The early development of acute leukaemias is assumed for the most part to be clinically silent and transient. But it now seems that ancestral precancerous cells are identifiable and persistent. See Article p.328
Aggressive leukaemias often present clinically out of the blue, without previous indications of cancer. But evolutionary models of cancer development posit a time-ordered, stepwise process involving the accumulation of mutations, proliferation of mutated cells into expanded clonal populations and selection of the fittest cells1. Such models imply that any seemingly sudden case of cancer will have arisen from 'silent' precursor cells that have no clinical impact. In this issue, Shlush et al.2 (page 328) provide compelling evidence that the early-stage cells of acute myeloid leukaemia are not completely outcompeted and rendered extinct by their more aggressively cancerous and numerous progeny, but instead persist and show defined genetic and functional properties.
The cancerous myeloid cells (a subset of white blood cells) of patients with acute myeloid leukaemia (AML) often have a mutation in the gene encoding the DNA-methyltransferase enzyme DNMT3a. In two studies of patients with DNMT3a-mutated AML — one involving 4 patients and the other 17 — Shlush and colleagues made the unexpected observation that, in 15 patients, the DNMT3a gene carried the same mutation at a low rate in T cells (a type of immune cell that belongs to the lymphoid system) in the blood. Strikingly, however, the T cells lacked other alterations present in the leukaemic cells of the same patients, including mutations in the gene NPM1. The DNMT3a mutation was also detected at variable frequency in other immune cells (B and NK cells) at the time of AML diagnosis. Finding this mutation in non-myeloid cells hinted at its occurrence in a precursor cell that gives rise to blood cells of both the lymphoid and myeloid lineages.
The frequency with which DNMT3a and NPM1 were mutated in leukaemic cells was equally high in all but two patients studied. This suggests that both mutations were probably present in a 'founder' clone population, from which the leukaemic population expanded. From this starting point, the authors astutely deduced that the first cell to acquire a leukaemic 'driver' mutation in DNMT3a in these patients was a haematopoietic stem cell (HSC) — the precursors of the lymphoid and myeloid lineages — and that descendants of this cell persisted as an expanded, competitive clonal population (Fig. 1). However, whether mutation of DNMT3a is by itself sufficient to both initiate and sustain the growth of a clonal population before the onset of leukaemia is uncertain, because Shlush and colleagues analysed only specific genes, and so the mutational complexity of the cancers was probably underestimated3. Although mutation of NPM1 can initiate AML in mouse models of the disease4, and is often seen in humans with AML, it seems to be a secondary alteration in most cases of AML that carry a DNMT3a mutation.
These data agree with the results of functional experiments5 indicating that DNMT3a normally promotes the differentiation of HSCs into other cell lineages at the expense of the HSCs' self-renewal. The mutation in DNMT3a in AML results in a loss of the enzyme's catalytic activity6. But the mutated protein can repress the function of the normal protein encoded by the non-mutated copy of DNMT3a, and this is expected to increase self-renewal7. Shlush et al. found mutant DNMT3a in a relatively high percentage of HSCs from each AML patient (up to 30%), suggesting a high degree of competitive self-renewal of the DNMT3a-mutant cells. However, they also observed some normal differentiation of the DNMT3a-mutant HSCs into lymphoid and myeloid cell lineages.
The authors confirmed this functionally by repopulating immune-cell-deficient mice with blood cells from two patients with AML. In around 75% of these mice, multiple cell lineages were established. Ten of 12 such mice were studied further, and were found to have a high frequency of the DNMT3a mutation but to have normal NPM1, indicating a functional dominance of the DNMT3a-mutated HSC population. Only a minority of the mice had a myeloid leukaemic population with mutations in both DNMT3a and NPM1.
Shlush and co-workers further showed that DNMT3a-mutant HSCs and their differentiated progeny persisted in patients' blood even when AML was in remission following chemotherapy, indicating that at least some of these pre-leukaemic ancestral cells were resistant to treatment. This may be because a high fraction is quiescent — as in other types of leukaemia and cancer. Collectively, the data highlight the persistence of benign pre-leukaemic clones. Other data point to a similar scenario, including the observation8 of another mutation often found in AML, the ETO–RUNX1 fusion, in HSCs and B cells. A further study9 reported the identification of ostensibly normal HSCs, myeloid and lymphoid cells that had a subset of the mutations present in the immature AML myeloid cells of the same patient. Interestingly, and in contrast to Shlush and colleagues' findings, that study provided evidence of a linear, or evolutionary, order of several mutations in the HSC-derived pre-leukaemic cells.
More generally, these studies2,8,9 contribute to an emerging portrait of the complexity of clonal evolutionary pathways that lead to AML3,10,11 (Fig. 1). The premalignant phase, arising predominantly in HSCs, can be highly variable, both clinically and biologically, depending on the mutations, or combination of mutations, involved11 and their functional impact. Common secondary mutations, such as those in NPM1, may result in more stringent arrest of differentiation, or more vigorous cell proliferation — steps that lead to acute-phase disease. Secondary mutations, in contrast to the founder (or very early) alterations, often seem to arise in more-lineage-committed myeloid progenitors, as demonstrated for NPM1 by Shlush et al., and previously in chronic myeloid leukaemia12. A study13 of acute lymphoblastic leukaemia has similarly suggested that secondary mutations arise in more-differentiated progeny cells, and this might be expected to be a feature of other types of cancer.
As Shlush and colleagues point out, their data have several clinical implications. One is that, as a persistent and likely founder, the DNMT3a mutation could be both a therapeutic target and a marker for tracking residual disease. By contrast, therapies targeting secondary mutations, such as NPM1, will probably have only transient success. Another possibility suggested by the persistence of the pre-leukaemic clones and of mutant HSCs in remission is that these cells provide a cellular reservoir for relapse. The observation that some patients who present with DNMT3a–NPM1 double-mutant AML but, following treatment, relapse with cancers that retain only the DNMT3a mutation, is compatible with this idea14. Finally, it would be of interest to determine how often DNMT3a-mutant clones arise from HSCs in ageing adults and, where such clones emerge, how frequently the evolutionary transition to AML occurs, and over what time frame.
About this article
Journal of Translational Medicine (2014)