Oncohistones in brain tumors: the soil and seed

Abstract

Recurrent somatic mutations in histone H3 variants (termed oncohistones) have been identified in children and young adult high-grade gliomas and induce tumorigenesis through disruption of chromatin states. Oncohistones occur with exquisite neuroanatomical specificity and are associated with specific age distribution and epigenome landscapes. Here, we review the known intrinsic (“seed”) and the extrinsic (“soil”) factors needed for their optimal oncogenic effect and highlight the many unresolved questions regarding their effects on development and crosstalk with the tumor microenvironment. The “seed and soil” analogy, used to explain tumor metastatic niches, also applies to oncohistones, which mainly thrive and flourish in specific chromatin states during very narrow windows of development, creating exquisite vulnerabilities; which could provide effective therapies for these deadly cancers.

Oncohistones in brain tumors: a narrow window of opportunity in the right “soil and seed”

In 2012, two landmark papers described recurrent somatic mutations in hotspots within the histone 3 (H3) tail leading to amino acid substitutions at lysine 27 to methionine (K27M) and glycine 34 to arginine/valine (G34R/V) [1, 2]. These mutations, also known as oncohistones (see Glossary), give rise to deadly brain tumors [high-grade gliomas (HGGs)], in children and young adults, which are the most common type of central nervous system neoplasm and a leading cause of mortality and morbidity in childhood cancer. Outside of HGGs, oncohistones have been uncovered as the initiating driver event in other cancers. Glycine to tryptophan (G34W), and rarely to leucine, arginine, valine or methionine (G34L/R/V/M), in H3.3, constitute the singular driver mutation in >90% of giant cell tumours of bone [3]; H3 K36M, and more rarely to isoleucine (K36I), mutations occur in 95% of chondroblastomas [3], and also account for a subset of undifferentiated soft tissue sarcomas [4] and human papillomavirus (HPV)-negative head and neck squamous cell carcinomas (HNSCCs) [5]. Additionally, in osteosarcomas, ~6% of tumors carry either H3K27M, H3.3G34R or G34W hotspot mutations [6], while H3K27M/I also occur in rare cases of acute myeloid leukemia [7, 8] and Group A posterior fossa ependymomas (PFA-EPN), another deadly childhood brain tumor [9, 10]. Strikingly, K-to-M mutations target canonical H3.1 or H3.2 (cell-cycle dependent H3 variants) and non-canonical H3.3 (non-cycle dependent H3 variant), while amino acid substitutions on G34 exclusively occur on H3.3. Also, recent studies identified that the expression of the uncharacterized gene CXorf67 (now designated EZH2 inhibitory protein - EZHIP), which is normally restricted to germ cells, predominates in PFA-EPN and is mutually exclusive with H3K27M, while the rare H3 wild-type diffuse midline gliomas also exhibit EZHIP expression [11–13], suggesting convergence of effect between the H3K27M mutation and EZHIP.

Since the discovery of “oncohistones”, intense research efforts have led to significant understanding of their biochemical properties and their effect on chromatin-modifying proteins (reviewed in [14–16]). More recently, studies have examined the context dependency required for neoplastic transformation by histone mutations. The emergent paradigm supports a “seed and soil” model, in which the H3 mutations are acquired in a specific permissive cellular context (“cell-of-origin”) within the brain and interrupt epigenetic reprogramming necessary for further development and differentiation; this developmental arrest leads to eventual acquisition of additional partnering mutations and finally to neoplastic transformation. In this review, we explore how these intrinsic (seed) and extrinsic factors (soil) contribute to context-dependency of H3.3/H3.1/H3.2K27M, H3.3G34R/V and EZHIP in brain tumors, specifically discussing recent literature on cell-of-origin determinants and crosstalk with the tissue microenvironment in H3-mutant glioma. We will also describe ongoing and future therapeutic initiatives that will be fast tracked by novel, more relevant, pre-clinical models of oncohistones and the oncohistone mimic EZHIP.

Oncohistones show spatiotemporal specificity and a two-step oncogenic transformation through nonrandom partnership with genetic alterations in HGGs

H3-mutant and EZHIP-expressing HGG occur in non-overlapping patients and show exquisite specificity in clinicopathological features including anatomical location, age-of-onset, and in other co-occurring oncogenic mutations [1, 2, 15, 17–19] (Figure 1). These alterations are clonal in tumors being present in every tumor cell, and invariably associate with additional partner genetic alterations that are necessary to promote full blown tumorigenesis [20, 21].

Figure 1 – Spatiotemporal specificity of oncohistone and the oncohistone mimic EZHIP in childhood brain cancer.

Recurrent histone H3 mutations or expression of the oncohistone mimic EZHIP defines distinct brain tumor subgroups in children and young adults. Most high-grade gliomas (HGGs) occurring in the midline location (such as the thalamus, pons, and posterior fossa) carry the K27M mutation on either canonical histone H3.1 or non-canonical H3.3, or express the K27M oncohistone mimic endogenous gene EZHIP. In the thalamus, H3.3K27M mutations are found concurrently with either mutations in TP53 or FGFR, and altogether accounts for 60% of HGGs in this location. In the pons, H3.3K27M mutations are coupled to TP53 or PPM1D, while H3.1K27M tumors almost invariably bear ACVR1 activating mutations, altogether accounting for > 90% of pontine gliomas. Expression of EZHIP oncohistone mimic, or more rarely H3.1K27M mutation define 95% of posterior fossa ependymoma subgroup A (PFA-EP). In the cerebral cortex, H3.3G34R/V HGGs invariably bear mutations in TP53 and ATRX. Temporally, EZHIP positive PFA-EP are diagnosed in infants and toddlers, H3K27M in toddlers, young children, and more rarely in adults, while H3.3G34R/V are found in adolescent teens and young adults.

H3K27M (very rarely H3K27I), and EZHIP populate the brain midline (mainly pons, thalamus, and spine) where more than 95% of HGGs carry these genetic alterations and are clinically diagnosed as diffuse midline gliomas (DMGs), in the new 2021 World Health Organization (WHO) classification of tumors of the central nervous system [22] (Figure 1). The K27M mutation can occur on either canonical H3 variants H3.1 (encoded by HIST1H3B and more rarely HIST1H3C) or very rarely in H3.2, or on the non-canonical variant H3.3 encoded by H3F3A. H3.1/H3.2K27M-expressing tumors are restricted to the brain pons, affect young children between the age of 2–6 years, and often bear activating mutation in the growth factor receptor activin A receptor type I (ACVR1, reviewed in [18]). In contrast, H3.3K27M-expressing tumors are found across the brain midline, affect older children and adults (7 – 55 years of age) [23], and often partner with loss-of-function mutations in TP53 and genetic alterations in platelet-derived growth factor receptor (PDGFRA), mainly in the form of genomic amplifications [18, 24, 25]. In the absence of TP53 and PDGFRA alterations, H3.3K27M co-opt other growth factors in a brain location dependent manner. Indeed, activating genetic alterations in growth factor receptors, including EGFR and FGFR1, mainly in the thalamus, or in the PI3K signaling pathway, mainly in the pons, also associate with H3K27M or EZHIP and participate in the oncogenic process [18, 24]. Other genetic alterations including MET or NTRTK fusions, or BRAF mutations/fusions have also been identified in rare DMGs while sub-clonal mutations, especially in the PI3K signaling pathway, provide added intra-tumoral heterogeneity to the inter-tumoral heterogeneity [14, 26].

In contrast to the midline location of H3K27M mutations, H3.3G34R/V mutations are exclusive to the brain cortex, occurring in the temporo-parietal hemispheres in adolescent teens and young adults (12 to 35 year of age) where they account for ~30% of HGGs [27–29]. These mutations target exclusively H3.3, and, like the K27M amino acid substitution, require additional oncogenic partnership [1]. Indeed, H3.3G34R/V HGG invariably bear loss-of-function mutations on tumor suppressor genes TP53 and ATRX [1, 19, 30–32], and frequently couple to high expression and activating mutations in PDGFRA [28] and in rarer cases to other genetic alterations including MET fusions (Figure 1). Similarly, H3.3G34R/V mutations are clonally present in every tumor cell, and their partnership with other genetic alterations, when acquired, remains constant at all stages of the disease. However, while H3K27M are needed to maintain oncogenicity [33, 34], G34R/V mutations seem dispensable for tumor maintenance as their removal had limited effect on tumor growth [28]. Collectively, evidence suggest oncohistones are poorly oncogenic on their own in several models and need nonrandom partnership mutations to induce tumor formation [21, 35, 36], indicating a multistep oncogenic transformation process in HGGs.

Histone mutations alter the pre-existing epigenetic landscape

Despite only contributing to a minor proportion of the total histone pool (less than 20% of the total H3 pool), oncohistones alter the chromatin landscape, and subsequently affect transcriptional capacity. Biochemically, the K27M mutation acts as a potent dominant-negative mutant on Polycomb Repressive Complex 2 (PRC2) and inhibits global H3K27 methylation including its trimethylation (H3K27me3) [37], a histone modification associated with transcriptional repression of gene promoters (Figure 2). This mutation acts like a “molecular poison” and largely inhibits the catalytic function of EZH2, the PRC2 methyltransferase, especially the conversion of di- to tri-methylation, which is enzymatically tasking [38, 39]. Indeed, while catalysis of the mono and di-methylated forms of H3K27 is rapid, deposition of tri-methylation occurs with much slower kinetics and requires allosteric activation of EZH2 by the EED subunit of the PRC2 complex [40]. Deposition of H3K27me1 and me2 is less affected by H3K27M and the spread of H3K27me2 depends on H3K36me2 boundaries [41, 42] and cell cycling rate [43]. The oncohistone mimic EZHIP is an endogenous gene expressed in select tissues of placental mammals, with generally poor conservation across species, except for a highly conserved 12 amino acid sequence in its C-terminus [13, 44]. This conserved peptide sequence remarkably mirrors the sequence surrounding K27 on the H3 tail, and, except for a methionine in position 406, has the calculated “optimal” substrate sequence previously defined for PRC2. Notably, M406 mimics the K27M mutation on the H3 tail, and the resulting peptide sequence in both EZHIP and K27M-mutant H3 acts as a potent inhibitor of PRC2 catalytic activity [13, 44]. Indeed, mass spectrometry data indicate K27M or EZHIP expression lead to severe reduction of global H3K27me3 and, to a lesser extent, H3K27me2, while acetylation of H3K27 (H3K27ac) is increased in H3K27M HGGs [13, 34, 43, 45, 46].

Figure 2 – Histone mutations alter the pre-existing epigenetic landscape.

The trans-acting K27M mutation, or the K27M-mimic EZHIP, acts as a dominant negative mutant on Polycomb Repressive Complex 2 (PRC2), which deposits the H3K27 methylation. In K27M- or EZHIP-expressing cells, this leads to dramatic loss of H3K27me3, and me2/me1 to a lesser extent. On the genome, H3K27me3 is limited within the confinement of CpG Islands, which are nucleation sites for PRC2. H3K27me2/me1 domains are contracted, and allows for the inappropriate propagation of activating modifications, such as H3K27ac, to spread in intergenic regions. In contrast, H3.3G34R/V mutations act only in cis – impairing the adjacent K36 methylation by enzymes such as SETD2 and possibly NSD1/2. As the non-canonical histone H3.3 is deposited in selected regions of the genome (e.g. active genes/enhancers, repetitive DNA) this leads to localized loss of H3K36me2/3 in H3.3G34R/V-expressing cells. Furthermore, the loss of the H3.3 chaperone ATRX in H3.3G34R/V tumors may lead to preferential deposition of the mutant histones in actively transcribed regions and enhancers.

H3K27me3 is not completely absent in H3K27M expressing cells – and how remnant H3K27me3 is deposited in the presence of this oncohistone has been the subject of controversy (reviewed in [16]). Current data suggest that the PRC2 complex is normally recruited to its nucleation sites, predominantly unmethylated CpG islands (CGI) gene promoters. From these CGIs, this complex is unable to spread the H3K27me3 mark beyond these sites [34, 43]. H3K27me2, an abundant mark (50–70% of all H3) that coats intra and intergenic regions, is depleted in the presence of K27M oncohistones, especially in rapidly cycling cells. Contraction of H3K27me2/me3 initiates a downstream effect on epigenetic alterations, with the accumulation of the activating H3K27ac mark in intergenic areas [43, 46] and spreading of H3K36me2 domains in regions that no longer bear H3K27me2/3 [41, 42, 48]. Genes and repetitive elements embedded within the edge of H3K27me3 domains are subsequently more susceptible to aberrant activation in the presence of the K27M mutation, whereas genes at nucleation sites are tightly repressed [41, 46] (Figure 2). Thus, the specific H3K27ac enhancer/promoter landscape in K27M-mutant cells reflects the lineage of origin and is not a consequence of the oncogenic effect of the mutation [43, 46]. Notably, genes that become transcriptionally dysregulated in K27M-mutant gliomas depend on the pre-existing landscape of H3K27me3 and other chromatin modifications in the cell-of-origin. Last, despite severely decreased H3K27me3 deposition in K27M and EZHIP mutant gliomas, there are limited changes in chromatin architecture and 3D patterning of HOX genes [43]. In summary, the unifying action of K27M mutations is to restrict H3K27me3 at PRC2 landing sites, while other epigenetic changes are mainly contingent on the cell-of-origin chromatin state and cycling rate. How these other epigenetic changes participate in the oncogenic process requires further investigation.

In contrast to the trans acting mechanism of K-to-M mutations, H3.3G34R/V mutations impair the modification and recognition of the adjacent H3K36 residue in cis. H3K36 methylation exists as multivalent forms which are functionally distinct. Tri-methylation of H3K36 (H3K36me3) is deposited by SETD2 at transcribing gene bodies, whereas di-methylation of H3K36 (H3K36me2) is deposited at intergenic regions by multiple enzymes including NSD1/2 (reviewed in [49]) (Figure 2). All H3K36 lysine methyltransferases share the highly conserved catalytic SET domain, which recognizes the H3K36 residue through a narrow channel - and thus glycine substitution to bulkier amino acid e.g. arginine/valine potentially impairs entry of the histone tail and H3K36 methylation. Accordingly, H3.3G34R/V histone tails show severe loss of H3K36 tri-methylation, and to a lesser extent di-methylation [50]. As H3.3G34R/V mutation occur exclusively on H3.3 and acts predominantly in cis, the epigenetic perturbation of this histone is likely intimately linked to the deposition pattern of the histone variant it targets. H3.3 is incorporated into chromatin in a replication-independent manner by H3.3 chaperone complexes. The co-occurring loss of ATRX in H3.3G34R/V gliomas may therefore skew deposition of H3.3G34 mutant histones into active gene bodies to synergize inhibition of SETD2-mediated H3K36me3, a hypothesis supported by the presence of HGG with truncating loss-of function mutations of SETD2, which are mutually exclusive with H3.3G34R/V and occur in similar brain locations [51]. However, the transcriptional consequences of H3.3G34R/V mutations remain incompletely understood as does the requirement for ATRX and TP53 in promoting tumor formation.

Cell-of-origin context potentiates oncogenic transformation by H3 mutations

Multiple oligodendroglial progenitors as cells-of-origin for H3K27M glioma

The exquisite and non-overlapping anatomical/temporal specificity and partnership mutations of H3.1/H3.2K27M, H3.3K27M and H3.3G34R/V gliomas suggest distinct cell(s)-of-origin (Figure 3). Experimental modeling of K27M in mice, has shown that the K27M mutation leads to embryonic lethality in totipotent stem cells (zygote - morula), but is tolerated in a neural progenitor cell context [21]. This early neural progenitor context, together with synergistic action of additional oncogenic mutations, can lead to neoplastic transformation and HGG [38, 52]. More recently, bulk and single cell transcriptome profiling of primary K27M tumors have refined this origin within a committed progenitor in the oligodendroglial lineage - K27M tumors bear a cellular hierarchy resembling the developmental trajectory of neural progenitor cells (also known as radial glial cells, RGCs) to oligodendroglial progenitor cells (OPCs) to oligodendrocytes [26, 43, 53] (Figure 3). In normal development, OPCs originate in neuroepithelial zones surrounding the ventricles, and are specified by lineage transcription factors including OLIG1 and OLIG2 - which are highly expressed in K27M tumors [19]. Unlike other cell types in the brain, OPC specification occurs in waves that are spatially separated in the fetal brain - with the first originating in the ventral subventricular zone (SVZ) and shifting dorsally with the lateral wall in the second, and dorsal SVZ in the third wave, coinciding with the myelination of the brain from midline structures outwards [54]; this potentially explains the midline location of K27M tumors, but it remains unknown how K27M-mediated epigenetic perturbation prevents normal differentiation of oligodendrocytes.

Figure 3 – Cell-of-origin context potentiates oncogenic transformation by histone mutations.

The K27M midline gliomas likely originate in distinct committed oligodendroglial progenitors (OPC) in the midbrain and hindbrain, showing high expression of canonical OPC markers such as OLIG2 and PDGFRA. The H3.3K27M thalamic HGGs express the developmental patterning factors (e.g. GBX2) found in prosomere 2 (one of the segmental units that forms the developing diencephalon) in the midbrain, while pontine H3.3/H3.1K27M HGGs express patterning of rhombomeres 5–8 in the hindbrain. More specifically in the hindbrain, H3.3K27M pontine tumors show resemblance to neuroepithelial progenitors from the dorsal progenitor zone expressing PAX3/7, while H3.1K27M, ACVR1 pontine gliomas show similarities to ventral oligodendroglial progenitors expressing NKX6–1. In contrast, the H3.3G34R/V HGGs are derived from a forebrain interneuron progenitor in the lateral ganglionic eminence, expressing FOXG1, DLX1/2. Oncohistone-mediated oncogenic transformation in these lineage remains incompletely understood, but likely relies on synergistic partnering mutations such as ACVR1 and PDGFRA. RGC: radial glial cells.

Notably, K27M-mutant gliomas faithfully maintain chromatin configuration at developmental genes consistent with anatomically distinct oligodendrocyte-precursor-like cells [43]. H3.3K27M thalamic gliomas are derived from prosomere 2, which is one of the three segments that develops rostro-caudallly to form the diencephalon. It is likely that within this unique developmental origin, the K27M tumors acquire alterations in resident growth factors (e.g. PDGFRA, FGFR, EGFR), that may be necessary for optimal oncogenic effect. In turn, H3.1/H3.2K27M ACVR1-mutant pontine gliomas uniformly mirror early ventral NKX6–1+/SHH-dependent brainstem OPCs, while H3.3K27M gliomas in the pons frequently resemble dorsal PAX3+/BMP-dependent progenitors. This suggests a context-specific vulnerability in H3.1/H3.2K27M-mutant SHH-dependent ventral OPCs, which rely on acquisition of ACVR1 mutations to drive aberrant BMP signaling required for oncogenesis [43]. Indeed, the K27M mutation likely synergizes with oncogenic partners for OPC differentiation blockade - ACVR1 mutations have been shown to induce expansion of OLIG2+ progenitor pool in the neonatal brain through hyperactive BMP signaling [36], while amplification of PDGFRA, the main growth factor in the oligodendroglial lineage, has been shown to promote gliomagenesis [21, 55].

While H3K27M-mutant DMGs are observed in adults (below 50), these tumors are comparatively rare in this age group, and mainly occur in the thalamus. A recent study indicate that, even though they also arise in OPC-like cells, they have a mesenchymal signature potentially induced by specific microenvironmental cues associated with age [26]. Further studies are needed to understand how developmental processes altered by the H3K27M oncohistone can also manifest, even if more rarely, in adulthood.

Interneuron progenitor origin enables oncogenic co-option in H3.3G34R/V glioma

Recent data showed that H3.3G34R/V gliomas in fact originate in interneuron progenitors and masquerade as glial tumors [28]. These HGGs uniquely express a developmental gene expression program resembling a forebrain interneuron progenitor, which likely reflects the cell-of-origin for this glioma subgroup [28, 56, 57] (Figure 3). H3.3G34R/V exert distinct transcriptional/phenotypic effect in a brain region-specific manner and seem to depend on the forebrain-specific identity factor FOXG1 for oncogenicity [57]. Specifically, H3.3G34R/V gliomas express a transcription factor network of homeobox genes, including GSX2, that normally specifies the interneuron fate commitment and actively prevent differentiation towards the oligodendrocyte lineage in the developing brain. Importantly, H3.3G34R/V gliomas likely depends on this forebrain interneuron progenitor identity for neoplastic transformation as they can hijack there the PDGFRA oncogene through a pre-existing chromatin conformation of the adjacent GSX2 lineage factor to promote its aberrant expression [28].

In all, integrating single-cell transcriptomics and epigenomic datasets on oncohistone mutant tumors [26, 28, 43, 58] and other pediatric brain tumors [58–60] indicate a common trend: tumor cells are stalled in development by an intiating epigenetic driver event and are unable to differentiate further. In the context of HGGs, this seems to be the first needed step to oncogenesis. We speculate that acquisition of the nonrandom partnership genetic alterations associated with oncohistones in HGGs is aimed to maintain/stabilize this progenitor state and to promote full blown gliomagenesis. We also propose that H3K27M and EZHIP can mainly exert their effects on early progenitors that have not yet acquired and spread large amounts of H3K27me3, while H3.3G34R/V are oncogenic only in specific interneuron progenitors possibly because they can co-opt PDGFRA signaling in this cell lineage. Both concepts require further validation in future models.

Oncohistone tumor microenvironment in HGGs

Besides intrinsic contingencies of cell state and origin, extrinsic biological processes may promote survival and more aggressive behavior of oncohistone-mutant HGGs. Neuronal activity has emerged as a fundamental factor (reviewed in [61]). Indeed, glutamatergic neuronal activity promotes glial precursor cell proliferation [62], and more recently, it was reported that glial cancer cells are able to form functional synapses with neurons promoting growth, migration and brain colonization [63, 64]. The pivotal connection between brain networks and tumor formation needs further investigation as it may provide needed therapeutic avenues that can modulate tumor growth as recently shown for the AMPAR-blocking anti-epileptic drug perampanel [63]. Further investigations are needed to assess the extent on which HGGs rely on neural network for their growth.

It is widely recognized that immune cells play a major role in tumor establishment and progression including in the brain (reviewed in [65]). However, limited information regarding the tumor immune microenvironment exists for pediatric gliomas, including oncohistone-mutant tumors. The few published studies focused on profiling K27M DMG use limited immune markers and report the presence of rare T-cells and mainly infiltration by tumor-associated macrophages, similar to what has been observed in adult H3-wild type HGG [66, 67]. More in-depth characterization of the microenvironment in H3K27M, including evaluation of differences (if any) between H3.1 and H3.3K27M HGGs, and immune profiling of G34R/V tumors are needed. To this effect, more recent technologies, such as spatial transcriptomic and imaging mass cytometry, are expected to advance our understanding of intercellular communication of tumors. When modeling HGGs, most studies to date have used patient-derived xenografts in immunocompromised murine models; while these are extremely helpful for the tumor profiling, the immune landscape cannot be properly evaluated or modulated in these models.

The recent cell/lineage-of-origin identification of oncohistone-mutant tumors will promote the generation of more relevant and effective pre-clinical models including syngeneic mouse models that will allow better exploration of the effect of these mutations and how they interact with the microenvironment they arise in.

Targeting oncohistone-mutant HGGs

While overall survival rates for childhood CNS tumors have substantially improved, HGGs, including DMG, still have dismal overall survival [18, 19, 32]. Clinical trials in HGGs and DMGs are expanding with the knowledge accrued from recent studies; we will review here ongoing trials and a few targets that show promise in oncohistone-mutant gliomas (summarized in Figure 4).

Figure 4 -. Microenvironment and targeted therapies are investigation avenues for oncohistone mutant tumors.

Oncohistone H3 mutant tumors are regulated and driven by different oncogenic pathways, such as PDGFRa in H3.3G34R mutant tumors, and BMP signaling in H3K27M tumors; and represent potential targets for future therapies. Epigenetics events, such as DNA methylation and histone acetylation/methylation, are found to be dysregulated in these tumors, and the use of chromatin modifiers such as BAF, EZH2, HDAC and DNA methylation inhibitors are potential approaches. Lastly, a comprehensive understanding of tumor dependencies, immune infiltration and regulation, cancer-brain cells interaction, and appropriate model systems are a promising area of focus to be addressed for these tumors. ACVR1: Activin A Receptor Type 1; ADAM10: A disintegrin and metalloproteinase 10; ALK2: Activin receptor-like kinase-2; AKT: Protein kinase B; BAF: BBRG1-associated factors inhibitor; BMP: Bone morphogenetic proteins, BNDF: Brain-derived neurotrophic factor; CAR: Chimeric antigen receptor; EGFR: Epidermal growth factor receptor; ERK: Extracellular signal- regulated kinases; EZH2i: Enhancer of zeste homolog 2 inhibitor; FGFR: Fibroblast growth factor receptors; HDACi: Histone deacetylase inhibitor; MEK: Mitogen-activated protein kinase kinase; NLGN3: Neuroligin-3; p: phosphorylated; PDGFRA: Platelet- derived growth factor receptor alpha; RAF: Rapidly Accelerated Fibrosarcoma; RTK: Receptor tyrosine kinases; STAT: Signal transducer and activator of transcription.

Epigenetic modulation

Despite the drastic decrease in H3K27me3 levels in H3K27M-mutant HGGs, further EZH2 inhibition has been suggested to impair tumor growth [68, 69], possibly by inducing cell-cycle arrest, through relieving H3K27me3 mediated suppression of the cell-cycle regulator, CDKN2A [68]. Notably, EZH2 inhibitors have shown positive effect in childhood rhabdoid tumors of the central nervous system that have mutations in SMARCA1 and more rarely in SMARCA4, which are core components of the BAF (SWI/SNF) chromatin remodeling complex. This is of interest, especially since recent studies using CRISPR lethality screens showed that H3K27M-gliomas are dependent on SMARCA4 and other members of this complex for survival [29, 70, 71]. Panobinostat, a multi-HDAC inhibitor prolonged survival of mice in a patient-derived tumor xenograft model [72], and a phase I trial is ongoing for patients with DMG (NCT02717455). In another study, H3K27M was suggested to prime cells for viral mimicry, as loss of H3K27me3 deposition leads to pervasive intergenic H3K27 acetylation and baseline expression of endogenous retroviral elements (ERVs). The use of panobinostat in combination with DNA demethylation in mouse models led to further increases in the expression of these ERVs and affected cell growth in vitro and in vivo [46]. These studies need further validation in relevant pre-clinical models including immune competent models to support their therapeutic potential in DMG.

Growth receptor targeted therapies

Somatic activating ACVR1 mutations associate with ~ 20% of HGGs in the pons, especially H3.1K27M-mutant gliomas. These mutations are clonal and represent a potential therapeutic strategy for these tumors, especially as the use of targeted inhibitors to this receptor has shown promise in pre-clinical models [35, 36]. PDGFRA is another attractive target. It is amplified in a large number of DMGs [17, 18, 73], and overexpressed/mutated in close to 17% of hemispheric HGGs [73]. The exceptional high frequency (50% at diagnosis, 80% at recurrence) of PDGFRA mutations in H3.3G34R/V glioma is potentially clinically actionable [28]. Until recently, the poor blood brain barrier penetration of first generation PDGFRA inhibitors have limited their use in CNS cancers. Newer PDGFRA inhibitors including avapretinib have improved brain penetration and are being tested in clinical trials which include oncohistone-mutant HGGs (NCT04773782). One important consideration when targeting PDGFRA-mutant HGGs is that most mutations occur in the extracellular domain, which may render the mutant receptor insensitive to several clinically available PDGFR inhibitors [74]. Other receptors including EGFR, FGFR MET, and NTRK can be identified in oncohistone-mutant gliomas and targeted with available inhibitors. Last, activation of the MAPK pathway in most HGGs and the PI3K pathway especially in DMGs can be targeted using specific inhibitors. The effectiveness of these in prolonging survival remains unknown, especially when used in monotherapy. Future studies here as well are required to test the most optimal combinations based on the molecular profile of the histone-mutant HGG.

Microenvironment modulation & Immunotherapy

HGGs have been reported to be dependent on neuronal growth factors, such as brain-derived neurotrophic factor (BDNF) and neuroligin-3 (NLGN3), and their modulation would be an attractive therapeutic strategy to limit DMG and HGG progression [63, 64]. NLGN3 is secreted by OPCs and neurons and cleaved by the metalloproteinase enzyme ADAM10. To this effect, a phase I clinical trial has been initiated in HGGs (NCT04295759). Peptide vaccines against human H3K27M, using MHC-humanized mouse model expressing HLA-A*0201, HLA-DRA*0101, and HLA-DRB1*0101, successfully induced specific CD4+ and CD8+ T cell responses [75]. Clinical trials using peptide-based vaccines, alone or combined, are underway (NCT02960230 and NCT01130077). However, this promising approach is restricted to children carrying the HLA-A2 allele, and mid- and long-term survival data remains to be acquired. Chimeric antigen receptor (CAR)-T therapies have shown promising results against K27M mutant gliomas, and several phase I clinical trials are underway. However, few patients have been included for now and these preliminary results require further validation on a higher number of patients and with longer follow-up [76–78].

In summary, there are multiple novel therapeutic avenues that stem from recent studies and that would benefit from more stringent pre-clinical validation. As it is highly unlikely that monotherapies or one single avenue will be able to cure HGGs, the future is for combination of drugs/ cellular or immune approaches used synchronously and/or asynchronously. There is an urgent need for novel adaptative clinical trials that would tease out optimal therapeutic combinations informed by the genetic makeup of a given tumor and of its microenvironment, and where relevant biomarkers assess in real time the effectiveness of the chosen therapies.

Concluding remarks and future perspectives

The long-standing concept of seed and soil for cancer can potentially be applied in the context of oncohistones. In this analogy, the soil would reflect the specific epigenomic state of the developing brain cell, which is the fertile ground for tumorigenesis induced by mutations in histone variants, which are the seed. In contrast, other epigenomic states from different cell types or perhaps the same cell type at a different stage of differentiation may not be as fertile for tumorigenesis. Thus, the “seed and soil” reflects how oncohistones can only thrive in specific chromatin states, in exquisite cells/lineage of origin, during narrow developmental windows. Indeed. the rapid and evolving understanding of the histone-mutant tumors at the molecular level has revolutionized our understanding of their oncogenic actions, and more recently unraveled developmental origins that may be permissive to transformation by these mutations. How the underlying cell states and microenvironment contribute to malignant transformation are not yet fully elucidated. Many questions remain unanswered (see Outstanding questions). These tumors are composed of heterogeneous cell populations, suggesting the possibility of additional and complex mechanisms controlling their establishment, maintenance, and progression. While there have been some advances in other tumor types, many gaps remain in our understanding of the nature of the immune contribution within the context of onco-histone mutant brain tumors. With the increased understanding of genetic and epigenetic landscapes promoted by these histone mutations, and the availability of reliable preclinical models able to dissect developmental windows, cell-of-origin, and the optimal TME needed to promote oncogenicity, uncovering specific vulnerabilities in histone-mutant HGG is becoming a real possibility. This, will lead to effective targeted therapies that ultimately benefit patients with these deadly tumors.

Outstanding Questions Box.

• How do histone mutations affect development in the relevant lineage-of-origin? Specifically, why OPCs in K27M tumorigenesis? Are OPCs derived from different anatomical regions (e.g. forebrain, midbrain, hindbrain) equally susceptible to oncohistones? Would H3.3G34R/V mutation alone impair development of the interneuron lineage?

• How are oncohistone mutant tumors being fueled by their partnering mutations? What are the functional roles of ATRX and P53 alterations in H3.3G34R/V context?

• Why does EZHIP and H3K27M require additional mutations/genetic alterations to promote oncogenicity in HGGs, but not in PFA-ependymoma?

• What is the role of the pediatric brain microenvironment in tumor intiation and maintenance? How do tumor-initiating histone mutant cells evade from immune surveillance?

Highlights.

Epigenetics have a profound impact on cell fate and organization. Oncohistones alter the pre-existing epigenome landscape in the cell-of-origin and this modulates gene expression for pre-neoplastic transformation.

Histone-mutant brain tumors’ pathological mechanisms are dependent on the intrinsic (cell-of-origin – seed) and extrinsic (tissue microenvironment - soil) contexts. Oncogenesis additionally relies on other partnering mutations.

The ‘seed and soil’ paradigm supports a model in which H3 mutations are acquired in a specific permissive cellular context and narrow window during development which interrupt epigenetic reprogramming necessary for further differentiation, and eventually leads to neoplastic transformation.

The manner in which the microenvironment (immune and non-immune cells) recognizes the tumor initiating cells alter their development and progression. New experimental approaches that recognize and explore this crosstalk should be addressed in future studies

Glossary

Chimeric antigen receptor (CAR)-T cell

synthetic receptors inserted into the surface of a T lymphocyte. CAR-T cells are able to redirect T cells to a specific tumor antigen, promoting tumor lysis

Diffuse intrinsic pontine glioma (DIPG)/diffuse midline glioma (DMG)

aggressive high-grade glioma in the pontine area of the brainstem. DIPG was recently reclassified as DMG by the World Health Organization

Epigenetic reprogramming

global remodeling of epigenetic marks during development that may result in changes in structure and positioning of chromatin and histones. Enzymes that erase or add marks are responsible for these complex events, such as DNA methyltransferases, histone methyltransferases, and histone deacetylases

High-grade gliomas

tumors of glial cells of the brain. According to the World Health Organization, it includes anaplastic astrocytoma (grade III) and glioblastoma multiforme (grade IV)

Interneurons

multipolar neurons found between sensory and motor neurons, making up more than 99% of all neurons in the body. They are involved in all higher functions, including learning, memory, cognition, and planning

Oncohistones

somatic heterozygous mutations in histone 3 encoding genes that are associated with different tumor types. The oncohistones described, so far, as cancer drivers include mutations in histone 3 lysine 27 and lysine 36 to methionine or isoleucine (H3K27M/I; H3K36M/I), and histone variant 3.3 glycine 34 to arginine, valine or tryptophan (H3.3G34R/V/W)

Oligodendrocyte

glial cells originated from oligodendrocyte precursor cells (OPCs) that arise in specific regions of the brain and spinal cord and then disperse through the central nervous system. They are responsible for the myelination of axons

Peptide vaccine

peptides used to stimulate the immune system to destroy a target of choice, such as a tumor