ARTICLE doi:10.1038/nature09650 Genetic variegation of clonal architecture and propagating cells in leukaemia Kristina Anderson 1 , Christoph Lutz 2 , Frederik W. van Delft 1 , Caroline M. Bateman 1 , Yanping Guo 2 , Susan M. Colman 1 , Helena Kempski 3 , Anthony V. Moorman 4 , Ian Titley 1 , John Swansbury 1 , Lyndal Kearney 1 , Tariq Enver 2 { & Mel Greaves 1 Little is known of the genetic architecture of cancer at the subclonal and single-cell level or in the cells responsible for cancer clone maintenance and propagation. Here we have examined this issue in childhood acute lymphoblastic leukaemia in which the ETV6–RUNX1 gene fusion is an early or initiating genetic lesion followed by a modest number of recurrent or ‘driver’ copy number alterations. By multiplexing fluorescence in situ hybridization probes for these mutations, up to eight genetic abnormalities can be detected in single cells, a genetic signature of subclones identified and a composite picture of subclonal architecture and putative ancestral trees assembled. Subclones in acute lymphoblastic leukaemia have variegated genetics and complex, nonlinear or branching evolutionary histories. Copy number alterations are independently and reiteratively acquired in subclones of individual patients, and in no preferential order. Clonal architecture is dynamic and is subject to change in the lead-up to a diagnosis and in relapse. Leukaemia propagating cells, assayed by serial transplantation in NOD/SCID IL2Rc null mice, are also genetically variegated, mirroring subclonal patterns, and vary in competitive regenerative capacity in vivo. These data have implications for cancer genomics and for the targeted therapy of cancer. Recent genome-wide scrutiny of cancer cells has revealed extraordinary complexity, with substantial numbers of both potential ‘driver’ and neutral or ‘passenger’ mutations per case 1,2 . Informative though these screens are, they probably reflect predominant or composite genetic landscapes that obscure the existence of subclonal heterogeneity of disease 3 . Intraclonal genetic diversity is a common feature of cancer 4 and is probably, from a Darwinian, natural selection perspective, the essential substrate for clonal evolution, disease progression, relapse or metastasis. Subclonal genetic complexity might also be an important consideration for therapeutic targeting. Furthermore, if a subset of ‘stem-like’ cancer cells, or, as we refer to, propagating cells, are the basis of sustained clonal expansion and disease progression 5 then, in principle, they should be genetically diverse if selection and passage through evolu- tionary bottlenecks is to occur. Identifying intraclonal genetic architecture requires genetic scrutiny of single cells or clonal foci, and there are limited examples of this so far 6 ; nevertheless, they testify to the existence of significant heterogeneity. The genetic diversity of cancer propagating cells is, as yet, unexplored. We elected to address this issue in lymphoblastic leukaemia. The substantial advantage of this cancer, in addition to its amenability to single-cell analysis, is that it is minimally deranged or unstable, genetically, and the broad, temporal sequence of genetic events is known. For the B-cell precursor subset of childhood acute lymphoblastic leukaemia (ALL) with ETV6–RUNX1 fusion studied here, the latter genetic lesion is predominantly a prenatal and pre- sumed initiating event 7 . It is coupled with a modest number (3–6) of recurrent, genomic copy number alterations (CNA) 8 . These accrue as secondary and, most likely, postnatal lesions 9 in genes that, predomi- nantly, regulate the cell cycle or B-cell differentiation 8 . Subclonal diversity of genotypes in ALL We initially selected 60 cases of ETV6–RUNX1-positive ALL and in which ETV6 was also deleted (15–85% of cells) as detected by fluorescence in situ hybridization (FISH). Of these, 30 were further selected (see Supplementary Table 1) that also had (by FISH) deletion of PAX5 (n 5 15) or CDKN2A (also called p16)(n 5 12) or deletions of both PAX5 and CDKN2A (n 5 3) in at least 10% of cells. All 30 cases were then scrutinized using a multiplexed combination of distinctive fluorochrome-labelled bacterial artificial chromosome (BAC) probes. Two-hundred cells with the ETV6–RUNX1 fusion signal (that is, the reference founder mutation present in all leukaemic cells) were evaluated for each case and each individual cell designed an allele status (that is, mono- or bi-allelic deletion) for ETV6 and PAX5 (three colour) or ETV6 and CDKN2A (three colour) or ETV6, PAX5 and CDKN2A (four colour). The use of an ETV6–RUNX1 probe also allowed us to detect duplication of the fusion gene (in 15 out of 30 cases) or an extra copy of chromosome 21q (via RUNX1 signal copy number; in 21 out of 30 cases). The latter is a common genetic abnormality in ALL and an assumed driver event 10 . Cutoff levels (%) for scoring genetically dis- tinctive subclones were determined using normal blood controls and varied depending upon probe set combination. A threshold was set at 2% for cells with a single CNA (in addition to ETV6–RUNX1 fusion) and 1% for cells with two or more CNA (see Methods and Sup- plementary Table 2). Enumeration of CNA in individual cells in reference to ETV6– RUNX1 fusion—the universal marker of all the leukaemic cells— allowed us to identify distinctive genetic signatures of subclones and their relative frequencies. From this we could infer the most likely evolutionary or ancestral relationships between the subclones and derive a clonal architecture. The genetic architectures that were observed were very diverse. The simplest of these genetic architectures that we identified (in 6 cases) involved two or three subclones that could be aligned in a linear sequence (Fig. 1a); however, these were cases with the lowest complement of CNA (only two or three out of the possible total of the seven pre-selected) in addition to the ETV6–RUNX1 fusion. All 1 Section of Haemato-Oncology, The Institute of Cancer Research, Sutton SM2 5NG, UK. 2 MRC Molecular Haematology Unit, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS, UK. 3 Paediatric Malignancy Unit, Great Ormond Street Hospital & UCL Institute of Child Health, London WC1N 3JH, UK. 4 Leukaemia Research Cytogenetics Group, Northern Institute for Cancer Research, Newcastle University, Newcastle upon Tyne NE1 4LP, UK. {Present address: University College London Cancer Institute, London WC1E 6BT, UK. 356 | NATURE | VOL 469 | 20 JANUARY 2011 Macmillan Publishers Limited. All rights reserved ©2011