Evaluation of Histone 3 Lysine 27 Trimethylation (H3K27me3) and Enhancer of Zest 2 (EZH2) in Pediatric Glial and Glioneuronal Tumors Shows Decreased H3K27me3 in H3F3A K27M Mutant Glioblastomas

Abstract

H3F3A mutations are seen in ∼30% of pediatric glioblastoma (GBMs) and involve either the lysine residue at position 27 (K27M) or glycine at position 34 (G34R/V). Sixteen genes encode histone H3, each variant differing in only a few amino acids. Therefore, how mutations in a single H3 gene contribute to carcinogenesis is unknown. H3F3A K27M mutations are predicted to alter methylation of H3K27. H3K27me3 is a repressive mark critical to stem cell maintenance and is mediated by EZH2, a member of the polycomb‐group (PcG) family. We evaluated H3K27me3 and EZH2 expression using immunohistochemistry in 76 pediatric brain tumors. H3K27me3 was lowered/absent in tumor cells but preserved in endothelial cells and infiltrating lymphocytes in six out of 20 GBMs. H3K27me3 showed strong immunoreactivity in all other tumor subtypes. Sequencing of GBMs showed H3F3A K27M mutations in all six cases with lowered/absent H3K27me3. EZH2 expression was high in GBMs, but absent/focal in other tumors. However, no significant differences in EZH2 expression were observed between H3F3A K27M mutant and wild type GBMs, suggesting that EZH2 mediated trimethylation of H3K27 is inhibited in GBM harboring K27M mutations. Our results indicate that H3F3A K27M mutant GBMs show decreased H3K27me3 that may be of both diagnostic and biological relevance.

Introduction

DNA is organized into nucleosomes containing approximately 147 DNA base pairs wrapped around histones. Each nucleosome consists of an octomer comprised of H2A, H2B, H3 and H4 histone proteins. The N‐terminal tails of histones contain lysine (K) and arginine (R) residues that can undergo posttranslational modifications that regulate transcription and alterations in these processes are thought to play a central role in the pathogenesis of cancer. These modifications such as acetylation, methylation, phosphorylation, ubiquitination and SUMOylation may result in changes in DNA function and transcription by regulating access to cellular transcriptional machinery (reviewed in 2 26). For example, histone 3 bears lysine residues at positions K4, K9, K27 and K36 that can undergo methylation and these amino acids are highly conserved in all 16 human genes that encode histone 3 proteins 8. Methylation can be an activator or suppressor of transcription depending on the histone residue that is methylated. Methylation of H3K9 and H3K27 is associated with silencing of transcription. In contrast, methylation of H3K4 and H3K36 is associated with activation of transcription (reviewed in 2, 5).

EZH2 belongs to the polycomb‐group family of proteins that suppresses stem cell differentiation and promotes maintenance and self‐renewal of stem cells by trimethylation of H3K27 3, 17, 23. EZH2 mutations have been noted in lymphoma and myeloid malignancies (reviewed in 24). While EZH2 mutations have not been observed in other malignancies to date, EZH2 is deregulated in various cancers. For example, EZH2 overexpression is observed in advanced or metastatic breast and prostate cancer and is associated with poor prognosis 4, 15, 30. EZH2 is overexpressed in glioma stem‐like cells and adult glioblastoma patient samples 18, 19, 36. Inhibition or knockdown of EZH2 results in decreased self‐renewal of glioma stem‐like cells and glioma cell proliferation and migration 19, 29.

Genomic analyses of pediatric glioblastomas, anaplastic astrocytomas and diffuse intrinsic pontine gliomas (DIPG) have demonstrated recurrent mutations in H3F3A gene (encoding H3.3) where the lysine residue at position 27 is replaced by methionine (K27M) or the glycine residue at position 34 is replaced by arginine or valine (G34R/V) 14, 22, 27, 35. A small percentage of DIPG cases showed HIST1H3B encoding the histone H3.1 where the lysine residue at position 27 is replaced by methionine (K27M) 35. Because lysine residues in the histone 3 tail are subject to posttranslational modifications such as methylation, we hypothesized that trimethylation of H3K27 may be altered in pediatric glioblastomas (GBMs) bearing the H3F3A K27M mutation. We used immunohistochemistry (IHC) to detect overall levels of H3K27me3 in 76 pediatric gliomas and glioneuronal tumors. Our goal was to determine whether mutation in a single H3 encoding gene H3F3A could alter H3K27me3 IHC patterns in mutant pediatric GBMs in comparison to wild type GBMs or other pediatric glial and neuroglial tumors. Further, it is not known if EZH2 expression is altered in H3F3A K27M mutation GBMs. We evaluated the H3K27 methylating enzyme EZH2 in all the tumor samples. Our results indicate that H3F3A K27M mutant GBMs show significant decreases in overall H3K27me3 without significant changes in EZH2 expression.

Methods

Cohort

Cases were obtained from the Children's Hospital of Philadelphia and Children's hospital of Los Angeles after approval from both the institutional review boards. All cases were de‐identified prior to analysis. Sixty‐four cases were contained in previously well‐characterized tissue microarrays 28. Only cores containing more than 95% tumor were considered in this analyses and included a total of 64 tumors. The tumors consisted of 22 pilocytic astrocytomas (World Health Organization grade I), eight dysembryoplastic neuroepithelial tumors (DNET, grade I), 12 gangliogliomas (GG, grade I), six grade II astrocytomas (four fibrillary astrocytomas, one pilomyxoid astrocytoma and one pleomorphic xanthoastrocytoma, all grade II), eight oligodendrogliomas (five grade II oligodendrogliomas and two grade III anaplastic oligodendrogliomas, grade not available in one case) and eight GBMs (grade IV). Additionally, 12 GBM cases were evaluated as full sections. The cohort demographics are indicated in Table 1.

Table 1.

Demographics of cases. DNET = dysembryoplastic neuroepithelial tumors; GBM = glioblastoma; GG = ganglioglioma

Tumor type	Number of cases	Sex		Age (months)
		Male	Female	Mean	Median
DNET	8	5	3	136	150
GG	12	6	6	145	158
Pilocytic astrocytoma	22	12	10	110	117
Astrocytoma	6	2	4	132	142
Oligodendroglioma	8	6	2	156	199
GBM	20	12	8	150	141

IHC and automated scoring

Immunohistochemical studies were performed as previously published 16, 32. In brief, immunostaining was performed using the Discovery XT processor (Ventana Medical Systems, Tucson, AZ, USA). Tissue sections were blocked for 30 minutes in 10% normal goat serum in 2% bovine serum albumin (BSA) in phosphate buffered saline (PBS). Sections were incubated for 5 h with 0.1 μg/mL of the rabbit polyclonal anti‐H3K27me3 (07–449, Millipore, Billercia, MA, USA; 1 μg/mL) or rabbit polyclonal anti‐EZH2 (5246, Cell Signaling, Danvers, MA, USA; 1:200) antibodies. Tissue sections were then incubated for 60 minutes with biotinylated goat anti‐rabbit IgG (Vector labs, Burlingame, CA, USA; PK6101) at 1:200 dilution. Blocker D, Streptavidin‐ HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions. H3K27me3 is expressed in embryonic tissues and whole mount mouse embryos were used as positive controls 1. EZH2 is upregulated in prostatic carcinoma and tissue sections of prostatic adenocarcinoma were used as positive controls 20, 30. To ensure homogeneity of IHC, we compared staining in 20 randomly selected cases contained in tissue microarrays with their corresponding full sections selected from the same blocks that were used to generate the TMA for both H3K27me3 and EZH2. For both markers, staining results were identical in tissue microarrays and full sections.

For automated scoring, each slide was scanned using an Aperio Scanscope Scanner (Aperio Vista, CA, USA) and viewed through Aperio ImageScope software program. An individual blinded to the experimental design captured JPEG images from each core or full section (circular area of 315 cm² corresponding to the entire core or randomly chosen equivalent area from full sections) at 10 × magnification on the Aperio ImageScope viewing program. Quantification of immunostaining on each JPEG was conducted using an automated analysis program with Matlab's image processing toolbox based on previously described methodology 33. The algorithm used color segmentation with RGB color differentiation, K‐Means Clustering and background‐foreground separation with Otsu's thresholding. To arrive at a score, the number of extracted pixels were multiplied by their average intensity for each core. The final score for a given case and marker was calculated by averaging the score of two cores or two random areas chosen from full sections vis à vis that marker.

Sequencing

Genomic DNA from GBMs was extracted from formalin fixed paraffin embedded blocks from two 10μ slices using the Formapure kit (Agencourt, Beverly, MA, USA) in a 96‐well format, using a modified version of the manufacturers' method, and in a semi‐automated fashion. H3F3A coding exons were sequenced using Sanger sequencing following polymerase chain reaction (PCR) using previously described methods 12. Specific primers (see supplementary methods) were designed using Primer 3 (http://frodo.wi.mit.edu/). M13 tails were added to facilitate Sanger sequencing. PCR reactions were set up in 384 well plates, in a Beckman Coulter Biomek® FX, and run in a Duncan DT‐24 water bath thermal cycler, with 10–40 ng of genomic DNA as template, HotStart Taq (Kapa Biosystems, Woburn, MA, USA), 250 μM dNTPs, 1X PCR buffer, 6% dimethyl sulfoxide (DMSO) and 0.2 μM primers. A “touchdown” PCR method was used, which consisted of: 1 cycle of 95°C for 5 minutes; three cycles of 95°C for 30 s, 64°C for 30 s, 72°C for 60 s; three cycles of 95°C for 30 s, 62°C for 30 s, 72°C for 60 s; three cycles of 95°C for 30 s, 60°C for 30 s, 72°C for 60 s; 37 cycles of 95°C for 30 s, 58°C for 30 s, 72°C for 60 s; one cycle of 70°C for 5 min. Amplified DNA was purified using AMPure (Agencourt Biosciences, Beverly, MA, USA). The purified PCR reactions were split into two, and sequenced bidirectionally with M13 forward and reverse primer and Big Dye Terminator Kit v.3.1 (Applied Biosystems, Foster City, CA, USA), at Agencourt Biosciences. Dye terminators were removed using the CleanSEQ kit (Agencourt Biosciences), and sequence reactions were run on ABI PRISM 3730xl sequencing apparatus (Applied Biosystems). The sequences were analyzed using an automated mutation detection pipeline developed by the MSKCC Bioinformatics Core.

Statistical analysis

Statistical analyses were performed using Prism (version 5, La Jolla, CA, USA). Student's t‐test was used to assess differences in H3K27me3 or EZH2 expression between H3F3A K27M and wild type GBMs.

Results

H3K27me3 immunoreactivity in pediatric gliomas and glioneuronal tumors

H3K27me3 was examined using IHC in a cohort of gliomas and glioneuronal tumors consisting of 22 pilocytic astrocytomas, eight DNET, 12 GG, six astrocytomas, eight oligodendrogliomas and 20 GBM. The demographics of the cases are presented in Table 1. Six out of 20 GBM cases showed lowered or absent H3K27me3 staining in tumor cells but preserved staining in endothelial cells and infiltrating lymphocytes. All other tumors examined showed strong H3K27me3 nuclear staining (Figures 1 and 2).

Figure 1.

H3K27me3 in pediatric gliomas and glioneuronal tumors. Representative images (20×) of H3K27me3 in: A. Pilocytic astrocytoma, B. Grade II diffuse astrocytoma, C. Grade II oligodendroglioma, D. Grade III anaplastic oligodendroglioma, E. Grade I ganglioglioma (arrows indicating ganglion cells), blue amorphous areas represent eosinophilic granular bodies, F. Dysembryoplastic neuroepithelial tumor, grade I (arrows indicate floating neurons).

Figure 2.

H3K27me3 in decreased in H3F3A K27M mutant GBM compared to wild type GBM. Representative images of H3K27me3 in: A, C and E. Three H3F3A wild type GBM cases (A −10×, C −20×, E −40×), B, D and F. Three H3F3A K27M mutant GBM cases (B −10×, D −20×, F −40×, arrows indicate retained H3K27me3 staining in endothelial cells and infiltrating lymphocytes).

H3K27me3 is significantly lower in pediatric GBM bearing H3F3A K27M mutations compared to wild type tumors

Genomic DNA was extracted and sequenced from the GBM cases. Six out of 20 (30%) GBM cases showed H3F3A K27M mutations (Figure 3). These cases were predominantly midline in location (four out of six thalamic and one out of six spinal cord) compared to wild type GBM, which were mainly cortical (10 out of 14) in location (Supporting Information Table S1) consistent with observations from others 27. No H3F3A G34R/V mutations were detected. GBMs that bore H3F3A K27M mutations corresponded to the six cases that showed lowered/absent H3K27me3 staining. Quantitative analysis showed a significant decrease in H3K27me3 staining in H3F3A K27M mutant compared to wild type GBM (Figure 4, p = 0.01). Some cases of GG, astrocytoma and oligodendrogliomas showed sparse tumor cellularity resulting in a lower overall H3K27me3 value (Figure 4A). However, tumor cells in these cases showed preserved nuclear H3K27me3 staining. Lowered/decreased H3K27me3 nuclear staining was specific to H3F3A K27M mutant GBM tumor cells. Further, H3K27me3 values depicted in Figure 4A are an overall representation of tumor cells and endothelial cells/lymphocytes. As H3F3A K27M mutant GBMs showed retained H3K27me3 staining in endothelial cells/lymphocytes, the values depicted in Figure 4A in H3F3A K27M mutant GBMs reflect the presence of staining in these cells.

Figure 3.

Mutational analyses of GBM cases. Coding exons were sequenced using Sanger sequencing. Representative plots from: A. A H3F3A wild type GBM case showing lysine (AAG, arrow) at position 27. B. A H3F3A K27M mutant GBM case showing methionine (ATG, arrow) at position 27.

Figure 4.

Quantification of H3K27me3 and EZH2. H3K27me3 A. and EZH2 expression B. were quantified in pediatric gliomas and glioneuronal tumors. Differences between H3F3A K27M and wild type GBM with either marker were analyzed using Student's t‐test. *p = 0.01. DNET = dysembryoplastic neuroepithelial tumors, GG = ganglioglioma, Pilocytic = pilocytic astrocytoma.

EZH2 expression does not differ between H3F3A K27M mutant and wild type gliomas

We assessed EZH2 expression in all tumor samples. EZH2 expression was negative in all grade I (pilocytic astrocytomas, DNET and GG) and grade II (astrocytomas and oligodendrogliomas) tumors irrespective of tumor subtype (Figures 4 and 5). Focal EZH2 expression was noted in grade III anaplastic oligodendroglioma (Figures 5B). Strong EZH2 expression was noted in GBMs. Comparison of EZH2 expression in H3F3A K27M mutant and wild type GBMs showed no significant differences (p = 0.2, Figures 4 and 5).

Figure 5.

EZH2 expression is strongly expressed in GBMs and shows no difference between H3F3A K27M mutant or wild type GBM. Representative images (20×) of EZH2 in: A. Pilocytic astrocytoma, B. Grade III anaplastic oligodendroglioma, C. H3F3A wild type GBM, D. H3F3A K27M mutant GBM.

Discussion

Pediatric brain tumors are the most common solid malignancy in childhood and are a cause of significant morbidity and mortality. Advances in the molecular characterization of pediatric brain tumors have enhanced our understanding of these diseases and have revealed the importance of epigenetic events in the pathogenesis of these tumors 7, 21, 34. More recently, discovery of histone H3F3A mutations at the K27 and G34 positions in pediatric glioblastomas and DIPG have brought epigenetics to the forefront of pediatric brain tumor research 14, 22, 27, 35. Our goal was to determine the effects of the H3F3A K27M mutations on trimethylation of H3K27 in pediatric GBMs in comparison to wild type GBMs and other glioneuronal tumors.

H3K27me3 in H3F3A K27M mutant GBMs

Among the H3F3A mutations, K27M mutations carry an overall poor prognosis and were seen in 30 out of 50 DIPG and seven out of 36 of non‐brainstem pediatric GBM 35, nine out of 48 pediatric GBM 22 and 30 out of 42 DIPG 14 in three different reports. However, the effect of this mutation on overall trimethylation of H3K27 is not known. On first principles, these mutations would not be expected to affect overall H3K27me3 as histone H3 is encoded by over 16 human genes and H3F3A comprises only a small fraction of total cellular H3 8. Surprisingly, we found that H3K27me3 was lower or absent in tumor cells, but retained in endothelial cells and infiltrating lymphocytes (serving as internal control) in six out of 20 H3F3A K27M mutant GBMs. Wild type GBMs (14 out of 20) showed significantly higher H3K27me3 staining compared to mutant GBMs. Lowered H3K27me3 could thus serve as a diagnostic surrogate for H3F3A K27M mutations in differentiating wild type from H3F3A K27M gliomas. Future studies expanding the sample size are needed to confirm this observation.

The mechanism(s) by which lowered H3K27me3 in H3F3A K27M mutant tumors contribute to glioma pathology are not known and are avenues for future studies. Trimethylation of H3K27 plays a role in repression of lineage‐regulatory genes during pluripotency in embryonic stem cells 31. Further, H3K27me3 is also known to affect DNA methylation 9. Global loss of trimethylation of H3K27 may contribute to pathogenesis in H3F3A K27M mutant pediatric GBM by affecting differentiation pathways. Indeed, gene expression patterns and DNA methylation analyses from H3F3A mutant GBMs show deregulation of genes related to brain development and promoter methylation of genes such as OLIG1/2 (in G34R/V GBMs) and FOXG1 (in K27M GBMs) which supports this hypothesis 22, 27. The finding that overall H3K27me3 is specifically decreased in GBMs bearing H3F3A K27M mutations may thus be of significance from both biological and diagnostic perspectives.

EZH2 in H3F3A K27M mutant GBMs

EZH2 is the histone methyltransferase responsible for H3K27 methylation. The above results suggest that either EZH2 is not expressed or it has been rendered nonfunctional in the pediatric H3F3A K27M mutant GBMs we examined. To distinguish between these possibilities, the expression of EZH2 was examined with IHC. There was no significant difference in EZH2 expression between H3F3A K27M mutant and wild type GBMs. This suggests the hypothesis that H3F3A K27M mutations may function in a dominant negative fashion to suppress EZH2 trimethylation of H3K27 in a yet unknown manner in pediatric GBMs bearing H3F3A K27M mutations.

EZH2 and H3K27me3 in other gliomas and glioneuronal tumors

We found that EZH2 was not expressed or was minimally expressed in all grade I and grade II tumors regardless of tumor subtype. EZH2 expression showed focal positivity in grade III (anaplastic oligodendroglioma) tumors. In contrast, EZH2 was expressed at higher levels in all GBMs suggesting that increased EZH2 expression was restricted to higher‐grade tumors similar to observations from others 6. All glioneuronal tumors and other glial tumors examined showed H3K27me3 staining. Interestingly, overall levels of H3K27me3 were quite similar between GBMs and pilocytic astrocytomas. Pilocytic astrocytomas show BRAF mutations 10, 25, which is known to alter the cellular methylome in melanomas and colon cancer 11, 13. One possibility for further study raised by these results is that alterations in BRAF may contribute to H3K27me3.

Summary

In summary, we show that H3F3A K27M mutations in GBMs correlate with decreased overall trimethylation of H3K27 without affecting EZH2 levels. Decreased H3K27me3 was specific to H3F3A K27M mutant in GBMs as all other tumor subtypes including H3F3A wild type GBMs showed strong H3K27me3 staining. These findings have both diagnostic and biological significance and are avenues for future research.

Supporting information

Supplementary methods.

Table S1. Demographics, location and H3F3A genotype of GBMs.