Identification of Prognostic Markers in Argentinian Acute Lymphoblastic Leukemia Pediatric Patients Through Transcriptome Analysis: Differential Gene Expression, Mutational Status and Immune Microenvironment

Abstract

PURPOSE Acute lymphoblastic leukemia (ALL) is the most incident pediatric cancer. Although considerable progress has been made on treatment efficacy and survival rates, 15%-30% of patients relapse and/or die. We aimed to identify molecular profiles and microenvironment makeup in leukemic bone marrow samples that could help to better predict disease outcome. METHODS We performed RNA-seq on bone marrow samples from pediatric ALL patients at diagnosis (n = 37). Patients were recruited under a multi-center clinical protocol in Argentina (median follow-up: 31 months). We analyzed differential gene expression, gene set variations, mutations in candidate genes, and fusion genes among clinico-pathological features. The abundance of immune-cell populations was inferred by digital cytometry (MIXTURE) and a “cytolytic score” based on the expression of five genes specific to cytotoxic cells. RESULTS We detected: (1) 37 differentially expressed genes (DEG) between poor vs good responders to prednisone; (2) 71 DEG between high vs standard-risk groups; (3) 13 DEG between patients who died/relapsed vs those who did not; and (4) 35 DEG between patients that developed severe therapy-related acute toxicity vs those who did not (|log2FC|> 1; P.adj < .05). We observed that 15%-30% of the DEG corresponded to lncRNAs. We found six differentially expressed pathways relevant to cancer and leukocyte biology among risk groups ( P < .01). We identified 17 mutations and three fusion genes in 44% of patients; the presence of these mutations shortened the relapse-free survival (Cox- P-val = .006). High cytolytic score was associated with activated CD8+T cells, immune cell trafficking, and bone marrow niche signaling genesets, suggesting potential candidates for immunotherapies. Higher CD8+T-cell/NK at diagnosis was marginally associated with worse event-free survival (hazard-ratio = 5.39; Cox- P-val = .08). CONCLUSION Transcriptomic sequencing allowed the analysis and integration of multiple molecular features that might improve pediatric ALL outcome prediction.