Molecular Profiling of High-Risk Pediatric Acute Myeloid Leukemia

Background: Several landmark genomic profiling studies have dramatically advanced our understanding of the origin, progression and clonal evolution of adult acute myeloid leukemia (AML) and directly impacted clinical care. However, very little is known about the mutational landscape of pediatric AML, a distinct entity that shares few genetic and clinical characteristics with adult AML. To investigate potential drivers of high-risk pediatric AML, comprehensive genomic profiling was performed on high-risk AML samples as part of a prospective clinical next-generation sequencing program.

Methods:Samples obtained from patients with known high-risk features at diagnosis or with refractory or relapsed AML were selected for molecular profiling. Comprehensive testing included whole-exome sequencing (WES) of matched tumor (bone marrow or chloroma tissue) and normal tissue (peripheral blood or buccal swab) samples and transcriptome analysis (RNAseq). Targeted sequencing of 467 cancer-associated genes was used when tumor tissue was limited. Sequencing was performed on Illumina's HiSeq 2500 with 150X and 500X average coverage for WES and targeted sequencing, respectively. Variants were filtered to select alterations in cancer-related genes or genes relevant for patient care.

Results:Fifteen patients with AML (mean age 7.7 yrs; range 0.75-19 yrs) met high-risk criteria (high-risk features at diagnosis = 4, relapsed disease = 8, refractory disease = 3) and were selected for profiling. WES and RNAseq were performed on 11 samples, WES only on 3 samples and targeted DNA sequencing on 1 sample. The median number of variants was 60 (range 14- 5950) per case. After filtering, 54 mutations were identified in 35 genes with a mean of 3.6 mutated genes per patient sample (range 0-14); two samples only carried a fusion gene with no other genetic alterations. At least one driver genetic alteration was detected in each patient sample. Thirteen samples carried mutations in at least one gene known to be altered in AML (e.g. IDH1, WT1, TP53, NRAS) (mean, 2; range, 1-6) and 5 samples carried novel mutations in 15 genes not previously implicated in AML (e.g. CARD9, CHD9, Axin1). Mutations in 11 AML related genes were detected in more than one sample including NRAS in 4, TP53 in 3 and KRAS, PTPN11 PHF6, JAK3 in 2 samples each; genes not previously implicated in AML were only mutated in single patients. Of note, mutations in genes encoding members of the RAS pathway occurred in 60% of cases (9/15 samples). RNAseq identified gene fusions in 7/11 samples (63%). Four fusions involving KMT2A and core binding factor genes were also detected by FISH while three fusions were detected by RNAseq only: NUP98-NSD1 in two patients and CBFA2T3-GLIS2 in one patient. Samples carrying driver gene fusions had the lowest number of mutated genes (0-1) compared to samples lacking a gene fusion (1-5 mutated genes), with one exception of a patient with history of infant ALL who later developed KMT2A-AFF1AML with the highest number of mutated genes (n=14). There was no correlation between the number of mutated genes and age, clinical characteristics, initial risk classification at diagnosis or intensity of therapy prior to sequencing.

Conclusion:Our study provides an initial overview of the genetic alterations that characterize high-risk, chemo-resistant pediatric AML. Analysis of the data highlights the overall low genetic complexity of high-risk AML despite the aggressive clinical behavior and exposure to intense chemotherapy, including stem cell transplant. Of interest, similar to adult AML, we found that mutations leading to aberrant activation of the RAS pathway were also very frequent in our cohort of pediatric high-risk AML, while genes typically mutated early in the process of leukemogenesis in adult AML, such as NPM1, DNMT3A, FLT3, IDH1, IDH2 were not affected. Such findings suggest that distinct, age-specific mechanisms of leukemogenesis might exist. Furthermore, our data also highlights the important role of RNA sequencing in complementing current standard diagnostic tools, allowing the identification of driver fusion genes in samples for which no other driver event is detected. Larger studies, preferably including diagnostic samples and utilizing broader approaches, are needed to better understand the mechanisms responsible for the initiation and progression of childhood AML.