Next-Generation Sequencing and Detection of Minimal Residual Disease in Acute Myeloid Leukemia: Ready for Clinical Practice?

Abstract

Acute myeloid leukemia (AML) represents a heterogeneous disease, both with respect to molecular pathogenesis and clinical outcome. Although dose-intensive chemotherapy and allogeneic stem cell transplantation have improved outcomes in AML, there remains significant heterogeneity in clinical outcome such that approximately 20% of patients are cured with existing therapies, 20% have therapy-refractory disease from the time of diagnosis, and 50% relapse and die from refractory disease after an initial response to leukemia therapy.1-4 The challenge is how to best determine prognosis and identify which patients will have a substantive chance of cure, and which patients will likely relapse or present with refractory disease. Current standard of care uses clinical, cytogenetic, and molecular factors for risk stratification; however, there remains a pressing need for better approaches to prognostication in AML. In recent years large sequencing studies of AML have identified a spectrum of recurrent somatic alterations that contribute to tumor initiation and maintenance.5 Most importantly, this knowledge can be used to improve risk stratification, and prognostic algorithms that incorporate focused molecular profiling have been developed and validated in large AML patient cohorts.6-11 However, many questions remain regarding the use of molecular profiling in clinical practice in AML and in other malignancies. First, is there value in the use of more extensive molecular studies, such as the use of large molecular panels or exome or genome sequencing? Second, does intrapatient heterogeneity affect prognosis or the response to AML therapies? Third, and most important, can dynamic analysis of the AML genome during and after therapy be used to inform prognosis? In this issue of JAMA, Klco and colleagues address these questions and provide critical insight into the role of next-generation sequencing techniques in the clinical context.12 The authors analyzed diagnostic samples from 71 patients with de novo AML treated at a single center with integrative genomic studies, including whole-genome sequencing or whole-exome sequencing, RNA sequencing, microRNA sequencing, and methylation arrays. The 71 patients were divided into 3 groups: patients with refractory disease with relapse within the first 6 months after the start of therapy, those with relapse within 6 to 12 months, and those with remission for more than 12 months. The authors did not observe significant differences between these groups with respect to the numbers of coding or total genomic mutations. Moreover, the presence of specific mutations in the coding sequence of known or novel genes, or of select noncoding mutations, was not associated with clinical outcome. The authors concluded that genomic profiling done at a single time point (eg, at diagnosis) did not sufficiently predict outcome in AML. Although there are important limitations to the accuracy and robustness of prognostic models based on molecular profiling,13 Klco and colleagues analyzed a relatively small, single-institution cohort, and as recognized by the authors, this factor significantly limits their ability to inform the use of molecular profiling as a prognostic tool in AML. It will be critical to perform similar, high-quality studies in larger AML cohorts, ideally from multicenter clinical trials with a homogeneous therapeutic approach, to ascertain whether whole-genome or whole-exome sequencing or other state-of-the-art genomic approaches at AML diagnosis can better predict outcome than current prognostic schema that use limited mutational studies, cytogenetic analyses, and clinical parameters. Klco and colleagues also sought to address a previously unanswered question: Can serial studies of mutations during therapy inform outcome in AML? The authors used 2 different approaches to address this question: enhanced exome sequencing with high-coverage sequencing of known leukemia genes for high-quality banked samples and a targeted, short amplicon sequencing platform for samples preserved in paraffin. This allowed the investigators to accurately and quantitatively follow mutational burden before, during, and after therapy in patients with AML in their cohort. The authors used these platforms to assess whether morphologic complete remission was associated with a clearance of AML mutations (variant allele frequency ≤2.5%) and the relationship between molecular responses and clinical outcomes. The authors assessed 50 patients in morphologic remission on day 30, which allowed them to identify different groups of patients. One group of patients (26 of 50 patients) showed a complete eradication of somatic mutations detected in the diagnosis sample at the time of clinical remission. By contrast, a second group of patients (24 of 50 patients) retained at least 1 mutation present in the diagnosRelated article page 811 Opinion