Brain tumors are one of the most common childhood malignancies. Diffuse high-grade gliomas represent approximately 10% of pediatric brain tumors. Exon sequencing has identified a mutation in K27M of the histone H3.3 gene (H3F3A K27M and G34R/V) in about 20% of pediatric glioblastomas, but it remains to be seen whether these mutations can be considered specific for pediatric diffuse high-grade astrocytomas or also occur in other pediatric brain tumors. We performed a pyrosequencing-based analysis for the identification of H3F3A codon 27 and codon 34 mutations in 338 pediatric brain tumors. The K27M mutation occurred in 35 of 129 glioblastomas (27.1%) and in 5 of 28 (17.9%) anaplastic astrocytomas. None of the other tumor entities showed H3F3A K27M mutation. Because H3F3A K27M mutations occur exclusively in pediatric diffuse high-grade astrocytomas, analysis of codon 27 mutational status could be useful in the differential diagnosis of these neoplasms.
Tumors of the central nervous system (CNS) are one of the most common childhood malignancies and a leading cause of cancer-related morbidity and mortality.1 Although diffuse high-grade gliomas (HGGs) are the most common CNS tumors in adults, they represent only approximately 10% of all pediatric CNS tumors. The overall prognosis of glioblastomas (GBMs) in children seems slightly better than in adults, with a higher percentage of long-term survivors; nevertheless, prognosis remains poor in general despite aggressive treatment with surgery, chemotherapy, and radiotherapy.2 In contrast to adult diffuse high-grade astrocytomas, in which key genetic alterations (such as IDH1/2 mutation and EGFR amplification) have been identified, molecular alterations in pediatric cases have not been fully elucidated. Although alterations of the p53/MDM2/p14 pathway are detectable at a similar frequency compared with adults, pediatric GBMs show frequent gains of chromosome 1q and only rarely gains of chromosome 7 and 10q losses.3 The most frequent focal amplifications also differ, with PDGFRA and EGFR predominating in childhood and adult populations, respectively.3
Histones are basic nuclear proteins responsible for the nucleosome structure within eukaryotic cells. Among the 5 classes of histones, some are expressed only during the S phase, whereas others are replication independent and referred to as replacement histones, which are expressed in quiescent or terminally differentiated cells. H3.3 is a replacement histone subclass that is encoded by 2 distinct genes, H3.3A (H3F3A) and H3.3B.4-6 Presently, pediatric cases of GBM are being investigated by genomic sequencing, leading to interesting new findings. Exon sequencing has identified a K27M driver mutation in H3F3A in 20% of pediatric cases of GBM but not in adult glioblastomas.7-9 Diffuse intrinsic pontine gliomas (World Health Organization [WHO] grades II-IV) have also been shown to carry H3F3A K27M mutations.10 Finally, mutations at codon 34 of H3F3A (either G34R or G34V) have been reported in pediatric GBMs.7 Although the glycine at codon 34 (G34) is not subjected to posttranslational modifications, it is in close proximity to a lysine at position 36 (K36), whose methylation status is involved in transcriptional elongation.6,7
The H3F3A K27M mutation is more frequent than G34V/R mutations in pediatric diffuse high-grade astrocytomas7-9 and could therefore be used as a potential diagnostic marker for the identification of these tumors analogous to the use of IDH1/2 mutations in the diagnosis of the adult diffuse gliomas. However, it is still unclear whether these mutations, particularly the K27M mutation, can be considered as specific for pediatric diffuse high-grade astrocytomas or also occur in other pediatric brain tumors. To answer this question, we performed a pyrosequencing-based analysis for H3F3A K27M mutation in 338 pediatric brain tumors.
Materials and Methods
Fresh-frozen or formalin-fixed paraffin-embedded (FFPE) tissue specimens from 338 pediatric brain tumors were analyzed. The series includes 129 GBMs (WHO grade IV), 28 anaplastic astrocytomas (AAs; WHO grade III), 6 anaplastic oligoastrocytomas (WHO grade III), 6 anaplastic (pilocytic) astrocytomas (APAs; WHO grade III), 15 pilocytic astrocytomas (PAs; WHO grade I), 5 pilomyxoid astrocytomas (WHO grade II), 12 desmoplastic infantile gangliogliomas/astrocytomas (DIGs/DIAs; WHO grade I), 2 oligodendrogliomas (WHO grade II), 4 oligoastrocytomas (WHO grade II), 6 diffuse astrocytomas (WHO grade II), 9 anaplastic ependymomas (WHO grade III), 8 ependymomas (WHO grade II), 19 medulloblastomas (MBs; WHO grade IV), 14 atypical teratoid/rhabdoid tumors (ATRTs; WHO grade IV), 26 primitive neuroectodermal tumors (PNETs) of the CNS (WHO grade IV), 12 ependymoblastomas (EPBLs; WHO grade IV), 10 gangliogliomas (GGs; WHO grade I), and 10 choroid plexus carcinomas (CPCs; WHO grade III). All tumors were diagnosed according to the WHO classification of tumors of the CNS11 using standard histologic and immunohistochemical methods. No stereotactic biopsy specimen material was included in this study.
Fragments chosen for DNA extraction were checked histologically (either frozen or permanent section) to ensure that they consisted of at least 80% tumor cells. Extraction of DNA from frozen tissue was carried out using proteinase K digestion followed by phenol/chloroform extraction described elsewhere.12 DNA from FFPE tumor tissues was extracted using the QIAamp DNA Mini Tissue Kit (Qiagen GmbH, Düsseldorf, Germany) according to the manufacturer's instructions. We screened the hotspot codons 27 and 34 of the H3F3A gene for mutations using a pyrosequencing assay. For this purpose, single-stranded DNA templates were immobilized on streptavidin-coated Sepharose highperformance beads (GE Healthcare, Uppsala, Sweden) using the PSQ Vacuum Prep Tool and Vacuum Prep Worktable (Biotage, Uppsala, Sweden) according to the manufacturer's instructions and then incubated at 80°C for 2 minutes and allowed to anneal to a 0.4-mmol/L sequencing primer at room temperature. Pyrosequencing was performed using PyroGold Reagents (Biotage) on the Pyromark Q24 instrument (Biotage) according to the manufacturer's instructions. Controls in which the sequencing primer or template were omitted were used to detect background signal. Pyrogram outputs were analyzed by the PyroMark Q24 software (Biotage) using the allele quantification software to determine the percentage of mutant vs wild-type alleles according to percentage relative peak height. In this study, a 240–base pair fragment of exon 2 of H3.3 containing the coding region was amplified in a polymerase chain reaction (PCR) using 25 ng genomic DNA as the template and the following primer set for H3.3: forward, 5'-TGTTTGTAGTTGCATATGGG-3', and reverse, 5'-biotin-TACAAGAGAGACTTTTGTCC-3'. PCR amplicons of H3.3 containing both mutation hotspots (codons 27 and 34) were analyzed on 2% agarose gels and subjected to pyrosequencing reactions using the pyrosequencing primer H3.3 (-Py-5'-CAAAAGCCGCTCGCA-3') with the following nucleotide dispensation order: GAT GAGTGCGCTCTACTCGAGCGTGTGA. As a negative control, DNA derived from normal brain tissue was used.
Our series of 163 diffuse pediatric HGGs included 77 male and 86 female patients. The mean age at surgery was 9.48 years (age range, newborn to 18 years). Among these 163 HGGs, we found 40 cases harboring an H3F3A K27M mutation (24.5%). GBM showed a K27M mutation in 35 cases (27.1%; 15 males, 20 females) (specificity, 87.5%; sensitivity, 27.1%), whereas 5 of 28 (17.9%; 3 males, 2 females) AAs presented the same mutation. With regard to the classification of a tumor to the group of diffuse malignant astrocytic tumors (ie, GBM and AA), the sensitivity was 25.5%, but the specificity was 100%. Among the 6 anaplastic oligoastrocytomas, no H3F3A mutation could be detected. The G34R mutation (not shown) was observed in only 7 of 129 of the pediatric GBMs (5.4%), as well as in 1 case of anaplastic oligoastrocytoma, but not in AA. Our study included 3 cell lines from patients with pediatric GBM (SF 188, CHLA-200, and UKB-pGBM1). They contained exclusively wild-type alleles. All other tumor entities investigated in this study showed no evidence of an H3F3A K27M mutation. The results are summarized in Table 1. Exemplary pyrograms are shown Figure 1.
Central Nervous System Tumors and Cell Lines Investigated for H3F3A K27M Mutation Status
Histones are eukaryotic nuclear proteins that play an important role in the regulation of DNA replication, transcription, and storage by changing the nucleosome structure depending on their posttranslational modifications. Among the 5 classes of histones, H3.3 is a replacement histone subclass that is encoded by 2 different genes, H3.3A (H3F3A) and H3.3B.5H3.3 is incorporated into chromatin independently of the cell replication cycle.5 Both H3F3A mutations found in GBMs are located in a region undergoing posttranslational modifications.7,9 In particular, the amino acid exchange from lysine to methionine at position 27 (K27M) prevents posttranslational methylation, associated with polycomb-mediated gene repression6 and K27 acetylation, present at active promoters and enhancers.6 Because of its significant frequency in pediatric diffuse high-grade astrocytomas, the H3F3A K27M mutation could be potentially used as a molecular diagnostic marker for these tumors. Molecular markers are emerging as an important tool in routine diagnostic neuropathology, notably in pediatric tumors such as ATRT (INI-1 mutation) and in PA (KIAA1549-BRAF fusions and BRAF V600E mutation). For high-grade gliomas, IDH-1 mutations represent the best-known diagnostic markers,13,14 but because of the low IDH-1 mutation rate in pediatric glial tumors,15 they unfortunately have only a limited utility in this age group. In contrast to IDH-1 mutations, mutations in H3F3A have been frequently observed in pediatric astrocytomas (pediatric GBMs and pediatric diffuse intrinsic pontine gliomas).7-10 In GBM, we found an overall mutation rate of 27.1%, which is in line with the frequency described in other studies.7,9 With regard to the classification of a tumor to the group of diffuse malignant astrocytic tumors (ie, GBM and AA), the sensitivity was 25.5%, but the specificity was 100%. Because this group of tumors is treated according to the same protocols in pediatric oncology, the secure assignment to this group is important. Although H3F3A K27M mutations occur only in one fourth of these tumors, we did not find this mutation in any other pediatric brain tumor, including low-grade gliomas and nondiffuse high-grade tumors. Because the H3F3A K27M mutation is significantly more frequent than the G34R/V mutation (ratio of 6:1 in our series) in pediatric diffuse high-grade astrocytomas, the former is suitable as a molecular marker for this tumor group. In addition, G34R/V mutations seem to be less specific in pediatric tumors because they also occur in CNS-PNET.16
Although previous studies have shown the absence of H3F3A mutations in adult HGGs and in extracerebral pediatric solid tumors (such as rhabdomyosarcoma, osteosarcoma, and neuroblastoma),9 only a limited number of the more common pediatric CNS tumors, such as PAs, MBs, and ependymomas, have been analyzed for the presence of H3F3A mutations.7,9 We did not find any mutations in low-grade gliomas (DIG, DIA, PA, GG), embryonal tumors (MB, ATRT, CNS-PNET, EPBL), (anaplastic) ependymomas, or CPCs. Our analysis thus confirms that H3F3A K27M mutations are restricted to pediatric diffuse high-grade astrocytomas.
From a diagnostic point of view, the absence of H3F3A K27M mutations in PAs, APAs, and CNS-PNETs is noteworthy.
Histologically, APAs are defined by the presence of hypercellularity, moderate to severe cytologic atypia, and brisk mitotic activity with or without necrosis. Nevertheless, they have a better prognosis than diffuse astrocytomas and GBM. At the molecular level, APAs do not show the typical molecular features of GBM (ie, PDGFRA or EGFR) amplifications but frequently present PI3K/AKT pathway alterations.17 Based on practical experience, the distinction between GBM and APA, as well as in some cases between high-grade astrocytoma and PA, may be difficult in small tumor biopsy specimens. Because these entities also share similar neuroradiological features, reliable molecular markers useful in this differential diagnosis are particularly welcome. The combined analysis of the presence of KIAA1549-BRAF fusion and H3F3A K27M mutation in pediatric cases of ambiguous histology should be helpful in distinguishing PA and APA from GBM. However, KIAA1549-BRAF fusions also have been found in single adult GBM,18 but they have been not shown in pediatric cases of GBM.
By routine histopathologic diagnostics, CNS-PNETs are also difficult to distinguish from diffuse high-grade gliomas and often represent an “exclusion diagnosis” based on the identification of a noncerebellar embryonal neoplasm with MB-like histology, frequently associated with expression of neuronal markers.19 However, a definitive distinction between a CNS-PNET and a GBM remains sometimes impossible: in such cases, the final diagnosis depends on the expertise of the pathologist or relies on additional institution-defined criteria, which may vary from laboratory to laboratory.20 Unfortunately, definitive molecular markers for CNS-PNET, in contrast to other embryonal tumors (such as INI-1 alterations in ATRTs or chromosome 19q13.14 amplification in EPBLs), are not available. In this view, the presence of an H3F3A K27M mutation in a histologically undifferentiated tumor reliably excludes the diagnosis of a CNS-PNET.
In conclusion, H3F3A K27M mutation occurs exclusively in diffuse high-grade pediatric astrocytomas, and therefore analysis of H3F3A codon 27 could provide significant help in the differential diagnosis of these neoplasms.
We thank S. Albrecht, MD, Department of Pathology, McGill University Health Centre, Montreal, Canada, for careful reading and editing of the manuscript.