Metabolic Signature-Based Subtypes May Pave Novel Ways for Low-Grade Glioma Prognosis and Therapy

Abstract

Metabolic signatures are frequently observed in cancer and are starting to be recognized as important regulators for tumor progression and therapy. Because metabolism genes are involved in tumor initiation and progression, little is known about the metabolic genomic profiles in low-grade glioma (LGG). Here, we applied bioinformatics analysis to determine the metabolic characteristics of patients with LGG from the Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA). We also performed the ConsensusClusterPlus, the CIBERSORT algorithm, the Estimate software, the R package "GSVA," and TIDE to comprehensively describe and compare the characteristic difference between three metabolic subtypes. The R package WGCNA helped us to identify co-expression modules with associated metabolic subtypes. We found that LGG patients were classified into three subtypes based on 113 metabolic characteristics. MC1 patients had poor prognoses and MC3 patients obtained longer survival times. The different metabolic subtypes had different metabolic and immune characteristics, and may have different response patterns to immunotherapy. Based on the metabolic subtype, different patterns were exhibited that reflected the characteristics of each subtype. We also identified eight potential genetic markers associated with the characteristic index of metabolic subtypes. In conclusion, a comprehensive understanding of metabolism associated characteristics and classifications may improve clinical outcomes for LGG.

Keywords: immune characteristics; low-grade glioma; metabolic signature; prognosis; subtypes.

FIGURE 1

Metabolic subtypes in LGGA. (A) The intersection of prognostic metabolic signatures between TCGA and CGGA. (B) CDF curve of the TCGA cohort. (C) CDF delta area curve. The horizontal axis represents the category number k and the vertical axis represents the relative change in area under the CDF curve. (D) Sample clustering heatmap, k = 3. (E) Overall survival curve of three metabolic subtypes in the TCGA cohort. (F) PFS curve of metabolic subtypes in TCGA cohort. (G) OS curves of three subtypes in the CGGA cohort.

FIGURE 2

The genomic alteration of the three clusters in the TCGA cohort. (A–E) Comparisons of aneuploidy score, fraction altered, number of segments, tumor mutation burden and homologous recombination defects in the TCGA cohort. (F) The somatic mutations and copy number mutations of the three clusters.

FIGURE 3

Comparison of differences in classical immune cell typing and immune cell composition analysis. (A) Comparative analysis of the metabolic molecular subtypes in TCGA and the reported six classical subtypes. (B) Comparative analysis of the metabolic molecular subtypes in TCGA and the previous six pan-cancer immune molecular subtypes. (C) Immune cell composition and proportion in the TCGA cohort given by ESTIMATE software. (D) Immune cell composition and proportion in the CGGA cohort given by ESTIMATE software. (E) Immune cell composition and proportion in the TCGA cohort given by CIBERSORT software. (F) Immune cell composition and proportion in the CGGA cohort given by CIBERSORT software.

FIGURE 4

The immunotherapy response difference between the three clusters. (A) TIDE scores of all three metabolic subtypes in the TCGA cohort. (B) T cell dysfunction scores of all three metabolic subtypes in the TCGA cohort. (C) T cell rejection scores of different metabolic subtypes of TCGA. (D) Immune response status in different metabolic subtypes of TCGA. (E) TIDE scores of different metabolic subtypes in the CGGA cohort. (F) T cell dysfunction scores of different metabolic subtypes in the CGGA cohort. (G) T cell rejection scores of different metabolic subtypes in the CGGA cohort. (H) Differences of immune response status in different metabolic subtypes in the CGGA cohort. (I) Different immunotherapy sensitivity in programmed cell death protein 1 inhibitor therapy in the TCGA cohort. (J) Different immunotherapy sensitivity in programmed cell death protein 1 inhibitor therapy in the CGGA cohort.

FIGURE 5

Construction of metabolic subtype characteristic score. (A) The relationship between two key metabolic signatures and metabolic subtypes in the TCGA cohort. (B) The metabolic subtype signature scores of different subtypes in the TCGA cohort. (C) The ROC curve for metabolic subtype signature scores in the TCGA cohort. (D) The relationship between two key metabolic signatures and metabolic subtypes in the CGGA cohort. (E) The metabolic subtype signature scores between different subtypes in the CGGA cohort. (F) The ROC curve for metabolic subtype signature scores in the CGGA cohort.

FIGURE 6

Co-expressed gene modules identification. (A) Clustering tree of each sample. (B) The scale-free fit index for various soft-thresholding powers (β). (C) The mean connectivity for various soft-thresholding powers. (D) Dendrogram of all differentially expressed genes/lncRNAs, clustered based on a dissimilarity measure. (E) Co-expression module gene statistical results. (F) Correspondence between each module and clinical information. (G) Scatter diagram for module membership vs gene significance for MC1 in the green module. (H) Scatter diagram for module membership vs gene significance for MC2 in the cyan module. (I) Scatter diagram for module membership vs gene significance for MC3 in the blue module.

FIGURE 7

Functional enrichment analysis of metabolic co-expression gene module. (A) Correlation analysis between module feature vector and metabolic subtype feature index. (B) The correlation between the feature vector of the green module and the feature index of metabolic subtypes. (C) Correlation between the cyan module and the feature index of metabolic subtypes. (D) Correlation between the blue module feature vector and the metabolic subtype feature index. (E–G) Functional enrichment analysis results for green, cyan, and blue modules.

FIGURE 8

Hub genes identification of metabolic co-expressed gene module. (A) The protein interaction network between the key genes of the modules; the different colors of the network nodes indicate different modules. (B) Venn diagram of key genes. (C) Kaplan-Meier prognostic curve of marker genes related to metabolic subtype feature score.