Instruction of haematopoietic lineage choices, evolution of transcriptional landscapes and cancer stem cell hierarchies derived from an AML1-ETO mouse model

The t(8;21) chromosomal translocation activates aberrant expression of the AML1-ETO (AE) fusion protein and is commonly associated with core binding factor acute myeloid leukaemia (CBF AML). Combining a conditional mouse model that closely resembles the slow evolution and the mosaic AE expression pattern of human t(8;21) CBF AML with global transcriptome sequencing, we find that disease progression was characterized by two principal pathogenic mechanisms. Initially, AE expression modified the lineage potential of haematopoietic stem cells (HSCs), resulting in the selective expansion of the myeloid compartment at the expense of normal erythro- and lymphopoiesis. This lineage skewing was followed by a second substantial rewiring of transcriptional networks occurring in the trajectory to manifest leukaemia. We also find that both HSC and lineage-restricted granulocyte macrophage progenitors (GMPs) acquired leukaemic stem cell (LSC) potential being capable of initiating and maintaining the disease. Finally, our data demonstrate that long-term expression of AE induces an indolent myeloproliferative disease (MPD)-like myeloid leukaemia phenotype with complete penetrance and that acute inactivation of AE function is a potential novel therapeutic option.


Supplementary Figure 1A: The R26/AE mouse model directs GFP transgene expression to a subset of HSC and blood cell progenitors
Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for ten days are depicted. FACS plots showing LT-and ST-HSC were gated on 7AAD /L K + S + /CD48 /Flt3 /CD150 + cells and subclassified for the expression of CD34 (LT-HSC) and CD34 + (ST-HSC). The CLP fraction was gated for 7AAD /L K int S int /IL7receptor + cells. GMP, CMP and MEP were gated on 7AAD /L K + S / IL7 receptor cells and subclassified based on the expression of Fc receptor and CD34. Numbers indicate the percentage of GFP + cells within each population. The table at the bottom shows the number of animals that were analysed, mean percentages of GFP + and GFP cells, standard deviations and standard errors of mean for LT-, ST-HSC and lineage restricted blood cell progenitors.

Supplementary Figure 1B: Median fluorescence of GFP + haematopoietic stem cells and lineage-restricted blood cell progenitors
Median fluorescence values of GFP + LT-, ST-HSC, CMP, GMP and MEP populations from four R26/AE reconstituted C57BL/6 mice that have been DOX-induced for ten days are d e p i c t e d i n t h e g r a p h o n t h e t o p . L T -a n d S T -H S C w e r e g a t e d o n 7AAD /L K + S + /CD48 /Flt3 /CD150 + /GFP + c e l l s a n d s u b c l a s s i f i e d f o r t h e e x p r e s s i o n o f CD34 ( L T -H S C ) a n d C D 3 4 + ( S T -H S C ) . G M P , C M P a n d M E P w e r e g a t e d o n 7AAD /L K + S / IL7 receptor /GFP + expression and subclassified based on the expression of Fc receptor and CD34. The median fluorescence is represented as the relative intensity value below which 50% of the events were found within each population. The table at the bottom shows the number of animals that were analysed, median GFP fluorescence values, standard deviations and standard errors.

Supplementary Figure 2: Peripheral blood parameters in control and AE-expressing mice
The bar diagrams show mean peripheral blood parameters ±SD from R26/AE reconstituted RAG2 control mice without DOX exposure (white bars) and reconstituted animals that have been DOX-induced during nine months (black bars). RBC, red blood cell count; HB, haemoglobin; WBC, white blood cell count; PLT, platelet count; MO, monocyte count; EO, eosinophil count; BA, basophil count; LUC, large myeloperoxidase-negative cell count; NE, neutrophil count; LY, lymphocyte count. The table at the bottom shows the number of animals that were analysed, mean percentages, standard deviations and standard errors of mean, p value summaries and p values for each peripheral blood cell parameter.

Supplementary Figure 3: Histological sections through haematopoietic organs from control and AE-expressing mice
Representative H&E-stained sections from R26/AE reconstituted RAG2 control mice without DOX exposure (upper panel) and reconstituted animals that have been DOX-induced during nine months (lower panel) are shown for the thymus, the lymph nodes and the spleen. For each organ the general appearance (x4) together with a higher magnification (x40) are shown. The apparent increase in erythropoiesis in the red pulp of the AE-expressing animal is indicated by black arrowheads.

Supplementary Figure 4: Flow cytometric analysis of extramedullary haematopoiesis in control and AE-expressing spleens
The bar diagrams show total and relative percentages of erythrocytes, granulocytes, B and L K + S + cells from R26/AE reconstituted RAG2 control mice without DOX exposure (white bars) and reconstituted animals that have been DOX-induced during nine months (black bars). In the upper row of plots the absolute percentages of cells within the population of nucleated spleen cells is shown. The lower plots depict the relative percentages of cells within each cell type that was analysed. The tables at the bottom contain information about the number of animals that were tested, mean absolute and relative percentages, standard deviations, standard errors of mean, p value summaries and p values for each cell type.

Supplementary Figure 6: Flow cytometric analysis of control and AE-expressing mice reveals B and T cell maturation defects (A)
Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted on the left. FACS plots showing B cell maturation were gated on 7AAD cells and subclassified into pre-pro (B220 + /IgM ), immature (B220 + /IgM low ) and mature (B220 + /IgM high ) B cell maturation stages. The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the three analysed B cell populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower  table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for each maturation stage. (B) Representative FACS profiles of thymic cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted on the left. FACS plots showing initial T cell maturation were gated on 7AAD /CD4 /CD8 cells and subclassified into DN1 (CD44 + /CD25 ), DN2 (CD44 + /CD25 + ) and DN3 (CD44 /CD25 + ) thymocytes. The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the three analysed T cell populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean  relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for each maturation stage. (C) Representative FACS profiles of thymic cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted on the left. FACS plots showing T cell maturation were gated on 7AAD cells and subclassified into DP (CD4 + /CD8 + ), CD4 (CD4 + /CD8 ) and CD8 (CD4 /CD8 + ). The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the three analysed T cell populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for each maturation stage. Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted. FACS plots showing CLP were gated on 7AAD /L K int S int /IL7 receptor + . The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the CLP population with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for CLP. (B) Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted. GMP, CMP and MEP were gated on 7AAD /L K + S / IL7 receptor expression and subclassified based on the expression of Fc receptor and CD34L K int S int . The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the three analysed progenitor populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed standard deviations, standard errors of mean, p value summaries and p values for each progenitor population. Figure 9: Analysis of GFP a n d G F P + L T -, S T -H S C a n d C L P populations reveals no significant pool size differences between control and AEexpressing mice (A) Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted. FACS plots showing LT-and ST-HSC were gated on 7AAD /L K + S + /CD48 /CD150 + cells and subclassified for the expression of CD34 (LT-HSC), CD34 + (ST-HSC) and GFP. The graph shows relative percentages of control (white squares) and DOX-exposed GFP (black dots) and GFP + (green dots) cells within the LT-and ST-HSC populations with each square and dot representing one individual mouse. The tables at the right show the number of animals that were analysed, mean percentages of control, GFP and GFP + cells, standard deviations and standard errors of mean, p value summaries and p values for LT-HSC (upper table) and ST-HSC (lower table). (B) Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted. FACS plots showing CLP were gated on 7AAD /L K int S int /IL7 receptor + and subclassified for the expression of GFP. The graph shows relative percentages of control (white squares) and DOX-exposed GFP (black dots) and GFP + (green dots) cells within the CLP population with each square and dot representing one individual mouse. The table indicates the number of animals that were analysed, mean percentages of control, GFP and GFP + cells, standard deviations and standard errors of mean, p value summaries and p values for CLP.

Supplementary Figure 10: Relative percentages of GFP + but not GFP GMP and MEP pools are altered in AE-expressing mice
Representative FACS profiles of BM cells from R26/AE reconstituted RAG2 mice that have been DOX-induced for nine months are depicted. The first plot shows the percentages of GFP and GFP + cells in the BM of a DOX-induced mouse. In the other plots, control, DOXexposed GFP and GFP + GMP, CMP and MEP are shown that were gated on 7AAD /L K + S / IL7 receptor expression and subclassified based on the expression of Fc receptor, CD34 and GFP. The graph at the left bottom shows relative percentages of control (white squares), DOX-exposed GFP (black dots) and GFP + (green dots) cells within each progenitor population with each square and dot representing one individual mouse. The tables indicate the number of animals that were analysed, mean relative percentages of control, GFP and GFP + cells, standard deviations and standard errors of mean, p value summaries and p values for CMP, GMP and MEP.

S Supplementary Figure 11: Immature granulocytes are increased in the BM of long-term AE-expressing mice
Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for eighteen months are depicted on the left. FACS plots showing granulocytes were gated on 7AAD cells and subclassified into immature (CD11b + Gr1 low ) and mature (CD11b + Gr1 high ) stages. Th e graph sh ows rel ati v e percen tages of con tr ol (whi te squares) and DOX-induced (black dots) cells within the two analysed granulocytic populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for each maturation stage.

Supplementary Figure 12: Long-term AE expression does not alter the pool of LT-and ST-HSC in leukaemic mice
Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for eighteen months are depicted. FACS plots showing LT-and ST-HSC were gated on 7AAD /L K + S + /CD48 /Flt3 /CD150 + cells and subclassified for the expression of CD34 (LT-HSC) and CD34 + (ST-HSC). The graph on the right shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the LT-and ST-HSC populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for LT-and ST-HSC.

Supplementary Figure 13: AE expression alters the relative percentages of CMP, GMP and MEP (A)
Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for eighteen months are depicted. FACS plots showing CLP were gated on 7AAD /L K int S int /IL7 receptor + and subclassified for the expression of GFP. The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the CLP population with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for CLP. (B) Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for eighteen months are depicted. GMP, CMP and MEP were gated on 7AAD /L K + S / IL7 receptor expression and subclassified based on the expression of Fc receptor, CD34 and GFP. The graph shows relative percentages of control (white squares) and DOX-induced (black dots) cells within the three analysed progenitor populations with each square and dot representing one individual mouse. The two tables on the right indicate mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for each maturation stage. Representative FACS profiles of BM cells from R26/AE reconstituted C57BL/6 mice that have been DOX-induced for eighteen months are depicted. The first plot shows GFP and GFP + cells in the BM of a leukaemic mouse with numbers indicating the percentages of GFP and GFP + cells within the population of nucleated BM cells. In the contour plots, control, DOX-exposed GFP and GFP + G M P , C M P a n d M E P a r e s h o w n t h a t w e r e g a t e d o n 7AAD /L K + S / IL7 receptor expression and subclassified based on the expression of Fc receptor, CD34 and GFP. The table below indicates the number of animals that were analysed, mean relative percentages of control, GFP and GFP + cells, standard deviations, standard errors of mean, p value summaries and p values for CMP (upper table), GMP (table  in the middle) and MEP (lower table). Figure 16: Flow cytometric analysis of BM granulocytes reveals the reduction of immature myeloid cells in two previously leukaemic mice following the DOX switch (A) Histogram plots of Gr1-expressing BM granulocytes for a control animal (black outline and filled in grey), a mouse continuously receiving DOX (+DOX, red line), a mouse shifted to a DOX-free diet presenting signs of progressing leukaemia ( DOX, orange line) and an animal that has been shifted to a DOX-free diet presenting signs of recovery and partial restoration of granulocytic maturation ( DOX, blue line). Note the right shift of the blue profile indicating the reduction of immature Gr1 low and the increase of Gr1 high cells. The table on the right indicates the relative percentages of GFP + mature and immature granulocytes for each individual mouse that is shown in the histogram. (B) The graph on the left shows relative percentages of immature (Gr1 low ) and more mature (Gr1 high ) cells within the BM population of granulocytes for control mice (white squares), animals continuously receiving DOX (+DOX, red dots), animals that had been shifted to a DOX-free diet presenting signs of progressing leukaemia ( DOX, orange dots) and mice that had been shifted to a DOX-free diet presenting signs of recovery ( DOX, blue dots). The tables on the right indicate the mean absolute and relative percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for all immature (upper tables) and mature granulocytes (lower tables).

Supplementary Figure 17: Post reconstitution flow cytometric analysis demonstrates that L-HSC reconstitute both lympho-and myelopoiesis whereas L-GMP only reconstitute myelopoiesis
The upper panel shows representative FACS plots for an L-HSC-and the lower panel for an L-GMP-reconstituted mouse. Numbers indicate the percentages of GFP + cells in the BM (left plots) or in the myeloid, B and T cell fraction of peripheral blood samples. Note the absence of GFP-expressing T and B cells in the L-GMP-reconstituted mouse.

Supplementary Figure 18: Flow cytometric analysis reveals differences in granulocytic maturation between L-HSC-and L-GMP-reconstituted mice (A)
Histogram plots of CD11b/Gr1-expressing granulocytes for a control animal (black outline and filled in grey), an L-HSC-(red line) and an L-GMP-(orange line) reconstituted mouse revealing the presence of GFP-expressing granulocytes in the BM of reconstituted mice. The table below indicates the relative percentages of GFP+ granulocytes for each individual mouse that is shown in the histogram. The graph on the right shows increased relative percentages of CD11b + /Gr1 + BM granulocytes in both L-HSC-(red dots) and L-GMP-(orange dots) reconstituted mice when compared to control mice (white squares). The two tables on the right indicate the mean relative cell percentages, number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for L-HSC-reconstituted (upper table) and L-GMP-(lower table) reconstituted animals. (B) Contour plots depicting immature (CD11b + Gr1 low ) and mature (CD11b + Gr1 high ) granulocytes of control, L-HSC-and L-GMP-transplanted mice. Thick-framed boxes highlight immature granulocytes. The graph on the right shows relative percentages of cells for control (white squares), L-HSC-(red dots) and L-GMP-(orange dots) transplanted mice within the immature and mature granulocytic population with each square and dot representing one individual mouse. The two tables at the bottom indicate the mean absolute (upper table) and mean relative (lower table) cell percentages, the number of animals that were analysed, standard deviations, standard errors of mean, p value summaries and p values for control L-HSC-and L-GMP-transplanted mice. Note that immature CD11b + /Gr1 low granulocytes are augmented in both L-HSC and L-GMP-transplanted mice whereas the percentage of more mature CD11b + /Gr1 high cells is only significantly decreased in L-HSCreconstituted animals but not in L-GMP-transplanted mice. Table S1 and Table S2: Whole transcriptome and mRNA profiles of Ctrl-, ST-and L-GMP populations are provided in Table S1 and Table S2.   Table S4 contains the list of human orthologous genes corresponding to the top 150 transcripts that were highly up-regulated or exclusively expressed in murine ST-GMP and in murine L-GMP indicating the rank in the gene list, the metric score and the individual enrichment score found in a cohort of 61 t(8;21) CBF AML patients.  Table S5 contains the list of 17 human genes that were found to be on average two fold or more increased in a group of 61 t(8;21) patients as compared to a cohort of 69 healthy donors. In addition, these genes were silent or only very poorly transcribed in sorted primary human CD34+/CD38-cord blood HSC, CD34+/CD38low/CD36-cord blood HSC, primary human CD34+/CD38+ haematopoietic progenitors and unsorted Lin+/-human primary cord blood cells.

Flow cytometry analysis and sorting
For staining of differentiated blood cell lineages we used antibodies against CD4, CD8, CD25, CD44, B220, IgM, Gr1, CD11b, Ter119, CD71, CD41 and CD3. Erythrocytes were removed by lysis with ACK buffer (Lonza) for 5 minutes at room temperature. Stem cells and progenitors were stained with a lineage cocktail containing antibodies against CD3, B220, CD11b, TER119 and Ly-6G followed by staining with antibodies against Fc RII/III, IL7R , CD34, CD150, c-Kit, Sca-1, CD48 and/or Flt-3. A list of all antibodies that were used in this study is shown below.

Quantitative real-time PCR
The Kasumi-1 cell line (Asou et al, 1991)  Differences in cDNA input between samples were compensated by normalizing to the expression of hprt ( C t ). The fold-change value between each test sample was calculated according to Schmittgen (Schmittgen, 2001), calibrated relative to the amount of mRNA in the Kasumi-1 cell line and displayed as 2 Ct . Primers and cycling details are shown below.

QPCR primers
AML1-ETO forward primer 5'GCCCCGAGAACCTCGAAA3' AML1-ETO reverse primer 5'TCCACACGTGAGTCTGGCATT3' hprt forward primer 5'GCTCGAGATGTCATGAAGGAGAT3' hprt reverse primer 5'AGCAGGTCAGCAAAGAACTTATAGC3' BM transplantation for the analysis of phenotypes following short-and long-term DOX induction BM cells were obtained from non-induced compound R26/AE donors. For the analysis of mice that were exposed to DOX for 8-10 months, 5 x 10 5 whole BM donor cells were injected into the tail vein of irradiated RAG2 -/animals (4.5 Gy). For long-term DOX-induction experiments (DOX exposure 16-18 months) involving peripheral blood films, spleen weight, FACS analysis, histology of spleen, thymus, lymph nodes, liver, kidney and lung we used C57BL/6 recipients (irradiation 2 x 5.5 Gy) that were tail vein injected with 5x10 5 donor cells from non-induced compound R26/AE mice and DOX-exposed for either 16 months (N=5) or 18 months (N=4). Spinal cord sections for histo-pathological analysis ( Figure 3D) were obtained from RAG2 -/reconstituted recipients (4.5 Gy) that were injected with 5x10 5 BM cells from non-induced compound R26/AE donors and exposed to DOX for 20-22 months (N=4). In all experiments cohorts of mice that were reconstituted in the same fashion but not

RNA-Seq
Cell sorting: GMP (4 x 10 4 L -K + S + IL7R -CD34 + Fc RII/III + cells) from non-induced compound R26/AE controls (Ctrl-GMP) or C57BL/6 recipients (2 x 5.5 Gy) that were injected with 5 x 10 5 non-induced R26/AE BM cells and exposed to DOX for ten days (ST-GMP) were sorted using a FACSAria cell sorter (BD). To obtain long-term DOX-induced leukaemic GMP (L-GMP), 5 x 10 5 BM cells from R26/AE-reconstituted C57BL/6 recipients that had been exposed to DOX for 18 month were injected into secondary C57BL/6 recipients (2 x 5.5 Gy) and the DOX regime was continued for an additional four month. At this point manifest leukaemia was confirmed by inspection of peripheral blood films (presence of circulating blasts), high WBC counts and splenomegaly. For obtaining L-GMP cells for the following RNA-Seq experiments, 4 x 10 4 L -K + S + IL7R -CD34 + Fc RII/III + cells were preparatively sorted using a FACSAria cell sorter (BD). In all cases sorting efficiencies were confirmed by re-analysis of each sample (purity > 97%).
Next Bioinformatics and data pre-processing: Raw output data was pre-processed according to the Illumina standard protocol including filtering for low quality reads and de-multiplexing.
RNA-Seq data acquisition: Sequence reads were aligned to the mm9 reference mouse genome sequence (Waterston et al, 2002) using bowtie (Langmead et al, 2009). Alignment coordinates were compared to the exon coordinates of RefSeq transcripts (Pruitt et al, 2007) and for each transcript the counts of overlapping alignments were recorded. Sequence reads that were not aligned to the genomic sequence were aligned to a database of all possible exon-exon junction sequences of the RefSeq transcript deposit. Finally, counts of reads aligning to splice junctions were aggregated with the respective transcript counts and normalized to RPKM (number of reads which map per kilobase of exon model per million mapped reads (Mortazavi et al, 2008)). The count data was analyzed for differential expression using the R/Bioconductor package DESeq (Anders & Huber, 2010) and normalized by edgeR (Robinson et al, 2010). Pathway analysis was performed with the Ingenuity IPA software and scatter plots were generated using ANTHEUS (Build 3.3.66).

Isolation and functional characterization of LSC
To obtain long-term DOX-induced L-HSC and L-GMP, 5 x 10 5 BM cells from R26/AEreconstituted C57BL/6 recipients that had been exposed to DOX for 18 month were injected into secondary C57BL/6 recipients (2 x 5.5 Gy) that were kept on DOX. Four month later all animals were sacrificed and presented signs of leukaemic disease indicated by the presence of circulating blasts, high WBC counts and splenomegaly. To isolate L-HSC (defined as L -K + S + CD150 + ) a n d L -G M P ( d e f i n e d a s L -K + S + IL7R -CD34 + Fc RII/III + ) w h o l e B M o f t h e sacrificed animals was subjected to preparative FACS using a FACSAria cell sorter (BD). In all cases sorting efficiencies were confirmed by re-analysis of each sample (purity > 97%). To test the potential of L-HSC and L-GMP for initiating and maintaining leukaemia (LSC-assay), we injected 1 x 10 3 L-HSC (9 recipient mice) or 2 x 10 4 GMP (10 recipient mice) together with 2.5x10 5 supportive C57BL/6 BM cells into C57BL/6 (2 x 5.5 Gy) animals that were continued on DOX. Recipient mice were bled seven weeks post transplantation and reconstitution was confirmed by analysing GFP expression in peripheral granulocytes, B and T cells. Five month after the transfer all animals were analysed. FACS sorted L-HSC and L-GMP from age-matched R26/AE BM-reconstituted mice that were treated in the same manner as described above but had never been exposed to DOX were used as control cells in the LSCassay.

Acute ablation of AE expression in leukemic mice
In these experiments 5 x 10 5 BM cells from R26/AE-reconstituted C57BL/6 recipients that had been exposed to DOX for 18 month were injected into secondary C57BL/6 recipients (2 x 5.5 Gy) and continued on DOX. Four month later all animals were sacrificed and presented signs of leukaemic disease indicated by the presence of circulating blasts, high WBC counts and splenomegaly. To be able to investigate the effect of AE ablation in mice that individually progressed towards leukaemia and thus might have acquired different secondary mutations, we used three different sacrificed mice as BM donors and ten individual mice as recipients. In each case 2.5 x 10 5 leukemic BM cells were co-injected together with 2.5 x 10 5 supportive C57BL/6 BM cells into a single C57BL/6 mouse (2 x 5.5 Gy) and all animals were continued on DOX for an additional three month. At this point seven mice were switched to a DOX-free diet and three mice were continued on DOX. Four month later all animals were analysed.
Age-matched R26/AE BM-reconstituted mice that were treated in the same manner as described above but had never been exposed to DOX were used as controls.

Colony-forming unit assay
Bone marrow cells (1x10 4 ) of transplanted control and eight months AE-expressing recipients were seeded into methylcellulose-containing media (MethoCult 3434; StemCell Technologies). The cultures were plated in triplicate and incubated with 5% CO 2 at 37°C.
After eight days, colonies were scored according to the manufacturer's instructions. In each case three control and three AE-expressing animals were analysed.

Routine haematologic assays and histopathology
Peripheral blood counts were performed on an ADVIA120 Haematology Analyzer (Bayer) or an HAEMAVET HV950 (Drew). Peripheral blood films were air dried and subsequently stained with Giemsa. Organs were formalin-fixed (4% PBS buffered formaldehyde) embedded in paraffin, sectioned and stained with haematoxylin and eosin according to standard protocols. Images were generated using a BX45 Microscope equipped with a C4040 digital camera (Olympus) and imported into Photoshop 7.0 (Adobe Systems).

Conversion and comparison of murine ST AE-and leukaemia-associated transcripts with human t(8;21) CBF AML and non-leukaemic profiles
For converting mouse RefSeq identifiers to orthologous Human Genome U133 Plus 2.0 Affymetrix IDs BioMart was used (http://www.biomart.org). To compare mouse and human gene expression profiles, confirmed t(8;21) AML and non-leukaemic gene expression signatures were extracted from GSE17855 (Balgobind et al, 2011), GSE6891 andGSE22056 (de Jonge et al, 2010;Verhaak et al, 2009) and GSE15061 (Mills et al, 2009) and normalized using the RMA procedure (Irizarry et al, 2003). To select L-GMP-specific genes that were absent or poorly expressed in human HSC, progenitors and normal blood cells, gene expression signatures from GSE30377 (Eppert et al, 2011) were extracted and normalized as above. For gene set enrichment analysis we applied the GSEA software package (Subramanian et al, 2005) and for determining the expression of genes in normal embryonic and adult tissues the Gene Enrichment Profiler program (Benita et al, 2010) was used.