Oncogenic Kras Initiates Leukemia in Hematopoietic Stem Cells

How oncogenes modulate the self-renewal properties of cancer-initiating cells is incompletely understood. Activating KRAS and NRAS mutations are among the most common oncogenic lesions detected in human cancer, and occur in myeloproliferative disorders (MPDs) and leukemias. We investigated the effects of expressing oncogenic KrasG12D from its endogenous locus on the proliferation and tumor-initiating properties of murine hematopoietic stem and progenitor cells. MPD could be initiated by KrasG12D expression in a highly restricted population enriched for hematopoietic stem cells (HSCs), but not in common myeloid progenitors. KrasG12D HSCs demonstrated a marked in vivo competitive advantage over wild-type cells. KrasG12D expression also increased the fraction of proliferating HSCs and reduced the overall size of this compartment. Transplanted KrasG12D HSCs efficiently initiated acute T-lineage leukemia/lymphoma, which was associated with secondary Notch1 mutations in thymocytes. We conclude that MPD-initiating activity is restricted to the HSC compartment in KrasG12D mice, and that distinct self-renewing populations with cooperating mutations emerge during cancer progression.


Introduction
Self-renewal is integral to the malignant phenotype [1]. In principle, the ability of cancer cells to self-renew may be intrinsic to the compartment in which the tumor-initiating mutation occurs, or may be acquired as a consequence of mutations in more differentiated cells. The hematopoietic system has proven highly informative for addressing how cancer-associated mutations and cell of origin interact to establish malignant self-renewing populations. Accumulating evidence supports the idea that many hematopoietic malignancies exist in a hierarchy of differentiation with only a minor population capable of propagating and maintaining the disease in vivo [2]. These cells are termed leukemiainitiating cells or leukemia stem cells (LSCs), and manifest some biologic properties of normal hematopoietic stem cells (HSCs). However, the precise relationship between these populations is uncertain and appears to depend, in part, on both the leukemia subtype and on the effects of specific mutations. For example, overexpressing MLL fusion proteins found in human acute myeloid leukemia transforms both murine HSCs and more differentiated progenitors [3,4]. By contrast, inactivation of the JunB transcription factor must occur in the HSC compartment for initiation of myeloid malignancies [5]. These proof-of-concept experiments underscore the importance of understanding how oncogenes and tumor suppressors that are commonly mutated in human cancers perturb self-renewal and growth control. Importantly, the functional characteristics of LSCs that distinguish them from HSCs and how these properties are modulated by oncogenes are poorly understood.
Chronic and juvenile myelomonocytic leukemias (CMML and JMML) are aggressive myeloid malignancies that are classified as myeloproliferative disorders (MPDs) [9]. Both diseases are characterized by leukocytosis with excess monocytes in blood and bone marrow, and by significant infiltration of malignant myeloid cells into the liver, spleen, and other organs. Hyperactive Ras is strongly implicated in the pathogenesis of JMML and CMML. Somatic NRAS and KRAS mutations are found in ;40% of CMML specimens [10,11], and ;85% of JMML patients have mutations in KRAS, NRAS, NF1, or PTPN11 (reviewed in [12]). The latter two genes encode proteins that regulate Ras-GTP levels. Importantly, children with germline mutations in NF1 or PTPN11 are at markedly increased risk of developing JMML, which argues strongly that deregulated Ras signaling can initiate this MPD [12]. This hypothesis is further supported by studies of Nf1, Kras, and Ptpn11 mutant mice, all of which develop MPDs that resemble JMML and CMML [13][14][15][16]. Mutations that alter other signaling molecules also cause MPD. For example, the BCR-ABL fusion gene is the hallmark of chronic myeloid leukemia, and JAK2 mutations are found in nearly all cases of polycythemia vera (reviewed in [17]). While mutations affecting signaling molecules are found in nearly all types of MPD, additional cooperating mutations are rare. MPDs are therefore genetically straightforward and tractable malignancies for understanding how aberrant signal transduction contributes to cancer by perturbation of stem and progenitor cell fates.
Mice expressing oncogenic Kras G12D in hematopoietic cells develop a fatal MPD with 100% penetrance that is characterized by leukocytosis, splenomegaly, and anemia, with death at approximately 3 mo of age [13,14]. In this system, injecting Mx1-Cre, Kras LSL-G12D mice with polyinosinic-polycytidylic acid (pIpC) induces expression of Cre recombinase, which removes an inhibitory loxP-STOP-loxP (LSL) element and activates the Kras G12D allele (we hereafter refer to Mx1-Cre, Kras LSL-G12D mice that have been treated with pIpC as Kras G12D mice). Mx1-Cre, Kras LSL-G12D mice provide a robust experimental system for investigating how expressing oncogenic Kras from the endogenous promoter affects HSCs and their progeny in vivo. Here, we identify tumor initiating cells in the Kras G12D model of MPD and find that, unlike some myeloid oncogenes, Kras G12D does not confer aberrant selfrenewal properties to committed progenitor cells. However, small numbers of primitive Kras G12D cells can initiate MPD. This population shows excessive proliferation and rapidly dominates multilineage hematopoiesis in vivo. These data indicate that hyperactive Ras signaling is sufficient for the competitive advantage demonstrated by mutant HSCs and further implicate the HSC compartment as critical for therapy of JMML and CMML. Transplanted Kras G12D HSCs efficiently initiate T lineage acute lymphoblastic leukemia/ lymphoma (T-ALL), which is associated with Notch1 mutations and with acquisition of LSC activity in differentiating thymocytes. These results further demonstrate that distinct self-renewing populations can arise through cooperating oncogenic mutations during cancer progression.

Results
A Small Kras G12D Population Dominates Both Primitive and Differentiated Hematopoietic Compartments in Mx1-Cre, Kras LSL-G12D Mice Although basal Mx1-Cre activity is low [18,19], many Mx1-Cre, Kras LSL-G12D mice that are not injected with pIpC ultimately succumb with MPD ( [14] and unpublished data). This observation suggests that HSCs and/or progenitor cells that activate Kras G12D expression have a substantial proliferative advantage in vivo. To assess the kinetics of this process, we analyzed recombination in the myeloid progenitors of young Mx1-Cre, Kras LSL-G12D mice that were not treated with pIpC. Bone marrow cells from 3-5-wk-old animals were plated in methylcellulose medium to enumerate granulocytemacrophage colony forming unit progenitors (CFU-GM), and individual myeloid colonies were genotyped by PCR. Surprisingly, nearly all CFU-GM in untreated animals were recombined as early as 3 wk of age ( Figure 1A). We then examined more primitive populations in Mx1-Cre, Kras LSL-G12D mice crossed to a ROSA26 yellow fluorescent protein (YFP) reporter strain ( Figure 1B) [20]. In these mice, cells expressing Cre are identified by YFP expression. The frequency of YFPþ cells in wild-type (WT) mice was within the expected range of background Cre expression [18,19]. However, we found a much higher incidence of YFP expression in bone marrow cells of Mx1-Cre, Kras LSL-G12D , ROSA26-YFP mice. This result was consistent among all populations analyzed, including the primitive Flk2 À Lin À/lo Sca1 þ c-kit þ (Flk2 À LSK) compartment, which is highly enriched for HSCs [21]. To confirm that YFP expression correlated with Kras G12D expression, we also directly genotyped colonies formed by single Flk2 À LSK cells. This method again revealed a predominance of Kras G12Dexpressing cells in naïve Mx1-Cre, Kras LSL-G12D mice ( Figure  1C). Together, these data are consistent with an advantage for Cre-expressing cells in untreated Mx1-Cre, Kras LSL-G12D mice.

Increased Cell Cycle Entry of Kras G12D HSCs
The apparent outgrowth of Kras G12D Flk2 À LSK cells in mice that were not injected with pIpC suggested that Kras G12D expression might increase proliferation in this compartment. To test this hypothesis, we stained bone marrow cells collected from Kras G12D mice and WT littermates 2 wk after pIpC treatment with antibodies to cell surface proteins and with dyes that stain DNA and RNA (7-aminoactinomycin D [7-AAD] and pyronin Y, respectively). These studies revealed a significant reduction in the number of quiescent Flk2 À LSK cells in Kras G12D animals, which are identified by having a 2n DNA content and low pyronin Y staining (Figure 2A and 2B). Whereas the Flk2 À subset of WT LSK contains roughly 80% cells in the G 0 phase of the cell cycle, Kras G12D Flk2 À LSK are only 50% quiescent. As an initial exploration of mechanisms

Author Summary
Ras proteins act as molecular switches that relay growth signals from outside the cell. This mechanism is often subverted in cancer, and Ras proteins are activated directly by RAS gene mutations in approximately one-third of human malignancies. We have modeled this in mice engineered to have a Ras mutation. These mice develop a disease similar to chronic leukemias in humans called myeloproliferative disorders. It is marked by a fatal accumulation of mature and immature cells in the blood and bone marrow. We investigated whether some or all of these neoplastic cells were immortal. In agreement with the ''cancer stem cell'' hypothesis, we found that immortal cells were extremely rare in the bone marrow of diseased mice. They were found only in the same cell populations that contain normal bone marrow stem cells. However, these cells had high rates of replication and produced large numbers of daughter cells. Furthermore, many mice went on to develop acute lymphoid leukemia after acquiring additional mutations in maturing lymphoid cells. These studies exemplify the evolution of malignant stem cells during cancer progression. They also highlight the importance of rare, long-lived cells in the genesis and, potentially, therapy of highrisk chronic leukemias caused by abnormal Ras proteins.
regulating cell cycle progression, we analyzed D-and E-type cyclin expression by quantitative PCR in sorted Flk2 À LSK cells from Kras G12D and WT mice. Kras G12D Flk2 À LSK cells had significantly higher expression of cyclin D1 ( Figure 2C).

Self-Renewal Is Restricted to Primitive Hematopoietic Cells in Kras G12D Mice
Bone marrow mononuclear cells from Kras G12D mice display a hypersensitive pattern of CFU-GM progenitor growth, which is a cellular hallmark of JMML [12][13][14]. We found that this abnormal CFU-GM activity resides primarily in the common myeloid progenitor (CMP) compartment, and that these cells demonstrate enhanced proliferation in vivo ( [22] and unpublished data). To investigate if this population could initiate and maintain MPD, we collected Lin À/lo Sca1 À c-kit þ CD34 þ FccRII/III À CMPs from 5-wk-old Kras G12D mice and WT littermates by FACS, and transferred 10,000 of these cells into lethally irradiated recipients with 10 6 WT marrow cells for radioprotective support. Transplanted tester cells, recipient cells, and support cells were marked by expression of different isoforms of CD45, allowing them to be distinguished by flow cytometry. Transplanted Kras G12D or WT CMPs demonstrated robust day 8 spleen colony-forming unit (CFU-S 8 ) activity, with colony size somewhat larger for Kras G12D input CMPs ( Figure 3A). However, we detected less than 0.1% of circulating myeloid cells derived from transplanted Kras G12D or WT CMPs 1 mo after transplantation ( Figure 3B and 3C). As expected from previous studies [23,24], Kras G12D and WT CMPs made minor contributions to the circulating B cell compartment with a statistically insignificant trend towards greater B cell production from Kras G12D cells. Taken together, these data indicate that Kras G12D CMPs do not initiate a hematologic disease. By contrast, transferring 500 Flk2 À LSK cells from Kras G12D animals into lethally irradiated recipients rapidly resulted in durable multilineage reconstitution. Kras G12D cells dominated the T cell and B cell compartments sooner and to a higher degree than the progeny of WT Flk2 À LSK cells in control animals. The myeloid series demonstrated a more variable time course, but Kras G12D derived cells also eventually out-competed WT cells ( Figure 4).
Recipients of Kras G12D Flk2 À LSK cells that were euthanized 3 mo after transplantation had mild to moderate MPD, manifested as leukocytosis and splenomegaly with myeloid and erythroid infiltration ( Figure 5). As discussed below, we also found that all recipients of Kras G12D Flk2 À LSK cells developed T-ALL 2-4 mo after adoptive transfer. Although early mortality from T-ALL precluded analyzing recipient mice beyond 3 mo, Kras G12D Flk2 À LSK cells recapitulate the essential features of MPD seen in the original Mx1-Cre, Kras LSL-G12D model [13,14].

HSCs Are Less Frequent in Kras G12D Mice but Have Enhanced Repopulating Ability
We next asked how Kras G12D expression affects HSC function. Immunophenotypic analysis revealed a 2-fold reduction in the number of marrow Flk2 À LSK cells in Kras G12D mice 2 wk after pIpC injection ( Figure 6A). This was mostly offset by an increased number of splenic Flk2 À LSK cells. The reduction in marrow Flk2 À LSK cells persisted in older animals ( Figure S2). We also performed limit dilution studies to assess functional HSC activity. In these experiments, lethally irradiated recipients received decreasing numbers of whole bone marrow cells from either Kras G12D mice or WT littermates that had been injected with pIpC 2 wk earlier. Recipients were bled monthly, and flow cytometry was performed to assess whether CD45.1 þ donor cells were able to provide durable (.2 mo) multilineage (myeloid, B cell, and T cell) engraftment. These studies demonstrated a striking 10fold decrease in the number of long-term repopulating stem cells in Kras G12D animals compared to the WT littermate controls ( Figure 6B; Table S1).
Progression of MPD despite a reduction in the size of the HSC compartment suggests an increased production of mature cells by each Kras G12D HSC. To address this possibility, we examined the patterns of reconstitution from either Kras G12D or WT HSCs in mice that received the limit dilution dose (1 3 10 5 Kras G12D and 1 3 10 4 WT bone marrow cells). As ;50% of these recipients were engrafted with donor cells, the Poisson distribution predicts that approximately 70% of the engrafting mice received a single HSC and ;24% received two HSC. By 2 mo after transplantation, donor Kras G12D marrow cells made a markedly greater contribution to recipient hematopoiesis than WT cells ( Figure 6C).
We were able to compare repopulation of the stem cell compartment by WT and Kras G12D HSCs in a few lethally irradiated recipient mice before the onset of T-ALL. In these experiments, lethally irradiated recipients (CD45.2) were reconstituted with equal numbers of Kras G12D (CD45.1) and WT (CD45.1/CD45.2) Flk2 À LSK cells, as well as 10 6 CD45.2 whole bone marrow cells (CD45.2) for radioprotection. Whereas only one of six recipients euthanized 2 mo after transplantation showed a clear bias towards Kras G12D -derived Flk2 À LSK cells, mice that survived for 3 mo without evidence of diffuse T-ALL demonstrated an overwhelming bias towards Kras G12D -derived Flk2 À LSK cells, myeloid progenitors, and mature myeloid cells ( Figure S3).

T-ALL Arises from Transplanted Kras G12D HSCs in Thymocytes That Acquire Somatic Notch1 Mutations
Recipients that were injected with Kras G12D Flk2 À LSK cells, either alone or in a 1:1 ratio with WT Flk2 À LSK cells, uniformly became moribund 8-14 wk after transplantation. 3.14. p-Values by unpaired t-test are indicated: ***, p , 0.001; **, p , 0.01; *, p , 0.05. (C) RNA from doubly sorted Flk2 À LSK cells was isolated, and quantitative PCR performed on cDNA to test expression levels of selected cyclins (n ¼ 3 or 4 as shown; bar shows geometric mean). Results are expressed as fold change in expression compared to WT Flk2 À LSK cells, after normalization to b-actin expression. These results represent three or four independent experiments as shown, each performed with pooled bone marrow from three to five animals. The geometric mean of cyclin D1 expression is 2.4-fold over WT (95% confidence interval 1.2-4.9). Purity of sorted cells is shown in Figure S1. doi:10.1371/journal.pbio.1000059.g002 Examination of euthanized mice revealed massive thymic enlargement with an arrest in T cell development at the CD4/ CD8 double positive stage, and variable infiltration of blast cells within the liver, spleen, and bone marrow ( Figure 7A). Importantly, two of three animals from the bone marrow limit dilution transplantation assay that were engrafted with a single Kras G12D repopulating unit developed T-ALL that was identical to that seen in animals repopulated with 500 Kras G12D Flk2 À LSK cells (unpublished data).
Next, we asked if T-ALL arose within the bone marrow or the thymus of mice transplanted with Kras G12D HSCs. To do this, we took advantage of our prior observation that recipients conditioned with sublethal irradiation fail to engraft with Kras G12D bone marrow [13]. Therefore, sublethally irradiated recipients can exclusively select for hematologic malignancies with more aggressive biologic behavior. We isolated bone marrow cells and thymocytes from primary recipients of Kras G12D Flk2 À LSK cells that developed T-ALL, and injected each population into sublethally irradiated secondary recipients ( Figure 7B). As expected, no animals demonstrated multilineage engraftment or MPD. Animals transplanted with thymocytes quickly succumbed with an identical T-ALL as primary recipients; however, none of the mice transplanted with bone marrow developed leukemia. Thus, whereas bone marrow-derived Kras G12D HSCs efficiently give rise to T-ALL, the T-ALL LSC population is initially restricted to the thymus in primary recipient mice.
These results suggested that one or more secondary mutations might have developed in a novel T-lymphoid clone. Somatic NOTCH1 mutations are common in human and murine T-ALL [25][26][27][28]. To determine if a similar mechanism might contribute to the evolution of Kras G12D HSCs to T-ALL LSCs, we performed Western blot analysis to detect cleaved (activated) Notch1 protein. Cleaved Notch1 was observed in thymocytes from diseased primary recipients, but not in bone marrow cells ( Figure 7C), a finding that is consistent with the secondary transplant data. Direct sequencing around the PEST domain uncovered frameshift mutations in exon 34 of Notch1 in thymocytes from five of six animals transplanted with Kras G12D Flk2 À LSK cells that developed T-ALL, but no mutations in thymocytes from control mice that received WT HSCs alone and remained well. The presence of a somatically acquired Notch1 mutation in a large fraction of the tumor provides a molecular indication of clonality.

Thymic Abnormalities in Kras G12D Mice
The propensity of Kras G12D HSCs to generate T-ALL led us to investigate the effects of Kras G12D expression in early T-  lineage cells. Flow cytometry of the bone marrow revealed no expansion of Lin À Flk2 þ IL-7Ra þ c-kit int Sca1 int common lymphoid progenitors ( Figure 8A) [29]. However, Kras G12D mice demonstrated consistent thymic enlargement compared to age-matched littermate controls, even prior to the onset of T-ALL ( Figure 8B). Immunophenotyping of primary thymocytes demonstrated an essentially normal distribution of CD4þ and CD8þ expression, with a slight trend toward an increased number of CD4/CD8 double negative (DN) cells (after excluding Mac1þ and Gr1þ infiltrating myeloid cells) ( Figure 8B). By contrast, further examination of the DN compartment using the cell surface markers CD44 and CD25 uncovered skewed development in Kras mutant mice ( Figure  8C). Together, these data demonstrate that oncogenic Kras G12D perturbs thymic homeostasis, particularly in early stages of thymocyte maturation.

Discussion
We find that oncogenic Kras G12D expression in HSCs confers a strong in vivo growth advantage, increases proliferation, and results in MPD and T-ALL. In MPD, as in normal marrow, stem cell activity is restricted to the Flk2 À LSK population, which represents less than 0.1% of nucleated marrow cells. While pathologic behaviors of more mature cells may contribute to tissue infiltration, anemia, and organomegaly, self-renewal is confined to this very primitive population. Therefore, hyperactive Ras signaling promotes excess proliferation in multiple hematopoietic compartments without immortalizing non-self-renewing cells. Similar data have been described in murine models of MPD based on BCR-ABL overexpression or loss of JunB, both of which also deregulate cytoplasmic signaling networks [5,30]. By contrast, recent experiments have provided direct evidence that some oncogenic transcription factors allow committed myeloid progenitors to acquire self-renewal ability [3,4]. Taken together, studies of myeloid oncogenes performed to date support the general idea that mutations that predominately alter cytoplasmic signaling networks and those that affect transcription factors controlling cell fate decisions comprise discrete complementation groups for the fully transformed phenotype [31,32].
Our limit dilution transplantation data demonstrate that oncogenic Kras confers a dramatic growth advantage in the HSC compartment. Under stringent conditions in which the contribution of a single WT HSC can barely be detected, the progeny of one (or at most three) Kras G12D HSC comprise a substantial fraction of the hematopoietic compartment. These studies provide direct experimental evidence that the outgrowth of malignant cells in MPD can be attributed to hyperactive Ras signaling in HSC. Our data suggest a pathogenic model in which JMML or CMML is initiated by a somatic mutation that deregulates Ras signaling in a single HSC. This idea is consistent with limited data from human patients and xenograft studies that implicate the HSC as the cell of origin for JMML [33][34][35][36][37][38].
To begin to address the mechanism by which mutant HSCs outgrow their WT counterparts, we analyzed the cell cycle in Kras G12D Flk2 À LSK cells and found they are preferentially in cycle. The overexpression of cyclin D1 in Kras G12D cells we observed is consistent with many prior studies in cultured cell lines engineered to overexpress oncogenic Ras [39,40]. Intriguingly, HSCs in mice lacking D-type cyclins demonstrate severe proliferative defects and accumulate in the G 0 and/or G 1 phases of the cell cycle [41]. If increased cyclin D1 levels conversely result in excessive proliferation of HSC, then MPD in Kras G12D mice may be mediated in part by cyclin D1, similar to the requirement for cyclin D1 in a model of Ras-mediated breast cancer [42].
We demonstrate a substantial early growth advantage of HSCs that express Kras G12D ; however, it is also possible that oncogenic Ras expression has a negative long-term impact on HSC function. Increased proliferation or oncogenic stress may ultimately detract from self-renewal capacity. Reduced HSC fitness was observed in Pten À/À mice, in which phosphatidylinositol 3-kinase signaling is hyperactive [43,44]. Similar effects of Kras G12D are suggested by the reduced numbers of HSCs in Kras G12D mice, although this finding could also reflect changes in the composition of the Flk2 À LSK population or cell-extrinsic effects related to alteration of the marrow microenvironment. The rapid demise of primary Kras G12D mice from MPD, and of transplant recipients from T-ALL, precluded serial transplantation experiments to test the long term fitness of Kras G12D HSC. In addition, the 5-fold discrepancy between HSC numbers that were measured by flow cytometry versus limit dilution transplantation suggests a defect in engraftment of Kras G12D HSC. This idea is consistent with a prior report in which retroviral transduction of mutant NRAS appeared to reduce engraftment potential [45], and with extensive data showing that proliferating HSCs fail to engraft efficiently [46][47][48][49][50].
Despite the reduced number of HSC in the bone marrow of Kras G12D mice, our data are not entirely consistent with the idea that Kras G12D is a cell-intrinsic negative regulator of HSC self-renewal. If it were, we would expect specific loss of Kras G12D cells and outgrowth of WT or Kras LSL-G12D cells, because conditional models using Mx1-Cre typically retain a small pool of cells with the unrearranged locus [51,52]. However, we observed preferential retention of Kras G12D cells, even within the diminishing Flk2 À LSK compartment. This observation suggests that residual WT HSCs are unable to compensate for the reduced HSC number. Therefore, we favor the hypothesis that the reduction in HSC number in Kras G12D mice reflects a disordered bone marrow microenvironment with reduced supportive capacity rather than a purely cell-intrinsic effect of Kras G12D in HSC.
The natural history of hematologic disease was different in primary Kras G12D mice than in recipients of transplanted Kras G12D HSC. A subtle but important finding is that MPD is established earlier in primary Kras G12D mice than in transplanted recipients. (Figure 5 and [13]; also see [16,53]). There are several possible explanations for this observation. The hematopoietic microenvironment in young mice may be more permissive for MPD than the irradiated bone marrow of an adult. Additionally, pIpC administration in Mx1-Cre mice may quickly create a field of Kras G12D myeloid progenitors that contributes to the rapid evolution of MPD in primary mice through cytokine-mediated autocrine and/or paracrine mechanisms [54]. It is also possible that nonhematopoietic stromal cells in Mx1-Cre, Kras LSL-G12D mice express K-Ras G12D and contribute to the rapid onset of MPD in primary Kras G12D mice.
The kinetics of MPD development relate directly to the high frequency of T-ALL we observed in transplanted recipients as compared to primary Kras G12D mice. The apparent incidence of T-ALL is highly subject to selection bias, because animals that die from MPD cannot be evaluated for subsequent emergence of T-ALL. For example, we have observed ;10%-15% of Kras G12D mice develop T-ALL on an inbred C57BL/6 strain background (unpublished data), but median time to death from MPD is shorter than the typical latency of T-ALL, and lymphoid tumors exclusively appear in mice with a relatively late onset of MPD. Similarly, we have not observed spontaneous T-ALL in F1 (C57BL/6 3 129Sv/Jae) mice, which die from MPD at a younger age than the C57BL/6 strain described here ( [13] and unpublished data). In transplant recipients, aggressive T-ALL arose in mice that also had evidence of underlying MPD that was not yet severe enough to kill the animal. Together, these observations suggest that the attenuation of MPD in transplant recipients was central to the apparent increase in the incidence of T-ALL in the transplant setting.
Cooperation of hyperactive Ras and deregulated Notch signaling in T-ALL has recently been shown [55][56][57]. Our data extend these studies by demonstrating the remarkable efficiency with which Kras G12D HSCs can initiate T-ALL, and delineating how multiple cell types may participate in the stepwise acquisition of oncogenic mutations in hematologic cancers. In the Mx1-Cre, Kras LSL-G12D model, T-ALL is initiated by oncogenic Kras expression in HSC, but full transformation occurs when cooperating Notch1 mutations arise in a T- lineage cell. In this sense, both Kras G12D HSCs and the fully transformed thymocytes can be considered different types of malignant stem cells with distinct leukemogenic potentials.
One potential implication of these results is that the initiating Kras mutation creates conditions favorable for acquisition of cooperating mutations by increasing the size of susceptible lymphoid progenitor pools and/or conferring resistance to apoptotic signals during thymic selection. Kras G12D appears to most greatly affect the DN population that is characteristically undergoing TCR rearrangement, selection, and proliferation [58,59]. Interestingly, K-Ras G12D protein expression may substitute for the pre-T cell receptor rearrangement at this critical checkpoint, thereby allowing propagation of thymocytes that would normally be edited [60]. Consistent with this idea, a patient with impaired lymphoid homeostasis and multiple lymphoid malignancies was recently reported to have a germline NRAS G13D mutation, and oncogenic NRAS suppressed apoptosis of lymphocytes after cytokine withdrawal [61]. By contrast, Kindler et al. recently reported reduced thymic cellularity in Mx1-Cre, (B) Primary recipients (n ¼ 5) received 500 Kras G12D Flk2 À LSK cells with or without an equal number of WT Flk2 À LSK cells after lethal irradiation (950 rad). These primary recipients were euthanized 2-3 mo later, and 10 6 bone marrow cells or thymocytes were transferred into sublethally irradiated (450 rad) secondary recipients (two per primary mouse). Sublethal irradiation selectively permits transfer of acute leukemia but not Kras G12D HSCs or MPD [13]. Thymocytes, but not bone marrow cells, transferred T-ALL. (C) Cell lysates from thymocytes and bone marrow of animals euthanized 3 mo after transplantation with Kras G12D or WT HSCs were blotted with an antibody specific for cleaved Notch1. Three independent primary recipients are shown. Sequence analysis demonstrates frameshift mutations near the PEST domain of Notch1 in thymocytes from five of six mice that received Kras G12D HSC, but not in recipients receiving WT HSCs alone (reference sequence from GenBank [http://www.ncbi.nlm.nih.gov/Genbank] accession number NM_008714). doi:10.1371/journal.pbio.1000059.g007 Kras G12D mice [56]. We speculate that the proliferative effects of K-Ras G12D in the T cell compartment were obscured in those studies by the short interval between pIpC injection, which induces systemic interferon production, and histologic analysis.
The idea that patients may harbor a variety of genetically distinct LSC is consistent with studies of patients with chronic myeloid leukemia in blast transformation [62], and has important therapeutic implications. The need to eliminate partially transformed but self-renewing cells, like Kras G12D HSCs, will depend on their propensity to initiate a life-threatening disease. Targeted therapies that are directed against onco-proteins such as K-Ras G12D will effectively eliminate premalignant clones only if the targeted lesions are initiating rather than secondary mutations. For example, inhibition of Notch signaling is an attractive therapeutic strategy for T-ALL that is being investigated in the clinic. However, if these cancers arise from aberrant HSCs that do not contain a NOTCH1 mutation and are not eradicated by treatment, relapse could occur through the acquisition of distinct cooperating mutations in a self-renewing preleukemic population. Consistent with this idea, studies of human T-ALL suggest that NOTCH1 mutation occurs as a secondary mutation in at least some cases, with some patients developing recurrent disease having distinct NOTCH1 alleles [63].
Finally, our data have implications for understanding the nature of cancer stem cell populations in nonhematopoietic malignancies. KRAS is the most frequent target of dominant oncogenic mutations in human cancer, and it is particularly important in carcinomas of the lung, pancreas, and colon. Analogous cancers arise in strains of mice expressing conditional oncogenic Kras alleles in these tissues [64][65][66][67][68]. Importantly, whereas oncogenic Kras expression efficiently initiates tumorigenesis in murine lung and pancreas, colon cancer is observed only when the tumor suppressor Apc is inactivated as well [68]. These data are consistent with studies of human patients, which imply that KRAS mutation occurs early in pancreatic cancer but typically after APC mutation in colon carcinoma [69][70][71]. Lung cancer in Kras G12D mice appears to be initiated in a distinct bronchio-alveolar stem cell population [72]. On the basis of these observations and our data, we speculate that, like HSCs, cells initiating pancreatic and lung cancer will possess inherent self-renewal potential, and that KRAS mutations only contribute to colon tumorigenesis in cells that have already acquired a mutation that enhances self-renewal. Uncovering specific proteins and pathways that are essential for the self-renewal and survival of Kras mutant cancer stem cell populations may reveal novel targets for therapeutic intervention in a variety of human cancers.

Materials and Methods
Ethics statement. All animals were handled in strict accordance with good animal practice as defined by the relevant national and local animal welfare bodies, and all animal work was approved by the institutional animal research committee (University of California, San Francisco IACUC).
Mouse strains. Animals were housed in a barrier facility at the University of California, San Francisco. All mice were of the C57BL/6 strain. As described [13], 21-d-old Mx1-Cre, Kras LSL-G12D were injected IP with 250 lg pIpC (Sigma). ROSA26 LSL-YFP reporter mice were a gift from C. Lowell. Ptprc (CD45) congenic mice were from Jackson Labs.
Prior to sorting, c-kit þ cells were enriched with anti-CD117 microbeads and an AutoMACS (Miltenyi). Cell sorting was performed on a FACSAria and flow cytometry on an LSRII, both using FACSDiva software (BD). Data were analyzed using FlowJo software (TreeStar).
Transplantation. Recipients received a single fraction of 950 rads for lethal irradiation, or 450 rads for sublethal irradiation, from a cesium source. Donor cells were prepared in 100 to 200 ll of FSB, and injected retro-orbitally into anesthetized mice. Recipients received water with neomycin and polymyxin for 2 wk. Blood counts were monitored monthly using a Hemavet 950FS (Drew Scientific).
Limit dilution was performed as described [73]. Lethally irradiated recipients were transplanted with 1 3 10 4 , 3 3 10 4 , or 1 3 10 5 nucleated bone marrow cells along with 10 6 unfractionated WT bone marrow cells for radioprotection. Peripheral blood was analyzed monthly by flow cytometry; mice with detectable engraftment in myeloid, B and T lineages 2 mo after transplantation were scored as positive and L-Calc software (StemCell Technologies) was used for statistical analysis.
For the CFU-S 8 assay, lethally irradiated mice received 10 5 CMPs without support cells. Spleen colonies were observed 8 d later by gross examination at harvest and after 48 h fixation in 10% formalin. Sections were stained with hematoxylin and eosin.
Quantitative real-time PCR. The assay was performed as described [50]. Flk2 À KLS cells from animals pooled by genotype were doublesorted directly into RNA binding/lysis buffer from the RNEasy kit (Qiagen), and total RNA was extracted per instructions. First strand cDNA synthesis was performed using a SuperScript III kit (Invitrogen) per manufacturer's instructions. Reactions were performed in an ABI-7900 sequence detection system using SYBR green according to manufacturer's instructions (Applied Biosystems). Each amplification was performed in 10 ll with a template cDNA equivalent of 100 sorted HSCs. Each sample was tested in triplicate with each primer pair, and normalized to b-actin expression. Due to limiting numbers of doubly sorted cells, final cell purity was not analyzed; however the staining characteristics of c-kit-enriched and singly sorted cells are presented in Figure S1.
Poisson statistics. To analyze repopulation by a single HSC, we identified a cohort of mice receiving a cell dose yielding engraftment in only 50% of recipients. The Poisson distribution indicates that the probability of a mouse receiving k HSCs is given by f ðk; nÞ ¼ n k e Àn k! ; where n is the average HSC number per mouse. The average that yields k ¼ 0 HSC at a rate of 0.5 is given by 0.5 ¼ e Àn , which is solved to give n ¼ 0.693. Using this value for the average number of HSC per mouse, the Poisson distribution can be used to estimate the likelihood of any mouse receiving a given number of HSC: f(0) ¼ e Àln 2 ¼ 0.5; f(1) ¼ (ln 2)e Àln 2 ¼ 0.346; f(2) ¼ 1/2 (ln 2) 2 e Àln 2 ¼ 0.120; f(3) ¼ 1/6 (ln 2) 3 e Àln 2 ¼ 0.028. The proportion of mice expected to receive four or more HSCs is the remainder, 1 À (0.5 þ 0.346 þ 0.12 þ 0.028) ¼ 0.006, or 0.6%.