Mutational Landscape of Pediatric Acute Myeloid Leukemia: A Report of the AML-BFM Study Group with a Targeted NGS Approach in 525 Patients Integrating De Novo, Relapsed and Secondary AML

Background:

Recent studies in adults revealed the complexity of the genetic profiles of AML blasts that are used for fine-tuning of the risk-adapted treatment regimens. It is known, that recurrent translocations detected at diagnosis are largely preserved at relapse in the AML blasts. However, the molecular landscape and clonal evolution of small indels and missense alterations in the blasts assessed by NGS has not been well analyzed yet for pediatric AML patients, neither for primary and relapsed children and adolescents nor for patients with secondary AML. Here, we present the implications of our amplicon-based targeted sequencing efforts in a large cohort of pediatric patients with de novo AML or secondary AML at diagnosis and relapse.

Patients and Methods:

In total, samples from 525 AML patients (age <18 years) registered by the AML-BFM study group in Germany from 11/2000 until 04/2019 were analyzed here (de novo AML, n=449; secondary AML, n=13; first relapse n=63). Paired diagnostic and relapse samples were available from 45 patients. Sequencing was performed using the TruSight Myeloid panel (Illumina) on a MiseqDX device detecting somatic alterations in 54 genes associated with myeloid malignancies. On average, we achieved a 500X coverage in ≥95% of the target regions. We excluded intronic, synonymous and variants with an allele frequency below 5% and a read depth below 50 reads. Five-year probabilities of overall survival (pOS) and event-free survival (pEFS) were determined by Kaplan-Meier analysis in SPSS (IBM SPSS Statistics version 25).

Results:

In this large population of 525 pediatric patients, 632 mutations were detected in 40 out of 54 genes (74%). In total, 448 (80%) mutations in 9 genes (ASXL1, CEBPA, FLT3, KIT, KRAS, NPM1, NRAS, PTPN11 and WT1) were present in ≥3% of de novo AML patients. In ≥3% of relapsed patients, 48 (78%) mutations have been identified in 10 genes (CEBPA, FLT3, KIT, NPM1, NRAS, PHF6, PTPN11, RUNX1, TP53 and WT1). Although patient numbers were limited for secondary AML, data indicate a largely overlapping profile of affected genes. The median number of mutations per patient was low (n=1) in all 3 subgroups. However, 20% of the patients with de novo AML and 12% of relapsed patients had ≥ 3 mutations. Patients with more than 4 mutations have been only identified in de novo AML. The number of mutations per patient did not have a prognostic impact, analyzed in an uniformly treated subgroup of patients (pEFS: p=0.358; pOS: p=0.261). The comparison of paired samples (diagnosis/relapse) revealed the presence of three categories: Mutations only present at diagnosis for de novo AML samples, mutations that persisted through relapse and mutations that were newly gained at relapse. 39% of initially detected mutations (28 out of 72) persisted in the respective relapse samples with the highest stability in CEBPA, FLT3, KIT and WT1. In contrast, 61% (44 out of 72 initial mutations) were not detected and 45% (25 out of 54 relapse mutations) have been gained within the disease progress. The genetic profile changed from de novo AML to relapse in most patients (n = 34; pOS= 41 ± 15%). However, it has been found to be stable in 24% (n = 11; pOS 21±17%; p= 0.49) of the patients.

Conclusions:

This analysis elucidates the mutational landscape of both pediatric de novo and secondary AML as well as the genetic evolution of pediatric relapsed AML. Very high numbers of mutations per patient only occurred in de novo AML patients. In contrast to the assumed stability of recurrent translocations, the mutational landscape changed during disease progression in most of the patients. Whether these altered clones detectable at relapse were already present at the initial diagnosis, albeit at low frequency, or occurred de novo during treatment is very challenging to answer in pediatric AML due to the often very limited sample sizes. Either new single-cell sequencing approaches and/or analysis after expanding the cells in xenotransplantation models will provide more insights. Of note, modification of the sequencing panel would be beneficial for future studies due to the fact, that 26% of genes included were not affected by mutations.