miR-9 upregulation leads to inhibition of erythropoiesis by repressing FoxO3

MicroRNAs (miRNAs) are emerging as critical regulators of normal and malignant hematopoiesis. In previous studies of acute myeloid leukemia miR-9 overexpression was commonly observed. Here, we show that ectopic expression of miR-9 in vitro and in vivo significantly blocks differentiation of erythroid progenitor cells with an increase in reactive oxygen species (ROS) production. Consistent with this observation, ROS scavenging enzymes, including superoxide dismutase (Sod2), Catalase (Cat), and glutathine peroxidase (Gpx1), are down-regulated by miR-9. In addition, miR-9 suppresses expression of the erythroid transcriptional regulator FoxO3, and its down-stream targets Btg1 and Cited 2 in erythroid progenitor cells, while expression of a constitutively active form of FoxO3 (FoxO3-3A) reverses miR-9-induced suppression of erythroid differentiation, and inhibits miR-9-induced ROS production. Thus, our findings indicate that aberrant expression of miR-9 blocks erythropoiesis by deregulating FoxO3-mediated pathways, which may contribute to the ineffective erythropoiesis observed in patients with hematological malignancies.

Over the last decade it has become apparent that non-coding RNAs function as regulators of crucial cellular processes and play a critical role in human diseases 1 . MicroRNAs are a class of evolutionally conserved, single-stranded, small (approximately 19−23 nucleotides) non-coding RNAs 2,3 . Mature miRNAs regulate gene expression by inducing targeted mRNA degradation, translational inhibition or through sequestration of targeted mRNA from the translational machinery 4 . Studies indicate that microRNAs are key regulators of normal hematopoiesis 5 and deregulation of microRNAs has been implicated in the pathogenesis of hematological malignancies 6 . MiR-9 regulates normal neurogenesis during brain development [7][8][9] . Notably, miR-9 acts as a tumor suppressor and oncogene in a cell-context dependent manner 10 . The upregulation of miR9 expression has been noted in various hematological malignancies, including Hodgkin's lymphoma 11 , AML [12][13][14] , ALL 15 and MLL-AF9-induced leukemia 16 while it also appears to be down-regulated in some cases of ALL 17 and in pediatric AML with t(8;21) 18 . Anemia occurs frequently in patients with hematologic malignancies, such as acute leukemia and myelodysplastic syndromes, and is caused by a variety of mechanisms 19 . The role of miR-9 in erythropoiesis has not been reported previously.
In this study, we find that forced expression of miR-9 inhibits erythroid progenitor differentiation both in vitro and in vivo, FoxO3 is a critical regulator of erythropoiesis and is required for regulation of oxidative stress in erythroid progenitor cells 20,21 . We observe that ectopic expression of miR-9 induces reactive oxygen species (ROS) production in erythroid progenitor cells and inhibits their differentiation through regulation of FoxO3 expression. Our findings suggest that a low level of miR-9 expression in erythroid progenitor cells is critical for maintenance of normal erythropoiesis. may interfere with erythroid differentiation. We first assessed the consequences of miR-9 overexpression on differentiation of G1ER cells. The G1ER cell line, derived from GATA-1 knockout murine embryonic stem cells, is widely used for analysis of erythroid differentiation in vitro. In this cell line, the coding region of GATA-1 is fused to the ligand-binding domain of the human estrogen receptor (ER). Exposure of G1ER cells to β-estradiol causes rapid activation of GATA-1 expression, inducing erythroid differentiation 22,23 . The miR-9 precursor was subcloned into the retroviral vector Migr1, which expresses EGFP. Expression of miR-9 in G1ER cells was confirmed by qRT-PCR (Fig. 1A). EGFP+ cells transduced by Migr1 or Migr1-miR-9 were sorted, and then differentiation was induced by treatment with β-estradiol for 48 hours. Hemoglobin expression, determined by benzidine staining, is a hallmark of terminal erythroid differentiation. As shown in Fig. 1B,C, the percentage of benzidine positive cells was significantly lower in miR-9-transduced G1ER cells than in Migr1-transduced G1ER cells (20% in miR-9-transduced vs. 40% in Migr1-transduced), suggesting that ectopic expression of miR-9 inhibits erythroid differentiation of G1ER cells.
As the terminal differentiation in vitro of erythroblasts derived from fetal liver closely mimics erythroid cell differentiation in vivo 24 , we next examined whether ectopic expression of miR-9 affects the erythroid differentiation of fetal liver erythroblasts. In fetal liver cells, five distinct populations of cells, corresponding to erythroblasts at different developmental stages, R1 (CD71 med Ter119 low ), R2 (CD71 high Ter119 low ), R3 (CD71 high Ter119 high ), R4 (CD71 med Ter119 high ), and R5 (CD71 low Ter119 high ), can be identified by their expression of characteristic cell surface markers. R1 cells are the least differentiated, while the R2-R5 populations are progressively more differentiated 24 . Ter119 − fetal liver cells were isolated from E14.5 embryos and infected with retrovirus expressing either Migr1 or Migr1-MiR-9. We observed that retroviral transduction of fetal liver cells with Migr1-MiR-9 results in a 2-fold increase in miR-9 expression ( Fig. 2A). We then performed flow cytometric analysis on gated GFP-positive cells. One day after retrovirus infection, the majority of erythroblasts were in the early development stages R1, R2, and R3 (left panels, Fig. 2B,C). The proportion of R1 and R2 populations in the Migr1-miR-9 transduced erythroblasts was significantly higher than that observed in the Migr1-transduced erythroblasts; whereas, the proportions of R3 and R4 populations were significantly reduced in the Migr1-miR-9 transduced erythroblasts compared to Migr1-transduced erythroblasts (left panel, Fig. 2B,C). Erythroblasts were induced to differentiate at day 2 after infection. Three days after induction of differentiation (day 4), the majority of erythroblasts were differentiated (comprising R3, R4, and R5 populations). Consistently, the proportion of cells in the R1, R2, and R3 populations were higher in Migr1-miR-9 transduced erythroblasts than was observed in the Migr1-transduced erythroblasts; whereas, the proportions of cells in the R4 and R5 populations were significantly lower in Migr1-miR-9 transduced erythroblasts than in the Migr1-transduced erythroblasts (Right panel, Fig. 2B,D). In summary, these studies suggest that miR-9 overexpression inhibits differentiation of fetal liver-derived erythroblasts in vitro.
In the erythroid lineage, burst-forming unit-erythroid (BFU-E) is recognized as the earliest committed progenitor, which further differentiates into colony-forming unit-erythroid (CFU-E). Upon stimulation by erythropoietin (Epo), CFU-E progenitor cells develop into erythoblasts 25,26 . To further evaluate the effects of miR-9 overexpression on the erythroid differentiation potential of fetal liver cells, we performed BFU-E and CFU-E assays using EGFP + sorted fetal liver cells that had been transduced with either Migr1 or the Migr1-miR-9 vector. Notably, miR-9-transduced fetal liver cells gave rise to 2-3-fold less BFU-E and CFU-E colonies as compared to Migr1-transduced fetal liver cells (Fig. 3A,B). The miR-9-transduced BFU colonies also appeared much smaller than Migr1-transduced BFU colonies (Fig. 3C) and there was a significant reduction in the number of mature erythroid cells, as determined by benzidine staining (Fig. 3D). To determine whether the reduced BFU and CFU colony-forming ability of miR-9-transduced fetal liver cells is a consequence of increased cell death or apoptosis of these cells, we analyzed the status of cell death and apoptosis in both vector or miR-9-transduced fetal liver cells. Surprisingly, forced expression of miR-9 reduced the frequency of cell death (DAPI + ) of fetal liver cells, but had a marginal effect on early apoptosis (Annexin V + , DAPI − ) ( Fig. 3E-F). Together, these results suggest that miR-9 overexpression inhibits differentiation of erythroid progenitors.

miR-9 blocks erythroid differentiation in vivo.
To evaluate whether ectopic expression of miR-9 also affects erythroid differentiation in vivo, we generated chimeric mice reconstituted with BM cells expressing miR-9/EGFP or EGFP. Lin -BM cells isolated from C57B/L6 mice were infected with Migr1 or Migr1-miR-9 retrovirus, followed by transplantation into lethally-irradiated wildtype recipient mice. At 7 weeks post-transplantation, erythroid differentiation was examined using flow cytometric analysis of EGFP + BM cells for CD71 and Ter119 cell-surface expression. In BM cells, erythroblasts can be separated into R1 (Ter119 − CD71 + ), R2 (Ter119 + CD71 + ), R3 (Ter119 + CD71 med ), and R4 (Ter119 + CD71 − ) populations, corresponding to proerythroblasts, basophilic erythroblasts, late basophilic and polychromatophilic erythroblasts, and orthochromatophilic erythroblasts, respectively. As shown in Fig. 4, the frequency of R1 and R2 erythroblasts was significantly higher in miR-9-EGFP + transduced BM cells than it was in EGFP + control transduced BM cells. In contrast, when compared to mice transplanted with control Migr1 cells, mice transplanted with miR-9-EGFP+ cells had a significantly lower frequency of late stage erythroblasts (R4 population, Ter119 + , CD71 + ). As observed in vitro, these data indicate that miR9 overexpression blocks erythroid differentiation in vivo. MiR-9 regulates ROS production in erythroid progenitors. Oxidative stress plays a critical role in the maturation of erythroid progenitors 27 . We next examined ROS production in fetal liver erythroid progenitors and G1ER cells expressing either control vector or miR-9. In fetal liver cells, miR-9-transduced erythroid progenitors had a two-fold increase in ROS levels compared with control cells. However, incubation with the ROS scavenger N-acetyl-L-cysteine (NAC) inhibited miR-9-induced ROS production (Fig. 5A,B). Similarly, in G1ER cells, forced expression of miR-9 also significantly increases ROS production, which is inhibited by NAC (Fig. 5D,E). Scavenging enzymes have an important role in controlling cellular ROS levels. To determine whether ROS scavenging enzymes are reduced in miR-9-transduced erythroid progenitors, we performed qRT-PCR analysis of these genes in both vector and miR-9-transduced erythroid progenitors and G1ER cells. As shown in Fig. 5C,F, expression of superoxide dismutase (Sod2), catalase (Cat) and glutathine peroxidase (Gpx1) enzymes

miR-9 down-regulates FoxO3 in erythroid progenitors and co-expression of miR-9 and FoxO3 reverses miR-9 induced inhibition of erythroid differentiation. Previously we demonstrated that
miR-9 down-regulates FoxO3 expression by directly binding the 3′ UTR region of FoxO3 in BM cells 10 . To determine whether miR-9 inhibits FoxO3 expression in erythroid progenitors, we examined FoxO3 expression in miR-9-and vector-transduced fetal liver erythroid progenitors and G1ER cells. As determined by both Western Blot and qRT-PCR analysis, forced expression of miR-9 significantly inhibits FoxO3 expression in both cell types (Fig. 6A, Supplementary Fig. 1). Concomitantly, Btg1 and Cited 2, known down-stream targets of FoxO3, were also markedly down-regulated in miR-9-transduced erythroid progenitors and G1ER cells as compared to vector-transduced cells (Fig. 6A). As FoxO3 is a known regulator of erythroid differentiation 21,28 , we hypothesized that the effects of miR9 on erythroid differentiation may be at least partially mediated by FoxO3. We therefore examined the induction of erythroid differentiation in fetal liver cells expressing either control vector, miR-9, a constitutively active form of FoxO3 (FoxO3-3A), or miR9 and FoxO3-3A. In line with a previous report 28 , forced expression of FoxO3-3A reduced the frequency of R2 and R3 erythroid blasts and increased the frequency of R5 (late) erythroid blasts. Notably, erythroid progenitors co-expressing miR-9 and FoxO3-3A and control vector-transduced erythroid progenitors had comparable frequencies of erythroid blasts at both early and late stages (Fig. 6B,C), suggesting that constitutive expression of FoxO3 reverses the inhibitory effects of miR-9 overexpression on erythroid progenitor differentiation.
FoxO3-3A inhibits miR-9-induced ROS production. As FoxO3 regulates oxidative stress in erythropoiesis 28 , we determined whether FoxO3-3A suppressed the increase in ROS production induced by miR-9 overexpression. Of note, ROS levels were comparable in erythroid progenitors expressing either control vector or both miR-9 and FoxO3-3A vectors (Fig. 6D). Thus, it appears that the miR-9-induced increase in ROS levels is mediated by FoxO3 suppression.

Discussion
Here, we find that ectopic expression of miR-9 inhibits erythroid differentiation of G1ER cells and erythroid progenitor cells derived from murine fetal liver. Consistent with these results, our studies show that forced expression of miR-9 through retroviral transduction of murine BM cells also significantly inhibits erythroid progenitor differentiation in vivo. In a previous study it was shown that miR-9 directly binds the 3′ UTR region of the FoxO3 gene, suppressing its expression in myeloid progenitor cells 10 . Here we demonstrate that miR-9 also downregulates FoxO3 mRNA and protein levels in erythroid progenitor cells, and that expression of FoxO3-3A, a constitutively active form of FoxO3, reverses miR-9-induced inhibition of erythroid differentiation. Thus, we provide the first evidence that miR-9 regulates erythropoiesis via modulation of the FoxO3 pathway.
Importantly, miR-9 has a previously unrecognized role in the regulation of ROS production during erythropoiesis. While ROS are normal byproducts of cell metabolism, excessive ROS production leads to oxidative stress that can damage biomolecules, and it is associated with hematopoietic malignancies 29 . Tight regulation of oxidative stress is critical for erythropoiesis and especially during maturation of erythrocytes as they accumulate hemoglobin 27,30 . Increased miR-9 expression upregulates ROS production in erythroid progenitor cells, likely through downregulation of ROS scavenging enzymes such as Sod2, Cat and Gpx1. FoxO3 is an important regulator of oxidative stress during erythropoiesis, and loss of Foxo3 inhibits maturation of erythroid progenitors exposed to excessive oxidative stress 20,21 . Notably, we found that the constitutive mutant of FoxO3 (FoxO3-3A) reverses miR-9-induced ROS production. Thus, it appears that miR-9 induces ROS production by dysregulating FoxO3-mediated mitigation of ROS stress.
miRNAs are known to have multiple downstream targets 31 . The function of miR-9 is cell-context dependent, and likely mediated by a number of downstream pathways. For instance, miR-9 regulates the development and function of neural progenitors by targeting genes including TLX 32 , FoxG1 33 and Hes1 8,34 while it induces metastasis of breast cancer cells by targeting LIFR and E-cadherin 35,36 . Although we demonstrate that FoxO3 mediates miR-9 function in erythroid progenitor cells, our studies do not rule out the possibility that alternate miR-9 downstream targets mediate other functions for miR-9 during erythropoiesis.

Erythroid differentiation in vitro. Total fetal cells were labeled with biotin-conjugated anti-Ter119
antibody, and the Ter119-cells were isolated by Dynabeads as per the manufacturer's instructions (Dynabeads Biotin Binder, Invitrogen), followed by retroviral infection. Two days after infection, the cells were cultured in erythroid-culture differentiation medium (StemPro-34, Invitrogen) containing 2 mM L-glutamine, 2% penicillin/ streptomycin, 10 −4▯ M ▯β-mercaptoethanol, 0.5 mg/ml transferrin, 40 ng/ml IGF and 10 U/ml Epo. BM Transplantation. The donor BM cells were collected from C57BL/6 mice injected with 5-Fluorouracil (5-FU) intraperitoneally 5 days before collection. The recipient mice received a lethal dose of irradiation (10 Gy). After 12 hours, the mice were then administered 0.5-1.0 × 10 6 retrovirally-infected BM cells by retroorbital injection. All of the animal studies were performed in accordance with the guidelines of the Animal Care Committee of the University of Illinois at Chicago and all procedures were approved by the Animal Care Committee of the University of Illinois at Chicago. Flow Cytometry Assays. Single-cell suspensions from BM were stained with the indicated fluorochrome-conjugated antibodies (eBioscience, USA). For the detection of apoptosis, fetal liver cells were stained with Anti-Annexin V antibody (eBioscience, USA) and DAPI. For the detection of cell cycle, fetal liver or G1ER cells were labeled with BrdU overnight, and stained with anti-BrdU-APC anibody (eBioscience, USA) and DAPI. For the ROS detection, fetal liver or G1ER cells were stained with hydroethidine (10 mM)(Molecular probe) following the manufacturer's instructions. Flow cytometry was performed at the University of Illinois at Chicago core facility using CyAn flow cytometer. All data were analyzed by FlowJo software (TreeStar).
Colony-Forming Assays. E14.5 D fetal liver cells were spinfected with the MSCV-puromycin retrovirus and the cells were selected by 1 μg/ml puromycin for 3 days. To detect CFU-E colonies, 1 × 10 4 selected E14.5 fetal liver cells were plated in duplicate in semisolid medium (The Mouse Methylcellulose Complete Media; R&D systems) containing 4 U/ml erythropoietin according to the manufacturer's protocol. To detect BFU-E colonies, 0.1 × 10 4 selected E14.5 fetal liver cells were plated in duplicate in methylcellulose medium containing 4 U/ml erythropoietin and 100 ng/ml SCF. The CFU-E and BFU-E colonies were counted at 3 days or 14 days after plating respectively.
Western Blotting. The Western Blot was performed as described previously 38 . Rabbit FoxO3 antibody and Tubulin antibodies were obtained from Cell signaling technology and Millipore corporation respectively. Benzidine staining. Erythroid differentiation of G1ER and fetal liver cells was also evaluated by hemoglobin content. The concentration of hemoglobin was estimated by the standard methodology using benzidine staining. Staining solution was freshly prepared by adding 20 μl of 33% hydrogen peroxide to 1 ml of freshly prepared solution of 2% benzidine in 3% acetic acid (w/v). The solution was diluted 1:1 with the cell suspension. After 3 mins, the percentage of benzidine-positive cells (blue in color) was assessed under a light microscope. The percentage of positive cells was counted from a total of 200 cells. Each experiment was performed in triplicate. Statistical analysis. Statistical significance was calculated using the two tailed Student's t test.