Characteristics of diffuse hemispheric gliomas, H3 G34-mutant in adults

Abstract Background Diffuse hemispheric gliomas, H3 G34-mutant (DHG H3G34-mutant) constitute a distinct type of aggressive brain tumors. Although initially described in children, they can also affect adults. The aims of this study were to describe the characteristics of DHG H3G34-mutant in adults and to compare them to those of established types of adult WHO grade IV gliomas. Methods The characteristics of 17 adult DHG H3G34-mutant, 32 H3.3 K27M-mutant diffuse midline gliomas (DMG), 100 IDH-wildtype, and 36 IDH-mutant glioblastomas were retrospectively analyzed. Results Median age at diagnosis in adult DHG H3G34-mutant was 25 years (range: 19–33). All tumors were hemispheric. For 9 patients (56%), absent or faint contrast enhancement initially suggested another diagnosis than a high-grade glioma, and diffusion-weighted imaging seemed retrospectively more helpful to suspect an aggressive tumor than MR-spectroscopy and perfusion MRI. All cases were IDH-wildtype. Most cases were immunonegative for ATRX (93%) and Olig2 (100%) and exhibited MGMT promoter methylation (82%). The clinical and radiological presentations of adult DHG H3G34-mutant were different from those of established types of adult grade IV gliomas. Median overall survival of adult DHG H3G34-mutant was 12.4 months compared to 19.6 months (P = .56), 11.7 months (P = .45), and 50.5 months (P = .006) in H3.3 K27M-mutant DMG, IDH-wildtype, and IDH-mutant glioblastomas, respectively. Conclusions Adult DHG H3G34-mutant are associated with distinct characteristics compared to those of established types of adult WHO grade IV gliomas. This study supports considering these tumors as a new type of WHO grade IV glioma in future classifications.

gliomas. Median overall survival of adult DHG H3G34-mutant was 12.4 months compared to 19.6 months (P = .56), 11.7 months (P = . 45), and 50.5 months (P = .006) in H3.3 K27M-mutant DMG, IDH-wildtype, and IDH-mutant glioblastomas, respectively. Conclusions. Adult DHG H3G34-mutant are associated with distinct characteristics compared to those of established types of adult WHO grade IV gliomas. This study supports considering these tumors as a new type of WHO grade IV glioma in future classifications.

Key Points
• Adult diffuse gliomas G34-mutant display a misleading radiological presentation.
• Adult diffuse gliomas G34-mutant differ from other grade IV gliomas.
• Adult diffuse gliomas G34-mutant have a poor outcome.
Recurrent somatic gain-of-function mutations in the genes encoding the histone H3 variants H3.3 and H3.1 are encountered in and characterize 2 different types of high-grade gliomas that have initially been described in the pediatric population but can also affect adults. [1][2][3][4] Lysine to methionine substitution at codon 27 in the H3F3A or in the HIST1H3B/C genes (H3.3 K27M mutation) are associated with aggressive diffuse midline gliomas and define "Diffuse midline glioma, H3.3 K27M-mutant" which corresponds to grade IV in the revised 2016 World Health Organization (WHO) classification. 5 In contrast, glycine to arginine or valine substitutions at codon 34 in the H3F3A gene (H3.3 G34 mutation) are associated with non-midline hemispheric high-grade gliomas. [6][7][8] H3.3 K27M and H3.3 G34 mutations are mutually exclusive and exclusive with IDH mutations. 1,9,10 H3.3 G34 mutations are about twice less frequent than H3.3 K27M mutations. They are encountered in less than 1% of all gliomas but in 15% of high-grade gliomas in adolescents and young adults. 7,11,12 In the 2016 WHO classification, H3.3 G34-mutant diffuse gliomas were not considered as a distinct tumor type and most cases are currently classified as IDH-wildtype glioblastomas. 5 However, in future classifications, H3.3 G34mutant diffuse gliomas will be individualized. In its sixth update, cIMPACT-NOW (the consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy-Not Official WHO) indeed recommended to consider H3.3 G34mutant diffuse gliomas as a new WHO grade IV tumor type that should be referred to as "Diffuse hemispheric glioma, H3.3 G34-mutant" and the grade will be given in arabic. 13 In the fifth edition of the CNS WHO classification, which is in progress, the suggested name is consequently "Diffuse hemispheric glioma, H3G34-mutant, " grade 4. Since the median age at diagnosis in diffuse hemispheric gliomas, H3 G34-mutant is around 18-19 years, half of the patients are followed by adult neuro-oncologists. 3,4,6,10 Yet, the characteristics and the outcome of adult diffuse hemispheric gliomas, H3 G34-mutant remain to be fully described. [14][15][16] Consequently, the aim of this study was to report the characteristics of diffuse hemispheric gliomas, H3 G34-mutant in adults and compare them to those of established types of adult WHO grade IV gliomas.

Importance of the Study
In future classifications, diffuse gliomas H3 G34-mutant will be considered as a new WHO grade IV tumor type. Since the median age at diagnosis in diffuse hemispheric gliomas, H3 G34-mutant is around 20 years, half of the patients are followed by adult neurooncologists. This study provides a detailed analysis of clinical, radiological, and survival of a series of 17 adult diffuse hemispheric gliomas, H3 G34-mutant. The results highlight that adult diffuse hemispheric gliomas, H3 G34-mutant frequently display a misleading radiological presentation that can lead to a potentially detrimental diagnostic delay. Although these gliomas have a poor prognosis, similar to that of H3.3 K27M-mutant diffuse midline gliomas and IDH-wildtype glioblastomas, they display distinct characteristics compared to established types of WHO grade IV adult gliomas.

Neuro-Oncology Advances
Nice, Paris, and Toulouse) between January 2012 and May 2018. Their clinical, radiological, histological, and molecular characteristics were retrospectively reviewed.
The characteristics of adult diffuse hemispheric gliomas, H3 G34-mutant were compared to those of retrospective contemporaneous series of H3.3 K27M-mutant diffuse midline gliomas (DMG), IDH-wildtype, and IDH-mutant glioblastomas diagnosed in Lyon University Hospital according to the revised WHO 2016 classification. The inclusion criteria were the following: (1) age ≥ 18 years, (2) available MRI for radiological review, (3) and available TERT promoter mutation status for IDH-wildtype glioblastomas.
Brain MRI scans were independently reviewed by 3 investigators (T.P., C.I., and L.P.B.) for the location of T2/fluidattenuated inversion recovery (FLAIR) hyperintensity, the presence and type (fainted, nodular, or ring-like) of parenchymal contrast enhancement, the presence of apparent diffusion coefficient (ADC) restriction, as well as abnormalities on proton magnetic resonance spectroscopy (1H-MRS) and, dynamic contrast-enhanced perfusion MRI. For patients with available MRI follow-up, tumor progression was defined according to RANO criteria. 17

Molecular Analysis
DNA was extracted from formalin-fixed paraffin-embedded samples using a standard protocol (QIAmp DNA mini Kit, Qiagen). After PCR amplification, the mutations of codon 132 of IDH1, codon 172 of IDH2, codons 27 and 34 of H3F3A, codon 28 of HIST1H3B, codon 600 of BRAF, as well as mutational hotspots C228 and C250 of the TERT promoter were investigated by single-nucleotide primer extension with the ABI PRISM SNaPshot Multiplex kit, Applied Biosystems. 18 MGMT promoter methylation was assessed by pyrosequencing, which was performed using the PyroMark Q96 MGMT kit (Qiagen) on a PSQTM96 MA system (Biotage). 19 EGFR gene amplification was detected by using SYBR Green real-time quantitative polymerase chain reaction (qPCR) analysis (Absolute SYBR Green Rox Mix; Abgene) with the following set of primers EGFR-F: GTGCAGATCGCAAAGGTAATCAG, EGFR-R: GCAGACCGCATGTGAGGAT; hydrolysis probe: FAM CCCCTCCCCCGTATCTC. 20

Statistical Analysis
Categorical comparisons were performed using the Fisher's exact test or the Chi2 test when the conditions of application were met. The nonparametric Wilcoxon test was used for quantitative variable as the hypothesis of normality of distribution was not verified. Overall survival and progression-free survival were represented with Kaplan-Meier plots and compared using the log-rank test.
The statistical tests were bilateral and the level of significance was set at 5% (P < .05). Statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc.).

Standard Protocol Approvals and Registrations
Study design and manuscript organization were guided by the STROBE statement on cohort studies. The study was approved by the Institutional Review Board (local number . Informed consents were obtained from alive patients.

Radiological Presentation
Initial MRI scans were available for review for 16/17 patients. Radiological characteristics are summarized in Table  1. One patient (patient 6) underwent a brain MRI which was normal, 3 years before diagnosis. At diagnosis, all tumors had a hemispheric location and were monocentric. Midline involvement was observed in 4 patients but always as an extension of a primarily hemispheric tumor. Most cases demonstrated no or only a faint contrast-enhancement (n = 11/15, 73%) and presented as cortico-subcortical (n = 13/16, 81%), poorly delimited (n = 9/16, 69%), infiltrative lesions that were mostly located in the frontoparietal (n = 11/16, 69%) or in the temporo-insular lobes (n = 3/16, 19%). All but one tumor (n = 12/13, 92%) harbored areas of ADC restriction on diffusion-weighted imaging. Only half of the cases in which perfusion MRI and/or MR spectroscopy were performed presented an elevated (>1.7) intratumoral relative cerebral blood flow (50%, n = 4/8) or a metabolic profile suggestive of an aggressive tumor (elevated Choline/N-Acetyl-Aspartate ratio > 2 and/or of lipid/ lactate peaks, 44%, n = 4/9).
For 6 patients, the initial radiological presentation led to suspect a high-grade glioma. For 10 cases, however, the young age of the patient, absent or faint contrast enhancement (n = 9) or the presence of an isolated hemorrhage (n = 1) led to suspect another diagnosis ( Figure 2). In these patients, the initially suspected diagnoses were the following: low-grade glioma (n = 3), primary CNS lymphoma (n = 2), multiple sclerosis (n = 2), limbic encephalitis (n = 1), metabolic disease (n = 1), and ruptured arterio-venous malformation (n = 1). Subsequently, in 8 patients, the decision was made to perform an MRI follow-up leading to a median delay of 2.3 months before histological diagnosis. Retrospectively, in all but one of these patients, the initial MRI demonstrated focal ADC restriction that could have led to suspect an aggressive glioma. In contrast, an elevated intra-tumoral relative cerebral blood flow and an aggressive metabolic profile on MR spectroscopy were both present in only 2/8 patients (25%).
Among the 11 patients with initially absent or faint contrast-enhancement, 8 underwent a repeated MRI before (n = 4) or after (n = 4) histological diagnosis that demonstrated the apparition of a ring-like or nodular contrast enhancement in all of the patients after a median interval of 2.6 months (range: 0.8-5.3 months).

Treatment and Outcome
A total of 12 patients underwent a biopsy (70%), 1 underwent a partial (6%) and 4 a complete resection (24%). One patient presented with a brain hematoma leading to neurological worsening after biopsy and one patient presented a lethal refractory raised intracranial pressure after biopsy. In the 15 remaining patients, postsurgical course was uneventful. Postoperative oncological treatment consisted of temozolomide radio-chemotherapy (n = 13, 82%), radiotherapy alone (n = 1, 6%), radiotherapy followed by PCV (procarbazine, CCNU, and vincristine) chemotherapy (n = 1, 6%), and temozolomide chemotherapy alone (n = 1, 6%). No severe adverse effect was reported during the oncological treatment and no patient developed drugresistant epilepsy.

Comparison of Adult Diffuse Hemispheric Gliomas, H3 G34-mutant With Established Types of Adult WHO Grade IV Gliomas
As given in Table 2, several characteristics of adult diffuse hemispheric gliomas, H3 G34-mutant differed significantly from those of established types of adult WHO grade IV gliomas. Compared to H3.3 K27M-mutant DMG, IDH-mutant and IDH-wildtype glioblastomas, diffuse hemispheric gliomas, H3 G34-mutant occurred in younger patients, more frequently involved the parietal lobe, more frequently presented as a hemorrhagic tumor at diagnosis, less frequently demonstrated contrast-enhancement and necrosis, more frequently displayed ADC restriction and less frequently expressed Olig2. In addition, TERT promoter mutations were lacking in diffuse hemispheric gliomas in contrast to IDH-wildtype glioblastomas while MGMT promoter methylation was more frequent in diffuse hemispheric gliomas, H3 G34-mutant than in H3.3 K27Mmutant DMG and IDH-wildtype glioblastomas. Median overall survival of adult diffuse hemispheric gliomas, H3 G34-mutant was 12.4 months compared to 19.6 months (P = .56), 11.7 months (P = .45), and 50.5 months (P = .006) in H3.3 K27M-mutant DMG, IDH-wildtype glioblastomas, and IDH-mutant glioblastomas, respectively ( Figure 3). The comparison yielded similar results when the analysis was restricted to H3.3 K27M-mutant DMG and IDH-mutant glioblastomas with comparable age (ie, aged < 35 years, data not shown). In IDH-wildtype glioblastomas, the number of patients aged < 35 years was too limited to allow this analysis.

Discussion
Diffuse hemispheric gliomas, H3 G34-mutant are rare brain tumors that affect children and adolescents but also adults. 4,10,12,15 In this study, that is the first dedicated to adult cases, we show that adult diffuse hemispheric gliomas, H3 G34-mutant frequently displayed a misleading radiological presentation, had distinct characteristics compared to established types of WHO grade IV adult gliomas and a poor prognosis.

Radiological Presentation
The radiological characteristics of adult diffuse hemispheric gliomas, H3 G34-mutant in the present series are consistent with those reported in previous studies, LGG/MS although these studies were not dedicated to adults (Table  3). In the present series, adult diffuse hemispheric gliomas, H3 G34-mutant were constantly located within cerebral hemispheres, 3,4,6,8-10,14 frequently affected the parietal and temporal lobes, 4,[6][7][8]11,15,21,22 frequently demonstrated no or faint contrast-enhancement, 7,15 not rarely presented a massive initial hemorrhage 23 and, frequently presented an unusual presentation for a high-grade glioma. 7,15,23 Actually, less than half of cases fulfilled the imaging criteria of highgrade glioma which led in some cases to a potentially detrimental diagnostic delay. Patients for whom the radiological presentation was misleading typically presented with an MRI aspect that has been previously described as "gliomatosis cerebri-like, " consisting of a non-contrast enhancing ill-defined infiltrating lesion that led to suspect another diagnosis than a high-grade glioma because of the young age of the patients. 15 Nontumoral diagnoses, including inflammatory, infectious, vascular and metabolic diseases, could be initially suspected. Retrospectively, a striking feature in these patients that could have raised suspicions toward an aggressive hypercellular tumor 15,24 was the presence of areas of profound focal ADC restriction. In addition to diffusion-weighted-imaging, dynamic 18F-Fluoro-Ethyl-L-Tyrosine positron emission tomography (FET-PET) may be useful to preoperatively identify noncontrast enhancing diffuse hemispheric gliomas, H3 G34mutant as aggressive gliomas. 23

Histo-molecular Presentation
At the histo-molecular level, it was evidenced herein that adult diffuse hemispheric gliomas, H3 G34-mutant could present a PNET-like or a GBM-like morphology, lacked Olig2 expression, were most frequently immunonegative

Neuro-Oncology Advances
for ATRX, immunopositive for TP53, frequently demonstrated a methylated MGMT promoter, and lacked TERT promoter mutations. These findings are consistent with those reported in previous series that included both pediatric and adult cases (Table 3). 3,4,6,7,11,15,21,22,25,26 The absence of Olig2 expression is a hallmark of diffuse hemispheric gliomas, H3 G34-mutant. Yet, as observed in our series, reactive glial cells may wrongly suggest positivity in some cases. As highlighted in children, 27 due to a frequent PNET-like morphology, adult diffuse hemispheric gliomas, H3 G34-mutant could have previously been misdiagnosed as CNS PNET, a now obsolete entity. Actually, in a young adult presenting with an IDHwildtype hemispheric PNET-like tumor, the most likely diagnosis is a diffuse hemispheric glioma, H3 G34-mutant since other entities that can present as PNET-like tumors occur in younger patients. 28 Importantly, in the present series, despite the young age of the patients and a high rate of MGMT promoter methylation, adult diffuse hemispheric gliomas, H3 G34-mutant had a poor prognosis. Despite many similarities, it has been shown that pediatric and adult H3.3 K27M-mutant DMG present significant differences. 29 Compared to pediatric cases, adult H3.3 K27M-mutant DMG more frequently affect the thalamus and less frequently the pons, almost exclusively harbour H3F3A mutations and have a better prognosis. 29 It remains to be determined whether this is also the case in diffuse hemispheric gliomas, H3 G34-mutant, however, these gliomas seem to be a more homogeneous entity than H3.3 K27M-mutant DMG. Age at diagnosis is narrower in H3 G34-mutant gliomas than in H3.3 K27Mmutant DMG and in a previous study neither histological appearance nor molecular characteristics were associated with age. 6 Also consistent with this hypothesis, as detailed in Table 3, the characteristics of adult diffuse hemispheric gliomas, H3 G34-mutant in the series herein seemed very   Table 3. Summary of H3 G34-mutant gliomas characteristics in the present and in previously reported series* Present series Chen C et al. 28 Roux et al. 12 Mackay et al. 22 Vettermann et al. 23 Mackay et al. 21 Puntonet et al. 15 Neumann et al. 11 Korshunov et al. 6 Sturm

Outcomes and Therapeutic Opportunities
The median overall survival of adult diffuse hemispheric gliomas, H3 G34-mutant herein is consistent with the median overall survival ranging from 12 to 36 months reported in previous series (Table 3). 3,4,6,12,21,22,26 Although several studies suggested that diffuse hemispheric gliomas, H3 G34-mutant had a better prognosis than H3.3 K27M-mutant DMG 4,10,30,31 the prognosis of these two tumor types was similar in the present series and was also similar to that of IDH-wildtype glioblastomas. However, the clinical, radiological and histo-molecular characteristics of diffuse hemispheric gliomas, H3 G34-mutant were very different from those of H3.3 K27M-mutant DMG, IDH-mutant, and IDH-wildtype glioblastomas. These findings support that diffuse hemispheric gliomas, H3 G34-mutant should be considered as a new WHO grade IV tumor type in future classifications. 13 Beside gliomas, H3.3 G34 mutations are found in the majority of giant cell tumors of the bone (92%) and some osteosarcomas (5%). 32,33 Although the oncogenesis of diffuse hemispheric gliomas, H3 G34-mutant remains to be fully understood, some therapeutic opportunities have already been identified. Histones pack DNA and are involved in transcriptional regulation and telomeres stability 10,34 through post-translational modifications. 1,2,14,[34][35][36] Mutations of some histone residues, such as the H3.3 K27 and K36 residue, generate conformational changes or steric inhibition that prevent post-translational modifications and trigger deep epigenetic reprogramming. [1][2][3]14,34 In contrast to the H3.3 K27 residue, the H3.3 G34 residue does not undergo post-translational modifications.
However, the H3.3 G34 mutations alter accessibility of H3 K36 to histone methyltransferases affecting H3 K36 methylation and subsequently gene expression. 37 H3.3 G34 mutations have been shown to result in the upregulation of transcription factor MYCN 38,39 and to impair H3 K36 methylation by SETD2 which promotes aberrant PRC2 activity and drives tumor progression. 40 H3.3 G34 mutations also promote genomic instability by blocking the interaction between H3 K36 and the mismatch repair protein MutSα. 41 In addition, it has been shown that diffuse hemispheric gliomas, H3 G34-mutant are characterized by a global DNA hypomethylation and ATRX/DAXX mutations, resulting in alternative lengthening of telomeres (ALT). 3,4,8,25,30 Recently, it has been shown that although considered gliomas, H3 G34-mutant tumors arise in specific interneuron progenitors, in which G34 mutations prevent neuronal differentiation and that PDGFRA signalling plays a major role in these tumors. 26 PDGFRA activating mutations (mostly in the extracellular domain) which are present in 40% of the cases at diagnosis and, in 80% of the cases at recurrence, promote astroglial features and are potent oncogenic drivers. 26 In the future, targeting MYCN, 38 DNA hypomethylation, 4,8 ALT, 3,25,30 and PDGFRA signaling 26 could constitute new treatment strategies in diffuse hemispheric gliomas, H3 G34-mutant. Targeting MYCN upregulation could be achieved by inhibiting kinases such as CHK1 and AURKA which stabilize MYCN 38 or by using BRD3 and BRD4 inhibitors which play essential roles in the transcription of MYCN. 42,43 Regarding PDGFRA signaling inhibition, given the high rate of extracellular mutations, it remains to be determined whether it could be best achieved through PDGFRA inhibitors or through inhibitors targeting downstream MAPK activation. 26 In H3.3 K27M-mutant DMG histone deacetylase inhibitors and the selective antagonist of dopamine receptor DRD2/3 ONC201 are promising therapeutic strategies that are currently being evaluated in clinical trials. 44,45 Whether these strategies may benefit patients with H3 G34-mutant gliomas remains to be determined.

Study Limitations
Other than the limited sample size and its retrospective design, this study is limited by the absence of a comprehensive molecular analysis, the absence of multimodal imaging in all of the patients, and the absence of comparative pediatric series. In addition, the results will need to be validated in an independent and larger series. Nevertheless, the study provides for the first time a picture of adult diffuse hemispheric gliomas, H3 G34-mutant characteristics.

Conclusions
Adult diffuse hemispheric gliomas, H3 G34-mutant are associated with distinct characteristics compared to those of established types of adult WHO grade IV gliomas. This study supports considering these tumors as a new type of WHO grade IV glioma in future classifications. If