Fax: + 81-11-611-2136.
Gene transfer of the CD40-ligand to human dendritic cells induces NK-mediated antitumor effects against human carcinoma cells
Article first published online: 4 JAN 2007
Copyright © 2006 Wiley-Liss, Inc.
International Journal of Cancer
Volume 120, Issue 7, pages 1491–1498, 1 April 2007
How to Cite
Tomihara, K., Kato, K., Masuta, Y., Nakamura, K., Tanaka, T., Hiratsuka, H. and Hamada, H. (2007), Gene transfer of the CD40-ligand to human dendritic cells induces NK-mediated antitumor effects against human carcinoma cells. Int. J. Cancer, 120: 1491–1498. doi: 10.1002/ijc.22518
- Issue published online: 30 JAN 2007
- Article first published online: 4 JAN 2007
- Manuscript Accepted: 26 OCT 2006
- Manuscript Received: 31 MAR 2006
- Ministry of Education, Culture, Sports, Science and Technology of Japan; Ministry of Health and Welfare of Japan
- Japan Leukemia Research Fund
- adenoviral vector;
- dendritic cells;
- NK cells;
The CD40-ligand (CD40L) is a key molecule for the activation of dendritic cells (DCs), followed by the induction of DC maturation and cytokine production. Here we found that DC infected with adenovirus vector encoding human CD40L (CD40L-DC) displayed significantly higher levels of immune accessory molecules and IL-12 production than did uninfected cells, and that CD40L-DC produced much higher levels of IFN-γ. To investigate whether CD40L-DC-derived these soluble factors could stimulate NK cells without physical cell-to-cell contact, we cocultured NK cells with CD40L-DC in transwell culture plates. NK cells showed up-regulated cytotoxic activity toward various squamous oral cell carcinoma (OSC-70, HSC-2, HSC-3), and we determined that both IL-12 and IFN-γ contributed to the CD40L-DC-mediated NK cell activation. NK cells stimulated with CD40L-DC resulted in the induction of the cell surface expression of TRAIL, the production of IFN-γ and intracellular accumulation of granzyme B. The cytotoxic activity of NK cells stimulated with CD40L-DC could be mostly inhibited by neutralizing antibody for TRAIL and completely abrogated by the combination of antibody and exocytosis inhibitor, indicating that this was mainly mediated by a TRAIL-TRAIL-receptor interaction and granule exocytosis. Moreover, CD40L-DC-activated NK cells could induce up-regulation of a death-receptor TRAIL-R2 (DR5) and down-regulation of a decoy receptor TRAIL-R3 (DcR1) on carcinoma cells. Overall, these results have revealed that CD40L-DC could activate an innate immune reaction by stimulating NK cells followed by carcinoma cells, supporting that administration of CD40L-DC may have potential as an anticancer therapy. © 2006 Wiley-Liss, Inc.
Dendritic cells (DC) are known to enhance the tumoricidal activity of natural killer (NK) cells, and it has been reported that murine DC activated murine NK cells to enhance cytotoxic activity and IFN-γ production in vitro.1 In that report, DC-induced NK cytotoxic activity and IFN-γ secretion required DC-NK cell-to-cell contact. However, the molecular mechanisms of the DC and NK cell interaction have not been fully clarified. It has also been reported that human DC induced human NK cytotoxic activity, which similarly required DC-NK cell-to-cell contact. Thus, IFN-α- or lipopolysaccharide (LPS)-pretreated human DC induced cytotoxic activity in cocultured human NK cells.2, 3, 4 More recently, Borg et al. reported that DC-mediated NK cell activation required the formation of a synapse, which led to IL-12 polarization in DC.5
In addition to the DC to NK cell contact-dependent mechanism, IL-12 or IFNs secreted from DC are also involved in initiating the activation of NK cells.
It has been reported that precursor DC have the ability to produce type1 IFNs,6 and that precursor DC activated by viral infection are induced to increase the production of type1 IFNs, which activate NK cells.7
CD40L is essential for the activation of DC, followed by the induction of cytokine production. CD40L-stimulated mature DC increase the production of IL-12, which is essential for the activation of T cells by DC.8
Several reports have shown that adenovirus-mediated CD40L gene transduction to DC induced antitumor effect following generation of tumor specific cytotoxic T cells (CTLs).9, 10 Although much evidence clearly defines an important role of CD40L-stimulated DC in eliminating tumor cells followed by the activation of CTLs, the role in the innate immune response has not been entirely clear.
In this study, we investigated whether DC transduced with the human CD40L gene (CD40L-DC) could induce NK-mediated antitumor effects and demonstrated that cell-to-cell contact was not required for the CD40L-DC-mediated NK cell activation.
Materials and methods
Human peripheral blood mononuclear cells (PBMC) were donated by healthy donor from Hokkaido Red Cross Blood Center. In brief, plastic nonadherent cells were depleted, and adherent cells were cultured in complete RPMI-1640 medium [containing 1 mM sodium pyruvate, antibiotics, and 10% fetal bovine serum (FBS)] supplemented with recombinant human GM-CSF and IL-4 (Osteogenetics, Germany) at a concentration of 50 ng/ml for 7 days.
NK cells were isolated from human PBMC of a healthy donor by using the NK CELL isolation Kit II (Miltenyi Biotec, Germany) according to the manufacturer's instructions, and cultured in RPMI-1640.
The human oral squamous-cell carcinoma cell line, OSC-70 was established from metastatic oral squamous carcinomas as described previously.11
The human oral cancer cell lines HSC-2 (JCRB0622) and HSC-3 (JCRB0623) were purchased from the Human Science Research Resources Bank (HSRRB; Osaka, Japan). All cell lines were maintained in RPMI-1640 medium containing 10% FBS.
To coculture NK cells with DC or tumor cells, we used a dual-chamber transwell plate separated by insertion of a 0.4-μm-pore size semipermeable membrane (12-well plate, Corning, NY). In all of the assays, NK cells were cocultured with DC at an NK/DC ratio of 1:1 for 24 hr.
All the recombinant adenoviral vectors used in this study were based on the E1- and E3-deleted serotype 5 adenovirus with a modified fiber F/RGD, harboring an integrin-binding RGD-motif within the HI loop of its knob protein.12 In the first step, a 14,896-bp EcoR I fragment (including the right side of the adenoviral genome) of pWEAxKM-F/RGD12 was joined with a 24,505-bp EcoR I fragment (including the left side of the adenoviral genome) of pLRI,13 generating the cosmid vector pWEAx-F/RGD.
Human CD40L cDNA was cloned by RT-PCR using total RNA extracted from activated human T cells of healthy donors. The human CD40L primer sequences were: Forward, 5′-CGG AAT TCA GCA TGA TCG AAA CAT ACA ACC AAA C-3′; Reverse, 5′-CGG GAT CCT CAG AGT TTG AGT AAG CCA AAG GAC G-3′.
The human CD40L cDNA was inserted between the EcoR I and Bgl II sites in the pCAcc vector, and the resulting plasmid was designated pCAhCD40L. Then, pCAhCD40L was digested with Cla I, and the CD40L expression unit was inserted into the Cla I site of pWEAx-F/RGD, resulting in the cosmid vector pWEAxCAhCD40L-F/RGD. To generate recombinant adenovirus, Pac I-digested cosmid was transfected into 293 cells with Lipofectamine2000 reagent (Invitrogen, Carlsbad, CA). Resulting plaques were isolated and then evaluated by restriction enzyme digestion of the viral genome. Similarly, AxCAZ3-F/RGD that carried lacZ as a reporter gene was generated. The resulting adenoviral vectors, AxCAhCD40L-F/RGD (Adv-CD40L) and AxCAZ3-F/RGD (Adv-LacZ) were amplified in 293 cells and purified by cesium chloride ultracentrifugation.14 Purified viruses were dialyzed against phosphate-buffered saline (PBS) with 10% glycerol and stored at −70°C until use. To determine the viral concentration (pt/ml), the viral solution was incubated in 0.1% sodium dodecyl sulfate (SDS) and its absorbance (A) was measured at 260 nm.15 The concentration was defined as pt/ml = A260 × (1.1 × 1012). Before use, contamination of the viral stocks with replication-competent viruses was ruled out by PCR analysis using primers specific for E1A and E1B.16
DCs were infected with adenovirus using 1,000 pt/cell to achieve nearly 100% infection efficiency. After 48 hr of adenovirus infection, DCs were evaluated for maturation markers by flow cytometry. In all of the assays, DCs were used 48 hr postinfection.
Cells were washed and then suspended in staining media [PBS, 2% FBS, 0.05% NaN3 and 1 μg/ml propidium iodide (PI)] containing saturating amounts of fluorochrome-conjugated mAbs or biotinylated mAbs, followed by phycoerythrin (PE)-conjugated streptavidin. Finally, the cells were washed with staining media and analyzed by flow cytometry using a FACS-Calibur® (Becton Dickinson, San Jose, CA). Dead cells staining with PI were excluded from the analysis. The relative expression of surface antigen is described as the mean fluorescence intensity ratio (MFIR). MFIR equals the MFI of cells stained with a fluorochrome-conjugated antigen-specific mAb divided by the MFI of cells stained with a fluorochrome-conjugated isotype control mAb. Fluorescein-conjugated mAbs specific for human CD54 were purchased from Caltag (Burlingame, CA). Fluorescein isothiocyanate (FITC)-conjugated mAb specific for human CD80, CD86, CD95, CD154 (CD40L), or HLA-DR, DP, DQ (HLA-class II) and phycoerythrin (PE)-conjugated mAb specific for CD56, NKG2D and PE-conjugated streptavidin were purchased from BD-PharMingen (San Diego, CA). Fluorescein-conjugated mAb specific for human TRAIL-R1 (DR4) or TRAIL-R2 (DR5) were purchased from Alexis (San Diego, CA). Purified mAb for human TRAIL-R3 (DcR1) or -R4 (DcR2) were kindly provided by Dr. Hideo Yagita (Juntendo University, Tokyo, Japan). FITC-conjugated mAb specific for human CD69 and biotinylated mAb specific for human TRAIL or human FasL were purchased from eBioscience (San Diego, CA). PE-conjugated mAb specific for human CD83 was purchased from Immunities (Marselle, France).
Flow cytometric analysis of intracellular perforin and granzyme B was performed as follows: Human peripheral blood NK cells cocultured with supernatants of lacZ-infected DC or CD40L-infected DC for 24 hr. Last 4 hr, the cells were cultured in the presence of the transport inhibitor GolgiStop™ (monensin, BD Pharmingen), and stained with FITC-conjugated anti-CD56 mAb and then fixed and permeabilized with Cytofix/Cytoperm™ solution for 20 min at 4°C. After washing with Perm/Wash™ solution, the cells were stained with PE-conjugated anti-perforin or anti-granzyme B mAb for 30 min at 4°C. Alternatively, the cells were stained with PE-isotype control antibody. After washing twice, NK cells expressing CD56 and cytotoxic molecules were analyzed by flow cytometry using FlowJo software.
Supernatants of DC were obtained 48 hr after infection with either Adv-LacZ or Adv-CD40L. Cytokine production was measured in the supernatants with ELISA kits specific for IL-12 (R&D Systems, Minneapolis, MN) and IFN-γ (eBioscience) according to the manufacturer's instructions.
NK cells transwell cocultured with CD40L-DC were added to tumor cells in a 96- well plate at several effector/target cell (E/T) ratios. After 4 hr, LDH release activity was evaluated by an LDH release assay kit (Promega, WI), and the mean percentage of cytotoxicity was calculated.
The mean percentage of cytotoxicity was calculated by the following formula: % cytotoxicity = [experimental − effector spontaneous − target spontaneous] × 100/[target maximum − target spontaneous].
In some experiments, the cytotoxicity assay was performed in the presence of anti-human IFN-γ antibody (Anti-IFN-γ Ab, 100 ng/ml, eBioscience), anti-human IL-12 antibody (Anti-IL-12 Ab, 100 ng/ml, eBioscience), anti-human TRAIL antibody (Anti-TRAIL Ab, 100 ng/ml, eBioscience) and concanamycin A (CMA, 100 nM, Sigma-Aldrich, MO).
The statistical significance of the data was analyzed using a Student t test. A p value of less than 0.05 was considered to be significant.
DC infected with Adv-CD40L (CD40L-DC) showed enhanced surface antigen expression and cytokine production
DC infected with Adv-CD40L (CD40L-DC) was evaluated for enhanced expression of various DC-lymphocyte costimulatory molecules by flow cytometry. In contrast to immature DC (im-DC) and DC infected with Adv-LacZ (LacZ-DC), CD40L-DC expressed high levels of CD40L, CD80, CD83, CD86 and CD54 antigens (Fig. 1a). We next examined the cytokine production in cultures of CD40L-DC by ELISA. CD40L-DC produced much higher levels of IL-12 and notably IFN-γ compared with im-DC and LacZ-DC (Fig. 1b) (n = 3, p < 0.05). Collectively, these results imply that gene transfer of CD40L by adenovirus vector efficiently induce the maturation of DC.
CD40L-DC induced NK cytotoxic activity
Previous reports indicated that activation of NK cells by mature DC treated with TNFα or LPS required cell-to-cell contact. To investigate whether CD40L-DC could induce NK cytotoxic activity without physical interactions, we examined the cytotoxic activity of NK cells cocultured with CD40L-DC in a transwell plate for 24 hr (CD40L-DC-activated NK cells). The cytotoxicity of NK cells against human various squamous-cell carcinoma cells (OSC-70, HSC-2 and HSC-3) was evaluated by the LDH-release cytotoxicity assay. As shown in Figure 2, CD40L-DC-activated NK cells displayed increased cytotoxicity against these carcinoma cells. In contrast, NK cells cocultured with LacZ-DC (LacZ-DC-activated NK cells) did exhibit a faint cytotoxic activity.
CD40L-DC-elicited NK tumoricidal activity was mediated via IFN-γ and IL-12
Our results showed that CD40L-DC produced much higher IFN-γ and IL-12 levels than did LacZ-DC (Fig. 1b) and could up-regulate NK cytotoxicity without DC-NK cell-to-cell contact (Fig. 2). These data suggested that the CD40L-DC-secreted IFN-γ or IL-12 might be responsible for the up-regulation of NK tumoricidal activity. In the presence of neutralizing antibody to either IFN-γ or IL-12, NK cells were cocultured with CD40L-DC, followed by evaluation of the cytotoxicity of NK cells against OSC-70 cells. In Figure 3a, it is shown that when NK cells were cocultured with CD40L-DC in the presence of either IFN-γ or IL-12 antibody, the cytotoxicity of NK cells was weaker than the cytotoxicity in the absence of neutralizing antibodies (n = 3, p < 0.05, for all E/T ratios). Furthermore, when NK cells were cocultured with CD40L-DC in the presence of both IFN-γ and IL-12 antibodies, the cytotoxicity of NK cells was mostly abolished (n = 3, p < 0.05, for all E/T ratios). To confirm the contribution of IFN-γ and IL-12 produced from CD40L-DC, we examined whether the addition of these cytokines to NK cells exhibit same cytotoxic activity as the coculture with CD40L-DC against OSC-70. Consistent with previous data of neutralizing antibodies, mixture of recombinant IFN-γ and IL-12 resulted in the enhancement of NK cytotoxicity (Fig. 3b), indicating that both IL-12 and IFN-γ mainly contributed to the CD40L-DC-mediated NK cell activation. However, cytotoxic activity was not completely restored at same levels of CD40L-DC-activated NK cells. Further studies are required to determine whether the unknown factors in the supernatant of CD40L-DC contribute to functional activation of NK cells.
CD40L-DC up-regulated TRAIL expression and IFN-γ production of NK cells
To investigate the mechanism of NK cell activation by these soluble factors secreted by CD40L-DC, we examined phenotypic alterations by assessing surface expressions of CD69 and CD56 on NK cells and IFN-γ production. The CD69 is an antigen that is induced on the surface of activated NK cells,17 while CD56 is present in increased amounts on activated NK cells.18, 19 In contrast to LacZ-DC-activated NK cells, CD40L-DC-activated NK cells showed increased number of CD69-positive cells and the enhanced expression levels of CD56 on their cell surface (Fig. 4 a). Next, we examined the expression of the tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL) and Fas ligand (FasL) on CD40L-DC-activated NK cells. Recently, TRAIL was shown to be expressed on NK cells and T cells,20 where it strongly affects these cells' ability to eliminate tumor cells.21, 22 FasL, which is also an inducer of cell death, has been shown to be expressed on the cell surface of effector cells.23, 24 NK cells with CD40L-DC up-regulated TRAIL expression, but not FasL expression, on their cell surface comparable with the levels found on NK cells without DC or with LacZ-DC (Fig. 4 a). To assess another indicator of NK activation by CD40L-DC, we determined IFN-γ production by CD40L-DC-activated NK cells. After cocultivation of purified NK cells with CD40L-DC in transwell, NK cells were separated from CD40L-DC and incubated 2 more days. As shown in Fig. 4 b, CD40L-DC-activated NK cells, but not LacZ-DC-activated NK cells, secreted high levels of IFN-γ (n = 3, p < 0.05).
To clarify the functional activity of TRAIL induced on the surface of NK cells, we next investigated the cytotoxic activity of CD40L-DC-activated NK cells toward oral squamous cell carcinoma in the presence of blocking antibody to TRAIL. As shown in Figure 5, the blocking antibody inhibited significantly the enhanced cytotoxicity of CD40L-DC-activated NK cells (n = 3, p < 0.05, for all target tumors and E/T ratios). In contrast to HSC-2, cytotoxic activity of CD40L-DC-activated NK cells against OSC-70 and HSC-3 was inhibited partially by an anti-TRAIL antibody alone. We further examined whether another cytolytic molecules related in NK cytotoxicity were also altered under CD40L-DC stimulation. As shown in Figure 6a, both LacZ-DC- and CD40L-DC-activated NK cells constitutively expressed NKG2D and perforin but intracellular accumulation of granzyme B was up-regulated in CD40L-DC-activated NK cells. To determine the contribution of these cytolytic molecules, we examined the effect of anti-TRAIL antibody and exocytosis inhibitor, concanamycin A (CMA) on the cytotoxic activity of CD40L-DC-activated NK cells (Fig. 6b). Cytotoxic activity against OSC-70 was inhibited significantly by the anti-TRAIL antibody or CMA alone and more profoundly by the combination of these inhibitors, indicating that the cytotoxic activity of CD40L-DC-activated NK cells toward carcinoma cells was mediated by up-regulated both TRAIL-TRAIL-receptor interaction and exocytosis of granzyme B. Overall, these findings indicate that CD40L-DC have great potential in inducing activating NK cells in anticancer immunotherapy.
DC are able to induce an antitumor immune response by generating tumor-specific CTLs.25 Most of the DC-mediated immunotherapies have been dependent on eliciting the specialized APC function in DC. In addition to function as APC, DC have been shown to regulate the innate immune response by activating NK cells, which show cytotoxic effects toward tumor cells without MHC-class I restriction.1, 2, 3, 4, 5
CD40L ordinarily exists as a 39-kDa type II transmembrane glycoprotein and subsequently cleaved from cell surface, releasing a soluble fragment, sCD40L. Although sCD40L released from activated helper T cells apparently forms homotrimers, lyophilized recombinant human CD40L protein forms primarily monomers, exhibiting only faint biological activity to stimulate CD40-positive cells, such as DC and leukemic B cells.26, 27, 28 To overcome the insufficient function of sCD40L to stimulate DC, we investigated whether adenovirus-mediated transduction of the human CD40L gene to DC (CD40L-DC) is efficient approach for the induction of mature DC that produce an activation of NK cells and a subsequent enhanced antitumor activity. NK cells after cocultivation in a transwell culture plate with CD40L-DC showed enhanced cytotoxic activity to human oral squamous cell carcinoma. This result suggested that the enhancement of NK cytotoxicity to tumor cells could be caused by soluble factors secreted by CD40L-DC, because DC are major producers of cytokines. Thus, mature DC can produce IL-12, which is important for NK activation by DC. As shown in our data, CD40L-DC produced a much higher level of IL-12 than did nonactivated DC. Many reports have shown that for the DC-mediated NK cell activation, IL-12 was important but not sufficient without DC-NK cell-to-cell contact. However, our data demonstrated that NK cells cocultured with CD40L-DC in the transwell culture plate acquired enhanced cytotoxic effect toward tumor cells without cell-to-cell contact. This result led us to believe that other contributors, such as IFNs, could be responsible for NK cell activation by CD40L-DC. It has already been reported that NK cells stimulated with IFNs (α and γ) displayed an up-regulated cytotoxic activity to tumor cells.29 Furthermore, it has been reported that DC secreting type1 IFNs could activate NK cells.6, 7 While previous reports clearly have shown immune reactions caused by type1 IFNs derived from DC, biological effects of IFN-γ derived from DC have not been proved convincingly. In this study, we have shown that CD40L-DC, in addition to produce IL-12, produce a much higher level of IFN-γ than do LacZ-DC (shown in Fig. 1b), TNFα-treated DC or sCD40L-treated DC (Tomihara, et al. unpublished data), indicating that this approach should be useful in eliciting systemic NK cell activation without cell-to-cell contact.
We also found that CD40L-DC-activated NK cells induced phenotypic alterations of tumor cells. OSC-70 carcinoma cells up-regulated HLA-class II and ICAM-1 expressions in response to cocultivation with CD40L-DC-activated NK cells without NK cell–tumor cell contact (data not shown). The expression of these antigens on tumor cells was shown to be critical for tumor cell recognition by immune effector cells including helper and cytotoxic T cells. It has been shown that a reduction or loss of expression of these antigens was observed and suggested to be related to immune escape in several types of tumor cells.30, 31 We have observed that CD40L-DC increased the expressions of these antigens on tumor cells, mediated by soluble factors (Tomihara, et al., unpublished observation). In the present study, we found that CD40L-DC-activated NK cells caused alterations in the surface phenotype of tumor cells. These results may indicate an advantage for DC- or NK-mediated tumor rejection.
In addition to the induction of HLA and adhesion antigen expression, CD40L-DC-activated NK cells increased TRAIL-R2 expression on tumor cells. TRAIL-R1 (DR4) and TRAIL-R2 (DR5) are important for induction of TRAIL-mediated cell death in tumor cells, and they are expressed only on tumor cells.32, 33, 34 In contrast to the up-regulation of death receptor on tumor cells, CD40L-DC-activated NK cells decreased the expression of TRAIL-R3 (DcR1) that have been shown to be decoy receptor to inhibit TRAIL-induced apoptosis in malignant and nonmalignant cells.35, 36 Consistent with these findings, CD40L-DC-activated NK cells significantly induced the susceptibility to recombinant TRAIL in tumor cells (data not shown). NK cells also produced significantly greater levels of IFN-γ in response to cocultivation with CD40L-DC. This result and previous reports37, 38 of the synergistic effect of IFN-γ in TRAIL-mediated apoptosis of tumor cells led us to believe that NK cells secreting IFN-γ elicited the susceptibility to TRAIL in tumor cells by modulating cellular distribution.
In this study, we have demonstrated that CD40L-DC-activated NK cells increased surface expression of TRAIL and intracellular accumulation of granzyme B. It has been reported that NK cells and CTLs increased expression of TRAIL on their cell surface in response to several cytokines.20 Furthermore, TRAIL on these cells has been shown to play an important role for an antimetastatic effect.21, 22 Hence, to further confirm that CD40L-DC-activated NK cells, which showed increased TRAIL expression on their cell surface, had enhanced cytotoxicity to tumor cells via TRAIL, we carried out the NK cytotoxicity assay in the presence of blocking antibody to TRAIL. The cytotoxicity of CD40L-DC-activated NK cells was predominantly blocked by neutralizing antibody to TRAIL and completely by the combination with CMA. The result suggested that induction of TRAIL and granzyme(s) contributed to the up-regulated cytotoxicity of activated NK cells by CD40L-DC. When these results are taken into account, our results have indicated that a very useful candidate source for functional TRAIL is inducible on the cell surface of NK cells activated by CD40L-DC. In conclusion, adenovirus-mediated gene transduction of CD40L into DC may be a powerful inducer of an antitumor immune response mediated by NK cells, and this strategy may also be applicable in cancer therapy.
We thank Hokkaido Red Cross Blood Center for providing us with human peripheral blood, and Dr. Hiroaki Uchida, Dr. Katsunori Sasaki, and Dr. Jianhua Huang for technical assistance and Dr. Tomoko Sonoda for the assistance in statistical analysis.
This work was supported by Grant-in-Aid for Scientific Research on Priority Areas “Cancer” from the Ministry of Education, Culture, Sports, Science and Technology (K. Kato, K. Nakamura and H. Hamada), by a part of Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan (K. Tomihara), by Grant-in-Aid for Cancer Research from the Ministry of Health and Welfare of Japan (K. Kato and H. Hamada) and by Grant-in-Aid from Japan Leukemia Research Fund (K. Kato).