High Frequency of H3 K27M Mutations in Adult Midline Gliomas

Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 https://doi.org/10.1007/s00432-018-02836-5

ORIGINAL ARTICLE – CANCER RESEARCH

High frequency of H3 K27M mutations in adult midline gliomas

Azadeh Ebrahimi1,2,8,10 · Marco Skardelly3,4,5,8 · Martin U. Schuhmann3 · Martin Ebinger9 · David Reuss2,10 · Manuela Neumann1,8 · Ghazaleh Tabatabai4,5,6,7,8 · Patricia Kohlhof‑Meinecke11 · Jens Schittenhelm1,8

Received: 11 October 2018 / Accepted: 27 December 2018 / Published online: 4 January 2019 © Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract Purpose Diffuse midline gliomas, H3 K27M-mutant were introduced as a new grade IV entity in WHO classification of tumors 2016. These tumors occur often in pediatric patients and show an adverse prognosis with a median survival less than a year. Most of the studies on these tumors, previously known as pediatric diffuse intrinsic pontine glioma, are on pediatric patients and its significance in adult patients is likely underestimated. Methods We studied 165 cases of brain tumors of midline localization initially diagnosed as diffuse astrocytomas, oligo- dendrogliomas, pilocytic astrocytomas, supependymomas, ependymomas and medulloblastomas in patients with an age range of 2–85. Results We identified 41 diffuse midline gliomas according WHO 2016, including 12 pediatric and 29 adult cases, among them two cases with histological features of low grade tumors: pilocytic astrocytoma and subependymoma. 49% (20/41) of the patients were above 30 years old by the first tumor manifestation including 29% (11/41) above 54 that signifies a broader age spectrum as previously reported. Our study confirms that H3 K27M mutations are associated with a poorer prognosis in pediatric patients compared to wild-type tumors, while in adult patients these mutations do not influence the survival significantly. The pattern of tumor growth was different in pediatric compared to adult patients; a diffuse growth along the brain axis was more evident in adult compared to pediatric patients (24% vs. 15%). Conclusion H3 K27M mutations are frequent in adult midline gliomas and have a prognostic role similar to H3 K27M wild- type high-grade tumors.

Keywords Diffuse midline glioma · H3F3A · H3 K27M mutations

Introduction of the H3F3A (encoding histone H3.3), HIST2H3C (encoding histone H3.2) and HIST1H3B/C (encoding histone H3.1) Diffuse midline gliomas H3 K27M mutant are aggressive genes (Castel et al. 2015; Hoffman et al. 2016). They also brain tumors of short median survival, usually less than a show a distinct pattern of global DNA methylation placing year, despite current multimodal therapies (Buczkowicz them into a separate entity (Sturm et al. 2012). Replacement et al. 2014a; Robison and Kieran 2014). These gliomas are of lysine by methionine at codon 27 of the gene encoding mainly located in thalamus, brain stem and spinal cord and histone variant H3 is the most frequent and typical muta- are attributed mostly to the pediatric age group. However, tion in these tumors that can be detected by sequencing they have been reported in young adults and even rarely in as well as immunohistochemistry (IHC) either directly or patients older than 50 years (Buczkowicz et al. 2014a; Daoud through its consequences, i.e. the global loss of trimeth- et al. 2018). These gliomas are characterized by specific ylation (H3K27me3) on this residue (Venneti et al. 2014). genetic and epigenetic alterations such as somatic mutations Initial studies have associated diffuse midline gliomas with astrocytic high-grade morphology (i.e. high mitotic activity, endothelial proliferation and necrosis) and midline localiza- * Azadeh Ebrahimi tion. However, these tumors can show a broader spectrum [email protected]‑heidelberg.de of histological entities (Joyon et al. 2017; Pages et al. 2016; * Jens Schittenhelm Solomon et al. 2016). The midline localization has been sug- [email protected]‑tuebingen.de gested to be associated with specific molecular pathways Extended author information available on the last page of the article

Vol.:(0123456789)1 3 840 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 in these tumors other than those known for classical dif- brain stem and also interhemispheric tumors with expan- fuse gliomas of other localizations; these molecular changes sion to the midline structures) in the Department of Neuro- make the histology of midline gliomas in treatment and sur- surgery, University Hospital of Tuebingen, between 1996 vival prediction rather insignificant (Paugh et al. 2011; Puget and 2018, including three outpatient consultation cases from et al. 2012). So far, the majority of studies on diffuse midline Department of Pathology, Stuttgart. The study was author- gliomas have been performed on pediatric age group while ized by the ethics board of University Hospital of Tuebin- a group of adult patients also present with H3K27M mutant gen (permission number 494/2016BO2). The patients with midline gliomas (Daoud et al. 2018; Laigle-Donadey et al. a minimum age of 18 years at diagnosis were considered for 2008; Meyronet et al. 2017). The clinically significant and adult cohort. Histological diagnosis and tumor grading in all histopathological characteristics of H3 K27M mutations in samples were reviewed primarily on full slides and revised adult vs. pediatric patients have not yet been studied com- according to the 2016 WHO classification system for CNS prehensively. Current diagnostic guidelines according to tumors by at least two experienced neuropathologists (Louis recommendations of WHO classification of CNS tumors et al. 2016). 2016 on diffuse midline gliomas have extended the relevant molecular examinations for histone mutations from pediatric Immunohistochemistry patients to adult patients younger than 54 years old based on a few findings in adults according to previous studies (Foster H3K27M, ATRX, IDH1R132H and BRAF V600E immu- et al. 2016; Louis et al. 2018). In order to address the impact nostains were performed on an automated immunohisto- of H3F3A mutation in adult patients compared to pediat- chemistry system (BenchMark, Ventana Medical Systems, ric patients, we retrospectively investigated a cohort of 165 Strasbourg, France), as previously described (Schittenhelm patients with neuroepithelial brain tumors of midline locali- et al. 2011). Settings for the immunostains were as follows: zation, aging from 2 to 85 years at first diagnosis. The status OptiView method: CC1 pretreatment for 40 min, primary of H3K27M mutation along with other relevant molecular antibody incubation for 20 min at 42 °C, counterstaining markers such as ATRX, BRAF V600E and IDH1 and IDH2 with hematoxylin. Antibodies were used as listed in Table 1 mutations, as well as methylation status of MGMT promoter, and all other staining reagents were provided from Ventana were investigated in this study. In a subset of these tumors, Medical Systems, Strasbourg, France. The immunohisto- NGS panel sequencing and DNA methylation array analy- chemical staining was performed when sufficient material sis were carried out. We characterized the histopathological was available or when after histological review a relevant aspects of these tumors along with clinical significance of differential diagnosis had to be excluded. 17 out of 165 cases H3 K27M mutations in pediatric vs. adult patients. were analyzed as tissue microarrays (TMA). TMAs were provided as previously described (Ebrahimi et al. 2016) and used for BRAF V600E immunostaining. These cases Materials and methods included ependymomas of midline localization. All 17 tissue cores contained more than 95% representative tumor with at Tissue samples least 80% tumor cell content.

165 neuroepithelial brain tumor samples were enrolled in Molecular studies this study. The samples were obtained from the patients that underwent surgery for neuroepithelial tumors in mid- Further molecular analyses were performed when ade- line structures. In order to avoid a histology bias, the cases quate tissue for DNA extraction was available. Using a were selected by their location based on preoperative imag- BlackPREP FFPE kit (Analytik Jena, Germany), DNA ing studies (thalamus, brainstem, spinal cord, cerebellar of adequate quality and quantity was extracted from the vermis, third and fourth ventricles with expansion to the tumor tissue according to the manufacturer’s instructions.

Table 1 List of antibodies used for immunohistochemistry Antibody Antibody dilution Incubation time Provider

Rabbit anti human Histone H3 (K27M mutant) 1:500 32 min Merck Millipore, Billeria, MA, USA Mouse IgG2a IDH1 R132H (H09) LOT 13211/19 1:400 32 min Dianova, Hamburg, Germany Rabbit anti-ATRX 1:400 40 min Sigma St. Louis, MO, USA Supernatant (Klon VE1), Heidelberg-BRAFV600E 1:5 30 min DKFZ, Dr. Capper Rabbit anti human Tri-Methyl-Histone H3 (K27) (C36B11) 1:200 32 min Cell signaling, Cambridge, UK

1 3 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 841

Tissue was selected from the regions of paraffin blocks MGMT promoter methylation calling was 50%. We defined that presented sufficient tumor content in microscopy a cut-off value of 9% to classify MGMT methylated versus (minimum of 50% tumor cell content). All samples lack- nonmethylated cases (Quillien et al. 2016). ing IDH1R132H mutation based on immunohistochemical staining were further sequenced for IDH1 and IDH2 muta- Generation of DNA methylation array data and gene tions in patients younger than 55 years due to further prob- panel sequencing ability for other IDH mutations in this age group (Chen et al. 2014), or in samples with ATRX loss. H3 K27M Thirteen samples, each containing 200–500 ng of tumor sequencing was performed for all H3 K27M IHC positive DNA, were processed for DNA methylation analysis in tumors when adequate tissue was available. Analysis of department of neuropathology Heidelberg. The Infinium MGMT promoter methylation status was performed for HumanMethylation450 Bead-Chip (450 k) array (Illumina, tumors with anaplastic features (WHO grade III and IV) Carlsbad, California, USA) was used to determine the DNA and all H3 K27M mutated tumors. For IDH1 and IDH2, methylation status of 482,421 CpG sites according to the H3 K27M and MGMT promotor pyrosequencing, the manufacturer’s instructions at the Genomics and Proteom- amplified single-stranded DNA templates were immobi- ics Core Facility of the German Cancer Research Center lized on streptavidin-coated Sepharose high-performance (DKFZ). Methylation results were compared with reference beads and then incubated at 80 °C for 2 min and allowed classes provided by https://www.molecularneuropathol ogy. to anneal to a 0.3 mmol/L sequencing primer at room tem- org (Capper et al. 2018). 11 samples were further sequenced perature. Pyrosequencing was performed using PyroGold in the Department of Neuropathology, Heidelberg, using Reagents on the Pyromark Q24 instrument (all Qiagen, enrichment/hybrid-capture-based next-generation sequenc- Hildesheim, Germany) according to the manufacturer’s ing gene panel analysis covering 130 genes of particular instructions. Raw data were imported into the PyroMark relevance in brain tumors, as previously described (Sahm Q24 software (Version 2.0.7 Qiagen, Hildesheim, Ger- et al. 2016). Tumors with midline glioma mutations other many) and allelic regions of interest were quantified to than H3F3A K27M were stained for H3K27me3 (Table 1). determine the percentages of mutant vs. wild-type alleles according to the percentage relative peak height. Follow- Statistical analysis ing primers were used for amplification in a polymerase chain reaction (PCR), using 150 ng genomic DNA as the Quantitative and statistical analyses were performed using template: JMP 7.0 (SAS Institute, Cary, NJ, USA). For correlation IDH1 fwd: ggtcttcagagaagccatt (biotinylated) analyses we performed unpaired, two-tailed Student’s t test IDH1 rev: gcaaaatcacattattgccaac and the Fisher’s exact test to identify possible significant IDH2 fwd: tgtcctcacagagttcaagc associations or differences between two pairs. For multiple IDH2 rev: ctaggcgaggagctccagt (biotinylated) comparison, ANOVA (one- or two-way) or Kruskal–Wal- H3F3A fwd: tgtttggtagttgcatatgg lis testing was used. To assess the potential interobserver H3F3A rev: tacaagagagacctttgtcc (biotinylated). variability, we performed a Pearsons Chi square test on the Pyrosequencing was performed using following sequenc- data collected by the two neuropathologists and calculated ing primers: Cohen’s kappa coefficient for agreement (Cohen 1968). IDH1: tgatccccataagcat Kaplan–Meier test was performed for survival analysis. IDH2: agcccatcaccattg Univariate analyses of the different variables were obtained H3F3A: caaaagccgctcgca. with 95% confidence intervals (CIs) followed by Cox mul- The level of detection for the pyrosequencing was 5%. tivariate logistic regression analyses and hazard ratio (HR) For MGMT pyrosequencing, a total of 250 ng of genomic calculation. A p value < 0.05 was considered as significant. DNA was subjected to bisulfite conversion, using the “Pro- mega MethylEdge Bisulfid Conversion System (Promega, Madison, WI, USA) according to the manufacturer’s pro- Results tocol. PCR amplification and pyrosequencing covering five CpG islands in exon 1 of the MGMT locus on chromosome A cohort of 165 samples from patients of age between 2 and 10q26 from position 131,265,519 to 131,265,537 were per- 85 years including 42 pediatric and 123 adult patients with formed with the PyroMark Q24 CpG MGMT kit (Qiagen, a history of a neuroepithelial tumor in midline location was Hildesheim, Germany) using nucleotide dispensation order analyzed. The cohort also included a variety of histological GTCGTTAGTCAGTTCGTATCAGTCGTCA. The mean diagnosis as listed in Table 2. Through histological review, percentage of methylated alleles at all five CpG loci was immunohiostochemical staining, confirmative H3F3A used for the analysis. The minimal tumor cell content for pyrosequencing, and further DNA methylation array analysis

1 3 842 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 b 0/5 (0%) 0/7 (0%) 0/9 (0%) BRAF V600E mutation 1/3 (33%) 0/1 (0%) 0/1 (0%) – 0/6 (0%) 0/3 (0%) 0/7 (0%) 0/5 (0%) – 1/47 (2%) 9/29 (31%) 21/59 (36%) 3/24 (12.5%) H3K27M mutation 6/20 (30%) 0/1 (0%) 0/1 (0%) 0/1 (0%) 0/10 (0%) 0/4 (0%) 0/7 (0%) 1/6 (16.7) 0/3 (0%) 40/165 (24.2%) 0/15 (0%) 0/44 (0%) 0/8 (0%) IDH2 mutated 0/9 (0%) 0/0 (0%) 0/1 (0%) 0/0 (0%) 0/1 (0%) 0/1 (0%) 0/0 (0%) 0/2 (0%) 0/1 (0%) 0/82 (0%) 1/15 (7%) 0/45 (0%) 0/9 (0%) IDH1 mutated 3/9 (33%) 0/0 (0%) 0/1 (0%) 1/1 (100%) 0/1 (0%) 0/1 (0%) 0/0 (0%) 0/2 (0%) 0/1 (0%) 5/84 (6%) 1/27 (3.7%) 0/50 (0%) 0/24 (0%) IDH1R132H positive 0/19 (0%) 0/1 (0%) 0/1 (0%) 1/1 (100%) 0/10 (0%) 0/4 (0%) 0/7 (0%) 0/6 (0%) 0/2 (0%) 2/152 (1.3%) 6/29 (21%) 12/59 (20%) 1/21(5%) ATRX loss ATRX 7/17 (41%) 1/1 (100%) 0/1 (0%) 0/1 (0%) 0/9 (0%) 0/4 (0%) 0/7 (0%) 0/6 (0%) 0/3 (0%) 27/158 (17.1%) 20/9 33/26 13/11 Sex (m:f) Sex 14/6 0/1 1/0 0/1 6/4 2/2 3/4 3/3 2/1 97/68 4–75; 43 4–85; 42.1 3–44; 16.1 Age (min.– Age max.; mean value) 14–73; 47.2 35 8 68 2–82; 44.1 2–30; 19.3 17–48; 31.4 9–60; 40.2 6–25; 13.7 2–85; 37.7 1 1 1 4 7 6 3 29 59 24 20 10 165 N III IV I a WHO grade II III II II III I I IV ) a 2016 Molecular and histopathological features of the cohort present features Molecular and histopathological cytoma This case had also H3K27M mutation There is no WHO grade dedicated to pilocytic astrocytomas with the to anaplasia according is no WHO gradeThere current pilocytic astrocytomas dedicated to WHO book of CNS tumors

2 Table a b Tumor histology (WHO histology Tumor astrocytoma Anaplastic Glioblastoma Pilocytic astrocytoma Pilocytic astrocytoma with anaplasia Pilocytic astrocytoma Diffuse astrocytoma Anaplastic pleomorphic- Anaplastic xanthoastro Oligodendroglioma Ependymoma Anaplastic ependymoma Anaplastic Myxopapillary ependymoma Myxopapillary Subependymoma Medulloblastoma All cases

1 3 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 843 and panel sequencing 41 samples were detected as harbor- focal increased proliferation activity. The tumor revealed his- ing the relevant mutations of midline gliomas and classified tologically a pilocytic pattern with loose-textured multipolar as H3 K27M mutant midline gliomas (Table 3: pediatric cells and microcysts and microvascular proliferations in the patients, Table 4: adult patients). 13 samples contained presence of high proliferative activity. The tumor was well sufficient tissue for DNA methylation analysis and a DNA demarcated, located in thalamus and was hyperintense in T2 methylation profile was available in these tumors including and hypointense in T1 mode in MRI that reminded of a low- nine pediatric and four adult patients. All cases matched grade glioma. Parts of the tumor revealed contrast enhance- the established methylation class diffuse midline glioma H3 ment. The contrast-enhancing areas revealed a rapid growth K27M mutant and further confirmed our diagnosis based on in short-term follow-up within 23 days. In both cohorts of H3 K27M IHC and pyrosequencing of the H3 K27 region. pediatric and adult patients, thalamus was the most common These cases included a H3 K27M IHC negative, H3 K27M tumor location. However, the pattern of tumor growth was wild-type, ACVR1 and Tp53 mutant recurrent sample of a different in pediatric compared to the adult patients; espe- pediatric patient with thalamic tumor and histological diag- cially a diffuse growth along the brain axis was more evident nosis, glioblastoma WHO grade IV and loss of H3K27me3. in adult compared to pediatric patients based on the preop- Two H3 K27M mutant pediatric diffuse midline gliomas erative imaging studies [7/29 (24%) vs. 2/13 (15%)] (Fig. 3). revealed low-grade histological features including a sub- Kaplan–Meier survival analysis of the H3 K27M IHC ependymoma (Fig. 1) and a pilocytic astrocytoma (Fig. 2). positive tumors revealed that pediatric patients with H3 The subependymoma showed a diffuse growth pattern with- K27M IHC positive tumors had an inferior prognosis out contrast enhancement and a small exophytic component compared to patients with H3 K27M IHC negative tumors in MRI. In spectroscopic examination, the tumor was also (median survival 80 vs. 482 days, Log-rank p = 0.002); while compatible with a low-grade glioma. In addition, another in adult patients the presence of the H3 K27M IHC positive pediatric diffuse midline glioma showed an initial histologi- tumor did not significantly affect the survival (median sur- cal diagnosis of pilocytic astrocytoma with anaplasia due to vival 164 in wild type tumors vs. 121 days in mutant tumors,

Table 3 Clinical, histopathological and molecular features of the pediatric cohort with H3 K27M mutation or methylation class diffuse midline glioma H3 K27M mutant DM DM DGM DM DM DM DM DM DM DM DM DM G G G G 10 G G G G G G 11 G 12 7 3 4 8 Paent 1 2 5 6 9

Sexf mmfmfmff mf f Age at diagnosis4 78999101215151616 Mitosis / HPF 7142 1025 1123 Endothelial proliferaons Necrosis Tumor locaonThThThIMBSBSSTh Th BS Th Th, 3.V, P Histological diagnosis GBM GBM GBM GBM PA SE AA PA GBM GBM GBM GBM Inial WHO tumor gradingIVIVIVIVIIIII Ia IV IV IV IV Recurrence ATRX IDH1R132H-IHC H3.K27M-IHC H3F3A p.K27M b b b HIST1H3B ACVR1 p.G328E p.G328E p.G328E IDH1 IDH2 BRAFV600E H3.G34 TP53 DNA Methylaon class MGMT

Present Not recurrent Diﬀuse midline glioma H3 K27M mutant Negave Wild type Methylated Loss

Absent Recurrent Not availablePosive Mutated Not methylated Retained GBM glioblastoma, PA pilocytic astrocytoma, SE subependymoma, S spinal, Th thalamus, IM intramedullary, BS brainstem, P pons, V ventricle a This tumor revealed focally up to five mitoses per high-power field b Due to insufficient tissue/DNA for further molecular analysis, other possible mutations could not be ruled out

1 3 844 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850

Table 4 Clinical, histopathological and molecular features of adult cohort with H3 K27M mutation or methylation class diffuse midline glioma H3 K27M mutant DMG 13 DMG 14 DMG 15 DMG 16 DMG 17 DMG 18 DMG 19 DMG DMG 21 DMG 22 DMG 23 DMG 24 DMG 25 DMG 26 DMG 27 DMG 28 DMG 29 DMG 30 DMG 31 DMG 32 DMG 33 DMG 34 DMG 35 DMG 36 DMG 37 DMG 38 DMG 39 DMG 40 DMG 41

Paent 20

Sex mfmmm fmmmmm fmffmmmm fmff ffmmmm Age at diagnosis1820212223252729293030313434374041444957585859616167727373 Mitosis / HPF 11523501 5011 4211 301001011100 5 Endothelial proliferaons Necrosis Th, T, 3.V, IM, Th, Th, Th, Tumor locaon CBSThThIMThThBSThH3.V Th 4.V Th BS BS, BG, IM Th T, IM Th Th IM IM BS BS lat.V BS BG Th Th Histological GBMGBM GBMAAGBM GBMDAAAGBM DA AA GBMGBM GBMAAGBM GBMPAGBM GBMDAAAAADAAAAADADAGBM diagnosis WHO grade based on IV IV IV III IV IV II IIIIVII III IV IV IV IIIIVIVIIV IV II III IIIIIIII IIIIIIIIV histological diagnosis Recurrence ATRX IDH1R132H-IHC H3.K27M-IHC H3F3A p.K27M a a a HIST1H3B ACVR1 IDH1 IDH2 BRAFV600E H3.G34 TP53 DNA Methylaon class MGMT

Present Not recurrent Diﬀuse midline glioma H3 K27M mutant Negave Wild type Methylated Loss

Absent Recurrent Not availablePosive Mutated Not methylated Retained GBM glioblastoma, AA anaplastic astrocytoma, DA diffuse astrocytoma, Th thalamus, IM intramedullary, BS brainstem, P pons, V ventricle, BG basal ganglia, C cerebellum a Due to insufficient tissue/DNA for further molecular analysis, other possible mutations could not be ruled out

Log-rank p = 0.8). This difference in pediatric patients histological approach and also recent studies with molecu- remained significant even when only tumors with diffuse lar approach are focused on pediatric patients. The clinical growth pattern and features of anaplasia (high mitotic rate, significance of H3 K27M mutant tumors in adult patients necrosis, endothelial proliferation) were considered for anal- has not been well defined yet (Bechet et al.2014 ; Korshu- ysis (median survival 91 vs. 543 days, Log-rank p = 0.02) nov et al. 2015). There are a few studies on adult brain stem (Fig. 4). The patients’ survival was independent of the spe- gliomas, in most of which the presence of histone muta- cific tumor location (data not shown). MGMT-promoter was tions have not been assessed and the diagnosis has been not methylated in any of gliomas in our cohort. made based on the localization and clinical course of the tumor (Theeler et al. 2015). In one study of adult DIPG, patients of age between 18 and 65 years were studied from Discussion which 5 patients aged between 22 and 65 years showed H3 K27M mutations (Reyes-Botero et al. 2014). Among these Recent molecular discoveries in glial brain tumors have five patients, two had supratentorial tumors in frontal and changed their definition dramatically (Sturm et al. 2012). fronto-insular lobe instead of midline structures. In the pre- Diffuse midline gliomas, H3 K27M mutant, has been intro- sent study, we selected the tumors by their location covering duced in the latest WHO classifications of CNS tumors a broad spectrum of age ranges. We intentionally included and characterized as a grade IV neoplasm independent of non-diffusely growing neuroepithelial tumors to avoid a histological appearance. Most of the previous studies with selection bias introduced by histological preselection and

1 3 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 845

Fig. 1 Diffuse midline glioma with initial histological diagnosis of of glial cell processes with frequent micro cystic changes. It lacked subependymoma grade I; the tumor was located in brain stem and mitotic figures, necrosis and endothelial proliferation histologically consisted of clusters of small uniform nuclei in a matrix possible diagnostic changes used in older WHO classifica- immunohistochemistry and all cases were negative (Schafer tion schemes. Surprisingly, we found out that H3 K27M et al. 2018). Together, these results clearly indicate that H3 mutations in adult midline tumors are frequent (24% of all K27M mutations in CNS tumors are location-specific but adult cases fulfilled the criteria for a diffuse midline glioma, not age-related. Still, there are some tumors in our cohort H3 K27M mutant). Meyronet et al report the frequency of that would qualify by high-grade histology, clinical charac- H3 K27M mutations about 13% in adult patients (Meyronet teristics and imaging as diffuse midline gliomas, although et al. 2017), while Daoud et al reported 28% (7 out of 25 the H3 K27M mutations were absent in these cases. Some patients) in a smaller collective of patients compared to our authors, therefore, propose to introduce a H3 K27M wild- study (Daoud et al. 2018). Previously, we have screened type tumor class into the WHO classification of brain tumors more than 150 supratentorial gliomas for H3 K27M by (von Bueren et al. 2018). In our study, thalamus was the

1 3 846 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850

Fig. 2 Diffuse midline glioma with initial histological diagnosis of high-power field and endothelial proliferations. A piloid aspect with pilocytic astrocytoma grade I; the tumor was located in thalamus and bipolar elongated thin processes was evident in intraoperative cyto- histologically revealed a loose myxoid texture with moderate cellu- logical smear larity and multipolar cells as well as focally up to five mitoses per most frequent location for these tumors; however, it should et al. 2014; Taylor et al. 2014; Wu et al. 2014). ACVR1 be taken into consideration that many brain stem tumors are mutations have been shown to constitutively activate the not biopsied at all when the imaging findings are typical and BMP-TGF-ß signaling pathway. However, the relevance of in these cases the patients receive treatment without histo- these mutations with tumorigenesis in midline gliomas has logical confirmation (Walker et al. 2013). In accordance with not yet been well addressed. previous reports the H3F3A K27M mutation is the most fre- A variety of histological patterns in diffuse midline glio- quent mutation found in diffuse midline gliomas (95% in our mas have been described in previous case series and reports. cohort) and has a clear-cut clinical consequence on patients’ The most frequently reported patterns include diffuse astro- survival. The remaining cases (5%) harbored HIST1H3B cytomas consisting of high grade and low grade (Buczkow- K27M mutation that was also reported in association with icz et al. 2014a). Primitive neuroectodermal tumor (PNET) diffuse midline gliomas. ACVR1 is the third most common like morphology has been described in DIPGs or gliomas of mutation after p53 mutation in 20–32% of diffuse midline midline localization according to a few studies (Buczkowicz gliomas (Buczkowicz et al. 2014b; Taylor et al. 2014; Wu et al. 2014a; Sufit et al. 2012; Zagzag et al. 2000). How- et al. 2014). In our cohort, the case with absent K27M muta- ever, this morphology is not uncommon in glioblastomas, tions that clustered into the midline glioma group was also resulting from aberrant MYC amplifications or treatment ACVR1 mutant and had a loss of H3K27me3 expression. with immune-checkpoint inhibitors. Therefore, it is possible The ACVR1 gene encodes the receptor serine/threonine that other factors than the H3 K27M mutations contribute kinase ALK2 and was found to harbor nonsynonymous to this tumor appearance. Similarly, some H3 G34-mutant heterozygous somatic mutations at specific residues includ- pediatric tumors have been misdiagnosed as CNS-PNETs ing R206H, Q207E, R258G, G328E, G328V, G328W and in the past (Korshunov et al. 2016). Ganglioglioma has G356D substitutions (Buczkowicz et al. 2014b; Fontebasso been reported as a histological entity in H3 K27M mutated

1 3 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 847

Fig. 3 Tumor location in adult (a) and pediatric (b) patients

tumors recently (Okuda et al. 2018). In a study of 54 gan- tumors with low-grade histology: a case of pilocytic astro- gliogliomas of midline location, 9.3% revealed a H3 K27M cytoma of brain stem and a subependymoma WHO grade mutation (Pages et al. 2016). There was a co-occurrence I of lateral ventricle again indicating that these mutations of BRAFV600E mutation in all H3 K27M mutated mid- are not restricted to tumors with high-grade histological line gliomas in this study (Pages et al. 2016). The status of features. These two low-grade histologies have been previ- BRAFV600E mutation was available in 16 H3 K27M IHC ously reported rarely as single cases (Morita et al. 2018; positive diffuse midline gliomas in our study and one of Orillac et al. 2016; Reers et al. 2017). According to the rec- them revealed an additional BRAF V600E mutation (pedi- ommendations of the cIMPACT consortium, tumors lacking atric case DMG11), suggesting that H3 K27M mutations a diffuse growth pattern should not be classified as diffuse are independent of co-occurring MAPK-pathway activation. midline glioma, H3 K27M-mutant despite the presence of Co-occurrence of these two mutations in pediatric thala- H3 K27M mutations (Louis et al. 2018). In an autopsy-based mus gliomas has been shown in another report too in which study of DIPG among 44 patients 38.6% of cases revealed a H3K27M is shown to be the dominant prognostic indicator leptomeningeal spread of the tumor at autopsy, while 25% (Ryall et al. 2016). However, BRAF V600E mutant diffuse had tumor cell-infiltrating structures beyond the brainstem midline gliomas might be considered for therapy with BRAF (spinal cord and/or thalamic involvement) or tumor cells inhibitors or other MAPK-pathway interactors, though there diffusely infiltrating as far rostrally as the frontal lobe; how- are no established studies on efficacy of these therapeutic ever, no correlation between the tumor dissemination and options in diffuse midline gliomas yet. In addition to the histological grade of the tumor was noticed (Buczkowicz broader age range, our study revealed two H3 K27M mutant et al. 2014a). In the mentioned study, localizations in brain

1 3 848 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850

have been missed (Cage et al. 2013; Roujeau et al. 2007). However, some of these tumors might even show the characteristics of low-grade tumors in imaging studies too, similar to the H3 K27M mutant midline gliomas with pilocytic astrocytoma and subependymoma histology in our cohort. Histone 3 mutations in midline gliomas have been shown to be associated with poor prognosis and low survival in adult and pediatric patients in previous studies (Daoud et al. 2018; Feng et al. 2015; Karremann et al. 2018). Accord- ing to our study, H3 K27M mutant diffuse midline gliomas impose a worse prognosis in pediatric patients compared to wild-type tumors irrespective of histology, while the prognostic role of these mutations in adult patients seems to be similar to K27M wild-type high-grade tumors. Recent reports also indicated that H3 K27M mutant gliomas show an unmethylated MGMT promotor status as opposed to the hypermethylation phenotype of IDH-mutant tumors (Banan et al. 2017). In our study, we were able to confirm these results; all H3 K27M mutant tumors analyzed for their MGMT status in our cohort displayed an unmethylated MGMT promotor. In conclusion, H3 K27M mutations can be found with higher frequency in adult midline gliomas as previously thought; however, the presence of this mutation in midline glial tumors imposes a worse prognosis rather in pediatric than in adult patients and this effect is independent of tumor histology and tumor localization. Limitations of our study are the retrospective nature of our cohort with divergent treatments due to the different initial diagnoses and grading Fig. 4 Kaplan–Meier’s survival analysis of pediatric and adult patients; pediatric patients with H3K27M mutation had a signifi- based on pure histological features in the pre H3 K27M-IHC cantly lower survival than H3K27M wild type, when histological fea- era as well as limited material availability for confirmation tures of anaplasia including high mitotic rate, necrosis and endothe- of mutations via sequencing in some cases. lial proliferation were present (tumors with diffuse growth pattern and initial WHO grade III and IV) (median survival 91 vs. 543 days, Log- Funding JS is supported by a Grant from the Else Uebelmesser Foun- p p rank = 0.02), * value < 0.05 dation for Applied Cancer Research, Tuebingen, Germany. GT served on Advisory Boards of BMS, Roche Switzerland, MSD Switzerland, received travel grants from Roche Switzerland, MSD Switzerland, stem as well as the poor clinico-pathological course of the Medac; received research support/Grants from Roche Diagnostics and disease were considered as the diagnostic criteria (Buczko- Medac. wicz et al. 2014a). None of our H3 K27M mutant midline gliomas had a leptomeningeal spread, possibly a result of Compliance with ethical standards early-onset biopsy. However, a diffuse tumor growth along Conflict of interest the axis was more frequent in adult than pediatric patients. The authors have no conflict of interest. Previous studies have shown that neither the histological grade of diffuse midline gliomas (Joyon et al. 2017; Pages et al. 2016; Solomon et al. 2016) nor the tumor localization References (Karremann et al. 2018) should be regarded as the reference for survival estimation of patients. The presence of Banan R, Christians A, Bartels S, Lehmann U, Hartmann C (2017) a K27M mutation in a low-grade diffuse glioma in a small Absence of MGMT promoter methylation in diffuse midline gli- specimen is most likely due to the technical restriction in oma, H3 K27M-mutant. Acta Neuropathol Commun 5:98. https obtaining tissue during the biopsy procedure. Since these ://doi.org/10.1186/s40478-017-0500-2 tumors are mainly located in vital anatomical structures, the Bechet D et al (2014) Specific detection of methionine 27 mutation in histone 3 variants (H3K27M) in fixed tissue from high-grade tumor cannot be reached easily and the biopsied tissue might astrocytomas. Acta Neuropathol 128:733–741. https://doi. not be representative and possible higher-grade regions may org/10.1007/s00401-014-1337-4

1 3 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850 849

Buczkowicz P, Bartels U, Bouffet E, Becher O, Hawkins C (2014a) Louis D, Ohgaki H, Wiestler O, Cavenee WK (2016) World health Histopathological spectrum of paediatric diffuse intrinsic pontine organization classification of tumours of the central nervous glioma: diagnostic and therapeutic implications. Acta Neuropathol system. In: Bosman F, Jaffe E, Lakhani S, Ohgaki H (eds) World 128:573–581. https://doi.org/10.1007/s00401-014-1319-6 health organization classification of tumours revised, 4th edn. Buczkowicz P et al (2014b) Genomic analysis of diffuse intrinsic pon- IARC, Lyon tine gliomas identifies three molecular subgroups and recurrent Louis DN et al (2018) cIMPACT-NOW update 2: diagnostic clari- activating ACVR1 mutations. Nat Genet 46:451–456. https://doi. fications for diffuse midline glioma, H3 K27M-mutant and dif- org/10.1038/ng.2936 fuse astrocytoma/anaplastic astrocytoma, IDH-mutant. Acta Cage TA et al (2013) Feasibility, safety, and indications for surgical Neuropathol 135:639–642. https://doi.org/10.1007/s0040 biopsy of intrinsic brainstem tumors in children. Childs Nerv Syst 1-018-1826-y ChNS Off J Int Soc Pediatric Neurosurg 29:1313–1319. https:// Meyronet D et al (2017) Characteristics of H3 K27M-mutant gliomas doi.org/10.1007/s00381-013-2101-0 in adults. Neurooncology 19:1127–1134. https://doi.org/10.1093/ Capper D et al (2018) DNA methylation-based classification of cen- neuonc/now274 tral nervous system tumours. Nature 555:469–474. https://doi. Morita S et al (2018) Brainstem pilocytic astrocytoma with H3 K27M org/10.1038/nature26000 mutation: case report. J Neurosurg 129:593–597. https://doi. Castel D et al (2015) Histone H3F3A and HIST1H3B K27M mutations org/10.3171/2017.4.JNS162443 define two subgroups of diffuse intrinsic pontine gliomas with dif- Okuda T et al (2018) Pediatric ganglioglioma with an H3 K27M muta- ferent prognosis and phenotypes. Acta Neuropathol 130:815–827. tion arising from the cervical spinal cord. Neuropathology. https https://doi.org/10.1007/s00401-015-1478-0 ://doi.org/10.1111/neup.12471 Chen L et al (2014) Predicting the likelihood of an isocitrate dehy- Orillac C et al (2016) Pilocytic astrocytoma and glioneuronal tumor drogenase 1 or 2 mutation in diagnoses of infiltrative glioma. with histone H3 K27M mutation. Acta Neuropathol Commun Neurooncology 16:1478–1483. https://doi.org/10.1093/neuonc/ 4:84. https://doi.org/10.1186/s40478-016-0361-0 nou097 Pages M et al (2016) Co-occurrence of histone H3 K27M and BRAF Cohen J (1968) Weighted kappa: nominal scale agreement with pro- V600E mutations in paediatric midline grade I ganglioglioma. vision for scaled disagreement or partial credit. Psychol Bull Brain Pathol. https://doi.org/10.1111/bpa.12473 70:213–220 Paugh BS et al (2011) Genome-wide analyses identify recurrent Daoud EV et al (2018) Adult brainstem gliomas with H3K27M muta- amplifications of receptor tyrosine kinases and cell-cycle regula- tion: radiology, pathology, and prognosis. J Neuropathol Exp Neu- tory genes in diffuse intrinsic pontine glioma. J Clin Oncol Off rol 77:302–311. https://doi.org/10.1093/jnen/nly006 J Am Soc Clin Oncol 29:3999–4006. https://doi.org/10.1200/ Ebrahimi A et al (2016) ATRX immunostaining predicts IDH and JCO.2011.35.5677 H3F3A status in gliomas. Acta Neuropathol Commun 4:60. https Puget S et al (2012) Mesenchymal transition and PDGFRA amplifica- ://doi.org/10.1186/s40478-016-0331-6 tion/mutation are key distinct oncogenic events in pediatric dif- Feng J et al (2015) The H3.3 K27M mutation results in a poorer fuse intrinsic pontine gliomas. PLoS One 7:e30313. https://doi. prognosis in brainstem gliomas than thalamic gliomas in adults. org/10.1371/journal.pone.0030313 Hum Pathol 46:1626–1632. https://doi.org/10.1016/j.humpa Quillien V et al (2016) Validation of the high-performance of pyrose- th.2015.07.002 quencing for clinical MGMT testing on a cohort of glioblastoma Fontebasso AM et al (2014) Recurrent somatic mutations in ACVR1 in patients from a prospective dedicated multicentric trial. Onco- pediatric midline high-grade astrocytoma. Nat Genet 46:462–466. target 7:61916–61929. https://doi.org/10.18632 /oncot arget .11322 https://doi.org/10.1038/ng.2950 Reers S, Krug D, Stummer W, Hasselblatt M (2017) Malignant pro- Foster SA et al (2016) Activation mechanism of oncogenic deletion gression of a histone H3.3 K27M-mutated spinal pilocytic astro- mutations in BRAF, EGFR, and HER2. Cancer Cell 29:477–493. cytoma in an adult. Clin Neuropathol 36:83–85. https://doi. https://doi.org/10.1016/j.ccell.2016.02.010 org/10.5414/NP300990 Hoffman LM et al (2016) Spatial genomic heterogeneity in diffuse Reyes-Botero G et al (2014) Molecular analysis of diffuse intrinsic intrinsic pontine and midline high-grade glioma: implications for brainstem gliomas in adults. J Neurooncol 116:405–411. https:// diagnostic biopsy and targeted therapeutics. Acta Neuropathol doi.org/10.1007/s11060-013-1312-2 Commun 4:1. https://doi.org/10.1186/s40478-015-0269-0 Robison NJ, Kieran MW (2014) Diffuse intrinsic pontine glioma: a Joyon N et al (2017) K27M mutation in H3F3A in ganglioglioma reassessment. J Neurooncol 119:7–15. https://doi.org/10.1007/ grade I with spontaneous malignant transformation extends the s11060-014-1448-8 histopathological spectrum of the histone H3 oncogenic pathway. Roujeau T et al (2007) Stereotactic biopsy of diffuse pontine lesions Neuropathol Appl Neurobiol 43:271–276. https://doi.org/10.1111/ in children. J Neurosurg 107:1–4. https://doi.org/10.3171/ nan.12329 PED-07/07/001 Karremann M et al (2018) Diffuse high-grade gliomas with H3 K27M Ryall S et al (2016) Targeted detection of genetic alterations reveal the mutations carry a dismal prognosis independent of tumor loca- prognostic impact of H3K27M and MAPK pathway aberrations tion. Neurooncology 20:123–131. https://doi.org/10.1093/neuon in paediatric thalamic glioma. Acta Neuropathol Commun 4:93. c/nox149 https://doi.org/10.1186/s40478-016-0353-0 Korshunov A et al (2015) Integrated analysis of pediatric glioblastoma Sahm F et al (2016) Next-generation sequencing in routine brain reveals a subset of biologically favorable tumors with associated tumor diagnostics enables an integrated diagnosis and identifies molecular prognostic markers. Acta Neuropathol 129:669–678. actionable targets. Acta Neuropathol 131:903–910. https://doi. https://doi.org/10.1007/s00401-015-1405-4 org/10.1007/s00401-015-1519-8 Korshunov A et al (2016) Histologically distinct neuroepithelial tumors Schafer S et al (2018) Low FoxG1 and high Olig-2 labelling indi- with histone 3 G34 mutation are molecularly similar and comprise ces define a prognostically favourable subset in isocitrate dehy- a single nosologic entity. Acta Neuropathol 131:137–146. https:// drogenase (IDH)-mutant gliomas. Neuropathol Appl Neurobiol doi.org/10.1007/s00401-015-1493-1 44:207–223 https://doi.org/10.1111/nan.12447 Laigle-Donadey F, Doz F, Delattre JY (2008) Brainstem gliomas in Schittenhelm J, Mittelbronn M, Meyermann R, Melms A, Tatagiba M, children and adults. Curr Opin Oncol 20:662–667. https://doi. Capper D (2011) Confirmation of R132H mutation of isocitrate org/10.1097/CCO.0b013e32831186e0 dehydrogenase 1 as an independent prognostic factor in anaplastic

1 3 850 Journal of Cancer Research and Clinical Oncology (2019) 145:839–850

astrocytoma. Acta Neuropathologica 122:651–652. https://doi. Venneti S et al (2014) A sensitive and specific histopathologic prog- org/10.1007/s00401-011-0885-0 nostic marker for H3F3A K27M mutant pediatric glioblastomas. Solomon DA, Wood MD, Tihan T, Bollen AW, Gupta N, Phillips JJ, Acta Neuropathol 128:743–753. https://doi.org/10.1007/s0040 Perry A (2016) Diffuse midline gliomas with histone H3-K27M 1-014-1338-3 mutation: a series of 47 cases assessing the spectrum of mor- von Bueren AO et al (2018) A suggestion to introduce the diagnosis phologic variation and associated genetic. Alter Brain Pathol of “diffuse midline glioma of the pons, H3 K27 wildtype (WHO 26:569–580. https://doi.org/10.1111/bpa.12336 grade IV)”. Acta Neuropathol. https://doi.org/10.1007/s0040 Sturm D et al (2012) Hotspot mutations in H3F3A and IDH1 define 1-018-1863-6 distinct epigenetic and biological subgroups of glioblastoma. Can- Walker DA et al (2013) A multi-disciplinary consensus statement con- cer Cell 22:425–437. https://doi.org/10.1016/j.ccr.2012.08.024 cerning surgical approaches to low-grade, high-grade astrocyto- Sufit A et al (2012) Diffuse intrinsic pontine tumors: a study of primi- mas and diffuse intrinsic pontine gliomas in childhood (CPN Paris tive neuroectodermal tumors versus the more common diffuse 2011) using the Delphi method. Neurooncology 15:462–468. https intrinsic pontine gliomas. J Neurosurg Pediatr 10:81–88. https:// ://doi.org/10.1093/neuonc/nos330 doi.org/10.3171/2012.3.PEDS11316 Wu G et al (2014) The genomic landscape of diffuse intrinsic pontine Taylor KR et al (2014) Recurrent activating ACVR1 mutations in dif- glioma and pediatric non-brainstem high-grade glioma. Nat Genet fuse intrinsic pontine glioma. Nat Genet 46:457–461. https://doi. 46:444–450. https://doi.org/10.1038/ng.2938 org/10.1038/ng.2925 Zagzag D et al (2000) Primitive neuroectodermal tumors of the brain- Theeler BJ et al (2015) Adult brainstem gliomas: correlation of clini- stem: investigation of seven cases. Pediatrics 106:1045–1053 cal and molecular features. J Neurol Sci 353:92–97. https://doi. org/10.1016/j.jns.2015.04.014

Affiliations

1 Department of Neuropathology, Institute of Pathology 6 Center for Personalized Medicine, Eberhard Karls University and Neuropathology, University Hospital of Tuebingen, of Tuebingen, Tuebingen, Germany Eberhard Karls University of Tuebingen, 72076 Tuebingen, 7 German Consortium for Translational Cancer Research Germany (DKTK), DKFZ Partner Site Tuebingen, Tuebingen, 2 Clinical Cooperation Unit Neuropathology, German Cancer Germany Consortium (DKTK), German Cancer Research Center 8 Center for CNS Tumors, Comprehensive Cancer Center (DKFZ), Heidelberg, Germany Tuebingen‑Stuttgart, University Hospital of Tuebingen, 3 Department of Neurosurgery, University Hospital Eberhard Karls University of Tuebingen, Tuebingen, of Tuebingen, Eberhard Karls University of Tuebingen, Germany 72076 Tuebingen, Germany 9 Department of General Pediatrics, Hematology/Oncology, 4 Interdisciplinary Division of Neurooncology, Departments University Children’s Hospital, 72076 Tuebingen, Germany of Vascular Neurology and Neurosurgery, University 10 Department of Neuropathology, Institute of Pathology, Hospital of Tuebingen, Eberhard Karls University University Hospital of Heidelberg, Im Neuenheimer Feld of Tuebingen, 72076 Tuebingen, Germany 224, 69120 Heidelberg, Germany 5 Laboratory for Clinical and Experimental Neurooncology, 11 Department of Pathology, Katharinenhospital Stuttgart, Hertie-Institute for Clinical Brain Research, Tuebingen, Stuttgart, Germany Germany

1 3