Genomic Landscapes in Elderly AML

Acute Myeloid Leukemia (AML) in patients over the age of 60 (elderly AML) remains a clinical challenge. Despite advances in the field, outcomes for elderly AML patients remains poor regardless of treatment administered. We hypothesize that elderly AML is characterized by the acquisition of somatic mutations that enhance leukemogenic potential and associate with worse clinical outcomes. To test this hypothesis, we determined the landscape of somatic mutations in de novo AML patient samples from the phase III prospective ECOG-ACRIN Cancer Research Group clinical trial 3999 (E3999; NCT00046930; ages 60-93). DNA was isolated from patient specimens processed to isolate blast-enriched cells (CD3 and CD19 depletion using Miltenyi magnetic beads) and patient-matched CD3-positive cells (flow cytometry) as germline controls. Exome capture was performed in paired disease and germline samples from 210 patients. Samples were sequenced on a HiSeq 2500 obtaining a median coverage of 150x bases per position. Low-coverage whole genome sequencing was used to identify copy number alterations (CNAs) in 202 of the samples. Indexed Illumina libraries were sequenced to a coverage of approximately 3 million reads per sample.

The Cancer Genome Project pipeline guidelines were utilized to determine somatic mutation events in the whole exome data. We aligned the reads with BWA mem, determined somatic copy number aberrations using CNVKit, short insertions and deletions (including FLT3 Internal Tandem Duplication) using Pindel, and substitutions using CaVEMan. We then used classical heuristics including strand bias, mapping quality and high rate of presence in the ExAC database to remove artifacts and non-somatic variants. Finally, we used COSMIC and published data in the literature to triage events into known oncogenic events, unknown but likely oncogenic events and variants of unknown significance. CNAs were determined from the whole genome sequencing data by partitioning the human genome into 60 thousands bins (i.e. resolution of 50kb) and counting uniquely mapped reads within the genomic bins/intervals. Read counts within bins were subsequently processed and transformed into integer copy number states.

We found a range of 1 to 37 mutational events per sample (median of 15), including 0 to 7 oncogenic or likely oncogenic events (median of 2). Most commonly mutated genes were NPM1 (25.7%); epigenetic modifier genes: TET2 (24.7%), DNMT3A (20.5%), ASXL1 (12.9%), and IDH2 (11.4%); FLT3 (20%); spliceosome complex genes: SRSF2 (19.5%), U2AF1 (10%) and STAG2 (10%); RUNX1 (15.7%); TP53 (12.4%), and NRAS (11.9%). Comparing our AML cohort to the mutation landscape reported by Papaemmanuil et. al., NEJM 2016 (age range: 18 to 84 years), we found that the elderly group was enriched in oncogenic mutations in the SRSF2, TET2, U2AF1 and ASXL1 and depleted for IDH1 and FLT3 (Fisher exact test adjusted p-value <0.05) genes. Furthermore, the elderly AML patient cohort was enriched in spliceosome/cohesin modifier and TP53-complex karyotype subtypes (Fisher exact test: p<2.8*10^-8 and p<10^-5 respectively).

Detailed analyses of the mutational spectra in the E3999 clinical trial suggests enrichment for mutations in epigenetic modifiers, spliceosome factors, and cytogenetic events associated with poor prognosis in AML. These may be underlying the aggressive phenotype in this age range and contribute to disease pathogenesis.

Authors contributed equally: Ari Melnick and Ross Levine

Co-corresponding authors: Maria Kleppe and Francine E. Garrett-Bakelman