|Home | About | Journals | Submit | Contact Us | Français|
Leukemias and other cancers possess self-renewing stem cells that help to maintain the cancer1,2. Cancer stem cell eradication is thought to be critical for successful anti-cancer therapy. Using an acute myeloid leukemia (AML) model induced by introducing the leukemia-associated monocytic leukemia zinc finger (MOZ)-TIF2 fusion protein, we show here that AML can be cured by the ablation of leukemia stem cells. The MOZ-fusion proteins interacted with PU.1 to stimulate the expression of macrophage-colony stimulating factor receptor (M-CSFR, also called CSF1R/c-FMS/CD115). Analysis using PU.1-deficient mice demonstrated that PU.1 was essential for MOZ-TIF2 to establish and maintain AML stem cells. Cells expressing high levels of CSF1R (CSF1Rhigh cells), but not those expressing low levels of CSF1R (CSF1Rlow/− cells), showed potent leukemia-initiating activity. Using transgenic mice expressing a drug-inducible suicide gene controlled by the CSF1R promoter, AML was cured by ablation of the CSF1Rhigh cells. Induction of AML was suppressed in CSF1R-deficient mice. CSF1R inhibitors slowed the progress of MOZ-TIF2–induced leukemia. Thus, CSF1Rhigh cells contain leukemia stem cells, and the PU.1-mediated upregulation of CSF1R may be a useful therapeutic target for MOZ leukemia.
Chromosomal translocations that involve the monocytic leukemia zinc finger (MOZ) gene3 are typically associated with the FAB-M4 or -M5 subtype of human acute myeloid leukemia (AML) and often predict a poor prognosis4. While MOZ is essential for the self-renewal of hematopoietic stem cells5,6, MOZ-fusion proteins enable the transformation of non–self-renewing myeloid progenitors into leukemia stem cells7. We generated a mouse model for AML by introducing c-KIT+ bone marrow (BM) cells infected with MSCV-MOZ-TIF2-ires-EGFP retrovirus into lethally irradiated mice, as previously reported8.
To identify leukemia-initiating cells (LIC), we investigated the BM cells of AML mice for various cell surface markers by fluorescence-activated cell sorting (FACS) analysis. CSF1Rhigh and CSF1Rlow/− cells were present in the AML mouse BM (Fig. 1a) and expressed equivalent levels of MOZ-TIF2 proteins (Fig. 1b). To determine their LIC activity, CSF1Rhigh and CSF1Rlow/− cells were isolated by cell sorting, and limited numbers (10–104 cells) were transplanted into irradiated mice. One-hundred CSF1Rhigh cells were sufficient to induce AML in all mice transplanted (Fig. 1c). Conversely, no mice developed AML after 103 CSF1Rlow/− cells were transplanted per mouse, and only half of the mice developed AML with delayed onset when 104 CSF1Rlow/− cells were transplanted (Fig. 1d). Thus, the CSF1Rhigh cells displayed a > 100-fold stronger LIC activity than CSF1Rlow/− cells.
FACS analysis indicated that the CSF1Rhigh-LICs were c-Kit+ Sca-1− CD16/32+ Mac-1low Gr-1+ (Fig. S1A). Comparison of the CSF1Rhigh and CSF1Rlow/− cells indicated that Mac-1 expression was lower in CSF1Rhigh-LICs than in CSF1Rlow/− cells (Fig. 1e). However, significant differences were not observed between the CSF1Rhigh and CSF1Rlow/− cells in their cell morphology (Fig. 1f), colony-forming ability in methylcellulose medium (Fig. 1g), cell cycle distribution (Fig. S1B), or HoxA9 expression levels (Fig. S1C). To investigate if downstream pathways of CSF1R were activated, we measured the phosphorylation levels of STAT5 and ERK in CSF1Rhigh and CSF1Rlow/− cells. STAT5 was highly phosphorylated in CSF1Rhigh cells but not in CSF1Rlow/− cells (Fig. 1h), while ERK was phosphorylated in both CSF1Rhigh and CSF1Rlow/− cells.
Side population (SP) cells, a feature of some normal and malignant stem cells, were present in the BM of MOZ-TIF2-induced AML mice (Fig. S2A). While most SP cells were CSF1Rhigh cells, non-SP cells contained both M-CSFRhigh and CSF1Rlow/− cells (Fig. S2B). The LICs were ~ 10-fold more enriched in the SP fraction than in the non-SP fraction (Fig. S2, C and D). Since the SP population was very small (~ 0.12%), the population of LICs in the SP fraction was also small (~ 1% of all LICs) and most of the LICs were present in the non-SP fraction (~ 99%).
To determine if a high level of CSF1R expression also occurs in human AML cells with MOZ translocations, we investigated CSF1R expression in BM cells from an AML patient with a t(8;16) translocation expressing MOZ-CBP9. FACS analysis indicated that CSF1Rhigh and CSF1Rlow/− cells were also observed in the AML cells with t(8;16) (Fig. 1i). MOZ-CBP fusion transcripts were detected in both CSF1Rhigh and CSF1Rlow/− cells (Fig. 1j).
The above results suggest that leukemia stem cells express a high level of CSF1R, indicating that leukemia might be cured by inducing apoptosis of CSF1Rhigh cells. To test this, we used transgenic mice expressing a drug-inducible FKBP-Fas suicide gene and EGFP under control of the CSF1R promoter10 (Fig. 2a). The suicide gene products are inactive monomers under normal conditions, but can be activated by injection of the AP20187 dimerizer that induces the apoptosis of cells expressing high levels of CSF1R10. The c-KIT+ BM cells of transgenic mice were infected with the MOZ-TIF2 retrovirus, and transplanted into lethally irradiated wild-type mice. These mice developed AML ~ 2 months after transplantation, in which morphologically indistinguishable CSF1Rhigh and CSF1Rlow cells were observed and endogenous CSF1R expression was proportional to EGFP and FKBP-Fas expression levels (Figs. 2b and S3A).
Next, we transplanted the BM cells of these AML mice (105 cells/ mouse) into secondary recipient mice. Seven days after transplantation, the mice were injected with AP20187 or a control solvent, as described10. An increase in the number of CSF1Rhigh cells (Fig. 2c) and splenomegaly (Fig. 2d) were observed in the control-treated mice three weeks after transplantation. However, neither CSF1Rhigh cells nor splenomegaly was detected in the AP20187- treated mice after a one-week course of treatment with AP20187 (Figs. 2c and 2d). Although CSF1Rlow cells were observed in the BM and peripheral blood after the one-week treatment course, these cells were not detected after three months (Figs 2c and S3B). All control mice developed AML 4–6 weeks after transplantation, but none of the AP20187-treated mice died of AML within 6 months of transplantation (Fig. 2e). These results indicate that ablation of the CSF1Rhigh cells was sufficient to cure MOZ-TIF2–induced AML, and that a high expression level of CSF1R is a key element for leukemia stem cell potential. Since it has been reported that N-Myc overexpression rapidly causes AML in mice11, we also generated AML mice with MSCV-N-Myc-ires-EGFP using the BM cells of suicide gene–expressing transgenic mice as control animals. In AML mice with N-Myc, the GFP+ leukemia cells were Mac1+ Gr1+ CSF1R− blast cells (Fig. S4, A and B). Treatment of mice with AP20187 did not affect AML induction (Fig. S4C).
To investigate the role of CSFR in the development of MOZ-TIF2–induced AML, wild-type and Csf1r−/−12 mouse fetal liver cells of E16.5 littermate embryos were infected with the MOZ-TIF2 virus and transplanted into lethally irradiated mice. All mice transplanted with wild-type cells developed AML within three months. In contrast, AML induction was initially suppressed in mice transplanted with Csf1r−/− cells, but half of the mice developed AML after a longer latency period (Fig. 3a). The suppression of AML was rescued by coinfection with the MSCV-CSF1R retrovirus (Fig. 3b). STAT5, which was highly phosphorylated in CSF1Rhigh cells but not in CSF1Rlow/− cells, was phosphorylated in Csf1r+/+ cells but not in Csf1r−/− cells (Fig. S5). We also generated AML mice with MSCV-N-Myc-ires-EGFP, using Csf1r+/+ and Csf1r−/− fetal liver cells as controls. All of the mice transplanted with either Csf1r+/+ or Csf1r−/− cells expressing N-Myc developed AML (Fig. S4D). These results indicate that CSF1R is important for AML induction by MOZ-TIF2.
The above results suggest that signaling through CSF1R may be a therapeutic target for kinase inhibitors in MOZ leukemogenesis. To test this, we used the CSF1R-specific inhibitor Ki2022713 and the tyrosine kinase (including CSF1R1) inhibitor Imatinib mesylate (STI571; GlivecR)14–16. Oral administration of Ki20227 and Imatinib inhibited MOZ-TIF2-induced splenomegaly (Fig. 3C) and slowed MOZ-TIF2–induced AML onset (Fig. 3d). However, they did not affect the progress of N-MYC–induced AML (Fig. 3e).
The monocyte-specific expression of CSF1R is reportedly regulated by transcription factors such as AML1, PU.1, and C/EBP17. We previously found that MOZ interacts with AML1 and PU.1, but not with C/EBPα or C/EBPε, to stimulate transcription of their target genes5,18. Deletion analysis indicated that PU.1 interacts with the N-terminal and central regions of MOZ (Figs. 4a and S6), and that the acidic amino acid–rich region (DE region) of PU.1 was required for its high-affinity interaction with MOZ (Figs. 4a and S7A–D). While several deletions in the PU.1 protein prevent binding to N-terminal MOZ (1–513) (Fig S7C), considerable binding is retained with the full length protein (Fig S7B), suggesting there may be other PU.1-binding sites in MOZ and/or associated proteins. A pull-down assay using E. coli-produced GST-PU.1 or GST-AML1 and in vitro-produced N-terminal MOZ indicated a direct interaction between PU.1/AML1 and MOZ (Fig. S8). However, these interactions are weak in comparison to those observed in coimmunoprecipitates, suggesting that other factors may facilitate these interactions in vivo. Reporter analysis using a CSF1R promoter-luciferase construct showed that MOZ, MOZ-CBP, and MOZ-TIF2 activated the CSF1R promoter in the presence of PU.1, but not in the presence of AML1 (Fig. 4b). MOZ, MOZ-TIF2, and MOZ-CBP did not activate a CSF1R promoter mutant lacking PU.1-binding sites (Fig. 4c). These results suggest that MOZ and MOZ-fusion proteins activate CSF1R transcription in a PU.1-dependent manner. Deletion analysis indicated that the DE, Q, and ETS domains of PU.1, as well as the H15 and the central PU.1-binding domains of MOZ/MOZ-fusions, are required for the activation of CSF1R transcription (Figs. S7E and S9). The MOZ mutant lacking the C-terminal region (1–1518) failed to activate the transcription, indicating that the transcriptional activity of MOZ-TIF2/MOZ-CBP requires the TIF2 or CBP sequence Hoogenkamp et al.19 recently reported that, although chromatin reorganization of Csf1r requires prior PU.1 expression together with AML1 binding, once the full hematopoietic program is established, stable transcription factor complexes and active chromatin can be maintained without AML1 at the Csf1r locus. This might explain why AML1 was not required for the MOZ-TIF2-mediated activation of Csf1r.
To test for the expression of endogenous CSF1R, we used PU.1−/− myeloid progenitors expressing the PU.1-estrogen receptor fusion protein (PUER). Upon restoration of PU.1 activity by exposure to 4-hydroxytamoxifen (4-HT), PUER cells can differentiate into macrophages20. We infected PUER cells with MSCV-MOZ-TIF2-ires-GFP or control retroviruses, and sorted and cultured GFP+ cells in the presence of 4-HT. The results of FACS (Fig. 4d) and quantitative RT-PCR (Fig. S10) analyses indicated that CSF1R expression was induced after exposure to 4-HT, and that MOZ-TIF2 enhanced the PU.1-induced upregulation of CSF1R. Importantly, five days after exposure to 4-HT, CSF1Rhigh and CSF1Rlow cells were detected in PUER cells expressing MOZ-TIF2 but not in control cells. CSF1R expression was not induced before addition of 4-HT, even in PUER cells expressing MOZ-TIF2, indicating that functional PU.1 is required for MOZ-TIF2–induced CSF1R expression. Chromatin-immunoprecipitation (ChIP) analysis indicated that PU.1 and MOZ/MOZ-TIF2 were recruited to Csf1r in the BM cells of MOZ-TIF2-induced AML mice (Fig. S11A). In PU-ER cells expressing MOZ-TIF2, recruitment of MOZ/MOZ-TIF2 was detected after 4-HT treatment, but not before the treatment (Fig. S11B), suggesting that recruitment of MOZ/MOZ-TIF2 is dependent upon functional PU.1.
To determine if PU.1 is essential for the development of MOZ-TIF2–induced AML, PU.1+/+ and PU.1−/− fetal liver cells of E12.5 litter mates were infected with retroviruses for MOZ-TIF2 or N-Myc as a control, and were transplanted into irradiated mice. Although mice with PU.1+/+ cells expressing MOZ-TIF2 developed AML 8–14 weeks after transplantation, mice with PU.1−/− cells were quite healthy for at least 6 months (Fig. 4e). In contrast, all mice transplanted with either PU.1−/− or PU.1+/+ cells expressing N-Myc developed AML 6–10 weeks after transplantation (Figure 4f). When both PU.1 and MOZ-TIF2 were introduced into PU.1-deficient fetal liver cells, the mice developed leukemia (Fig. 4g). However, introduction of either PU.1 or MOZ-TIF2 alone was not sufficient for AML induction in mice. Thus, we conclude that PU.1 is required for the initiation of MOZ-TIF2–induced AML.
To determine if PU.1 is required for the maintenance of MOZ-TIF2-induced AML, fetal liver cells of PU.1 conditional knock-out mice (PU.1flox/flox expressing ER-Cre) were infected with MOZ-TIF2 to induce AML. The BM cells of AML mice were again transplanted into irradiated mice, and half of the mice were then treated with tamoxifene to induce PU.1 deletion (Fig. 4h). All of the control mice died of AML within 6 weeks, but none of the tamoxifene-treated mice developed AML for at least for 6 months. These results indicate that PU.1 is also required for the maintenance of MOZ-TIF2–induced AML stem cells.
Taken together, our results indicate that MOZ and its leukemia-associated fusion proteins activate the PU.1–mediated transcription of monocyte-specific Csf1r. MOZ-fusions might constitutively stimulate high Csf1r expression levels to induce AML (Fig. 4i). In contrast, we previously found that MOZ-fusions inhibited AML1-mediated activation of granulocyte-specific Mpo gene transcription18. Since MOZ-fusions are associated with monocytic leukemia, the lineage commitment may be determined by differential regulation of the target genes by MOZ fusions (i.e., upregulation of monocyte specific genes such as Csf1r and downregulation of granulocyte-specific genes such as MPO). It is also likely that normal MOZ modulates Csf1r expression to an appropriate level to regulate normal hematopoiesis (Fig. 4i), since Csf1r expression was impaired in MOZ−/− fetal liver cells (Fig. S12).
We observed that AML induction was suppressed in mice transplanted with Csf1r−/− cells, but half of these mice developed AML, albeit at a longer latency. Since MOZ-TIF2 can induce AML in the absence of Csf1r, MOZ-TIF2 can provoke rapid induction of AML not alone in a CSF1R-dependent manner, but also in a CSF1R-independent AML at longer latency. There are several possibilities for why CSF1R-independent AML cells escape CSF1R-dependency. Increased HoxA9 expression was observed in both CSF1Rhigh and CSF1Rlow/− cells (Fig. 1h). HoxA9 overexpression is reportedly not sufficient to induce AML, but requires additional mutations or oncogene activations21,22. Thus, it is possible that CSF1R-independent cells require additional mutations to escape CSF1R-dependency. Since we used a retrovirus vector to introduce MOZ-TIF2, it is also possible that oncogene activation by retroviral integration may mediate AML pathogenesis. To investigate whether downstream pathways of CSF1R were activated, we checked the STAT5 phosphorylation levels in CSF1R+/+ AML cells and CSF1R−/− AML cells, and found that STAT5 was highly phosphorylated in CSF1R+/+ AML cells but not in CSF1R−/− AML cells (Fig. S5).
In conclusion, our results indicate that PU.1-mediated upregulation of Csf1r is critical for MOZ-leukemia stem cell potential. Association of CSF1R with AML has been suggested by several previous findings. CSF1R upregulation has been reported in human23–25 and mouse26 AML, and CSF1R is also known as the protooncogene c-Fms. Multilineage hematopoietic disorders are induced by the transplantation of BM cells expressing the v-fms oncogene27. It was recently reported that CSF1R is involved in a chromosomal translocation associated with AML, which results in expression of a novel fusion of RBM6 to CSF1R28. These results suggest that CSF1R is important for not only MOZ-leukemia, but also for a subset of AMLs.
C57BL/6 mice were purchased from CREA Japan (Tokyo). NGF-FKBP-Fas transgenic mice10 (Jackson Lab.), CSF1R-deficient mice12, PU.1-null/conditional deficient mice29, and CreERT2 mice (TaconicArtemis GmbH)30 were maintained on a C57BL/6 genetic background. Mouse experiments were performed in a specific pathogen-free environment at the National Cancer Center animal facility according to institutional guidelines and with approval of the National Cancer Center Animal Ethics Committee.
MSCV-MOZ-TIF2-ires-GFP, MSCV-N-MYC-ires-GFP, MSCV-CSF1R-pgk-pac, and MSCV-PU.1-pgk-pac were generated by inserting their cDNAs into the respective vectors. The constructs were transfected with PLAT-E31 cells using the FuGENE 6 reagent (Roche Diagnostics), and supernatants containing retrovirus were collected 48 h after transfection. The c-Kit+ cells (1 × 105 cells), which were selected from BM or fetal liver cells using CD117 MicroBeads (Militenyi Biotec), were incubated with the retrovirus using RetroNectin (Takara Bio) for 24 h in StemPro-34 SFM medium (Invitrogen) containing cytokines (20 ng/mL SCF, 10 ng/mL IL-6, 10 ng/mL IL-3). The infectants were then transplanted together with BM cells (2 × 105) into lethally irradiated (9 Gy) 6- to 8-week-old C57BL/6 mice by intravenous (IV) injection. Secondary transplants were performed by intravenous injection of BM cells from the primary AML mice into sublethally irradiated (6 Gy) C57BL/6 mice.
AP20187 (gift from Ariad Pharmaceuticals; 10 mg/kg) was administered daily by IV injection for 5 d, and then 1 mg/kg AP20187 was administered every 3 d thereafter as described previously10. Mice were orally administrated Imatinib mesilate (Novoltis Pharma; 100 mg/kg), Ki2022713 (gift from KIRIN Pharma; 20 mg/kg), or solvent twice daily from 7 d after transplantation.
Bone marrow cells from AML mice were preincubated with rat IgG, and then incubated on ice with the following staining reagents: anti-CD115(CSF1R)-PE (eBioscience), anti–Mac-1 (M1/70)-PE-Cy7 (eBioscience), anti–Gr-1(RB6-8C5)-APC (BD Pharmingen), and anti-c-Kit-APC (2B8)-APC (BD Pharmingen). Flow cytometric analysis and cell sorting were performed using the cell sorter JSAN (Baybioscience), and the results were analyzed using FlowJo software (Tree Star).
CSF1R-luciferase constructs were generated by ligation of wild-type and PU.1-lacking CSF1R promoter32 with pGL4. For reporter analysis, SaOS2 cells were transfected with CSF1R-luc and phRL-CMV together with various expression constructs in 24-well plates, and luciferase activity was assayed 24 h after transfection using the microplate luminometer GLOMAX (Promega). Results of reporter assays represent the average values for relative luciferase activity generated from at least three independent experiments that were normalized using the activity of the enzyme from phRL-CMV as an internal control.
For immunoprecipitation experiments, cells were lysed in a lysis buffer containing 250 mM NaCl, 20 mM sodium phosphate (pH 7.0), 30 mM sodium pyrophosphate, 10 mM NaF, 0.1? NP-40, 5 mM DTT, 1 mM PMSF, and protease inhibitor. Cell lysates were incubated with anti-FLAG antibody-conjugated agarose beads (Sigma) and slightly rotated at 4°C overnight. The absorbed beads were washed 3 times with lysis buffer. Precipitated proteins were eluted from the beads by FLAG peptide and dissolved with the same volume of 2X SDS sample buffer. When immunoprecipitation was not performed, total protein lysates were prepared in 2X SDS sample buffer. Antibodies were detected by chemiluminescence using ECL plus Detection Reagents (Amersham Biosciences, Buckinghamshire, United Kingdom). The primary antibodies used in this study were anti-FLAG (M2) (Sigma), anti-HA (3F10) (Roche), and anti-MOZ18 antibodies.
We performed unpaired two-tailed Student's t-tests for comparisons and a log-rank test for survival data using JMP8 software (SAS Institute).
We would like to thank Dr. D. E. Zhang for the CSF1R promoter mutant lacking PU.1-binding sites, Drs. Y. Kamei and A. Iwama for MOZ-TIF2 cDNA, Dr. H. Ichikawa for N-MYC cDNA, and Dr. A. Kuchiba for help with statistical analyses. This work was supported in part by Grants-in-Aid for Scientific Research from the Ministry of Health, Labor, and Welfare and from the Ministry of Education, Culture, Sports, Science, and Technology (I.K), by the Program for Promotion of Fundamental Studies from the National Institute of Biomedical Innovation of Japan (I.K), and by NIHealth grants CA32551 and 5P30-CA13330 (E.R.S.).