Molecular Monitoring of Acute Myeloid Leukemia: Patterns at Presentation and Post-Therapy

Background: Given the limitations of targeted molecular assays (such as translocation detection) and flow cytometry for all cases of acute myeloid leukemia (AML), an approach targeting just the most commonly mutated areas of a larger gene set could provide a more cost-effective and broadly applicable approach for residual disease monitoring. Molecular monitoring of response to treatment in AML using a next-generation sequencing panel is appealing as a rapid, comprehensive method that can be employed across the wide diversity of genomic subtypes of AML. In order to assess the potential utility and pitfalls of routine molecular monitoring in AML, we used a gene panel containing 29 myeloid-associated genes to test bone marrow samples from newly diagnosed AML patients

Methods: Bone marrow aspirate samples from 55 consecutive cases of newly diagnosed AML that were submitted for gene studies were assessed, with 20 of these cases also having post-induction samples studied. Sequencing of genomic DNA was performed using a TruSeq custom-designed amplicon method on the Miseq platform (Illumina). Included were the commonly mutated areas of 29 myeloid-associated genes. Results were analyzed using SeqNext software, with somatic mutations being defined as typical frameshift or nonsense alterations, previously characterized missense changes, or novel missense changes with high mutation probability based on read level and predicted structural features. Sensitivity was set at 5%, with some indels showing reduced detection below 10%. Variance in quantitation of mutation calls was assessed using inter-assay comparisons of mutated reads percentage and dilution curves in some cases. Single nucleotide polymorphisms (SNPs) were utilized to assist in normalizing levels of mutations by comparing diagnostic and post-therapy samples.

Results: Among 55 cases of newly diagnosed AML, the number of definitive somatic mutations present at diagnosis ranged from 1 to 5 (mean 2.8 mutations/case) with every case having at least one mutation detected. The most frequently mutated genes were DMNT3A (20.9% of cases) RUNX1 (18.9%), NRAS (18.9%), ASXL1 (16.3%), CSF3R (16.3%) and NPM1 (16.3%). However NRAS (8 cases), TP53 (5), TET2 (3), and WT1 (2) were the most frequently mutated genes with only 1 identified mutation. DMNT3A, RUNX1, NPM1 and FLT3 were commonly mutated genes that were never present as the sole mutation event. Eleven cases (20%) had normalized mutation read levels that were different for individual mutations (beyond assay variance) consistent with multiple or subclonal abnormal hematopoietic populations. Among the 20 cases with post-induction samples for comparison, 9 cases showed complete disappearance of the mutated clone post-induction which correlated with clinical/hematologic remission; 6 cases showed no significant change in the level of the mutated clone post-induction, 1 had a 20-fold reduction in the mutated clone at 6 weeks before complete molecular response at 12 weeks, and 1 showed increase in the mutated clone level with acquisition of a CEBPA mutation. The remaining 3 cases showed more complex patterns, such as disappearance and emergence of newly identified mutations in the followup samples indicating presence of multiple abnormal hematopoietic populations.

Conclusions: Molecular monitoring using a 29-gene panel identified mutation markers in all cases of AML at diagnosis. NRAS and TP53 appear to be important drivers of leukemogenesis in AMLs with low mutation density. In most cases, multiple markers were present and showed parallel changes post-therapy, facilitating straightforward interpretation. However, interpretation of post-treatment results can be complex in the smaller subset of cases that show disappearance and emergence of new mutations or subclonal diversity.