Clinical Features and Molecular Markers on Diffuse Midline Gliomas With H3K27M Mutations: A 43 Cases Retrospective Cohort Study

Purpose Diffuse midline gliomas (DMG) with H3K27M mutations have been identified as a rare distinctive entity with unique genetic features, varied molecular alterations, and poor prognosis. The current study aimed to evaluate the clinical characteristics and profile of molecular markers on patients with a DMG harboring H3K27M mutations, and explore the impact of this genetic makeup on overall survival. Methods We retrospectively analyzed 43 consecutive patients diagnosed with a DMG harboring H3K27M mutations (age range 3 to 75 years) and treated in a tertiary institution within China between January 2017 to December 2019. Various clinical and molecular factors were evaluated to assess their prognostic value in this unique patient cohort. Results The median overall survival (OS) was 12.83 months. Preoperative Karnofsky Performance Score (KPS) and adjuvant radiotherapy were found to be independent clinical parameters influencing the OS by multivariate analysis (p = 0.027 and p < 0.001 respectively). Whereas extent of tumor resection failed to demonstrate statistical significance. For molecular markers, P53 overexpression was identified as a negative prognostic factor for overall survival by multivariate analysis (p = 0.030). Conclusion Low preoperative KPS, absence of radiotherapy and P53 overexpression were identified as predictors of a dismal overall survival in patients with DMG and H3K27M mutations.


INTRODUCTION
Diffuse midline glioma (DMG) with histone H3K27M mutation is a novel entity that was added to the 2016 World Health Organization (WHO) classification of tumors of the Central Nervous System (CNS), on the principle of integrated diagnosis with combinations of characteristics from both histological and molecular aspects (1,2). This distinct entity is an aggressive tumor that corresponds to grade IV glioma, regardless of their histological grade, and is described as an infiltrative glial neoplasm with predominately astrocytic differentiation and a K27M mutation in the histone H3 gene (either H3F3A or HIST1H3B/C) (1,3).
In accordance to the diagnostic criteria outlined by the WHO, DMG with H3K27M mutations have a diffuse growth pattern and are restricted to the midline structures, such as the thalamus, brainstem, cerebellum and spinal cord (4). These tumors are most frequently observed in the pediatric population, but adults are also commonly affected (5). As an aggressive malignant tumor, prognosis is considerably poor, and the 2-year survival rate is less than 10% for patients receiving standard treatment including surgery and adjuvant chemoradiation (6).
Given the rarity and novelty of DMG with H3K27M mutation in adults (7), several studies have sought to describe the incidence and clinical behavior of these tumors (8,9). However, not all reported cases adequately fulfill the diagnostic criteria for DMG with H3K27M mutation. For example, nonmidline gliomas with H3K27M mutation or circumscribed H3K27M-mutant cases have also been included in prior studies (6,10). Owing to the unique deep-seated location of DMG, surgical treatment is often challenging and might lead to suboptimal resection or even failure to obtain an adequate biopsy specimen. As such, limited surgical outcomes data has been reported and further investigation of relevant molecular markers and associated genetic alterations is still needed.
In the current study, we retrospectively explored the clinical and molecular characteristics of 43 patients who were diagnosed with a DMG with H3K27M mutation between January 2017 and December 2019 at our institution. Our findings highlight some of the factors that influence patient survival and the potential cooperative genetic alterations for patients with DMG with H3K27M mutations.

Patient cohort and Identification
This retrospective study was reviewed and approved by our Institutional Review Board (IRB). The relevant demographic and clinical data on all glioma cases were obtained and identified from the institutional electronic medical record (EMR) system. For each case, the diagnosis and pathological classification were confirmed according to the 2016 WHO classification of CNS tumors. Consecutive cases of diffuse midline gliomas were selected from January 2017 to December 2019, as H3K27M immunohistochemistry (IHC) began to be regularly performed on glioma patients in our hospital starting in January 2017. Pre-operative Magnetic resonance imaging (MRI) was reviewed by two neurosurgeons (P-G Ji and J-H Liu) to determine the location of the gliomas. The primary tumor location was reported and categorized as thalamus, midbrain, pons, medulla oblongata, basal ganglia, spinal cord or cerebellum. Patients were followed under the standard treatment protocol for all CNS tumor patients at our institution and the clinical status of all patients at last followup was recorded.
For evaluation of the extent of resection, gross total resection (GTR) was defined as no residual tumor remnant (100%) seen with postoperative MRI imaging. Subtotal resection (STR) was defined as less than 90% extent of resection, while the partial resection (PR) was defined as less than 50% resection of the tumor. Postoperative computed tomography (CT) and MRI were applied within 48 h postoperatively.
In total, 43 cases of diffuse midline gliomas with H3K27M mutation were included for analysis.

Immunohistochemistry and Pathology Review
Tissue blocks embedded with formalin-fixed paraffin (FFPE) and hematoxylin and eosin (H&E)-stained slides were retrieved from the Department of Pathology. Five to eight unstained slides with 4mm-thick sections were prepared for further IHC evaluation for each case. IHC staining was performed for the differential diagnosis. IHC analysis included H3K27M, trimethylation of Lys27 (H3K27Me3), O-6-methylguanine DNA methyltransferase (MGMT), IDH-1, ATRX, p53, fibrillary acidic protein, Oligo2, nestin, NeuN, Syn, S-100, and Ki-67. IHC was interpreted by a neuro-pathologist and stains were registered as positive, negative or uninterpretable.

DNA Extraction and Mutation Detection
Sanger sequencing (SINOMD, Beijing, China) was used to determine the frequency of mutations in IDH1 and IDH2. Sanger sequencing were performed using an ABI-3130 DNA Analyzer (Applied Biosystems).
Methylation-specific quantitative PCR (qMSP) was applied to detect the status of MGMT promoter methylation. Genomic DNA was extracted from paraffin section and converted by bisulfite.

Statistics
Continuous variables were reported by mean and standard deviation or median and interquartile range (IQR) and analyzed using the unpaired Student t test or Wilcoxon Sum Rank test, while H3K27M mutations and related categorical variables were summarized using Pearson Chi-square or Fischer's exact test. Kaplan-Meier curves and log-rank test were used to compare survival between patients with or without changes. In this study, GraphPad Prism (Version 7.10, La Jolla, California, USA) and SPSS statistics (version 22.0, IBM corp., Armonk, New York, USA) were applied for the statistical analysis.

RESULTS
In our present study, 917 cases of histologically confirmed glioma (January 2017 to December 2019) were retrieved from the Tangdu Hospital Medical Record System, including circumscribed and diffused gliomas. By retrieving the data, the incidence of pediatric (<21 years old) glioma is 131/917 (14.3%), and incidence of adult (≥21 years old) glioma is 786/917 (85.7%). Of these, 43 cases were diagnosed as DMG with H3K27M mutation (n = 43/917, 4.69%). According to the diagnostic criteria of diffuse midline gliomas with H3K27M mutant, WHO IV, all 43 cases were retrospectively evaluated.

Demographic Features of DMG Patients With H3K27M Mutations
Of the 43 cases confirmed as DMG with H3K27M mutant, WHO IV, 26 (60.5%) were males and 17 (39.5%) were females. At the time of diagnosis, the patient's age ranged from 3 to 75 years, with the median age of 38 (IQR: 17-50). In our cohort, most of the patients with DMG were adults (aged≥21 years old, 30/43, 69.77%), with 13 (30.23%) children (<21 years old) included in the present study.
The baseline demographic and clinical features of these cases are summarized in Tables 1 and 2.

Surgical Treatment and Follow-Up
PR and STR was performed in the majority of patients given the high risk of iatrogenic morbidity associated with GTR of thalamic and other midline lesions. For lesions located in the posterior fossa, including the brainstem and spine, surgical debulking or biopsy was considered as the primary surgical option. Ultimately, GTR was achieved in 8 patients (18.60%), STR was performed in 12 patients (27.91%), and PR was in 19 cases (44.19%), while 4 patients received stereotactic biopsy (9.30%). Representative images of extent of tumor resection are listed in Supplementary Figure 1.
All patients were followed postoperatively according to the standard protocol at our institution. Clinical status was recorded at last follow-up. Follow-up duration or time to death were recorded for all patients. In our cohort, there was no loss of follow-up and the median follow-up duration was 8.03 (0.03-41.87) months. Adjuvant chemoradiation was advised for all patients. However, only 20 patients (46.51%) received prescribed adjuvant both RT and temozolomide (TMZ)-based chemotherapy. Four patients (9.30%) completed RT only, and three patients (6.98%) received TMZ only. Two patients (4.65%) received other treatments, including traditional Chinese medicine and herbal therapy. Unfortunately, there were 14 patients (32.56%) who did not receive any additional treatment after surgery. At last follow-up, 20 (46.51%) patients were still alive, while 23 (53.49%) patients were deceased.

IHC Features and Molecular Characteristics
Pathologic assessment including IHC was performed per institutional protocol for all tumor specimens. According to the 2016 WHO classification of CNS tumors, typical morphological features of glioma were presented in all cases. However, H3K27me3 and MGMT status was not available for every patient in this cohort with IHC.
In accordance with the 2016 WHO classification, all patients in our cohort presented with a H3K27M mutation, while the status of H3K27me3 expression was checked in 19 cases, in which 13 (68.42%) cases were found to have loss of expression, implying a decrease of H3K27Me3. p53 was overexpressed in 23 of 43 cases (53.49%), and loss of ATRX (loss of expression) was found in 38 of 43 cases (88.37%) ( Figure 1, Table 3, and Supplementary Table 1).
No specimen demonstrated a IDH1 or IDH2 mutation according to Sanger sequencing. MGMT expression was negative in 18 of 35 cases (51.43%) by IHC. 10 of 43 cases confirmed MGMT promoter methylation by PCR (23.26%).

Clinical and Molecular Associations With Survival Outcome
The impact of various clinical factors such as age, sex, molecular parameters on overall survival (OS) was analyzed and presented using the log-rank test and Cox regression analysis ( Table 3). Of note, some patients could not complete their serial postoperative MRIs due to poor baseline and postoperative KPS, resulting in censored data for the assessment of progression-free survival (PFS). Therefore, we focused on OS and the relationship between the clinical and molecular factors.
The median OS was 12.83 months for the whole cohort, and the 12-, 24-month OS rates were 53.8% and 40.2%, respectively. According to both univariate and multivariate analyses, both better preoperative KPS (≥70) (p = 0.017, 95% CI: 0.11-10.32, HR = 0.29) and treatment with adjuvant radiotherapy (p < 0.001, 95% CI: 0.02-0.23, HR = 0.07) were significantly associated with a better OS prognosis. Age, sex, location of the tumor and the extent of surgical resection failed to show statistical significance upon univariate analysis.
On assessment of the patients' molecular markers, status of MGMT promoter methylation and H3K27me3 status did not influence prognosis in our cohort. P53 overexpression demonstrated a worse prognosis by univariate analysis (p = 0.002). Using a multivariate model, P53 retention remained as a significantly favorable predictor of longer OS (p = 0.008, 95% CI: 0.08-0.68, HR = 0.23). Loss of ATRX failed to demonstrate a significant impact on OS (p = 0.17, 95% CI: 5.89-18.51) by univariate analysis. However, it was determined as an independent factor associated with the poor prognosis by multivariate analysis (p = 0.021, 95% CI: 0.01-0.69, HR = 0.09).
The Kaplan-Meier survival analysis of OS, stratified by KPS, adjuvant radiotherapy, ATRX, and P53, respectively, are shown in Figure 2. Multivariate analyses are presented in Figure 3. In particular, we noticed the cooperative alterations for p53 and ATRX. Both ATRX loss and p53 overexpression was observed in 20 cases (46.51%) in the cohort, which also implied a worse survival according to Kaplan-Meier analysis (Figure 4).
Adult and pediatric cases were also stratified and analyzed (Supplementary Figure 2 and Supplementary Table 2). Based on our current data, there was no significant difference in OS between adult and pediatric groups (p = 0.282). Representative IHC images are shown in Supplementary  Figure 3.

DISCUSSION
As the knowledge of molecular markers in the context of malignancy increases, our understanding of gliomas has dramatically increased. Molecular pathology has become of utmost importance in the diagnosis and treatment of various CNS tumor subtypes. Diffuse midline glioma with H3K27M mutation is a novel entity defined by the 2016 WHO classification system (1,4). Evidence has suggested that tumors with a H3K27M mutation are associated with dismal survival compared with H3K27M wildtype tumors (2, 11). As a distinct entity of glioma, DMG with H3K27M mutation typically occurs in children and corresponds to a grade IV lesion (6,9). To date, analysis of DMG with H3K27M mutation has been mainly limited to the pediatric population. Our present study uniquely explores the clinical and molecular characteristics of 43, mostly adult, cases of diffuse midline glioma with H3K27M mutations treated as a large, tertiary medical center in China. Based on our results, a higher preoperative KPS, use of adjuvant radiotherapy, and P53 retention are found to be individual positive prognostic factors for better overall survival in this unique patient cohort.

The Clinical Features of DMG With H3K27M Mutation
Since the revision of WHO 2016 classification, H3K27M status was introduced as a routine test at our institution in January 2017. According to our EMR, the number of diagnosed cases of DMG with H3K27M mutation only accounts for 4.69% for glioma patients. The median age of our cohort is 38 years old, which is in line with previous reports indicating a more frequent occurrence in children and young adults (2). Although males make up the majority of cases, sex is not associated with prognosis. The lesion most commonly invaded thalamus and midbrain in our cohort, which is consistent with previous reports of a thalamic predominance of DMGs in adults (9). Likewise, thalamic-related symptoms are most commonly observed, which are the manifestations of damage to the vital structures, and may result to motor function deficits (12). Preoperative KPS is closely related to tumor location and severity of the patients' baseline clinical function. In line with prior studies on glioma, our results suggest that a poor pre-op KPS is significantly associated with poor prognosis.
Extent of resection is typically a strong predictor of survival prognosis for most gliomas, especially the high grade gliomas, such as glioblastoma (12,13). However, since DMGs are deepseated lesions that are located in eloquent CNS regions, the ability to achieve a safe maximal resection is substantially limited. As a result, the historical diagnosis of DMG has been traditionally made by radiographic assessment alone. With the advancement of intraoperative technologies, safe, maximal resection and/or biopsy is now possible for this unique and challenging patient population. In our study, PR and STR are preferred for most lesions, with GTR limited to specific lesions. For example, brainstem gliomas are generally treated with biopsy or surgical debulking. This surgical strategy was adopted according to the high potential risk of surgical morbidity and the dismal clinical behavior of tumors located within the midline structures that carry a H3K27M mutation (14,15). Evidence has indicated that a greater extent of tumor resection did not provide a prognostic benefit for DMG with H3K27M mutation (2). Our current research does not find strong relevance between the extent of resection and OS for DMG cases as well. Thus, we do not consider that a more aggressive surgical intervention should be attempted for these cases. However, given the limited number of patients in this study, future studies are needed with a larger patient cohort in order to substantiate or dispute this experience.
Although adjuvant chemoradiation is typically advised after surgery, only 24 patients (RT only and RT+TMZ) in our present series received RT. Given the limited treatment options for DMG, our limited data suggests that patients may have a survival benefit with adjuvant RT (Figure 2). Various factors might impact the choices for the adjuvant therapy, such as socioeconomic background, education and personal expectation (15). In our cohort, we find that low postoperative *Since the tumor invades multiple brain regions, the sum of the tumor location is greater than the total of case number. SD, standard deviation; GTR, gross total resection; STR, subtotal resection; PR, partial resection; PFS, progression-free survival; OS, overall survival; RT, radiotherapy; TMZ, temozolomide; KPS, Karnofsky performance status. No. 16 and No. 27. After surgical intervention and formal pathological diagnosis, family members tend to have much lower psychological expectation, which may lead them to be less active in following treatments. This factor may also lead to

Molecular Features of DMG With H3K27M Mutation
Histone gene mutations in H3F3A or HIST1H3B/C, encoding for histone H3.3 and H3.1 respectively, are considered an overall dismal prognostic indicator for patients with a DMG (3,10). TP53 is a pivotal tumor suppressor gene and TP53 mutation has a significant overlap with H3K27M mutations according to previous studies (16) and is in line with our study. For diffuse intrinsic pontine gliomas (DIPG) with H3K27M mutation, about 60-80% of TP53 mutations have been associated with an aggressive course (5). Similar results were seen in our cohort andp53 overexpression was found to be associated with worse survival prognosis for both pediatric and adult patients (Supplementary Table 2 and Supplementary Figure 2). For pediatric patients, the mutation and loss of ATRX expression occur in about 1/3 of specimens and is even more common in DIPG (15). Our series found an exceptionally higher (38/43, 88.37%) incidence of ATRX loss in both pediatric (10/13) and adult (28/30) patients, which is prevalent in the thalamus and brainstem. ATRX loss was found to be an independent predictor of poor prognosis in our series when included in the multivariate model. Based on this finding, future larger studies are warranted.
Interestingly, evidence suggests that there are cooperative genetic alterations for DMG with H3K27M mutation, particularly in relation to TP53 and ATRX mutations without IDH-1 mutation (17). ATRX loss and p53 overexpression in our cohort was associated with a dismal survival expectation ( Figure  4). However, these results should be interpreted with extreme caution given the limited number of cases included in the present study.
H3K27me3 functions as transcriptional repressor and is closely related to posttranslational modifications (18). Previous researches have shown that reduced H3K27me3 levels lead to extensive transcriptional modification of tumor cell regulation, and has been associated with an unfavorable prognosis in several cancers (19). The expression of H3K27me3 (13/19, 68.42%) is  decreased in most H3K27M mutant cases. Although not statistically significant in our cohort, evidence suggests that the cooperative alteration of K27M-mediated loss of H3K27me3 is context-independent (20).
MGMT promoter methylation is comparatively rare for DMG patients (21). In our present study, the IHC results suggest a high rate of MGMT negative expression (18/35 cases). Meanwhile, qMSP confirms that MGMT methylation is not rare (10/43) in our cohort. In addition, we noticed a discrepancy regarding the MGMT status between IHC and qMSP. Evidence has suggested that the predictive power of IHC for OS appears weaker than other methylation assays (22) and the clinical utility of MGMT IHC is still controversial. Thereby, we adopted qMSP results for the analysis.
DMG with H3K27M mutation is still a fatal disease with dismal prognosis. Recently epigenetic studies have greatly broadened our understanding of its biological behavior and underlying mechanism. CSF analysis has also shown promise for the diagnosis and monitoring treatment response (23). In order to better understand and optimize survival in this unique and challenging patient population, further study is needed.

Study Limitations
Several limitations inherent to the retrospective nature of our present study must be carefully considered. First, as a rare entity in glioma, the incidence of DMG with H3K27M mutation is substantially low. The small number of patients found for our cohort implies that the statistical power is limited in analyzing the relevant clinical and molecular factors. Second, IHC staining was utilized for most analyses in our study, and sequencing was only partially applied. Therefore, we could not further evaluate more detailed histone mutations, such as H3.1 and H3.3 mutations, which limited us to explore other genetic biomarkers and potential therapeutic effects.

CONCLUSIONS
Despite these limitations, we retrospectively summarized the clinical and molecular features of 43 patients diagnosed with a diffuse midline glioma with H3K27M mutation over a three-year period at our institution. DMG with H3K27M mutation is associated with poor survival, with a median OS of 12.83 months in our study. Poor preoperative KPS, absence of adjuvant radiotherapy, and P53 overexpression were found to be independent prognostic factors related to worse overall survival. In addition, cooperative genetic alterations, such as loss of ATRX and p53 overexpression without IDH-1 mutation, might help us to better understand the pathogenesis and appropriate treatment strategies for gliomas with H3K27M mutation.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.