Advertisement

Oncogenic transcription factor Evi1 regulates hematopoietic stem cell proliferation through GATA‐2 expression

Hiromi Yuasa, Yuichi Oike, Atsushi Iwama, Ichiro Nishikata, Daisuke Sugiyama, Archibald Perkins, Michael L Mucenski, Toshio Suda, Kazuhiro Morishita

Author Affiliations

  1. Hiromi Yuasa1,
  2. Yuichi Oike1,
  3. Atsushi Iwama2,
  4. Ichiro Nishikata3,
  5. Daisuke Sugiyama4,
  6. Archibald Perkins5,
  7. Michael L Mucenski6,
  8. Toshio Suda*,1 and
  9. Kazuhiro Morishita*,3
  1. 1 Department of Cell Differentiation, The Sakaguchi Laboratory, School of Medicine, Keio University, Tokyo, Japan
  2. 2 Department of Cellular and Molecular Biology, Graduate School of Medicine, Chiba University, Chiba, Japan
  3. 3 Department of Biochemistry, Medical College, University of Miyazaki, Miyazaki, Japan
  4. 4 Division of Cellular Therapy, Advanced Clinical Research Center, Institute of Medical Science, University of Tokyo, Tokyo, Japan
  5. 5 Department of Pathology, Yale University School of Medicine, New Haven, CT, USA
  6. 6 Division of Pulmonary Biology, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, USA
  1. *Corresponding authors: Department of Cell Differentiation, The Sakaguchi Laboratory, School of Medicine, Keio University, Tokyo, Japan. Tel.: +81 3 5363 3475; Fax: +81 3 5363 3474; E‐mail: oike{at}sc.itc.keio.ac.jp or E‐mail: sudato{at}sc.itc.keio.ac.jpDepartment of Biochemistry, Faculty of Medicine, Miyazaki Medical College, University of Miyazaki, 5200 Kihara, Miyazaki 889‐1692, Japan. Tel.: +81 3 5363 3473; Fax: +81 3 5363 3474; E‐mail: kmorishi{at}med.miyazaki-u.ac.jp
  1. These authors contributed equally to this work

Abstract

The ecotropic viral integration site‐1 (Evi1) is an oncogenic transcription factor in murine and human myeloid leukemia. We herein show that Evi1 is predominantly expressed in hematopoietic stem cells (HSCs) in embryos and adult bone marrows, suggesting a physiological role of Evi1 in HSCs. We therefore investigate the role and authentic target genes of Evi1 in hematopoiesis using Evi1−/− mice, which die at embryonic day 10.5. HSCs in Evi1−/− embryos are markedly decreased in numbers in vivo with defective self‐renewing proliferation and repopulating capacity. Notably, expression rate of GATA‐2 mRNA, which is essential for proliferation of definitive HSCs, is profoundly reduced in HSCs of Evi1−/− embryos. Restoration of the Evi1 or GATA‐2 expression in Evi1−/− HSCs could prevent the failure of in vitro maintenance and proliferation of HSC through upregulation of GATA‐2 expression. An analysis of the GATA‐2 promoter region revealed that Evi1 directly binds to GATA‐2 promoter as an enhancer. Our results reveal that GATA‐2 is presumably one of critical targets for Evi1 and that transcription factors regulate the HSC pool hierarchically.

Introduction

The ecotropic viral integration site‐1 (Evi1) was first identified as a common site of retroviral integration in murine myeloid tumors (Morishita et al, 1988; Mucenski et al, 1988). Rearrangements involving human Evi1 on chromosome 3q26 often activate its expression in human acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) (Morishita et al, 1992b; Mitani et al, 1994). Evi1 is a member of the SET/PR domain family of zinc‐finger transcription factors and it contains two separated zinc‐finger DNA‐binding domains, which recognize different DNA sequences (Matsugi et al, 1990; Morishita et al, 1995). However, the authentic target genes of Evi1 have not yet been identified.

The expression of Evi1 in adult mice is restricted to the kidney and ovary (Morishita et al, 1990), whereas Evi1 is expressed abundantly in several embryonic tissues in the developing mouse (Perkins et al, 1991), suggesting that Evi1 plays a role in normal development. In fact, mice lacking Evi1 tend to exhibit widespread hypocellularity, hemorrhaging and a disruption of the paraxial mesenchyme and, as a result, they die at approximately 10.5 days postcoitus (d.p.c.; E10.5). In addition, homozygous embryos show defects in the heart, somites, cranial ganglia and the peripheral nervous system (Hoyt et al, 1997). These findings indicate that Evi1 plays an important role in cell proliferation and/or differentiation in a variety of cell types. Moreover, Evi1 has been shown to be expressed in bone marrow hematopoietic stem cells (HSCs), implying that Evi1 plays a role in the regulation of HSCs (Phillips et al, 2000; Park et al, 2002; Shimizu et al, 2002). On the other hand, no detailed expression profiles of Evi1 or its physiological roles in primary hematopoietic cells have yet been published. As a result, we attempted to clarify the role of Evi1 in the development of HSCs during embryonic hematopoiesis.

In this study, we performed a detailed analysis of hematopoiesis in mice lacking Evi1 and found that the development and expansion of definitive HSCs is severely impaired in homozygous (Evi1−/−) embryos. This hematological phenotype is reminiscent of that observed in GATA‐2‐deficient embryos. The expression of Evi1 or GATA‐2 in HSCs from Evi1−/− embryos could prevent the failure of in vitro development and the proliferation of HSCs through an upregulation of GATA‐2 expression. Since the GATA‐2 mRNA levels in Evi1−/− embryos decreased markedly, we localized the cis‐element to the GATA‐2 promoter region through which Evi1 directly regulates GATA‐2 transcription in vitro. Our data provide the first evidence demonstrating that Evi1 plays a physiological role in normal hematopoiesis, while also identifying Evi1 as a transcriptional regulator epistatic to GATA‐2, which controls both HSC maintenance and proliferation.

Results

Evi1 expression in embryonic and adult hematopoietic stem cells

We first examined the expression pattern of Evi1 mRNA in various hematopoietic lineage cells. Quantitative RT–PCR revealed that Evi1 mRNA is predominantly expressed in both the embryonic and adult HSC fractions, that is, CD34+CD45+c‐Kit+ cells from mouse embryos and c‐Kit+Sca‐1+Lin cells from adult mouse bone marrow, respectively (Figure 1A). In addition, we found an abundant Evi1 expression in both the yolk sac and para‐aortic splanchno‐pleural (P‐Sp) region, where primitive (embryo‐type) and definitive (adult‐type) hematopoiesis occurs, respectively (Godin et al, 1995). These findings led to a detailed analysis aimed at determining the role of Evi1 in normal hematopoiesis by analyzing Evi1 homozygous mutant mice.

Figure 1.

Expression pattern of Evi1 in hematopoietic cells and gross appearance of E9.5 Evi1−/− embryos. (A) Quantitative RT–PCR analysis of Evi1 mRNA in hematopoietic cells from mouse embryos and adult bone marrow. The embryonic cells analyzed are CD45Ter119+ primitive erythrocytes, CD45+Ter119 hematopoietic cells other than primitive erythrocytes, CD45+c‐KitCD34 mature hematopoietic cells, a CD45+c‐Kit+CD34+ HSC‐enriched population and para‐aortic splanchnopleura (P‐Sp). The bone marrow cells analyzed are CD4+/CD8+, T cells; B220+, B cells; CD11b+, macrophages and monocytes; Gr‐1+, granulocytes; Ter119+, erythrocytes; CD41+, megakaryocytes; CD45+Lin+, mature hematopoietic cells; CD45+Lin, immature hematopoietic cells; c‐Kit+Sca‐1+Lin, HSC‐enriched population. A mixture of anti‐Mac‐1, ‐Gr‐1, ‐B220, ‐CD4, ‐CD8 and ‐Ly‐6 antibodies was used as a lineage marker (Lin). Evi1 mRNA was assayed by real‐time PCR (Materials and methods). The data were normalized to GAPDH mRNA and calibrated to the Evi1/GAPDH ratio (ΔCT) in CD45Ter119+ cells in mouse embryos and CD45+Lin+ in adult bone marrow. The data are the mean and standard deviation of 2ΔΔCt in triplicate assays. (B) (a) The yolk sac from Evi1−/− embryos shows severe anemia and defective large vessel development, while large vessels developed and were organized in the yolk sac of Evi1+/ (b) and wild‐type embryos at E9.5 (c). (d–f) Gross appearance of Evi1−/− (d), Evi1+/−(e) and wild‐type (d) embryos at E9.5. The Evi1−/− embryo is smaller and much paler than the wild‐type littermate, while also showing pericardial effusion and hemorrhaging.

In this study, we analyzed 126 wild‐type, 264 heterozygous (Evi1+/−) and 124 homozygous (Evi1−/−) embryos at E9.5, which together demonstrated a proportion of 1:2:1 in keeping with Mendel's first law. By E11.0, all homozygous Evi1−/− embryos were dead, but living embryos could still be recovered at E9.5. The yolk sac at E9.5 of Evi1−/− embryos was paler than those of wild‐type and heterozygous Evi1+/− embryos (Figure 1Ba–c). Although the total number of blood cells from the yolk sac decreased to 20% of that in the littermates in Evi1−/− embryos (Supplementary Figure 1A), ε and β hemoglobin staining revealed no maturational defect of primitive erythroid cells in the yolk sac (data not shown). In addition, yolk sac cells from Evi1−/− embryos showed no alteration in the erythroid and granulocyte–macrophage (GM) colony formation in comparison to the wild‐type counterparts (Supplementary Figure 1B). Since Evi1−/− embryos exhibited pericardial effusion with hemorrhaging (Figure 1Bd–f), the reduction in yolk sac erythrocytes is probably due to a loss of blood cells. At E9.5, the body of Evi1−/− embryos was smaller than that of the wild‐type littermates, but the ‘turning’ of the embryos with 21–29 pairs of somites had finished, suggesting that they had developed normally.

Decreased in vitro proliferation of hematopoietic stem cells in Evi1/ mutants

We next determined the capacity for definitive hematopoiesis, which occurs in the P‐Sp region at E9.5. Although the recovered number of cells from the P‐Sp region of Evi1+/+, Evi1+/− and Evi1−/− embryos was almost equal among the three lineages (Supplementary Table 1), the number of CD45+ hematopoietic cells dramatically decreased in the P‐Sp region from the Evi1−/− embryos, in comparison to that from Evi1+/− or wild‐type Evi1+/+ (Figure 2Aa). Accordingly, the P‐Sp region of Evi1−/− embryos contained significantly fewer CD34+c‐Kit+CD45+ HSCs than in the controls, although the percentages of HSCs in CD45+ cells showed almost the same levels (70–80%) for the Evi1+/+, Evi1+/− and Evi1−/− embryos (Figure 2Ab and c). These findings suggest that Evi1 possibly plays a role in the development and/or expansion of HSCs.

Figure 2.

Defects in the in vitro proliferation and differentiation of CD45+c‐Kit+CD34+ HSCs from E9.5 Evi1−/− embryos. (A) Incidence of HSCs in E9.5 embryos. The cells from the P‐Sp region of E9.5 embryos were stained with anti‐CD45, ‐CD34 and ‐c‐Kit mAbs and analyzed by flow cytometry. (a) CD45+ cells were gated (percentages of CD45+ cells are indicated in the panel). (b) CD45+ cells were examined for the expression of CD34 and c‐Kit (percentages of each fraction are indicated in the upper right quadrant). (c) The number of CD45+c‐Kit+CD34+ cells in P‐Sp from Evi1−/−, Evi1+/− and wild‐type embryos at E9.5 was determined. The columns represent the mean values±s.d. (n=5). (B) In vitro hematopoiesis in P‐Sp cultures. The cells from the P‐Sp cultures on OP9 cells were harvested at 7 days (a–c) and 9 days (d, e) in culture. (a) CD45+ cells that developed in culture were gated (the percentages of CD45+ cells are indicated in the panel). (b) CD45+ cells were examined for the expression of CD34 and c‐Kit (the percentages of each fraction are indicated in the upper right quadrant). (c) The numbers of CD45+c‐Kit+CD34+ cells in P‐Sp culture using Evi1−/− and wild‐type P‐Sp were determined. The columns represent the mean values±s.d. (n=5). (d, e) The number of progenitor cells that gave rise to erythroid (d) and GM colonies (e) was determined after 7 days of culture. (C) In vivo hematopoietic reconstitution by transplanted cells. One‐embryo‐equivalent cells from E9.5 P‐Sp were injected into a conditioned newborn recipient (open circle). Five‐embryo‐equivalent cells from Evi1/ P‐Sp were transplanted into a recipient mouse (filled circle). At 8–12 weeks after the transplantation, the proportion of the donor‐derived cell population in the recipient granulocytes was determined (%).

We next examined the capacity of HSCs from Evi1−/− embryos to proliferate using an in vitro P‐Sp explant/OP9 stromal cell coculture system (Takakura et al, 2000). After P‐Sp explants were cultured on OP9 cells for 7 days, fractions of hematopoietic cells were analyzed by fluorescence activated cell sorting (FACS). The majority of CD45+ hematopoietic cells recovered from wild‐type P‐Sp cultures exhibited the CD34+c‐Kit+CD45+ HSC phenotype (64.7% in Figure 2Bb). In contrast, the number of CD34+c‐Kit+CD45+ HSCs in the Evi1−/− P‐Sp cultures was significantly lower (25.6% in CD45+ cells) than that obtained in the controls, indicating a failure of expansion and maintenance of HSCs in Evi1−/− P‐Sp culture (Figure 2Ba–c). To examine the differentiation capacity of the Evi1‐deficient hematopoietic cells, hematopoietic cells derived from P‐Sp explants were subjected to in vitro colony assays in a semisolid agar culture (Iscove et al, 1974) supplemented with stem cell factor (SCF), interleukin 3 (IL‐3) and erythropoietin (Epo). Cells from the Evi1−/− P‐Sp culture from each P‐Sp region gave rise to 33.3 and 22.1% erythroid and GM colonies, respectively, relative to those from wild‐type embryos (Figure 2Bd and e). An analysis of the differentiation markers after P‐Sp explants on OP9 cells for 7 days demonstrated that the differentiation capacity of HSCs in Evi1−/− mice was not grossly impaired in terms of the erythroid and GM lineages (Supplementary Figure 2). These results again indicated the inefficient expansion of hematopoietic stem and progenitor cells in Evi1−/− P‐Sp cultures with a differentiation capacity to erythroid and myeloid cells.

Defective activities of hematopoietic stem cells in P‐Sp of the Evi1/ embryos

We next investigated long‐term repopulation capacity of HSCs in the Evi1−/− P‐Sp region. Using an irradiated adult mouse as a recipient, the definitive HSC activity was first detectable in the Aorta‐Gonad‐Mesonephros (AGM) region of the embryo at day 10.5 by the transplantation of donor test cells with competitor cells (Medvinsky and Dzierzak, 1996). Since the Evi1−/− embryo died on E10.5, we could not use the same strategy. Instead, we transplanted the donor hematopoietic cells from the E9.5 P‐Sp region into the livers of sublethally conditioned newborn mice, and this strategy has been reported to detect the HSC activity as early as on day 8 or 9 in both yolk sac and P‐Sp (Yoder et al, 1997).

One‐embryo‐equivalent cells from E9.5 P‐Sp of Evi1+/+, Evi1+/ or Evi1/ embryos (Ly5.2+) were transplanted into a busulfan‐conditioned newborn recipient (Ly5.1+). The average number of hematopoietic cells from each P‐Sp region was almost the same among the three lineages with over 90% viability (Supplementary Table 1). At 3 months post‐transplant, the donor‐derived Ly5.2+ cells could be detected in the peripheral blood of the recipients that received P‐Sp cells from the Evi1+/+ (2.56% on an average) and Evi1+/ (0.62% on an average) but not from the Evi1/ embryos (Figure 2C). Since the number of HSCs from Evi1−/− P‐Sp at E9.5 was one‐fifth of those from wild‐type P‐Sp (Figure 2A), we further combined and transplanted five‐embryo‐equivalent cells from P‐Sp of Evi1−/− into a busulfan‐conditioned newborn recipient. But no donor‐derived cells could be detected in the peripheral blood of the three recipients until 3 months (Figure 2C, closed circles). As demonstrated above, the differentiation capacity of hematopoietic cells from the Evi1−/− P‐Sp is not grossly impaired to make GM colonies on OP9 cells in vitro (Figure 2Be). However, no differentiated granulocytes derived from the Evi1−/− embryo were detected in the recipient mice. Since the GM colony‐forming cells are mainly committed macrophage progenitor cells, they do not show the in vivo reconstitution activity for any lineages, including granulocytes. Thus, Evi1 is essential for the long‐term repopulation capacity of HSCs, which is impaired in an Evi1 dose‐dependent manner in Evi1+/− and Evi1/ embryos. Correspondingly, definitive HSC in the yolk sac from Evi1/ embryos (E9.5) did not contribute to reconstitution at all in the conditioned newborn recipients (Supplementary Figure 3).

Defects of Evi1/ mutant embryos in vascular remodeling and angiogenesis

With respect to hematopoiesis and angiogenesis, we recently showed that HSCs play an important role in embryonic angiogenesis due to their secretion of an angiogenic factor, angiopoietin‐1 (Ang‐1) (Takakura et al, 2000). Because the number of HSCs significantly decreased in the Evi1−/− embryos, we next determined whether the vascular system of the Evi1−/− embryos was normally organized. The pattern of the vascular system of the Evi1−/− embryo was studied by whole mount staining of E9.5 embryos using the anti‐PECAM‐1 monoclonal antibody, which detects differentiated endothelial cells. In comparison to wild‐type embryos (Figure 3Aa and c), we found that whole Evi1−/− embryos failed to form an organized vascular network at E9.5 (Figure 3Ab and d). The vascular network of the Evi1−/− embryos was generally less complex than that of the wild‐type embryos (Figures 3Ac and d). In contrast, the homozygous mutants exhibited plexus‐like vascular networks and vessels in the brain periphery, which either did not develop or tended to degenerate, suggesting that gross defects in the vascular branching were observed, especially in the head. Because such embryos have already finished ‘turning’ at that stage, it is suggested that the abnormalities in the vascular system are possibly not caused by the growth retardation of Evi1−/− embryos. In particular, extensive remodeling, resulting in the formation of large and small vessels, occurred in the yolk sac of the wild‐type embryos (Figure 3Ae). However, although the honeycomb‐like vascular plexus appeared to form normally at E8.5 in Evi1−/− embryos, it remained largely unchanged across most of the surface of the yolk sac at E9.5 (Figure 3Af), suggesting that the remodeling of the vessels did not completely occur in the Evi1−/− embryo.

Figure 3.

Defects in vascular remodeling and network formation in E9.5 Evi1−/− embryos. (A) Whole‐mount PECAM‐1‐stained wild‐type (a) and Evi1−/− (b) embryos at E9.5. Panels (c) and (d) are high‐power views of panels (a) and (b), respectively. The Evi1−/− embryo shows pericardial effusion (arrowhead in (b)). The arrows and arrowheads in (c) indicate the remodeled and organized arteries (arrows) and venous vessels (arrowheads). Highly branched small capillaries and network forming vessels were observed in wild type. Evi1−/− counterparts in (d) show no remodeled and organized arteries or veins, and a smaller caliber change in the vessels. (e, f) Yolk sac vascularization. Highly branched small capillaries were evident in wild type (e), while no remodeled or caliber changed vessel was observed in the yolk sac of Evi1−/− (f). (B) Requirement of HSC development for angiogenesis in vitro. P‐Sp explants from Evi1−/− (a) and wild‐type (b) embryos were dissected at E9.5 and cultured on OP9 cells. P‐Sp cultures were stained with anti‐PECAM‐1 mAbs. The number of hematopoietic cells (round cells) in Evi1−/− P‐Sp cultures (Evi1−/−) (b) was significantly less than that observed in wild‐type (WT) cultures (a). Defects in the vascular network (vn) in vitro are evident in Evi1−/− P‐Sp cultures (b) in comparison to WT cultures (arrows in (a)). The bar indicates 0.1 mm. (c) Development of vascular network stained with anti‐PECAM‐1 mAbs in Evi1−/− P‐Sp culture with HSCs enriched from GFP‐positive adult bone marrow (Evi1−/−+GFP‐BM). (d) Detection of GFP‐positive cells by fluorescence microscopy in (c). The development of vascular network in Evi1−/− P‐Sp culture with HSCs from E9.5 WT embryos (Evi1−/−+HSC E9.5) (e) or with the addition of 200 ng/ml recombinant Ang‐1 (Evi1−/−+Ang‐1). (f) Recombinant Tie2‐Fc fusion protein was added with HSCs from E9.5 WT embryos and inhibited the endothelial cell growth of Evi1/ cells by HSC (Evi1−/−+HSC E9.5/Tie2‐Fc). (g) Recombinant CD4‐Fc fusion protein was added with HSC and did not inhibit the effect by HSC as a control HSC (Evi1−/−+HSC E9.5/CD4‐Fc) (h). The bar indicates 0.1 mm.

Ang‐1 secreted from HSCs prevents defective angiogenesis in P‐Sp culture from Evi1/ embryos

To analyze the vascular defects in Evi1−/− embryo, we used an in vitro P‐Sp culture analysis on OP9 cells to determine the in vitro differentiation of endothelial cells and vascular network formation (Takakura et al, 2000). Explants from wild‐type embryos developed many round cells adhering to the presumptive vascular network area (Figure 3Ba). In contrast, Evi1−/− P‐Sp explants generated a small number of hematopoietic cells (Figure 3Bb). After immunostaining with anti‐PECAM‐1 antibody, it was shown that the wild‐type explant generated vascular beds (vb) and networks (vn) (Figure 3Ba); on the other hand, poor vascular network formation was observed in the cultures of Evi1−/− explants (Figure 3Bb). In proportion to the number of hematopoietic cells, the area of the vascular network was reduced in culture of Evi1−/− explants (Figure 3Bb). Since the network formation is normally promoted by Ang‐1 secreted from developing HSCs in P‐Sp cultures (Takakura et al, 2000), the defects in the development and proliferation of HSCs in Evi1−/− P‐Sp cultures have been suggested to possibly be one of the reasons for the failure in vascular remodeling.

To test the hypothesis that hematopoietic cells could promote angiogenesis in Evi1−/− P‐Sp cultures, HSCs from the adult bone marrow of transgenic mice harboring green fluorescent protein (GFP) (Okabe et al, 1997) were sorted by flow cytometry and then added to a P‐Sp culture of Evi1−/− embryos. As shown in Figure 3Bc, a fine vascular network stained by anti‐PECAM‐1 antibody was developed and coincided with the region where HSCs formed a colony. Under fluorescence microscopy, all GFP‐positive cells demonstrated a round shape (Figure 3Bd), suggesting that exogenous adult HSCs did not differentiate into elongated endothelial cells. Moreover, the addition of CD45+c‐Kit+CD34+ cells from E9.5 wild‐type embryos or recombinant Ang‐1 (200 ng/ml) to Evi1−/− P‐Sp cultures also promoted the vascular network (Figure 3Be and f). Next, to determine whether Ang‐1 secreted from HSCs could restore the defective angiogenesis in Evi1−/− embryos, we added HSCs from E9.5 wild type either with the recombinant Tie2‐Fc fusion protein, which contains the Tie2 ectodomain and thereby acts as an inhibitor by sequestering Ang‐1, or with recombinant CD4‐Fc as a control to Evi1−/− P‐Sp cultures. The vascular networks were restored by HSCs with CD4‐Fc (Figure 3Bh), but they were not restored by addition of HSCs with Tie2‐Fc (Figure 3Bg), suggesting that the defective angiogenesis in Evi1−/− embryos is possibly related to a defect in HSCs due to the production of Ang‐1.

Decreased expression of Tie2/angiopoietin signaling molecules and GATA gene family in the Evi1/ P‐Sp region

Next, expression of regulator molecules in angiogenesis at the Evi1−/− P‐Sp region was examined by quantitative RT–PCR. The expression of Ang‐1, Ang‐2 and Tie2 in P‐Sp from Evi1−/− embryos decreased markedly, while the expression of vascular endothelial growth factor receptor (VEGFR)‐1 and VEGFR‐2 did not significantly change (Figure 4), suggesting that loss of angiopoietin signaling likely causes the defects in vascular remodeling seen in Evi1−/− embryos (Figure 3B). On the other hand, it was reported that the angiopoietin signaling is not required for the differentiation and proliferation of embryonic and fetal hematopoiesis (Puri and Bernstein, 2003). Since GATA‐2 is essential for the proliferation and maintenance of HSCs and hematopoietic progenitors (Tsai et al, 1994), and GATA‐1 plays a role in the differentiation of hematopoietic cells, the expression of GATA‐1 and GATA‐2 was determined and it was found to markedly decrease in the P‐Sp region of the Evi1−/− embryos (Figure 4). In contrast, the expression of SCL, which is essential for the development of definitive HSCs, did not significantly change, suggesting that the decreased expression of GATA‐1 or/and GATA‐2 is probably related to the defective HSC activity in the Evi1−/− embryo.

Figure 4.

Decreased expression of angiopoietin signaling molecules and GATA gene family in the Evi1/ P‐Sp region. An analysis of mRNA expression of several genes related to hematopoiesis or angiogenesis in P‐Sp from Evi1−/−, Evi1+/− and the wild‐type embryos at E9.5 by quantitative RT–PCR. The mRNA expression of each gene was normalized to that of GAPDH mRNA and calibrated to the gene/GAPDH ratio (ΔCT) in wild‐type embryos. The relative expression rate in each gene is presented as the mean and standard deviation of 2ΔΔCt in quadruplicate assays. *P<0.001; **P<0.01, relative to controls.

In vitro prevention of defective proliferation of Evi1/ HSCs via GATA‐2

To further support the hypothesis that Evi1 plays a role in regulating the expression of GATA‐1 or GATA‐2, we examined whether the Evi1 expression prevents a decreased expression of GATA family in Evi1−/− P‐Sp. P‐Sp explants alone were infected with either an Evi1 retrovirus (Evi1‐IRES‐EGFP) or a mock EGFP retrovirus (EGFP) as a control (Kaneko et al, 2001) and were then collected for a quantitative RT–PCR analysis on day 4. A two‐fold increase in the expression level of GATA‐2 mRNA was detected in Evi1−/− P‐Sp infected with Evi1 retrovirus (Figure 5A), while the GATA‐2 mRNA levels remained lower in Evi1−/− P‐Sp infected with control EGFP virus. There were no alterations in the expression levels of GFP among P‐Sp explants, which were all infected with either Evi1 or EGFP alone (Supplementary Figure 4A). Along with the GATA‐2 expression, the expression of Tie2 and GATA‐1 was upregulated by the Evi1 expression, but the expression of Ang‐1 and Ang‐2 did not significantly change (Figure 5A).

Figure 5.

Prevention of defective hematopoiesis in Evi1−/− P‐Sp cultures by GATA‐2. (A) Evi1 maintains the expression of GATA‐2 mRNA in E9.5 Evi1−/− P‐Sp. P‐Sp explants from both Evi1−/− (Evi1−/−) and wild‐type (WT) E9.5 embryos were infected with either an Evi1 retrovirus (Evi1; open square) or a mock EGFP retrovirus (GFP; closed square) as a control. The cells from P‐Sp cultures were harvested at 4 days in culture. The mRNA expression of each gene was assayed by quantitative RT–PCR, and then it was normalized to that of GAPDH mRNA and calibrated to the gene/GAPDH ratio (ΔCT) in wild‐type P‐Sp cells using a control expression vector (EGFP). The relative expression rate in each gene is presented as the mean and standard deviation of 2ΔΔCt in quadruplicate assays. *P<0.001; **P<0.01, relative to controls. (B) P‐Sp explants (Evi1−/− or WT) on OP9 cells were infected with GATA‐1 (GATA‐1), GATA‐2 (GATA‐2), Evi1 (Evi1) or a mock EGFP (GFP) retrovirus as a control. The columns represent the ratio of the number of CD45+c‐Kit+CD34+ cells in P‐Sp cultures relative to those in the wild‐type P‐Sp cultures infected with a mock EGFP retrovirus (100%). Three independent experiments were performed.

Moreover, we examined whether the enforced expression of Evi1, GATA‐1 or GATA‐2 enables the proliferative capacity of HSCs derived from Evi1−/− embryos in vitro to return to its original level. P‐Sp explants on OP9 cells were infected with Evi1, GATA‐1, GATA‐2 or a mock (EGFP) retrovirus as a control. On day 7 of culture, the cells were subjected to FACS analysis and the number of CD45+c‐Kit+CD34+ cells was counted. Many GFP+ cells were generated from Evi1−/− P‐Sp infected with the Evi1 (+Evi1) or the GATA‐2 retrovirus (+GATA‐2) and most of them were CD45+c‐Kit+CD34+ cells (Supplementary Figure 4B). On the other hand, only a few GFP+CD45+c‐Kit+CD34+ cells were generated from Evi1−/− P‐Sp infected with the GFP retrovirus (+GFP). As shown in Figure 5B, the Evi1 or GATA‐2 expression, but not that of GATA‐1, in Evi1−/− P‐Sp was able to increase the number of CD45+c‐Kit+CD34+ cells up to that of wild‐type P‐Sp. Therefore, these results showed that the enforced expression of Evi1 or GATA‐2 helps to maintain the proliferative capacity of HSCs and/or early hematopoietic progenitors in Evi1−/− embryos. Taken together, these data suggested that Evi1 plays a crucial role in both maintaining and proliferating HSCs in the P‐Sp region by upregulating GATA‐2 expression.

Reduced expression of GATA‐2 in each HSC from Evi1/ mutant embryos

To determine whether Evi1 plays a role in regulating the expression of GATA‐2 in vivo, we assayed the expression of the GFP reporter driven by the GATA‐2 7.0IS promoter (kindly provided by Dr Masayuki Yamamoto, Tsukuba University, Japan) in Evi1−/− embryos. The murine GATA‐2 gene utilizes two different first exons. The distal IS exon is specifically used in hematopoietic and neural cells, whereas the proximal IG exon is used in a variety of cells (Minegishi et al, 1998). In transgenic mice containing the 7.0ISGFP construct, which includes the 7.0 kb 5′ flanking region of the GATA‐2 IS exon, the GFP expression shows the endogenous expression profile of GATA‐2 in hematopoietic and neural cells (Minegishi et al, 1999). These mice enabled us to analyze the expression of GATA‐2 in vivo. GFP‐positive cells were easily detected in P‐Sp, liver rudiment and the vitelline artery (VA) in 7.0IS‐GFP/Evi1+/+ mouse embryos, while the 7.0IS‐GFP/Evi1−/− mouse embryos showed a markedly decreased number of GFP‐positive cells (Figure 6A and B). Moreover, the intensity of GFP significantly decreased in each sorted CD45+c‐kit+CD34+ cell in P‐Sp from 7.0IS‐GFP/Evi1−/− embryos in comparison to that from 7.0IS‐GFP/Evi1+/+ embryos (Figure 6C), suggesting that the GATA‐2 expression was selectively impaired in each purified CD45+c‐Kit+CD34+ cell in P‐Sp of Evi1−/− mice. These findings strongly support the hypothesis that Evi1 plays a role in the transcriptional regulation of GATA‐2 in HSCs.

Figure 6.

Loss of GATA‐2 mRNA in P‐Sp of E9.5 Evi1−/− embryos. The decreased transcription of GATA‐2 in E9.5 Evi1−/− embryos. The expression of the transgene driven by the GATA‐2 7.0IS promoter was analyzed in Evi1−/− embryos. 7.0IS‐GFP transgenic mice were mated with Evi1+/− mice to create 7.0IS‐GFP/Evi1+/− transgenic mice. After the crossing of 7.0IS‐GFP/Evi1+/− transgenic mice and Evi1+/− mice, the GFP expression in 7.0IS‐GFP/Evi1−/− transgenic mice was compared to that seen in 7.0IS‐GFP/Evi1+/− transgenic mice. (A) GATA‐2‐GFP‐positive cells were observed in P‐Sp (arrow), liver rudiment (arrow head), the vitelline artery (VA, open arrow) and the dorsal artery (DA) in 7.0IS‐GFP/Evi1+/+ transgenic mice (upper panel), while the frequency of GATA‐2‐GFP‐positive cells was markedly decreased in the 7.0IS‐GFP/Evi1−/− transgenic mice (lower panels). (B) Higher magnification of lower trunk region, which includes P‐Sp of 7.0IS‐GFP/Evi1+/+ transgenic mice (upper panel) and 7.0IS‐GFP/Evi1−/− transgenic mice (lower panels). The arrowheads indicate the P‐Sp region. Few GFP+ cells were detected in the P‐Sp of 7.0IS‐GFP/Evi1−/− transgenic mice. (C) GFP fluorescence intensity. Histograms of the GFP‐positive cells for the CD45+c‐Kit+CD34+ cell populations, which are included in P‐Sp of 7.0IS‐GFP/Evi1+/+ transgenic mice (upper panel) and 7.0IS‐GFP/Evi1−/− transgenic mice (lower panels). The data are representative of three independent experiments.

Regulation of Evi1 in activation of the GATA‐2 IS promoter

To determine whether Evi1 directly regulates the transcription of GATA‐2, we first searched for the DNA‐binding sequences of Evi1 in the 7.0 kb 5′ promoter region of the GATA‐2 IS exon, which is necessary for the endogenous expression of GATA‐2 in hematopoietic and neural cells (Minegishi et al, 1998, 1999). Evi1 protein contains two independent DNA‐binding domains, 1 and 2 (D1 and D2), with seven and three zinc‐finger repeats, respectively. Each domain binds to a different consensus sequence, similar to the GATA sequences in the case of D1 (D1‐CONS: [GAC/TA] N0–6 [GAT/CA]) and the ETS sequences in the case of D2 (D2‐CONS: GAAGAT or GATGAG) (Morishita et al, 1995). We found four possible binding sequences for D1 (Supplementary Figure 5A and Supplementary Table 2A) and eight for D2 (Supplementary Figure 5D and Supplementary Table 2B) in the GATA‐2 IS promoter region. To examine the DNA‐binding capacity in vitro, each GST‐fused DNA‐binding domain for Evi1 (Evi1‐D1 or Evi1‐D2) was used in an electrophoretic mobility shift assay (EMSA) (Morishita et al, 1995). Among the four possible binding sites for D1, only one site (GACATGGACA, b in Figure 6A) could bind to Evi1‐D1 and it showed a similar binding affinity for the original D1‐CONS (Supplementary Figure 5A). On the other hand, five of eight possible binding sites for D2 (Figure 7A) could bind to Evi1‐D2 with the same binding affinity for the original D2‐CONS (Supplementary Figure 2E and F).

Figure 7.

Evi1 directly binds to the promoter region of GATA‐2 and thus enhances the GATA‐2 transcription. (A) Structure of reporter plasmids of the GATA‐2 promoter region. Various lengths (1.6, 3.3, 6.0 and 7.0 kb) of the promoter region and 7.0 kb of the GATA‐2 promoter region containing mutations in the binding site (b) (from GACATGGACA to GACATGGAGA) (7mb) or in the binding site (f) (from GATGAG to TATGAG) (7mf) were inserted upstream of the luciferase gene in a promoter‐less reporter plasmid (pGL3‐Basic). The top line shows six possible binding sites (b, f–h, j and l) and four unique restriction sites (KpnI, XhoI, XbaI and NotI) located upstream of the IS exon. (B) GATA‐2 promoter activity and transactivation by exogenous Evi1 in HEL cells. Various GATA‐2 IS promoter‐LUC reporter constructs (pGL3‐B1.6, 3.3, 6.0 and 7.0) were transfected into HEL cells with (black bar) or without (white bar) an Evi1 expression vector. A mock vector (pGL3‐B) was used as a control (Vec). (C) GATA‐2 promoter activity and transactivation by exogenous Evi1 in EML C1 cells. The experiment was performed as in HEL cells. (D) Specific DNA binding of Evi1 to site b detected by ChIP. Five genomic DNA fragments containing each possible DNA‐binding site (D1‐a or D1‐b) were amplified from the genomic DNA from fixed EML C1 cells after immunoprecipitation with normal rabbit serum (NRS) or with anti‐Evi1 antibody (αEvi1). The localization of the amplified regions in GATA‐2 promoter is shown in Supplementary Figure 6A. For controls, each genomic region was amplified from purified DNA with the indicated dilutions after formaldehyde fixation and sonication (Input).

We next examined whether Evi1 could regulate the transcription of GATA‐2 through direct binding to the binding sequence b for D1 (D1‐b) or five other binding sequences for D2 (D2‐f, ‐g, ‐h, ‐j and ‐l) (Figure 7A) using hematopoietic cell lines. Various genomic fragments of the GATA‐2 IS promoter region were isolated and inserted upstream of luciferase cDNA (Luc) in the pGL3‐Basic vector (Figure 7A). Each reporter plasmid was transiently transfected into EML C1, a murine stem cell factor‐dependent multipotent progenitor cell line (Tsai et al, 1994), or HEL, a human erythroleukemia cell line (Martin and Papayannopoulou, 1982). GATA‐2 mRNA is highly expressed in both cell lines, while the Evi1 mRNA level is higher in HEL but relatively low in EML C1 (data not shown). In HEL cells, the luciferase activity was detected only with the pGL3‐B7.0 plasmid, suggesting that it was primarily driven by DNA derived from a 1 kb fragment located −7.0 to −6.0 kb upstream of the GATA‐2 IS promoter (Figure 7B). When each reporter plasmid was transfected with the Evi1 expression plasmid, the luciferase activity of pGL3‐B7.0 increased from five‐ to eight‐fold. In EML C1 cells, the luciferase activity gradually increased as we used constructs from −1.6 to −7.0 kb 5′ of the IS promoter, until reaching a 10‐fold maximal activation (Figure 7C). When each reporter plasmid was transfected with the Evi1 expression plasmid, the luciferase activity with the highest transactivation of 2.5‐fold by exogenous Evi1 was obtained with the pGL3‐B7.0 plasmid. In this experiment, the activation of GATA‐2 promoter by the enforced Evi1 expression was only around two‐fold that seen in EML C1 and HEL cells, probably because Evi1 mRNA is already expressed in both cell lines and/or both cell lines may not have a transcriptional setting similar to that of P‐Sp in terms of GATA‐2 gene transcription.

In these experiments, the region between −7.0 and −6.0 kb showed the highest promoter activity in both cell lines and its promoter activity was enhanced by exogenous Evi1, suggesting that Evi1 may bind to the region between −7.0 and −6.0 kb of the GATA‐2 IS promoter. Two possible Evi1 binding sites, D1‐b and D2‐f, are located within this region. In order to determine the critical site for the transcriptional regulation of GATA‐2, we generated a reporter plasmid containing a genomic fragment, 7.0 kb 5′ of the IS exon containing a mutated D1‐b site (pGL3‐B7.0mb) (GACATGGACA to GACATGGAGA) or a mutated D2‐f site (pGL3‐B7.0mf) (GATGAG to TATGAG). In both cell lines, the luciferase activity of pGL3‐B7.0mb decreased to the same level as that of pGL3‐B6.0 and exogenous Evi1 failed to transactivate the promoter, whereas the activity of pGL3‐B7.0mf did not change in comparison to that of pGL3‐B7.0 (Figure 7B and C). These findings indicate that the D1‐b site in the region between −7.0 and −6.0 kb is a critical cis‐element through which Evi1 regulates the GATA‐2 promoter activity.

To support the hypothesis that the GATA‐2 gene is a direct target of Evi1 in vivo, we undertook a chromatin immunoprecipitation (ChIP) analysis. After the fixation of EML C1 cells, a sonicated DNA–protein mixture was immunoprecipitated by anti‐Evi1 antibody. The genomic region was amplified by PCR using primers specific for the D1‐b site with other four different regions in the upstream region of GATA‐2 (Supplementary Figure 6A). As shown in Figure 7D, a 227 bp band containing the D1‐b site (D1‐b) was amplified from genomic DNA precipitated by anti‐Evi1 antibody. However, a 244 bp band of 2 kb upstream of the D1‐b site (Up), a 228 bp band of the D1‐a site (D1‐a), a 245 bp band of 2 kb downstream of the D1‐b site (Down) and β‐actin promoter region (β‐actin) used as a control were not amplified from the same DNA. We finally observed that the forced expression of Evi1 induced GATA‐2 expression in a hematopoietic cell line, EML C1, along with GATA‐1, Ang‐1, Ang‐2 and Tie2 at 2 months after transfection (Supplementary Figure 6B). Taken together, these results support the possibility that GATA‐2 is one of the direct target genes of Evi1.

Discussion

In this study, we showed that Evi1 is expressed at a high level in definitive hematopoietic stem cells of the P‐Sp region detected by quantitative RT–PCR. A detailed analysis of Evi1 homozygous mutant mice clearly showed a marked reduction in CD45+CD34+c‐Kit+ HSCs in the Evi1−/− P‐Sp region, while also showing defects in the proliferation and maintenance of HSCs in Evi1−/− P‐Sp cultures in vitro. Moreover, Evi1−/− P‐Sp cells did not exhibit HSCs showing a repopulation capacity.

Among the transcription factors crucial for the development of definitive HSCs in embryos, we showed that the expression of Evi1 correlates positively with that of GATA‐2 in both the P‐Sp region and CD45+CD34+c‐Kit+ HSCs. Moreover, the defect in Evi1−/− HSCs was most reminiscent of that observed in GATA‐2 homozygous mutant mice (Tsai et al, 1994). We found the GATA‐2 expression to significantly decrease in Evi1−/− embryos. Through an extensive GATA‐2 promoter analysis in vitro, we demonstrated that Evi1 directly regulates the GATA‐2 transcription by binding to the sequence lying between −6127 and −6108 of the 5′ flanking region of the IS (specific) first exon. Furthermore, the constitutive expression of Evi1 or GATA‐2, but not GATA‐1, in Evi1−/− P‐Sp cells helped to prevent the failure of HSC expansion. These results indicate that GATA‐2 is one of the authentic targets of Evi1 in HSCs. In GATA‐2 homozygous mutant mice, severe defects in both the yolk sac and definitive hematopoiesis were seen (Tsai et al, 1994). On the other hand, yolk sac hematopoiesis appeared intact in Evi1−/− embryos, suggesting that a likely differential hierarchical regulation of HSCs exists between the yolk sac and definitive hematopoiesis. Therefore, it is of interest to investigate whether Evi1 also maintains the long‐term repopulating HSCs in bone marrow niche, and to also clarify the authentic targets of Evi1, since we herein revealed that Evi1 mRNA is abundantly and selectively expressed in c‐Kit+Sca‐1+Lin HSCs in adult mouse bone marrow.

We recently showed that HSCs play an important role in angiogenesis during embryogenesis by secreting an angiogenic factor, Ang‐1 (Takakura et al, 2000). Defects in vascular remodeling are phenotypes commonly seen in mice lacking SCL, Runx1 or c‐Myb, in which mutant mice show a defective development of definitive HSCs. These mice show embryonic lethality at midgestation with severe hemorrhaging and/or defects in vascular development (Mucenski et al, 1991; Okuda et al, 1996; Visvader et al, 1998). Adding normal HSCs or Ang‐1 to P‐Sp/OP9 cocultures prevents the defective vascular remodeling in Runx1 homozygous mutant mice, supporting the role of Ang‐1 and HSCs with the secretion of Ang‐1 in angiogenesis (Takakura et al, 2000). Indeed, the expression of Ang‐1 and its cognate receptor, Tie2, significantly decreased in Evi1−/− embryos. Taken together with the marked reduction in HSC numbers, the vascular phenotype of Evi1−/− embryos could be explained in part by inappropriate Ang‐1 signals from HSCs.

Several underlying molecular mechanisms of Evi1 expression have been proposed, including an enhanced AP‐1 activity (Tanaka et al, 1994), the antagonizing growth‐inhibitory effects of TGFβ (Kurokawa et al, 1998) and a differentiation block (Morishita et al, 1992a; Kreider et al, 1993). Since GATA‐2 strongly blocks hematopoietic cell differentiation and stimulates immature cell proliferation (Minegishi et al, 2003), it is possible that the ectopic expression of Evi1 induces an aberrant GATA‐2 expression, which is in some part responsible for the oncogenic activity of Evi1. In this respect, it will be interesting to analyze the contribution of GATA‐2 to the leukemic transformation as a downstream target gene of the Evi1 oncogene.

Materials and methods

Animals

Evi1 mutant mice (Hoyt et al, 1997) and 7.0IS‐GFP transgenic mice (Minegishi et al, 1999) were used in this study. Mice were housed in environmentally controlled rooms of the Laboratory Animal Research Center of Keio University under the guidelines for animal and recombinant DNA experiments of Keio University and of the Research Center for Frontier Bioscience of University of Miyazaki.

In vitro culture of P‐Sp

P‐Sp explants derived from E9.5 Evi1−/−, Evi1+/− and wild‐type embryos were cultured on a layer of OP9 stromal cells at 37°C in humidified 5% CO2 air. Maintenance of OP9 cells and P‐Sp explant cultures for differentiation to endothelial cells were carried out as described previously (Takakura et al, 1998; Oike et al, 1999).

Transplantation of P‐Sp cells into busulfan‐treated neonates

Transplantation of cells into busulfan‐treated neonatal mice was performed as described previously (Yoder et al, 1997) with a slight modification.

Preparation of cDNA from P‐Sp of mice and quantitative RT–PCR analysis

Total RNA was isolated by TRIZOL Reagent (Invitrogen) and 1 μg of DNase‐treated total RNA was reverse transcribed using the Advatage® RT‐for‐PCR Kit (BD Biosciences Clontech). A 10 ng portion of the single‐stranded cDNA products was used in each real‐time PCR reaction (TaqMan 2 × Universal PCR Master Mix, PE Applied Biosystems). All samples were normalized to GAPDH mRNA abundance and measured using an ABI Prism 7700 Sequence Detector analyzed by the Sequence Detection System software (version 1.7a) (PE Applied Biosystems).

Retroviral infection of P‐Sp

The murine GATA‐1, GATA‐2 or Evi1 cDNA was subcloned into a site upstream of an IRES‐EGFP construct in the retroviral vector pGCsam (GCsam‐GATA‐1‐IRES‐EGFP, GCsam‐GATA‐2‐IRES‐EGFP or GCsam‐Evi1‐IRES‐EGFP) as described previously (Kaneko et al, 2001). To produce recombinant retrovirus, plasmid DNA was transfected into 293gp cells along with the VSV‐G expression plasmid by CaPO4 co‐precipitation. In expression analysis, P‐Sp explants on fibronectin‐coated dishes were infected with either an Evi1 retrovirus (GCsam‐Evi1‐IRES‐EGFP) or a mock retrovirus (GCsam‐IRES‐EGFP) on days 1 and 3 of culture. On day 4 of culture, cells were collected for RT–PCR analysis, because P‐Sp explants could not survive for over 5 days in this culture condition. For in vitro P‐Sp explants/OP9 stromal cell coculture analysis, P‐Sp explants were infected with each retrovirus on days 1 and 3 of culture and were subjected to FACS analysis on day 7.

Production of recombinant Tie2‐Fc and CD4‐Fc fusion protein

Recombinant fusion proteins of the extracellular domain of murine surface molecules (Tie2 or CD4) and the Fc part of human Ig were designed (Tie2‐Fc or CD4‐Fc) as described previously (Takakura et al, 1998).

Reporter gene assay of the GATA‐2 promoter region

A 7 kb fragment with the major transcription initiation site of mouse GATA‐2 (IS exon) was isolated by PCR and four different plasmids for reporter gene assays of the GATA‐2 gene upstream regions (1.6, 3.3, 6.0 and 7.0 kb) were constructed as shown in Figure 7D. Site‐directed mutagenesis of the D1‐like sequence site (b) and D2‐like sequence site (f) was carried out using a QuickChange Site‐Directed Mutagenesis Kit (Stratagene).

Chromatin immunoprecipitation analysis

A ChIP analysis was performed as described previously (Weinmann et al, 2001). Briefly, EML C1 (1 × 108) cells were crosslinked with 1% formaldehyde and sonicated. After the chromatin fraction was incubated with anti‐Evi1 antibody, immune complexes were bound to protein A Sepharose 4B beads (Amersham) twice. Eluted DNA samples were then analyzed by PCR using specific primer pairs (Supplementary Table 3B) containing ‘Up’ (2 kb upstream of D1‐b), ‘Down’ (2 kb downstream of D1‐b), the D1‐like sequence D1‐a or D1‐b at GATA‐2 promoter region or murine β‐actin promoter as shown in Figure 7D. Rabbit polyclonal anti‐Evi1 antibody was immunized and made by GST fusion Evi1 protein as described previously (Morishita et al, 1995).

Supplementary data

Supplementary data are available at The EMBO Journal Online.

Supplementary Information

Supplementary Material and Methods [emboj7600679-sup-0001.pdf]

Supplementary Figures [emboj7600679-sup-0002.pdf]

Supplementary Material and Methods [emboj7600679-sup-0003.doc]

Supplementary data [emboj7600679-sup-0004.pdf]

Acknowledgements

We thank Dr Schickwann Tsai for providing EML and BHK/MKL cells, Dr Masayuki Yamamoto for providing 7.0IS‐GFP transgenic mice and Dr Masafumi Onodera for GCsam‐IRES‐EGFP. This work was supported by Grants‐in‐Aid for Scientific Research on Priority Areas and for 21st Century COE program (Life Science) from the Ministry of Education, Culture, Sports, Science and Technology of Japan.

References