Integrative multi-omic analysis of NLRP3 inflammasome dysregulation and subtyping for personalized treatment in acute myeloid leukemia

Abstract

Acute myeloid leukemia (AML) is a devastating form of blood cancer characterized by uncontrolled growth and impaired maturation of myeloid precursor cells in the bone marrow. Despite advancements in treatment strategies, the prognosis for AML patients remains poor. The NLRP3 inflammasome, a multi-protein complex involved in innate immunity and inflammation, has been implicated in various diseases; however, its role in AML development and progression is not well understood. In this study, we analyzed genomic, bulk-, and single-cell transcriptomic data to assess the contribution of NLRP3 inflammasome genes to AML. Results suggested that 28 NLRP3 inflammasome genes with clinical implications were dysregulated in AML. Notably, we identified seven prognosis-related genes: CASP1, CPTP, MEFV, NFKB2, PANX1, PYCARD, and SIRT2. To further investigate the functional relevance of NLRP3 inflammasome genes, we developed an NLRP3 score (Nscore) based on the expression levels of these seven genes. We identified four dysregulated gene clusters that distinguished high- and low-Nscore groups, enabling the identification of two distinct AML subtypes, with subtype 2 exhibiting worse overall survival than subtype 1. Additionally, using a network-based approach, we identified 62 NLRP3 inflammasome-related genes and constructed a risk score model with COL2A1, ITGB2, and SRC genes, providing a comprehensive assessment of patient risk stratification. We revealed a positive correlation between the Nscore and macrophage M1, suggesting potential drug response mechanisms. Based on the identified AML subtypes and their distinct mutational landscapes, we predicted potential treatment options, with Paclitaxel, Afuresertib, and Mitoxantrone emerging as potential therapeutic agents for the AML subtypes.

Keywords: Acute myeloid leukemia; Drug response; NLRP3 inflammasome; Prognosis; Tumor immune microenvironment.

Fig. 1

Dysregulation of NLRP3 inflammasome genes with clinical implications in LAML. A PCA plot comparing TCGA-LAML and healthy control samples, demonstrating distinct clustering patterns. B Heatmap illustrating the expression levels of NLRP3 inflammasome genes in TCGA-LAML samples compared to healthy control samples. C Forest plot of the uni-variable Cox regression model, displaying the hazard ratios and their corresponding 95% confidence intervals for NLRP3 inflammasome genes. Orange nodes indicate statistically significant associations (p < 0.05). D Forest plot of the multi-variable Cox regression model, presenting the hazard ratios and their corresponding 95% confidence intervals for NLRP3 inflammasome genes after adjusting for relevant covariates. E Violin plots depicting the expression differences of NLRP3 inflammasome genes across CALGB categories. F Dot plots showing the expression differences of NLRP3 inflammasome genes across FAB categories

Fig. 2

Identification of prognostic NLRP3 inflammasome genes and four gene clusters with distinct functions. A KM curve comparing overall survival between patients with high- and low-Nscore. B Volcano plot displaying the dysregulated protein-coding genes between patients with high- and low-Nscore. C Heatmap illustrating the expression patterns of dysregulated protein-coding genes between patients with high- and low-Nscore. D Barplot showing the functional enrichment results of the four identified gene clusters

Fig. 3

Four NLRP3 inflammasome-related gene clusters defined two LAML subtypes. A Consensus clustering result of two distinct TCGA-LAML subtypes. B Clinski criterion plot depicting the evaluation of the consensus clustering result, aiding in the determination of the optimal number of subtypes. C KM survival curve comparing overall survival between the identified LAML subtypes. D Barplot illustrating the dysregulated functional gene sets associated with the identified LAML subtypes

Fig. 4

Identification and evaluation of NLRP3-related gene prognostic index. A Network-based identification of 62 NLRP3 inflammasome-related prognostic genes, displayed with colors representing the log2 fold change values, indicating the degree of differential expression between high- and low-Nscore groups. B Network-based identification of 62 NLRP3 inflammation-related prognostic genes, displayed with colors representing the HR values from uni-variable Cox survival analysis, indicating the association with patient survival. C LASSO regression coefficients, representing the contributions of each variable in the model. D Cross-validation for the LASSO regression, aiding in the selection of the optimal penalty parameter. E KM curve comparing overall survival between high- and low-risk patients in the TCGA-LAML cohort. F Area under the curve (AUC) values of the Risk Score prediction model for different time periods in the TCGA-LAML cohort, indicating the predictive accuracy of the model. G, H KM survival curves comparing overall survival between high- and low-risk patients in the GSE71014 and GSE37642 cohorts, respectively. I, J AUC values of the risk score prediction model for different time periods in the GSE71014 and GSE37642 cohorts

Fig. 5

LAML subtypes with higher Nscore was positively correlated with macrophages abundance. A Heatmap showing the immune cell abundance differences between the two TCGA-LAML subtypes. B Correlation plot demonstrating the relationship between Nscore and immune cell abundances in the TCGA cohort. C Uniform Manifold Approximation and Projection (UMAP) plot displaying the cell type clustering of scRNA-seq cohort. D Expression levels of four prognostic NLRP3 inflammasome genes in the scRNA-seq cohort. E Violin plot depicting the expression differences of the four prognostic NLRP3 inflammasome genes across different cell types in the scRNA-seq cohort. F Correlation curve illustrating the relationship between immune score and Nscore. G Correlation curve illustrating the relationship between MHC score and Nscore. H Differences in T cell dysfunction score between the high- and low-Nscore groups

Fig. 6

AML subtypes with higher Nscore exhibited distinct drug responses and potential treatment strategies. A Oncoplot depicting genes with high-frequency somatic mutations in TCGA-LAML subtype 1. B Oncoplot depicting genes with high-frequency somatic mutations in TCGA-LAML subtype 2. C Top co-occurrence somatic mutation of genes in two subtypes (top left: subtype 2, bottom right: subtype 1). D Oncogenic pathway mutation plot illustrating the mutation status of oncogenic pathways in the two subtypes (top: subtype 2, bottom: subtype 1). E Volcano plot displaying drugs with dysregulated IC50 values between high- and low-Nscore groups in the GDSC dataset. F Correlation plot illustrating the correlations between Nscore and IC50 values of drugs