E4F1 deficiency results in oxidative stress–mediated cell death of leukemic cells

To study E4F1 functions during tumorigenesis, we used a recently developed genetically engineered mouse model containing E4F1-null and E4F1 conditional KO alleles (E4F1^−/flox) based on the Cre/Lox-P technology (Fig. 1 A and Fig. S1 A; Le Cam et al., 2004; Lacroix et al., 2010). E4F1^−/flox mice were crossed with RERT mice, a knockin (KI) strain expressing the 4-hydroxy-tamoxifen (4OHT)–inducible Cre-ER^T2 fusion protein under the control of the ubiquitously active promoter of the RNA polymerase II large subunit gene (Guerra et al., 2003). In vivo recombination efficiency of the E4F1^flox allele was assessed in several organs of E4F1^−/flox; RERT^KI/KI mice by quantitative PCR (qPCR) on genomic DNA, RT-qPCR on purified RNA, or Western blot analyses of total protein extracts (Fig. 1 B and Fig. S1 A). Upon 4OHT administration, high recombination efficiency was detected in several organs, including liver, spleen, and lungs, as well as in peripheral white blood cells and total or purified lineage-negative (Lin⁻) BM cells (Fig. 1 B). These analyses validated our model and confirmed that E4F1 is efficiently deleted upon Cre activation in the hematopoietic compartment. However, analyses of E4F1^−/flox; RERT^KI/KI mice also revealed critical roles of E4F1 in other organs, limiting long-term investigations of phenotypes resulting from E4F1 loss in the hematopoietic compartment in that mouse model.

To address E4F1 functions during tumorigenesis, we next crossed E4F1^flox; RERT^KI/KI animals with Ink4a/Arf KO mice. The latter strain of mice harbors a deletion of exons 2 and 3 of the Ink4a/Arf locus and therefore expresses neither p16^INK4a nor p19^ARF. Ink4a/Arf-null animals are tumor prone and were previously described to develop with high penetrance essentially B and T cell lymphomas and soft tissue sarcomas with an expected mean age of tumor appearance around 30 wk (Serrano et al., 1996). To avoid defects resulting from E4F1 inactivation in the whole organism and investigate E4F1 functions during tumorigenesis, we transplanted fetal liver hematopoietic stem cells (HSCs) isolated from E4F1^−/flox; RERT^KI/KI; Ink4a/Arf^−/− embryos or their control E4F1^+/flox; RERT^KI/KI; Ink4a/Arf^−/− littermates into lethally irradiated WT recipient mice (herein referred to as E4F1 KO or CT; Ink4a/Arf KO mice, according to the genotype of the transplanted cells). We used this strategy to generate several cohorts of tumor-prone mice in which E4F1 could be acutely inactivated specifically in the hematopoietic system and in tumors derived from those cells upon administration of 4OHT (Fig. S1 B).

We then carefully monitored our transplanted mice for perturbations of the hematopoietic system and for tumor susceptibility. Strikingly, we found that our cohorts of transplanted animals, regardless of E4F1 genotype and in the absence of 4OHT administration, developed with full penetrance HSs, a tumor type originating from the monocytic/macrophagic lineage, but no B nor T cell lymphomas. These tumors arose as soon as 10 wk after transplantation, and postmortem histological analyses indicated that all transplanted animals showed diffuse and/or nodular neoplastic infiltrations in the spleen, liver, or lungs. Pathological analyses of tumors recovered from reconstituted animals confirmed that they exhibited all phenotypic and immunological features reminiscent of human HSs, according to the international World Health Organization classification of human tumors. HS tumor cells appeared as large cells with abundant eosinophilic cytoplasm and pleiomorphic nuclei with large nucleoli (Fig. 1 C) and showed positive immunoreactivity with the histiocytic cell surface markers F4/80 and/or Mac2 and heterogenous staining for the proliferation marker Ki67 (Fig. 1 D and Fig. 2 B). All tumors stained negative for B and T cell markers (Fig. S1 D), indicating that none of the transplanted animals developed the previously described B or T cell lymphoma spectrum found in Ink4a/Arf KO animals (Serrano et al., 1996). Increased number of F4/80-positive HS cells was also detected by flow cytometry in the peripheral blood between 10 and 20 wk after transplantation and indicated the progression of the disease (Fig. S1 F). Finally, most of these transplanted animals exhibited profound anemia and body weight loss at late stages of tumor progression (unpublished data). Death of those animals occurred between 3 and 11 mo after transplantation, according to E4F1 status (Fig. 2 A).

Altogether, these data indicated that we established a novel murine HS model based on transplantation of Ink4a/Arf-null fetal liver HSCs. Tumors that developed in that animal model also harbored the E4F1 conditional KO allele, allowing us to address the roles of E4F1 during tumor development.

Next, we addressed the consequences of E4F1 inactivation in transplanted animals. For evaluation of tumor susceptibility, recipient mice were transplanted with fetal HSCs harvested from genetically matched E4F1^+/flox or E4F1^−/flox; RERT^KI/KI; Ink4a/Arf^+/+ or Ink4a/Arf^−/− mice (herein referred to as E4F1 CT or KO; Ink4a/Arf WT or KO, according to the genotype of the transplanted cells). In the first experimental group, 4OHT was administered in reconstituted animals 10 wk after transplantation to allow full reconstitution of the hematopoietic system before Cre-mediated recombination. Administration of 4OHT was then repeated on a monthly basis until the death of the animals (Fig. S1 B).

As depicted in Fig. 2 A, E4F1 inactivation significantly increased the lifespan of animals transplanted with Ink4a/Arf KO cells, with a median survival time (50% survival) of 28 wk and 32 wk in E4F1 CT and KO animals, respectively. Of note, ∼20% of E4F1 KO; Ink4a/Arf KO animals exhibited a much longer lifespan, with death occurring up to 12 wk later than the last E4F1 CT; Ink4a/Arf KO animal. Necropsy and histological experiments of all animals, performed at the time of their natural death, indicated that loss of E4F1 reduced tumor infiltration in the main sites of HS dissemination, including lungs, liver, and spleen (Fig. 2 B, Fig. S2 B, and not depicted). This reduction of tumor infiltration was also illustrated by a 2.5-fold decrease of the median spleen weight (Fig. 2 C) in E4F1 KO; Ink4a/Arf KO when compared with control animals.

These pathological analyses suggested that E4F1 inactivation delayed tumor development in animals transplanted with Ink4a/Arf KO cells. To further confirm E4F1 impact on tumorigenesis, we set up an independent experimental group for longitudinal experiments in which tumor progression was monitored on individual animals by in vivo positron emission tomography (PET) imaging. We performed quantitative analyses on four independent animals of each genotype (E4F1 CT or KO; Ink4a/Arf KO). Besides classical PET background observed in heart (Fig. S2 A) that precluded analysis of tumor development in lungs, we efficiently followed tumor progression in the liver and spleen. As expected, we observed increased PET signal in both the liver and spleen of E4F1 CT; Ink4a/Arf KO animals during the time course of the experiment, illustrating the normal progression of the disease. In striking contrast, we found that E4F1 inactivation, upon 4OHT administration to E4F1 KO; Ink4a/Arf KO animals, resulted in a strong decrease of PET signal in both the liver and spleen (Fig. 3, A and B; and Fig. S2 C). Consistent with this tumor regression, we observed on tissue sections prepared from E4F1 KO; Ink4a/Arf KO animals that Mac2-positive HS tumor cells exhibited increased TUNEL (terminal deoxynucleotidyl transferase biotin-dUTP nick end labeling) staining, which is a hallmark of cell death (Fig. 3 C). Altogether, these analyses revealed that E4F1 inactivation delayed HS development, increased lifespan in this tumor-prone animal model, and induced tumor regression in established HSs.

Our results prompted us to analyze the cellular consequences of E4F1 inactivation in several HS cell lines that we established from our HS murine model. Flow cytometry and immunofluorescence analyses indicated that the established cell lines expressed predominantly the Mac2 or F4/80 surface markers, with some cell lines expressing both markers, confirming their histiocytic origin (Fig. S4, A and B). Conditional E4F1 inactivation was induced by the addition of 4OHT in the culture medium of E4F1^−/flox HS cell lines (hereafter referred to as E4F1 CT or KO, according to the absence or presence of 4OHT in the culture medium, respectively), resulting in efficient depletion of the E4F1 protein (Fig. 4 A). E4F1 inactivation in vitro resulted in a dramatic decrease of viable cells 3 d after 4OHT addition. Importantly, E4F1^+/flox HS cell lines remained grossly unaffected by 4OHT treatment, confirming that the decreased number of viable cells observed in E4F1^−/flox cells was not the consequence of 4OHT toxicity but resulted from E4F1 depletion (Fig. 4 A and Fig. S4 C). E4F1 KO HS cells exhibited no alteration of their proliferation rate or of their mitotic index 4 d after 4OHT addition (Fig. 4 B). However, we observed massive cell death upon E4F1 inactivation, as shown by increased annexin V staining (Fig. 4 C). Consistent with this result, E4F1 inactivation abrogated the formation of colonies in anchorage-independent assays (Fig. 4 D). Altogether, these data indicated that E4F1 depletion resulted in massive cell death in murine HS cell lines established from primary tumors.

Time-lapse video microscopy analyses of E4F1 KO HS cells showed the apparition of large vacuoles reminiscent of autophagy (unpublished data) 3–4 d after 4OHT treatment, just before cell death occurred. Autophagy is a self-degradation process implicating the lysosomal pathway in which intracellular membrane structures engulf altered organelles or aggregated proteins. We sought additional experimental evidences of autophagy in E4F1 KO HS cells by several complementary approaches. Transmission electron microscopy indicated the presence of ultrastructural characteristics of autophagic cells, such as double-membraned autophagosomes and autolysosomes, in E4F1 KO HS cells (Fig. 5 A). Moreover, we observed by immunoblot that E4F1 inactivation resulted in the conversion of cytoplasmic LC3 (LC3-I) protein to the autophagosome membrane-bound form of LC3 (LC3-II), which is a hallmark of autophagy (Fig. 5 B). Finally, increased formation of autophagosomes and autolysosomes in E4F1 KO HSs was visualized by fluorescence microscopy after anti-LC3 or acridine orange (AO) stainings, respectively (Fig. 5, C and D).

To determine whether induction of autophagy was the main cause of E4F1 KO HS cell death, cells were treated with the autophagy inhibitor 3-methyladenine (3MA), which blocks an early step controlling autophagosome formation (Levine and Kroemer, 2008; Mizushima, 2009). 3MA efficiently blocked autophagy and cell death in E4F1 KO cells, as shown by decreased LC3-II expression and reduced number of annexin V–positive cells (Fig. 5, B and E). We also evaluated whether apoptosis contributed to cell death occurring upon E4F1 inactivation. The z-VAD caspase inhibitor had a minor but significant effect on cell death of E4F1 KO HS cells (Fig. S5 A). Altogether, these results showed that E4F1 KO HS cells died mainly through autophagic cell death with a minor implication of apoptosis.

Oxidative stress is a well-known inducer of autophagy in a variety of cell types including macrophages (Xu et al., 2006). Therefore, we wondered whether autophagy observed in E4F1 KO HS cells resulted from increased ROS levels. E4F1 KO HS cells exhibited increased staining with the 2′,7′-dichlorofluorescein diacetate (DCFDA) probe, a cell-permeable fluorescent dye that reacts with a broad spectrum of ROS (Fig. 6 A). Importantly, DCFDA staining in E4F1 KO HSs was not modified by the autophagy inhibitor 3MA, indicating that increased ROS levels resulting from E4F1 inactivation were not a consequence of autophagic cell death (Fig. S6 A). Consistent with the development of an early and massive oxidative stress in these cells, a robust antioxidant response was detectable early after E4F1 inactivation, as illustrated by an increased expression of several antioxidants factors, including the transcription factor NRF2, catalase, HO1 (heme oxygenase 1), and NQO1 (NAD[P]H dehydrogenase quinone 1; Fig. S6, C–F).

Incubation with probes that detect specific ROS subtypes such as MitoSOX and OxyBURST confirmed that the oxidative stress that resulted from E4F1 inactivation included increased levels of superoxide anions of mitochondrial origin (Fig. 6 A). These data suggested that this organelle could be the main source of ROS detected in E4F1 KO HS cells. As a read-out of potential mitochondrial alterations in E4F1 KO HS cells, we next determined their in situ oxygen consumption rates (OCRs) and ATP levels. Compared with control cells, E4F1 KO HS cells exhibited a strong increase in O₂ consumption (at all levels, i.e., basal respiration, oligomycin C–sensitive respiration, and maximum respiratory capacity; Fig. 6 B). Strikingly, this strong increase in mitochondrial O₂ consumption was not associated with efficient ATP production because E4F1 KO HS cells exhibited decreased ATP levels (Fig. 6 C). Because it is well established that various alterations of the electron transport chain functions increase ROS production (Koopman et al., 2010), this strongly suggests that E4F1 KO HS cells display mitochondrial alterations that impinge on the efficient coupling between oxygen consumption and ATP production, ending in increased ROS production.

The direct consequence of a massive increase in ROS levels is the oxidation of the cellular components, including DNA. Accordingly, we detected increased levels of 8-hydroxy-2′-deoxyguanosine (8-OHdG) in E4F1 KO HS cells, which is a stable marker of oxidatively damaged DNA, as shown by nuclear relocalization and increased intensity of the 8-OHdG staining (Fig. 6 D; Struthers et al., 1998).

These data led us to further investigate whether these ROS were the main cause of death that resulted from E4F1 inactivation. Treatment of E4F1 KO HS cells with the superoxide anion scavenger Tiron (4,5-dihydroxy-1,3-benzene disulfonic acid-disodium salt) (a) resulted in reduction of the conversion of LC3-I to LC3-II (Fig. 6 E), (b) diminished nuclear labeling of 8-OHdG (Fig. 6 D), (c) and led to a marked decrease in the number of annexin V–positive cells (Fig. S5 B). Similar results were obtained with 3H-1,2 dithiole-3-thione (D3T), a chemical compound which increases cellular antioxidant defenses by increasing NRF2 activities (Fig. 6 E and Figs. S5 B and S6 H; Zhu et al., 2006). In addition, treatment of E4F1 KO HS cells in culture with the ROS scavenger N-acetyl-l-cysteine (NAC) rescued cellular viability 3 d after 4OHT addition (Fig. 7 A). Finally, NAC treatment of E4F1 KO; Ink4a/Arf KO animals strongly altered the massive cell death occurring in established HSs upon E4F1 inactivation in vivo, as shown by decreased TUNEL staining performed on tissue sections prepared from these animals (Fig. 7, B and C). Collectively, our data support the notion that E4F1 inactivation in HS cells results in mitochondrial defects and oxidative stress that lead to autophagic cell death, both in vitro and in vivo.

Careful monitoring of E4F1 KO or CT; Ink4a/Arf WT animals indicated that E4F1 depletion did not dramatically affect the different lymphoid and myeloid populations in the peripheral blood (not depicted) and did not alter the lifespan of those animals over a 12-mo observation period in that experimental setting (Fig. 2 A). Therefore, we investigated whether cell death resulting from E4F1 inactivation in HS cells was specific of transformed cells. To address this question, we inactivated E4F1 ex vivo in purified intraperitoneal primary murine macrophages. Although efficient E4F1 depletion was obtained upon 4OHT addition ex vivo, neither increased oxygen consumption, increased ROS levels, nor induced cell death were detected in these cells (Fig. 8). These data showed that E4F1 inactivation resulted in cell death selectively in transformed cells originating from the monocytic/macrophagic lineage but not in normal primary macrophages.

The consequences of E4F1 depletion in HS cell survival prompted us to evaluate whether the observed defects could be extended to human leukemic cell lines of myeloid origin. We first evaluated the consequences of E4F1 depletion in U937 cells, a cell line originally isolated from pleural effusion of a patient with HSs (Sundström and Nilsson, 1976). E4F1 depletion was achieved by transduction of cells with lentiviruses encoding two independent short hairpin RNAs (shRNAs) directed against human E4F1 or a control irrelevant shRNA (Fig. 9 A). E4F1 depletion in U937 cells resulted in increased ROS levels. Comparable results were obtained in other leukemic cell lines, including the acute promyelocytic leukemia HL60, the acute monocytic leukemia THP1, and the erythroleukemic HEL cell lines (Fig. 9 B and Fig. S8 D). Interestingly, similar to murine HS cells, E4F1 depletion in HEL and HL60 cells resulted in increased oxygen consumption (Fig. 9 C). These defects led to massive cell death, as shown by increased annexin V staining (Fig. 9 D).

Altogether, our data suggest that the essential roles of E4F1 in mitochondrial functions and cell survival observed in murine HS cells extend to human leukemic cell lines. Based on this conclusion, we next conducted a pilot experiment aimed at evaluating E4F1 expression levels in samples isolated from leukemic patients.

HS is a rare human tumor precluding the analysis of a significant number of samples. Because we found that E4F1 depletion induced comparable defects in several leukemic cell lines of myeloid origin, we evaluated E4F1 expression levels in AML, a more common myeloid malignancy. E4F1 protein level was assessed by quantitative immunoblotting on protein extracts prepared from 39 BM biopsies harvested from patients diagnosed with AML. Normal BM samples were used as controls. The vast majority (35/39) of AML expressed E4F1 at higher levels than controls, with a subgroup (9/39) that we defined as strong overexpressors (4–13-fold increase above controls; Fig. 9 E and Fig. S8 C). These data were consistent with our observation that murine HS cell lines exhibit a higher amount of E4F1 messenger RNA (mRNA) and protein levels compared with primary macrophages (Fig. S8, A and B). Thus, this pilot experiment suggested that leukemic cells overexpress E4F1 protein, which is consistent with the notion that E4F1 is essential for their survival.

Although the multifunctional E4F1 protein is involved in essential oncogenic pathways, its exact functions during transformation remain poorly understood. In this study, we evaluated E4F1 functions in tumor development in vivo, using an original mouse model based on mice transplanted with genetically engineered HSCs harboring Ink4a/Arf-null and E4F1 conditional KO alleles. This strategy resulted in the generation of a tumor-prone mouse model with unexpected tumor spectrum. Although other murine models based on the Ink4a/Arf KO allele have previously been shown to develop mainly lymphomas and sarcomas (Serrano et al., 1996), we failed to detect any pathological evidences of lymphoma development in our transplanted mice. The absence of lymphoma in our experimental model did not reflect differences in genetic backgrounds because genetically matched donor mice developed lymphomas with the expected frequency (Fig. S1 E). Strikingly, all of our transplanted mice developed HSs, a rare neoplasm of the histiocytic lineage with poor prognosis and aggressive clinical course in humans. Most human HS cases previously described were originally misclassified and are now recognized as diffuse large B cell lymphomas with associated reactive macrophages. However, extensive morphological and immunophenotypical analyses confirmed that the hematopoietic tumors arising in our model are bona fide HSs, according to criteria published in the World Health Organization classification of human tumors. The high penetrance of this tumor type raised the question about the molecular etiology and cell of origin of HSs. Consistent with previous studies, our data support the cooperating role of functional alterations of the Ink4a/Arf locus and the phosphoinositide 3-kinase (PI3K) pathway in HS pathogenesis in murine models (Eischen et al., 2002; Lund et al., 2002; Carrasco et al., 2006). Indeed, although PTEN protein remained expressed in all HS cell lines established from our murine model, we found clear evidence for a deregulation of the PI3K pathway, as illustrated by a marked increase in the activating phosphorylation of AKT on Ser473 (Fig. S3). Because only few relevant HS mouse models exist so far (Martín-Caballero et al., 2001; Eischen et al., 2002; Carrasco et al., 2006; Khoo et al., 2007), the characterization of this new model exhibiting full penetrance and apparent specificity for the development of HS might help to better understand the etiology of this aggressive disease.

Using this tumor-prone mouse model, we found that E4F1 inactivation resulted in the massive cell death of transformed cells, tumor regression, and increased life span of the transplanted animals. Nevertheless, E4F1 KO; Ink4a/Arf KO animals still died earlier than E4F1 KO; Ink4a/Arf WT mice. The cause of death of these animals remains unclear. Our data do not support the idea that E4F1 inactivation induced severe hematopoietic defects compromising the lifespan of those transplanted animals. Detailed necropsy analyses performed with trained pathologists indicated that some but not all E4F1 KO; Ink4a/Arf KO animals exhibited lesions that could reflect massive and rapid cell death of tumor cells in essential organs such as lungs and liver. These defects may have resulted in a severe tissue disorganization and animal death.

To decipher E4F1 functions in HS cell survival, we next performed extensive analyses in established HS cell lines derived from our animal models. Our previous study indicated that E4F1 KO embryos display mitotic progression defects followed by cell death at the peri-implantation stage, as illustrated by increased mitotic index and aberrant mitotic figures (Le Cam et al., 2004). Such defects were not observed in HS cells upon acute inactivation of E4F1, suggesting that E4F1 has different functions in various cell types.

Compelling evidence indicates that E4F1 inactivation in HS cell lines turned on a switch toward autophagic cell death, a process which likely reflects the well documented propensity of cells originating from the macrophage lineage to undergo autophagy in stress conditions (Xu et al., 2006; Huang et al., 2009). Seeking the stress that had initiated this response, we found that acute E4F1 inactivation in HS cells led to mitochondrial defects and increased ROS levels. Importantly, rescue experiments based on ROS scavengers or chemical compounds that increase cellular antioxidant defenses indicated that these ROS preceded the autophagic cell death response in E4F1 KO HS cells. Increased ROS levels observed in these cells likely resulted in deleterious oxidation of many cellular constituents, including proteins, DNA, and lipids. Accordingly, we observed that E4F1 inactivation in HS cells associated with massive genomic DNA oxidation that was largely abrogated upon treatment with ROS scavengers. Furthermore, administration of NAC to E4F1 KO Ink4a/Arf KO animals prevented the massive cell death that was observed in vivo upon E4F1 inactivation in established HSs. Preliminary analyses performed by PET scan on a limited number of NAC-treated animals also suggested that ROS scavengers prevented tumor regression induced by E4F1 inactivation (unpublished data).

Those results raised the question about the source of ROS production in E4F1 KO HS cells. Because E4F1 has been described as an interactor and a regulator of the tumor suppressor p53 (Le Cam et al., 2006), one possibility was that E4F1 depletion had modified some p53 functions leading to increased ROS levels and/or activation of an autophagic response. Indeed, beside its well documented functions on cell cycle and apoptosis, p53 has been recently shown to directly modulate DRAM (damage-regulated autophagy modulator), a component of the autophagic machinery (Crighton et al., 2006). Furthermore, transcriptional independent activities of cytoplasmic p53 have also been proposed to regulate autophagy (Tasdemir et al., 2008). Finally, p53 has been described to mediate pro- as well as antioxidant effects, likely depending on the intensity of the p53 response. Paradoxically, both have been suggested to participate in its tumor suppressor functions (Vousden and Ryan, 2009). We currently do not favor the hypothesis that activation of p53 is involved in increased ROS levels found in E4F1 KO HS cells because those cells already harbor an altered p53 pathway resulting from genetic inactivation of the Ink4a/Arf locus. In addition, shRNA-mediated depletion of endogenous p53 failed to modulate increased ROS levels or cell death in E4F1 KO HS cells (Fig. S7). Our data rather point to a more direct role of E4F1 in the regulation of mitochondrial activities that impinge on ROS levels. Indeed, increased ROS levels were detected by several ROS-sensitive fluorescent probes, including dyes that detect superoxide anions of mitochondrial origin. Our data support the importance of superoxide anions as a byproduct of mitochondrial defects in the induction of cell death observed in E4F1 KO HS cells. Consistent with that notion, we found that E4F1 inactivation resulted in perturbed mitochondrial oxygen consumption and decreased ATP production, and cell death was partly rescued upon treatment with Tiron, which allows detoxification of superoxide anions into less toxic ROS. This is of particular interest in light of our recent data showing that E4F1 inactivation in murine fibroblasts perturbed expression of several mitochondrial components involved in energetic and metabolic processes (unpublished data). These data support the notion that high ROS levels in E4F1 KO cells are the direct consequence of mitochondrial dysfunctions, which in turn induce cell death in HS cells. Furthermore, shRNA-mediated depletion of E4F1 in several human transformed cell lines of myeloid origin resulted in the same sequence of events, suggesting that the role of E4F1 in those processes is conserved in murine and human leukemic cells.

The role of ROS in tumorigenesis remains controversial. On one hand, many studies describe ROS as inducers of tumorigenesis (Klaunig et al., 2010; Ziech et al., 2011). On the other hand, previous data have also shown that some cancer cells, including leukemic cells, are more sensitive to increased ROS levels than their nontransformed counterparts, opening a window of opportunity for pro-oxidant anticancer treatments (Huang et al., 2000; Trachootham et al., 2006, 2008; Valko et al., 2007). Our data are consistent with the latter strategy because we found that E4F1 inactivation induces ROS-mediated cell death in HS transformed cells but not in normal primary macrophages. The molecular mechanisms that are responsible for the higher sensitivity of cancer cells to increased ROS levels still remain unclear. Several explanations have been proposed, implicating the lower antioxidant defenses in some tumor cells or the paradoxical role of AKT in sensitizing cancer cells to ROS-induced cell death (Huang et al., 2000; Nogueira et al., 2008). Based on our data showing that transformed but not normal cells display mitochondrial defects upon E4F1 depletion, we also do not exclude the possibility that transformed cells exhibit deregulation of some yet unidentified mitochondrial activities that sensitize cancer cells to E4F1 depletion. Together with our finding that overexpression of E4F1 protein level, ranging from moderate (two- to fourfold) to high (4–13-fold), occurs in a large subset of AML samples, these observations raise an interesting hypothesis that some leukemic cells become addicted to the prosurvival functions of E4F1 during the process of cell transformation. In conclusion, our findings may have important implications for novel anticancer therapies, in particular for treatment of leukemias that have been shown to respond to chemotherapeutic agents that regulate the redox status.

E4F1-null and E4F1^flox mice were generated previously (Le Cam et al., 2004; Lacroix et al., 2010). Ink4a/Arf (Serrano et al., 1996) and RERT (Guerra et al., 2003) mice were provided by M. Serrano and M. Barbacid (Centro Nacional de Investigaciones Oncológicas, Madrid, Spain), respectively. E4F1-null, E4F1^flox, RERT, and Ink4a/Arf mice were intercrossed to obtain appropriate genotypes: E4F1^+/flox or E4F1^−/flox; RERT^KI/KI; Ink4a/Arf^+/+ or Ink4a/Arf^−/−. The mice were maintained on a mixed 129Sv/J/DBA/C57BL/6 background. E4F1^flox; RERT mice were genotyped by PCR on tail genomic DNA using the following primers. (a) E4F1 WT (E4F1⁺) and conditional KO flox alleles (E4F1^flox): 5′-CCTTGAGCACGGAGGAGAGC-3′ and 5′-GCCCTAGCCTGCTCTGCCATC-3′. (b) E4F1 constitutive KO allele (E4F1⁻): 5′-CACTGCCTTGGAGGACTTTG-3′ and 5′-CCTCTGTTCCACATACACTTCATTC-3′. (c) WT and KI RERT alleles: 5′-GTCAGTACACATACAGACTT-3′, 5′-TGAGCGAACAGGGCGAA-3′, and 5′-TCCATGGAGCACCCAGTGAA-3′. (d) Ink4a/Arf alleles: WT, 5′-ATGATGATGGGCAACGTTC-3′ and 5′-CAAATATCGCACGATGTC-3′; and KO, 5′-CTATCAGGACATAGCGTTGG-3′ and 5′-AGTGAGAGTTTGGGGACAGAG-3′.

In vivo recombination of the E4F1^flox allele was obtained by topical skin applications of 4OHT (Sigma-Aldrich) on shaved back skin of the transplanted recipient animals (2 mg/mouse/application; three applications). Cre-mediated E4F1 recombination efficiency was assessed by immunoblotting or by qPCR analysis on genomic DNA using primers specific for the E4F1 locus, as described in Fig. S1 A: primer A, 5′-GGCTGCTGCGTGGATTTC-3′; B, 5′-GCTAGGTAGGGTAGGAGGCTGTCT-3′; C, 5′-TTCGGTATAGTGTTGAGG-3′; and D, 5′-AGGGGCTGGGCTACAATGG-3′.

For survival analyses, 4OHT administration started 10 wk after transplantation and was repeated on a monthly basis throughout the entire life of the animals. For PET scan analyses, 4OHT (2 mg/mouse/application; three applications) was administrated around 20 wk after transplantation, when animals exhibited obvious signs of tumor development. For NAC-based in vivo experiments, NAC was administrated at the final concentration of 40 mM in the drinking water (changed every day) for 7 d before 4OHT administration and maintained afterward during the total duration of the experiment (15 d). All animal husbandry and experiments were approved by and performed in accordance with the guidelines of the Institute of Molecular Genetics Ethics Committee (agreement no. B-34-172-16).

Congenic C57BL/6-Ly5.1 mice were lethally irradiated (11 Gy) with a cobalt radiation source and used as recipient mice. 5 × 10⁶ fetal liver donor HSCs isolated from Ly5.2 E4F1^−/flox or E4F1^+/flox; RERT^KI/KI; Ink4a/Arf^+/+ or Ink4a/Arf^−/− embryonic day (E) 14.5 embryos, were injected i.v. into the tail vain of recipient (Ly5.1) mice. Peripheral blood was obtained from retroorbital sinuses of recipient mice under isoflurane anesthesia and analyzed every 4 wk after transplantation by flow cytometry with several combinations of lineage-specific antibodies: APC-CD19 for B cells, PerCP-CD3 for T cells, and APC-Gr1, FITC-F4/80, or PE-Mac2 for myeloid cells. PE-CD45.1 (Ly5.1) and FITC-CD45.2 (Ly5.2) antibodies were used to determine the level of reconstitution by identifying the proportion of donor cells in recipient mice. Fluorochrome-conjugated antibodies used for flow cytometry analyses (FACSCalibur flow cytometer; BD) were purchased from BD and eBioscience. Flow cytometry data were analyzed with FlowJo software (Tree Star).

E4F1 mRNA expression was evaluated in different hematopoietic compartments by RT-PCR. Cells were lysed in TRIZOL reagent (Invitrogen), and total RNAs were isolated according to the manufacturer’s recommendations. cDNAs were synthesized from 1 µg of total RNA using random hexamers and SuperScript III Reverse transcription (Invitrogen). Real-time qPCR was performed on a LightCycler 480 SW 1.5 apparatus (Roche) with Platinum Taq DNA polymerase (Invitrogen) and an SYBR Green mix containing 3 mM MgCl₂ and dNTPs 30 µM each; 45 cycles of 95°C for 4 s, 62°C for 10 s, and 72°C for 30 s. Results were quantified with a standard curve generated with serial dilutions of a reference cDNA preparation. PCR products were always loaded on an agarose gel to verify fragment size and purity of amplicons. RPL13A and HPRT transcripts were used for normalization. Primers sequences were as follows: E4F1 forward, 5′-CCAAAGCCTACCTGCTCAAG-3′; and reverse, 5′-CTGGGCATTCTTGGTTTTGT-3′; RPL13A forward, 5′-GAGGTCGGGTGGAAGTACCA-3′; and reverse, 5′-TGCATCTTGGCCTTTTCCTT-3′; and HPRT forward, 5′-AAGCCTAAGATGAGCGCAAG-3′; and reverse, 5′-TTACTAGGCAGATGGCCACA-3′.

Formalin-fixed tissues were embedded in paraffin, sectioned, and processed for routine hematoxylin and eosin (H&E) staining and IHC. Anti-Ki67 (SP6) and -MPO (Ab-1) antibodies were obtained from Thermo Fisher Scientific. Anti–CD3-ε (M-20) and -Pax5 (C-20) antibodies were obtained from Santa Cruz Biotechnology, Inc., and anti-F4/80 (BM8) and -Mac2 (M3/38) antibodies were obtained from eBioscience. IHC was performed on 4-µm tissue sections using the appropriate primary antibodies and the corresponding biotinylated secondary antibody coupled to streptavidin–peroxidase complex (ABC Vectastain kit; Vector Laboratories). Revelation was performed using the peroxidase substrates DAB (3,3′-diaminobenzidine; brown coloration) or VIP (dark purple coloration) from Vector Laboratories. TUNEL staining was performed according to the manufacturer’s (Roche) instructions.

PET scan analyses were performed at the Animage imaging department in Lyon (Centre d’Exploration et de Recherche Multimodal et Pluridisciplinaire, France). Animals analyzed by PET imaging belong to an experimental group composed of transplanted animals (4 mo after transplantation) that displayed palpable tumors in the spleen and a high percentage of F4/80-positive HS cells in the peripheral blood. Quantitative PET analyses of four independent animals of each genotype were performed on spleen and liver before (day 0) and after (day 15) E4F1 inactivation. Clinical grade radiolabeled fluorodeoxyglucose ([¹⁸F]FDG) was produced locally using a cyclotron. [¹⁸F]FDG uptake was calculated from a whole body acquisition of 30-min length on anesthetized animals performed with a high-resolution small animal PET scanner (Raytest ClearPET; Crystal Clear Collaboration) after i.v. injection of 300 µCi [¹⁸F]FDG in 100 µl of 0.9% NaCl solution in tail vein. Acquisition started 2 h after [¹⁸F]FDG injection to reduce tracer uptake in the bladder caused by renal excretion. PET scan images were reconstructed and quantified according to the injected amounts of [¹⁸F]FDG using the freeware Amide.

AO, 3MA, 4OHT, DAPI, propidium iodide (PI), Tiron, collagenase, Hepes, Na pyruvate, Tween 20, and NAC were purchased from Sigma-Aldrich. D3T was purchased from Axxora Platform, and Click-it EdU flow cytometry assay kit, CM-H₂DCFDA, MitoSOX, OxyBURST, and FITC-conjugated avidin were purchased from Invitrogen. Annexin V–FLUOS (herein referred to as annexin V) and BSA were purchased from Roche. RPMI Glutamax-I and FBS were purchased from Invitrogen and Biowest, respectively. For experiments on HS cell lines and primary macrophages, the working concentrations of the following chemical reagents were used: 1 µg/ml AO, 1 mM 3MA, 3 µM 4OHT, 100 µM Tiron, 200 µM NAC, and 10 µM D3T.

Viability was determined by manual counting after trypan blue exclusion. The proliferation rate of HS cells was assessed after incubation for 2 h with EdU by FACS analysis with the EdU flow cytometry assay kit (Invitrogen), according to the manufacturer’s recommendations. Mitotic cells were identified with the Ser10 phospho–histone H3 antibody (Cell Signaling Technology). Cell death was evaluated by flow cytometry upon annexin V/PI staining. Total and mitochondrial ROS levels were evaluated by flow cytometry or immunofluorescence microscopy, upon staining with CM-H₂DFCDA, OxyBURST, and MitoSOX probes or the MitoSOX probe only, respectively, according to the manufacturer’s recommendations. As shown in Fig. 6 A (right), the kinetics of MitoSOX staining were evaluated over a 75-min period. Quantitative data represent the fold increase of time-dependent changes in mean fluorescence intensity of MitoSOX staining measured by flow cytometry. Annexin V/PI and CM-H₂DCFDA/MitoSOX analyses were performed on freshly trypsinized live cells on a flow cytometer (FACSCalibur; BD).

Four HS cell lines (two E4F1^+/flox and two E4F1^−/flox) were established ex vivo out of nine lungs and zero out of nine livers harvested from unrecombined E4F1 KO or CT; Ink4a/Arf KO animals, as follows: fresh HSs of appropriate genotype were harvested from transplanted recipient mice of ∼20 wk of age, mechanically dissociated, digested in collagenase for 2 h at 37°C, and then cultured in RPMI Glutamax-I supplemented with 10% IFBS, 2.5 g/l glucose, 10 mM Hepes, and 1 mM Na pyruvate in a humidified 5% CO₂, 37°C incubator. Primary macrophages were harvested by injection of 10 ml PBS in the intraperitoneal cavity of mice of appropriate genotype, centrifuged, and plated for 2 h. Nonadherent cells were removed by repeated washing with culture medium. Purity (>90%) of the adherent macrophage population was evaluated by flow cytometry with Mac2 antibody. Primary macrophages were cultured in the same conditions and medium than HS cells lines. E4F1 inactivation in HS cell lines and primary macrophages was obtained by the addition of 4OHT to the culture medium, 2 h after plating.

Oxygen consumption was determined by two independent methods. The XF24 Extracellular Flux Analyzer (Seahorse Bioscience), based on fluorimetric sensors, was used for real-time in situ measurement of the OCR in E4F1 HS cells and primary intraperitoneal macrophages. Analyses were conducted on 80,000 cells/well in triplicate for each cell type (E4F1 CT or KO HS cells or primary macrophages). Mean values of OCRs were then calculated after correction to the total amount of cellular protein per well. For these experiments, cells were incubated for 1 h at 37°C in HCO₃-free DME containing 25 mM glucose and 1 mM pyruvate. Measurement of OCR was performed over 2 min in three measurement intervals to assess basal metabolic rate, oligomycin C (1 µg/ml final)–sensitive OCR associated to ATP production, and maximal respiratory capacity (upon FCCP [0.3 µM final] followed by rotenone [0.1 µM final] administration). Oxygen consumption was also measured using a temperature-regulated Clark-type O₂ electrode chamber (Strathkelvin Instruments) in murine E4F1 HS cells and human leukemic cell lines. Oxygen consumption was measured on 5 × 10⁶ cells for 5–15 min (until oxygen consumption ceased) at 37°C in the presence of 2.5 mM malate and 10 mM pyruvate, upon repeated ADP injection (three injections of ADP, 1 mM final). Respiration values were normalized to cell number. For ATP measurements, 20,000 cells were plated in quadruplicates in a 96-well plate, and ATP was measured 3 h later by using the Cell Titer Glo (Promega) luminescent assay. In brief, cells were incubated in the lysis buffer, and the luminescence was determined 10 min later. ATP levels were normalized to total protein levels.

Single cell suspensions of HS cells were mixed with RPMI containing 0.4% noble agar and plated on the top of a layer containing 1% noble agar in 6-well plates in triplicate (100,000 cells per plate). Fresh medium, with or without 4OHT, was added every 3 d. Colonies were stained after 3 wk of culture with a 0.5% crystal violet solution and counted.

In all experiments, HS cells were grown on a glass culture chamber (Falcon; BD). For LC3 immunostaining, cells were fixed with 2% paraformaldehyde for 5–10 min at room temperature, permeabilized by 0.1% Triton X-100 for 5 min at room temperature, incubated for 1 h with a blocking solution of PBS/1.5% BSA, and then stained for 1 h at 37°C with anti-LC3 antibody (Sigma-Aldrich) or overnight at 4°C with anticatalase (Rockland), anti-HO1 (Stressgen), NQO1 (Abcam), and anti-NRF2 (gift from E.E. Schmidt, Montana State University, Bozeman, MT) antibodies. After three washes in PBS/1% BSA/0.1% Tween 20, the slides were incubated for 45 min at 37°C in the dark with FITC-conjugated anti–rabbit secondary antibody. To visualize oxidized DNA (8-OHdG staining), cells were fixed with methanol at −20°C for 20 min and then incubated with 15 µg/ml FITC-conjugated avidin for 1 h at 37°C, as previously described (Radisky et al., 2005). To detect mitochondrial ROS, living cells were first incubated with MitoSOX (5 µM final) for 10 min at 37°C and then fixed with 2% paraformaldehyde. Acidic vesicular organelles were visualized by staining cells with 1 µg/ml AO for 10 min at 37°C. All fixed cells were costained with DAPI for 5 min and mounted in Vectashield imaging medium (Vector Laboratories) before observation with an upright microscope (Axioimager Z1/apotome; Carl Zeiss). MitoSOX images were acquired with a gamma adjustment of 0.55.

Total protein extracts were prepared by lysing HS cells or intraperitoneal primary macrophages in Triton X-100 lysis buffer (50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 50 mM NaF, 5 mM EDTA, 40 mM β-glycerophosphate, 1 mM Na orthovanadate, 10⁻⁴ M PMSF, 10⁻⁶ M leupeptin, 10⁻⁶ M pepstatin A, and 1% Triton X-100), separated by SDS-PAGE, and transferred to nitrocellulose membranes. Membranes were blocked in TBS containing 5% nonfat milk for 1 h at room temperature and incubated overnight at 4°C with primary antibodies. The following antibodies were used: E4F1 affinity-purified rabbit polyclonal antibody that was generated against the human full-length (aa 1–784) E4F1 protein, antiactin, and anti-LC3 antibodies (Sigma-Aldrich), anti-GAPDH (Santa Cruz Biotechnology, Inc.), anti–phospho-Akt (S473) and anti–total Akt (Cell Signaling Technology), and anti-PTEN (A2B1; Santa Cruz Biotechnology, Inc.). Quantitative analyses with the infrared imaging system (Odyssey; LiCor) was performed using DyLight-conjugated secondary antibodies from Thermo Fisher Scientific. Semiquantitative enhanced chemiluminescence–based immunoblotting was performed using horseradish peroxidase–conjugated secondary antibodies purchased from GE healthcare.

Adherent HS cells were fixed in 3.5% glutaraldehyde in 0.1 M Sorensen’s phosphate buffer, pH 7.4, for 1 h at room temperature. The cells were then scrapped and maintained overnight at 4°C in this fixative solution. After washes in Sorensen’s buffer, cells were postfixed in a 1% osmic acid plus 0.8% potassium ferrocyanide for 2 h at room temperature in the dark. After two washes in Sorensen’s buffer, the cells were dehydrated in a graded series of ethanol solutions (30–100%) and embedded in EmBed 812 resin. Thin sections (85 nm) were cut on an ultramicrotome (Ultracut E; Leica) and collected at different levels of each block. These sections were counterstained with uranyl acetate and examined on a transmission electron microscope (model 7100; Hitachi).

Viral particles were produced in HEK 293T cells by standard procedures after transfection of the following constructs: pLKO1 encoding shRNAs directed against human E4F1 (shE4F1#1, Sigma-Aldrich mission clone 1740s1c1; shE4F1#2, clone 2411s1c1), murine p53 (sh p53, Sigma-Aldrich mission shRNA clone 1526s1c1), or an irrelevant shRNA (sh Ct, clone shc002). 48 h after transfection, viral particles were harvested in the supernatant and added overnight in the culture medium of E4F1 HS cells or U937, HEL, HL60, or THP1 human cell lines in the presence of polybrene (8 µg/ml final).

Anonymous normal and AML BM samples were provided by the Laboratoire d’Hématologie, Hôpital St. Eloi, Centre Hospitalier Universitaire (collection no. DC-2010-1185 approved by the French minister of health). Frozen vials were thawed in complete medium, and whole-cell extracts were prepared in Laemmli buffer containing protease inhibitors. E4F1 and actin protein expression levels were determined by quantitative immunoblotting.

Unless otherwise indicated, the unpaired Student’s t test was used in all analyses, data in bar graphs are represented as mean ± SEM, and statistical significance was expressed as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Fig. S1 describes the HS mouse model used to decipher E4F1 functions during leukemic development. Fig. S2 illustrates tumor regression occurring upon E4F1 inactivation. Fig. S3 shows deregulation of the PI3K pathway in HS cell lines. Fig. S4 shows the immunophenotypic characterization of HS cell lines. Fig. S5 shows the impact of caspase inhibitor and antioxidant molecules on cell death of E4F1 KO HS cells. Fig. S6 shows the induction of a cellular antioxidant response upon E4F1 inactivation in HS cells. Fig. S7 shows the role of p53 in cell death occurring in E4F1 KO HS cells. Fig. S8 shows the overexpression of E4F1 in murine HS cell lines and human AML samples.

We thank the Réseau d’Histologie Expérimentale de Montpellier histology facility and Patricia Cavelier for processing our animal tissues, Animage in vivo imaging core facility (Lyon, France) for PET scan analyses, Montpellier Rio Imaging for cell imaging, and the Centre Régional d’Imagerie Cellulaire facilities for ultrastructure microscopy (Montpellier, France). We also thank Eric Jouffre, Marc Plays, and Karim Chebli for technical help in mouse handling, M. Barbacid and M. Serrano for providing us with the Ink4a/Arf KO and RERT KI strains of mice, S. Pattingre and Lucille Espert for valuable discussions on autophagy, and S. Thevenin for statistical analyses. We are grateful to all members of the L. Le Cam, C. Sardet, and L. Fajas laboratories for helpful discussions and critical readings of the manuscript.

This work was supported by the Agence Nationale pour la Recherche, the Association pour la lutte contre le Cancer (ARC), the Leukemia Program from the Fondation de France, the British Association for International Cancer Research (AICR) foundation, and la Fondation pour la Recherche Médicale (C. Sardet équipe labelisée 2007) and the institutional supports of the Institut National de la Santé et de la Recherche Médicale Avenir Program (L. Le Cam) and the Centre National de la Recherche Scientifique (C. Sardet). E. Hatchi and J. Caramel are supported by an ARC PhD and AICR postdoctoral fellowship, respectively.

The authors have no conflicting financial interests.