A Novel Identified Necroptosis-Related Risk Signature for Prognosis Prediction and Immune Infiltration Indication in Acute Myeloid Leukemia Patients

Abstract

AML ranks second in the most common types of leukemia diagnosed in both adults and children. Necroptosis is a programmed inflammatory cell death form reported to be an innate immune effector against microbial and viral pathogens and recently has been found to play an eventful role in the oncogenesis, progression, and metastasis of cancer. This study is designed to explore the potential value of necroptosis in predicting prognostic and optimizing the current therapeutic strategies for AML patients. We collected transcriptome and clinical data from the Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) databases and selected necroptosis-related genes with both differential significance and prognostic value. Six genes (YBX3, ZBP1, CDC37, ALK, BRAF, and BNIP3) were incorporated to generate a risk model with the implementation of multivariate Cox regression. The signature was proven to be an independent prognostic predictor in both training and validation cohorts with hazard ratios (HRs) of 1.51 (95% CI: 1.33-1.72) and 1.57 (95% CI: 1.16-2.12), respectively. Moreover, receiver operating characteristic (ROC) curve was utilized to quantify the predictive performance of the signature and satisfying results were shown with the area under the curve (AUC) up to 0.801 (3-year) and 0.619 (3-year), respectively. In addition, the subtyping of AML patients based on the risk signature demonstrated a significant correlation with the immune cell infiltration and response to immunotherapy. Finally, we incorporated risk signature with the classical clinical features to establish a nomogram which may contribute to the improvement of clinical management. To conclude, this study identified a necroptosis-related signature as a novel biomarker to improve the risk stratification, to inform the immunotherapy efficacy, and to indicate the therapeutic option of targeted therapy.

Keywords: acute myeloid leukemia; immune infiltration; multivariate Cox regression; necroptosis; prognostic model.

Figure 1

Overall flow chart of research design.

Figure 2

Filtering of differentially expressed necroptosis-related genes of prognostic value. (A) Molecular interactions of 111 necroptosis-related regulators depicted by the STRING platform. (B) Venn plot shows integration of 3 groups of genes and the overlapping part consists of 22 genes. (C) 6 genes were eventually screened through the multivariate Cox regression to generate the prognostic model. p values are showed as: * p < 0.05; ** p < 0.01; **** p < 0.0005. (D) The real-time PCR results demonstrated the mRNA levels of 6 genes in the risk model in healthy donors (n = 3) and AML samples (n = 4). (Student t-test, * p < 0.05; ** p < 0.01; **** p < 0.0005).

Figure 3

Assessment of ability of the risk signature to predict prognostic in both training and validation cohorts. (A,B) Kaplan-Meier survival curves reveal the difference of prognostic in two risk subgroups in TCGA cohort (A) and GSE37642 cohort (B). (C,D) ROC curves showing the predictive power of the risk model on the survival rate in TCGA cohort (C) and GSE37642 cohort (D).

Figure 4

Two risk subgroups generated based on the median risk score. (A–D) Distribution of risk score, survival time and survival status for AML patients in TCGA cohort (A,C) and GSE37642 cohort (B,D). (E,F) Principal component analysis of the risk score to distinguish the two subgroups in TCGA cohort (E) and GSE37642 cohort (F). (G,H) Comparison of the expression of 6 genes in the risk model between the two subgroups in TCGA cohort (G) and GSE37642 cohort (H). (Wilcox test, ** p < 0.01; *** p < 0.001, **** p < 0.0005).

Figure 5

Characteristics of classical clinical features in two risk subgroups. (A) Heatmap visualizing the expression of 6 genes and clinicopathological features at different risk levels in TCGA cohort. (B) Alluvial diagram demonstrating a dynamic network of the interrelated clinical features in TCGA cohort.

Figure 6

Functional enrichment analysis performed on the DEGs identified between the two risk subgroups in TCGA cohort. (A) Volcano plot displaying DEGs between the two risk subgroups by the threshold of |log2FC| ≥ 0.4 and adjusted p-value < 0.01. Specific DEGs involved in the selected pathways are displayed. (B) KEGG pathways significantly enriched in the high-risk group in the bubble plot form. (C,E,F) Selected canonical biological process associated GSEA pathways in the high-risk group. (D) Representative GO terms enriched in terms of biological process (BP), cellular component (CC) and molecular function (MF) respectively in the high-risk group. p value is showed as: * p < 0.05.

Figure 7

Landscape of immune activity and relevant infiltrating immune cells in the two risk subgroups. (A,B) Heatmap showing the ssGSEA enrichment scores of 40 immune components including immune cells, relative immune factors and pathways in TCGA (A) and GSE37642 (B) cohorts. (C,D) Comparisons of proportion of 22 immune cells between the two risk subgroups in TCGA (C) and GSE37642 (D) cohorts (Wilcox text, * p < 0.05; ** p < 0.01; *** p < 0.001, **** p < 0.0005).

Figure 8

Evaluation of potential value of the risk model in the ICB immunotherapy in TCGA cohort. (A–F) Correlation of the 6 genes in the signature with TMB. (G) Differences of expression profiles of common checkpoint genes in the two risk subgroups (Wilcox text, ** p < 0.01; *** p < 0.001, **** p < 0.0005). (H) Correlation between the risk score and TIDE score (Spearman test, p < 0.05).

Figure 9

Construction of the OS-predictive nomogram for clinical application in TCGA cohort. (A,B) Univariate (A) and multivariate (B) Cox regression analysis of the risk signature to verify its independent predictive efficacy of OS. (C) ROC curves for the risk score and major clinical features including gender, age, white blood cell (WBC) count, blast cells percentage in bone marrow (BM) and cytogenetic risk. (D) Nomogram based on the combination of clinical features and risk score predicting 1-year, 3-year and 5-year survival. (E) ROC curves showing the predictive power of nomogram. (F) Calibration plot showing the predictive reliability of nomogram. p values are showed as: * p < 0.05; ** p < 0.01; *** p < 0.001.