Multilineage hematopoiesis depends on rare multipotent BM-resident hematopoietic stem cells (HSCs; Orkin and Zon, 2008). HSCs possess multiple cell fate choices: they can remain quiescent, self-renew, undergo apoptosis, or differentiate into blood lineages. Strict regulation of these fates is critical for HSC maintenance, and dysregulation of the balance between these fates is a common feature of blood malignancies (Lobo et al., 2007). Because of their unique ability to sustain life-long multilineage hematopoiesis, HSCs rely on mechanisms safeguarding their integrity and protecting them from acquiring mutations, which could lead to their malignant transformation. Although HSC quiescence has been proposed to play protective functions against stem cell exhaustion and against the acquisition of mutations leading to malignant transformation (Lobo et al., 2007; Orford and Scadden, 2008), the role of autophagy in these processes remains unknown. Autophagy is a catabolic pathway characterized by the formation of a double-membrane vesicle, called the autophagosome, which engulfs cytoplasmic components and delivers them to lysosomes for degradation (Klionsky, 2007). The pathway is highly conserved in eukaryotes and is regulated both developmentally and by environmental factors such as nutrient/energy availability, hypoxia, and reactive oxygen species (ROS; Scherz-Shouval and Elazar, 2007). Formation of the autophagosome requires two ubiquitin-like conjugation systems, in which Atg12 (autophagy-related gene 12) is covalently linked to Atg5, and Atg8 is conjugated to phosphatidylethanolamine (Geng and Klionsky, 2008). Atg7 is a necessary catalyst in both conjugation systems and is therefore essential for autophagy (Tanida et al., 1999, 2001). Autophagy is considered a cell survival pathway that plays roles in development (Yue et al., 2003), immunity (Deretic and Levine, 2009), and cell death (Maiuri et al., 2007) and has, as such, been implicated in neurodegeneration, autoimmunity, and cancer (Levine and Kroemer, 2008). However, the role of autophagy in cancer remains controversial, in that it appears to have both tumor-promoting and -suppressive functions. Autophagy is induced by metabolic stress, which is commonly present in tumors and therefore acts as a survival factor for the tumor cells (White et al., 2010). In this study, we examine the role of the essential autophagy gene Atg7, which has no described autophagy-unrelated functions, in HSC survival and function, within the adult hematopoietic system. For that purpose, we conditionally deleted Atg7 throughout the hematopoietic system (Vav-Atg7−/− mice; Mortensen et al., 2010a), revealing a critical cell-autonomous requirement for autophagy in the maintenance of HSC integrity and demonstrating that autophagy suppresses myeloproliferation. RESULTS As homozygous knockout of Atg7 is neonatally lethal in mice (Komatsu et al., 2005), we conditionally deleted Atg7 in the hematopoietic system (Vav-Atg7−/− mice). Vav-Atg7−/− mice develop a progressive anemia, splenomegaly, and lymphadenopathy and survive for a mean of only 12 wk (Mortensen et al., 2010a). Mechanisms underlying the progression of anemia over time remained unexplained. In this study, we hypothesize that the lack of Atg7 in earlier stages of hematopoiesis could be responsible for the progressive and severe anemia found in Vav-Atg7−/− mice. Cell-intrinsic defects caused by the absence of mitochondrial autophagy (mitophagy) were found to cause both the lymphopenia and anemia of Vav-Atg7−/− mice. However, although anemia was still observed when the deletion of Atg7 was restricted to the erythroid lineage, it was milder and nonprogressive (Mortensen and Simon, 2010; Mortensen et al., 2010a). The phenotypic difference between pan-hematopoietic and erythroid knockouts of Atg7 was partly caused by the less efficient excision driven by the erythroid-specific ErGFP-Cre line (Heinrich et al., 2004) when compared with Vav-iCre (Mortensen et al., 2010a). However, this provided an incomplete explanation for the different phenotypes observed. Importantly, the erythropoietin receptor promoter that drives Cre expression in ErGFP-Cre mice is active only in erythroid progenitors (Heinrich et al., 2004), whereas the Vav gene regulatory elements (used to drive the expression of iCre in Vav-iCre mice) are active in all nucleated hematopoietic cells (Ogilvy et al., 1998, 1999b), including HSCs (Ogilvy et al., 1999a; de Boer et al., 2003). We therefore investigated the role of Atg7 in the maintenance of hematopoietic stem and progenitor cells (HSPCs). Atg7 is essential for HSC activity Atg7 expression analysis showed that it is uniformly expressed in long-term HSCs (defined as Lin−Sca-1+c-Kit+ [LSK] CD34−Flt3−), short-term HSCs (LSK CD34+Flt3−), and lymphoid-primed multipotent progenitors (LMPPs; LSK CD34+Flt3+; Fig. 1 A). To investigate a functional requirement for Atg7 in adult hematopoiesis, we analyzed Vav-Atg7−/− mice. We confirmed excision of Atg7 in sorted Vav-Atg7−/− BM lineage-negative cells enriched in HSPCs (Fig. S1 A). The role of Atg7 in the activity of HSPCs was first addressed by performing colony-forming cell (CFC) assays, in which BM cells from Vav-Atg7−/− mice generated a similar number of colonies compared with BM cells from WT littermates but failed to efficiently form secondary colonies after replating (Fig. 1, B and C). Figure 1. HSCs from Vav-Atg7−/− BM fail to reconstitute the hematopoietic system of lethally irradiated mice. (A) Relative Atg7 messenger RNA (mRNA) expression in murine long-term HSCs (LT-HSCs), short-term HSCs (ST-HSCs), and LMPPs was measured by real-time Q-PCR. Data are mean ± SEM (n = 3). (B) CFC assay performed on total BM cells from WT (Vav-iCre+; Atg7Flox/WT or Vav-iCre−; Atg7Flox/Flox) and Vav-Atg7−/− (Vav-iCre+; Atg7Flox/Flox) mice. The graph shows the mean number of CFC colonies ± SD counted on day 12 (n = 3). (C) The methylcellulose cultures shown in B were replated in methylcellulose 12 d after initial plating. The total number of colonies was counted after 10 d in culture. Mean values ± SD (n = 3) are shown. (D) Competitive repopulation assay. CD45.2+ BM cells from WT (left) or Vav-Atg7−/− (right) 9-wk-old mice were mixed 1:1 with CD45.1+ WT competitor BM cells and transplanted into lethally irradiated CD45.1+ WT recipients (n = 6 per group). Representative dot plots illustrating the CD45.2+ population frequency of donor-derived cells in peripheral blood of the recipient mice 16 wk after transplantation are shown. (E) Mean percentage (±SEM) of donor-derived CD45.2+ cells in peripheral blood of CD45.1+ recipients (n = 6 per group) described in D analyzed at 4, 12, and 16 wk after transplantation. (F) Lethally irradiated recipients were transplanted with 2 × 106 WT BM cells alone (n = 3), Vav-Atg7−/− BM cells (1.8 × 106 cells) in a 10:1 ratio with CD45.1+ competitor BM (0.2 × 106 cells; n = 4), or 2 × 106 Vav-Atg7−/− BM cells alone (n = 4). The mean percentage (±SEM) of CD45.2+ cells (gated as shown in D) in peripheral blood of the recipient mice is shown. Analysis was performed 4, 8, and 12 wk after transplantation. § indicates that recipients of Vav-Atg7−/− BM cells alone had to be culled 4 wk after transplantation because of poor health, and no further analysis could therefore be performed. (G) Kaplan-Meier survival curves of recipient mice of the in vivo reconstitution assays described in F. (H) Kaplan-Meier survival curves of lethally irradiated recipients of either 104 WT or Vav-Atg7−/− flow-sorted BM LSK cells. (I) Lethally irradiated recipients were transplanted with 2 × 106 WT (n = 5) or Vav-Atg7−/− (n = 6) FL cells. The mean percentage (±SEM) of CD45.2+ donor cells in peripheral blood of the recipient mice over time is shown. The percentages of CD45.2+ myeloid (CD11b+Gr1+), B (B220+CD19+), and T cells (CD4+ and CD8+) are indicated (populations gated as shown in Fig. S1 F). Four out of six of the Vav-Atg7−/− FL recipients had to be culled 7 wk after transplant because of their poor state. However, for simplicity, their percent blood populations are shown combined with those of the remaining recipient mice, which were bled 8 wk after transplantation. Two-tailed Mann-Whitney tests were performed on the indicated datasets (ns, nonsignificant; **, P 20% myeloid blasts in the BM, and (c) the myeloproliferation can be transplanted to recipients of Atg7−/− BM, FL, or LSK cells. Autophagy may indeed function as a protective mechanism against leukemogenesis. Interestingly Beclin 1 (Atg6) is monoallelically deleted in some cancers (Aita et al., 1999; Liang et al., 1999), implying that autophagy may be protective against tumor formation. In mouse models, autophagy has also been shown to suppress tumors by preventing the accumulation of DNA damage (Mathew et al., 2007) and p62 aggregates (Komatsu et al., 2007; Mathew et al., 2009). Furthermore, ageing Beclin 1+/− mice have a significantly higher incidence of spontaneous tumor development (Qu et al., 2003; Yue et al., 2003) than age-matched WT mice. Vav-Atg7−/− LSK cells show higher ROS levels and DNA damage and proliferate more than normal. We hypothesize that high ROS levels in HSPCs may lead to genetic changes and could eventually confer Vav-Atg7−/− HSPCs with a malignant phenotype. There is evidence that ROS-related damage causes mainly smaller/point mutations (Feig et al., 1994), which can have a major impact on tumorigenesis (Harada et al., 2004). Consistent with our hypothesis, we failed to identify copy number variations of large genomic regions in the increased CD11b+Gr1+ population in Vav-Atg7−/− mice by array comparative genomic hybridization and karyotyping analyses (unpublished data). Yet, we found small regions with copy number variations encompassing a few genes with links to leukemia (unpublished data). The atypical myeloid proliferation in Vav-Atg7−/− mice is likely to arise from an HSPC, as mice with LysM-Cre–mediated deletion of Atg7 in myeloid progenitors and myeloid cells (Clausen et al., 1999; Ye et al., 2003) survived normally and showed no signs of myeloid proliferation at autopsy (unpublished data). Moreover, despite the lack of hematopoietic repopulation, atypical myeloid cells were found in the lethally irradiated recipients of BM LSK cells from Vav-Atg7−/− mice within 2 wk of transplantation, suggesting that cells initiating these symptoms are present among LSK cells. CD11b+Gr1+ cells were found increased in the blood of Vav-Atg7−/− mice, and we show that this particular subset displays higher proliferation rates and up-regulates CD47. CD47 is expressed on HSPCs and leukemic cells and confers protection from phagocytosis by interacting with its receptor, SIRP-α, on macrophages (Jaiswal et al., 2009). Interestingly, the noncompetitive transplantation of Vav-Atg7−/− BM cells into either immunocompromised or lethally irradiated recipients led to myeloid infiltrates only at peripheral sites without BM involvement. However, BM involvement could be found in recipients of Vav-Atg7−/− FL. This could suggest that because the HSPCs from adult Vav-Atg7−/− mice have lost their repopulation ability, the atypical myeloid cells within that compartment have equally lost the ability to invade the BM. In contrast, the HSPCs of FL origin still able to reconstitute lethally irradiated mice can also form myeloproliferative foci in the BM. Relevance to human MPD and myelodysplastic syndrome (MDS) Based on our findings and on reports that autophagy levels decrease with age (Cuervo, 2008), we propose that failure to remove mitochondria, leading to the accumulation of DNA mutations in HSPCs, may account for the increased incidence of MPD/MDS in older patients. The accumulation of mitochondrial iron deposits in ringed sideroblasts has indeed been suggested as evidence of mitochondrial dysfunction in MDS (Fontenay et al., 2006). It was also recently found that erythroid precursors from high risk MDS patients (those likely to progress to leukemia) have lower autophagy levels compared with low risk MDS patients, highlighting a role for mitophagy in the progression of MDS to leukemia (Houwerzijl et al., 2009). Vav-Atg7−/− mice therefore represent a novel model for hematopoietic defects, which could be particularly beneficial for understanding the importance of mitochondrial quality control in the prevention of these diseases. In conclusion, we provide genetic evidence that Atg7 is an essential regulator of adult HSC maintenance. We propose that quiescent HSCs require the efficient process of autophagy, which controls mitochondrial mass, ROS levels, and genomic integrity, to maintain normal HSC functions and sustain multilineage hematopoiesis. The relationship between autophagy and leukemic transformation remains an open question meriting future investigation. MATERIALS AND METHODS Mice. Mice were bred and housed in the Department of Biomedical Services, University of Oxford in individually ventilated cages. Atg7Flox/Flox mice were crossed to Vav-iCre mice (from D. Kioussis, Medical Research Council National Institute for Medical Research, London, England, UK) to obtain Vav-iCre; Atg7Flox/Flox. Genotyping was performed on ear genomic DNA as described previously (de Boer et al., 2003; Komatsu et al., 2005). Male and female mice were used equally in all experiments. Vav-iCre−; Atg7Flox/Flox and Vav-iCre+; Atg7Flox/WT littermates were used equally as littermate controls. All animal experiments were approved by the local ethical review committee and performed under a Home Office license. Quantitative PCR (Q-PCR). RNA extraction and Q-PCR reactions were performed as previously described (Kranc et al., 2009). All experiments were performed in triplicate. Differences in input cDNA were normalized with a combination of Gapdh and Ubc expression. CFC assays. MethoCult GF M3434 medium (STEMCELL Technologies Inc.) was used to enumerate mouse CFCs. Three replicates were used per group in each experiment. Colonies were tallied at days 10–14. In vivo transplantation experiments. In competitive in vivo repopulation assays, CD45.1+ competitor BM cells were mixed with CD45.2+ test donor BM cells in a 1:1 or 1:10 ratio, whereas in noncompetitive repopulation assays, only CD45.2+ cells were transplanted into CD45.1+ lethally irradiated hosts. The competitor cell numbers for each experiment are stated in Table S1. Overall, 2 × 106 total BM cells, 2 × 106 FL cells, or 104 sorted LSK cells were injected intravenously into lethally irradiated (9 Gy) B6SJL CD45.1+ recipients. FL cells (CD45.2+) were obtained from 14.5 d postimplantation embryos. LSK cell BM and FL transplant recipients were bled 4, 8, 12, and 16 wk after transplantation, and multilineage reconstitution was monitored in peripheral blood. In the leukemia transplantation experiment, 2 × 106 BM cells were injected intravenously into sublethally irradiated (4.5 Gy) Rag-1−/− hosts. All transplantation experiments were terminated according to UK Home Office regulations. Determination of total BM counts. Tibias and femurs of both hind legs were taken from each mouse. These were crushed using a pestle and mortar, and a BM suspension was obtained. The nucleated cells within each suspension were then counted by the Trypan blue dye exclusion test of cell viability. Flow cytometry. Flow cytometry experiments were performed on CyAn or LSRII instruments (Dako), unless otherwise stated, and data were analyzed with FlowJo 9.1 for Mac (Tree Star, Inc.). Single cell suspensions from BM, spleen, and peripheral blood were surface stained with the indicated antibodies. In most cases, Lin markers were stained using unconjugated rat anti–mouse CD4 (RM4-5), CD5 (53–7.3) CD8a (53–6.7), CD11b (M1/70), B220 (CD45R), Ter119 (TER-119), and Gr1 (RB6-8C5), followed by staining with Cy5–R-PE–conjugated goat anti–rat IgG (Invitrogen). The same antibody clones were used for all lineage marker stainings. For staining of the BM myeloid compartment, lineage marker staining was performed with antibodies against CD4, CD5, CD8a, CD11b, B220, and Gr1 (clones as above) and was followed by staining with Qdot605-conjugated goat anti–rat IgG (Invitrogen). In the CLP staining, lineage markers were stained using APC-conjugated anti–mouse CD3e (145-2C11), CD4, CD8a, Gr1, B220 (RA3-6B2; BD), and CD19 (1D3; BD). In the NKP staining, lineage markers were stained using PECy5-conjugated anti–mouse CD4 (BD), CD8a (BD), CD19 (MB19-1), Ter119, Gr1, and CD11b. All antibodies were obtained from eBioscience, unless indicated otherwise. Other antibodies used for surface staining were FITC anti–mouse CD11b (M1/70), eFluor 450 anti–mouse CD11b (M1/70), FITC anti–mouse TCR-β (H57-597), PE or Pacific blue anti–mouse NK1.1 (PK136), APC anti–mouse CD19, PE anti–mouse CD16/32 (93), APC anti–mouse Gr1 (Ly-6G), APC-eFluor 780 anti–mouse CD117 (c-Kit), FITC anti–mouse CD47 (miap301), PECy7 anti–mouse CD41 (MWReg30), PECy5.5 anti–mouse Ter119, PE anti–mouse CD127 (A7R34), APC anti–mouse CCR9 (all from eBiosciences); Pacific blue anti–mouse CD45.2 (104), Pacific blue anti–mouse Sca-1 (E13-161.7), PE anti–mouse CD49b (DX5), APC or PECy7 anti–mouse CD150 (TC15-12F12.2), biotin anti–mouse CD105 (MJ7/18; BioLegend); streptavidin PETxRed, PE anti–mouse CD135 (Flt3, A2F10.1), FITC anti–mouse CD122 (TM-b1), and FITC anti–mouse Sca-1 (BD). Dead cells were always excluded using either DAPI or 7AAD. For active caspase 3 staining, cells were first surface stained, fixed and permeabilized (Fixation and Permeabilization kit; eBioscience), and stained with FITC-conjugated anti–active caspase 3 monoclonal antibody (BD). Alternatively, cells were fixed and permeabilized by incubation at −20°C for 2 h in 70% ethanol before staining with PE-conjugated anti-Ki67 antibody (BD) or with rabbit anti-53BP1 antibody (NB100-304; Novus Biologicals), followed by anti–rabbit IgG and IgM Alexa Fluor 488 (Invitrogen). Mitochondrial stains were performed after surface marker staining by incubating cells at 37°C for 30 min with 100 nM MitoTracker green, 100 nM NaO, 100 nM tetramethylrhodamine methylester, or 5 µM MitoSOX red (all from Invitrogen) and directly analyzed without fixing. Absolute cell numbers in peripheral blood were determined using TruCount tubes (BD); samples were then analyzed on a FACSCalibur machine (BD). Histology, tissue staining, and immunostaining. Full autopsies were performed on six Vav-Atg7−/− mice (one 9-wk-old male, one 10-wk-old male, and four 10-wk-old females), together with three WT controls (one 9-wk-old male and two 10-wk-old females). All major organs were examined macroscopically, harvested, and fixed in 4% neutral buffered formalin, before processing to paraffin. Frozen material was retained from approximately half of the mice. 4-µm sections were stained with hematoxylin and eosin (H&E) using standard techniques. Immunostaining was performed either manually or using an OptiMax automated staining machine (BioGenex). Monoclonal rat anti–mouse primary antibodies raised against CD205 (MCA949; clone CC98), CD3 (MCA500G), CD19 (MCA1439), c-kit (2B8), polyclonal rabbit anti–Pax-5 (Abnova), Ly6G/Gr1 (RB6-8C5; eBioscience), and polyclonal rabbit anti-ankyrin (Abcam) were used on frozen sections, with WT mouse lymph node as a positive control. Monoclonal rat anti–mouse CD45R (B220; R&D Systems) and polyclonal rabbit anti–mouse/human CD3 (A0452; Dako) were used on formalin-fixed, paraffin-embedded sections, with WT mouse lymph node as a positive control. Monoclonal rat anti–mouse primary antibodies raised against CD68 (MCA1957; AbD Serotec), CD11b (M1/70), and polyclonal rabbit anti–mouse serum raised against myeloperoxidase (Ab45977; Abcam) were used on frozen sections, with WT mouse lymph node as a positive control. VECTASTAIN ABC kits against rat or rabbit or the VECTASTAIN mouse on mouse immunodetection kit (Vector Laboratories) was used for primary antibody detection. Slides were counterstained with Mayer’s hematoxylin and mounted in DePex mounting medium (VWR International). Sections were examined by a specialist hematopathologist and photographed with a camera (DS-FI1; Nikon) with a control unit (DS-L2; Nikon) and a microscope (BX40; Olympus). Finally, BM smears were performed by smearing onto a slide 5 µl of BM suspension obtained by crushing tibias and femurs in 100 µl PBS. Slides were allowed to air dry and then May-Wright-Giemsa stained on a Hematek machine. Statistics. Statistical analyses were performed using Prism 4 for Mac (GraphPad Software, Inc.). Error bars represent SEM, and p-values were calculated with a two-tailed Mann–Whitney test unless stated otherwise. Online supplemental material. Fig. S1 shows in vivo reconstitution assays. Fig. S2 shows that Vav-Atg7−/− mice present myeloid infiltrates in a wide range of organs. Fig. S3 shows that myeloid cells from Vav-Atg7−/− mice express higher levels of the myeloid leukemia marker CD47. Table S1 lists organs presenting myeloid infiltrates in 9–10-wk-old Vav-Atg7−/− mice at autopsy. Table S2 lists histological features of sternal BM in the six Vav-Atg7−/− mice analyzed for myeloid infiltrates. Table S3 shows the transplantability of the myeloproliferation from Vav-Atg7−/− in the different transplantation settings. Online supplemental material is available at http://www.jem.org/cgi/content/full/jem.20101145/DC1.