Immunological profiles of human oligodendrogliomas define two distinct molecular subtypes

Abstract

Background: Human oligodendroglioma presents as a heterogeneous disease, primarily characterized by the isocitrate dehydrogenase (IDH) mutation and 1p/19q co-deletion. Therapy development for this tumor is hindered by incomplete knowledge of somatic driving alterations and suboptimal disease classification. We herein aim to identify intrinsic molecular subtypes through integrated analysis of transcriptome, genome and methylome.

Methods: 137 oligodendroglioma patients from the Cancer Genome Atlas (TCGA) dataset were collected for unsupervised clustering analysis of immune gene expression profiles and comparative analysis of genome and methylome. Two independent datasets containing 218 patients were used for validation.

Findings: We identified and independently validated two reproducible subtypes associated with distinct molecular characteristics and clinical outcomes. The proliferative subtype, named Oligo1, was characterized by more tumors of CNS WHO grade 3, as well as worse prognosis compared to the Oligo2 subtype. Besides the clinicopathologic features, Oligo1 exhibited enrichment of cell proliferation, regulation of cell cycle and Wnt signaling pathways, and significantly altered genes, such as EGFR, NOTCH1 and MET. In contrast, Oligo2, with favorable outcome, presented increased activation of immune response and metabolic process. Higher T cell/APC co-inhibition and inhibitory checkpoint levels were observed in Oligo2 tumors. Finally, multivariable analysis revealed our classification was an independent prognostic factor in oligodendrogliomas, and the robustness of these molecular subgroups was verified in the validation cohorts.

Interpretation: This study provides further insights into patient stratification as well as presents opportunities for therapeutic development in human oligodendrogliomas.

Funding: The funders are listed in the Acknowledgement.

Keywords: Immunological profiling; Molecular subtype; Multi-omics; Oligodendrogliomas; Prognosis.

Fig. 1

Immune gene profile of oligodendrogliomas yields two subtypes in TCGA cohort. a Flow chart of this study. A total of 355 oligodendroglioma samples are used in this study. TCGA cohort is used as training set, CGGA and POLA cohorts are collected as validation sets. b Heatmap of two immune subtypes defined in TCGA cohort. 40 genes of centroid derived from PAM classifier are showed. Patients are arranged based on the subtypes. Molecular and clinical information are also annotated for each patient. CL, classical; MES, mesenchymal; PN, proneural. c Principal component analysis (PCA) of two subtypes using whole transcriptome data. d Survival analysis of immune subtypes based on OS. P value is calculated by the log-rank test between subtypes.

Fig. 2

External validation of the immune class in the publicly available CGGA and POLA datasets. a Heatmap shows the immune subtypes of CGGA cohort predicted by a PAM classifier trained on the TCGA cohort. Patients are arranged based on the predicted immune subtypes. The 40 centroid genes and clinical information are displayed. CL, classical; MES, mesenchymal; PN, proneural. b Principal component analysis (PCA) of two subtypes using whole transcriptome data in CGGA cohort. c The Kaplan–Meier analysis (log-rank test) of immune subtypes based on OS. d Heatmap shows the immune subtypes of POLA cohort predicted by a PAM classifier trained on the TCGA cohort. The 40 centroid genes and clinical information are also displayed. e PCA of immune subtypes using whole transcriptome data in POLA cohort. f Survival analysis of immune subtypes in POLA cohort. P value is calculated by the log-rank test between subtypes.

Fig. 3

Comparison of the DNA alterations between immune subtypes in TCGA cohort. a Boxplots show the difference of DNA damage measures between subtypes of TCGA cohort (Wilcon rank-sum test). Error bars show standard error of the mean, and the middle bar represents the median level of corresponding features. ∗P < 0.05, ∗∗∗P < 0.001 b Oncoprint of mutation status and copy number variations between immune subtypes. Fisher test is adopted for comparison analysis. The differentially altered genes annotated to Wnt signaling, regulation of neurogenesis, and semaphorin-plexin pathway are displayed.

Fig. 4

Comparison of the DNA methylation between subtypes in TCGA and POLA cohorts. a Heatmap of TCGA samples ordered according to subtypes using the top differential 637 probes defined by Wilcoxon test. b Gene Ontology (GO) analysis of the differentially methylated genes between subtypes in TCGA cohort. c Heatmap of POLA samples ordered according to subtypes using the top differential 500 probes defined by Wilcoxon test. d Functional annotation of the differentially methylated genes between subtypes in POLA cohort.

Fig. 5

Tumor immune infiltration dissection of two subtypes in TCGA cohort. a Comparison of immune, stromal, and tumor purity scores from ESTIMATE for immune subtypes (Wilcon rank-sum test). b Comparison of lymphocytes and macrophage proportion (from CIBERSORT) for immune subtypes (Wilcon rank-sum test). Lymphocytes consist of B cells, CD4 cells, CD8 cells, follicular helper T cells, Tregs, gamma delta T cells, NK cells, and plasma cells. c Relative abundance fractions of the immune cell population in two subtypes using CIBERSORT tool. d Heatmap shows enrichments of immune-related signatures in two immune subtypes. P values are labelled. e Boxplots display the expression levels of inhibitive checkpoint genes. ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001, Wilcon rank-sum test.

Fig. 6

Immunostaining analyses of immune cells and checkpoints between subtypes. a 12 Oligo1 and 12 Oligo2 cases were collected from CGGA cohort. NOTCH1, TIGIT, HAVCR2, CD206, CD163, CD14, and CD3 antibodies were used to assess cell proportions. b Boxplots shows the positive cell proportion evaluated by immunostaining. Scaled bar, 50 μm ∗∗P < 0.01, ∗∗∗P < 0.001, Student t-test.

Fig. 7

Association between the acquired immune subtypes and metabolism-relevant signatures. a and b Heatmaps show the differential enrichments of metabolism-related signatures in the TCGA and CGGA cohorts. Amina acid, carbohydrate, lipid, vitamin, and other metabolism signatures are presented. The statistical difference is compared through the Wilcon rank-sum test, and the P value < 0.05 is considered significant.

Fig. 8

Overview of immune subtypes of human oligodendrogliomas. We identified two expression-based subgroups in oligodendrogliomas, and the clinical characteristics, somatic variations, DNA methylation data, transcriptional data, immune infiltration, and metabolic enrichments of these two immune subtypes were explored.