Increased 5-hydroxymethylcytosine and decreased 5-methylcytosine are indicators of global epigenetic dysregulation in diffuse intrinsic pontine glioma

Tumors of the central nervous system are the second most common malignancy in pediatric patients. Diffuse intrinsic pontine gliomas (DIPGs) comprise approximately 10% of pediatric brain tumors and are universally fatal[1]. Due to their location, DIPGs have rarely been biopsied at diagnosis outside of recent clinical trials. No improvements in DIPG outcome have been noted in the past 20 years, perhaps due to the paucity of patient-derived samples and lack of molecular insights leading to novel treatments. Encouragingly, recent molecular and proteomic analyses of autopsy specimens have identified key genetic alterations in DIPG, including amplifications in genes encoding receptor tyrosine kinases (PDGFRA, MET)[2], PDGFRA mutations[3], as well as distinct DIPG subgroups based on Hedgehog (SHH) and MYCN pathway activation[4]. Mutations in the ACVR1 gene, encoding the transforming growth factor-beta (TGF-beta) superfamily member activin, have been reported in approximately 20% of DIPGs[5, 6].

Epigenetic research has added to our understanding of how chromatin remodeling by methylation and acetylation of histones affects gene expression in tumors[7]. Among brain tumors, glioblastomas (GBMs) can be subdivided into 6 groups based on DNA methylation patterns. DIPGs are classified within one group that has relatively hypomethylated DNA and are associated with mutations in genes encoding for histone proteins[8‐11]. Approximately 60% of DIPGs have a mutation in the H3F3A gene which encodes the variant histone 3.3 proteins[11], which is also associated with a worse prognosis[12]. Less frequently, DIPGs have mutations in HIST1H3B, encoding for histone H3.1, while other extrapontine pediatric high-grade gliomas may have alternative mutations in H3F3A (G34R or G34V)[13, 8, 9]. As a result of the common missense mutation (H3F3A), lysine at position 27 (H3K27) is changed to methionine (K27M) in the amino terminal tail of histone 3.3[10]. Post-translational modification of this critical histone regulates gene expression, DNA repair, and maintenance of centromeres/telomeres[8]. Methylation of the H3K27 is mediated by polycomb repressive complex 2 (PRC2), which binds to polycomb group (PcG) target genes and induces the PRC2 component enhancer of zeste homologue 2 (EZH2) methyltransferase to methylate H3K27. Histone methylation creates a repressed state by inhibiting transcription, altering chromatin compaction, and affecting recruitment of DNA methyltransferases (DNMTs)[8, 14]. The H3K27M mutant histone inhibits PRC2 by interacting with EZH2 and suppressing its function[13]. Therefore, as a result of the K27M mutation, histone methylation is decreased at this site[8, 9, 13, 10]. H3.3 is primarily found at transcription sites and telomeres, and the H3K27M mutant H3.3 is associated with transcriptionally active (open) chromatin[8]. Inhibition of histone K27 tri-methylation (H3K27me3) leads to global activation of transcription[8, 13].

One of the most studied epigenetic alterations is methylation of DNA. Mammalian DNA methylation occurs primarily at the 5-position of cytosines (5mC) in CpG dinucleotides. Methylated cytosines at gene promoters with high GC content (CpG islands) are usually associated with transcriptional silencing. Loss of 5mC leads to a redistribution of PRC2 complexes which suggests that 5mC could affect interactions between PRC2 and chromatin[15]. Aberrant recruitment of PRC2 to DNA not usually associated with H3K27me3 may shift PRC2 away from original targets and again promote an active transcriptional state. H3K27me3 is not uniformly increased in these hypomethylated DNA regions, which provides evidence that DNA methylation is one of many epigenetic factors affecting PRC2 function[15, 10]. A hypomethylated genome, in addition to the H3F3A mutation, likely accentuates the transcriptionally active state of DIPGs by disrupting histone methylation at H3K27me3.

The methylation of cytosine in CpG islands is a modification produced by DNMTs. Reversal of 5mC methylation is accomplished in a multistep enzymatic process using Ten Eleven Translocation (TET) enzymes, thymine DNA glycosylase (TDG), and base excision repair (BER)[16, 17]. TET enzymes can convert 5mC in a reaction dependent on alpha-ketoglutarate (α-KG) to 5-hydroxymethylcytosine (5hmC)[16, 18]. 5hmC can then either be further processed by TDG and BER or persist as 5hmC in mammalian genomes[19]. Decreased 5hmC levels have been described in a variety of cancers[20], as well as in high-grade gliomas[21]. 5hmC is often associated with the gene bodies of actively transcribed genes and is considered an epigenetic mark in its own right[17].

These studies raise the possibility that unregulated loss of H3K27me3 through H3K27M mutation and elevated 5hmC could shift normal development into a pathologic state. The association between loss of H3K27me3 and elevated 5hmC in neural development also suggests a regulatory cross talk between these two pathways. Histone 3 lysine 9 methylation (H3K9me3) is another important histone methylation mark implicated in the development of gliomas. Methylation at this site affects global DNA methylation, chromatin compaction, and transcription[22].

In this study we compared epigenetic alterations between DIPG and GBM with respect to patient age and tumor geographical location using archival formalin-fixed paraffin-embedded material to gain a better understanding of epigenetic alterations specific to DIPG.

We determined by IHC relative levels of H3K27me3, 5mC, and 5hmC in DIPG, pediatric GBM, and adult GBM tissue. Our data are consistent with previously reported low levels of H3K27me3 due to the H3K27M mutation in the majority of DIPG. H3K9me3 in DIPG was not significantly different from H3K9me3 immunoreactivity in extrapontine GBM and unlikely to be playing a role in the tumorigenicity of DIPG. Surprisingly, 5hmC immunoreactivity was significantly elevated in DIPG, in contrast to 5mC, suggesting that imbalance between 5hmC and 5mC plays a global role in the biology of DIPG. In general, high levels of 5hmC have been shown to be a feature of terminally differentiated cells[20]. In numerous solid tumors including carcinoma of the breast, prostate, colon, melanoma, and gliomas, 5hmC levels appear to be greatly reduced in neoplastic cells[20]. Our group has previously shown that in adult GBM and anaplastic astrocytoma high levels of 5hmC are associated with a less aggressive phenotype[21]. However, the global DNA hypomethylation that occurs in DIPG in conjunction with relatively increased 5hmC may represent signs of a novel global epigenetic dysregulation state distinct from that of adult high grade glioma. The elevation of 5hmC in DIPG must be considered in the context of global loss of 5mC in these tumors. The striking imbalance between 5hmC and 5mC in DIPG may be another sign of the marked epigenetic dysregulation which underlies these aggressive tumors.

A limitation of our study is the relative lack of clinical data available for the patient samples. We do not know which, if any, chemotherapeutic agents patients have been treated with and whether they have received radiation. Our findings may be affected by prior treatment but the internal consistency of our findings within tumor groups supports their significance. The underlying cause for the shift in balance towards increased 5hmC expression relative to 5mC in DIPG compared to normal brain and extrapontine GBM is unclear. However, one possibility is altered activity of TETs. Preliminary evaluation of TET mRNA expression shows elevation of TET1 and TET3 in DIPG compared to control brain. In other cancer types, TET is mutated and inactivated, leading to globally decreased 5hmC and increased DNA methylation[17]. In contrast, a possible result of the loss of global DNA methylation could be deposition of the 5hmC intermediary and epigenetic imbalance.

During differentiation, global levels of 5hmC and H3K27me3 are tightly co-regulated and in most solid tumors both marks are concordantly reduced[24]. The data presented here suggests that DIPGs have a unique epigenetic fingerprint featuring high 5hmC and loss of H3K27me3. This epigenetic constellation may contribute to DIPG tumor initiation and maintenance. The alteration in H3.3 which inactivates EZH2 and global DNA hypomethylation suggests that two regulatory pathways that communicate to maintain normal brain development are altered in this tumor type, leading to an unopposed active transcriptional state (Figure 7). Elevated 5hmC and depressed H3K27me3 in normal brain development are associated with neural differentiation and affect neural migration through inhibition of EZH2 (part of the PRC2 complex)[20, 18]. We hypothesize that migration is aberrant in DIPG since EZH2 is no longer able to function at Lysine 27 in Histone 3.3. Elevated 5hmC is found in differentiating cells and usually in conjunction with H3K27me3 loss. There is a possibility that the Histone 3 and cytosine methylation pathways cross-talk and co-regulate each other. However, in DIPG, aberrant TET enzyme expression may result in loss of 5mC and increased 5hmC. This may lead to aberrant PRC2 sequestration from primary targets, a transcriptionally active state, and tumorigenicity.

Histone and cytosine methylation dysregulation is unique to DIPG. When neither pathway is able to function in its regulatory role in neural development, DIPG develops. This epigenetic imbalance may underlie DIPG tumor formation and resistance to treatment. Our work supports recent findings indicating that H3K27me3 loss, through the dominant negative H3K27M mutation, is associated with DNA hypomethylation and an increase in transcriptionally upregulated genes. This suggests that there is crosstalk between histone methylation pathways and DNA methylation pathways leading to changes in transcriptional activity[10].

The H3.3 Lysine 27 is a critical site for the deposition of an inhibitory epigenetic mark and is mutated in the majority of DIPG, suggesting that targeting epigenetics could be one therapeutic approach for this highly aggressive pediatric tumor. Understanding the epigenetic landscape of DIPG opens up the opportunity for epigenetic modifiers, which could potentially shift the active genome of this deadly tumor into a silent and regulated state. The role of epigenetic modifiers such as methylation inhibitors and histone deacetylase inhibitors[7] in treating DIPG needs to be reevaluated as the epigenetic mechanisms of DIPG are better understood.