Abstract: In conclusion, we constructed an inflammation related risk model using internal scRNA data and external bulk RNA data, which can accurately distinguish survival outcomes in AML patients.
Introduction
Acute Myeloid Leukemia (AML) is characterized by significant clinical and genetic heterogeneity. While some patients achieve long-term remission with standard induction therapy, others face rapid relapse and poor outcomes. Current prognostic tools often overlook the complex interactions within the bone marrow microenvironment, particularly the role of inflammation. Chronic inflammation is a known driver of tumor progression and immune evasion. This study aimed to bridge this gap by developing a robust inflammatory-related gene signature that can accurately predict patient survival and guide personalized treatment strategies.
Methods and Model Construction
The researchers employed a multi-step "conjoint analysis" approach, integrating high-resolution single-cell transcriptomics with large-scale bulk RNA sequencing data.
- Single-Cell Discovery: scRNA-seq was performed on bone marrow samples from 11 AML patients and 2 healthy controls. This analysis identified nine distinct cell clusters. Using single-sample Gene Set Enrichment Analysis (ssGSEA), the team identified four "core" cell types—hematopoietic stem cell progenitors (HSC/Prog), granulocyte-macrophage progenitors (GMP), myelocytes, and neutrophils—that were most significantly associated with the disease state.
- Gene Filtering and Signature Development: 1,594 marker genes from these core cells were cross-referenced with bulk RNA-seq data from the GSE114868 dataset. Through Univariate Cox regression and LASSO (Least Absolute Shrinkage and Selection Operator) analysis, the researchers narrowed the list to a 5-gene signature: CALR, KDM1A, SUCNR1, TMEM220, and ADM.
Key Insights: The Inflammation Paradox and Core Drivers
A central finding of the study is the "Inflammation Paradox." High-risk patients, as defined by the 5-gene signature, exhibited significantly higher ImmuneScores and StromalScores. While a high immune presence is often associated with better outcomes in solid tumors, in the context of AML, a pro-inflammatory, "hot" bone marrow microenvironment appears to support leukemia progression and contribute to drug resistance.
The prognostic power of this signature is deeply rooted in the biology of specific progenitor cells, particularly GMPs and HSCs. This suggests that the inflammatory signature reflects the primitive, stem-like state of the leukemia cells, which are often the most resistant to therapy.
Results: Clinical Utility and Drug Sensitivity
The 5-gene model successfully stratified patients into high- and low-risk groups across multiple independent datasets, including TCGA_LAML and GSE106291. High-risk patients consistently showed significantly shorter overall survival (OS). The model's predictive accuracy, measured by the Area Under the Curve (AUC), remained consistently above 0.6 for 1- to 4-year survival.
Beyond prognosis, the model also predicted drug sensitivity. High-risk patients showed increased sensitivity to agents like Vinblastine and Paclitaxel. Notably, the expression of KDM1A—a key component of the signature and a known epigenetic regulator—correlated with sensitivity to Vincristine and Bortezomib, suggesting potential pathways for risk-stratified therapeutic interventions.
Limitations and Future Directions
While the results are promising, the authors note that the initial single-cell discovery was based on a relatively small cohort of 11 patients. Furthermore, as a retrospective study based on existing datasets, prospective clinical trials are required to validate the model's utility in real-world clinical decision-making. Future research should also focus on the specific molecular mechanisms by which these five genes modulate the inflammatory response within the AML niche.
Summary updated