Integrating bulk RNA-seq and scRNA-seq data to explore diverse cell death patterns and develop a programmed cell death-related relapse prediction model in pediatric B-ALL

Abstract

Acute B-lymphoblastic leukemia (B-ALL) is a hematologic malignancy with diverse mechanisms of PCD influencing its progression. This study aimed to identify PCD-related biomarkers and develop a predictive model for relapse in pediatric B-ALL patients. Initially, we examined the activity of 16 PCD patterns in B-ALL patients using scRNA-seq. Following this, we employed both univariate and multivariate Cox regression analyses to identify relapse-related PCD patterns and constructed a relapse prediction model comprising seven key PCD-related genes: Bcl-2-interacting killer (BIK), translocator protein (TSPO), BCL2L2, PIP4K2C, mixed-lineage kinase-like (MLKL), STAT2, and WW domain-containing oxidoreductase (WWOX). Based on the optimal cut-off value derived from the cell death index(CDI) model, patients were categorized into high-CDI and low-CDI groups. Additionally, we evaluated the association between CDI scores and immune cell infiltration, tumor microenvironment (TME) characteristics, and drug sensitivity. Nine PCD patterns, encompassing ferroptosis, autophagy, necroptosis, entotic cell death, alkaliptosis, apoptosis, netotic cell death, oxeiptosis, and NETosis, exhibited strong associations with relapse in B-cell acute lymphoblastic leukemia (B-ALL). The CDI model, validated across multiple cohorts, demonstrated substantial predictive power for relapse-free survival (RFS) and was identified as an independent risk factor. This study offers a comprehensive analysis of PCD patterns in pediatric B-ALL, yielding valuable insights into potential novel therapeutic strategies and opportunities for personalized treatment approaches.

Keywords: Acute lymphoblastic leukemia; Prognosis; Programmed cell death; Single-cell RNA sequencing.

Fig. 1

Workflow of the integrative bioinformatics analyses.

Fig. 2

Landscape of PCD patterns in B-ALL patients. (A) Univariate Cox analysis of 9 relapse-associated PCD patterns. (B) Kaplan-Meier survival curves of 9 PCD patterns associated with relapse.

Fig. 3

Investigation of various PCD patterns in single-cell RNA sequencing (scRNA-seq). (A) The distribution of cell types. (B-J) The distribution of activity scores for: apoptosis (B), autophagy (C), entotic cell death (D), ferroptosis (E), necroptosis (F), alkaliptosis (G), netotic cell death (H), NETosis (I), and oxeiptosis (J).

Fig. 4

Construction of the CDI model. (A,B) LASSO regression analysis for identifying candidate genes. (C-D) The Kaplan-Meier survival curves and survival status of different CDI groups are illustrated in the training cohort (C) and testing cohort (D). (E-F) The time-dependent ROC curves of the training cohort (E) and testing cohort (F).

Fig. 5

(A) Univariate Cox regression analysis and (B) multivariate Cox regression analysis demonstrated that the CDI model was an independent prognostic factor. (C) Univariate Cox regression analysis and (D) multivariate Cox regression analysis in high-risk group. (E) A nomogram was constructed using the CDI score and clinicopathological factors to predict 1-year, 3-year, and 5-year relapse-free survival rates in high-risk group. (F) Calibration plots for 1-year, 3-year, and 5-year relapse probabilities in high risk group.

Fig. 6

The relationship between CDI score and tumor microenvironment. (A) Differences in the infiltration of 22 immune cells between high-CDI and low-CDI patient groups. (B) Correlation between immune cells and 7 model genes. (C) Correlation between CDI score and immune checkpoints. (D-E) Correlation between CDI score with monocytes (D) and macrophages M2 (E). (F) Differences in stromal cell score, immune cell score and ESTIMATE score between high-CDI and low-CDI patient groups.

Fig. 7

Correlation between CDI score and the enrichment scores of HALLMARK pathways.

Fig. 8

Differences in chemotherapy sensitivity between high-CDI and low-CDI groups. (A-E).