The evolving molecular genetics of low-grade glioma

Abstract

Low-grade gliomas (LGG) constitute grade I and grade II tumors of astrocytic and grade II tumors of oligodendroglial lineage. Although these tumors are typically slow growing, they may be associated with significant morbidity and mortality due to recurrence and malignant progression, even in the setting of optimal resection. LGG in pediatric and adult age groups are currently classified by morphologic criteria. Recent years have heralded a molecular revolution in understanding brain tumors, including LGG. Next generation sequencing has definitively demonstrated that pediatric and adult LGG fundamentally differ in their underlying molecular characteristics, despite being histologically similar. Pediatric LGG show alterations in FGFR1 and BRAF in pilocytic astrocytomas and FGFR1 alterations in diffuse astrocytomas, each converging on the MAP kinase-signaling pathway. Adult LGG are characterized by IDH1/2 mutations and ATRX mutations in astrocytic tumors and IDH1/2 mutations and 1p/19q codeletions in oligodendroglial tumors. TERT promoter mutations are also noted in LGG and are mainly associated with oligodendrogliomas. These findings have considerably refined approaches to classifying these tumors. Moreover, many of the molecular alterations identified in LGG directly impact on prognosis, tumor biology, and the development of novel therapies.

Introduction

Gliomas constitute the most common primary central nervous system tumor variants. The current World Health Organization (WHO) classification system uses two basic morphologic criteria to delineate individual diagnostic entities. The first defines tumor type on the basis of presumed histiogenesis into either astrocytic or oligodendroglial groupings. Further classification by grade, anticipating biological behavior in the absence of treatment, yields the final classification scheme: Low-grade gliomas - (1) grade I astrocytomas – pilocytic astrocytomas, (2) grade II diffuse astrocytomas, and (3) grade II oligodendrogliomas; High-grade gliomas - (1) Grade III anaplastic astrocytomas, (2) Grade III anaplastic oligodendrogliomas and (3) grade IV glioblastomas (astrocytic tumors). Additional morphologic subtypes, such as angiocentric glioma, have also been described, although their underlying pathogenesis remains relatively obscure.

The past decade has seen a molecular revolution driven by high-throughput sequencing technology, yielding penetrating insights into glioma pathogenesis. Several new driver mutations have been described in gliomas and the field has rapidly grown to reflect the emerging complexity of these tumors. Many of these insights carry prognostic and therapeutic implications and have impacted how we approach the diagnosis and classification of gliomas. As such, histopathologic criteria may no longer be solely sufficient to appropriately classify these tumors and it is increasingly recognized that molecular information should be integrated into standard diagnostic interpretations (1).

While glioblastomas have undergone extensive genomic characterization, low-grade gliomas (LGG) remain comparatively less understood. In contrast to high-grade gliomas, LGG frequently exhibit extend periods of relative indolence in their growth and clinical behavior. Nevertheless, LGG that cannot be surgically resected are often associated with significant morbidity and mortality, with diffuse variants invariably progressing through recurrence to high-grade status. Thus, better molecular characterization of these tumors is critical to developing novel therapeutic targets for effective treatment. Recent years have witnessed significant advances in understanding the molecular characteristics of LGG. Moreover, while histopathologic features characterizing gliomas in adults and children are similar, it is becoming increasingly apparent that tumors in these two age groups have distinct underlying biological foundations. In this article we review the various molecular alterations identified and characterized in pediatric and adult LGG and discuss the implications of recent discoveries from both diagnostic and biologic perspectives.

Molecular aspects of pediatric low-grade gliomas

In recent years, findings from a number of studies have greatly clarified the molecular events likely driving pediatric gliomas. These discoveries include histone mutations in pediatric high-grade astrocytomas and alterations in BRAF, FGFR1 and MYB in astrocytic tumors (2–9). LGG are generally characterized by a lower frequency of somatic mutations and structural variations per tumor, suggesting that their pathogenesis is driven by fewer genetic changes overall. We discuss these molecular events below. Since many of these findings represent new or relatively new discoveries, data regarding prognosis and potential mechanism of pathogenesis are just beginning to emerge

BRAF alterations in LGG

BRAF-KIAA1549 fusions

The most frequent alterations in pediatric LGG involve the v-raf murine sarcoma viral oncogene homolog B1 (BRAF). BRAF is member of the Raf kinase family of proteins involved in the Mitogen-activated protein (MAP) kinase pathway. Pilocytic astrocytomas (grade I) show tandem duplications at chromosome 7q34 resulting in the formation of a fusion gene between the kinase domain of BRAF and KIAA1549 (BRAF-KIAA1549 fusion gene) (4–9). Rare BRAF fusions such as BRAF-FAM131B, BRAF-RNF130, BRAF-CLCN6, BRAF-MKRN1 and BRAF-GNAI1 have also been reported but at a much lower frequency (9, 10) (11). Interestingly, pilocytic astrocytomas show variable rates of BRAF fusion depending on their location in the central nervous system (CNS). For example ~75% of cerebellar pilocytic astrocytomas exhibit BRAF-KIAA1549 fusions. By contrast, only 33% of supratentorial and ~50% of optic nerve pilocytic astrocytomas harbor BRAF fusions (12–14). Data regarding the prognostic effects of BRAF fusion in pediatric LGG are variable, with no clear consensus, and thus may warrant further investigation (12, 13, 15, 16).

BRAF V600E mutation

Nonsynonymous point mutations in BRAF resulting in a valine to glutamic acid substitution at position 600 (V600E) were first described in melanocytic lesions. Subsequently, BRAF V600E mutations were identified in specific subtypes of gliomas. BRAF V600E mutations are most frequent in pleomorphic xanthoastrocytomas (PXA, ~70%) and gangliogliomas (GG, ~20%) and occur at lower frequencies in pilocytic astrocytomas, diffuse astrocytomas and pilomyxoid astrocytomas (8, 17–21). BRAF V600E mutant LGG exhibited a trend towards increased risk for progression (16) and in gangliogliomas was associated with shorter recurrence-free survival (22).

Receptor tyrosine kinase (RTK) FGFR1 alterations in LGG

RTK are cell surface receptors that play a key role in signal transduction. These proteins bear an extracellular domain, a transmembrane domain and an intracellular tyrosine kinase domain (TKD) (23), and have been extensively implicated in the pathogenesis of both adult and pediatric high-grade glioma (24). Two groups simultaneously reported LGG alterations in fibroblast growth factor receptor 1 (FGFR1), an RTK and member of the FGF receptor family that binds to the fibroblast growth factor family of proteins (8, 9).

In diffuse cerebral LGG, intragenic duplications of a portion of the TKD of FGFR1 were noted in 24% of cases (8). These alterations are thought to constitutively activate the receptor independent of ligand activation. Jones et al reported mutations affecting the TKD of FGFR1 in 14/141 (two sample sets including 5/96 and 9/45 tumors negative for BRAF alterations) pilocytic astrocytomas (9). Mutations involved two hotspot codons, Asn546 and Lys656. Moreover, FGFR1 mutant pilocytic astrocytomas tended toward extracerebellar and midline localization.

Interestingly, rare alterations involving the kinase domain of NTRK2 (TrkB), an RTK that binds BDNF, were noted in 2/49 pilocytic astrocytomas resulting in the formation of the fusion genes (QKI-NTRK2 and NACC2-NTRK2) (9). Similar to FGFR1 alterations, the fusion proteins include receptor dimerization domains and are hypothesized to dimerize independent of ligand.

MYB and MYBL1 mutations in LGG

MYB (V-Myb Avian Myeloblastosis Viral Oncogene Homolog) copy number alterations were initially described in 2 diffuse astrocytomas (WHO grade II) and 1 angiocentric glioma (WHO grade II) using Affymetrix SNP arrays and interphase FISH analyses (25). Subsequent efforts using whole genome sequencing, transcriptome and targeted high-throughput sequencing showed rearrangement of MYBL1 (V-Myb Avian Myeloblastosis Viral Oncogene Homolog-Like 1) in one diffuse astrocytoma and rearrangement or copy number alterations of MYB in 5 diffuse astrocytomas, 2 angiocentric gliomas and 1 oligodendroglioma (8). Together, MYBL1 and MYB alterations were present in 25% of diffuse grade II LGG. Alterations in MYB manifested as episome formation, deletion of the negative regulatory region of MYB or deletion of microRNA-binding sites (8).

Another group identified gains on chromosome 8q13.1 involving MYBL1 in 28% (5/18) of pediatric diffuse astrocytomas (26), where they resulted in tandem duplication and truncation of the negative-regulatory C-terminal domain of the protein. MYBL1 truncation products, but not wild type MYBL1, when transduced into NIH-3T3 cells, were able to generate tumors in vitro in soft agar and as xenografts in vivo. Interestingly, this group did not identify MYB alterations in diffuse astrocytomas, although two angiocentric gliomas showed 6q23.3 deletions resulting in truncated MYB.

Genetic alterations in BRAF and FGFR1 give rise to aberrant Mitogen-activated protein kinase (MAP Kinase) pathway signaling

The MAP kinase pathway is critical to normal development and deregulated in a multitude of cancers. Situated downstream of receptor tyrosine kinase activation, pathway signaling is initiated by recruitment of the G-protein Ras (through Shc and Sos), which in turn activates the Raf/MEK/ERK cascade, ultimately leading to protein phosphorylation and activation of multiple protein targets (FIG. 1.).

FIG. 1. IDH 1/2, ATRX, and TERT promoter mutations and 1p/19q codeletion in adult LGG.

IDH1/2 mutations are thought to be an early event in adult LGG pathogenesis in a common precursor cell. Mutations in IDH1 (cytoplasmic) and IDH2 (mitochondrial) result in the generation of the oncometabolite 2-hydroxyglutarate (2-HG). 2-HG is thought to inhibit α-ketoglutarate dependent demethylases resulting in histone and DNA-CpG island methylation (G-CIMP phenotype). Mutations in ATRX are seen in astrocytic tumors. ATRX is a helicase belonging to the SWI/SNF family involved in H3.3 deposition (along with its partner DAXX). Its deficiency induces alternative lengthening of telomeres. 1p/19q codeletion is seen in oligodendrogliomas. CIC and FUBP1 alterations are associated with 1p/19q codeletion in a variable percentage of oligodendrogliomas. TERT promoter mutations are also noted in oligodendrogliomas and are thought to be important for telomere maintenance. Text in red indicates mutations.

In the context of LGG, BRAF fusion proteins and the BRAF V600E mutation each result in aberrant activation of the MAP kinase pathway. BRAF fusion proteins bear the kinase domain, but not the auto-inhibitory N-terminus and thus harbor constitutive kinase activity (27) (FIG. 1.). Similarly, intragenic duplications involving FGFR1 result in a constitutively active receptor protein, with downstream activation of the MAP kinase pathway. Indeed, expression of TKD-duplicated FGFR1 in vitro resulted in activation of the MAP kinase and phosphoinositide 3-kinase (PI3K) pathways and was reversed by FGFR1 inhibitors. Furthermore, Tp53-null astrocytes bearing TKD-duplicated FGFR1 when xenografted into mouse brains generated high-grade gliomas. The authors hypothesize that TKD-duplicated FGFR1 brings two TKD together overriding the requirement for receptor dimerization by FGF for signal transduction (8).

That deregulated MAP kinase pathway signaling serves as a major driver in these tumors is underscored by the identification in LGG, albeit at lower frequency, of mutations in other core pathway components. For instance, mutations in the tyrosine phosphatase PTPN11 were noted in 2/49 pilocytic astrocytomas with FGFR1 mutations (9). PTPN11 encodes a RAS-MAPK–related adaptor protein, suggesting cooperativity with activating FGFR1 alterations (FIG. 1.). Along these lines, mutations in RAS and NF1 (encoding the neurofibromin protein, a negative regulator of RAS) were noted at similar frequencies (8, 9). Together, these findings suggest that MAP kinase pathway activation is required for development of these tumors and, moreover, serves as a unifying pathogenic concept in the broad classification of well-encapsulated WHO grade I neuroepithelial neoplasms like PXA, GG, and pilocytic astrocytoma. Also consistent with these conjectures, conditionally deleted Mek1/2 significantly reduced glial progenitors in mice, leading to failed gliogenesis. Moreover, animals that survived to postnatal stages were nearly devoid of astrocytes and oligodendroglia (28). Similarly, conditional deletion of ERK1/2 in mice resulted in abnormal glial development and reduced progenitor proliferation (29). These factors underscore the importance of MAP kinase signaling in glial development and suggest that aberrant activation of this pathway mediates tumorigenesis in pediatric LGG.

Molecular characteristics of adult low-grade gliomas

IDH 1/2 mutations

The discovery of isocitrate dehydrogenase (IDH) 1/2 mutations in gliomas heralded the genomic era of glioma research. Large profiling studies have identified IDH1/2 mutations in >70% of grade II and grade III gliomas and more than 90% of secondary GBM (30–32). Wild type IDH proteins form core components of the TCA cycle, where they catalyze the conversion of isocitrate to α-ketoglutarate (α-KG) in the cytoplasm (IDH1) and mitochondria (IDH2). IDH mutations are invariably missense and heterozygous, with IDH1 mutations predominating (more than 90%), and involve active site arginine residues, either R132 in IDH1 or R172 in IDH2 (31–34). Mutant IDH1/2 catalyze the generation of the oncometabolite D-2HG from α-KG (35, 36) (FIG. 2.), whose physiologic effects are profound (see below). IDH-mutant gliomas tend to be occur more frequently in young adult patients (31, 37–40). Moreover, the association of IDH1/2 mutation with astrocytomas, oligodendrogliomas and oligoastrocytomas strongly suggests that the event arises early in the pathogenesis of LGG (37, 40, 41) (FIG. 2.). IDH-mutant tumors confer favorable prognosis relative to wild type counterparts regardless of WHO grade (42–46).

FIG. 2. FGFR1 and BRAF alterations in pediatric LGG converge on the MAP kinase pathway.

Alterations in FGFR1 result in constitutive activation of the receptor resulting in activation of the MAP kinase pathway. A subset of pediatric LGG also shows mutations in the receptor tyrosine kinase NTRK2. Rare mutations involve other members of this pathway including RAS, NF1 (negative regulator of RAS) and PTPN11, a tyrosine phosphatase adaptor protein. BRAF V600E mutations and the BRAF-KIAA1549 fusion (other rare fusions not illustrated) also result in constitutive kinase activity and aberrant MAP kinase activation. Text in red indicates mutations/alterations.

A subset of gliomas exhibits widespread DNA hypermethylation across the genome, referred to as the CpG island hypermethylator phenotype (G-CIMP) (47)). G-CIMP is strongly associated with mutations in IDH1/2 in LGG (47–51). Moreover, in both immortalized astrocytes and colon cancer cell lines, expression of IDH1 R132H, the most common glioma-associated IDH mutation, fully recapitulates G-CIMP (50, 51). The oncometabolite D-2HG generated in IDH-mutant tumors inhibits a variety of α-KG-dependent enzymes (52, 53) involved not only in DNA demethylation but also in carnitine synthesis, hypoxic sensing, collagen modifications and histone modification (reviewed in (54)). Indeed, astrocytic cell lines expressing mutant IDH1 R132H, and IDH-mutant oligodendrogliomas show increased trimethylation of histone marks such as H3K9, H3K27 and H3K36 (50, 55, 56) (FIG. 2.). Both DNA and histone hypermethylation, occurring as a consequence of elevated D-2HG, are thought to arrest cellular differentiation by repressing a broad spectrum of target genes (50, 55) (FIG. 2.).

CIC and FUBP1 mutations

Co-incident whole-arm loss of chromosomes 1p and 19q is observed in approximately 70% of oligodendrogliomas and is a significantly favorable prognostic factor (57–59). 1p/19q codeletion results from an unbalanced translocation involving the centromeric regions of chromosomes 1p and 19q (58) (FIG. 2.). Precisely how these genetic abnormalities contribute to oligodendroglial pathogenesis remains unestablished. However, it has been postulated for some time that these regions might harbor potential tumor suppressor genes.

Mutations in CIC (homolog of the Drosophila gene capicua) on chromosome 19q and FUBP1 (FUSE binding protein 1) on chromosome 1p have been recently described in oligodendrogliomas (60–63) (FIG. 2.). CIC mutations were observed in 17–79% of tumors diagnosed at WHO grade II oligodendrogliomas (5/14 (60), 7/9 (63), 11/62 (61), 8/21 (62)). This frequency rose to 25–75% when only 1p/19q codeleted tumors were considered (3/4 (60), 7/9 (63), 9/36 (61), 7/12 (64)). Consistent with a loss-of-function phenotype, CIC mutations are distributed throughout the coding region of the gene (with a predilection for exon 5) and include nonsense, insertions/deletions, missense and frame shift variants. Also consistent with loss-of-fuction, FUBP1 mutations are mainly frameshift and nonsense variants, and occur at lower frequencies (14–22%) than CIC mutations in low-grade oligodendrogliomas (2/14 (60), 2/9 (63), 3/21 (62), 3/17 (64)). CIC is known to functionally repress genes normally activated by RTK signaling by a mechanism called “default repression” (65), and FUBP1 is a DNA-binding protein that activates c-MYC transcription (66). Thus, both CIC and FUBP1 appear to serve as negative regulators of established oncogenic pathways. Nevertheless, establishing the precise mechanisms by which either CIC or FUBP1 mutations contribute to oligodendroglioma pathogenesis will require further study.

ATRX mutations

ATRX (α thalassemia/mental retardation syndrome X-linked) is a DNA helicase and chromatin remodeling protein, belonging to the SWI/SNF family (67). Germline loss-of-function mutations in ATRX are associated with alpha thalassemia mental retardation X-linked (ATR-X) syndrome (68). A primary function of ATRX is incorporation of histone H3.3 monomers into chromatin in collaboration with the histone chaperone protein DAXX (Death-associated protein 6) (69, 70) (FIG. 2.). In 2012, ATRX mutations were described in adult and pediatric astrocytic gliomas (3, 62, 71–73) where they exhibited a strong association with the alternate lengthening of telomeres (ALT) phenotype, a pathological telomere maintenance mechanism though to promote cellular immortality (3, 71, 74, 75) (FIG. 2.). In total, ATRX mutations were found in 33%–67% of grade II astrocytic tumors, and occurred in 75–80% of IDH-mutant LGG that did not also exhibit 1p/19q codeletion (62, 71, 73). In fact, ATRX mutation was mutually exclusive with 1p/19q codeletion in glioma and strongly associated with TP53 mutation (3, 62, 71–73) (FIG. 2.). These data suggest that ATRX mutation, together with TP53 mutation, may delineate a distinct pathogenic route operative in the majority of diffuse astrocytic LGG. That being said, the precise mechanism(s) by which ATRX mutations drive gliomagenesis remain unclear.

TERT promoter mutations

Point mutations in the promoter region of the telomerase reverse transcriptase (TERT) gene were first discovered in melanoma, and are thought to increase telomerase expression, thereby maintaining telomere length and enabling repeated cell division (76, 77). These mutations have been identified in many CNS tumors, including glioblastomas, medulloblastomas and LGG. They occur exclusively at positions −228 and −250 in the TERT promoter region, substituting a cytosine for a thymine in either case to unmask a binding site for ETS family transcription factors (78–82). In LGG, TERT promoter mutations are predominantly observed in oligodendrogliomas (63–78%, 12/19 (78); 25/34 (79) and 29/37 (83)) and less frequently (0–32%) in diffuse astrocytomas (0/8 (78), 10/52 (79) 8/25 (83)). Intriguingly, astrocytic tumors with TERT promoter mutations show an inverse relationship with IDH mutations (83). Additionally, TERT promoter mutations are tightly associated (98–100%) with 1p/19q co-deletion and are mutually exclusive with ATRX mutations in LGG (78–80, 83) (FIG. 2.). These findings emphasize the importance of pathological telomere maintenance in LGG, whether by way of TERT promoter mutations in 1p/19q codeleted tumors (predominantly oligodendroglioma), or ALT in ATRX-mutant tumors (predominantly astrocytoma).

Diagnostic and therapeutic implications of molecular genetics in pediatric and adult LGG

Diagnostic implications

As detailed above, high-throughput molecular profiling has dramatically altered conceptions of glioma biology and, in doing so, has led to refinements of well-established classification schemes (FIG. 1 and 2). Perhaps most notably, the identification of IDH1/2 mutations in both low and high-grade adult gliomas is now of considerable importance, due to the significant prognostic benefit conferred by the genomic alteration. The standard initial approach of many pathology practices is immunohistochemical, using an antibody specifically directed against IDH1 R132H (accounting for more than 95% of all glioma-associated IDH mutations), followed by sequencing-based genotyping in immunonegative cases. The availability of a robust immunohistochemical reagent recognizing most IDH-mutant tumors also facilitates the differentiation of true glial neoplasms from non-neoplastic glioma mimics such as reactive astrogliosis (33, 84–88).

The mutual exclusivity of ATRX mutation and 1p/19q codeletions in LGG has prompted the proposal that all diffuse gliomas be classified on the basis of IDH and ATRX mutational status—or a negative staining pattern by IHC (FIG. 3A)—combined with 1p/19q codeletion. This approach may lend better clarity to the often-subjective diagnosis of mixed lineage gliomas (oligoastrocytomas), as these tumors have been shown to nicely segregate into ATRX-mutant and 1p/19q codeleted subgroups (62, 71, 73). Moreover, prognostic stratification among gliomas delineated by these molecular criteria outperforms that seen following conventional histopathological classification (62) (89). CIC and FUBP1 mutations do not bear 100% concurrence with 1p/19q codeletions (60–63). Thus, from a diagnostic and prognostic view, assessment of 1p/19q deletion remains superior to CIC and FUBP1 genotyping in the establishment of oligodendroglial lineage. As an aside, the near universal occurrence of TERT mutations in 1p/19q-codeleted LGG suggests that combined IDH1/2 and TERT genotyping might also support the rendering of an oligodendroglioma diagnosis (78).

FIG. 3. Immunohistochemical assessment of mutations in ATRX and BRAF.

A, Loss-of-function mutations in ATRX lead to loss of nuclear expression. H & E stained and ATRX-immunostained micrographs of ATRX-mut and ATRX-wt tumors are shown. B, positive BRAF V600E immunostaining in a mutant-harboring PXA. H& E staining also shown.

In pediatric LGG, the identification of BRAF alterations—by molecular techniques or the BRAF V600E antibody (FIG. 3B)—is of significance due to the potential for targeted therapies (see below). Identifying BRAF alterations may also help in diagnostically challenging cases by designating encapsulated WHO grade I astrocytic variants (e.g. pilocytic astrocytoma) from other neoplastic entities (19, 90). That being said, more studies are required to fully assess the impact of BRAF, FGFR1 and MYB/MYBL1 alterations on current classification of pediatric LGG.

Therapeutic implications

The advances in the molecular characterization of LGG discussed above have provided numerous insights into potential pathogenic mechanisms. Indeed, these studies have yielded an array of therapeutic targets that can be leveraged to design novel therapies. Many of these efforts are still in experimental stages. However, pharmaceutically targeting BRAF V600E has already achieved considerable success in melanoma (91–93) and V600E inhibitors were effective in preclinical animal models of high-grade glioma (94, 95) and a single case of a 12-year-old patient with GBM (96). However, these successes should be treated with cautious optimism due to the established existence of multiple mechanisms of resistance to BRAF inhibitors in other tumor types (97). Recent studies have also shown that mutant IDH1 inhibition is partially effective in xenograft glioma models, although blood-brain barrier permeability remains an issue (98). Alternatively, vaccine-based approaches against IDH1 R132H have elicited anti-tumor immune responses in tumors bearing the mutation (99). The utility of these and other therapeutic strategies in targeting LGG remains unclear and will be the subject of extensive research effort in the immediate future.