Hypoxia-inducible factor-1α (HIF-1α) mediates adaptive responses to changes under tissue hypoxia in carcinoma cells by controlling the expression of various target genes. Previous studies have demonstrated that HIF-1α is associated with adverse clinical outcome in breast carcinoma patients, but details of HIF-1α's role have remained largely unknown. Therefore, in this study, we examined the expression profiles of HIF-1α-induced genes in 10 breast carcinoma cases using microarray data. As a result, we demonstrated that the status of hexokinase II (HKII) was associated with carcinoma recurrence in patients with these genes. The enzyme HKII is involved in the first, and rate-limiting, step of glycolysis, but its clinical significance has not yet been examined in breast carcinoma. Therefore, we immunolocalized HKII in 118 breast carcinomas, and HKII immunoreactivity was detected in 44% of the cases. It is significantly associated with histological grade, Ki-67 labeling index and HIF-1α immunoreactivity. Also, HKII status is significantly associated with increased risk of recurrence and adverse clinical outcome in breast cancer patients. Subsequent multivariate analysis demonstrated that HKII status was an independent prognostic factor for disease-free survival of patients. These results all suggest that HKII is induced by HIF-1α and plays important roles in the proliferation and/or progression of breast carcinoma possibly through increased glycolytic activity. The status of HKII is therefore considered a potent prognostic factor in human breast cancer patients.
Breast cancer is one of the most common malignancies in women. Invasive breast cancer is generally regarded as a disease that metastasizes in an early phase, and adjuvant therapy, such as endocrine therapy (e.g. tamoxifen or aromatase inhibitors) and chemotherapy, is frequently used after surgical treatment. However, parts of these carcinomas acquire clinical resistance and recur despite the therapy. Distant recurrence in patients treated with tamoxifen alone after surgery has been reported as 15% at 10 years, and results of 11 adjuvant chemotherapy trials revealed that 25% of patients who received adjuvant chemotherapy developed distant recurrence. Therefore, it is very important to examine the molecular mechanisms of clinical recurrence in breast carcinoma to improve clinical outcome of patients.
Hypoxia is known to contribute to multidrug-resistance and is associated with a poor clinical outcome for the breast cancer patients regardless of treatment modality.[3, 4] Hypoxia-inducible factor-1α is a transcription factor that mediates adaptive responses to changes under tissue hypoxic conditions in carcinoma cells, and its overexpression is significantly associated with an adverse clinical outcome in breast cancer patients.[5, 6] Therefore, HIF-1α may have important therapeutic potential, and its inhibitors, including echinomycin, have been clinically attempted. Under hypoxic conditions, HIF-1α is stabilized and translocated in the nucleus, then forms a complex with HIF-1β and binds to hypoxia-responsive elements on target genes.[8-10] At this juncture, more than 60 direct HIF-1α-target genes have been identified, and these are involved in angiogenesis, cell survival, glucose metabolism, invasion and drug resistance. Various functions of HIF-1α have been characterized by the expression patterns of these genes, but the molecular pathway of HIF-1α associated with clinical recurrence of breast cancer has remained largely unknown. Therefore, in this study, we studied the expression profiles of HIF-1α-induced genes in breast cancer tissue based on microarray data and demonstrated that HKII expression is most associated with recurrence.
Hexokinase catalyzes the conversion of glucose, resulting in the production of glucose-6-phosphase; this is the first, and rate-limiting, step in the glycolytic pathway. A glycolytic enzyme, HKII is one of four isoforms of hexokinase family denoted as HKI–IV in mammalian tissue.[11, 12] Its overexpression has been reported in several human carcinomas,[13-16] suggesting HKII has an important role in providing carcinoma cells with energy. Immunolocalization of HKII was reported in breast cancer by Brown et al., but its clinical significance has not yet been examined. Therefore, in this study, we immunolocalized HKII in human breast carcinoma tissue to clarify its clinicopathological significance.
Materials and Methods
Patients and tissues
Two sets of breast carcinoma specimens were evaluated in this study. The first set consisted of 10 specimens of invasive ductal carcinoma, not other specified, obtained from Japanese female patients (age range, 48–74 years) who underwent surgical treatment in 2001 or 2002 in the Department of Surgery, Tohoku University Hospital (Sendai, Japan). The Allred score of ER was 6–8 in these cases. All patients received endocrine therapy after surgery, and four patients also received adjuvant chemotherapy. Recurrence was evaluated based on whether first locoregional recurrence or distant metastasis was detected within the follow-up period after surgery (range, 15–82 months). These specimens were stored at −80°C for microarray analysis.
The second set of specimens consists of 118 cases of invasive ductal carcinoma, not other specified, obtained from Japanese female patients who underwent surgical treatment from 2004 to 2008 in the Department of Surgery, Tohoku University Hospital. The mean age of these patients was 57 years (range, 27–87 years). A review of the charts revealed that 91 patients received adjuvant endocrine therapy and 54 patients received adjuvant chemotherapy following surgery. The clinical outcome of the patients was evaluated by disease-free and breast cancer-specific survival in this study. Disease-free survival was defined as the time from surgery to the date of the first locoregional recurrence or first distant metastasis within the follow-up period after surgery. Breast cancer-specific survival was defined as the time from surgery to death from breast cancer. The mean follow-up period was 57 months (range, 3–84 months). All specimens had been fixed in 10% formalin and embedded in paraffin wax. As shown in Table S1, clinicopathological characteristics of the patients examined in this study were not markedly different from those previously reported.
Research protocols for this study were approved by the Ethics Committee at Tohoku University School of Medicine (2010-569).
Laser capture microdissection and microarray analysis
Gene expression profiles of breast carcinoma cells in the first set (n = 10) were examined using microarray analysis. Gene expression profile data was assembled previously.[21, 22] Briefly, approximately 5000 breast carcinoma cells were laser transferred from the frozen section, and total RNA was subsequently extracted. Sample preparation and processing were performed as described in the GeneChip Expression Analysis Manual (Affymetrix, Santa Clara, CA, USA), except that the labeled cRNA samples were hybridized to the complete human U133 GeneChip set (Affymetrix), containing U133A (22 215 genes) and U133B (22 577 genes). We focused on expression of 64 HIF-1α-induced genes reported by Semenza et al. as well as on HIF-1α in this study.
Rabbit monoclonal antibody for HKII (C64G5) were purchased from Cell Signaling Technology (Danvers, MA, USA), and other monoclonal antibodies for Ki-67 (MIB1), VEGF (JH121), MDR-1 (JSB1), CD31 (JC70A) were purchased from Dako (Carpinteria, CA, USA), Thermo Scientific (Fremont, CA, USA), Abcam (Tokyo, Japan) and Dako, respectively. Goat polyclonal antibody for HIF-1α (C19) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA). A Histofine Kit (Nichirei Biosciences, Tokyo, Japan), which employs the streptavidin-biotin amplification method, was used in this study. Antigen retrieval was performed by heating the slides in an autoclave at 120°C for 5 min in citric acid buffer (2-mM citric acid and 9-mM trisodium citrate dehydrate [pH 6.0]) or in antigen retrieval solution (pH 9.0; Nichirei Biosciences) for HKII, Ki-67 and CD31 immunostaining. For VEGF, antigen retrieval was performed in the citric acid buffer in a microwave for 15 min. Dilutions of primary antibodies used in this study were as follows: HKII, 1/100; Ki-67, 1/300; HIF-1α, 1/100; VEGF, 1/100; MDR-1, 1/40: and CD31, 1/40. The antigen–antibody complex was visualized with 3,3′-diaminobenzidine solution (1-mM 3, 3′-diaminobenzidine, 50-mM Tris–HCl buffer [pH 7.6], and 0.006% H2O2) and counterstained with hematoxylin. For negative controls in HKII immunostaining, we used normal rabbit IgG instead of the primary antibody or no secondary antibody in this study. No specific HKII immunoreactivity was detected in these sections.
Immunohistochemistry for ER (CONFIRM anti-ER [SP1]) and PR (CONFIRM anti-PR [1E2]; Roche Diagnostics Japan, Tokyo, Japan) was performed with Ventana Benchmark XT (Roche Diagnostics Japan), and that for HER2 was performed by HercepTest (Dako).
Scoring of immunoreactivity and subgroup definition of the breast carcinoma
We detected HKII immunoreactivity in the cytoplasm of breast carcinoma cells, and the cases that had more than 10% positive carcinoma cells were considered positive for HKII.[13, 16] Immunoreactivity for ER and PR was detected in the nucleus, and cases with an Allred score >3 were considered ER- and PR-positive breast carcinoma. We scored HER2 immunostaining according to the standardized HercepTest scoring system (score 0–3+; Dako), and strongly circumscribed membrane immunoreactivity of HER2 present in more than 10% of carcinoma cells (score 3+) were considered positive. Also, HER2 gene amplification was investigated by FISH in intermediate scoring (score 2+) cases, and the score 2+ cases that were positive were considered positive for HER2. Ki-67 immunoreactivity was detected in the nucleus, and Ki-67 LI (%) was determined by counting the positive cells in more than 1000 carcinoma cells at the hot spot.[23, 24] We classified HIF-1α, VEGF, and MDR-1 immunoreactivity into two groups according to previous reports.[25, 26] Briefly, the percentage of positive cells in each case was scored as follows: 0 (0%); 1 (<10%); 2 (10–50%); 3 (50–80%); 4 (>80%). Immunointensity was scored as follows: 0, negative; 1, weak; 2, moderate; and 3, strong. These scores were multiplied (range, 0–12) and then classified into negative (multiplied score 0–3) or positive (score 4–12) group.[25, 26] In each case, MVD was evaluated as the greatest number of CD31-positive microvessels per high-power field (×200).
Intrinsic subtypes of breast carcinoma were defined according to 2011 St. Gallen surrogate definition: luminal A (ER and/or PR positive, HER2 negative, Ki-67 LI <14%), luminal B (ER and/or PR positive, HER2 negative, Ki-67 LI ≥14% [HER2 negative], or ER and/or PR positive, HER2 positive [HER2 positive]), HER2 positive (ER and PR negative, HER2 positive), and triple negative (ER, PR, HER2 negative).
To evaluate HKII status and clinicopathological factors, Student's t-test or a cross-table using the χ2 test were used. Disease-free and breast cancer-specific survival curves were generated according to the Kaplan–Meier method, and statistical significance was calculated using the log-rank test. Univariate and multivariate analyses were evaluated by a proportional hazard model (Cox). P < 0.05 and 0.05 ≤ P < 0.10 were considered significant and borderline-significant in this study, respectively. The statistical analyses were performed using the JMP Pro version 9.02 (SAS Institute, Inc., Cary, NC, USA).
Expression profiles of HIF-1α-induced genes associated with recurrence of breast carcinoma patients
We first examined associations between both the expression of HIF-1α-induced genes and HIF-1α and the recurrence of the breast carcinoma in 10 cases using microarray analysis. When the expression ratio of a particular gene in the recurrence group compared to that in the non-recurrence group was >1.5 or <0.5, we tentatively determined that the gene was predominantly expressed in either the recurrence or non-recurrence group. As shown in Fig. 1(a), among 65 genes examined, the number of genes predominantly expressed in the recurrence group (n = 5) was six (9%), while that in the non-recurrence group (n = 5) was three (5%). A great majority of the genes (56 genes [86%]) had a similar expression level between the recurrence and non-recurrence groups (ratio, 1.5–0.5). Among these genes, HK2 showed the highest ratio (1.9), suggesting the possible involvement of HKII in the recurrence of breast carcinoma after surgery.
Ki-67 antibody recognizes cells in all phases of the cell cycle except G0 (resting) phase, and immunohistochemical Ki-67 LI is closely correlated with proliferative activity of breast cancer. When we classified these cases into two groups (n = 5, respectively) according the median value of the Ki-67 LI (23.5%), 12 out of 65 genes (18%) were predominantly expressed in breast carcinomas with higher Ki-67 (Fig. 1b). In addition, four out of six genes, HK2, ADRA1B, GPI and ADM, demonstrated predominant expression in the recurrence group (Fig. 1a). The list of genes and their relative expression ratios are summarized in Table 1.
Table 1. List of 65 genes with relative expression ratio
Bold indicates genes predominantly expressed in the recurrence or higher Ki-67 group (>1.5 fold). Italics indicates genes predominantly expressed in the non-recurrence or lower Ki-67 group (<0.5 fold). †Cases were classified into two groups according the median value of the Ki-67 LI. ‡Genes performed immunohistochemistry in this study. HIF-1α, hypoxia-inducible factor-1 alpha; HKII, hexokinase II; MDR-1, multidrug resistance protein 1; VEGF, vascular endothelial growth factor.
We detected HKII immunoreactivity in the cytoplasm of breast carcinoma cells (Fig. 2a), and in some case, it was marked in the invasive edge of the breast carcinoma (Fig. 2b). In this study, HKII was positive in 52 out of 118 breast carcinomas (44%) and negative in 66 cases (56%; Fig. 2c). In contrast, HKII immunoreactivity was focally and weakly detected in the epithelium of non-neoplastic mammary glands (Fig. 2d). When we immunolocalized HKII in 16 benign breast disease tissues, which included papilloma (Fig. 2e), fibroadenoma (Fig. 2f), sclerosing adenosis, and usual ductal hyperplasia (n = 4, respectively), obtained from pure benign breast disease patients, HKII immunoreactivity was focally and weakly detected in the epithelium, similar to the non-neoplastic glands.
Associations between HKII immunohistochemical status and various clinicopathological parameters in breast cancer cases are summarized in Table 2. Specifically, HKII status was positively associated with histological grade (P = 0.038), Ki-67 LI (P = 0.0084) and HIF-1α immunoreactivity (P = 0.034). It was also marginally associated with pT (P = 0.058), although this association did not reach statistical significance. No significant association was detected between HKII and other factors, such as age, menopausal status, stage, lymph node metastasis, ER status, PR status, HER2 status, intrinsic subtype, VEGF immunoreactivity, MDR-1 immunoreactivity and MVD in this study.
Table 2. Association between HKII immunohistochemical status and clinicopathological parameters in 118 breast carcinomas
Positive (n = 52)
Negative (n = 66)
P < 0.05 (bold) and 0.05 ≤ P < 0.10 (italics) are considered significant and borderline-significant, respectively. †Data are presented as mean ± SEM. All other values represent the number of cases and their percentage of all the cases (n = 118). ‡Intrinsic subtype was defined according to 2011 St. Gallen surrogate definition. ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HIF-1α, hypoxia-inducible factor-1 alpha; HKII, hexokinase II; LI, labeling index; MDR-1, multidrug resistance protein 1; MVD, microvessel density; PR, progesterone receptor; pT, pathological tumor size; VEGF, vascular endothelial growth factor.
Association between HKII status and clinical outcome of breast cancer patients
As shown in Fig. 3(a), HKII status was significantly associated with an increased incidence of recurrence (P = 0.027 by log-rank test) in the 118 breast cancer patients examined. The association between HKII status and breast cancer-specific survival is summarized in Fig. 3(b). A significant association was also detected between HKII status and adverse clinical outcome of these patients (P = 0.046 by log-rank test). When we further categorized the HKII-positive cases into two groups according to the HKII immunoreactivity in the invasive front of tumor (+/+, HKII-positive cases with marked HKII immunoreactivity in the invasive front [n = 15]; +/−, HKII-positive cases without marked HKII immunoreactivity in the invasive front [n = 37]), the +/+ group showed a significantly worse prognosis for disease-free survival than the +/− group (P = 0.0031; Fig. 3c), but not for breast cancer-specific survival (P = 0.52; Fig. 3d). A significant association between HKII status and prognosis was also detected in stage I–II cases (P = 0.022 for disease-free survival; P = 0.021 for breast cancer-specific survival; Fig. 3e,f), but this was not detected in the stage III cases (P = 0.70 for disease-free survival; P = 0.89 for breast cancer-specific survival; data not shown). The tendency between HKII status and prognosis was observed regardless of lymph node status of the cases (lymph-node negative group: P = 0.066 for disease-free survival, P = 0.10 for breast cancer-specific survival; lymph-node positive group: P = 0.29 for disease-free survival, P = 0.27 for breast cancer-specific survival; data not shown). The association between HKII status and disease-free and breast cancer-specific survival according to the intrinsic subtypes of tumor was as follows: luminal A, P = 0.16 (disease-free); luminal B: P = 0.17 (disease-free, Fig. 3g) and P = 0.073 (breast cancer-specific, Fig. 3h); and triple negative, P =0.90 (disease-free) and P =0.76 (breast cancer-specific). For breast cancer-specific survival among patients with luminal A subtype, the P-value was not calculated because no patients died. Among the HER2 positive subtype, P values not calculated because no patients recurred or died in this group.
A similar tendency was also detected in the group that received adjuvant endocrine therapy (n = 91; P = 0.046 for disease-free survival; Fig. 3i) and P = 0.082 for breast cancer-specific survival (Fig. 3j) or adjuvant chemotherapy (n = 54: P = 0.29 for disease-free survival; P = 0.079 for breast cancer-specific survival; Fig. 3k,l).
Univariate analysis of disease-free survival by Cox (Table 3), Ki-67 LI, pT, lymph node metastasis, histological grade, MVD and HKII were demonstrated to be significant prognostic parameters for disease-free survival in 118 breast carcinoma patients. A multivariate analysis revealed that Ki-67 LI (P = 0.025), pT (P = 0.044) and HKII (P = 0.031) were independent prognostic factors with relative risks over 1.0 (Table 3). However, multivariate analysis of breast cancer-specific survival revealed no independent prognostic factor with relative risk over 1.0 in this study (Table 4). In the stage I–II cases (n = 99), pT (P = 0.025), HKII (P = 0.0091) and MDR-1 (P =0.041) were significant prognostic parameters for breast cancer-specific survival according to univariate analyses. Multivariate analysis revealed HKII (P = 0.0029) and MDR-1 (P = 0.013) as the independent parameters (data not shown).
Table 3. Univariate and multivariate analyses of disease-free survival in 118 breast cancer patients
Relative risk (95%CI)
Data considered significant (P <0.05) are in bold. †Data were evaluated as continuous variables. All other data were evaluated as dichotomized variables. ‡Significant (P <0.05) and borderline-significant (0.05 ≤ P <0.10) values were examined in the multivariate analyses in this study. 95%CI, 95% confidence interval; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HIF-1α, hypoxia-inducible factor-1 alpha; HKII, hexokinase II; LI, labeling index; MDR-1, multidrug resistance protein 1; MVD, microvessel density; pT, pathological tumor size; VEGF, vascular endothelial growth factor.
Table 4. Univariate and multivariate analyses of breast cancer-specific survival in 118 breast cancer patients
Relative risk (95%CI)
Data considered significant (P <0.05) are in bold. †Data were evaluated as continuous variables. All other data were evaluated as dichotomized variables. ‡Significant (P < 0.05) and borderline-significant (0.05 ≤ P < 0.10) values were examined in the multivariate analyses in this study. 95%CI, 95% confidence interval; ER, estrogen receptor; HER2, human epidermal growth factor receptor 2; HIF-1α, hypoxia-inducible factor-1 alpha; HKII, hexokinase II; LI, labeling index; MDR-1, multidrug resistance protein 1; MVD, microvessel density; pT, pathological tumor size; VEGF, vascular endothelial growth factor.
Gene expression profiling is an important method for predicting the likelihood of recurrence of disease in breast cancer patients. Results of our present microarray analysis revealed that HIF-1α and five HIF-1α-induced genes were both potentially associated with the recurrence of breast carcinoma. Among these, adrenoceptor has been reported to be induced by cell migration, and α1b-adrenoceptor (gene symbol: ADRA1B) expression was associated with recurrence and poor prognosis. In addition, phosphoglucose isomerase/autocrine motility factor (gene symbol: GPI) induces epithelial-to-mesenchymal transition and has been correlated significantly with the progression and poor prognosis of breast cancer. Although HK2 demonstrated the highest expression ratio in this study, the clinical significance of HKII has not been elucidated in the breast cancer patients. In contrast, the expression level of a great majority of HIF-1α-induced genes (56 out of 64 genes; 86%) was not significantly changed between recurrence and non-recurrence groups in this study. Considering that HIF-1α has various biological functions such as angiogenesis, glucose metabolism and drug resistance, these genes may play more important roles in other functions than in recurrence specifically.
In our present study, HKII immunoreactivity was detected in 44% of breast carcinomas, and it was almost negligible in the normal or benign breast tissues. Previously, HK activity has been higher in malignant tissues than in benign tumors or normal tissues, and HKII immunoreactivity was detected in 17–21% of gastric carcinomas and 45–56% of hepatocellular carcinomas.[13-16] Moreover, Brown et al. reported that HKII was positive in 19 out of 24 (79%) breast carcinomas. Our present results suggest that HKII immunoreactivity is increased in a subset of breast carcinoma compared to normal or benign breast tissue, and the relatively wide distribution of HKII immunoreactivity also suggests biological importance of HKII in human breast carcinomas.
In our study, HKII immunoreactivity was significantly associated with Ki-67 LI, which reflected proliferative activity of the breast carcinoma. In addition, HKII immunoreactivity was sometimes marked in the invasive tumor front, which is widely considered the most biologically active part of the carcinoma and which may drive disease outcome. The involvement of HKII in the rate-limiting step of glycolysis provides cells with energy. The glycolytic pathway also provides precursors for biomolecules necessary for proliferation,[37, 38] and increased glycolytic character provides tumor protection and enhances invasion. Carcinoma cells have high glycolytic activity under conditions suitable for oxidative phosphorylation, which confers tumor cells with a survival advantage. Therefore, HKII is considered to play an important role in the cell proliferation of breast carcinoma, possibly by increased glycolytic activity.
In this study, a positive association was detected between HKII and HIF-1α immunoreactivities. Under hypoxic conditions, transcription factor HIF-1α binds to the upstream region of HKII to activate gene expression, and colocalization of HIF-1α and HK-II has been also reported in hepatocellular and gastric carcinoma tissues.[14, 42] Therefore, HKII expression may be regulated by HIF-1α in breast carcinoma. In contrast, we also detected HKII immunoreactivity in 19 out of 56 HIF-1α-negative breast carcinoma cases. These data suggest that HKII expression may be upregulated by a low or undetectable protein level of HIF-1α in these cases, but regulation of HKII expression by androgens, peroxisome proliferator activated receptor γ and microRNA-143 have been also reported.[44, 45] Therefore, factors other than HIF-1α may be involved in the expression of HKII in some breast carcinomas.
Results of our present study also demonstrate that HKII status was significantly associated with recurrence and poor prognosis of breast carcinoma, and a similar tendency was also detected in patients who had received adjuvant therapy. In addition, results of multivariate analysis showed that HKII status was indeed an independent prognostic factor for recurrence. Previously, Palmieri et al. reported that HKII overexpression was detected in 77% of brain metastasis tissues of breast carcinoma. Nakano et al. showed that a HKII inhibitor (3-bromopyruvate) diminished ATP-binding cassette transporter activity to restore drug retention in multiple myeloma cells, and Farabegoli et al. very recently demonstrated that MCF-Tam (MCF-7 breast carcinoma cells with acquired tamoxifen resistance) exhibited higher glycolytic activity than the parental MCF-7 cells. Therefore, residual carcinoma cells following surgical treatment in HKII-positive breast carcinomas may still have the potential to grow and/or metastasize more rapidly than HKII-negative cases, despite adjuvant therapy. However, mean follow-up period was 57 months in this study, and replication studies with a larger sample set with a longer-follow up period are needed to confirm the significance of HKII in the breast carcinoma.
In our present study, lymph node metastasis was a significant prognostic parameter for both disease-free and breast cancer-specific survival by univariate analyses, as expected, but it was not an independent prognostic factor according to multivariate analyses (Tables 3, 4). Considering that patient characteristics examined in this study were not so different from those in a previous report (Table S1), the prognostic power of lymph node metastasis may partially overlap with other factors used in the multivariate analysis, including HKII in the breast carcinoma.
In summary, we examined expression profiles of HIF-1α-induced genes using microarray analysis and demonstrated that HKII expression was most closely associated with breast cancer recurrence. A subsequent immunohistochemical analysis revealed that HKII immunoreactivity was detected in 44% of breast cancer cases and was significantly associated with histological grade, Ki-67 LI and HIF-1α immunoreactivity. In addition, multivariate analysis demonstrated that HKII status is an independent prognostic factor for disease-free survival. These findings suggest that HKII is, at least in a part, induced by HIF-1α and plays important roles in the proliferation and/or progression of breast carcinoma, possibly by increasing the glycolytic activity.
We appreciate the skillful technical assistance of Ms. Yayoi Takahashi, MT (Department of Pathology, Tohoku University Hospital, Sendai, Japan). This work was partly supported by Grant-in-Aid for Scientific Research (50400312) from the Japanese Ministry of Education, Culture, Sports, Science and Technology (Tokyo, Japan).