Highly efficient Runx1 enhancer eR1-mediated genetic engineering for fetal, child and adult hematopoietic stem cells

A cis-regulatory genetic element which targets gene expression to stem cells, termed stem cell enhancer, serves as a molecular handle for stem cell-specific genetic engineering. Here we show the generation and characterization of a tamoxifen-inducible CreERT2 transgenic (Tg) mouse employing previously identified hematopoietic stem cell (HSC) enhancer for Runx1, eR1 (+24m). Kinetic analysis of labeled cells after tamoxifen injection and transplantation assays revealed that eR1-driven CreERT2 activity marks dormant adult HSCs which slowly but steadily contribute to unperturbed hematopoiesis. Fetal and child HSCs which are uniformly or intermediately active were also efficiently targeted. Notably, a gene ablation at distinct developmental stages, enabled by this system, resulted in different phenotypes. Similarly, an oncogenic Kras induction at distinct ages caused different spectrums of malignant diseases. These results demonstrate that the eR1-CreERT2 Tg mouse serves as a powerful resource for the analyses of both normal and malignant HSCs at all developmental stages.


Introduction
There has been considerable interest in identifying a cis-regulatory element that targets gene expression to hematopoietic stem cells (HSCs). Characterization of such an enhancer, termed HSC enhancer, holds promise of generating important insights into the transcriptional programs responsible for "stemness", and will also provide a powerful tool for manipulation of HSCs in experimental animal models, such as mouse and zebrafish. A HSC enhancer, in combination with reporters such as EGFP or LacZ, can be used for the marking of HSCs [1][2][3]. This way of HSC identification is called genetic marking of HSCs. In addition, the use of fluorescence reporters, like EGFP, enables in vitro and in vivo live imaging of HSCs. Such a HSC enhancer can also drive the expression of any gene of interest, namely transgenesis in HSCs. Furthermore, when engineered to drive the expression of Cre recombinase gene, a HSC enhancer enables excision of a floxed allele, thereby allowing for genetic engineering, including gene targeting in a HSC-specific manner.
A series of candidate HSC enhancers have been identified. However, none of them have shown activities restricted to HSC compartment, but much broader expression including less immature progenitor fractions and endothelial cells which are major components in the bone marrow (BM) [4][5][6][7][8]. We and others have succeeded in identifying an intronic enhancer for Runx1, named eR1 (+24m), which is highly active in HSCs [2,9,10]. To exploit this HSC enhancer activity of eR1, eR1-CreER T2 transgenic (Tg) mouse was generated.

Results
Generation and selection of eR1-CreER T2 Tg mouse lines eR1 is a regulatory element, being active in HSCs and located at 24 kb downstream of mouse Runx1 P1 promoter ( Figure 1A). A Tg mouse with CreER T2 recombinase expression driven by eR1 was generated ( Figure 1B). Crossing with a reporter strain, Rosa26-LSL-tdTomato mice, selection of appropriate Tg line was conducted amongst 5 Tg lines, in all of which successful germline transmission of Tg was confirmed ( Figure S1A). Line #5-2 was selected for subsequent analyses for normal HSCs due to its highest labeling efficiency, 35.83% Tomato + cells in peripheral blood (PB) at 1 week after a single tamoxifen injection. For malignant HSC studies, another Tg line, #3-17 was selected for its lowest leakage (0.01%) in the absence of tamoxifen treatment ( Figure S1A and S1D). Optimization study for tamoxifen dosage and mouse age, using the eR1-CreER T2 Tg mouse line #5-2, demonstrated that the highest labeling efficiency with minimal toxicity can be achieved at 0.2 mg/g x 6 times on alternative days from 3 to 4 weeks old Tg mice ( Figure 1C-H, S1E and S1F). Using this dosing regimen, no apparent changes were found in hematological parameters including flow cytometry analyses for stem/progenitor fractions.

Highly efficient HSC marking in adult and child eR1-CreER T2 Tg mice
In tamoxifen-induced labeling experiment, successful HSC marking was evaluated according to the following three criteria: 1) Initially targeted cells are HSCs, 2) Labeling is traced in all hematopoietic lineages (lineage tracing), and 3) Labeling is maintained for a significantly long period as long-term label-retaining cells. To interrogate these points, analysis of eR1-CreER T2 Tg; Rosa26-LSL-tdTomato mice was conducted at 48 hours, 1 month and 1 year after 0.2 mg/g single tamoxifen injection ( Figure 1C). It is postulated that if initially targeted cells are HSCs, all the progeny cells derived from HSCs will retain the tdTomato signals in the Rosa26-LSL-tdTomato mouse system. Hence, one month after tamoxifen injection, all hematopoietic cells (HPCs) are expected to display tdTomato signals. The schematic diagram in Figure 1D depicts the expected dynamics of tdTomato + cells in the eR1-CreER T2 ; Rosa26-LSL-tdTomato mice.
In line with the reported eR1 activity in HSCs, the highest marking (40.5% tdTomato + cells) was observed in long term hematopoietic stem cells (LT-HSC) fraction at 48 hours after tamoxifen injection ( Figure 1E). The percentages of tdTomato + cells significantly decreased with differentiation in a stepwise manner and only 2.5% lineage positive (Lin + ) cells expressed tdTomato. These results suggest that the initially targeted cells by eR1-mediated labeling are largely LT-HSCs and, to a lesser extent, short term hematopoietic stem cells (ST-HSCs) and multipotent progenitor (MPP) cells.
At one month after tamoxifen injection, 72.5% Lin + cells demonstrated tdTomato signals ( Figure 1F) and labeled cells were observed throughout all hematopoietic lineages, suggesting that originally targeted cell population via eR1 activity were primarily HSCs. The percentage of tdTomato + cells in PB progressively increased until the longest time point of observation, one and half year after tamoxifen injection ( Figure 1G).
It has been recently reported that actively proliferating ST-HSCs and MPPs mainly contribute to steady state adult hematopoiesis, whereas dormant LT-HSCs usually do not but serve as a reservoir population for stressed hematopoiesis [11,12]. To further prove that the dormant adult HSCs were targeted by eR1-CreER T2 , 5-fluorouracil (5-FU) was injected into eR1-CreER T2 Tg; Rosa26-LSL-tdTomato mice one month after tamoxifen injection. In two different injection cohorts, the percentages of tdTomato + cells in PB increased rapidly after 5-FU challenge (dotted lines in Figure 1G). These results suggest that eR1-CreER T2 mediated labeled cells include dormant adult LT-HSCs, which minimally contribute to unperturbed adult hematopoiesis but play an active role in stressed hematopoiesis.
It is well recognized that HSC behavior is different at distinct developmental stages.
HSC population rapidly expands from E10.5 onwards until 4 weeks after birth, while HSCs are largely maintained in quiescence during the adult stage [13]. LT-HSCs were highly proliferative in 3-week-old mice as the percentage of tdTomato + cells in LT-HSCs increased from 40% to 70% in the period from 48 hours to 1 week after tamoxifen injection, whereas the percentage of tdTomato + cells increased marginally (45% to 50%) in LT-HSCs from 6-weekold mice, concordant with the notion that HSCs are maintained in a quiescent state in adult bone marrow (BM) ( Figure S1E and S1F). Collectively, these results suggest that the eR1-CreER T2 mouse model is useful for HSC marking at both adult (6 weeks or older) and childhood (1-4 weeks) stages.

eR1-CreER T2 system enables the best HSC enrichment
The most rigorous way to functionally define HSC is to perform transplantation assays that allow for the assessment of long-term multilineage engraftment. To further confirm HSCspecific targeting in eR1-CreER T2 Tg mice, long-term reconstitution experiments were performed, using the congenic Ly5.1/Ly5.2 system, where irradiated adult mice were used as recipients. Sorted BM cell suspensions were obtained from individual adult eR1-CreER T2 Tg; Rosa26-LSL-tdTomato mice at 48 hours after tamoxifen injection (Figure 2A). Two different sets of transplantation experiments were carried out, using either c-Kit + Sca1 + Lin -(KSL) or CD34 -Flt3 -KSL BM cells as donor cells. In both experiments, tdTomato + and tdTomatocells were sorted from each fraction ( Figure 2B). 1,000 KSL or 100 CD34 -Flt3 -KSL BM cells of either tdTomato + or tdTomatowere transplanted into irradiated recipient mice, with 2 x 10 5 non-Tg BM cells to provide short-term radioprotection. Donor cell engraftment was examined for 6 months after transplantation. tdTomato + cells from both KSL and CD34 -Flt3 -KSL fractions exhibited higher short-term (1-3 months) and long-term (4-6 months) chimerism in PB respectively as compared to their tdTomatocounterparts ( Figure 2C and 2D).
To ascertain the HSC enrichment by eR1, limiting dilution assay was conducted. The HSC frequency of tdTomato + and tdTomatopopulations in CD34 -Flt3 -KSL fraction was 1 in 7.96 and 1 in 26.77 respectively ( Figure 2E and 2F). This result suggests that eR1-CreER T2 mouse enables the robust HSC enrichment better than the CD34 -Flt3 -KSL fraction which is the current standard method to sort LT-HSCs. No bias towards specific lineages in reconstitution capability was seen in eR1-active (tdTomato + ) CD34 -Flt3 -KSL HSCs, whereas a moderate lymphoid-biased reconstitution was found in eR1-inactive (tdTomato -) CD34 -Flt3 -KSL HSCs ( Figure S2).
To determine the behavior of tdTomato + cells in the different bone marrow stem progenitor cell fractions and peripheral blood of eR1-CreER T2 Tg;Rosa26-LSL-tdTomato mice, a one-year kinetic study was carried out. Similar percentages of tdTomato + cells were observed in the ST-HSCs and LT-HSCs one month after tamoxifen injection (0.2 mg/g) in 6week-old mice ( Fig 2G). tdTomato + cells in the ST-HSCs increased steadily at 3 months and 6 months whereas the LT-HSC tdTomato + cells exhibited a slower increase than those in ST-HSCs during the same time period. However, tdTomato + cells in the LT-HSC fraction increased and reached equivalent levels with ST-HSC tdTomato + cells at the end of one year.
These results support that eR1-mediated labeling occurs in dormant LT-HSCs, the contribution of which to native hematopoiesis is limited when young, but progressively increases with age.

In situ identification of HSCs using eR1-CreER T2 Tg mouse
To explore the potential of eR1 system for in situ visualization of HSCs, bones of eR1-EGFP Tg mouse were subjected to immunohistological analysis. Half of the eR1-driven EGFP high cells were c-Kit + cells, and these c-Kit + EGFP high cells, which are likely to represent HSCs, were evenly distributed along the bone of 3-week-old eR1-EGFP Tg mouse ( Figure 3A arrow and 3B). In contrast, in 3-month-old mice, c-Kit + EGFP high cells were preferentially located at the trabecular region, which provides HSCs with an environment to preserve their dormant status ( Figure 3C arrow and D). This difference in the distribution of c-Kit + EGFP high cells, consistent with the results in earlier studies, suggests that eR1-driven fluorescence signal is useful for in situ visualization of HSCs. A similar assay was carried out using bones from 4month-old eR1-CreER T2 Tg; Rosa26-LSL-tdTomato mouse at 48 hours after tamoxifen injection. The immunoflourescent staining result showed the successful identification of c-Kit + tdTomato + cells both in metaphysis ( Figure 3E and 3G) and diaphysis ( Figure 3F). In trabecular region, a c-Kit + tdTomatocell adjacent to a c-Kit + tdTomato + cell was observed.
Such c-Kit + tdTomatocell might represent a daughter cell divided from c-Kit + tdTomato + cell by asymmetric cell division ( Figure 3H). Paired c-Kit + tdTomato + cells, which might be generated by symmetric division, were also found ( Figure 3I). These results illustrate that both Tg mouse models, eR1-EGFP Tg and eR1-CreER T2 Tg mice, enable in situ imaging of HSCs.
Fetal HSCs are marked in eR1-CreER T2 Tg mouse eR1 is shown to be active in hemogenic ECs and HPCs at embryonic stage E10.5 [2]. To investigate the eR1-driven CreER T2 expression pattern in newly generated Tg mouse at embryonic stage, Rosa26-LSL-tdTomato homozygous female mouse was crossed with eR1-CreER T2 Tg male mouse. The pregnant females were injected with 0.05 mg/g of tamoxifen on 9.5/ 10.5-and 0.05 or 0.1 mg/g 14.5-days post coitum (d.p.c.) ( Figure 4A). These tamoxifen injections generated 96.7%, 87.6% and 73.2% tdTomato + cells, respectively, in the PB of 3- week-old progeny mice ( Figure 4B). The percentages of marked cells in PB did not increase with age of mice, unlike those observed in the adult mice. Analysis of BM cells from 6-monthold progeny mice with tamoxifen injection on 14.5-d.p.c. showed nearly 100% marking in all the hematopoietic stem/ progenitor cell (HSPC) fractions ( Figure S3C).
To examine the initially labeled cells by tamoxifen injection on 9.5-d.p.c., eR1-CreER T2 Tg; Rosa26-LSL-tdTomato embryos were harvested at 24 hours after tamoxifen injection. tdTomato signals were observed in the main hematopoietic sites, such as aortagonad-mesonephros (AGM), vitelline artery (VA) , umbilical artery (UV) and fetal liver (FL), in the E10.5 embryos ( Figure S3A iii). This expression pattern is largely identical to that seen in the eR1-EGFP Tg embryos ( Figure S3A iv) and endogenous Runx1+ cells visualized by staining by anti-Runx1 antibody ( Figure S3D). Remarkably, no tdTomato signals were observed in any eR1-CreER T2 Tg; Rosa26-LSL-tdTomato E10.5 embryos without tamoxifen injection ( Figure S3A i and ii). Whole-mount immunostaining of the embryos further confirmed the presence of tdTomato + cells in the known hematopoietic sites ( Figure 4C). HPC clusters were preferentially accumulated in the middle of distal artery (DA) within the AGM region as compared to rostral and caudal regions ( Figures 4C and D). HPCs at E10.5 were derived from hemogenic ECs in DA. Some endothelial cell marker CD31 + ECs in DA were co-stained with tdTomato antibody, suggesting that hemogenic ECs were labeled as expected ( Figure   4E). tdTomato-expressing HPC clusters retained CD31 signals, indicating that the cluster forming cells were emerged from both dorsal and ventral artery walls ( Figures 4F and G). The ECs of heart also displayed strong tdTomato signals. In addition, tdTomato + cells were aligned between pairs of somites that most likely represent the notochord. The notochord marking was only observed in eR1-CreER T2 Tg mouse, but not in eR1-EGFP Tg mouse. The significance of this labeling awaits further investigation.
Labeling and gene targeting efficiency in eR1-CreER T2 Tg; Rosa26-LSL-tdTomato embryo was further analyzed by flow cytometry using antibodies against CD31, CD41, CD45 (pan-leukocyte marker) and Ter119 (erythrocyte marker). Consistent with the whole-mount immunostaining result, essentially all of the CD45 + CD41 + cells were tdTomato + cells, 11% of CD31 + ECs exhibited tdTomato signal but Ter119 + erythroid cells lacked tdTomato signal ( Figure 4H), suggesting that eR1 drives CreER T2 expression in a part of ECs, probably hemogenic ECs and HPCs, but not in primitive erythrocytes. Collectively, these results demonstrate that the eR1-CrER T2 Tg system is similarly useful for the labeling of hemogenic ECs and fetal HSCs.

Induction of hematological malignancy by eR1-CreER T2 Tg system
It is well documented that the introduction of an oncogenic event in the stem cell fraction can efficiently induce tumor development [14][15][16][17]. To establish an inducible leukemia mouse model, eR1-CreER T2 Tg; Kras-LSL-G12D mice were generated. Line #3-17 of eR1-CreER T2 Tg mouse was utilized for this study since this Tg line demonstrated no leakage of CreER T2 activity in the BM of 5-month-old mouse ( Figure S1D). Oncogenic Kras activation is shown to readily induce hematopoietic malignancy [14]. As expected, all eR1-CreER T2 Tg; Kras-LSL-G12D mice developed hematopoietic malignancies within 4 months (median survival 98 days) after a single tamoxifen injection, whereas no disease development was observed in uninjected eR1-CreER T2 Tg; Kras-LSL-G12D mice ( Figure 5A).
In Mx1 promoter (p)-Cre Tg mice, minimal Cre expression was reported to be around 5% before polyinosinic-polycytidylic acid (pIpC) injection [14]. However, this estimation was obtained using Rosa26-LSL-LacZ reporter mouse, which is far less sensitive than Rosa26-LSL-tdTomato mouse as a reporter strain. As shown in Figure 5B, Mx1(p)-Cre Tg; Rosa26-LSL-tdTomato mice exhibited around 20% tdTomato + cells in PB of 6-week-old mice without pIpC injection and the leakage gradually increased with age (data not shown). After pIpC injection, the frequency of tdTomato + cells in PB varied amongst individual mice. In contrast, eR1-CreER T2 Tg mice showed no leakage of Cre activity before induction and minimal variation in the percentage of tdTomato + cells in PB after induction. This result reveals that the Mx1(p)-Cre inducible system, currently the most widely used method for the induction of leukemia, has fundamental flaws due to its high leakage of Cre activity, rendering it a faulty inducible system. Therefore, the eR1-CreER T2 Tg mouse is the long-awaited improved inducible mouse model for the induction of malignancy in hematopoiesis eR1 is active in leukemia stem cells Cancer has been shown to harbor cells with stem cell like properties, termed cancer stem cells (CSCs) [18][19][20]. eR1 is anticipated to be active in CSCs in leukemia, namely leukemia stem cells (LSCs). To investigate the potential of LSC marking by eR1, tamoxifen-inducible eR1-CreER T2 Tg; Kras-LSL-G12D; Rosa26-LSL-tdTomato; eR1-EGFP Tg mice (henceforth referred to as 4G mice) were generated. The EGFP + signal mark cells carrying stemness, while tdTomato + signal serves as a surrogate marker for oncogenic Kras activation. The compound 4G mice developed hematopoietic malignancy after tamoxifen injection. BM cells were harvested from the diseased mice. EGFP + and EGFPcells were separately sorted from the tdTomato + population and transplanted into recipient mice ( Figure 5C). Hematopoietic malignancies were observed in mice transplanted with tdTomato + EGFP + (TG) cells, but not in those transplanted with tdTomato + EGFP -(T) cells ( Figure 5E), suggesting that eR1 is active in LSCs.
In attempt to further enrich LSCs, we employed a combination of c-Kit and EGFP intensity. A well-known HSC marker, c-Kit is often used to enrich LSCs [21]. In normal HSCs, eR1-driven EGFP high fraction was found to contain greater abundance of LT-HSCs compared with eR1-driven EGFP low population (Ng et al., 2010). c-Kit + tdTomato + EGFP low (TG low) and c-Kit + tdTomato + EGFP high (TG high) population, which was present at 1.2% in the BM of diseased 4G mouse, were isolated from diseased mice and transplanted into recipient mice ( Figure 5D). In order to examine the minimal number of LSCs required to cause hematopoietic malignancies, varying number of sorted cells were used for transplantation. At all cell dosages tested, hematopoietic malignancies were observed in those transplanted with TG high population, but not in those mice transplanted with TG low population ( Figure 5F).
Recipient mice transplanted with as low as 200 TG high cells were capable of developing hematopoietic malignancies, suggesting that LSCs are highly enriched within c-Kit + tdTomato + EGFP high population in the 4G mouse model.

Induction of Kras G12D in fetal, child and adult HSCs causes different malignancies
It has long been suspected through clinical observations that oncogenic potential and/or tumor spectrum of a certain oncogenic event is heavily dependent on the developmental stage in which it occurs. To experimentally interrogate this notion, we utilized the eR1-CreER T2 system to induce oncogenic Kras in HSCs at different developmental stages, such as fetus, child and adult. The results revealed that the survival curves and tumor spectrums of eR1-CreER T2 Tg; Kras-LSL-G12D diseased mice were distinct amongst the three cohorts ( Figure 6A and 6B). In child (3-week-old) induction cohort, 67% of diseased mice were biphenotypic lymphoma carrying a mixture of myeloid and T cell features. The remaining 22% of diseased mice also exhibited myeloid features as myeloproliferative disorder (MPD) or biphenotypic leukemia. In the biphenotypic lymphomas, abnormal cells expressed myeloid markers such as Mac-1 and CD61 in thymus and spleen ( Figure 6C and 6D) and immature cell marker, c-Kit ( Figure 6D). In striking contrast, 100% of diseased mice developed T-cell malignancies in the adult induction cohort, including 60% lymphoma and 40% leukemia without myeloid features. Within the fetal cohort, majority of the mice developed T cell lymphoma (74%). The rest of the diseased mice (26%) displayed features of MPD and biphenotypic lymphoma with T and myeloid cell features. It was also observed that the surface marker expression of TCRαβ was higher in the T cell lymphoma of fetal cohort versus the adult cohort suggesting that the T cell lymphoma of these cohorts differ from one another (data not shown). Evidently, the induction of the same driver oncogene at distinct developmental stages resulted in remarkably different disease spectra.

A gene ablation in fetal, child and adult HSCs lead to distinct phenotypes
To demonstrate the usefulness of eR1-CreER T2 Tg mouse for gene targeting in HSCs,  Figure 7B and 7C). These phenotypes are closely resemble to those in Evi1 cKO mice using Mx1(p)-Cre system, once again demonstrating that the eR1-CreER T2 Tg mouse is an effective new platform for common gene targeting for HSCs in hematopoiesis. Besides gene targeting in child HSCs, the Evi1 gene was excised in fetal and adult HSCs as well.
Interestingly, the phenotypes were far less pronounced. Analysis of LT-HSCs in eR1-CreER T2 Tg; Evi1 fl3/fl3 mice among the fetal, child and adult cohorts uncovered significant differences in the LT-HSC percentages in the child cohort at the end of 1 year after tamoxifen injection, as compared to the fetal and adult cohort ( Figure 7D). Concomitantly, the reduction in platelet count (thrombocytopenia) was exclusively evident in the childhood cohort in comparison to the other two cohorts ( Figure 7E). This result demonstrates that the conditional deletion of Evi1 gene at different developmental stages would give rise to divergent phenotypes.

Discussion
In this report, we describe the successful generation and characterization of an inducible mouse model for highly efficient genetic engineering in HSCs, by employing the recently identified intronic enhancer for Runx1, eR1. The specificity and efficiency of HSC targeting achieved in this model is superior to any other mouse models currently available. A leakage of Cre activity in the absence of inducer has been a concern in the current gold standard, the Whereas Mx1(p)-Cre Tg; Kras-LSL-G12D mice predominantly develop MPD after pIpC injection [23, 24], these mice, surprisingly, also developed MPD without pIpC injection, with 100% penetrance, with disease latency similar to that of pIpC-injected mice [24]. Therefore, the Mx1(p)-Cre Tg mice is not a suitable inducible system as a leukemia model. In clear contrast, the eR1-CreER T2 Tg mouse serves as a long-awaited leakage-free inducible system for HSC-specific genetic engineering in mice.
One caveat, however, is that the CreER T2 is an extremely ligand-sensitive system, where a very small amount of tamoxifen is sufficient to trigger CreER T2 activity. This sensitivity is particularly critical when the CreER T2 system is used for establishing a leukemia mouse model. We observed disease manifestation in uninjected eR1-CreER T2 Tg; Kras-LSL-G12D mice when housed with tamoxifen injected mice in the same cage, indicating crosscontamination via excretions of injected mice. No such diseased mice were observed when tamoxifen injected and non-injected mice were separately housed. Therefore, in order to avoid inadvertent activation of CreER T2 in eR1-CreER T2 Tg mouse, tamoxifen injected and non-injected mice should be housed in separate cages. It is well documented that many types of cancer show age-dependent prevalence. For instance, trisomy 21, the most common congenital chromosomal abnormality, is at high leukemia risk in fetal and childhood hematopoiesis, but not in adult bone marrow [26]. This difference is attributed to the inherent property of background cells, particularly of the HSCs.

Vav(p)-Cre
Here, we observed that the induction of oncogenic Kras in HSCs of different developmental stages led to distinct spectra of malignancies. Interestingly, the induction at child HSCs caused leukemia/lymphomas of myeloid features, whereas the induction in fetal and adult HSCs largely induced T-cell malignancies. eR1-CreER T2 mouse appears to serve as a promising experimental platform to recapitulate clinical observation, namely age-dependent tumorigenesis, thereby allowing for elucidation of its underlying molecular mechanism.
As exemplified above, many scientifically important questions can be addressed using the eR1 enhancer system. Besides being a powerful tool in the HSC biology, we recently found that eR1 is also active in stem cells in stomach [27] and seemingly in multiple other organs.
Extensive studies for individual tissue stem cells are currently underway. The common stemness machinery governing multiple tissue stem cells may be unveiled via the eR1 enhancer-related investigation and experimental tools, and would contribute to tissue engineering and regenerative medicine.

Statistical analysis
Student's two-tailed, unpaired t-test and F test were used to determine statistical significance of two sets of samples. Log-rank (Mantel-Cox) test was used to determine statistical significant of two survival curves. Kruskal Wallis and Mann Whitney t test was performed for comparing two non-parametric data sets. P-values less than 0.05 were considered significant.          eR1-CreER T2 Tg; Evi1 fl3/fl3 mice were sacrificed one month after TMX injection. Control mice are the TMX injected littermates of eR1-CreER T2 Tg; Evi1 fl3/fl3 mice without eR1-CreER T2 transgene.
(A) Genotype PCR for Evi1 gene to detect the presence of floxed or excised exon 3. Genomic DNA from BM cells were amplified by PCR using primers (Supplemental Table 2).