Acute myeloid leukaemia: a paradigm for the clonal evolution of cancer?

Abstract

Acute myeloid leukaemia (AML) is an uncontrolled clonal proliferation of abnormal myeloid progenitor cells in the bone marrow and blood. Advances in cancer genomics have revealed the spectrum of somatic mutations that give rise to human AML and drawn our attention to its molecular evolution and clonal architecture. It is now evident that most AML genomes harbour small numbers of mutations, which are acquired in a stepwise manner. This characteristic, combined with our ability to identify mutations in individual leukaemic cells and our detailed understanding of normal human and murine haematopoiesis, makes AML an excellent model for understanding the principles of cancer evolution. Furthermore, a better understanding of how AML evolves can help us devise strategies to improve the therapy and prognosis of AML patients. Here, we draw from recent advances in genomics, clinical studies and experimental models to describe the current knowledge of the clonal evolution of AML and its implications for the biology and treatment of leukaemias and other cancers.

Keywords: Acute myeloid leukaemia; Cancer; Clonal evolution; In vivo models of leukaemia; Mutation.

Fig. 1.

Mutation burden and cancer incidence. (A) Comparison of the mean number of non-coding mutations per genome in tumours of different tissues (raw data from Lawrence et al., 2013). Error bars show the standard error of the mean. (B) UK annual incidence of various malignancies [Cancer Registry Statistics, 2011; (www.ons.gov.uk) and the Cancer Research UK website (www.cancerresearchuk.org/cancer-info/cancerstats/)]. An asterisk (*) signifies that the incidence data refer to the tissue of origin, rather than the specific cancer subtype shown in A (e.g. lung cancer rather than lung squamous cell or adenocarcinoma). CLL, chronic lymphocytic leukaemia; DLBCL, diffuse large B-cell lymphoma.

Fig. 2.

Recurrent mutation groups in de novo AML. Genes recurrently mutated in AML belong to distinct functional groups or pathways. The most prominent functional groups and genes associated with these are listed. The proportion of AMLs with mutations affecting each of these groups is displayed (data obtained from The Cancer Genome Atlas Research Network, 2013).

Fig. 3.

Linear and branching clonal evolution. (A) Linear evolution. Sequential dominant clones (clonal sweep) result in a linear architecture with stepwise accumulation of driver mutations. The final tumour carries all mutations arising during the evolutionary history and overwhelms earlier clones carrying only some of the mutations. (B) Branching evolution. The final leukaemia might be dominated by a single clone, but clones arising through divergent mutational pathways are also evident. Small subclones might fall below the limit of detection, in which case the complexity of branching is underestimated. Smaller fitness effects of mutations and faster acquisition favour branching versus linear evolution. Numerals indicate the number of mutations in cells. Cells carrying identical mutations are represented in the same colour.

Fig. 4.

Common mutations in de novo and secondary AML. A number of clonal blood disorders with a myeloid phenotype are represented. Each of these disorders is characterised by recurrent mutations in specific genes, some of which are shared between several different phenotypes (e.g. TET2). All of these disorders can transform to secondary AML upon acquisition of additional somatic mutations. When AML arises in the absence of an antecedent clonal blood disorder, it is known as primary AML. aCML, atypical CML; CML, chronic myeloid leukaemia; ET, essential thrombocythaemia; IMF, idiopathic myelofibrosis; PV, polycythaemia vera; SM, systemic mastocytosis.

Fig. 5.

Order of acquisition: constraint or opportunity? In many cancers, including AML, specific driver mutations are usually acquired early in the process of clonal evolution, whereas others are acquired late. Here we use the simplified example of two mutations that represent the only two driver events in AML. During the evolution of AML, mutation X (red square) is recurrently acquired first and mutation Y (blue oval) second. This pattern could be due to a strict requirement for a specific order in which mutations are acquired (constraint, panel A) or could simply reflect the statistical likelihood that mutations are acquired in this order (opportunity, panel B). We speculate that the key variables behind the concept of ‘opportunity’ are the number and longevity of cells susceptible to transformation (clonal size, represented here by the number of cells that inherit the mutation) and the speed with which additional mutations are acquired (mutagenesis rate). In panel Bi, mutation X leads to a marked clonal expansion in the progeny of the leukaemia-initiating cell (LIC). In turn, this increases the likelihood of a cell subsequently acquiring mutation Y and the development of leukaemia (solid arrow). Nevertheless, the development of leukaemia is not inevitable (dashed arrow). In panel Bii, mutation Y is acquired first in the putative LIC and does not facilitate the generation of progeny susceptible to transformation, such that the subsequent acquisition of mutation X is unlikely (solid arrow), but not impossible (dashed arrow). In panel Biii, mutation X has a neutral effect on the generation of LIC progeny, but causes accelerated mutagenesis and thus makes the likelihood of subsequent acquisition of mutation Y higher. In panel Biv, mutation Y is acquired first and here it has a neutral effect on initial LIC clonal size, but does lead to subsequent cell loss (e.g. by accelerating senescence), therefore markedly reducing the opportunity for acquiring additional mutations. Again, this eventuality, although unlikely, is not impossible. Finally, in panel Bv, mutation X leads to both clonal expansion and accelerated mutagenesis, making the development of leukaemia very likely or even inevitable. By the same token, a mutation with the opposite effects (i.e. no LIC clonal expansion or enhanced cell loss, and low mutagenesis rate) would make leukaemia very unlikely or impossible.