Lethal myelofibrosis induced by Bmi1-deficient hematopoietic cells unveils a tumor suppressor function of the polycomb group genes

We previously reported that deletion of both Ink4a and Arf in Bmi1-deficient mice substantially restores the defective self-renewal capacity of HSCs (Oguro et al., 2006). In this study, we performed competitive repopulation assays using the same number of BM competitor cells as the test cells, and found that Bmi1^−/−Ink4a-Arf^−/− cells had higher repopulating activity in recipient mice than the wild-type and Ink4a-Arf^−/− control BM cells (Fig. 1 A). To evaluate the repopulating capacity of Bmi1^−/−Ink4a-Arf^−/− BM cells precisely, we then transplanted wild-type, Ink4a-Arf^−/−, and Bmi1^−/−Ink4a-Arf^−/− BM cells into lethally irradiated mice without competitor cells. At 4 mo after transplantation, the recipients repopulated with Bmi1^−/−Ink4a-Arf^−/− donor cells had significantly fewer lineage marker-negative (Lineage⁻) Sca-1⁺c-Kit⁺ (LSK) cells in BM than did the control recipients as we reported previously (Oguro et al., 2010; Fig. 1 B). However, they had twofold more megakaryocyte/erythroid progenitors (MEPs) than the controls (Fig. 1 B) and showed extramedullary hematopoiesis as evident from a significant increase in the number of LSK HSCs/multipotent progenitors (MPPs) in spleen (Fig. 1 B). All of the recipient mice repopulated with Ink4a-Arf^−/− BM cells eventually developed sarcomas or lymphomas and died by 11 mo after transplant, as reported with the Ink4a-Arf^−/− mice (Fig. 1 C; Serrano et al., 1996). Alternatively, the recipient mice repopulated with Bmi1^−/−Ink4a-Arf^−/− BM cells died much earlier than the Ink4a-Arf^−/− controls (Fig. 1 C) and displayed a more progressive thrombocytopenia (Fig. 1 D).

The recipient mice repopulated by Bmi1^−/−Ink4a-Arf^−/− BM cells had marked hepatosplenomegaly (not depicted) and hypoplastic BM with severe fibrosis (Fig. 2, A and B) at their terminal stage. The spleen structure was destroyed by extramedullary hematopoiesis accompanied by massive fibrosis (Fig. 2, A and B). Extramedullary hematopoiesis was also evident in the liver (Fig. 2 A). All these features resemble those of human primary myelofibrosis (PMF). Notably, the recipient mice infused with Bmi1^−/−Ink4a-Arf^−/− BM cells along with competitor cells in Fig. 1 A also developed lethal myelofibrosis in a similar fashion (not depicted).

PMF is the rarest and most severe chronic MPN (Tefferi et al., 2007; Levine and Gilliland, 2008). Abnormal megakaryocytosis in the BM has been proposed as the main causative factor for myelofibrosis. Deregulated stem cell signaling, resulting in part from mutated JAK2 and MPL, likely cause abnormal megakaryocytosis. Myelofibrosis is thought to be the consequence of an excessive release/leakage of growth factors within the BM by cells from pathological hematopoietic clones, especially by necrotic megakaryocytes. TGF-β1 is speculated to be one of the major causative growth factors that activate mesenchymal cells (Martyré et al, 1994). Although abnormal megakaryocytosis was obscure in BM and spleen because of severe fibrosis at the terminal stage of the disease, the mice at earlier time points after transplantation had marked megakaryocytosis in both BM and spleen (Fig. 2, C and D). These findings clearly implicate pathological megakaryocytosis in the development of lethal myelofibrosis, and indicate that lethal myelofibrosis induced by Bmi1^−/−Ink4a-Arf^−/− hematopoietic cells follows the natural course of human PMF.

To identify the genes responsible for PMF-like disease in the absence of Bmi1, we compared gene expression profiles of LSK HSCs/MPPs and common myeloid progenitors (CMPs). In total, 245 and 286 genes were derepressed by more than twofold, specifically in Bmi1^−/−Ink4a-Arf^−/− LSKs and CMPs, respectively (Fig. 3 A). We then compared the list of derepressed genes with a list of PMF-associated genes identified by gene expression profiling of CD34⁺ cells in human PMF patients (Guglielmelli et al., 2007). Hmga2 appeared to be commonly up-regulated in Bmi1^−/−Ink4a-Arf^−/− CMPs and PMF CD34⁺ cells. HMGA2 was found to be one of eight genes that can distinguish PMF CD34⁺ cells from normal CD34⁺ cells, and abnormal expression of HMGA2 was associated with the presence of JAK2^V617F (Guglielmelli et al., 2007). Moreover, overexpression of HMGA2 was reported in 12 of 12 patients with myelofibrosis with myeloid metaplasia, among which two patients had a chromosomal translocation involving the HMGA2 gene at 12q (Andrieux et al., 2004). We therefore focused on Hmga2. Hmga2 expression was up-regulated by 1.6- and 13.4-fold in Bmi1^−/−Ink4a-Arf^−/− LSKs and CMPs, respectively, in our microarray analysis (Fig. 3 A). We then quantified the Hmga2 expression in each progenitor fraction by quantitative RT-PCR. Because the BM environment of Bmi1^−/− and Bmi1^−/−Ink4a-Arf^−/− mice is defective in supporting HSCs (Oguro et al., 2006), we purified progenitors from wild-type recipient BM reconstituted with wild-type, Ink4a-Arf^−/−, or Bmi1^−/−Ink4a-Arf^−/− BM cells to exclude any environmental effects. Hmga2 expression was predominantly found in CMPs in wild-type mice, but in the absence of Bmi1, it was derepressed in Flt3⁻LSK HSCs/MPPs (2.4-fold compared with Ink4a-Arf^−/− cells) and markedly increased in myeloid-committed progenitors (CMPs, 14.1-fold; granulocyte/macrophage progenitors [GMPs], 268-fold; MEPs, 11.5-fold), but was barely affected in Flt3⁺LSK lymphoid-primed MPPs (LMPPs, 0.9-fold) or common lymphoid progenitors (CLPs, 0.74-fold; Fig. 3 B). To confirm that this derepression is in fact mediated by loss of Bmi1, we performed RT-PCR analysis on Lineage⁻c-Kit⁺ Bmi1^−/− progenitors and found that Hmga2, as well as the canonical targets Ink4a and Arf, are derepressed in a Bmi1-deficient setting (Fig. 3 C). These results suggest that Bmi1 functions in the silencing of Hmga2 in cell types ranging from HSCs to myeloid progenitors.

The high-mobility group A (HMGA) nonhistone chromatin proteins alter chromatin structure, and thereby regulate transcription. HMGA proteins have been implicated in both benign and malignant tumors through mechanisms that result in HMGA overexpression. Chromosomal rearrangements involving the region 12q13-15, in which the HMGA2 gene is located, are one of the major mechanisms causing deregulation of the HMGA2 gene, giving rise to a truncated transcript lacking the C-terminal tail and/or 3′-UTR. Expression of HMGA2 is negatively regulated by the let-7 family of microRNAs, which bind to the 3′-UTR of HMGA2 and restrict its expression. Thus, chromosomal rearrangements within the HMGA2 locus cause overexpression of a full-length or truncated HMGA2 with a preserved DNA-binding capacity (Fusco and Fedele, 2007; Young and Narita, 2007).

To examine whether Bmi1 directly represses transcription of Hmga2, we next characterized the Hmga2 promoter by conducting chromatin immunoprecipitation (ChIP) assays. To obtain enough cells, we used BM Lineage⁻c-Kit⁺ progenitors depleted of cells committed to the lymphoid, myeloid, and erythroid lineages. Binding of Bmi1 to the Hmga2 promoter was detected, but not to the promoter of the control gene, β-actin (Fig. 3 D), which was marked with H3K4me3, an active histone mark (not depicted). PRC1 catalyzes the monoubiquitination of histone H2A (H2Aub1) at lysine 119. The Hmga2 promoter was marked with H2Aub1 in wild-type BM Lineage⁻c-Kit⁺ progenitors (Fig. 3 D). The levels of Bmi1 binding and H2Aub were significantly reduced in Bmi1^−/−Ink4a-Arf^−/− Lineage⁻ progenitors compared with the levels in wild-type cells (Fig. 3 E). These results indicate that Bmi1 directly represses the expression of Hmga2 by marking its promoter with a repressive histone mark.

The specific up-regulation of HMGA2 expression in PMF, but not in other chronic MPNs, such as polycythemia vera and essential thrombocythemia, suggested a contribution of HMGA2 to the development and/or progression of PMF (Guglielmelli et al., 2007). To address this, we evaluated the proliferative and differentiation capacity of Hmga2-overexpressing CD34⁻LSK HSCs. In all experiments, transduction efficiency was ∼90% as determined by flow cytometry using GFP as a marker (unpublished data). In liquid cultures supplemented with SCF, TPO, IL-3, IL-6/IL-6 receptor fusion protein (FP6), and EPO, Hmga2-overexpressing HSCs showed a growth advantage compared with the control (Fig. 4 A), and the Hmga2 culture contained significantly more LSK cells than the control at day 10 of culture (Fig. 4 B). Morphological analysis unveiled a drastic enhancement in production of megakaryocytes upon the overexpression of Hmga2. The Hmga2 culture contained significantly more multinucleated megakaryocytes and mononuclear micromegakaryocytes (Fig. 4, C and D).

We then cultured transduced HSCs for 9 d (10 d of culture in total) in the presence of SCF and TPO, which supports the proliferation of HSCs and progenitors rather than their differentiation (Ema et al., 2000). The total number of colony-forming cells (CFCs) was significantly increased in the Hmga2 culture (Fig. 4 E). Among CFCs, the number of high proliferative potential (HPP)-CFCs, which generate a colony with a diameter >1 mm, was increased with Hmga2 compared with the control. These results indicate that overexpression of Hmga2 promotes expansion of progenitor cells and also facilitates the proliferation and differentiation of megakaryocytic cells in vitro.

We next examined the effect of Hmga2 overexpression on hematopoiesis in vivo. We performed competitive repopulation assays using 30 CD34⁻LSK HSCs transduced with Hmga2 (CD45.2) at day 3.5 of culture along with 2 × 10⁵ fresh unfractionated BM cells (CD45.1) for radioprotection. Hmga2-overexpressing cells showed a greater, albeit not statistically significant, contribution to the myeloid lineage in peripheral blood (PB) compared with the control (Fig. 5 A). Recipients repopulated with Hmga2-overexpressing cells had significantly more LSK cells and myeloid progenitors in BM compared with the control (Fig. 5 C). Furthermore, they also had mild splenomegaly with extramedullary hematopoiesis as evident from a marginal increase in LSK HSCs/MPPs in the spleen and a significant increase in the PB (Fig. 5 B and C). The recipients repopulated with Hmga2-overexpressing cells appeared to have more megakaryocytes in the BM (not depicted) and platelets in the PB (Fig. 5 D). These findings indicate that derepression of Hmga2 could induce an MPN-like state with enhanced megakaryocytopoiesis and correspond well with findings recently obtained with transgenic mice carrying a 3′-UTR–truncated Hmga2 cDNA (ΔHmga2 mice; Ikeda et al., 2011). ΔHmga2 mice developed MPN-like hematopoiesis with an increased number of megakaryocytes in the BM, although the number of platelets in the PB was not described. Together, these findings implicate Hmga2 in the PMF-like pathological hematopoiesis induced by Bmi1^−/−Ink4a-Arf^−/− BM cells.

Given that HMGA2 is highly derepressed in CD34⁺ stem and progenitor cells in PMF patients, PcG function could be compromised in PMF patients, partly by loss-of-function mutations of EZH2. Indeed, we confirmed that Ezh2 also regulates Hmga2 as Bmi1 does. Hmga2 was highly derepressed in Ezh2-deficient progenitor cells (unpublished data). ChIP assays demonstrated that the Hmga2 promoter is bound by Ezh2 and marked with H3K27me3 in wild-type BM Lineage⁻c-Kit⁺ progenitors (unpublished data). Interestingly, recipient mice repopulated with Ezh2-deficient BM cells showed an increase in the platelet count in PB similar to recipients repopulated with HSCs overexpressing Hmga2 (Mochizuki-Kashio et al., 2011). These data indicate that both Bmi1 and Ezh2 transcriptionally repress Hmga2. However, the frequency of EZH2 mutations in PMF has been reported to be only 13% (Nikoloski et al., 2010). Thus, it would be intriguing to examine other mechanisms that compromise PcG function in PMF patients without EZH2 mutations.

Nevertheless, it should be noted that Hmga2 is not the only oncogenic target of Bmi1 responsible for the establishment of PMF-like disease, as fibrosis was not seen in recipient BM that was repopulated with Hmga2-overexpressing cells at 6 mo after transplantation (unpublished data). As evident from the significant increase in platelet counts, overexpression of Hmga2 does not compromise the terminal differentiation of megakaryocytes. It has been proposed that necrotic megakaryocytes in BM are the main causative factor for myelofibrosis. Although derepression of Hmga2 induces an MPN-like state with enhanced megakaryocytopoiesis, it may require the derepression of other Bmi1 targets to eventually induce a fibrotic state after a preceding MPN-like state. Our microarray analysis also identified Hoxc4 and Foxa1 as being derepressed in Bmi1^−/−Ink4a-Arf^−/− cells (Fig. 3 A). We confirmed this derepression by RT-PCR (unpublished data), and Hoxc4 and Foxa1 also appeared to be direct targets of Bmi1 by ChIP assays (unpublished data). We tested the effects of forced expression of these genes in HSCs, but failed to induce an MPN-like disease in recipient mice (unpublished data).

Collectively, our findings indicate that Bmi1 antagonizes the development of MPN in the absence of its major tumor-suppressive targets, Ink4a and Arf, and we showed the tumor suppressive function of Bmi1 through epigenetic silencing of oncogenes. Although Bmi1-deficiency is self-limiting unless Ink4a and Arf are deleted first, INK4A and ARF are frequently inactivated by deletions or mutations, or transcriptionally repressed by DNA methylation at their promoters in human cancers. Therefore, the situations like the Ink4a/Arf-null background that we used in this study also happen in the initiation and progression of human cancers. Our findings suggest that in such situations, the tumor cells with impaired BMI1 function could outcompete cells with normal BMI1 function because the effects of derepressed oncogenes, such as HMGA2, appear to supersede the effects of derepressed tumor suppressor genes. Corresponding to our findings, PRC1 has been reported to exert tumor suppressor activity through epigenetic silencing of Notch and JAK-STAT signaling in Drosophila melanogaster eyes (Martinez et al., 2009; Classen et al., 2009). Furthermore, mice with hypomorphic mutations of Eed and Suz12 show enhanced hematopoiesis (Lessard et al., 1999; Majewski et al., 2008). All of these findings support the tumor suppressor function of EZH2 observed in human myelodysplastic syndrome and MPN, and are suggestive of a broad range of target genes of the PcG proteins, including both oncogenes and tumor suppressor genes. Although tumor suppressor genes have been stressed as PcG targets, our findings shed light on the role of PcG proteins in the gene silencing of oncogenes. Thus, the PcG proteins fine-tune the growth of hematopoietic cells in both a positive and a negative manner to maintain hematopoietic homeostasis.

Bmi1^+/− mice (van der Lugt et al., 1994) and Ink4a-Arf^−/− mice (Serrano et al., 1996) that had been backcrossed at least eight times onto a C57BL/6 (CD45.2) background were used. C57BL/6 (CD45.2) mice and C57BL/6 mice congenic for the Ly5 locus (CD45.1) were purchased from Japan SLC and Sankyo Laboratory Service, respectively. Littermates were used as controls in all experiments. All mice were bred and maintained in the Animal Research Facility of the Graduate School of Medicine, Chiba University in accordance with institutional guidelines. All experiments using mice received approval from the Chiba University Administrative Panel for Animal Care.

BM cells (10⁶) from 4-wk-old CD45.2 mice were mixed with the same number of unfractionated BM competitor cells (CD45.1) and transplanted into CD45.1 mice irradiated at a dose of 9.5 Gy. PB cells of the recipient mice were analyzed with a mixture of antibodies that included PE-Cy7–conjugated anti-CD45.1, Pacific blue–conjugated anti-CD45.2, PE-conjugated anti–Mac-1 and anti–Gr-1, APC-conjugated anti-B220, and APC-Cy7–conjugated anti-CD4 and anti-CD8α antibodies. Cells were analyzed on a FACSCanto II (BD). Donor cell chimerism in the recipient PB cells was evaluated as percent donor chimerism calculated as (percent donor cells) × 100/(percent donor cells + percent recipient cells). PB cell counts were made using an automated cell counter (Celltec α; Nihon Kohden).

Mouse HSCs (CD34⁻LSK cells) were purified from BM of 8-wk-old mice. Mononuclear cells were isolated on Ficoll-Paque PLUS (GE Healthcare). The cells were stained with an antibody cocktail consisting of biotinylated anti–Gr-1, Mac-1, IL-7Rα, B220, CD4, CD8α, and Ter119 monoclonal antibodies. Lineage⁺ cells were depleted with goat anti–rat IgG microbeads (Miltenyi Biotec) through an LD column (Miltenyi Biotec). The cells were further stained with FITC-conjugated anti-CD34, PE-conjugated anti–Sca-1, and APC-conjugated anti–c-Kit antibodies. Biotinylated antibodies were detected with APC-Cy7–conjugated streptavidin. Analysis and sorting were performed on a FACS Aria II (BD). CMPs, GMPs, MEPs, and CLPs were analyzed as CD34⁺FcγR^lowc-Kit⁺Sca-1⁻Lineage⁻IL-7Rα⁻ cells, CD34⁺FcγR^hic-Kit⁺Sca-1⁻Lineage⁻IL-7Rα⁻ cells, CD34⁻FcγR^lowc-Kit⁺Sca-1⁻Lineage⁻IL-7Rα⁻ cells, and IL-7Rα⁺c-Kit^lowSca-1^lowLineage⁻ cells, respectively.

Full-length mouse Hmga2 cDNA in the pMY-ires-EGFP retroviral expression vector was kindly provided by K. Nakashima (Nara Institute of Science and Technology, Ikoma, Nara, Japan). A recombinant vesicular stomatitis virus glycoprotein (VSV-G)-pseudotyped high-titer retrovirus was generated by a 293gpg packaging cell line that had been engineered to express the VSV-G protein under the control of a tetracycline-inducible system. The virus in supernatants of 293gpg cells was concentrated by centrifugation at 6,000 g for 16 h. Viral titers were determined by infecting Jurkat cells (a human T cell line).

CD34⁻LSK HSCs were transduced with the indicated retrovirus, as previously described, with minor modifications (Iwama et al., 2004). CD34⁻LSK HSCs were sorted into 96-well microtiter plates coated with the recombinant human fibronectin fragment CH-296 (RetroNectin; Takara Shuzo) at 100 cells per well, and then incubated in α-MEM supplemented with 1% FBS, 1% l-glutamine, penicillin, streptomycin solution (GPS; Sigma-Aldrich), 50 µM 2-mercaptoethanol (2-ME), 100 ng/ml mouse stem cell factor (SCF; PeproTech), and 100 ng/ml human thrombopoietin (TPO; PeproTech) for 24 h. Next, cells were transduced with the indicated retrovirus at a multiplicity of infection (MOI) of 1,500 in the presence of 1 µg/ml RetroNectin and 10 µg/ml protamine sulfate (Sigma-Aldrich) for 24 h. After transduction, cells were further incubated in the same medium for 9 d, subjected to in vitro colony assays or in S-Clone SF-O3 (Sanko Junyaku) supplemented with 0.2% BSA, 50 µM 2-ME, 1% GPS, 50 ng/ml SCF, and 50 ng/ml TPO for 2.5 d, and then subjected to competitive repopulation assays. The transduction efficiency was nearly 90%, as judged from GFP expression. Colony assays were performed using a methylcellulose medium (M3234; STEMCELL Technologies) supplemented with 20 ng/ml mouse SCF, 20 ng/ml mouse IL-3 (PeproTech), 50 ng/ml human TPO, and 3 U/ml human EPO (provided by Kyowa Hakko Kirin). Colony numbers were counted on day 10. CD34⁻LSK HSCs (CD45.2) transduced with the indicated retrovirus were also transplanted intravenously into 8-wk-old CD45.1 mice irradiated at a dose of 9.5 Gy, together with 2 × 10⁵ BM competitor cells from 8-wk-old CD45.1 mice.

IL-7Rα⁻LSK cells were purified from BM of 4-wk-old Ink4a-Arf^−/− and Bmi1^−/−Ink4a-Arf^−/− mice and CMPs from recipients’ BM at 4 mo after infusion of Ink4a-Arf^−/− or Bmi1^−/−Ink4a-Arf^−/− BM cells. Total RNA was isolated using TRIZOL LS Reagent (Invitrogen), and its integrity was confirmed using LabChip RNA 6000 Nano chips and a 2100 Bioanalyzer (Agilent Technologies). Target cRNA was prepared from the total RNA equivalent to 10,000 cells (IL-7Rα LSK) or 8,000 cells (CMPs) with a two-cycle cDNA synthesis kit and 3′-amplification reagents for IVT labeling (Affymetrix), and then hybridized to a GeneChip Mouse Genome 430 2.0 oligonucleotide microarray (Affymetrix) according to the manufacturer’s instructions. The expression value of each gene was calculated and normalized using GeneChip Operating Software version 1.4 (Affymetrix). All data are MIAME compliant, and the raw data were deposited in Gene Expression Omnibus (accession no. GSE19796 and GSE31086).

Total RNA was isolated using TRIZOL LS solution (Invitrogen) and reverse transcribed by the ThermoScript RT-PCR system (Invitrogen) with an oligo dT primer. Real-time quantitative PCR was performed with an ABI prism 7300 Thermal Cycler (Applied Biosystems) using FastStart Universal Probe Master (Roche). Hypoxanthine-guanosine phosphoribosyl transferase 1 (Hprt1) expression was used to calculate relative expression levels. The combination of primer sequences and probe numbers are as follows: Hmga2, probe #26, 5′-AAGGCAGCAAAAACAAGAGC-3′ and 5′-CCGTTTTTCTCCAATGGTCT-3′; Hprt1, probe #95, 5′-TCCTCCTCAGACCGCTTTT-3′ and 5′-CCTGGTTCATCATCGCTAATC-3′.

Lineage⁻ or Lineage⁻c-Kit⁺ BM cells (2 × 10⁵ cells/antibody) isolated by flow cytometry were cross-linked with 1% formaldehyde for 15 min at room temperature, and incubated for 5 min at 4°C after the addition of 0.125 M glycine. Cells were washed with PBS, lysed with cell lysis buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 150 mM NaCl, 0.5% SDS, and protease inhibitor cocktail [PIC; Roche]) on ice, and sonicated until the DNA fragments were 200–500 bp in mean size as measured by Bioruptor (Cosmo Bio). After centrifugation at 15,000 rpm for 10 min, sheared chromatin was diluted 10-fold in dilution buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1.1% Triton X-100, 0.11% sodium deoxycholate, and PIC), and precleared by addition of Dynabeads Protein G (Invitrogen) for 1 h at 4°C. For anti-Bmi1 (clone 8A9; provided by N. Nozaki, MAB Institute, Yokosuka, Kanagawa, Japan), precleared chromatin was immunoprecipitated overnight at 4°C with antibody/Dynabeads Protein G mix. For anti–ubiquityl-Histone H2A (H2Aub; clone E6C5, 05–678; Millipore), precleared chromatin was immunoprecipitated overnight at 4°C with anti-H2Aub, followed by the addition of anti–mouse IgMμ (12–488; Millipore)/Dynabeads Protein G mix and incubation for 2 h at 4°C. Beads were then sequentially washed with the following combination of wash buffers; twice each with low-salt wash buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, and 0.1% sodium deoxycholate), high-salt wash buffer (50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, and 0.1% sodium deoxycholate), and LiCl wash buffer (10 mM Tris-HCl, pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% NP-40, and 0.5% sodium deoxycholate). Bound chromatin was eluted and, together with input DNA, cross-linking was reversed in elution buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, and 1% SDS) by overnight incubation at 65°C. Immunoprecipitated DNA and input DNA were treated with RNaseA (Sigma-Aldrich) and proteinase K (Roche), and extracted with phenol:chloroform. Quantitative PCR analysis was performed using SYBR Premix Ex Taq II (Takara Bio), and the 7500/7500 Fast Real-Time PCR system (Applied Biosystems). Primer sequences are as follows: Hmga2, region 1, 5′-GGGAAGCCAGCAGAGGTAAGCC-3′ and 5′-CGAGCGCATTTGCACGGCTC-3′; region 2, 5′-CTGGCACCATCGTGTGTCTGG-3′ and 5′-TGCGCGCACACACACTTCACT-3′; region 3, 5′-GGTCGCTCTTTTCCCGGGGC-3′ and 5′-TCCACCGAGGGTTGCCCG-3′; region 4, 5′-GCCGCCTTCGAGGCAGTTGT-3′ and 5′-CAAGAGGAGGGGGCAGGCCA-3′; region 5, 5′-AAAACTTGGGCTCCGGGTGCAGA-3′ and 5′-GGGCGCCCAGCTCAGCTCTAG-3′; region 6, 5′-CGCTGGACGTCCGGTGTTGAT-3′ and 5′-AAGAGCGGCGAGAGCAGGCG-3′; Actb, 5′-CCCAACACACCTAGCAAATTAGAACCAC-3′ and 5′-CCTGGATTGAATGGACAGAGAGTCACT-3′.

Frozen spleen sections were prepared and immunostained according to the method described by Kawamoto (2003). After fixation with 100% ethanol-dry ice and blocking in MAXblock medium (Active Motif) for 1 h at room temperature, spleen sections were incubated with an FITC-conjugated anti-CD41 antibody (BD) for 12 h at 4°C. Spleen sections were then washed and incubated with the DNA marker DAPI for 5 min at room temperature.

We thank M. van Lohuizen and R.A. DePinho for providing Bmi1^+/− mice and Ink4a-Arf^+/− mice, respectively, N. Nozaki, K. Helin, T. Kitamura, and K. Nakashima for the anti-Bmi1 antibody (8A9), the anti-Ezh2 antibody (AC22), the pMYs-ires-EGFP vector, and the Hmga2 cDNA, respectively, and George Wendt for critical reading of the manuscript.

This work was supported in part by Grants-in-aid for Scientific Research (#21390289) and for the Global COE Program (Global Center for Education and Research in Immune System Regulation and Treatment), MEXT, Japan, a Grant-in-aid for Core Research for Evolutional Science and Technology (CREST) from the Japan Science and Technology Corporation (JST), and grants from the Takeda Science Foundation, Astellas Foundation for Research on Metabolic Disorders, and the Tokyo Biochemical Research Foundation. H. Oguro was supported by a postdoctoral fellowship from the Japanese Society for the Promotion of Science.

The authors have no conflicting financial interests.