Oncogenic RAS Enables DNA Damage- and p53-Dependent Differentiation of Acute Myeloid Leukemia Cells in Response to Chemotherapy

Acute myeloid leukemia (AML) is a clonal disease originating from myeloid progenitor cells with a heterogeneous genetic background. High-dose cytarabine is used as the standard consolidation chemotherapy. Oncogenic RAS mutations are frequently observed in AML, and are associated with beneficial response to cytarabine. Why AML-patients with oncogenic RAS benefit most from high-dose cytarabine post-remission therapy is not well understood. Here we used bone marrow cells expressing a conditional MLL-ENL-ER oncogene to investigate the interaction of oncogenic RAS and chemotherapeutic agents. We show that oncogenic RAS synergizes with cytotoxic agents such as cytarabine in activation of DNA damage checkpoints, resulting in a p53-dependent genetic program that reduces clonogenicity and increases myeloid differentiation. Our data can explain the beneficial effects observed for AML patients with oncogenic RAS treated with higher dosages of cytarabine and suggest that induction of p53-dependent differentiation, e.g. by interfering with Mdm2-mediated degradation, may be a rational approach to increase cure rate in response to chemotherapy. The data also support the notion that the therapeutic success of cytotoxic drugs may depend on their ability to promote the differentiation of tumor-initiating cells.


Introduction
Acute myeloid leukemia (AML) is a clonal disease with a heterogeneous genetic background. Besides age, cytogenetic alterations and molecular lesions such as mutations in the FLT-3 or nucleophosmin genes play a pivotal role in predicting treatment response ( [1]; reviewed in [2,3]). AML is treated with induction and post-remission chemotherapy, frequently including high-dose cytarabine [4]; reviewed in: [5]. Mutations in NRAS and KRAS protooncogenes (resulting in ''oncogenic'' RAS) occur in approximately 20% of AML cases (reviewed in: [6]).
It has been suggested that leukemic transformation depends on the occurrence of two genetic lesions in a susceptible progenitor cell. Class I mutations that affect genes encoding receptor tyrosine kinases (Flt-3 or Kit) or RAS are thought to induce myeloid proliferation. Class II lesions affect transcription factors such as nucleophosmin, C/EBPa, AML-ETO, MLL-ENL, PML-RARa and block differentiation (e.g. [7]; reviewed in: [3]). Supporting this notion, oncogenic RAS alone induces a myeloproliferative state in murine models [8][9][10][11][12] and in cooperation with nuclear oncogenes such as PML-RARa induces acute leukemia [13].
A previous landmark study showed that AML patients benefit from high-dose cytarabine as post-induction therapy, which has subsequently become the standard consolidation therapy in AML [4]. Using AML samples taken from this study, we have previously shown that AML patients harboring oncogenic RAS show significantly less cumulative incidence of relapse upon treatment with high-dose cytarabine in the post-induction chemotherapy (best group), when compared to AML patients with oncogenic RAS treated with low-dose cytarabine (worst group). In contrast, dose escalation had a much weaker effect on the response to cytarabine in patients that harbour wild type RAS (intermediate groups [26]). These data suggested that there is a genetic interaction between the dose of cytarabine and the presence of oncogenic RAS. Importantly, multivariate analysis revealed that the interaction of RAS with cytarabine dose escalation was independent of cytogenetic status of the leukemic blasts, suggesting that oncogenic RAS affects the response of AML blasts to cytarabine [26]. Moreover, since the beneficial effect of high-dose cytarabine with oncogenic RAS is observed especially in the post-induction therapy, the findings suggested that the number of clonogenic leukemiainitiating cells was reduced as the result of an interaction between oncogenic RAS and cytarabine.
To better understand the interaction of RAS with cytarabine, we expressed oncogenic RAS in primary mouse bone marrow stem cells that had been immortalized by a MLL-ENL oncogene [27]. In this tissue-culture system, MLL-ENL acts as the class II mutation, whereas supplementation of the medium with growth factors presumably substitutes for a class I mutation. We find that oncogenic RAS in combination with DNA-damaging agents such as cytarabine decreases the clonogenic potential of these cells and induces a myeloid differentiation program in a DNA damage checkpoint-and p53-dependent manner. Our data suggest that in AML patients with oncogenic RAS, high-dose cytarabine therapy is effective since it promotes the differentiation of tumor-initiating cells.

Immortalized bone marrow stem cells expressing oncogenic RAS do not show enhanced proliferation, apoptosis or senescence in response to cytarabine
We infected mouse bone marrow cells that had been immortalized with a conditional MLL-ENL-ER oncogene [27] with either control retroviruses (empty vector: ''EV'') or retroviruses expressing an oncogenic Ha-RasV12 protein (generating EV and Ras cells; Figure 1A). Immunoblots confirmed that Ras cells expressed elevated levels of the Ha-Ras protein and displayed activation of Ras proteins as determined by binding to a GST-Raf protein and phosphorylation of Erk, a downstream effector of Ras ( Figure 1B). Expression of RAS did not significantly affect the expression of MLL-ENL-ER ( Figure S1) and of meis1 and hoxa9, which are critical target genes of MLL-ENL [28] ( Figure 1C). Furthermore, withdrawing 4-hydroxy-tamoxifen (OHT) to switch off the function of the MLL-ENL-ER chimera resulted in differentiation of both control (EV) and Ras cells as shown by the increased expression of markers of monocytic (itgam encoding Mac 1) and granulocytic (ly6g encoding the Gr1 antigen) differentiation ( Figure 1D) and in downregulation of meis1 and hoxa9 expression ( Figure 1C), demonstrating that oncogenic RAS did Control and Ras cells were treated with 50 mM cytarabine for the indicated times. Cell lysates were either incubated with GST-Raf bound to glutathione beads and immunoblots of bound proteins probed with pan-Ras antibody (''active Ras'') or cells lysates were probed with antibodies against phosphorylated Erk (pTyr-204) and Ras. Cdk2 served as loading control in this panel and all subsequent immunoblots. (C) MLL-ENL-ER target genes are not influenced by oncogenic RAS. Control and Ras cells were cultured for 12 days in the presence or absence of 4-OHT. Expression of meis1 and hoxa9 was analysed by RQ-PCR. Each column represents the mean6SD in this panel and all subsequent RQ-PCRs. The double asterisk represents statistical significance (p,0.01) of differences between cells cultured in the presence and absence of 4-OHT, respectively. (D) Oncogenic RAS does not abrogate the differentiation due to withdrawal of 4-OHT. Control and Ras cells were cultured for 12 days in the presence or absence of 4-OHT and the expression of the differentiation markers ly6g (encoding Gr1) and itgam (encoding Mac1) was measured by RQ-PCR. Statistical significance refers to differences between cells cultured in the presence or absence of 4-OHT, respectively. doi:10.1371/journal.pone.0007768.g001 not influence the expression of these markers independent of MLL-ENL-ER.
Treatment with cytarabine (AraC) had no significant effect on expression of total or active Ras in either control or Ras cells ( Figure 1B). In response to treatment with either low (10 nM) or elevated (100 nM) concentrations of cytarabine, both control and Ras cells showed a comparable decrease in cell number (Figure 2A). Furthermore, cytarabine inhibited DNA replication in both cell types to a similar extent, as measured by BrdU incorporation in a two-dimensional FACS analysis ( Figure 2B). This analysis also showed that both cell types underwent apoptosis in response to either a transient (three hour pulse) or prolonged treatment (24 hours) with cytarabine (detected as cells with a subG1 DNA content in Figure 2B). Addition of cytarabine led to an increase in expression of two marker genes of senescence, ink4b and dec-1 [29], in both control and Ras cells; possibly, therefore, cytarabine can induce senescence in MLL-ENL cells ( Figure 2C). However, the cells were only weakly positive for a second marker of senescence, acidic b-galactosidase, and this was not affected by cytarabine ( Figure 2D), suggesting that induction of senescence does not account for the selective loss of clonogenicity observed in Ras cells (see below).

Ras cells show reduced colony formation potential and enhanced checkpoint activation after treatment with cytarabine
We next tested the ability of control and Ras cells to form colonies in semisolid medium after treatment with cytarabine. Cells were pre-treated in suspension for 24 hours with cytarabine at different concentrations, then plated in semisolid medium in the absence of the drug. Colony formation was scored after four days. No apparent difference was found between the two cell types in the absence or in the presence of lower doses of cytarabine (10 nM) ( Figure 2E); pre-incubation with a higher concentration of 100 nM cytarabine led to a moderate (3.5fold) decrease in the colony forming capacity of control cells; in contrast, the same treatment largely abolished the clonogenic potential of Ras cells ( Figures 2E,F). Thus, transient exposure to cytarabine reduces the clonogenic potential of Ras cells more strongly than that of control cells ( Figure S2A). Slightly elevated concentrations of cytarabine were required to suppress colony formation in a second, independently derived clone of MLL-ENL cells ( Figure S2B). Importantly, expression of RAS also enhanced the sensitivity to cytarabine in these cells, confirming that the difference in sensitivity was due to expression of RAS and not due to variations during the infection and selection of cells in culture ( Figure S2B).
Since we observed no difference between control and Ras cells in the total level of apoptosis or senescence upon exposure to cytarabine, we considered two mutually non-exclusive possibilities to account for the differences observed in the clonogenic assays: First, MLL-ENL and Ras cells -despite being clonal isolatesmight be heterogeneous with respect to clonogenic potential. Therefore, the response to cytarabine that is measured in the total cell pool may be determined by the majority of non-clonogenic cells and may not reflect the response of a potentially small pool of clonogenic cells; consistent with these suggestion, we observed that only a fraction of both cell types expressed the monocyte/ macrophage marker Mac1 or c-kit, which is a marker of stem and progenitor cells (Figures S3A,B) and that the clonogenic population of both control and Ras cells could be enriched by isolating either a Mac1-depleted or a c-kit-enriched population ( Figure S3C). Therefore, in both control and Ras cells, the cells enriched for a more immature phentype revealed similar colony forming capacity, arguing that they did not express a qualitative difference with regard to colony formation. Second, the clonogenic potential might be reduced by a process that is distinct from both apoptosis or senescence; this possibility was addressed in experiments described further below.
To identify the mechanisms underlying the impaired clonogenicity of Ras cells, we wondered whether either of two events that occur in response to oncogenic activation of Ras could account for this observation: first, oncogenic Ras induces the expression of p16 Ink4a and p19 Arf , leading to increased levels of p53, and p21 Cip1 [19,30]. Consistent with these observations, Ras cells expressed elevated levels of p16 Ink4a and p19 Arf , and this was independent of the exposure to the DNA-damaging agent cytarabine; consequently, levels of p21 Cip1 and p53 were elevated in Ras cells even in the absence of cytarabine ( Figures 3A,B). Second, oncogenic Ras can induce a DNA damage response in primary cells, which leads to activation of the Atm and Atr checkpoint kinases [24,25,31]. We did not observe elevated levels of phosphorylated Chk1 or H2a.x, which are targets of the Atm or Atr kinases in immunoblots of either control or Ras cells before treatment with cytarabine ( Figures 3C,D); furthermore, immunofluorescence did not reveal a difference in phosphorylated Atm in untreated cells ( Figure 3E). However, cytarabine strongly induced phosphorylation of Chk1 in Ras cells, whereas the phosphorylation of this protein was hardly detectable in control cells treated with cytarabine. Unphosphorylated Chk1 was not differentially expressed between control and Ras cells ( Figure 3C). This difference in checkpoint responses was not due to a reduced rate of DNA replication in control cells, which would lead to a reduced incorporation of cytarabine ( Figure 2B). Ras cells also displayed strongly elevated levels of phosphorylated H2a.x and Atm after incubation with cytarabine relative to control cells ( Figures 3D,E). Consistent with these observations, exposure to cytarabine led to a further increase in levels of p21 Cip1 in Ras cells, whereas induction of p21 Cip1 was hardly detectable in control cells ( Figure 3A). The data show that oncogenic RAS cooperates with cytarabine in activation of DNA damage-dependent checkpoints.
Both mutation and enhanced expression of RAS genes have been reported in AML [32]. To determine whether elevation of checkpoint responses is specific for oncogenic Ras or whether it reflects the elevated levels of total Ras present in Ras cells, we determined the level of Chk1 phosphorylation in two pools of MLL-ENL cells that had been infected with retroviruses expressing wild type Ha-RAS ( Figure S4). Enhanced phosphorylation of Chk1 was observed in cells expressing mutated RAS, but also -albeit to a significantly lesser extend -in cells expressing high-levels of wild type RAS. Most likely, therefore, both, the oncogenic mutation of RAS and -more weakly -the enhanced expression contribute to the altered checkpoint activity of Ras cells.
DNA-damaging agents such as daunorubicine and etoposide also reveal differential effects in Ras cells In order to elucidate if the observed effects on DNA-damage checkpoints and the impaired capacity to form colonies in vitro were specific for cytarabine, we investigated the effects of the topoisomerase II inhibitors daunorubicine and etoposide on colony formation and DNA-damage checkpoints in control and Ras cells and compared this to cytarabine. Both control and Ras cells showed a dose-dependent phosphorylation of Chk1 ( Figure 4A) and inhibition of colony formation ( Figure 4B) in response to cytarabine. Consistent with our previous data, Ras cells showed an enhanced checkpoint response at all cytarabine concentrations tested (for a quantitation, see Figure S5A) and were more sensitive with respect to colony formation ( Figure S5B  and Ras cells were treated with daunorubicine and etoposide, Ras cells appeared more sensitive towards incubation with these drugs with regard to phosphorylation of Chk1 and colony formation, respectively (Figures 4C and D: daunorubicine; 4E and F: etoposide). Thus, the differential response of Ras vs. control cells was not only seen after treatment with cytarabine, but also with daunorubicine and etoposide.

Oncogenic RAS induces myeloid differentiation after cytarabine treatment
We next asked whether the selective loss of clonogenic potential of Ras cells in response to cytarabine correlates with an increase in differentiation [14][15][16][17][18]33]. Control and Ras cells were incubated with cytarabine, and differentiation was determined by analyzing cellular morphology and the expression of markers of granulocytic (Gr1) and monocytic differentiation (Mac1). There was a significant enhancement of differentiation in cells expressing oncogenic RAS after treatment with cytarabine as observed in May-Grunwald-Giemsa-stained slides ( Figure 5A and Figure S6A) as well as in RQ-PCR and FACS analyses of ly6g (encoding the Gr1 antigen) and itgam (encoding Mac 1) ( Figure 5B and Figure S6B). Consistent with these observations, the mRNA levels of the stem cell marker kit were lower in Ras cells and expression became virtually undetectable upon treatment with cytarabine ( Figure 5C).
To test whether the enhanced differentiation observed in Ras cells depends on the elevated checkpoint activity in response to cytarabine, we incubated control and Ras cells for 24 hours with cytarabine in the presence or absence of caffeine, an inhibitor of the Atm and Atr kinases [34]. Initial experiments showed that incubation with 0.5 mM caffeine was sufficient to inhibit phosphorylation and thus activation of Chk1, as demonstrated by the reduced level of p53 ( Figure S7A+B). Notably, treatment with caffeine together with cytarabine markedly decreased the myeloid differentiation of Ras cells ( Figure 5D). In contrast, addition of Control and Ras cells were treated with 350 nM cytarabine for the indicated times and cell lysates were probed with the indicated antibodies. Cdk2 is shown as loading control. (B) Ras cells express elevated levels of p53. Control and Ras cells were either treated with 100 mM cytarabine for five hours or exposed to UVB for 3.5 min and subsequently incubated for five additional hours. Immunoblots of cell lysates were probed with antibodies against p53 and Cdk2 as shown. (C) Phosphorylation of Chk1 is increased in Ras cells after treatment with cytarabine. Control and Ras cells were treated with 100 mM cytarabine for the indicated times. The panels show immunoblots of cell lysates subjected to antibodies against Chk1 and phospho-Chk1 (pSer345). (D/E) Ras cells show elevated checkpoint activation upon cytarabine treatment. Control and Ras cells were treated as before. Panel D shows immunoblots of cell lysates that were probed with antibodies directed against phospho-H2a.x (pSer139) and Cdk2 (loading control). Panel E shows immunofluorescence pictures of control and Ras cells treated with 100 mM cytarabine for one hour and stained with antibodies directed against phosphorylated Atm (pSer1981) (green). DNA was counterstained with Hoechst (blue). doi:10.1371/journal.pone.0007768.g003 caffeine did not restore the clonogenic potential of Ras cells treated with cytarabine, potentially because caffeine showed significant toxicity in long-term experiments (not shown). Taken together, the data show that the cytarabine-induced loss of clonogenicity correlates with a checkpoint-dependent induction of cellular differentiation of Ras cells.
Importantly, inducing cellular differentiation by withdrawal of 4-hydroxy-tamoxifen and subsequent inactivation of the MLL-ENL-ER chimeric protein did not elevate checkpoint activation in the absence of oncogenic RAS ( Figure S8), suggesting that the effects of oncogenic Ras on checkpoint activity are upstream and independent of its effects on cellular differentiation.
Figures 5B and C show a higher spontaneous differentiation of Ras cells, even without in vitro treatment with cytarabine. In order to investigate if oncogenic Ras induced myeloid differentiation, which is potentiated by activating DNA-damaging agents such as cytarabine, may also be observed in primary AML, we analyzed primary AML samples using cDNA expression analysis. We took advantage of primary AML cases diagnosed within the AML-SHG Germany multicenter study group and selected 31 AML cases with inversion (16) to have a comparable genetic background. AML with oncogenic N-RAS mutations (N = 12) revealed a different expression signature as compared to AML lacking such mutations (N = 19). A Gene Set Enrichment Analysis demonstrated that genes characteristic for hematopoietic progenitors were expressed at higher levels in samples harbouring wild-type N-RAS while genes associated with mature blood cells rather than progenitor and stem cell compartments were strongly enriched (up-regulated) in the N-RAS mutant samples (not shown) [35,36]. We confirmed the more differentiated phenotype of N-RAS mutant AML using RQ-PCR for the SIAT10 gene, which is known to be progressively up-regulated during myeloid differentiation ( Figure 5E) [35]. The higher spontaneous differentiation observed for Ras cells in vitro was therefore also detected in primary AML samples.  Expression of p16 Ink4a -resistant Cdk4R24C does not abrogate myeloid differentiation Relative to control cells, Ras cells express elevated levels of p16 Ink4a (Figure 3A), which inhibits Cdk4 kinase activity, leading to activation of the retinoblastoma tumour suppressor protein (Rb). In addition, they express elevated levels of p19 Arf and show enhanced checkpoint activity, both leading to activation of p53. To identify which if any of these two proteins is a critical downstream effector of Ras that mediates the cytarabine-induced loss of clonogenicity, we generated Ras cells that express either Cdk4R24C, a melanomaderived mutant of Cdk4 that is resistant to inhibition by p16 Ink4a [37] or a dominant-negative allele of p53 [38].
Immunoblots revealed that Cdk4R24C was expressed in Ras cells and had no effect on the activation of Chk1 in response to treatment with cytarabine ( Figure S9A); furthermore, expression of Cdk4R24C led to elevated levels of phosphorylated pRb, consistent with its ability to negate the effect of p16 Ink4a ( Figure  S9B). Importantly, expression of Cdk4R24C had no effect of the cytarabine-induced differentiation of Ras cells (Figures S9C,D,E) and further suppressed their clonogenic potential in the presence of cytarabine ( Figure S9F), strongly suggesting that p16 Ink4a is not a critical mediator of Ras action in this setting.

Dominant negative p53 inhibits oncogenic RAS-and cytarabine-induced differentiation
To test whether the observed differentiation depends on p53, we expressed a dominant negative allele of p53 (p53DD) in Ras cells (generating Ras/p53DD cells) [38]. Immunoblots confirmed the expression of p53DD in these cells, and showed that cytarabineinduced phosphorylation of Chk1 is independent of p53 ( Figure 6A). In contrast, expression of p53DD abrogated cytarabine-induced p21 Cip1 expression, a downstream target of p53; furthermore levels of endogenous p53 were elevated in cells expressing p53DD, indicative of downregulation of the MDM2 gene ( Figure 6B and Figure S10). These data demonstrate that p53DD blocked p53 function in Ras cells.
Importantly, May-Grunwald-Giemsa staining showed that untreated Ras/p53DD cells had an immature phenotype in contrast to Ras cells, which showed a monocytic/macrophage like morphology ( Figure 6C). RQ-PCR analysis showed that expression of p53DD decreased the basal expression of ly6g and itgam mRNAs and abrogated the cytarabine-induced increase in expression of these markers of myeloid differentiation ( Figure 6D); very similar results were obtained in a FACS analysis of Gr1 and Mac1 protein expression ( Figure 6E; Figure S11). In order to analyse whether the Ras-induced inhibition of colony growth was also mediated by p53, Ras/p53DD cells were analysed in the colony test after pretreatment with cytarabine as before. Importantly, expression of p53DD, but not the inactive L344P mutant of p53DD, restored the clonogenic potential of Ras cells upon transient exposure to cytarabine ( Figure 6F and Figure S12). Similar results were obtained by inhibition of p53 with the small molecule pifithrin-a [39] (Figure S12).
Taken together these data strongly suggest that the loss of clonogenicity and increased differentiation observed in Ras cells in response to cytarabine is mediated by activation of p53.

Nutlin-3 enhances the biological effects induced by oncogenic RAS and cytarabine
As shown above, p53 is needed for the cytarabine and oncogenic RAS driven myeloid differentiation program in leukemia cells. Nutlin-3 activates endogenous p53, as it inhibits the Mdm2-induced degradation of p53 [40]. Nutlin-3 can induce p53-dependent apoptosis in acute leukemia cells [41][42][43]. Interestingly, a nutlin-dependent maturation program has recently been described in AML cells [44]. We therefore wanted to know whether incubation of Ras cells with cytarabine together with nutlin-3 lead to increased myeloid differentiation in these cells.
To address whether the biological effects of endogenous p53 could be enhanced using nutlin-3, we measured the expression of p21 Cip1 as a downstream effector of p53. Expectedly, expression of p21 Cip1 increased after incubation of Ras cells with cytarabine ( Figure 7A). Nutlin-3 incubation also induced, to a lesser degree, p21 Cip1 expression. Importantly, co-incubation of nutlin-3 together with cytarabine potentiated cytarabine-induced p21 Cip1 expression in Ras cells ( Figure 7A); in contrast, neither cytarabine nor nutlin-3 had any effect on p21 Cip1 expression in Ras/p53DD cells.
We next asked whether nutlin-3 was also able to enhance cellular differentiation in Ras cells. RQ-PCR analysis showed that expression of differentiation marker genes ly6g and itgam was significantly increased when Ras cells were co-incubated with nutlin-3 and cytarabine ( Figures 7B,C). Moreover, nutlin-3 was unable to stimulate differentiation of Ras/p53DD cells, demonstrating that it does not act via a p53-independent mechanism. Together, the data show that Mdm2-mediated degradation of p53 limits the differentiation observed in response to cytarabine and that combination of nutlin-3 enhances the differentiation-inducing effect of cytarabine.
Finally, we noted that withdrawal of 4-hydroxy-tamoxifen to switch off the MLL-ENL-ER chimera induced expression of ly6g and itgam in control and in Ras cells, but was unable to do so in Ras/p53DD cells ( Figure 7D), arguing that p53 is essential for granulocytic and monocytic differentiation, at least in the context of oncogenic Ras.

Discussion
Oncogenic RAS mutations are among the most frequent mutations observed in human cancers (reviewed in [6]). Until recently, the prognostic role of oncogenic RAS in AML was not well understood. Some studies reported a correlation with poor outcome [45][46][47], whereas others observed a better prognosis in AML with oncogenic RAS mutations [48][49][50]. We have recently extended these analyses and have demonstrated an interaction between oncogenic RAS and the dose of cytarabine used during postinduction-treatment with respect to the cumulative incidence of relapse and overall survival [26]. Patients, whose AML blasts revealed oncogenic RAS mutations and who had been treated with low-dose cytarabine, had the worst prognosis (highest incidence of relapse; worst survival) [26]. In contrast, those with oncogenic RAS randomly treated with high-dose cytarabine had the best prognosis of all groups (lowest incidence of relapse, best survival). Patients with wild type RAS had only little benefit from cytarabine doseescalation, and their prognosis was in between the patients with RAS mutations. The molecular basis for this observation remained unclear.
In order to molecularly understand the interaction of oncogenic RAS with cytarabine in AML, we took advantage of mouse bone marrow cells that had been immortalized using a conditional MLL-ENL-ER oncogene [27,28] and that were co-infected with either an empty vector (control cells) or a vector expressing oncogenic RAS (Ras cells). We chose leukemic cells expressing MLL-ENL for several reasons: first, MLL-ENL is a potent oncogene and is able to transform and immortalize progenitor cells at various levels of myeloid differentiation, although high-doses of growth factors are still needed for in vitro culture, most likely to substitute for a lacking class I mutation in these cells [27,51]. Second, MLL-ENL expressing progenitor cells define a leukemia-initiating cell population resembling acute myeloid leukemia in humans [52]; third, once MLL-ENL-ER is switched off, the cells differentiate and undergo apoptosis, demonstrating the importance of this fusion gene for maintaining self renewal and growth.
Although in vivo therapy with cytarabine underlies a complex pharmacodynamic and pharmacokinetic regulation which may not be reproduced in the cell culture system used in this study [53], the MLL-ENL cells showed phenomena in vitro which resemble the observations made in the clinical study [26]. Notably, there was no significant difference between the cell number, cell survival or apoptosis of Ras and control cells treated with cytarabine. This corresponds with the clinical situation where a significant difference between AML-patients with and without RAS mutations with regard to complete remission was not found. Instead, the clinical observation that patients with oncogenic RAS relapse less frequently after complete remission suggested that the number of clonogenic stem cells able to cause AML relapse may be significantly decreased in AML with oncogenic RAS mutations upon treatment [26]. This may correlate with the lower in vitro clonogenicity of Ras cells that have been transiently exposed to cytarabine (and, as shown in Figure 4, to other cytotoxic drugs such as daunorubicine and etoposide).
Our analysis provides several molecular correlates for these observations and suggests that the difference in response to cytarabine is due to known biological properties of oncogenic RAS. Notably, Ras is known to activate expression of several proteins associated with cellular senescence, such as p16 Ink4a , p19 Arf and p21 Cip1 [20] and we confirmed these observations for the leukemic cells studied here ( Figure 3A). Oncogenic RAS also activates a DNA damage response in primary fibroblasts [24,25,54]. While oncogenic RAS was unable to induce a DNA-damage response in MLL-ENL cells by itself, it strongly enhanced the DNA-damage response observed after incubation with DNA-damaging agents such as cytarabine. As a result, high levels of p53 are present in Ras cells treated with cytarabine ( Figure 3B) and mediate their loss of clonogenicity ( Figure 6F).
Oncogenic RAS synergizes with cytarabine to enhance differentiation of AML cells Activated DNA damage checkpoint frequently induce cell cycle arrest or apoptosis, yet we have not observed any difference in regard to apoptosis or proliferation after cytarabine treatment between Ras and control cells. It has been reported that cytarabine can induce myeloid differentiation (e.g. [33,55]). Numerous studies have also shown that oncogenic RAS also induces differentiation  [14][15][16][17][18]. We observed a very moderate induction of differentiation in control cells, when treated with cytarabine. This was significantly increased in Ras cells incubated with cytarabine ( Figure 5). Therefore, either molecular change alone was ineffective in inducing full differentiation, whereas the combination of both did. Notably, the increased differentiation was abolished by incubation with caffeine, an inhibitor of the Atm and Atr kinases, demonstrating that it depends on checkpoint activation ( Figure 5D). Taken together, the data suggest that oncogenic RAS in combination with higher doses of cytarabine induces differentiation in vitro and strongly decreases the clonogenic potential. Supporting this notion, Ras cells treated with cytarabine were much less likely to express the stem cell marker kit as compared to control cells ( Figure 5C). Also, primary AML cells with oncogenic N-RAS mutations revealed a higher expression of differentiation markers as compared to patients lacking such mutations ( Figure 5E).
Oncogenic RAS-and cytarabine-induced myeloid differentiation depends on p53 TP53 is frequently mutated in human cancer, but rarely in myeloid leukemias; therefore, the MLL-ENL transformed cells here mimic the p53 status found in AML. Importantly, our data identify p53 as a critical mediator of the enhanced differentiation and loss of clonogenicity observed in Ras cells upon exposure to cytarabine. This was supported by the observation that Ras cells expressing p53DD did not undergo differentiation upon treatment with cytarabine ( Figures 6D,E). Also, treatment of Ras cells with both cytarabine and nutlin-3, an inhibitor of Mdm-2 induced degradation of p53, caused a further increase in p21 Cip1 expression, which was paralleled by increased expression of differentiation markers ( Figures 7A,B,C) and is consistent with a recent report that nutlin-3 can cause maturation in AML cells [44].
Notably, Ras cells expressing p53DD did not differentiate even when the MLL-ENL-ER oncogene was switched off by withdrawing 4-hydroxy-tamoxifen ( Figure 7D), demonstrating that p53 is required for induction of differentiation of these cells. MLL-ENL has been shown to attenuate p53 function, since it antagonizes the interaction of p53 with the p300/CBP co-activator [56]; it is conceivable that this activity of MLL-ENL counteracts the induction of differentiation by low levels of active p53, such as might be present in cells expressing oncogenic RAS in the absence of cytarabine.
One caveat in our study is the use of H-Ras, while AML patients more frequently harbor mutations in the N-RAS or K-RAS genes [48]. However, the signaling pathways downstream of Ras proteins are conserved and H-Ras can substitute for K-Ras during embryonic development [57]. Furthermore, higher expression of differentiation markers in primary AML cells with oncogenic N-RAS mutations ( Figure 5E) strongly suggests that mutant H-Ras and N-Ras function similarly to promote differentiation in hematopoietic cells.
Our findings may have implications for understanding how chemotherapeutic drugs exert their effects. Conventionally, cytostatic drugs such as cytarabine are thought to act as inhibitors of proliferation and inducers of apoptosis. This study suggests that the critical function of a cytotoxic drug such as cytarabine may be to promote the differentiation of tumor-initiating cells. That induction of differentiation is an attractive goal for anticancer therapy and is associated with higher cure rates has been demonstrated elegantly in acute promyelocytic leukemia, where high-doses of retinoic acid, given concomitantly with chemotherapy, overcomes the repressive effect of PML-RARa in differentiation. We propose that induction of differentiation should therefore be seen as a broader goal in AML therapy, e.g. for in vitro screening of new compounds, and in the development of new treatment protocols. The success of such a procedure will depend on the genetic background of the respective cancer cells.

Retroviral transduction of mouse primary hematopoietic cells
High-titer retrovirus supernatants were produced by transient transfection of the packaging cell line Phoenix-E using a standard Ca 2+ -phosphate precipitation method. Viral titers usually reached approximately 5610 6 CFU/ml. Retroviral transduction of primary hematopoietic cells was performed as described previously [58]. All constructs (Ha-RasV12, Cdk4R24C-Flag, p53DD (aa 302-390), p53LP (p53DD with L344P mutation) were cloned into pMSCV retroviral vectors (Clontech, USA) and cells were cultured in the presence of either puromycin, hygromycin or blasticidin, respectively. In total, we performed three independent infections of MLL-ENL cells with either control viruses or viruses encoding oncogenic RAS, using two different clones of MLL-ENL cells. The results we obtained were consistent between these independent experiments.
To measure checkpoint responses in short-term assays, cells were treated either with up to 100 mM cytarabine (Merck, Darmstadt, Germany) for the indicated times or exposed to UV-B for 3.5 min and subsequently cultured at 37uC. Where indicated, cells were cotreated with 500 mM caffeine (Sigma Aldrich, Munich, Germany). For long-term assays, cells were exposed to 100 nM cytarabine unless indicated otherwise. For stabilization of p53 protein, Nutlin-3 (Sigma Aldrich, Munich, Germany) was added at a concentration of 5 mM. For inhibition of p53 protein cells were treated with 20 mM Pifithrin-a (Calbiochem, Gibbstown, NJ, USA).
For immunoblot analysis of phopho-Chk1, cells were exposed for 1 h to 1 mM, 10 mM and 50 mM daunorubicine (Pfizer, Berlin, Germany) and etoposide (Teva-Gry, Kirchzarten, Germany), respectively. Colony formation assays were performed with concentrations of 10 nM and 25 nM for daunorubicine and 50 nM and 250 nM for etoposide.

Colony formation
The assays were performed in methylcellulose medium and colonies were stained with INT (Iodonitrotetrazolium chloride, Fluka, Buchs, Switzerland) at a final concentration of 1 mg/ml. For cell counting, methylcellulose was diluted in PBS, cells harvested and live cells were counted in triplicates of each sample using exclusion of trypan blue as criterion. Pictures of the colonies were captured using a Leica MZ125 binocular together with Leica DC300 camera and Leica IM1000 software (Leica Microsystems, Switzerland).
Acidic b-galactosidase assay Cells were assayed for the senescence-associated b-galactosidase activity by x-gal staining as described in [59] and subsequently transferred to slides by cytocentrifugation. Pictures were taken by Leica DMLB microscope with a Leica DFC420 camera and Leica DFC Twain software (Leica Microsystems, Switzerland).

Morphological analysis
A cytocentrifuge was used to spin cells onto slides. Staining was performed using May-Grunwald and Giemsa (Sigma Aldrich, Munich, Germany) according to the manufacturers protocol.

Immunofluorescence
Samples were fixed in 3,7% paraformaldehyde for 10 min, followed by permeabilization in PBS-T (PBS containing 0.2% Triton X-100). Nonspecific protein binding was blocked by 5% FCS (Gibco, Karlsruhe, Germany). Cells were stained using a primary antibody against phospho-Atm (pSer1981, Chemicon, USA) and a FITC-conjugated a-rabbit in parallel to Hoechst. Pictures were taken by BD Pathway 855 High-Content Bioimager with BD AttoVision 1.6 software (BD Biosciences, Heidelberg, Germany).

Ras pull-down assay
The activity of Ras was analyzed with Ras Activation Assay Kit from Millipore (CA, USA) according to the manufacturers recommendations. Lysates of 6610 6 cells were used for the Ras pull-down assay.

Flow cytometric analysis
Antibodies for flow cytometric (FACS) analysis (isotype control, Gr-1/Ly6G/C, Mac-1/CD11b) were purchased from BD Biosciences (Heidelberg, Germany) and used according to the recommendations of the manufacturer. The histograms display the geometric mean (GeoMean) of the analysed sample. The BrdU-FACS was carried out with the FITC BrdU Flow Kit (BD Pharmingen, BD Biosciences, Heidelberg, Germany) according to the manufacturers instructions.
The separation of cells positive and negative for Mac1 and c-kit, respectively was performed via the MACS-system (Miltenyi, Bergisch-Gladbach, Germany). The separation was carried out using anti-PE microbeads and MS columns in combination with Mac-1/CD11b-PE and c-kit/CD117-PE antibodies from BD Biosciences (Heidelberg, Germany) according to the instructions of the manufacturer.

RQ-PCR and RT-PCR
RNA was isolated with PeqGold TriFast according to the manufacturers instructions (Peqlab, Erlangen, Germany). RNA (2 mg) was reverse transcribed with 200 U M-MLV-RT (Invitrogen, Karlsruhe, Germany). cDNA was amplified either by PCR using a Q-PCR kit (Immomix from Bioline, Luckenwalde, Germany) and the product detected on agarose gel, or by quantitative real-time PCR and the product detected with SYBR green using a Mx3000 (Stratagene, USA) detection system. Expression of rps16 mRNA was used as reference. Real-time PCR was performed in triplicates and error bars indicate standard deviation. 31 primary AML cases with inversion (16), diagnosed and treated within the Germany AML-SHG 96 trial, were provided by T.I. and analyzed by RQ-PCR for SIAT10 expression. Expression values were normalized to b-actin mRNA and HL60 cells as a reference.

Mutagenesis PCR
The mutagenesis of mutated RasV12 to wild type Ras was done with Quick Change Multi Site-Directed Mutagenesis Kit from Stratagene (CA, USA) according to the protocol of the manufacturer. The sequences of primers used for the PCR reaction were as follows: forward primer, GTTGTTGTTGG-CGCCGGTGGTGTGGGCAAGAGTG, reverse primer, CAC-TCTTGCCCACACCACCGGCGCCAACAACAAC.

Statistical analysis
Statistical analysis was performed with two-tailed Student's t test with Welch's correction. The panels show immunoblots of cell lysates that were probed with antibodies against phospho-Rb (pSer807 and pSer811), Rb, p16Ink4a and Flag (Cdk4R24C). Cdk2 served as loading control. The asterisk denotes a non-specific band. (C) Ras cells differentiate despite expression of Cdk4R24C. Ras and Ras/Cdk4R24C cells were treated with 350 nM cytarabine for 24 hours, washed and cultured for additional two days in the absence of cytarabine. Cells were subjected to FACS analysis using a-Gr1-PE antibodies. (D) Quantification of (C). (E) Cytarabine induces the expression of ly6g mRNA in Ras/Cdk4R24C. Ras and Ras/Cdk4R24C cells were treated with 350 nM cytarabine for 24 hours and expression of ly6g mRNA was analyzed by RQ-PCR. (F) Cdk4R24C does not enhance the clonogenic potential of Ras cells. The assay was performed as described in Figure 2.