Can Conventional MRI Features Predict H3K27M Mutation Status of Diffuse Midline Gliomas?

Purpose Pre-surgical prediction of H3K27M mutation in diffuse midline gliomas (DMG) on MRI is desirable. The purpose of the study is to elaborate conventional MRI (cMRI) of H3K27M-mutant DMGs and identify features that could discriminate them from WT (wild type)-DMGs. Methods cMRI features of 123 patients with DMG were evaluated conforming to the institutional research protocols. Multimodality MRI was performed on 1.5 or 3.0 Tesla MR Scanners with imaging protocol including T1w, T2w, FLAIR, diffusion-weighted, susceptibility-weighted and post- contrast T1w sequences. Pertinent cMRI features were annotated along the lines of Visually AcceSAble Rembrandt Images (VASARI) features and Intra Tumoral Susceptibility Signal score (ITSS) were evaluated. R software was used for statistical analysis. Results Sixty-one DMGs were H3K27M-mutant (mutant DMGs). The patients in the H3K27M-mutant DMG group were younger compared to the WT-DMG group (WT DMGs) (mean age 24.13+13.13 years vs. 35.79+18.74 years) (P= 0.016). The two groups differed on 5 cMRI features– i) enhancement quality (P=0.032), ii) thickness of enhancing margin (P=0.05), iii) proportion of edema (P=0.002), iv) denition of non-contrast enhancing tumor (NCET) margin (P=0.001) and v) cortical invasion (P=0.037). The mutant DMGs showed greater enhancement and greater thickness of enhancing margin while the WT DMGs exhibited signicantly larger edema proportion with poorly dened NCET margins and cortical invasion. ITSS was not signicantly different among the groups. Conclusion cMRI features like enhancement quality, thickness of the enhancing margin, proportion of edema, denition of NCET margin and cortical invasion can discriminate between the H3K27M-mutant and WT DMGs. cerebellum (c A 17-year male with H3K27M- mutant (a, rst DMG shows a well-dened, heterogeneous tumor involving right hemipons and MCP without edema and thick enhancing margin. A 34-year male with H3K27M wild-type DMG (b, second row) shows a poorly-marginated tumor involving left hemipons and MCP with no associated edema and no post contrast enhancement. A 6-year female child with H3K27M- mutant (c, third row) DMG shows a well-marginated, mildly heterogeneous tumor involving left cerebellar hemisphere with no perilesional edema. There is patchy solid intense post contrast enhancement within the tumor. A 23-year male with H3K27M wild-type DMG (d, last row) shows a poorly-marginated tumor involving right cerebellar hemisphere, mild perilesional associated edema and no post contrast enhancement


Introduction
Diffuse midline glioma, H3K27M-mutant, as a distinct set of tumors has been recently categorized in the revised 2016 World Health Organization (WHO) classi cation of the central nervous system (CNS) tumors that exploits an integrated diagnosis combining both histological features and molecular signature. The term assembles diffuse intrinsic pontine gliomas (DIPGs) and in ltrating high-grade glial tumors of the midline carrying the similar canonical mutation at the Lysine 27 residue of the N-terminal tail of histone H3 with its unique recurrent substitution by methionine in the histone H3 variants, H3.3 (approximately 75% cases) and H3.1 (approximately 25% cases). These histone variants are encoded by H3F3A and HIST1H3B/C genes, respectively [1][2][3]. This observation is in contrast to the hemispheric gliomas, where glycine to arginine substitution occurs in the H3F3A gene (H3G34R). The tumors harboring this mutation have been reported to be high-grade that clinically behave aggressively and portends an unfavorable outcome compared to their wild-type (WT) counterparts. Typically, these lesions display short median survival duration of approximately 9 to 11 months from the time of diagnosis regardless of the site of the tumor. They have, therefore, been designated by WHO as grade IV tumors irrespective of their histologic morphology [3][4]. Primary childhood brain neoplasms are rare lesions, with an incidence of ∼2,200 cases per year and DIPG makes up ∼20% of these tumors [5]. Being a rare class of tumors, the exact incidence of DMGs harboring H3K27M mutation is still relatively unknown [6].
As opposed to the previous terminology of DIPG, the term "Diffuse Midline Glioma" (DMG) speci es that these lesions are not solely centered in the pons/ brainstem but may also originate in the other midline structures like thalami, gangliocapsular region, cerebellum, cerebellar peduncles, third ventricle, hypothalamus, and the pineal region as well as in the spinal cord [2,[7][8][9]. Being a rare class of tumors, the exact incidence of DMGs harboring H3K27M mutation (mutant DMGs) remains uncertain. While historically recognized as a pediatric predilection, mutant DMGs can occur across all age groups [6,10]. Diffuse brainstem gliomas are, in general, more aggressive in the pediatric age group than in adults [2]. In children harboring mutant DMGs, the median survival of less than 12, with less than 10 percent 2-year survival rate, while some of the series of adult patients have shown variable median survival rates of up to 20 months [5,6].
The critical anatomical location and the in ltrative nature of the tumor in the thalamus or brainstem limit any meaningful surgical resection, and treatment usually employed is fractionated radiotherapy (RT).
Multiple monotherapy and combination chemotherapy regimens have also been tried with equivalently dismal outcomes [5]. Despite RT and chemotherapy, DMG in children had poor prognosis as a two-year overall survival rate of less than 10% [11]. Since the tumor involves eloquent areas of the brain, a biopsy is usually avoided, and diagnosis of DMGs hinges on clinical and imaging features [7,10]. Imaging modalities like computed tomography (CT) and magnetic resonance imaging (MRI) aid in localization, diagnosis, guiding treatment, and dynamically/ temporally monitoring the disease evolution and response to treatment. Conventional MRI (cMRI) is commonly used for brain tumor evaluation. The Visually AcceSAble Rembrandt Images (VASARI) MRI feature set has been designed to facilitate reliable depictions of gliomas through organized terminology and distinct visual imaging features. It is given by The Cancer Imaging Archive (TCIA) [12]. The use of the VASARI feature set has been adopted earlier for the molecular subtyping of gliomas and to predict IDH mutation status [13,14]. DMGs are frequently diagnosed based on CT or MRI features as part of these patients' initial evaluation. To the best of our knowledge, there are very few studies that has reported the use of conventional MRI features to differentiate the H3K27M mutant DMGs from their wild-type counterparts (WT DMGs). Further, none of the previous reports have utilized an extensive set of conventional imaging features used in the VASARI feature set for the distinction of mutational status in the DMGs. Therefore, we studied the demographic, clinical, pathological, and imaging features of DMG patients retrospectively. We adopted the imaging features similar to those described in VASARI but relevant to midline gliomas for annotating these tumors on MRI. Our study aimed at assessing the cMRI markers to predict the H3K27M mutation status in DMGs non-invasively.

Methods
This was a retrospective observational study performed at a tertiary referral center conforming to the institutional research protocols. We searched our hospital's database retrospectively for the gliomas involving various midline structures of the brain, including the septal region, hypothalamus, thalamus brainstem, cerebellum, middle cerebellar peduncles, and pineal region from June 2016 to April 2020.

Patient cohort
The cases included in the study had, i) MRI suggestive of DMG, ii) who had not received any therapy or had not undergone biopsy or surgical resection before the MRI examination, and iii) for whom the histopathology and immunohistochemistry (IHC) reports were available. All the patients underwent stereotactic biopsy or surgical resection within one month of the MRI. The exclusion criteria were: 1. patients with cerebral hemispheric mass lesions; 2. patients with spinal cord lesions; 3. those subjects where imaging or HPE was suggestive of WHO grade I pilocytic astrocytomas or tumor subtypes other than glioma; 4. inconclusive biopsy; 5. biopsy-proven diffuse midline gliomas with no MRI study; and 6. immunohistochemistry for H3K27M mutation status unavailable. A total of 123 consecutive cases ful lled the inclusion criteria and were nally enrolled in the study, of which 61 patients had mutant DMGs (age (years) = 24.13+13.13), and 62 patients had WT DMGs (age = 35.79+18.74) (in Online Resource 1).
Intratumoral susceptibility signal (ITSS) scoring was also done on the Venobold and SWI images.

Qualitative imaging features
All the MRI scans were reviewed on picture archiving and communication system (PACS) by two independent neuroradiologists. Our analysis adopted the broad range of imaging attributes as laid out in the VASARI dataset for a detailed characterization of the lesions on conventional MRI. The following parameters were evaluated: tumor location, side of lesion center, involvement of eloquent brain, enhancement quality, proportion contrast-enhancing tumor (CET), proportion non-contrast enhancing tumor (NCET), proportion necrosis, cysts, multifocal or multicentric or gliomatosis pattern, T1/FLAIR ratio, thickness of enhancing margin (maximum thickness), de nition of the enhancing margin, de nition of the non-enhancing margin, proportion of edema, hemorrhage, pial invasion, ependymal extension, cortical involvement, deep white matter invasion, NCET crosses midline, CET crosses midline and presence of satellites lesions as per the de nition [12]. Few parameters like overall tumor margin and presence of exophytic component, hydrocephalus, and mass effect were also evaluated other than the VASARI feature set.
ITSS scoring was also performed as speci ed by Park et al. [15]. ITSS was de ned as low signal tubular or dot-like structures with or without conglomeration within the tumor in high-resolution SWI. ITSS was divided into 4 grades: grade 0 (No ITSS); grade 1 (1-5 dot-like or tubular ITSS); grade 2 (6-10 dot-like or tubular ITSS), and grade 3 (> 11 dot-like or tubular ITSS).
Intergroup analysis of various features was done between overall mutant and WT DMGs irrespective of tumor grade and location. Furthermore, subgroup analysis for thalamic, brainstem, and grade IV mutant and WT DMGs was also performed.

Statistical analysis
Data was collated o ine in a Microsoft Excel 2007 spreadsheet in a de-identi ed manner. The analysis was conducted using R software version 3.5.2. Interval scale data were presented as means and standard deviations, and nominal scale data are presented as frequencies and percentages. Between-group analysis of interval scale data was conducted using the non-parametric Mann-Whitney U test. Normality of within-group data was observed qualitatively using histograms, and for conformity of analysis, nonparametric methods were chosen. Between-group analysis of nominal scale data was conducted using a Chi-square test with or without Yate's correction as appropriate. A P-value of < 0.05 was considered statistically signi cant.

Results
Sixty-one out of 123 cases (49.59%) harbored H3K27M mutation, while 62 patients (50.41%) had H3K27M WT status. The mean age of patients in the mutant DMG group was 24.13 + 13.13 years, while in the WT group was 35.79 + 18.74 years (P= 0.016). There were 62 female and 61 male patients, and gender distribution was not signi cantly different between the two groups. Mutant DMGs were graded as WHO grade IV tumors owing to positive H3K27M mutation status. Six (10%) out of 62 WT DMGs were diffuse astrocytomas (grade II), 31 (50%) were anaplastic astrocytomas (grade III), and 25 (40%) were glioblastomas (grade IV).
On intergroup analysis between the mutant and WT DMGs, ve of the studied conventional imaging features evaluated showed signi cant differences. These features were the i. enhancement quality (P=0.032); ii. the thickness of the enhancing margin (P=0.05); iii. proportion of edema (P=0.002); iv.
de nition of NCET margin (P=0.001) and v. cortical invasion (P=0.037). The mutant DMGs showed signi cantly greater enhancement as well as the greater thickness of enhancing margin or solid enhancement. The WT DMGs exhibited a signi cantly larger proportion of edema with a poorly de ned NCET margin and a larger degree of cortical invasion compared with the mutant DMGs (in Fig. 1-3). Few other parameters like tumor heterogeneity (P=0.07), overall tumor margin (P=0.074), and de nition of CET margin (P=0.073) tend to approach the level of signi cance with a higher trend of heterogeneity in the mutant group, and WT tumors were showing more ill-de ned margins (Table 2).

Discussion
As per the decision of Working Committee 3 of the cIMPACT-NOW, the term DMG, H3K27M-mutant should be earmarked for the tumors that are diffuse, i.e., in ltrating, midline (e.g., involving the thalamus, brain stem, spinal cord, etc.), gliomas (astrocytic lineage) and H3K27M mutant, and must not be used for other tumors (as ependymoma or pilocytic astrocytoma) that possess H3K27M mutation [16]. H3K27M mutation is a gain-of-function mutation that modi es gene expression by posttranslational alterations in histone three, leading to altered DNA methylation and gliomagenesis via epigenetic regulation [4]. Genomic analysis of mutant DMGs has shown many correlated genetic variations, including mutations in the receptor tyrosine kinase/RAS/PI3K pathway, p53 overexpression, and ATRX (alphathalassemia/mental retardation syndrome X-linked) loss [1,9].
In this study, approximately half of the midline located gliomas harbored the H3K27M mutation and were designated as DMG, H3K27M-mutant. Such incidence of the H3K27M mutation in midline gliomas has also been previously reported [10,17,18]. Mutant DMGs occur in the younger population as compared to WT DMGs [10]. In our study, the patients with mutated DMGs were approximately 12 years younger than WT DMG patients (P= 0.016). We did not nd any signi cant difference in the gender distribution between the mutant and WT groups. The same has been observed earlier in the studies by Chen et al. and Schreck et al. [6,10].
We found that most of the DMGs (79%) were centered in the thalamus (with no signi cant difference between the groups). However, the difference was signi cant when brainstem location was considered, and a greater number of mutant DMGs were distributed in the brainstem than WT-DMGs. A similar nding has been reported by Schreck et al. [6].
Our results revealed that certain cMRI features could be used to discriminate the H3K27M mutational status of DMGs. On intergroup and subgroup (brainstem and grade IV tumors) analyses, we found that the mutant gliomas demonstrated greater enhancement with thicker enhancing margins or solid enhancement as compared to the WT group. In comparison, the WT tumors exhibited more edema, poorly de ned NCET margins, and more cortical invasion compared with the mutant DMGs. In the radiological evaluation results in the HERBY Phase II trial of non-brainstem pediatric HGGs, the mutated DMGs have been reported to enhance avidly compared to WT DMGs [19]. The H3K27M mutation escalates the expression of growth factors like PDGF and VEGF by recurrent mutations and ampli cation of signaling genes like ACVR1, PI3K, and RTKs. The PIK3R1 mutations are reported to be obligatory to H3.3K27M mutations, and PIK3CA mutations are found in all DIPGs. These PIK3 mutations are the promoters of angiogenesis [20][21] that might account for the more orid enhancement of the mutant gliomas that we appreciated in this study.
Histopathology literature is found wanting as to why mutant DMGs demonstrated a lesser degree of edema compared with the WT tumors. However, the same nding has also been discussed by Qui et al. in their study of 66 patients with DMGs, where only 10 cases showed peritumoral edema [22]. Castel et al. have reported that the more aggressive H3.3mutant DMGs exhibited lesser extracellular edema as compared to the less aggressive H3.1 mutant DMGs [23]. On the other hand, some authors found no signi cantly different cMRI parameters (such as tumor location, margin, cysts, necrosis, hemorrhage, degree, and pattern of contrast enhancement and edema) that could distinguish the mutant from WT DMGs [2,6,10,23,24].
When we compared the cMRI features between thalamic mutant and WT tumors, mutant tumors showed more exophytic components. WT tumors showed signi cantly more edema; however, there was no difference in the enhancement and the de nition of tumor margin between the two groups. The presence of exophytic growth pattern (tumor growing outside the borders of the structure of origin) confers a superior prognosis in brainstem gliomas than patients with tumors showing intrinsic growth patterns [25][26][27]. However, the same has not been described for the thalamic gliomas. This could indicate that the in uence of this mutation in different anatomical locations may be variable, and these tumor subsets can be studied in a more comprehensive manner.
The limitations of this study include the lack of follow up and no information regarding the survival data of the patients. Also, we did not perform any measurements to quantitate the degree of enhancement, which could increase the reproducibility of the results. Subgrouping of the histone of the H3.3/H3.1 gene group that could have been interesting was not performed in our study. Future studies with more extensive clinical data are needed to recognize various other aspects of H3K27M mutant DMGs, like outcome and management strategies, which are not discussed in this paper.

Conclusion
Conventional MRI features such as enhancement quality, the thickness of the enhancing margin, proportion of edema, de nition of non-contrast enhancing tumor margin, and cortical invasion can help discriminate between the H3K27M-mutant and WT DMGs. Among the varied imaging phenotypic characteristics of DMGs, these ndings carry an important implication for treatment planning as well as designing future trials in this distinct group of neoplasms.