoma (~80%) (i.e. glioblastoma that arise following progression from a grade II or III diffuse glioma) while primary glioblastomas usually lack or show a very low frequency of IDH mutations (~5%) ^25–32. IDH gene family mutations, the most common of which are IDH1^R132H and IDH2^R172K, confer both loss and gain of function that impacts epigenetic regulation through accumulation of 2-hydroxyglutarate and inhibition of α-ketoglutarate-dependent deoxygenases ^{17, 33–35}. The impact of mutant IDH induces a hypermethylator phenotype ³⁶, the glioma-CpG island methylator phenotype (G-CIMP) ³⁷. G-CIMP is characteristic of grade II-III diffuse glioma, is associated with improved prognosis, and with a proneural molecular gene expression signature ^{38, 39}.

IDH family mutations are early ^{25, 40} and consistent ^{40, 41} molecular events in the development of a glioma and are complemented by subsequent mutually exclusive glioma lineage specific genetic alterations, such as TP53 and alpha thalassemia/mental retardation syndrome X-linked (ATRX) mutations. These latter mutations are associated with the astrocytic phenotype ^42–47. On the other hand, 1p/19q co-deletion ^{39, 43, 46} is associated with mutations in the homolog of Drosophila capicua gene (CIC) ^{42, 48} and/or far-upstream binding protein 1 gene (FUBP1) and with the oligodendroglial phenotype ^{42, 49}. These molecular markers strongly support the predominant monoclonal origin of mixed oligoastrocytomas demonstrated by microdissection studies. ^{41, 50, 51} These molecular markers may help to further classify this controversial mixed glioma category into specific subclasses of either astrocytoma or oligodendroglioma ⁵².

In addition to its important diagnostic role, IDH mutations also have a significant prognostic role in high-grade diffuse glioma (WHO grade III and IV) independent of WHO grade in some instances ^{29, 30, 32, 42, 53–56}. On the other hand, the prognostic role of IDH mutations in WHO grade II diffuse glioma has not been completely elucidated ^{40, 57–63}. These mutations do not appear to have prognostic significance in non-CNS malignancies ^{21, 23}. Similarly, the predictive role of IDH mutations has not been clarified. Very few studies suggest a potential advantage of the use of IDH mutations for determining treatment response in anaplastic diffuse gliomas. Two randomized phase III clinical trials [Radiation Therapy Oncology Group (RTOG) 9402 and European Organisation for the Research and Treatment of Cancer (EORTC) 26951] have now reported a similar improved overall survival benefit to a combined regimen using radiotherapy and PCV specifically in patients with anaplastic gliomas that are IDH mutant and 1p/19q non-co-deleted ^{64, 65}. The predictive role of IDH mutations remains to be further investigated along with the possible prognostic implication of the IDH driven G-CIMP in grade II-III gliomas ^{37, 38, 66, 67}. Of note, a promising IDH1^R132H specific inhibitor drug (AGI-5198) has shown significant activity in pre-clinical models ⁶⁸ and is currently in phase 1 clinical trials for solid tumors ^{69, 7069, 7069, 7069, 70}(69, 70)(69, 70)^{69, 70,29, 84}. Multiple other agents targeting mutant IDH are also under investigation.

IDH^R132H mutation can be easily detected clinically by immunohistochemistry (Figure 1C). The expression of mutated IDH1^R132H protein confirms mutation. Cases that are IDH1^R132H immunonegative can be further interrogated by DNA sequencing for other IDH1 or IDH2 mutations. Since mutations are mainly present in codons 132 and 172 respectively, sequencing can be limited to these codons ¹⁷. At our institution all grade II-III diffuse gliomas and glioblastomas in young patients (less than age 50) are interrogated in this manner ⁷¹.

Chromosomes 1p/19q loss of heterozygosity in diffuse gliomas

Another important molecular marker in diffuse glioma is the presence of 1p/19q co-deletion, the molecular signature of oligodendroglioma, initially described in 1994 ^{4, 39, 48, 72–74}. The proposed mechanism of formation of this chromosomal abnormality is a translocation between 1p and 19q leading to the derivative chromosome, der(1;19)(q10;p10). This derivative chromosome was demonstrated in a small number of tumors leading to the hypothesis that subsequent der(1;19)(p10;q10) formation leads to 1p/19q loss of heterozygosity ^{75, 76}. Importantly, 1p/19q co-deletion is only present if IDH mutations are present ⁷⁷; therefore, this implies an association with the G-CIMP and proneural expression phenotypes ^{38, 39}, an association that was demonstrated in grade II-III oligodendrogliomas ^{39, 66}. Several studies showed that 1p/19q co-deletion can aid in risk stratification of IDH mutant gliomas with IDH mutant, 1p/19q co-deleted gliomas having the best prognosis, followed by IDH mutant, 1p/19q non-co-deleted gliomas and lastly by IDH wild-type, 1p/19q non-co-deleted tumors with the worst outcome ^{39, 57, 63, 64, 78}.

Besides prognostic significance, 1p/19q co-deletion is a marker of chemotherapeutic response ^79–83. Patients with 1p/19q co-deleted diffuse gliomas responded better to adjuvant chemotherapy [either procarbazine (Matulane®, Sigma Tau Pharmaceuticals, Gaithersburg, MD)/lomustine (CCNU) (CeeNU®, Bristal-Myers Squib Company, Princeton, NJ)/vincristine (Oncovin®, Eli Lilly and Company, Indianapolis, IN) (PCV) regimen; or temozolomide (TMZ)] ^{79, 81–90}. The mechanism for the associated chemosensitivity remains unknown and whether 1p/19q co-deletion leads to loss of a chemoresistant gene(s) or is merely a marker of more chemosensitive clones remains to be determined.

One of the most popular methods for detection of 1p/19q co-deletion is fluorescent in situ hybridization (FISH). FISH permits pathologists to correlate chromosomal arm copy number findings with tissue morphology and does not require the use of normal control samples (Figure 2). Polymerase chain reaction-based loss of heterozygosity assays or array comparative genomic hydridization (CGH) are also utilized in different laboratory settings ^{91, 92}.

Figure 2.

Diffuse glioma with 1p/19q co-deletion. The locus-specific identifier (LSI) probe, 1p36 (SpectrumOrange) and corresponding LSI 1q25 (SpectrumGreen) show two normal green signals and single abnormal orange signal in a subpopulation of interphase cells suggestive of 1p loss (left). LSI probe, 19q13 (SpectrumOrange) and corresponding LSI 19p13 (SpectrumGreen) show two normal green signals and single abnormal orange signal in a subpopulation of oligodendroglioma interphase cells suggestive of 19q loss (right) (Photographs courtesy of Prof. Dr. Adekunle Adesina, Texas Children’s Hospital, Houston, TX, USA).

ATRX and telomerase reverse transcriptase (TERT) promoter mutations in gliomas

Two distinct telomere maintenance mechanisms have been recently described mainly in IDH mutant grade II-III diffuse glioma and primary glioblastoma. Telomeres are repetitive guanine-rich nucleotide sequences situated at each chromatid end. They are required for chromosome stability and shorten with each cell division. ⁹³ In cancer, the length of telomere sequences is maintained either by telomerase enzyme activity or by a mechanism independent of telomerase activity called alternative lengthening of telomeres (ALT). ^{94, 95} Two mutually exclusive telomere maintenance mechanisms appear associated with IDH mutant WHO grade II-III diffuse gliomas. An ALT mechanism may be triggered by loss of the normal ATRX protein function of maintaining chromatin integrity for DNA replication. ATRX or death-associated protein 6 (DAXX) mutations cause dysfunctional ATRX-DAXX protein complexes that are unable to carry their normal histone chaperone function, leading to chromatin breakage, and abnormal DNA replication. ^96–98 Telomeric DNA double-strand breakage may trigger ALT.^{43, 99–102} This mechanism is encountered in IDH mutant, 1p/19q non co-deleted grade II-III diffuse glioma ^{43, 44, 46, 103}.

The other telomere maintenance mechanism involves point mutations in the TERT promoter. TERT encodes the catalytic subunit of telomerase. Telomerase is a RNA-dependent polymerase composed of two subunits: TERT (the catalytic subunit) and TERC (the telomerase RNA component which serves as a template for telomere extension) ^{104, 105}. Two consistent and mutually exclusive TERT promoter point mutations (C228T and C250T) have been described in gliomas ^{77, 106}. These mutations have been also frequently found in other non-CNS tumors, with C228T being by far more common (~80%) than C250T (~20%) overall ^{103, 107, 108}. Point mutations in the TERT promoter region create binding sites for the E-twenty-six (Ets) family of transcription factors, ^107–110 which upon binding cause two to four fold increase in transcriptional activity ¹⁰⁷ with subsequent increased TERT mRNA expression ⁷⁷. This increased mRNA expression seems to be positively correlated with the tumor’s CIC mutational status ⁷⁷, likely because CIC regulates the Ets family of transcription factors ¹¹¹. Chen et al. demonstrated in vitro that the increase in TERT transcriptional activity is maintained under hypoxic conditions and under treatment with TMZ ¹¹². This second telomere maintenance mechanism is also characteristic of IDH mutant, 1p/19q co-deleted grade II-III diffuse glioma (and therefore of molecular oligodendroglioma) ^{103, 106, 113}. In the diffuse glioma category, it is not yet clear if TERT promoter mutations confer additional prognostic benefit to the presence of IDH mutations and 1p/19q co-deletion. Most studies demonstrate improved survival in patients with IDH mutant, TERT promoter mutated grade II-III diffuse gliomas ^{77, 113, 114} ; however after stratification for 1p/19q co-deletion status, TERT promoter mutations demonstrate a prognostic advantage only to the 1p/19q non-co-deleted subset ¹¹⁴. Interestingly, TERT promoter mutations are negative prognostic biomarkers in the IDH wild-type grade II-III diffuse gliomas subset ^{77, 113, 114} and in primary glioblastoma ^{77, 103, 112, 113}. In addition to obtaining IDH mutation and 1p/19q co-deletion status, identifying ATRX and TERT promoter mutations status should further enhance the stratification of diffuse gliomas.

IDH mutation– driven subgroups of WHO grade II-III diffuse glioma

There is evidence that the presence or absence of IDH mutation is an important branch point towards grade II-III diffuse glioma subclassification. Gorovets et al. subclassified 101 grade II and III diffuse gliomas of astrocytic morphology by IDH mutation status. IDH mutant tumors were enriched in TP53 mutations, PTEN promoter methylation, and gains of 8q. Based on expression signatures, IDH mutant grade II-III astrocytomas were also subdivided into two subgroups: neuroblastic and early progenitor-like, the former enriched in mature neuronal and the latter enriched in developmental gene signatures. The early progenitor-like subgroup components were associated with TP53 mutations and several chromosomal copy number abnormalities (gains of 7p and 15q, and losses of 4q34.3, 9p23, 11p, 12q21.33, 13q, and 19q).

On the other hand IDH wild-type grade II-III astrocytomas shared EGFR amplifications, PTEN losses, PI3K/AKT molecular pathway activation, and gains of 7p and losses of 9p and 10q ¹⁵. Partial or total loss of 10q and 9p loss have been previously associated with dismal prognosis in WHO II-III diffuse glioma ^115–117. This is important because the latter signatures are characteristic molecular markers of primary glioblastoma ^{18, 118–120}. This suggests that a subgroup of grade II-III astrocytomas confined to the IDH wild-type genetic subclass is biologically identical to glioblastoma. This same group also demonstrated that these specific tumors clustered within a separate, heterogeneous subgroup defined based on expression profiling that, not surprisingly, was called pre-glioblastoma ¹⁵. Similarly Yan et al., defined an IDH wild-type subgroup based on expression profiling that was also predominantly composed of primary glioblastomas and also clustered several grade II-III diffuse gliomas ¹⁶.

A proposed classification scheme based on a summary of the molecular analysis to date for WHO II-III diffuse gliomas is shown in Figure 3. This schema may more accurately reflect underlying biology and be more representative of patient outcome than the WHO grade alone. The Cancer Genome Atlas (TCGA) ¹²¹ group is currently working on analyzing a large cohort of WHO grade II-III diffuse gliomas of all morphologies. Extensive data derived from multiple molecular platforms were analyzed and their results are expected shortly. We can speculate that similar molecular findings will be reported as have been reported previously and that emphasize a significant difference between groups of tumors that are primarily separated by the IDH mutation status.

Figure 3.

Proposed molecular classification of WHO grade II-III diffuse glioma.

Susceptibility loci for the development of diffuse glioma

Several large genome-wide association studies (GWAS) have been performed to identify genetic variants associated with the risk of development of a glioma. These GWAS studies have identified several risk single nucleotide polymorphism (SNP) loci. The reported loci with the strongest glioma risk association were rs78378222 (TP53, 17p13), rs4295627 and rs55705857 (CCDC26, 8q24.21), rs2736100 (TERT, 5p15.33), rs1920116 (TERC, 3q26.2), rs4977756 (CDKN2B, 9p21.3), rs6010620 and rs2297440 (RTEL1, 20q13.33), and rs498872 (PHLDB1, 11q23.3) ^122–131. Of these, rs4295627, rs55705857 (CCDC26, 8q24.21) and rs498872 (PHLDB1, 11q23.3) are strongly associated with low-grade disease, IDH mutations ¹³² and 1p/19q co-deletions ¹²⁷. For the former two SNPs, the risk for developing oligodendroglioma was also shown by Jenkins et al ^{128, 129}. High-grade disease, IDH wild-type, EGFR amplification, CDKN2A p16INK4a homozygous deletion, 9p and 10q loss were linked to rs2736100 (TERT, 5p15.33) and rs6010620 (RTEL1, 20q13.33) ¹²⁷. An additional SNP locus associated with high-grade glioma was rs2297440 (RTEL1, 20q13.33) ¹²⁸.

Molecular features of well-circumscribed gliomas

BRAF genetic alterations are shared by pilocytic astrocytoma, ganglioglioma, and pleomorphic xanthoastrocytoma. BRAF-KIAA1549 fusion-duplication, a possibly prognostic marker ¹³³, is frequent in younger patients ¹³⁴ with cerebellar pilocytic astrocytomas (~60–80%), intermediate in frequency for brainstem, hypothalamic, and optic pathway tumors (~60%) ¹³⁵, and low (~20%) in supratentorial cortical tumors ¹³⁶. BRAF^V600E mutations are common in pleomorphic xanthoastrocytoma (~60–80%) involving the temporal lobes, ganglioglioma (~30–50%) ^137–139, and less commonly found in pilocytic astrocytoma (~2 to 30% depending on location). Most common locations include diencephalon, followed by cerebral cortex and brainstem, and least common in cerebellar tumors) ¹³⁷.

Location-specific subgroups of ependymomas with characteristic genetic signatures have been described. Supratentorial ependymomas (~70–75%) are enriched in C11orf95–RELA fusion, driver of NF-kB cell signaling ^{140, 141}. Posterior fossa ependymomas are comprised of two genetically and clinically distinct subgroups, group A (PFA) and B (PFB). The more aggressive PFA group is enriched in cancer-related signal transduction pathway gene signatures, exhibits the CIMP phenotype (distinct from G-CIMP), while the PFB group is CIMP negative and enriched in chromosomal number aberrations ^142–144.

Conclusions

Grade II-III diffuse gliomas are heterogeneous tumors. Based on current published data, IDH mutations and 1p/19q co-deletion are major drivers of gliomagenesis and in determining outcome for grade II-III diffuse gliomas. Diffuse glioma subgroups defined based on these two molecular alterations will require further characterization in order to identify additional important biomarkers as well as for therapeutic target discovery. Continued endeavors to further characterize grade II-III glioma characterization should not only offer important information regarding many accepted clinical observations but also provide mechanisms regarding these observations. For example, why are 1p/19q co-deleted tumors more chemosensitive? Why is the improved overall survival in those patients receiving chemoradiation observed late after the median overall survival has already been reached? How should 1p/19q non-co-deleted tumors be treated? It is anticipated in the near future that preclinical/laboratory efforts paired with clinical results obtained from large collaborative studies and clinical trials will likely provide answers to these important questions regarding appropriate patient management. In the near future, the use of combined molecular-histopathological criteria will improve glioma risk stratification, aid in trial design, and ultimately be used to guide therapy for patients with diffuse gliomas.