Foxo3 Is Essential for the Regulation of Ataxia Telangiectasia Mutated and Oxidative Stress-mediated Homeostasis of Hematopoietic Stem Cells*

Unchecked accumulation of reactive oxygen species (ROS) compromises maintenance of hematopoietic stem cells. Regulation of ROS by the tumor suppressor protein ataxia telangiectasia mutated (ATM) is critical for preserving the hematopoietic stem cell pool. In this study we demonstrate that the Foxo3 member of the Forkhead Box O (FoxO) family of transcription factors is essential for normal ATM expression. In addition, we show that loss of Foxo3 leads to defects in hematopoietic stem cells, and these defects result from an overaccumulation of ROS. Foxo3 suppression of ROS in hematopoietic stem cells is mediated partly by regulation of ATM expression. We identify ROS-independent modulations of ATM and p16INK4a and ROS-mediated activation of p53/p21CIP1/WAF1/Sdi1 tumor suppressor pathways as major contributors to Foxo3-null hematopoietic stem cells defects. Our studies demonstrate that Foxo3 represses ROS in part via regulation of ATM and that this repression is required for maintenance of the hematopoietic stem cell pool.


Unchecked accumulation of reactive oxygen species (ROS) compromises maintenance of hematopoietic stem cells. Regulation of ROS by the tumor suppressor protein ataxia telangiectasia mutated (ATM) is critical for preserving the hematopoietic stem cell pool. In this study we demonstrate that the Foxo3 member of the Forkhead Box O (FoxO) family of transcription factors is essential for normal ATM expression. In addition, we show that loss of Foxo3 leads to defects in hematopoietic stem cells, and these defects result from an overaccumulation of ROS. Foxo3 suppression of ROS in hematopoietic stem cells is mediated partly by regulation of ATM expression. We identify ROSindependent modulations of ATM and p16 INK4a and ROSmediated activation of p53/p21 CIP1/WAF1/Sdi1 tumor suppressor pathways as major contributors to Foxo3-null hematopoietic stem cells defects. Our studies demonstrate that Foxo3 represses ROS in part via regulation of ATM and that this repression is required for maintenance of the hematopoietic stem cell pool.
Reactive oxygen species (ROS) 3 such as superoxide anions and hydrogen peroxide are by-products of oxidative metabolism and are involved in many signaling processes (1). However, excess accumulation of ROS resulting from defects in ROS scavenging is believed to impact cellular aging and the senescence process (1), whereas the ability to withstand oxidative stress has been correlated with enhanced longevity in several species (1). In agreement with deleterious effects of ROS, abnormal ROS accumulation has been implicated in the pathogenesis of various diseases, including bone marrow failure and disorders with high susceptibilities to malignancies such as ataxia telangiectasia, Fanconi anemia, as well as malignancies of hematopoietic tissues (2)(3)(4)(5).
Regulation of ROS by ATM is essential for hematopoietic stem cell self-renewal, because loss of ATM leads to depletion of the hematopoietic stem cell pool and bone marrow failure in old mice in a ROS-dependent manner (2). ATM serine/threonine protein kinase is a critical enzyme in the regulation of stress response to DNA damage, specifically double strand DNA break (6). Loss of function mutations in ATM is associated with ataxia telangiectasia that is an autosomal disorder characterized by neuronal degeneration, immunodeficiencies, genomic instability, predisposition to lymphomas and other malignancies, and extreme sensitivity to ionizing radiation (7). Although ROS are found accumulated in several cell types and tissues of ataxia telangiectasic patients (8) and antioxidant enzymes are either up-regulated or down-regulated in ATM mutant cells, the mechanism of ATM regulation of ROS is not clear (6,9).
The forkhead homolog DAF-16 is a major mediator of defense against oxidative stress in Caenorhabditis elegans (10). FOXO1, FOXO3a, FOXO4, and FOXO6 are the human homologs of DAF-16 and belong to the forkhead FoxO family of winged helix transcription factors. Although FOXO6 is predominantly expressed in the brain, FOXO1, FOXO3a, and FOXO4 are ubiquitous. As with DAF-16 in C. elegans, FoxO transcription factors are critical regulators of oxidative stress in mammals (11,12). In particular, activated FOXO3a up-regulates ROS scavenging enzymes in cultured cells (13,14) and is required for optimum expression of these enzymes in primary mouse erythroid cells (12). FOXO3a responds to cellular stress (including but not limited to oxidative stress) by inducing cell cycle arrest at G 1 -S and G 2 -M checkpoints, repair of damaged DNA, and apoptosis via up-regulation of genes that control these processes (reviewed in Ref. 15).
FoxO proteins are negatively regulated by the highly conserved PI 3-kinase/AKT signaling pathway (15). In response to factors such as hematopoietic cytokines (16 -19) or oncogenic stimuli such as BCR-ABL oncoprotein (19), the AKT family of serine/threonine protein kinases phosphorylates nuclear FoxO and mediates their translocation to the cytosol where FoxO is unable to induce the expression of their transcriptional targets (16).
In response to oxidative stress, phosphorylation of FoxO mediated by mammalian sterile 20-like kinase-1 (MST1) (20) and Ral/Jun N-terminal kinase-dependent pathway results in FoxO nuclear translocation and activation (21). Identification of FoxO at sites of chromosomal translocations in human tumors, including acute myeloid leukemias, suggested a role for these factors in tumorigenesis.
Under certain conditions of stress, such as when food is scarce, DAF-16 is activated to promote the dauer formation. We found the characteristics of the dauer that is in "juvenile form, resilient, long lived, and reproductively immature" (10), to resemble in many respects the "quiescent" state of stem cells. Because quiescence is (a) a major characteristic of stem cells, (b) reminiscent of dauer formation, and (c) protected from apoptosis by antioxidant function of FoxO homolog DAF-16, we postulated that the Foxo3 (the mouse homolog of FOXO3a) transcription factor that is the most highly expressed FoxO in the bone marrow (12) may protect mammalian stem cells from oxidative stress.
Recently, conflicting data as to the function of Foxo3 in the regulation of hematopoietic stem cell activity have been reported (11,22). Studies by Tothova et al. (11) found a loss of hematopoietic stem cell activity only when all three Foxo1, Foxo3, and Foxo4 were deleted in mice. These authors did not find any hematopoietic stem cell abnormalities as a result of loss of Foxo3 alone (11). Conversely, Miyamoto et al. (22) found that Foxo3 is essential for the maintenance of hematopoietic stem cell activity. However, at steady state, these authors did not detect a significant defect of hematopoietic stem cell frequency in young mice (22). Here we provide our data showing that Foxo3 is essential for the regulation of hematopoietic stem cell frequency and activity in young mice. We further demonstrate the following: (a) Foxo3 is critical for the regulation of oxidative stress in hematopoietic stem cells, and (b) Foxo3 modulation of oxidative stress mediates Foxo3 regulation of hematopoietic stem cell activity. Our in vivo and ex vivo data demonstrate that loss of Foxo3 results in ROS accumulation that mediates activation of p53/p21 CIP1/WAF1/Sdi1 and in alterations of hematopoietic stem cell cycling, specifically exit from quiescence and G 2 /M arrest.
Furthermore, we demonstrate that Foxo3 is required for the expression of ATM gene that is an essential factor for hematopoietic stem cell self-renewal. Using genetic loss and gain of function approaches by RNA interference and retroviral transduction of primitive hematopoietic stem cells, we demonstrate that ATM deficiency participates in perturbations of ROS regulation in Foxo3-null hematopoietic stem cells.
These novel findings establish Foxo3 as a physiological regulator of oxidative stress in hematopoietic stem cells, demonstrate that Foxo3 modulation of ATM regulates oxidative stress in hematopoietic stem cells, and provide insight into the mechanism of Foxo3 regulation of stem cell cycling. We discuss what we believe is the source of discrepancies between our results and previous reports (11,22) and present our model of Foxo3 regulation of hematopoietic stem cells.

EXPERIMENTAL PROCEDURES
Mice-The generation and genotyping of mice were performed as described previously (23). Briefly, heterozygotes Foxo3 ϩ/Ϫ mice (129ϫFVB/n) were backcrossed to FVB/n (F3), and progenies (that were Cre Ϫ ) were intercrossed with littermates to generate the experimental cohort (23). Littermates were used as controls in all experiments. Protocols were approved by the Institutional Animal Care and Use Committee of Mount Sinai School of Medicine.
Retroviral Constructs and Plasmids-The full-length human ATM cDNA from pMAT1 (24) was inserted upstream of the IRES in the XhoI site of MSCV-IRES-GFP (MIG) vector to construct MIG-ATM. MIG-Foxo3 was described previously (19), to construct pL4-shRNA Foxo3, synthesized shRNA oligomers targeting Foxo3 were annealed and cloned into EcoRI/XhoI site downstream of H1 RNA polymerase III promoter of pL4 (Invitrogen) vector.
N-Acetyl-L-Cysteine Treatment-Mice were injected intraperitoneally with 100 mg/kg body weight (Sigma) in phosphatebuffered saline solution, pH 7.4, every day for the indicated time period. For cultured cells, NAC (100 M) was used for the indicated time.
Cell Cycle Analysis-Mice were injected intravenously with 2 mg of 5-bromo-2-deoxyuridine (BrdU). At 19 h post-injection, bone marrow cells were collected, stained for LSK, fixed, and stained with 7-amino-actinomycin D and anti-BrdU antibody (Pharmingen) for analysis. The G 0 /G 1 analysis was performed as described (28), and FACS-sorted LSK populations were collected in Hanks' buffered salt solution medium containing 10% FCS, 1 g/liter glucose, 20 mM Hepes, pH 7.2, 50 g/ml verapamil. Cells were then washed; Hoechst 33342 (20 g/ml, Invitrogen) was added, and cells were immediately incubated at 37°C for 45 min after which pyronin Y (1 g/ml, Sigma) was added. Cells were incubated for another 15 min at 37°C, washed in cold medium, and resuspended in medium containing 1 g/ml of propidium iodide. Samples were immediately analyzed by flow cytometry (BD Biosciences LSRII).

RNA Isolation and Quantitative Real Time (QRT)-PCR-
RNA isolation and QRT-PCR were conducted as described (12). In brief, total RNA was isolated using RNeasy mini kit (Qiagen) according to the manufacturer's instructions. First strand cDNA was synthesized using SuperScript (Invitrogen), QRT-PCRs were performed using SYBR Green JumpStart TM Taq ReadyMix (Sigma) in duplicate (Light Cycler 2.0, Roche Diagnostics). Gene-specific primers span the intron-exon boundary by Primer Express 2.0 (ABI). Relative quantification was achieved by normalization against endogenous ␤-actin. For primer sequence, see supplemental Table 1. In NIH 3T3 and MEF experiments, TRIzol (Invitrogen catalog number 15596-026) was used for total RNA extraction.

ROS-mediated Hematopoietic Stem Cell Deficiency in Foxo3-
null Mice-Foxo3 is the most highly expressed FoxO in the bone marrow (12). Nuclear localization is often used as a surrogate assay for FoxO activity. Hematopoietic stem cells are highly limited in numbers, and only few cells can be obtained from each mouse (average of 5000 cells from two mice). We analyzed FoxO expression by both QRT-PCR and immunostaining in hematopoietic stem cells. Analysis of FoxO subcellular localization showed that although both Foxo1 and Foxo3

Foxo3 Regulates ATM and ROS in Hematopoietic Stem Cells
are expressed at similar levels in hematopoietic stem cells (Fig. 1A), Foxo3 is almost entirely nuclear (over 95% nuclear as compared with 40% nuclear for Foxo1, see Fig. 1, B and C) suggesting that Foxo3 is presumably highly active (and is the most active FoxO) in hematopoietic stem cells. Importantly, the expression of Foxo1 was not up-regulated, and its subcellular localization was only increased by 15% in Foxo3-null hematopoietic stem cells (Fig. 1). Foxo4 transcript or protein was barely detectable in hematopoietic stem cell compartment (Fig. 1). These results suggest that Foxo3 is the prevalent active FoxO in hematopoietic stem cells, and Foxo1 or Foxo4 activity does not fully substitute for loss of Foxo3.
Thus, to address the potential evolutionary role of FoxO in the regulation of quiescence of hematopoietic stem cells, we focused on Foxo3 and investigated the hematopoietic stem cell compartment in Foxo3 Ϫ/Ϫ mice. We found a significant suppression of Foxo3-deficient hematopoietic stem cell numbers and activity (Fig. 1). In contrast to Miyamoto et al. (22), we found that the frequency ( Fig. 2A) and the size (Fig. 2B) of the Foxo3-null hematopoietic stem cell compartment was significantly reduced as compared with the wild type. The hematopoietic compartment that is depleted from mature cells (called lineagenegative population) and expresses cell surface markers Sca1 and high levels of c-Kit receptor (Lin Ϫ Sca1 ϩ c-Kit high , LSK) encompasses all stem cell activity, and this population was substantially reduced in young Foxo3-deficient mice as compared with wild type littermates (Fig. 2). Hematopoietic stem cell (LSK) compartment is heterogeneous and consists of distinct populations of long term HSC (LT-HSC) and short term (ST-HSC) repopulating HSC (29). LT-HSC (Lin Ϫ Sca1 ϩ c-Kit high CD34 Ϫ/low Flk2 low ) have extensive self-renewing potential and give rise to ST-HSC (Lin Ϫ Sca1 ϩ c-Kit high CD34 ϩ Flk2 low ) with more restricted self-renewal  potentials. Foxo3-deficient LT-HSC were the most highly affected in the hematopoietic stem cell compartment (Fig. 2B, p Ͻ 0.02) suggesting that the self-renewing capacity of Foxo3null hematopoietic stem cell was compromised. This was further supported by the suppression of Foxo3 Ϫ/Ϫ hematopoietic stem cell functional activity. The ability of Foxo3-deficient hematopoietic stem cell in generating progenitors after 4 weeks in long term culture (LTC-IC assay) in vitro (25) was notably decreased as compared with their wild type counterparts (Fig. 2C, p Ͻ 6.2 ϫ 10 Ϫ10 , n ϭ 30). Similarly, the ability of Foxo3-deficient hematopoietic stem cells to repopulate in vivo lethally irradiated animals over several weeks was also severely compromised ( Fig.  2D) suggesting that the suppression of stem cell activity was intrinsic to Foxo3-null hematopoietic stem cells. We futher evaluated the long term repopulating ability of Foxo3-null HSC by using highly purified HSC from Foxo3 Ϫ/Ϫ mice backcrossed to C57BL6 mice (F10) in our in vivo competitive repopulation assay. We transplanted 100 CD34 Ϫ Lin Ϫ c-kit ihigh Sca-1 ϩ that are highly enriched in long term HSC from either wild type or Foxo3 Ϫ/Ϫ (backcrossed to C56BL6, F10, CD45.1) along with 2 ϫ 10 5 total bone marrow (CD45.2) into lethally irradiated recipients (CD45.2) and compared the contribution to peripheral blood formation of these cells with that of wild type (CD45.1) cells. As shown in Fig. 2E, these experiments confirmed that Foxo3-null hematopoietic stem cells are highly compromised in their long term ability to reconstitute multiline blood formation in lethally irradiated recipient mice. Together these results strongly indicate that Foxo3 Ϫ/Ϫ mice are markedly deficient in their hematopoietic stem cell compartment (Fig.  2), specifically in cells with long term repopulating ability (Fig. 2B).
We next asked whether the defect in Foxo3-null hematopoietic stem cell compartment resulted from abnormalities in scavenging ROS (Fig. 3). Using a hydrogen peroxidesensitive probe, we found ROS to be highly accumulated in Foxo3-deficient populations of hematopoietic cells containing all stem cells (LSK cells) (Fig. 3, A and B). Increased ROS concentrations were specific to Foxo3 Ϫ/Ϫ hematopoietic stem cells (LSK), because accumulation of ROS was similar in total WT and Foxo3-deficient bone marrow cells (Fig. 3B, compare top and bottom panels).
In agreement with these results, several antioxidant enzymes, including catalase and mitochondrial superoxide dismutase 2 (SOD2, MnSOD), which are known FoxO targets, and glutathione peroxidase 1 (GPX-1), which is not known to be regulated by FoxO, were profoundly and specifically down-regulated in Foxo3-null LSK cells as analyzed by QRT-PCR (Fig. 3C). These results suggest that Foxo3 is the chief regulator of these antioxidant enzymes in hematopoietic stem cells, and potential activity of Foxo1 and/or Foxo4 does not fully complement loss of Foxo3 in hematopoietic stem cells.
Importantly, reducing ROS using NAC (100 M) that is a generic ROS scavenger rescued the Foxo3-deficient hematopoietic stem cell phenotypes and the numbers of hematopoietic stem cell-generated colonies over a 4-week period in culture (LTC-IC assay, Fig. 3D). These findings demonstrate that abnormal accumulation of ROS in Foxo3-null hematopoietic stem cells leads to the defective activity of hematopoietic stem cells. These combined findings strongly point to the deleterious impact of unchecked ROS as a result of loss of Foxo3 on inducing defective hematopoietic stem cell activity.
To evaluate whether loss of Foxo3 has any specific impact on hematopoietic stem cell quiescence, G 0 distribution of WT and Foxo3-deficient hematopoietic stem cells (LSK) was compared using measurements of RNA versus DNA contents. These studies showed that loss of Foxo3 results in a significant reduction of the proportion of bone marrow hematopoietic stem cells (LSK) in G 0 that is concomitant with the increased proportion of these cells in S/G 2 /M suggesting that Foxo3 promotes hematopoietic stem cell quiescence (Fig. 4, A and B). These results suggested that hematopoietic stem cells are in cycle and are producing progenitors. However, despite these results, Foxo3-deficient hematopoietic stem (LSK) cells cultured under optimum growth factor conditions showed significantly reduced production of total cells (Fig. 4C, upper panel) and progenitors (Fig. 4C, lower panel). To further investigate this observation, the cell cycle distribution of the hematopoietic stem cell compartment using the nucleotide analog BrdU uptake in vivo was performed. As shown in Fig. 4E, loss of Foxo3 did not significantly impact cell cycle distribution in the whole bone marrow using BrdU uptake. In contrast, Foxo3 deficiency resulted in an altered increase in G 2 /M phase and decreased proliferation as measured by reduced incorporation of BrdU into the S phase of Foxo3-deficient hematopoietic stem cells (LSK, Fig. 4, D and F). These results were in sharp contrast to the observations reported by Miyamoto et al. (22). In contrast to its impact on cell cycle distribution, loss of Foxo3 did not appear to significantly alter the apoptotic rate of hematopoietic stem cells as measured by annexin-V binding of freshly isolated LSK cells (data not shown) or by the analysis of caspase 3 cleavage in FACSsorted LSK cells ex vivo (supplemental Fig. 1). Together these studies suggest that loss of Foxo3 results in loss of quiescence of hematopoietic stem cells that is concomitant with alterations in the ability of hematopoietic stem cells to complete their cycling.
ROS-mediated Activation of p53 Tumor Suppressor Pathways in Foxo3-deficient Hematopoietic Stem Cells-To obtain insight into the Foxo3-null hematopoietic stem cell cycle alterations, we analyzed transcript levels of candidate molecules previously implicated in hematopoietic stem cell cycling and/or known to be transcriptional targets of Foxo3. This analysis indicated a significant reduction in expression of Foxo3 targets the cyclin-dependent kinase (CDK) inhibitor 1B Cdkn1b (p27 KIP1 ) transcript (17), Ccng2 (cyclin G2) (30), and Ccnb1 (cyclin B) (31) in Foxo3-null hematopoietic stem cells (LSK, Fig. 5A). Expression of p27 KIP1 and cyclin G2, both important regulators of hematopoietic stem cell cycling (28,32), was profoundly downregulated in Foxo3-deficient hematopoietic stem (LSK) cells (Fig. 5A). Similarly, Foxo3 transcriptional target cyclin B that is required for completion of cell cycle and mitosis was significantly reduced in Foxo3 Ϫ/Ϫ LSK cells (Fig. 5A). These results strongly suggest that Foxo3 is a key transcriptional regulator of p27 KIP1 , cyclin G2, and cyclin B in hematopoietic stem cells, and potential functional Foxo1 (or Foxo4) does (do) not complement for loss of Foxo3 in the regulation of these genes.
Unexpectedly, however, the expression of another Foxo3 transcriptional target CDK inhibitor Cdkn1a (p21 CIP1/WAF1/Sdi1 ), with critical functions in the repression of hematopoietic stem cell cycling (33), was robustly up-regulated in Foxo3 Ϫ/Ϫ hematopoietic stem (LSK) cells (Fig. 5B). p21 CIP1/WAF1/Sdi1 is an essential regulator of G 1 and G 2 checkpoints in response to oxidative stress and DNA damage either as part of a transcriptional program activated by p53 tumor suppressor or independently (34,35). Up-regulation of p21 CIP1/WAF1/Sdi1 was associated with a significant up-regulation of p53 expression (Fig. 5B) suggesting that enhanced expression of p21 CIP1/WAF1/Sdi1 was mediated by p53 activation in Foxo3-deficient hematopoietic stem cells. In addition, known stress-response targets of activated p53, such as the DNA repair gene growth arrest and DNA damage-inducible 45␣ (Gadd45a), also a transcriptional target of Foxo3, and the antioxidant gene sestrin 2 (Sesn2), were significantly up-regulated in Foxo3-deficient as compared with WT hematopoietic stem cells (Fig. 5B), further supporting that p53 is activated as part of an antioxidative stress response in Foxo3-deficient hematopoietic stem cells.
To evaluate the role of ROS in alterations of Foxo3-deficient hematopoietic stem cells (LSK), the gene expression pattern of hematopoietic stem cells treated with a ROS scavenger was examined. Treatment with NAC for 5 days in vivo followed by 2 days in vitro resulted in a significant down-regulation of both p21 CIP1/WAF1/Sdi1 and p53 expression in Foxo3-deficient but not WT LSK cells (Fig. 5C), further demonstrating that up-regulation of p21 CIP1/WAF1/Sdi1 transcription in Foxo3-null hematopoietic stem cells was mediated by an antioxidant response of activated p53. As anticipated, scavenging ROS in Foxo3-null hematopoietic stem cells did not rescue the expression of direct targets of Foxo3 such as cyclin G2 (Fig. 5C) or p27 KIP1 (data not shown) that were highly repressed in Foxo3-deficient hematopoietic stem cells (Fig. 5A). These results confirmed that cyclin G2 and p27 KIP1 are direct targets of Foxo3, and their expression is not mediated by ROS. Although the basal levels of cyclin G2 (Fig. 5A) or p21 CIP1/WAF1/Sdi1 (Fig. 5B) in wild type LSK were higher in ex vivo cultured cells as compared with levels found in freshly isolated wild type LSK (Fig. 5B, likely because of conditions in culture as compared with in vivo), the notable reduction upon scavenging ROS was only seen with p21 CIP1/WAF1/Sdi1 transcript but not with cyclin G2 and only in Foxo3-null and not in wild type cells (Fig. 5C).
ATM Lies Downstream of Foxo3-ATM is an essential regulator of stress response and ROS-induced DNA damage that is critical for hematopoietic stem cell self-renewal (2, 6). ATM activates p53 via multiple pathways, including direct phosphorylation. To our surprise and in contrast to the observed upregulation of p53 expression, ATM transcript expression was significantly and unexpectedly down-regulated in Foxo3-null hematopoietic stem cells (Fig. 6A, ϳ2-fold reduction, p ϭ 0.0035). Importantly, this reduction in transcript expression was associated with similar suppression of ATM protein expression in primitive Foxo3-deficient hematopoietic cells (Fig. 6B, over 2-fold reduction) suggesting that ATM lies downstream of Foxo3. The 2-fold reduction of ATM expression in Foxo3-null primitive hematopoietic cells is of major significance because ATM is required for genomic stability and ATM ϩ/Ϫ mice exhibit an altered stress response phenotype of analysis of three cDNAs each generated from a pool of two to three WT and Foxo3 Ϫ/Ϫ mice. C, wild type and Foxo3 Ϫ/Ϫ mice were treated in vivo with or without NAC (100 mg/kg) every day for 5-7 days. BM LSK were then isolated and incubated with NAC ex vivo for an additional 2 days after which total RNA was isolated for QRT-PCR analysis. Results are shown as mean Ϯ S.E. of analysis of at least three cDNAs each generated from pool of two to three wild type and Foxo3 Ϫ/Ϫ mice. (36) suggesting that loss of half of ATM may participate in the defective Foxo3-null hematopoietic stem cell activity or stress response. ATM-deficient hematopoietic stem cells in old mice exhibit an up-regulation of both p16 INK4a and p19 ARF tumor suppressors (2) that are generated from alternative open reading frames of the INK4a locus. p16 INK4a and p19 ARF inhibit cell cycle progression through inhibition of inhibitors (therefore activation) of retinoblastoma and p53 proteins, respectively. Although in Foxo3-deficient hematopoietic stem cells, p16 INK4a expression was significantly up-regulated, the expression of p19 ARF remained unchanged (Fig. 6A).
In contrast to the expression of p53/p21 CIP1/WAF1/Sdi1 , the levels of ATM or p16 INK4a transcripts in Foxo3-null hematopoietic stem cells did not change significantly in response to the NAC antioxidant treatment (Fig. 6C). These results further confirmed that reduced ATM expression may be a direct result of loss of Foxo3 and not ROS accumulation in primitive hematopoietic stem cells. In agreement with a functional ATM downstream of Foxo3 in hematopoietic stem cells, the phosphorylation of ATM serine 1981 that is a marker of ATM kinase activity (37) was slightly albeit significantly reduced at the steady state in Foxo3-null hematopoietic stem cell-containing population of LSK cells as compared with their wild type counterparts (Fig. 6D).
NAC treatment led to a significant down-regulation of p19 ARF in Foxo3-deficient hematopoietic stem cells (LSK; Fig.  6C). This was highly unexpected because p19 ARF transcript levels were similar in WT and Foxo3-null hematopoietic stem cells (Fig. 6A). These findings suggest that ROS mediate regulation of p19 ARF expression in Foxo3-deficient but not in WT hematopoietic stem cells. Because p19 ARF inhibits the degradation of p53 protein through sequestration of MDM2 (38), these results suggest that ROS may induce activation of p53 antioxidant function, both directly by stimulating the expression of p53 transcript and indirectly through up-regulation of p19 ARF . These combined findings suggest that ROS accumulation leads to activation of p19 ARF /p53 signaling pathway resulting in upregulation of p21 CIP1/WAF1/Sdi1 .
Foxo3 Is Required for Normal ATM Expression-We next asked whether ATM expression is modulated by Foxo3 and thus examined the expression of ATM in Foxo3-deficient primitive hematopoietic cells expressing Foxo3 ectopically. WT and Foxo3-null bone marrow mononuclear cells highly enriched in primitive hematopoietic cells as a result of a 4-day treatment with a high dose of 5-FU (that ablate proliferating hematopoietic cells (25,39)) were isolated and transduced with the bicistronic retroviral vector MSCV-IRES-GFP encoding for Foxo3 (Fig. 7A). Retroviral expression of Foxo3 and not the vector control (Fig. 7B) resulted in a significant up-regulation of ATM transcript in GFP ϩ -transduced Foxo3-null primitive hematopoietic cells suggesting that Foxo3 regulates ATM expression. Expression of Foxo3 also induced significant down-regulation of p53 and p16 INK4a (Fig. 7B) without modifying the expression of p19 ARF (data not shown) in Foxo3 Ϫ/Ϫ primitive hematopoietic cells.
As anticipated ectopic retroviral expression of Foxo3 and not the vector control induced the expression of SOD2 and catalase (Fig. 7C) in Foxo3-null primitive hematopoietic cells. In addition, overexpressed Foxo3 strongly and significantly induced the expression of SOD1 (cytoplasmic SOD, CuZn-SOD) and GPX-1 in primitive hematopoietic cells (Fig. 7C) suggesting that expression of these antioxidant enzymes is also regulated by Foxo3. The mechanism through which Foxo3 induces SOD1 and GPX1 is not known.
To further examine Foxo3 regulation of ATM, we tested ATM expression in NIH 3T3 cells overexpressing Foxo3 by both QRT-PCR and Western blot. As in Foxo3-null primitive hematopoietic cells, ectopic expression of Foxo3 and not the vector control in NIH 3T3 cells resulted in a significantly enhanced ATM expression as measured by both QRT PCR (Fig.  8A) and Western blot (data not shown). Importantly, targeted silencing of Foxo3 expression by RNA interference using two distinct shRNAs resulted in significant reduction of expression of both ATM transcript (Fig. 8B) and protein (Fig. 8C) in NIH 3T3 cells. A 50% reduction in Foxo3 expression resulted in over 25% reduced expression of ATM (Foxo3 shRNA3), whereas 80% reduction of Foxo3 resulted in 36% suppression of ATM protein (Foxo3 shRNA4). However, because less than 100% transfection efficiency may underestimate these results, we repeated these experiments and FACS-sorted GFP ϩ (all expressing shRNA targeting Foxo3) NIH 3T3 cells for analysis (Fig. 8D). These experiments further confirmed our previous findings and showed that over 95% reduction in Foxo3 expression leads to 70% reduction in ATM transcript expression further demonstrating that Foxo3 is required for ATM expression (Fig. 8D).

Foxo3 Regulates ATM and ROS in Hematopoietic Stem Cells
Foxo3 and ATM Regulate the Expression of Antioxidant Enzymes Independently-Although some ATM-null cells and tissues, including ATM Ϫ/Ϫ hematopoietic stem cells (2), accumulate ROS, depending on tissues and cells, antioxidant enzymes are up-or down-regulated in ATM-null cells, and the exact mechanism of ATM regulation of ROS is unknown (6,8,9). Thus, we examined whether relative loss of ATM contributes to the deficiency of antioxidant gene expression observed in Foxo3-null primitive hematopoietic cells (Fig. 3C). To this end, retroviral supernatant highly expressing ATM was prepared (Fig. 9A) and used to transduce Foxo3-null primitive hematopoietic cells (Fig. 9B). Ectopic retroviral expression of ATM and not the vector control corrected the transcript level of p16 INK4a , without much impact on p19 ARF or p53 transcript expression in Foxo3-null primitive hematopoietic cells suggesting that ATM may regulate p16 INK4a expression. These results further suggest that up-regulation of p16 INK4a expression is the result of the relative loss of ATM expression in Foxo3-null primitive hematopoietic cells. Interestingly, retroviral expression of ATM restored the expression of antioxidant enzymes in GFP ϩ Foxo3-null primitive hematopoietic cells (Fig. 9C) suggesting that ATM is sufficient for the expression of SOD1, SOD2, catalase, and GPX1 in the absence of Foxo3.
We next asked whether ATM is required for Foxo3 regulation of antioxidant enzymes. ATM-deficient or WT MEFs (40) retrovirally expressing either empty vector or Foxo3 or ATM were examined for gene expression of antioxidant enzymes. As shown in Fig. 10, SOD2 was significantly down-regulated in ATM-deficient MEFs suggesting that ATM is required for SOD2 expression. Ectopic expression of either Foxo3 or ATM induced marked up-regulation of SOD1, SOD2, and GPX-1 in ATM-deficient MEFs suggesting that Foxo3 regulation of these antioxidant enzymes does not require ATM expression (Fig.  10). Up-regulation of the expression of these enzymes by overexpression of ATM suggests, however, that although ATM is only required for normal expression of SOD2, ATM may also participate in the regulation of expression of SOD1, catalase, and GPX-1 antioxidant enzymes.
These combined findings demonstrate that Foxo3 is required for the regulation of ATM expression. In addition, these findings strongly suggest that both ATM-mediated and independent pathways regulate oxidative stress in Foxo3-null primitive hematopoietic stem cells.

Foxo3 Is a Physiological Regulator of Oxidative Stress in
Hematopoietic Stem Cells-Oxidative stress is broadly defined as an imbalance between the production of oxygen radicals and the ability to detoxify ROS to prevent oxygen radical-induced injuries. Although the deleterious impact of ROS accumulation on HSC has begun to be explored, little is known about the source of ROS and signaling pathways governing the regulation of ROS in hematopoietic stem cells.
Our key finding is that Foxo3 is a physiological regulator of oxidative stress that is essential for the regulation of hematopoietic stem cell fate. Foxo3 is required for the expression of ATM and antioxidant enzymes. By regulating expression of ATM tumor suppressor and antioxidant enzymes, Foxo3 maintains ROS below a harmful threshold.
Foxo3 Is Essential for Normal ATM Expression-FoxO functions parallel those associated with p53 tumor suppression, including induction of apoptosis, cell cycle arrest, and antioxidative stress function. In addition, Foxo3 and p53 share a growing list of common targets such as Fas ligand, PUMA, SOD2, Gadd45, and p21 CIP1/WAF1/Sdi1 (as in our studies) and undergo similar phosphorylation, acetylation, and ubiquitination that regulate their activities, albeit in a nonsimilar fashion (reviewed in Ref. 41). Cross-talk between Foxo3 and p53 has also been observed in model culture systems (42,43). Here we have shown that in Foxo3-null hematopoietic stem cells, p53 antioxidant response that is acti-  vated and in contrast to the studies by Miyamoto et al. (22), we observed a robust p53-mediated up-regulation of p21 CIP1/WAF1/Sdi1 . Activation of p53/p21 CIP1/WAF1/Sdi1 may shield Foxo3-deficient hematopoietic stem cells from damages that may cause clonal expansion of this population. Activation of p53 in response to oxidative stress in Foxo3null hematopoietic stem cells may suggest that p53 and Foxo3 play complementary functions counterbalancing each other as gatekeepers in hematopoietic stem cells.
ATM serine/threonine protein kinase is a critical enzyme in the regulation of stress response to DNA damage. ATM activation enhances p53 activity through several pathways, including direct phosphorylation of p53. ATM serine/threonine protein kinase is activated in response to stress, especially stress-induced by double strand DNA break that may also be ROS-induced (6,44). Activation of ATM involves its kinase domain and is not known to affect its level of expression (6). Activated ATM in turn results in accumulation and activation of p53 tumor suppressor.
Thus, down-regulation of ATM and ATM phosphoserine 1981 in Foxo3 Ϫ/Ϫ hematopoietic stem cells was not anticipated (Fig. 6) and clearly is not mediated by ROS (Fig. 6C). The over 2-fold reduction of ATM expression in Foxo3-null hematopoietic stem cells is of significance because ATM ϩ/Ϫ mice display a compromised stress response (36). Our results demonstrate that Foxo3 is essential for the regulation of ATM expression (Figs. 6 -8) as shown by the following: (a) ATM transcript and protein are highly and specifically down-regulated in Foxo3null hematopoietic stem cells (Fig. 6); (b) ATM activity (ATM phosphoserine 1981) is also down-regulated; (c) down-regulation of ATM expression in Foxo3-null hematopoietic stem cells is ROS-independent; (d) overexpression of Foxo3 rescues the expression of ATM in Foxo3-deficient primitive hematopoietic cells (Fig. 7) and enhances ATM expression in NIH 3T3 cells (Fig. 8A); and (e) shRNA silencing of Foxo3 in NIH 3T3 cells results in significant down-regulation of ATM transcript and protein (Fig. 8, B-D).
Furthermore, our results show that the relative ATM deficiency participates in accumulation of ROS in Foxo3-null LSK cells because of the following: (a) overexpression of ATM rescues expression of antioxidant enzymes in Foxo3-deficient primitive hematopoietic cells (Fig. 9C) as does ectopic expression of Foxo3 in these cells (Fig. 7C); (b) ATM like Foxo3 is required for full SOD2 expression (Fig. 10); and (c) ATM and Foxo3 enhance the expression of several antioxidant genes independently (Figs. 7, 9, and 10).
These data suggest that in Foxo3-deficient hematopoietic stem cells, multiple ROS-mediated signaling pathways lead to the up-regulation and activation of p53.
Although ROS accumulation mediated the activation of p53/ p21 CIP1/WAF1/Sdi1 in both Foxo3-null hematopoietic stem cells (Fig. 6) and in a subset of erythroid precursor cells (12), its impact on cell cycling of these populations was distinct. Loss of Foxo3 led to relative loss of quiescence and a mitotic delay (altered G 2 /M phase) in hematopoietic stem cells as compared with a G 1 arrest in erythroid precursor cells (that are about to exit the cell cycle and differentiate) resulting in reduced erythroid cell differentiation (12). These findings suggest that Foxo3 plays distinct functions in different cellular contexts (hematopoietic stem cell as compared with erythroid cell) and may itself be regulated differently in these cells.
We found that Foxo3-null LSK cells display a delay in the G 2 /M phase of cell cycle (Fig. 4, D and F) consistent with the notion that FOXO3a activity is required in G 2 /M for cell cycle completion (31). These findings may result from loss of Foxo3 directly, from accumulation of ROS, or from reduced expression and/or activity of ATM in Foxo3-null LSK cells. Induced nuclear localization of active FOXO3a fused to the ligand-binding domain of estrogen receptor in Rat-1 fibroblasts also resulted in a delay in the G 2 /M phase of cell cycle (45) further supporting that the function of FoxO may vary with cellular context and conditions.
Foxo3 Is an Essential Regulator of Hematopoietic Stem Cell Fate-We showed that Foxo3 is the principal active FoxO (Foxo4 expression is barely detectable and Foxo1 is mostly cytoplasmic and therefore presumably functionally mostly repressed) in hematopoietic stem cells (Fig. 1). Our studies determined that loss of Foxo3 alone results in severe reduction of expression of FoxO direct targets in hematopoietic stem cells such as cell cycle and antioxidant enzyme genes (Figs. 3C and 5), demonstrating the essential function of Foxo3 in the regulation of these genes in hematopoietic stem cells. In addition, we found substantial accumulation of ROS in Foxo3-null hematopoietic stem cells (Fig. 3) similar to what was observed in FoxO-null hematopoietic stem cells (11). We further demonstrated that this ROS accumulation mediates the loss of hematopoietic stem cell numbers and function in young Foxo3-null animals (Fig. 3). Moreover, in contrast to Tothova et al. (11) and as anticipated, we did not find up-regulation of expression of direct targets of FoxO such as p27 KIP1 and cyclin G2 in response to NAC treatment (Fig. 5). Whether the differences in gene regulation in response to antioxidant treatment indicate distinct regulation of FoxO direct targets in Foxo3-null as compared with FoxO-null hematopoietic stem cells require further investigation.
Together these findings strongly argue that Foxo3 (and not other FoxO) is the principal regulator of oxidative stress and cell cycle in hematopoietic stem cells. In addition, our studies demonstrate that the function of Foxo3 in the modulation of oxidative stress is essential for determining the fate of hematopoietic stem cells. These findings have significant consequences beyond stem cells of hematopoietic origin as they suggest potential important function for Foxo3 in other stem cell systems such as oocytes, where loss of Foxo3 leads to their depletion (23), and likely Foxo3 regulation of oxidative stress in nonhematopoietic stem cells.
The discrepancy between our study and studies reported by Tothova et al. (11) may be due to the timing of experiments relative to the excision of the knock-out alleles (4 weeks after excision) in the work of Tothova et al. (11). Because we observe a clear phenotype only after 8 -9 weeks of age, we believe it may take this long for loss of Foxo3 (and accumulation of ROS in Foxo3-null HSC) to make a deleterious impact on hematopoietic stem cell function. Analysis of conditional Foxo3 Ϫ/Ϫ mice in the studies of Tothova et al. (11) only 4 weeks after excision may not reveal the effect of loss of Foxo3 in hematopoietic stem cells. However, differences in the mouse strain (mixed 129ϫFVB versus mixed 129ϫC57BL6) may explain some of the divergence in our findings of Foxo3-null hematopoietic stem cell phenotype as compared with Miyamoto et al. (22). In addition to the findings by Miyamoto et al. (22), we found substantial ROS accumulation in Foxo3-null hematopoietic stem cells (Fig. 3, A and B) derived from young mice. We demonstrated that ROS mediates the repressed hematopoietic stem cell activity in Foxo3-null mice (Fig. 3D). We further identified several tumor suppressor pathways that are highly altered in Foxo3null hematopoietic stem cells via ROS-dependent and -inde-pendent mechanisms and lead to G 0 exit and G 2 /M arrest in Foxo3-null hematopoietic stem cells (Figs. 5 and 6).
Foxo3-deficient hematopoietic stem cells exhibit a phenotype less pronounced but similar to hematopoietic stem cells with deficiency in phosphatase and tensin homology that is a negative regulator of PI 3-kinase/AKT (46,47). Foxo3-and FoxO-deficient hematopoietic stem cells display distinct cell cycle and apoptosis phenotypes (Fig. 4) (11,22). It is noteworthy that although we cannot rule out additional loss through apoptosis, we did not detect significant apoptosis in freshly isolated or short term cultured Foxo3-deficient hematopoietic stem cells (supplemental Fig. 1). These combined findings indicate that relative levels of ROS and FoxO activity have important impacts on hematopoietic stem cell fate. Thus, tight regulation of phosphatase and tensin homology/PI 3-kinase/Fox signaling pathway is essential for preserving hematopoietic stem cell pool and homeostasis.
Our studies suggest a model in which Foxo3 together with ATM, p53, and p16 INK4a provide a strong tumor suppressor network in protecting stem cells from damage that may undermine their genomic stability and result in their clonal expansion (model Fig. 11). Sustained activation of these pathways may alternatively result in senescence leading to the defects seen in Foxo3-null hematopoietic stem cells derived from aged mice (22). Future elucidation of the mechanisms of regulation of Foxo3 in hematopoietic stem cells, in particular potential interactions of Foxo3 with coactivators and corepressors, as well as identification of Foxo3 targets in stem cells, will be of fundamental importance for better understanding of aging of hematopoietic stem cells and malignant transformation in cancer stem cells.
Note Added in Proof-While this manuscript was under review, a new report showing that Foxo3 directly interacts with ATM and regulates its activity in response to DNA damage was published (51). These findings further the link between Foxo3 and ATM.