Genomic profiling and prognostic factors of H3 K27M‐mutant spinal cord diffuse glioma

Abstract

H3 K27‐altered diffuse midline glioma is a highly lethal pediatric‐type tumor without efficacious treatments. Recent findings have highlighted the heterogeneity among diffuse midline gliomas with different locations and ages. Compared to tumors located in the brain stem and thalamus, the molecular and clinicopathological features of H3 K27‐altered spinal cord glioma are still largely elusive, thus hindering the accurate management of patients. Here we aimed to characterize the clinicopathological and molecular features of H3 K27M‐mutant spinal cord glioma in 77 consecutive cases. We found that the H3 K27M‐mutant spinal cord glioma, with a median age of 35 years old, had a significantly better prognosis than H3 K27M‐mutant brain tumors. We noticed a molecular heterogeneity of H3 K27M‐mutant spinal cord astrocytoma via targeted sequencing with 34 cases. TP53 mutation which occurred in 58.8% of cases is mutually exclusive with PPM1D (26%) and NF1 (44%) mutations. The TP53‐mutant cases had a significantly higher number of copy number variants (CNV) and a marginally higher proportion of pediatric patients (age at diagnosis <18 years old, p = 0.056). Cox regression and Kaplan–Meier curve analysis showed that the higher number of CNV events (≥3), chromosome (Chr) 9p deletion, Chr 10p deletion, ATRX mutation, CDK6 amplification, and retinoblastoma protein (RB) pathway alteration are associated with worse survival. Cox regression analysis with clinicopathological features showed that glioblastoma histological type and a high Ki‐67 index (>10%) are associated with a worse prognosis. Interestingly, the histological type, an independent prognostic factor in multivariate Cox regression, can also stratify molecular features of H3 K27M‐mutant spinal cord glioma, including the RB pathway, KRAS/PI3K pathway, and chromosome arms CNV. In conclusion, although all H3 K27M‐mutant spinal cord diffuse glioma were diagnosed as WHO Grade 4, the histological type, molecular features representing chromatin instability, and molecular alterations associated with accelerated cell proliferative activity should not be ignored in clinical management.

In a large cohort (n=77), we characterized the molecular, clinicopathological, and prognostic characteristics of H3 K27M‐mutant spinal cord diffuse glioma. ATRX mutation, CDK6 amplification, RB pathway alteration, a higher number of Copy number variant (CNV) events, chromosome (Chr) 9p deletion, and Chr 10p deletion is associated with worse survival. Glioblastoma histological type and a high Ki‐67 index (>10%) are associated with a worse prognosis. Interestingly, the histological type, an independent prognostic factor in multivariate Cox regression, can stratify molecular features of H3 K27M‐mutant spinal cord glioma, including the RB pathway and PI3K pathway.

1. INTRODUCTION

Glioma, which arises from the brain and spinal cord, is the most common primary malignant tumor of the central nervous system (CNS) [1]. The accurate diagnosis and classification of gliomas are essential for improving patient clinical outcomes [2, 3]. The WHO 2021 Classification of CNS tumors has integrated molecular data into the typing, subtyping, and grading of gliomas, thus improving the ability of clinicians to select the most appropriate therapies and design clinical trials [1, 4, 5]. However, these advances are mainly obtained from studies of brain gliomas [6, 7]. The molecular features of diffuse spinal cord glioma are still largely elusive, dramatically limiting the accurate diagnosis and treatment of this deadly and highly disabling tumor.

We and others have reported differences in the clinical and molecular characteristics between the spinal cord and brain diffuse gliomas [8, 9, 10, 11, 12, 13]. The most characterized molecular feature of spinal cord diffuse glioma is H3 K27M mutation, which was identified in 40%–60% of tumors, a rate much lower than that observed in brain stem diffuse glioma (approximately 80%) [6, 9, 14, 15, 16, 17, 18]. Most patients with H3 K27M‐mutant spinal cord diffuse glioma are adults, and their prognosis is better than that of patients with H3 K27M‐mutant brain stem tumors, mainly pediatric patients [6, 9, 17, 19]. Recently, distinct cellular origin and prognosis characteristics have been revealed in the H3 K27M‐mutant midline gliomas with different anatomic localization and ages [14, 20, 21, 22, 23, 24]. These findings gradually challenged the current situation that the clinical management of spinal cord diffuse glioma is mainly referred to guidelines of their brain counterparts [25, 26, 27]. However, because of the low incidence and high risk of surgical resection of spinal cord diffuse glioma, controversial results have been reported in different spinal cord astrocytoma studies with small sample sizes [16, 28, 29, 30, 31], which cannot provide enough evidence for improving clinical management of spinal glioma. Thus, it is urgent to study the molecular and clinicopathological characteristics of spinal cord diffuse glioma in a large cohort to refine their diagnosis and clinical management.

Here, we collected a large cohort of consecutive cases with H3 K27M‐mutant spinal cord diffuse glioma (n = 77) with basic clinicopathological information and survival follow‐up information. Based on this cohort, we characterized the molecular and clinicopathology features of H3 K27M‐mutant spinal cord diffuse glioma. Finally, we highlighted the molecular heterogeneity and prognostic factors of these tumors.

2. MATERIALS AND METHODS

2.1. Patients and samples

We enrolled 77 consecutive patients with spinal astrocytomas at the Department of Neurosurgery, Beijing Tiantan Hospital, China, from March 2011 to March 2022. All studies performed were approved by the Institutional Review Board of Beijing Tiantan Hospital (IRB KY2013‐017‐01). The characteristics of the cases are summarized in Table S1.

2.2. Treatment of patients with spinal cord astrocytic tumors

As previously reported [12, 13], the extent of resection was estimated by evaluating postsurgery magnetic resonance images and classified as gross total resection (≥90%), subtotal resection (≥50% and <90%), or open biopsy (<50%). Postoperative radiation therapy was administered at a total dose of 40–50 Gy. Temozolomide was administered as chemotherapy at a daily dose of 75 mg/m² during radiotherapy or 5 consecutive days per treatment cycle with 4 weeks per cycle at a dose of 150–200 mg/m² for at least three cycles.

2.3. Determination of histological grade

The histopathological type and grade were determined by routine evaluation of formalin‐fixed paraffin‐embedded (FFPE) slices stained with hematoxylin and eosin (H&E) by at least two experienced neuropathologists as previously reported [9]. Cases with tumor cells with cytological atypia alone were considered diffuse astrocytoma (A); cases that also showed features of anaplasia and mitotic activity were regarded as anaplastic astrocytoma (AA); and cases that additionally had microvascular proliferation and/or necrosis were diagnosed as glioblastoma (GBM).

2.4. Determination of molecular features

The targeted sequencing results determined the molecular features for cases with targeted‐sequencing data. For other cases, the status of IDH1 R132H mutation, TERT promoter mutation, and BRAF V600E mutation was determined by pyrosequencing. Immunohistochemistry was also used to detect the H3 K27M‐mutant protein (1:800, ABE419; Millipore, Billerica, MA, USA), H3 K27me3 status (1:200, C36B11; CST, Boston, MA, USA), and Ki‐67 expression (1:50, MIB‐1; Labvision, Fremont, CA, USA). MGMT  (O6‐methylguanine–DNA methyltransferase) promoter methylation was assessed by pyrosequencing, as previously described [32].

2.5. Targeted sequencing and data processing

Extracted DNA from FFPE samples was used to prepare the sequencing library via capture probes and KAPA kits (Roche Diagnostics GmbH, Mannheim, Germany). The purified library was sequenced using the NovaSeq 6000 platform (Illumina, Santiago, CA, USA). The sequencing results were aligned to the human reference genome hg38 by Sentieon. The somatic mutations were annotated using SnpEff, and copy number variations (CNVs) were identified with CNVkit. The target genes and chromosomes in the designed panel are summarized in Table S2. The landscape of mutations and CNVs was drawn by Complexheatmap package Version 2.12.

2.6. DNA methylation profiling

The genomic DNA extracted from FFPE samples was bisulfite converted using the EZ DNA Methylation kit (Zymo Research, Irvine, CA, USA) following the manufacturer's recommended protocol. Bisulfite‐converted DNA was amplified, fragmented, and hybridized to Infinium EPIC 850 k Human DNA Methylation BeadChips (Illumina), following the manufacturer's recommended protocol. The methylation data were preprocessed using the ChAMP package (v.1.28.4) in R Bioconductor (version 3.5.3). Quality control and normalization were performed according to the standard pipeline of ChAMP. The reference methylation data (450k) were obtained from GEO (GSE109381) [33]. The t‐distributed stochastic neighbor embedding (t‐SNE) analysis was performed using Rtsne package version 0.16, and the 428,230 probes common between the 450k and EPIC arrays were used for analysis. The input for the t‐SNE calculation is Pearson correlation, weighted by variance, and the clustering analyses were performed using the beta values of the top 32,000 most variably methylated probes by standard deviation.

2.7. Statistical analysis

Statistical analysis was performed using GraphPad Prism 7 (GraphPad Software, La Jolla, CA, USA) or Rv4.0.5 (https://www.r-project.org/). A nonparametric test was used to compare the age distribution between the two subgroups of gliomas, and Fisher tests were used to compare the distribution of other clinicopathological features. The Kaplan–Meier method with a two‐sided log‐rank test was used to compare patients' overall survival (OS) in different subgroups stratified by various clinicopathological features. Univariate and multivariate Cox regression analyses were performed to determine survival‐associated factors.

3. RESULTS

3.1. Characteristics of spinal cord astrocytomas

Here, we collected 77 cases to study the characteristics of H3 K27M‐mutant spinal cord diffuse glioma (Figure 1A and Table S3). Among these cases, 34 were profiled by targeted sequencing, 9 were evaluated by DNA methylation assay, and 74 had follow‐up information. Data on the Ki‐67 index, IDH mutation, TERT promoter, and BRAF V600E status were also obtained for most cases. As shown in Figure 1B, tumor to thoracic cord location was the most common (34%). The patients' median age is 35 years, with 76.6% in adults and 55% in males. Compared with previously published survival data of brain stem and thalamus H3 K27M‐mutant tumors [7], patients with spinal cord H3 K27M‐mutant astrocytoma had a significantly longer OS (p < 0.0001, log‐rank test) (Figure 1C).

FIGURE 1.

Clinical, molecular, and survival characteristics of cases included in this study. (A) Summary of H3 status, survival, target‐sequencing, methylation assay, and other selected molecular information of all cases in this study. (B) Distributions of location, age at diagnosis, and gender of spinal cord diffuse glioma. Cervical cord (C1–C3) and Thoracic cord (T2–T11). (C) Kaplan–Meier survival curves of patients with H3 K27M‐mutant spinal cord diffuse glioma, compared with the previous study in H3 K27M‐mutant brain midline diffuse glioma. (D) The t‐distributed stochastic neighbor embedding (t‐SNE) plot of methylation features of diffuse glioma with H3 K27M mutation. The case ID and potential subgroups of spinal cord astrocytic used in this study are labeled. The full name of the abbreviation is summarized in Table S7.

DNA methylation profiles have been used to classify CNS tumors [33, 34]. Here, we observed that our cohort's nine tumors with methylation data had a similar DNA methylation profile to previously reported H3 K27M‐mutant midline gliomas via unsupervised clustering (Figure 1D). We also verified our cases by the DKFZ MNP predictive classifier [33]. The results showed that seven of nine cases matched “the methylation class diffuse midline glioma H3 K27M mutant” with scores ≥0.98, and one of the rest two cases also was predicted as “the methylation class diffuse midline glioma H3 K27M mutant” with scores = 0.86 (Figure S1A). Immunohistochemistry and targeted sequencing also verify that the last case was an H3 K27M‐mutant astrocytoma (Figure S1B).

3.2. The molecular landscape reveals the molecular heterogeneity of H3 K27M‐mutant spinal cord diffuse gliomas

We next summarized the molecular characteristics of H3 K27M‐mutant spinal cord diffuse glioma (Figures 2A and S2A, Table S4). Approximately, 82.35% of spinal cord H3 K27M‐mutant diffuse glioma had TP53 (n = 20/34) or PPM1D mutations (n = 9/34), and these were mutually exclusive (p = 0.0012, Fisher's test). Interestingly, NF1 mutation occurred in 44.12% (15/34) of H3 K27M‐mutant diffuse glioma and was mutually exclusive to TP53 mutation (p = 0.0021, Fisher's test). Alterations in RTK/RAS/PI3K/MAPK pathway components in H3 K27M‐mutant tumors included PDGFRA mutation/amplification (16/34), KIT amplification (11/34), EGFR amplification (1/34), MET amplification (2/34), PTEN mutation/deletion (5/34), KRAS mutation/amplification (3/34), PIK3CA mutation (7/34), PIK3R1 mutation (2/34), NRTK2 amplification (3/34), NTRK3 mutation/amplification (5/34), and others. Alterations in RB1 signaling pathway factors were present in 44.12% of H3 K27M‐mutant tumors, including CDK4 amplification (3/34), CDK6 amplification (5/34), CDKN2A/B deletion (6/34), and RB1 deletion (9/34). MYC amplification (4/34) and MYCN amplification (5/34) occurred in 26.47% (9/34) of cases. In addition, chromosome (Chr) 7p, Chr 7q gain, Chr 10q deletion, and Chr 19q deletion were observed in more than 10% of cases, respectively.

FIGURE 2.

The molecular features of H3 K27M‐mutant spinal cord diffuse glioma. (A) The molecular landscape of H3 K27M‐mutant spinal cord diffuse glioma. (B) Comparison of numbers of genes and chromosome (Chr) arms with copy number alteration (copy number variant [CNV]) in TP53‐mutant (mut) and TP53‐wildtype (WT) cases. (C) The distribution age in cases with TP53‐mut and TP53‐WT tumors. (D) Kaplan–Meier survival curves of patients with TP53‐mut and TP53‐WT tumors.

The mutual exclusivity of TP53 and other mutations, such as NF1 and PPM1D mutations, suggested the potential molecular heterogeneity of H3 K27M‐mutant diffuse glioma. Furthermore, we found that TP53‐mutant cases had a significantly higher number of copy number variants (CNV), including RB1 deletion, Chr 7p amplification, Chr 10q deletion, and Chr 19q deletion. RB1 deletion/mutation and Chr 10 loss exclusively occurred in cases with TP53 mutation (Figure 2A,B and Table S5). Whereas NF1 mutation, PPM1D mutation, and MYCN amplification tended to be present in tumors without TP53 mutation. Interestingly, the DNA methylation profiles of TP53‐mutant (case70, case73, case74, and case77) and TP53‐wildtype cases (case59, case75, and case76) could also be separated by the t‐SNE plot (Figure 1D), suggesting the molecular disparities between these tumors.

We then compared the clinicopathological features of TP53‐mutant and TP53‐wildtype tumors. Interestingly, the TP53‐mutant tumors had a marginally higher proportion of pediatric patients (age at diagnosis <18 years old, p = 0.056; Figure 2C). There was no difference in gender distribution and histological type distribution between the two groups of tumors (Figure S2B,C). Patients in the TP53‐mutant and TP53‐widltype   subgroups had a similar OS (Figure 2D).

3.3. The prognostic value of molecular features in H3 K27M‐mutant spinal cord diffuse glioma

To evaluate the prognostic value of each molecular and pathway alteration in H3 K27M‐mutant spinal cord tumors, we performed a univariate analysis with molecular features and the OS of patients (Table S6). We found that Chr 9p deletion (hazard ratio [HR]: 3.71, and 95% confidence interval [95% CI]: 1.45–9.46), Chr 10p deletion (HR: 4.75, 95% CI: 1.46–15.48), CDK6 amplification (HR: 3.91, 95% CI: 1.22–12.60), ATRX mutation (HR: 3.45, 95% CI: 1.19–10.04), Chr 1p amplification (HR: 14.65, 95% CI: 1.33–161.63), and RB1 pathway alteration (HR: 1.54, 95% CI: 1.02–2.34) are risk factors for survival. Then, we further verified the prognostic value of molecular features which account for more than 10% of cases via Kaplan–Meier survival curves (Figure 3). The result showed that Chr 9p deletion (Figure 3A), Chr 10p deletion (Figure 3B), CDK6 amplification (Figure 3B), ATRX mutation (Figure 3C), a higher number of CNV events (≥3, Figure 3D), and RB pathway alteration (Figure 3F) are associated with worse survival of patients with H3 K27M‐mutant spinal cord tumors.

FIGURE 3.

The prognostic value of molecular features of H3 K27M‐mutant spinal cord diffuse glioma. (A–C) Kaplan–Meier survival curves of patients with or without Chr 9p deletion (A), Chr 10p deletion (B), and CDK6 amplification (C). (D) Kaplan–Meier survival curves of patients with or without ATRX mutation. (E) Kaplan–Meier survival curves of patients stratified by the number of copy number variant (CNV) alterations. (F) Kaplan–Meier survival curves of patients with or without RB pathway alteration.

3.4. Histological type can stratify the prognosis and molecular features of H3 K27M‐mutant spinal cord diffuse glioma

After characterizing the molecular features and their prognostic value, we want to explore the prognostic value of other clinicopathological features in this large cohort of H3 K27M‐mutant spinal cord astrocytomas. We performed a univariate Cox analysis of clinicopathological features in 74 cases (Table 1). We found that the high Ki‐67 index (≥10%, p = 0.011) and histological type of GBM (p = 0.0021) were risk factors for survival of patients with H3 K27M‐mutant spinal cord diffuse glioma. The HR and 95% CI for high Ki‐67 index and histological type of GBM was 3.39 (1.32–8.72) and 1.82 (1.25–2.69), respectively. Partial resection (p = 0.096), longer tumor length (p = 0.064), and female (p = 0.055) were only marginally associated with poorer survival. Kaplan–Meier curves also showed that patients with GBM had significantly shorter OS than A and AA cases (p < 0.0001; Figures S3 and 4A). Patients with high Ki‐67 index also had a shorter OS (p = 0.0073; Figure 4B). Multivariate Cox analysis indicated that only higher histological grade (HR [95% CI] = 2.39 [1.28–4.49], and p = 0.0066) was significantly associated with poorer prognosis of tumors (Figure 4C).

TABLE 1.

The univariate Cox analysis results for different clinicopathological features in spinal cord astrocytomas with H3 K27M mutant.

		HR	Low95	High95	p‐value
Gender	Male versus female	0.5962	0.3514	1.0114	0.0551
Age	Increasing	0.9981	0.9778	1.0188	0.8543
Location
	“C4–T1” versus “C1–C3”	0.8025	0.3145	2.0477	0.6452
	“T2–T11” versus “C1–C3”	0.7170	0.4462	1.1522	0.1693
	“T12–L1” versus “C1–C3”	0.9258	0.6602	1.2983	0.6550
Tumor length	Longer versus shorter	1.1572	0.9913	1.3508	0.0644
Extent of resection
	STR versus GTR	1.3519	0.5786	3.1587	0.4862
	PR versus GRT	1.3215	0.9517	1.8352	0.0961
Histological grade	GBM versus AA versus A	1.8294	1.2449	2.6884	0.0021
Ki‐67	≥10% versus <10%	3.3924	1.3192	8.7237	0.0112
Radiotherapy	With versus without	1.0953	0.6232	1.9248	0.7517
Chemotherapy	With versus without	1.1090	0.6434	1.9115	0.7097
TERT promoter	Mutant versus wildtype	0.7306	0.2833	1.8844	0.5162

Abbreviations: C, cervical cord (C1–C3); C–T, cervicothoracic cord (C4–T1, cervical enlargement); GTR, gross total resection; PR, partial resection; STR, subtotal resection; T, thoracic cord (T2–T11); T–L, thoracolumbar cord (T12–L1, conus medullaris).

FIGURE 4.

The prognostic factors of spinal cord astrocytic tumors with H3 K27M. (A–B) Kaplan–Meier survival curves of patients stratified by histological type (A) and Ki67 index (B) in patients with spinal cord diffuse glioma. (C) The multivariate Cox analysis of gender, histological type, the extent of resection, tumor length, and Ki67 index in patients with H3 K27M‐mutant spinal cord astrocytic tumors.

We noticed that patients with GBM also had significantly shorter OS (p = 0.02) in the 34 cases with molecular features (Figures 5A and S4). To reveal whether molecular features also contributed to the prognostic value of histological type in H3 K27M‐mutant spinal cord diffuse glioma, we explored the molecular differences between GBM and A/AA (Figures 5B and S5). Interestingly, we found alterations in KRAS/PI3K pathway are significantly enriched in GBM cases (p = 0.0044, Fisher's test), and KRAS, PIK3CA, and PIK3R1 alterations exclusively occurred in GBM. The chromosome arms CNV were also enriched in GBM cases (p = 0.0382, Fisher's test). Strikingly, we also observed that CDK6 amplification and CDKN2A/B deletion exclusively occurred in nine of 19 GBM cases, suggesting activation of cell proliferation in GBM. Consistently, we observed a higher proportion of GBM cases with a high Ki‐67 index (p = 0.0018, Fisher's test; Figure 5C and Table 2). Interestingly, GBM patients had a lower age at diagnosis than patients with A/AA (p = 0.0073, Figure 5D and Table 2). These findings show that histological type is also associated with critical molecular alterations in H3 K27M‐mutant spinal cord diffuse glioma.

FIGURE 5.

Molecular differences between H3 K27M‐mutant astrocytoma (A)/anaplastic astrocytoma (AA) and glioblastoma (GBM). (A) Kaplan–Meier survival curves of patients with panel sequencing data. (B) Comparison of molecular alteration frequencies between A/AA and GBM. Fisher's test calculated the p‐values. (C) The distribution of Ki‐67 index in H3 K27M mutant spinal cord A/AA and GBM. Fisher's test calculated the p‐value. (D) The age at diagnosis of patients with H3 K27M mutant spinal cord A/AA and GBM. The nonparametric test calculated the p‐value.

TABLE 2.

Clinicopathological features of H3 K27M‐mutant spinal cord A/AA and GBM.

		A/AA	GBM	p‐value
Number		44	33
Age (year)	Median (range)	36 (9–57)	28 (7–61)	0.0073
Gender
	Male	28 (64%)	14 (42%)	0.1049
	Female	16 (36%)	19 (58%)
Location
	C	5 (11%)	4 (12%)	0.7665
	C–T	14 (32%)	13 (39%)
	T	14 (32%)	11 (33%)
	T–L	11 (25%)	5 (15%)
Resection				1.000
	GTR	11 (25%)	8 (24%)
	STR	10 (23%)	7 (21%)
	OB	23 (52%)	18 (55%)
Radiotherapy				0.6186
	Yes	26 (59%)	21 (64%)
	No	15 (34%)	9 (27%)
	Unknown	3 (7%)	3 (9%)
Chemotherapy				1.000
	Yes	20 (45%)	15 (45%)
	No	21 (48%)	15 (45%)
	Unknown	3 (7%)	3 (9%)
IDH
	Wildtype	37 (84%)	28 (85%)	1.000
	Mutant	0 (0%)	0 (0%)
	Unknown	7 (16%)	5 (15%)
TERT promotor
	Wildtype	30 (68%)	24 (73%)	1
	Mutant	5 (11%)	4 (12%)
	Unknown	9 (20%)	5 (15%)
BRAF V600E				1
	Wildtype	35 (80%)	26 (79%)
	Mutant	0 (0%)	0 (0%)
	Unknown	9 (20%)	7 (21%)
Ki‐67				0.0018
	<10%	11 (25%)	0 (0%)
	≥10%	32 (73%)	31 (94%)
	Unknown	1 (2%)	2 (6%)

Abbreviations: A, astrocytoma; AA, anaplastic astrocytoma; C, cervical cord (C1–C3); C–T, cervicothoracic cord (C4–T1, cervical enlargement); GBM, glioblastoma; GTR, gross total resection; PR, partial resection; STR, subtotal resection; T, thoracic cord (T2–T11); T–L, thoracolumbar cord (T12–L1, conus medullaris). The p‐value less than 0.05 were marked in bold. 

4. DISCUSSION

Intensive studies have provided an insightful understanding of the tumorigenesis of pediatric brain stem H3 K27M‐mutant diffuse glioma, which have led to several potential targeted therapies, such as HDAC, JMJD3, and BET bromodomain inhibitors [35, 36, 37, 38]. However, distinct molecular features have been revealed among H3 K27M‐mutant diffuse gliomas with different anatomic localization [7, 14]. The molecular and clinicopathological features of H3 K27M‐mutant spinal cord diffuse gliomas still need to be clarified, thus impeding the accurate clinical management of patients [4, 39]. Here, we portrayed molecular and clinicopathological characteristics of H3 K27M‐mutant spinal cord diffuse glioma in a relatively large cohort. We highlighted the molecular heterogeneity of H3 K27M‐mutant spinal cord astrocytoma. Based on this, we identified critical molecular features associated with the prognosis of H3 K27M‐mutant spinal cord diffuse glioma. All H3 K27M‐mutant spinal cord diffuse glioma should be diagnosed as WHO Grade 4 regardless of histological grade according to the latest WHO classification [1]. However, our findings demonstrated that histological type was an independent prognostic factor for H3 K27M‐mutant spinal diffuse glioma. Moreover, we found that the histological type can also stratify the molecular features of H3 K27M‐mutant spinal cord diffuse glioma, especially of RB and KRAS/PI3K pathways. These findings provide essential information for the clinical management of H3 K27M‐mutant spinal cord diffuse glioma.

It has been reported that p53 alteration and H3K27me3 loss may play an essential role in the malignant progression of giant cell tumors of bone [40]. Here, we observed that TP53 mutation occurred in 58.8% of H3 K27M‐mutant spinal cord astrocytoma, suggesting a vital role of TP53 dysfunction in the pathogenesis of this tumor. Recently, the H3 K27M‐mutant brain stem diffuse glioma localized to the medulla and pons was proven to be two subgroups (H3‐Medulla and H3‐Pons) by DNA methylation profiling [14]. Interestingly, NF1 and PPM1D were more frequent in cluster H3‐Medulla with a better prognosis and older age, while the percentage of TP53 mutations was higher in H3‐Pons with a worse prognosis and younger age. Here, we also noticed that TP53 mutant was mutually exclusive with PPM1D and NF1 mutation in H3‐K27M‐mutant spinal cord diffuse glioma. We found that the TP53‐mutant cases had a significantly higher number of CNV, including KIT/PDGFRA amplification, CDK6 amplification, RB1 deletion, Chr 7p amplification, Chr 10q deletion, and Chr 19q deletion, which is consistent with the essential role of TP53 in maintaining chromosome stability. These findings suggest potentially different pathogenesis of TP53‐mutant and TP53‐wildtype H3 K27M‐mutant spinal cord diffuse glioma. However, unlike H3‐Medulla and H3‐Pons with different prognoses, the TP53‐mutant and TP53‐wildtype H3 K27M‐mutant spinal cord diffuse glioma had similar prognoses in our cohort, suggesting a potential difference between diffuse midline gliomas in different anatomical locations.

The chromatin dysregulation has been recognized as one of the critical pathogenesis for H3 K27M‐mutant pediatric H3 K27M‐mutant glioma [41, 42]. Here, we observed that ATRX‐mutant cases have shorter OS than ATRX‐wildtype cases. As a member of the SWI/SNF family of chromatin remodeling proteins, ATRX mutation is also associated with an alternative lengthening of telomeres phenotype in glioma [43]. We also found that a higher number of CNV events (≥3), Chr 9p deletion, and Chr 10p deletion are associated with poorer survival. These findings suggest that chromatin instability is a factor that is closely related to tumor survival and pathogenesis in spinal cord H3 K27M‐mutant diffuse glioma. Meanwhile, we found that CDK6 amplification and RB pathway alteration are also associated with worse survival. Furthermore, a high Ki‐67 index (>10%) is a risky prognosis factor in spinal cord H3 K27M‐mutant glioma. Together, these findings suggest that cell proliferating activity is another critical survival‐associated factor in spinal cord H3 K27M‐mutant glioma.

The limitation of this study included the retrospective analysis and the limited number of patients with DNA methylation data. Future studies with adequate follow‐up time and multi‐omics data would be helpful for further revealing molecular characteristics of spinal cord diffuse gliomas.

In conclusion, in the largest cohort to our knowledge, we successfully characterized the molecular, clinicopathological, and prognostic characteristics of H3 K27M‐mutant spinal cord diffuse glioma. We revealed the molecular heterogeneity between H3 K27M‐mutant spinal cord gliomas with and without TP53 mutation. We showed that molecular alterations representing chromatin instability and accelerated cell proliferative are risky prognostic factors for H3 K27M‐mutant spinal cord diffuse glioma. We also demonstrated that histological types are associated with the prognosis and molecular characteristics of H3 K27M‐mutant spinal cord diffuse glioma. Overall, our findings provide critical information for the accurate diagnosis and clinical management of this highly lethal and disabling tumor.

AUTHOR CONTRIBUTIONS

Rui‐Chao Chai, Yong‐Zhi Wang, and Wen‐Qing Jia designed the study. Rui‐Chao Chai wrote the manuscript. Rui‐Chao Chai and Hao‐Yan conducted the data analysis and information collection. Rui‐Chao Chai and Yu‐Qing Liu performed molecular testing. Song‐Yuan An, Bo Pang, and Yao‐Wu Zhang collected the clinical information. Hui‐Yuan Chen and Xing Liu performed the histological diagnosis. Quan‐Hua Mu, Ke‐Nan Zhang, and Zheng Zhao performed analysis for DNA methylation data. Tao Jiang, Yong‐Zhi Wang, and Wen‐Qing Jia supervised the study and contributed intellectually to this work. All authors read and approved the final manuscript.

FUNDING INFORMATION

This work was supported by the Beijing Nova Program (Z201100006820118) and the National Natural Science Foundation of China (81903078, 82172783).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Figure S1. (A) The MNP diagnosis results of spinal cord astrocytic tumors with H3 K27M. (B) The H&E images, H3 K27M staining images, mutational information, and copy number variations of case74 were presented.

Figure S2. (A) The mutational landscape of H3 K27M‐mutant spinal cord astrocytoma. (B,C).

Figure S3. Kaplan–Meier survival curves of patients stratified by histological type.

Figure S4. Kaplan–Meier survival curves of patients stratified by histological type.

Figure S5. (A) Molecular features of H3 K27M‐mutant tumors with different histological types.

Table S1. Summary of clinical‐pathological, target‐sequencing, and methylation assay information of all cases in this study

Table S2. Target list of targeted sequencing.

Table S3. Clinicopathological features of spinal cord astrocytomas with H3 K27M mutation.

Table S4. the molecular features of all cases with panel sequencing results.

Table S5. Comparison between molecular features of TP53‐mutant and TP53‐wildtype tumors.

Table S6. The univariate cox analysis of genetic alterations in H3 K27M‐mutant spinal cord diffuse gliomas.

Table S7. The abbreviation of methylation classification.

Chai R‐C, Yan H, An S‐Y, Pang B, Chen H‐Y, Mu Q‐H, et al. Genomic profiling and prognostic factors of H3 K27M‐mutant spinal cord diffuse glioma. Brain Pathology. 2023;33(4):e13153. 10.1111/bpa.13153

DATA AVAILABILITY STATEMENT

The reference methylation data could be available in GEO (GSE90496); the panel sequencing results have been shown in supplementary table 4.