Next-Generation Sequencing Analysis of 23 Therapy-Related Acute Myeloid Leukemia Transcriptomes

Abstract 850

Therapy-related acute myeloid leukemia/myelodysplastic syndrome (t-AML) is a fatal complication that occurs in up to 5% of patients treated with chemotherapy or radiation. As more cancer patients achieve longer remissions, t-AML has increased in incidence, yet remains refractory to current therapy, with a median survival of 8–10 months. T-AML is thought to be a direct product of genetic mutations induced by alkylating agents, topoisomerase-II inhibitors, and radiation. Two main subtypes of t-AML have been described that differ in their morphology, cytogenetic karyotype, and prognosis: alkylator-agent-induced and topoisomerase-II inhibitor-induced. To identify expressed genetic variants that distinguish the two subtypes of t-AML, we have sequenced the leukemic mRNA of 23 t-AML samples using Illumina paired-end technology, yielding over 1 billion base pairs of aligned sequence per sample. The alkylator agent-related group is comprised of 13 samples with deletions of part or all of chromosomes 5 and/or 7, or with complex karyotypes. 10 samples with other karyotypes, including normal karyotype and balanced translocations, constitute the topoisomerase-II-related group. In addition, 13 de novo AML patient samples with a spectrum of cytogenetic abnormalities similar to the t-AML samples have been similarly sequenced. We have found over 13,000 single nucleotide variants (SNVs) per sample. SNP array analysis on 2 samples demonstrates 92% sensitivity and 98% specificity for SNP identification. Phenotypically relevant SNVs are prioritized by population-level allele frequency estimates, evolutionary constraint estimates, and the predicted coding region change. This analysis has yielded approximately 200 rare, predicted deleterious SNVs per sample. Some of these variants include those previously reported to be aberrant in AML (for example, FLT3, P53, RUNX1, PTPN11, CBL, and TET2), validating our analysis pipeline. Putative disease-relevant SNVs are being compared between the two main subtypes of t-AML, as well as between therapy-related and de novo AML by analyzing the frequency of the aberrant gene or biological pathway within the groups. We have found that SNVs frequently occur in the same gene or pathway in multiple samples. These findings are being validated within tumor and paired normal DNA, as well as in a larger cohort of 100 t-AML leukemia samples and 50 de novo AML leukemia samples using single-base primer extension assays. Furthermore, comparisons of transcript expression level and isoform usage are being made between groups and cytogenetic types. This information will allow the identification of pathways altered in t-AML, correlation of mutations with clinical outcome, identification of potential biomarkers, and generation of candidate targets for leukemia therapeutics.