Patients with de novo acute myeloid leukemia (AML) typically come to clinical attention with symptoms of bone marrow failure in the absence of any prior hematologic condition. In many such patients, the diagnosis is sudden and unexpected, without any significant prodrome. The generally poor prognosis of AML and the decades-long impasse in generating effective therapeutic strategies other than conventional antracycline and cytarabine-based (“3+7”) induction chemotherapy has spurred intense investigation of its molecular and cellular origins. Pioneering work by numerous investigators led to an early understanding of the genetics of AML, particularly the identification of recurrent chromosomal abnormalities and mutations in the genes FLT3 and NPM1. However, it is only with the recent implementation of massively parallel next-generation DNA sequencing that the full spectrum of mutations in AML has been defined. These genome and exome resequencing efforts determined that there are five coding mutations on average in an individual case of de novo AML. This observation raises the important question of how these multiple mutations accumulate in a single lineage of cells?
More than a decade ago, Weissman and colleagues hypothesized that since hematopoietic stem cells (HSCs) are the only long-lived self-renewing cells in the myeloid lineage, mutations must be serially acquired in clones of HSC.1 In 2012, we reported direct evidence supporting this model from studies of rare, residual HSC present at the time of AML diagnosis.2 In a small cohort of patients, we demonstrated that residual HSCs could be prospectively isolated from diagnostic AML samples. Through targeted deep-sequencing, HSCs were found to harbor some, but not all, of the mutations present in the same patient’s leukemia cells. Moreover, through analysis of single cells, we showed that mutations were serially acquired in successive clones of HSCs, formally providing proof of the model described above. We termed these cells pre-leukemic HSCs that harbored pre-leukemic mutations.
During the last year, this work was extended further by studies from the laboratory of Dr. John Dick and our own group.3,4 Dr. Dick and colleagues focused on AML cases harboring mutations in DNMT3A and identified this as an early mutation occurring in pre-leukemic HSCs that were capable of contributing to lymphoid and myeloid progeny in patients, both at the time of diagnosis and in clinical remission. They further showed that these DNMT3A-mutant HSCs outcompeted normal HSCs in xenotransplantation assays, establishing them as pre-leukemic. Finally, they reported evidence that IDH2 mutations could also be identified in pre-leukemic HSCs. Our own work investigated patterns of mutational acquisition in AML determining that early pre-leukemic mutations occur primarily in genes that regulate the epigenome, while later mutations are enriched for genes involved in proliferative signal transduction. Like Dr. Dick and colleagues, we also demonstrated that pre-leukemic HSCs persist in remission where they contribute to normal lymphoid and myeloid progeny. Finally, we showed that there are multiple clonal paths to relapse, including the possible acquisition of novel mutations by pre-leukemic HSCs.
Collectively, these studies established the existence of pre-leukemic HSCs that undergo clonal evolution resulting in de novo AML and raise a number of important clinical issues. First, there is currently great interest in developing targeted therapies against specific AML mutations; however, this approach is limited by the possibility of leukemic subclones that do not carry that specific mutation. Pre-leukemic mutations are likely to be contained in all AML cells and are therefore the best therapeutic target. Second, the possibility that pre-leukemic HSCs may acquire novel mutations that lead to relapsed disease suggests that curative therapies will not only need to target frankly leukemic cells but also pre-leukemic HSC. Third, the use of highly sensitive sequencing methods to detect specific mutations as a measure of minimal residual disease (MRD) is an area of active biomarker research; however, as pre-leukemic HSCs persist in clinical remission, detection of such pre-leukemic mutations may lead to false-positive assessments of MRD. Finally, the existence of clinically silent pre-leukemic HSC raises the possibility of early detection prior to the development of frank AML. In fact, several large population-based studies have recently been published reporting that mutations found to be recurrent in hematologic malignancies, including DNMT3A and TET2, can be detected in the peripheral blood of patients with no history of hematologic disease.5,6 Indeed, long-term follow-up of such individuals indicates that they have a much higher future risk of developing a hematologic disease. Thus, it is possible that population-based screening might identify such individuals, eventually paving the way for AML prevention strategies.
