Gene Expression Profiling in Pediatric AML Patients with Febrile Neutropenia during Induction Therapy

Background: Febrile neutropenia (FN) is a common complication in acute myeloid leukemia (AML), affecting over 90% of patients with significant number of patients experiencing sever FN. One of the clinical challenges is the fact that 5-10% of FN cases progress to fatal sepsis, which can lead to death (reviewed PMID: 38961525). This complication is particularly severe in pediatric AML patients during induction 1 treatment, as the intensive high-dose regimens induce prolonged neutropenia. This extended period of neutropenia, significantly increases the risk of infectious complications and early mortality. Despite the high incidence and severe consequences of FN in pediatric AML, current risk stratification methods are limited in their ability to predict which patients are at highest risk for developing severe FN (grade 3 or more). Identifying patients at increased risk for severe FN could allow for targeted prophylactic strategies, for managing severe FN and reducing related morbidity, mortality, and healthcare burden. Towards this objective, we designed this study to investigate if gene expression profiles at diagnosis can predict the occurrence of severe FN (≥ grade 3) in pediatric AML patients during induction 1.

Methods: RNA sequencing (RNAseq) data was generated from bone marrow and peripheral blood samples collected from 1,089 pediatric AML patients enrolled in the AAML1031 trial (NCT01371981). This data was obtained through the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database. Among these patients, 836 had both RNAseq and toxicity data available during Induction 1, and they were included in this study. With respect to FN, 206 patients developed ≥ grade 3 during Induction 1 and 447 patients did not. The grades for febrile neutropenia were defined using the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Gene expression data were processed and batch-corrected using ComBat. After excluding genes with a mean expression level below 1 across all samples, the expression 14,000 transcripts were log2-transformed before any subsequent analysis. Differential gene expression analysis between patients with or without ≥ grade 3 FN was performed using the limma package in R software version 4.4.1

Results: Forty-three genes significantly differed in the expression levels between patients who did or did not experienced ≥ grade 3 FN (p<0.001) during induction 1. These included genes with significant roles in immune response, cell cycle regulation, and apoptosis pathways such as: IKZF5, a member of the IKAROS family, which is implicated in hematopoietic cell differentiation and immune regulation. IKZF5 has also been previously associated with hematologic malignancies and aberrant neutrophil chemotaxis; APOE, codes for apolipoprotein E with role in lipid metabolism and immune response regulation. Previous studies have shown it to be associated with infection susceptibility.; NR1D2, a transcriptional repressor involved in regulation genes in lipid metabolism and inflammatory response; CHD1, associated with chromatin remodeling, influencing gene expression involved in immune responses; TMEM41B, has been implicated in autophagy and mechanisms underlying viral infection; HSPA13, a heat shock protein involved in protein folding and stress response, impacting immune system function; SIRT1, which regulates aging, metabolism, and inflammatory responses; TBK1, a key player in innate immunity and antiviral responses, involved in interferon production; EIF2AK2, also known as PKR, crucial for cellular response to viral infection and stress; and REL, part of the NF-kB family, regulating immune and inflammatory responses.

Conclusion: Our exploratory study leveraged the publicly available transcriptomics data for its utilization to identify gene expression signatures predictive of adverse drug events. Ongoing studies are focused on further in-depth investigation of FN within specific treatment arms, investigating gene-expression signatures associated with other clinically relevant toxicities/adverse events in pediatric AML. Overarching goal of this work is to develop a predictive model that could enable personalized prophylactic strategies, reducing morbidity, mortality, and healthcare costs, while improving overall patient outcomes and quality of life.

No relevant conflicts of interest to declare.