|Home | About | Journals | Submit | Contact Us | Français|
Trimethylation of histone 3 lysine 9 (H3K9me3) is a marker of repressed transcription. Cells transfected with mutant isocitrate dehydrogenase (IDH) show increased methylation of histone lysine residues, including H3K9me3, due to inhibition of histone demethylases by 2-hydroxyglutarate. Here, we evaluated H3K9me3 and its association with IDH mutations in 284 gliomas. H3K9me3 was significantly associated with IDH mutations in oligodendrogliomas. Moreover, 72% of World Health Organization grade II and 65% of grade III oligodendrogliomas showed combined H3K9me3 positivity and 1p19q co-deletion. In astrocytic tumors, H3K9me3 positivity was found in all grades of tumors; it showed a significant relationship with IDH mutational status in grade II astrocytomas but not in grade III astrocytomas or glioblastomas. Finally, H3K9me3-positive grade II oligodendrogliomas but not other tumor subtypes showed improved overall survival compared to H3K9me3-negative cases. These results suggest that repressive trimethylation of H3K9 in gliomas may occur in a context-dependent manner and is associated with IDH mutations in oligodendrogliomas, but may be differently regulated in high-grade astrocytic tumors. Further, H3K9me3 may define a subset of grade II oligodendrogliomas with better overall survival. Our results suggest variable roles for IDH mutations in the pathogenesis of oligodendrogliomas vs. astrocytic tumors.
Epigenetic mechanisms such as DNA methylation and histone methylation play a significant role in the pathogenesis and progression of tumors. DNA methylation has been extensively studied in glioblastomas (1–3). In contrast, histone modifications in glioma pathogenesis have remained poorly understood. Recent studies have emerged implicating histone modifications in adult gliomas (4–6) and histone gene mutations in pediatric glioblastomas (7, 8). Histones are proteins that form octamers consisting of H2A, H2B, H3 and H4 around which DNA is wrapped. Histones have amino acid tails that are subject to various modifications such as acetylation, methylation, phosphorylation, ubiquitination and SUMOylation of arginine (R) and lysine (K) residues, resulting in changes in DNA function and transcription (9). Histone lysine methylation can be an activator or suppressor of transcription depending on the histone residue that is methylated. For example, methylation of H3K4, H3K36 and H3K79 is associated with activation of transcription, whereas methylation of H3K9, H3K27 and H4K20 is associated with silencing of transcription (9, 10).
Histone lysine methylation is regulated by histone lysine methyltransferases (KMTs) and demethylases (KMDs). One of the earliest KMTs to be characterized was SUV39H. SUV39H belongs to the SET domain family and specifically methylates H3K9(11). Subsequently, H3K9 may be the most extensively studied site of histone methylation (12). Trimethylation of H3K9 (H3K9me3) is associated with pericentric heterochromatin formation, transcriptional repression and DNA methylation (13–15).
Very little is known about H3K9 modifications in glioma pathogenesis. H3K9me3 is increased in cells expressing mutant isocitrate dehydrogenase 1 (IDH1 R132H) (4, 5, 16). The NADP-positive dependent enzyme IDH is mutated in ~70% of World Health Organization (WHO) grade II and grade III gliomas and secondary glioblastomas (17–19). Mutations in IDH result in neomorphic activity of the enzyme catalyzing the production of the oncometabolite 2-hydroxyglutarate (2-HG) from α-ketoglutarate (α-KG) (20, 21). 2-HG is a structural analogue of α-KG and can inhibit several α-KG-dependent oxygenases (22, 23). The α-KG-dependent oxygenases form a large family of enzymes that influence various cellular functions, such as collagen modifications, carnitine synthesis, hypoxic sensing and histone and DNA demethylation (24). For example, KDM4C (belonging to the Jumonji C family of KDMs) uses α-KG, oxygen and Fe(II) as cofactors to demethylate histone lysine residues. Immortalized human astrocytes or murine neurosphere cultures overexpressing IDH1 R132H show increases in H3K9me3 compared to cells overexpressing wild type IDH1 (4, 16)l, which is mediated in part by the inhibition of KDM4C by 2-HG (4). Similarly, heterozygous knock-in of IDH1R132H in HCT116 colon cancer cell lines results in increased H3K9me3 (5).
Based on these observations we hypothesized that IDH mutant gliomas would show increased H3K9me3. The goals of this study were to evaluate the relationship between H3K9me3 and IDH mutations in gliomas of different subtypes and grades, and to determine if H3K9me3 had any prognostic significance in gliomas. The results suggest that the effect of IDH mutations on histone methylation may differ between oligodendrogliomas and high-grade astrocytomas and may provide insight into the role played by mutant IDH in the pathogenesis of these tumors.
Cases were obtained from the University of California San Francisco, the University of Pennsylvania, the University of Pittsburgh, Children’s Hospital of Philadelphia and Memorial Sloan-Kettering Cancer Center (MSKCC) following approval from the respective institutional review boards. All cases were de-identified prior to analysis and included both whole tissue and tissue microarray sections (Supplemental Digital Content 1, Methods, http://links.lww.com/NEN/A432).
The cohort demographics consisted of a total of 284 gliomas (Table 1). The tumors consisted of 263 adult gliomas (53 WHO grade II oligodendrogliomas, 47 grade III anaplastic oligodendrogliomas, 6 grade I pilocytic astrocytomas (supratentorial), 28 grade II diffuse astrocytomas, 43 grade III anaplastic astrocytomas and 86 grade IV glioblastomas) and 21 grade I pediatric pilocytic astrocytomas (14 cerebellar and 8 supratentorial). Each case was reviewed individually by a neuropathologist at each institution and then confirmed by a neuropathologist at MSKCC. Loss of heterozygosity (LOH) data for 1p and 19q loci was available in 51/53 of grade II and 40/47 grade III oligodendrogliomas. In oligodendrogliomas, 84% (43/51) of grade II and 75% (30/40) of grade III contained co-deletions of chromosomal arms 1p and 19q.
Immunohistochemical studies were performed as previously published (4, 25, 26). In brief, immunohistochemical detection was performed using the Discovery XT processor (Ventana Medical Systems, Tucson, AZ). Tissue sections were blocked for 30 minutes in 10% normal goat serum in 2% BSA in PBS. Sections were incubated for 5 hours with 0.1 μg/ml of the rabbit polyclonal anti-H3K9me3 (ab8898, Abcam, Cambridge, MA) or mouse monoclonal anti-IDH1 R132H (DIA H09, Dianova, Hamburg, Germany; 1:30) antibodies. Tissue sections were then incubated for 60 minutes with biotinylated goat anti-rabbit IgG (PK6101, Vector Laboratories, Burlingame, CA) or goat anti-mouse IgG (Vector, BA9200) at 1:200 dilution. Blocker D, Streptavidin- HRP and DAB detection kit (Ventana Medical Systems) were used according to the manufacturer instructions.
Whole mount mouse embryos were used as positive controls for H2K9me3 (27). H3K9me3 was also assessed in normal cortical adult post mortem brain tissue obtained from 4 cases. H3K9me3 in all of those cases was negative in glia and in endothelial cells but showed scattered nuclear staining in neurons (Figure, part A, Supplemental Digital Content 2, http://links.lww.com/NEN/A433). Two individuals scored H3K9me3 and IDH1 R132H as positive or negative in a blinded manner. Only nuclear staining for H3K9me3 in tumor cells was considered positive. Preservation of antigenicity in tumor sections that were negative for H3K9me3 was confirmed by retained neuronal nuclear staining in surrounding brain tissues as an internal control when available for evaluation (Figure, part C, Supplemental Digital Content 2, http://links.lww.com/NEN/A433). Sections stained for IDH1 R132H were scored positive or negative based on described criteria (25, 28, 29).
Genomic DNA was extracted from formalin-fixed paraffin-embedded blocks from 2 10-μm-thick slices using the Formapure kit (Agencourt, Beverly, MA) in a 96-well format, using a modified version of the manufacturers’ method, and in a semi-automated fashion. IDH1/2 mutational analyses were performed on cases that were IDH1 R132H-negative using previously described protocols (26, 30). Due to limitations of tissue availability, cases from the tissue microarray containing glioblastomas and pilocytic astrocytomas were not available for mutational analyses.
Statistical analyses were performed in consultation with the MSKCC biostatistics core. SPSS (version 21, IBM, Chicago, IL) and Prism (version 5, La Jolla, CA) were used to analyze data. Fisher exact test was used to determine the odds ratios to evaluate the strength of association between IDH mutational status and H3K9me3 positivity or the association between 1p-19q deletion and H3K9me3 positivity in oligodendrogliomas. The log-rank (Mantel-Cox) test was used to examine the association of H3K9me3 positivity on overall survival within each tumor subtype. Multivariate survival analysis was performed using the Cox proportional hazards model. Differences were considered significant when p < 0.05 (95% confidence intervals).
Positive H3K9me3 nuclear staining in gliomas was seen in 83% (44/53) of grade II oligodendrogliomas and 79% (37/47) of grade III oligodendrogliomas (Table 1; Fig. 1). There was a significant relationship between H3K9me3 positivity and IDH mutations in grade II oligodendrogliomas (76% [40/53] of cases with IDH mutations and positive for H3K9me3, odds ratio 8.9, p = 0.01), and grade III oligodendrogliomas (75% [35/47] of cases with IDH mutations and positive for H3K9me3, odds ratio 11.7, p = 0.01) (Figs. 1, ,22).
H3K9me3 positivity was as follows in astrocytic gliomas: 48% (13/27) of pilocytic astrocytomas (Figure, part B, Supplemental Digital Content 2, http://links.lww.com/NEN/A433), 78% (22/28) of diffuse astrocytomas, 67% (29/43) of anaplastic astrocytomas, and 79% of glioblastomas (68/86) (Table 1; Figs. 2, ,3).3). As expected, all pilocytic astrocytomas were IDH1 R132H-negative. H3K9me3 showed a significant association with IDH mutations in diffuse astrocytomas (65% (18/28) of cases with IDH mutations and positive for H3K9me3, odds ratio 9, p = 0.04 (Figs. 2, ,3).3). No significant relationship between IDH mutations and H3K9me3 positivity was observed in anaplastic astrocytomas or glioblastomas (Figs. 2, ,33).
Loss of heterozygosity information for 1p-19q was available in 46/53 of grade II and 40/47 grade III oligodendrogliomas (Table, Supplemental Digital Content 3, http://links.lww.com/NEN/A434). 1p-19q co-deletions showed a strong correlation with IDH mutational status (grade II oligodendrogliomas: odds ratio = 41, p = 0.0001 and grade III oligodendrogliomas: odds ratio = 28.5, p = 0.01). Two grade II oligodendrogliomas had co-deletions but did not bear IDH mutations. All grade III anaplastic oligodendrogliomas with co-deletions bore IDH mutations. H3K9me3 positivity and 1p-19q co-deletions were noted in 72% (37/51) of grade II and 65% (26/40) grade III oligodendrogliomas. H3K9me3 was positive in 12% (6/51) of grade II and 17% (7/40) of grade III oligodendrogliomas with intact 1p-19q. H3K9me3 was negative in 14% (7/51) of grade II and 10% (4/40) of grade III oligodendrogliomas with 1p-19q co-deletions (Fig. 4; Table, Supplemental Digital Content 3, http://links.lww.com/NEN/A434). However, Fisher exact test did not reveal a statistically significant relationship between H3K9me3 and loss of 1p-19q in either grade II (odds ratio = 0.8, p = 1) or grade III (Odds ratio=2.8, p=0.3) oligodendrogliomas (Table, Supplemental Digital Content 4, http://links.lww.com/NEN/A435).
Overall survival was improved in H3K9me3 positive grade II oligodendrogliomas vs. their H3K9me3-negative counterparts (Log-rank (Mantel-Cox) test, hazards ratio = 5.9, p = 0.005) (Fig. 5a; Table 2). Multivariate analysis using the Cox proportional hazards model showed that H3K9me3 positivity was a significant prognostic factor (n = 53, p = 0.015), but age, sex, or IDH mutational status were not (Table, Supplemental Digital Content 4, http://links.lww.com/NEN/A435). 1p-19q LOH showed a trend towards improved overall survival (p = 0.052) (Table, Supplemental Digital Content 4, http://links.lww.com/NEN/A435). No significant difference in survival was noted in grade III oligodendrogliomas (Log-rank (Mantel-Cox) test, Hazards ratio = 1.6, p = 0.4, Fig. 5b; Table 2). No significant associations were observed in overall survival between H3K9me3-positive and -negative gliomas, regardless of grade in astrocytic tumors (Table 2).
IDH mutations are seen in ~70% of intermediate grade gliomas. How mutations in IDH contribute to brain tumor development is not known. Mutant IDH catalyzes the production of 2HG from α-KG (20, 21). 2HG is structurally similar to α-KG and inhibits α-KG dependent dioxygenases, such as the Jumonji C family of histone demethylases, resulting in increased histone methylation marks including H3K9me3 in vitro (4, 22, 23). We sought to determine the relationship between H3K9me3 and IDH mutations in gliomas. We found significant associations between the 2 factors in oligodendrogliomas and grade II astrocytomas. While 67% of anaplastic astrocytomas and 78% of glioblastomas showed H3K9me3 positivity, there was no significant association with IDH mutations. Finally, H3K9me3 showed a beneficial survival effect in grade II oligodendrogliomas but not in grade III oligodendrogliomas or in any grade of astrocytic tumors.
Because IDH mutations are closely associated with 1p-19q co-deletions (18, 31–33), we considered whether H3K9me3 was associated with 1p-19q deletions in oligodendrogliomas. Indeed, 72% of grade II and 65% of grade III oligodendrogliomas that contained 1p-19q deletions were H3K9me3-positive. This association was not, however, statistically significant, suggesting that 1p-19q deletions may not directly regulate histone methylation. Recurrent mutations in the CIC gene (homolog of the Drosophila gene capicua) on chromosome 19q and FUBP1 gene (FUSE binding protein 1) on chromosome 1p have been recently described in oligodendrogliomas (34, 35). These proteins are thought to influence the RAS/MAP-kinase pathway (CIC) and MYC activation (FUBP1) and thus may not directly methylate histones (34).
H3K9me3-positive grade II oligodendrogliomas, but not other glioma subtypes showed significantly improved overall survival. Future studies with a larger cohort are needed to confirm this observation but the prognostic significance of H3K9me3 needs to be interpreted with caution. While H3K9me3 showed a significant relation with IDH mutations in oligodendrogliomas and grade II astrocytomas, we did not observe a one-to-one correspondence. For example, 7% of diffuse astrocytomas, 11% of grade II oligodendrogliomas and 13% of anaplastic oligodendrogliomas were IDH mutant but H3K9me3-negative. Conversely, 14% of diffuse astrocytomas, 5% of grade II oligodendrogliomas and 4% of anaplastic oligodendrogliomas were H3K9me3-positive but IDH wild type. We thus did not expect the prognostic value of IDH mutations to entirely translate to H3K9me3. One of the limitations of this study is that we are restricted to assessing H3K9me3 by immunohistochemistry due to tissue availability. Nevertheless, these observations reflect the complexity of the pathogenesis of IDH mutant gliomas. Multiple mechanisms downstream of mutant IDH and 2-HG may contribute to the pathogenesis with H3K9me3 representing one piece of the puzzle (36). Nevertheless, the finding that trimethylation of H3K9 is differentially regulated in oligodendrogliomas and high-grade astrocytomas is a significant observation. It suggests that mutant IDH may have different roles in oligodendrogliomas vs. high-grade astrocytomas; thus, it may provide clues for understanding the pathogenesis of these tumors.
We found significant H3K9me3 positivity in anaplastic astrocytomas and glioblastomas without any significant associations with IDH mutations. Similarly, H3K9me3 exhibited no prognostic associations in these tumors. Methylation of H3K9 is regulated by a complex dynamic interplay involving the enzymatic activity of KMTs and KDMs. It is possible that the activity of these enzymes is deregulated in higher-grade astrocytomas and glioblastomas by means of a yet undefined process that is independent of the effect of mutant IDH. Further, DNA methylation can promote histone methylation and vice versa. IDH mutations result in DNA methylation and reprograming the methylome to the glioma-CpG island methylation phenotype (G-CIMP), which could in turn influence H3K9 methylation (16). DNA methyltransferases can also influence histone methylation (37–39). Because Dnmt1 and Dnmt3 are deregulated in glioblastomas (40, 41), it is possible that aberrant activity of these enzymes mediate trimethylation of H3K9 in an IDH mutant/ 2-HG independent manner. The data suggest the hypothesis that H3K9 can be trimethylated in high-grade astrocytomas and glioblastomas in both mutant IDH1-2HG dependent and independent manner. Furthermore, other epigenetic marks such as H3K27me3 may be altered in these tumors. Future experiments examining this question will address this issue.
A high frequency of positivity of H3K9me3 was observed in all subtypes or grades of gliomas including pilocytic astrocytomas. Pilocytic astrocytomas show alterations in BRAF (42), which is known to alter the methylome in melanomas and colon cancer (43, 44). Further, melanomas bearing the BRAF(V600E) mutation show amplification of SETDB1, a H3K9 KMT resulting in increased H3K9me3 (45). It is possible that BRAF alterations contribute to H3K9me3 in pilocytic astrocytomas. To date, few studies have examined H3K9 modifications in glioma pathogenesis or glial development. H3K9me3 is thought to play a role in differentiation (46–48). Neurosphere cultures overexpressing mutant IDH1 R132H show increased H3K9me3 accompanied by a decrease in glial fibrillary acidic protein expression compared to cells overexpressing wild type IDH1, suggesting that H3K9me3 may contribute to repressing glial differentiation in IDH mutant gliomas (4).
Histone methylation remains poorly understood in gliomas and may be vital to further our understanding of the pathology of gliomas. We hypothesize that repressive trimethylation of H3K9 can occur in gliomas in IDH mutant dependent or independent fashion. Repressive H3K9 trimethylation may be regulated differently in oligodendroglial and astrocytic tumors and may be associated with IDH mutations in oligodendrogliomas, but may be regulated by other, yet undefined mechanisms in astrocytic tumors. This difference in H3K9 trimethylation could represent one of the ways by which IDH mutations play different roles in oligodendrogliomas and astrocytomas. Moreover, IDH mutations show differential associations with other genetic alterations in oligodendrogliomas vs. astrocytomas. IDH mutations in oligodendrogliomas are associated with 1p19q codeletions, while in astrocytic tumors IDH mutations are closely associated with ATRX and TP53 mutations (8, 49–52). ATRX is a SWI/SNF chromatin remodeler and plays a role in deposition of histone variant H3.3 (53). How IDH mutations, histone methylation and ATRX mutations contribute to the pathogenesis of gliomas is not known. The interplay between these complex genetic alterations and histone methylation in oligodendrogliomas and astrocytomas are avenues for future experimentation.
We thank the MSKCC cytology core facility, the Geoffrey Bene Translational core and Jianan Zhang for expert technical assistance and Elyn Riedel from the MSKCC biostatistics core for help with data analysis. We thank Dennis Pozega, Patrick Ward and Tullia Lindsten for critical reading of the manuscript.
This work was supported in part by grants from the NCI and NIH to C.B.T. C.B.T is co-founder of Agios Pharmaceuticals and has financial interest in Agios.
The authors of this study declare no other potential conflicts of interest.