Chromatin Remodeling Defects in Pediatric and Young Adult Glioblastoma: A Tale of a Variant Histone 3 Tail

Abstract

Primary brain tumors occur in 8 out of 100 000 people and are the leading cause of cancer‐related death in children. Among brain tumors, high‐grade astrocytomas (HGAs) including glioblastoma multiforme (GBM) are aggressive and are lethal human cancers. Despite decades of concerted therapeutic efforts, HGAs remain essentially incurable in adults and children. Recent discoveries have revolutionized our understanding of these tumors in children and young adults. Recurrent somatic driver mutations in the tail of histone 3 variant 3 (H3.3), leading to amino acid substitutions at key residues, namely lysine (K) 27 (K27M) and glycine 34 (G34R/G34V), were identified as a new molecular mechanism in pediatric GBM. These mutations represent the pediatric counterpart of the recurrent mutations in isocitrate dehydrogenases (IDH) identified in young adult gliomas and provide a much‐needed new pathway that can be targeted for therapeutic development. This review will provide an overview of the potential role of these mutations in altering chromatin structure and affecting specific molecular pathways ultimately leading to gliomagenesis. The distinct changes in chromatin structure and the specific downstream events induced by each mutation need characterizing independently if progress is to be made in tackling this devastating cancer.

Primary brain tumors (originating in the brain) are the leading cause of cancer‐related death in children under the age of 20, now surpassing leukemia, and the third leading cause of cancer‐related death in young adults aged 20–39 years 61. Astrocytomas are the most common primary brain tumor across the lifespan, and the highest grade, glioblastoma multiforme (GBM), remains essentially incurable despite decades of concerted therapeutic efforts 14. Similar histology of pediatric and adult tumors has caused current childhood GBM treatments to be fueled by adult studies, and show, across the board, little therapeutic advance. One impediment to treatment is that GBM is diagnosed as a single entity by pathologists who cannot discriminate potential genetic drivers and molecular subtypes. This impacts the design and outcome of clinical trials and it likely contributes to the apparent inherent resistance of GBM to adjuvant therapies and poor progress in improving survival or the quality of life of patients and their families. Other crucial obstacles to progress are the lack of reliable in vitro and in vivo models for pediatric GBM based on the lack of identified genetic drivers specific to children until very recently. Consequently, better stratification of patients based on tumor biology, improved identification of relevant therapeutic targets and the design of experimental “companion” models to test compounds affecting specific genetic/molecular drivers are essential for therapeutic breakthroughs in this deadly disease. This review will provide a short overview of gliomas across the lifespan and recent molecular advances that led to the identification of recurrent mutations in H3F3A, which are genetic drivers specific to pediatric GBM, their role in chromatin remodeling and the presence of at least six distinct molecular subgroups in GBM across ages that need to be tackled separately to achieve future therapeutic success in this deadly cancer.

General Overview of Gliomas Across Age

Gliomas are heterogeneous and are classified by the World Health Organization (WHO) according to the presumed cell of origin 41. WHO grade I and II gliomas are commonly referred to as low‐grade gliomas (LGG), while WHO grade III (anaplastic) and IV (GBM) tumors are regarded as high‐grade gliomas 41. Tumor grade, age at diagnosis and degree of surgical resection are regarded as the most important prognostic factors across the lifespan 35. The frequency, anatomic location, progression mode and pathologic spectrum of gliomas differ in children and adults. Oligodendroglial differentiation and progression of lower grade gliomas to higher grade tumors are rare in children, and the vast majority of pediatric GBM occur de novo. A significant proportion of pediatric gliomas are low‐grade tumors with a high prevalence of grade I astrocytomas and rare grade II astrocytomas, whereas in adults, high‐grade gliomas represent the vast majority of primary central nervous system tumors. Tumors in the optic pathway, and in infra‐tentorial locations such as the cerebellum and brainstem are the most common localizations in children, whereas supra‐tentorial tumors are more frequent in adults and include locations such as the thalamus and cerebral cortex. This is of importance as the cell of origin, the microenvironment and chromatin conformation will differ based on age, cell type and anatomical localization and are, as we have shown, relevant in pathogenesis 8, 36, 40, 59, 60. We and others have shown that grade I tumors, which represent ∼25% of all pediatric brain tumors, are characterized by genetic alterations in the mitogen‐activated protein kinase (MAPK) pathway 27, 31, 32, 53 and constitute one biological paradigm 13, 24, 27, 28. Grade II astrocytomas are rare in children (<5% of brain cancers) and high‐grade astrocytomas (HGA, grades III and IV) account for ∼10%–15% of all pediatric brain tumors when combined with diffuse intrinsic pontine gliomas (DIPG), which are high‐grade gliomas that occur in the brainstem, with an incidence close to 0.6/100 000 in children aged 0–19 years and a dismal mortality rate close to 90% at 5 years 58. Therapeutic advances remain low and this high mortality occurs regardless of patient's age. This is despite numerous clinical trials combining surgery, radiotherapy, chemotherapy and, in recent years, the development of many therapies including temozolomide, and targeted agents such as receptor tyrosine kinase and angiogenesis inhibitors, suggesting that these efforts are not sufficient and/or adapted to counter tumor progression.

GBM Tumors in Children and Young Adults Are not a Single Disease and Result from Mutations in Genes Affecting Chromatin Rewiring

The Cancer Genome Atlas project (TCGA) and other groups performed genomic and epigenomic analysis and targeted sequencing of adult GBM samples 1, 7, 9, 48, 50, 68. They revealed GBM in adults to be highly heterogeneous and identified the crucial role for isocitrate dehydrogenase 1 or 2 (IDH) metabolic pathways in the genesis of secondary GBM, while EGFR amplification and gain of function mutations as well as PTEN and/or CDKN2A/B loss target de novo adult GBM. Indeed, recurrent IDH1 and 2 mutations are found in up to 80% of grade II and III adult gliomas and secondary GBM but rarely in GBM occurring de novo (primary GBM) 50, 68. Pediatric GBM are morphologically indistinguishable from adult GBM, which prompted clinicians and investigators to view them as the same disease. We have, however, disproved this assumption and our group and others demonstrated the importance of analyzing the unique biology of these tumors in children. Indeed, while mutations in TP53, CDKN2A and PIK3CA are common to all HGA, PTEN mutations and EGFR amplifications occur in less than 10% of pediatric GBM 54, 55, while IDH mutations are rare in children (less than 5% of GBM) and their frequency is higher in adolescents 56. Pediatric and adult HGA have different gene expression profiles and DNA copy number alterations 2, 20, 23, 52, 57 including a higher proportion of amplification of PDGFRA in children, especially after radiation therapy 52. Up to recently, DIPG, which occur mainly in children aged between 5 and 10 years and have a particularly grim prognosis with an overall survival rate at 2 years less than 10%, have been considered to be molecularly distinct from supra‐tentorial GBM 51, 69.

Using next‐generation sequencing technology, we recently identified a new molecular mechanism driving GBM in children, namely two recurrent heterozygous somatic mutations in H3F3A, which encodes the replication‐independent histone 3 variant H3.3 59. These mutations lead to amino acid changes in key residues (K27M, G34R/G34V) and were identified in 35% of supra‐tentorial pediatric and young adult HGA 59. This is the first report to identify mutations in a regulatory histone in humans and these mutations are the pediatric counterpart of the recurrent IDH mutations identified in young adult GBM, which indirectly affect these histone marks 50, 68. H3.3 mutations significantly overlapped with mutations in TP53 and in ATRX (α‐thalassemia/mental retardation syndrome‐X‐linked) 59 and less frequently in the ATRX heterodimer DAXX, which encodes subunits of a chromatin remodeling complex required for H3.3 incorporation at pericentric heterochromatin and telomeres 16, 39. Using an independent cohort of ∼790 gliomas across age, group and grade, we further showed H3.3 mutations to be specific to high‐grade tumors, to be prevalent in children (incidence <3.4% in adult HGA) and to be mutually exclusive with IDH mutations 59, 62. We further identified that K27M‐H3.3 characterize 71% of brainstem high‐grade gliomas (DIPG) 34 and 80% of thalamic pediatric GBM 62. G34R/V‐H3.3 in HGA and K27M‐H3.3 in HGA and DIPG were independently identified by another group, which also identified K27M in the related canonical histone H3.1 in 18% of DIPG 67. H3F3A and ATRX/DAXX were not part of the close to ∼600 genes initially sequenced in adult GBM by the adult consortia including TCGA. This and their relative specificity to children explain why these mutations were not previously identified despite their staggering frequency. Rather than selecting and sequencing genes thought to be important, the use of unbiased genome‐wide sequencing methods to cover all protein‐coding genes in samples from diseased and normal tissue, and the enrichment in pediatric samples helped in identifying these mutations. Next‐generation sequencing technologies have currently become the method of choice to interrogate a tumor genome based on coverage and cost‐effectiveness. Remarkably, H3F3A mutations seem for the time being specific to pediatric HGA and have not yet been identified in other pediatric tumors including other brain tumors or leukemia 17 or in available datasets on adult cancers. However, one needs to bear in mind that most of these samples are from adult cohorts and that possibly enrichment of a specific rare tumor subset may still reveal presence of these mutations that may otherwise be overlooked even in children.

H3F3A and IDH Mutations Arise in Distinct Anatomical Locations of the Brain and May Arise from Distinct Developmental Time Points and Cellular Origins

Our findings indicate neuroanatomical and age specificities of the mutation type and the combination of mutations in this setting relative to older age GBM and point to a developmental defect at the origin of pediatric and young adult HGA (Figure 1) 34, 59, 62. Indeed, K27M‐H3.3 mutations characterize midline GBM and occur in younger children with brainstem and thalamic tumors (70%–80% of all HGA cases in these regions, median age 10.5 years, range 5–23 years) 34, 59, 62. They overlap with TP53 mutations in 80% of cases regardless of location, while ATRX mutations occur in only 50% of cases and favor older children and thalamic location (only 20% of DIPG are mutant for ATRX) 34. G34R/V‐H3.3 mutations are mainly found in HGA located within the cerebral hemispheres 34, 59, 62. They occur in patients around the threshold between the adolescent and adult populations (median age 18 years, range 9–42 years) and almost always overlap with mutant TP53/ATRX. IDH alterations affect younger adults (median age 40 years, range 13–71 years) and occur in the cortex, mainly in the frontal lobes suggesting that these tumors also arise from a neural precursor population that is spatially and temporally restricted in the brain 38. IDH alterations are gain of function, heterozygous, somatic mutations and have been shown to be initiating events in gliomas 50, 56, 66. Intriguingly, they also seem to require association with other genetic events to achieve full‐blown tumorigenesis. They are associated with two mutually exclusive genetic alterations, TP53 alterations and 1p19q co‐deletions 5, 49 that respectively characterize astrocytic and oligodendroglial IDH‐mutant gliomas. We 34, 40, 59 and others 29, 33 recently showed that mutations in ATRX characterize IDH‐ and TP53‐mutant astrocytomas in young adults with low‐ and high‐grade tumors. Furthermore, ATRX alterations are specific to astrocytic tumors and are mutually exclusive with CIC mutations and loss of 1p19q, which characterize oligodendrogliomas. ATRX inactivating mutations thus characterize older children with GBM and adult IDH‐mutant gliomas of the astrocytic lineage arguing for the importance of an H3F3A & ATRX & TP53 or an ATRX & IDH1/2 & TP53 mutant phenotype in their early development and progression and have now been found in pediatric and adult GBM 25, 59, older patients with neuroblastomas 10, 45 and in pancreatic neuroendocrine tumors 29. Mutations in ATRX/DAXX may interfere with H3.3 incorporation at these loci, thus compromising the structural integrity of the chromosome. Supportive of this hypothesis is the presence of alternative lengthening of telomeres (ALT), a telomerase‐independent telomere maintenance mechanism, in tumors with ATRX mutations 25, 59. ALT seems to be promoted by loss of ATRX that may increase the rate of homologous recombination at telomeres and pericentric heterochromatin and thus lead to increased lengthening of the telomeres through an alternate mechanism to increased telomerase activity, providing cells with the capacity for unlimited cellular proliferation. Based on the role of ATRX‐DAXX in H3.3 deposition in specific areas of the chromatin, in telomere maintenance and in genomic stability 6, 26, these findings further support the role of chromatin modifications in the genesis of HGA in children and young adults.

Figure 1.

Neuroanatomical and age specificity of isocitrate dehydrogenase (IDH), K27M‐H3.3 and G34R/V‐H3.3 mutations in high‐grade astrocytomas (HGA). K27M‐H3.3 or H3.1 (yellow stars) occur mainly in brainstem HGA and K27M‐H3.3 mainly thalamic HGA (70%–80% of all GBM in these locations). This H3.3 mutation is inconsistently associated with ATRX mutations. G34R/V‐H3.3 occur mainly in the cerebral hemispheres similar to IDH mutations which have been previously shown to have a predilection for the frontal cortex. Both H3.3 mutations are significantly associated with ATRX mutations (purple filled circles) and TP53 mutations (not represented here for clarity issues). IDH and H3.3 mutations represent distinct disease entities that possibly arise from separate cellular origins as the result of largely non‐overlapping sets of molecular events. Note: the size of the shape illustrating each mutation is approximately proportional to the % identified in 34, 59, 62.

Alterations in the Histone Code Underlie Pediatric and Young Adult GBM and Induce Defects in Chromatin Remodeling

Virtually all DNA processes are regulated by the histone code, including replication and repair, regulation of gene expression and centromere and telomere maintenance 11. Accordingly, mutations in genes affecting histone post‐translational modifications (PTMs) are increasingly described in cancer 11. Histones package and organize DNA at the level of the fundamental unit of chromatin, the nucleosome. Nucleosomes can be modulated by a large variety of covalent PTMs mostly occurring in the N‐terminal tails of histones, and also by the incorporation of histone variants 18, 21, 22. H3.3 is a universal, replication‐independent histone predominantly incorporated into transcription sites and associated with active and open chromatin [reviewed in 18, 63]. H3.3 not only functions as a neutral replacement histone, but in addition participates in the epigenetic transmission of active chromatin states and is associated with chromatin assembly factors in large‐scale replication‐independent chromatin remodeling mechanisms 37, 42. Its role in histone replacement at active genes and promoters is conserved in the single histone H3 present in yeast, indicating its importance throughout evolution 18. This histone variant is actively loaded in the developing brain, replacing other resident histones 3.1 or 3.2 4. Several PTMs regulate histone function in the nucleosome. Reversible methylation of several histone lysine residues is mediated by distinct histone methyltransferases or demethylases to specific histone Lys (K) or Arg residues. Polycomb repressor complex 2 (PRC2) recruitment is associated with methylation at K27 that represses transcription and cell differentiation 3. Gain of function mutations in EZH2, the main K27 trimethyltransferase, has been identified in leukemias and lymphomas 44, 46, 70, and overexpression of this gene was also identified in many cancers including medulloblastoma 30, 47. Thus, altered methylation or loss of acetylation of K27 is a potential driver of gliomagenesis. Methylation of K36 has been widely associated with active chromatin but also with transcriptional repression, alternative splicing, DNA replication and repair, DNA methylation and imprinting, and the transmission of memory of gene expression from parents to offspring [reviewed in 4]. H3.3K36 methylation is potentially disrupted by the G34R/V‐H3.3 mutation. Indeed, glycine 34 of histone H3 (H3G34) lies in close proximity to lysine 36 (H3K36), a residue that regulates transcriptional elongation. The G34R/V‐H3.3 mutation may impact the ability of histone‐modifying complexes to methylate or acetylate H3K36, thereby altering the transcription of several target genes. Strikingly, our gene expression analysis revealed different gene expression patterns between samples with the K27M‐H3.3 mutation vs. samples with the G34R/V‐H3.3 mutation, suggesting that each mutation favors the expression of a specific genetic program. IDH mutations that give a neomorphic function to these enzymes enable the generation of high quantities of 2‐hydroxyglutarate 15, an oncometabolite. This oncometabolite, in turn, competitively inhibits the activity of histone demethylases 43, affecting chromatin structure through alteration of histone PTMs. Interestingly, the oncometabolite produced by IDH1 mutations impairs histone demethylation, affecting UTX (H3K27me) and KDM4A/JMJD2A (H3K36/K9me) resulting in a block to cell differentiation 12, further supporting a role in gliomas for altered methylation of these residues.

Integrated Epigenetic and Genetic Analysis Identifies at Least Six Biologically Distinct GBM Subgroups Across Ages

Altogether, findings from several groups converge to indicate that alterations of the histone code are at the origin of pediatric and young adult HGA. DNA methylation patterns are better correlated with histone lysine methylation patterns than with the underlying genome sequence context 19 and, in recent years, a distinct glioma‐CpG‐island methylator phenotype (G‐CIMP) was identified and found to correlate with IDH1‐mutant gliomas 48, 64. Our investigation of the genome‐wide DNA methylation patterns using the Illumina 450k methylation array in a cohort of 210 GBM from children and adult patients 62 sub‐classified GBM into at least six distinct epigenetic groups, indistinguishable by histological appearance, but correlating with molecular genetic factors as well as key clinical variables such as patient age and tumor location. Indeed, IDH1 mutations characterized, as expected, a mutation‐defined subgroup which included pediatric IDH‐mutant GBM, while each H3F3A mutation defined an epigenetic subgroup of GBM with a distinct global methylation pattern. The last three epigenetic subgroups were enriched for hallmark genetic events of adult GBM and/or established transcriptomic signatures. In line with reports of PDGFRA copy number alterations being more prevalent in childhood high‐grade gliomas 2, 52, 57, an epigenetic cluster, we named receptor tyrosine kinase (RTK) I “PDGFRA,” harbored a proportion of pediatric patients (median age 36 years, range 8–74 years). A mesenchymal cluster displayed a widespread age distribution (median age 47, range 2–85 years), while an RTK II “Classic” cluster was mostly composed of older adults (median age 58, range 36–81 years) patients. Interestingly, the two H3F3A mutations gave rise to GBMs with differential regulation of transcription factors—OLIG1, OLIG2 and FOXG1, which may reflect different cellular origins. Last, our results demonstrate differential survival for each of these six subgroups, with worse prognosis and rapid death for K27M tumors, which behave like DIPG regardless of their localization within the brain, and a slightly improved survival for IDH‐mutant GBM which show the longest overall survival followed by G34R/V‐H3.3 GBM, while wild‐type tumors for these mutations have the expected bad prognosis seen in GBM.

Conclusions and Future Perspectives

Currently, there are no effective treatments for pediatric HGA which carry the highest rate of morbidity and mortality in childhood cancer. Collectively, recent findings have changed the landscape for this disease, providing at last relevant targets that will drive new trial models based on well‐characterized gene subsets, thus providing hope for future change in the management and outcome of this deadly group of cancers. Defects in chromatin remodeling are central in the genesis of pediatric and young adult HGA, and age and brain‐location specific defects in chromatin structure underlie the genesis of these tumors. This provides a launching point for the identification of diagnostics and therapies specific to the molecular subtypes in children and young adults. Although histologically indistinguishable, there are at least six distinct disease entities that may arise from separate cell types of origin as the result of largely non‐overlapping sets of molecular events. The development of the normal brain is a very dynamic and complex process, involving numerous extracellular factors that are present at precise times in specific brain locations. This is of major significance as the human brain continues to develop post‐natally, reaching completion in the early to mid‐20s. What is unique about the developing brain that enables K27M‐H3.3 mutations to be tumorigenic mainly in children and G34R/V‐H3.3 and IDH mutations to affect mainly adolescents and younger adults? Why the need for ATRX mutations that seem to be mutation (G34R/V‐H3.3 and IDH more than K27M‐H3.3), and/or location (cortex and thalamus more than the brainstem) and/or age (older more than young children) specific? Why the need for added TP53 mutations? Is it possible that the different extracellular factors in the developing brain tissue contribute to the transformation of cells with K27M‐H3.3 and G34R/V‐H3.3 mutations, allowing them to form tumors? The timing, the contribution and the genetic program that are altered by these mutations/combinations of mutations need further investigation to help address and better target their effects in HGA. Further dissection of the effects downstream of each H3F3A mutation in primary tumors and the generation of much needed “in vitro” and “in vivo” models of these mutations and their interaction(s) with ATRX and TP53 alterations are thus warranted. This will help us better tackle this cancer as uncovering what they induce on the epigenetic/genetic level has major potential to improve our understanding of HGA, brain development and potentially other diseases affecting the epigenome. Optimal clinical management should account for the distinction between these GBM disease subtypes, and as routine histopathology is unable to distinguish these genetic subgroups, use of molecular tools would result in better stratification of patients and enable better therapeutic choices as they become available.

Conflict of Interest

The authors declare no conflict of interest.