The Neuronal Differentiation Factor NeuroD1 Downregulates the Neuronal Repellent Factor Slit2 Expression and Promotes Cell Motility and Tumor Formation of Neuroblastoma

NeuroD1 is a basic helix-loop-helix (bHLH) transcription factor. NeuroD1-deficient mice die as a consequence of severe hyperglycemia within 5 days of birth (1). This is attributed to differentiation deficit and death of β cells in the pancreas during embryogenesis (1). In addition, NeuroD1 is involved in neurogenesis. It can convert embryonic epidermal cells into fully differentiated neurons in Xenopus and promotes premature cell-cycle exit and differentiation of neural precursor cells (2). By crossing NeuroD1-deficient mice with either transgenic mice expressing NeuroD1 in β cells or (mammalian atonal homolog) MATH-deficient mice, Miyata and colleagues as well as Schwab and colleagues demonstrated that NeuroD1 deletion in the central nervous system leads to depletion of the cerebellar granule cells and loss of the dentate gyrus in the hippocampus (3, 4). Furthermore, NeuroD1 is involved in the differentiation of peripheral neurons. Thus, NeuroD1-deficient mice exhibit neuronal cell loss in the inner ear (5, 6). In the retina of NeuroD1-deficient mice, amacrine cell differentiation is perturbed and the number of bipolar interneurons decreases (7). Furthermore, it has recently been reported that NeuroD1 plays an essential role in the maintenance of neural precursor cells in adult neurogenesis: inducible stem cell–specific deletion of NeuroD1 causes fewer newborn neurons in the hippocampus and the olfactory bulb (8, 9).

Neuroblastoma (NB) is the most common extracranial pediatric solid tumor and is derived from the sympathetic neuron lineage of neural crest cells (10). A hierarchical expression of regulatory molecules, such as Sox2, BMP, Mash1, Phox2B, and Hand2, determines the fate of neural crest cells and their further differentiation to form sympathetic neurons (11). The molecular mechanism of the normal development of sympathetic neurons may influence the biology of NB. For example, the homeodomain DNA-binding transcription factor Phox2B is mutated in familial and sporadic forms of NB (12). Phox2B mutations predispose to NB by promoting the proliferation and dedifferentiation of cells in the sympathoadrenergic lineage (13). Phox2B induction in a NB cell line downregulates expression of the homeobox protein Msx1, whereas Msx1 upregulates expression of Notch3 and Hey1 and downregulates NeuroD1 expression (14). The bHLH transcription factor Hand2, the proto-oncogene MYCN, and tyrosine hydroxylase, which is essential for catecholamine metabolism, are expressed in proliferating precursors of sympathetic neurons (15). MYCN gene amplification is often observed in NB, and it is the strongest prognostic factor of NB (16). Furthermore, the bHLH transcription factor Mash1-deficient mice lose sympathetic neurons (17). Mash1 is expressed in multipotent neuron progenitors immediately after expression of neural stem cells during postnatal hippocampus granule neurogenesis, whereas NeuroD1 is expressed in more differentiated progenitors (18). Mash1 is expressed mostly in poorly differentiated neuroendocrine carcinoma, whereas NeuroD1 is expressed in well-differentiated neuroendocrine carcinoma (19).

Weiss and colleagues generated MYCN transgenic (Tg) mice, in which MYCN expression is targeted to the sympathetic neuron lineage by the rat tyrosine hydroxylase (20). These mice develop NB in the celiac sympathetic ganglion, which has histology indistinguishable from that of human NB (20). In particular, hemizygous mice develop tumors with a stepwise increase of the frequency of MYCN amplification (21). Therefore, this model mimics human NB development. Moreover, this model demonstrates a marked increase of neuroblast hyperplasia at approximately 2 weeks after birth that is hypothesized to represent a precancerous lesion (21).

In spite of the accumulating evidence showing the critical roles of NeuroD1 in neurogenesis, its involvement in cancer development has not been fully explored. As NB tumorigenesis is closely related to the normal development of the sympathetic neuron lineage, we hypothesized that NeuroD1 could be an important regulator of NB. To examine this hypothesis, we used MYCN hemizygous mice and found that NeuroD1 plays a critical role in NB tumorigenesis.

Thermal conditions for PCR included 28–35 cycles of denaturing at 94°C for 30 seconds, annealing at 55–56°C for 30 seconds, and elongation at 72°C for 30 seconds. The primers used and the expected product sizes are listed in Supplementary Table SI. Detection of NeuroD1 and Slit2 was performed by using the TaqMan Gene Expression Assays kit (Applied Biosystems) according to the manufacturer's instructions. The assay ID for NeuroD1 is Hs00159598_m1 and that for Slit2 is Hs00191193_m1.

Nontargeting short hairpin RNA (shRNA; Sigma) and specific shRNA for human NeuroD1 (Sigma), mouse NeuroD1 (Open Biosystems), and human Slit2 (Open Biosystems) were used. The shRNA sequences are listed in Supplementary Table SII. Together with the lentivirus packaging plasmids psPAX2 (Addgene plasmid 12260) and pMD2.G (Addgene plasmid 12259), all plasmids were cotransfected into HEK293T cells to produce infectious virus particles. The titer was evaluated by using a QuickTiter Lentivirus Titer kit (Cell Biolabs). Approximately 2.5 × 10⁹/mL virus particles were obtained. NB cells were infected in the presence of 8 μg/mL polybrene (Sigma). For the infection of tumor sphere cells, virus particles were concentrated by PEG-it Virus Precipitation Solution (System Biosciences) and suspended in a sphere growth medium without polybrene.

A human NeuroD1 cDNA clone was amplified from human HEK293 cell cDNA with attB-flanked primers. For shRNA-resistant expression constructs, the shRNA-targeting site on the cDNA was mutated at 6 nucleotides without affecting the amino acid sequence. Thereafter, following the instructions of Invitrogen Gateway Technology and using the ClonaseII Kit, the mutated NeuroD1 cDNA was inserted into the gateway destination vector CSII-CMV-RfA-IRES2-Venus (RIKEN BioResource Center, Ibaraki, Japan), which is a lentivirus-based cDNA over-expression vector. CSII-CMV-Venus (RIKEN) was used as the negative control.

Tissues were fixed in 4% paraformaldehyde, dehydrated, and embedded in paraffin according to the standard procedure. Then, 5-μm sections were stained with hematoxylin and eosin stain or incubated with anti-NeuroD1 polyclonal antibody (Cell Signaling Technology), which was diluted at a concentration of 1:50 in 3% normal goat serum for 2 hours at room temperature, and then incubated with biotin goat anti-rabbit secondary antibody (1:50; BD Pharmingen) for 30 minutes at room temperature. For Western blot analysis, anti-NeuroD1 (1:500; Santa Cruz Biotechnology), anti-Slit2 (1:500; Millipore), and anti-β-actin (1:10,000; Sigma) antibodies were used.

SH-SY5Y cells were cross-linked by 1% formaldehyde, and the reaction was quenched with 0.125 mol/L glycine. Cells were harvested by using PBS containing phenylmethylsulfonylfluoride (PMSF) followed by homogenization with hypotonic buffer. Then, DNA was digested to 150 to 900-bp fragments with micrococcal nuclease (New England Biolabs). The pellet obtained was resuspended in 300 μL of lysis buffer containing propidium iodide and sonicated. The lysates were centrifuged, and the supernatants obtained were diluted with dilution buffer. Furthermore, these lysates were incubated with control Immunoglobulin G1 (IgG1) antibody (R&D Systems) or anti-NeuroD1 (Santa Cruz Biotechnology) accompanied by rotation for 4 hours at 4°C this was followed by the addition of 15 μL of protein G Sepharose (GE Healthcare) and rotation for 1 hour at 4°C. The precipitants obtained were washed, eluted with elution buffer, and subjected to PCR. The chromatin immunoprecipitation (ChIP) primer sequences are listed in Supplementary Table SI.

A total of 2 × 10⁵ cells were plated in a 6-well plate, and, 24 hours later, infected with virus. For the wound-healing assay, another 48 hours later, a scratch on the confluent monolayers of cells was made using a plastic pipette tip. The cell-migration area was calculated by using Metamorph 6.1 software (Molecular Devices), and it was represented as the change in length (μm). For the Boyden chamber assay, cells were harvested 48 hours after virus infection and counted. Transwell plates (pore size: 8 μm; Corning) were used. The undersurface of the chamber was coated with 10 μg/mL collagen I (Upstate Biotechnology). Virus-infected cells were seeded in the upper chamber (4 × 10⁴ per well) with 1% bovine serum albumin. Thereafter, 10% FBS was added to the lower chamber. Cells were allowed to migrate for 6 hours. Nonmigrated cells in the upper chamber were removed by using a swab. Migrated cells attached to the lower side of chamber were stained with 4′,6-diamidino-2-phenylindole and counted in 4 randomly selected fields. Then, 4 fields at 100 times the magnitude per sample were selected. The experiment was performed twice, with each sample in triplicate.

The MYCN Tg mice (20) were maintained in our animal facility, where they were housed under a controlled environment and provided with standard nourishment and water. Tumor tissues from hemizygous MYCN Tg mice were dissected, minced, and then inoculated subcutaneously into 1-month-old 129/SVJ wild-type (WT) mice (the mouse strain of MYCN Tg mice). After approximately 15 such serial passages, a more invasive tumor was obtained. This constituted our allograft model in the present study. This study was approved by the Animal Care and Use Committee of Nagoya University Graduate School of Medicine, Nagoya, Japan.

A 0.5-cm³ tumor tissue was dissected and washed several times with Hanks' balanced salt solution (HBSS). After thorough mincing, cells were digested with 0.25% trypsin (5 mL) at 37°C for 10 minutes. The same volume of trypsin inhibitor (Sigma) was used to stop the digestion. After centrifuging and washing with HBSS, the tissue pellet was mixed with 5 mL HBSS and allowed to react for 10 minutes. Thereafter, the supernatant was carefully transferred to a new tube and centrifuged. The cell pellet was suspended in a growth medium for stem cells [Dulbecco's modified Eagle Medium (DMEM):F12K 1:1, epidermal growth factor 10 ng/mL, basic fibroblast growth factor 15 ng/mL) and seeded onto a 10-cm petri dish. The medium was changed 24 hours later to remove dead cells. Approximately 2 days later, tumor spheres were recognized. These tumor spheres were digested with trypsin and broken down into single cells and were passed through fresh culture every 3–4 days. For the allograft tumorigenesis experiment, 1 × 10⁴ shRNA-treated cells mixed with 30% Matrigel (BD Biosciences) were subcutaneously inoculated into 1-month-old 129/SVJ WT mice. Eighteen days later, the tumors were dissected and weighed.

SH-SY5Y cells were obtained from the American Type Culture Collection. IMR32 cells were procured from RIKEN. We monitored the morphology of these cells by microscopic examination and confirmed that their morphology was the same as that of the original cells. In addition to cell growth and morphology, we also confirmed that IMR32 cells strongly expressed MYCN whereas SH-SY5Y cells did not. No further validation was performed during these experiments. Stable NeuroD1-knockdown SH-SY5Y cells were maintained in DMEM with 10% FBS and 1 μg/mL puromycin (InvivoGen).

Results are expressed as the mean ± SD. Statistical significance was evaluated by Student's t test.

We examined the NeuroD1 expression profile during the tumorigenesis of NB in MYCN Tg mice. As previously reported (21), we found that neuroblast hyperplasia occurred in the celiac sympathetic ganglions approximately 2 weeks after birth in MYCN Tg mice but not in WT mice [Fig. 1A (a and b)]. There were several loci of neuroblast hyperplasia in a ganglion [Fig. 1A (b)]. Because the incidence of hyperplasia correlates with the incidence of tumor formation in MYCN Tg mice, neuroblast hyperplasia is considered to be the precancerous status of NB (21). We found that ganglion cells of WT mice, at 2 weeks after birth, showed only low-level expression of NeuroD1 [Fig. 1A (c)]. Surprisingly, NeuroD1 was strongly expressed in the hyperplastic loci in MYCN Tg mice [Fig. 1A (d), arrowheads], whereas only low-level expression of NeuroD1 was detected in the surrounding ganglion cells [Fig. 1A (d)]. The expression was strong in the majority of hyperplastic cells but not necessarily in all cells [Fig. 1A (d)]. We confirmed the presence of NeuroD1 mRNA expression in the hyperplastic ganglions of MYCN Tg mice by using RT-PCR, but the expression in WT ganglion was below the detection threshold (Fig. 1B). Furthermore, we found that there was continued NeuroD1 expression in advanced cancerous tissues of NB derived from MYCN Tg mice (Fig. 1C). Interestingly, NeuroD1 expression was not homogenous, but rather a portion of the total cancer cells strongly expressed it (Fig. 1C).

Hyperplasia in WT mice usually regresses approximately 2 weeks after birth (21). Therefore, we subsequently questioned whether hyperplasia in WT mice at an early age could also express NeuroD1. We found neuroblast hyperplastic regions in the celiac sympathetic ganglions at day 0 after birth in both WT and MYCN Tg mice, as previously reported [Fig. 1D (a and b); ref. 21]. While there was low-level expression of NeuroD1 in immature ganglion cells in both genotypes [Fig. 1D (a and b, white arrowheads], NeuroD1 expression was barely detectable in neuroblast hyperplastic regions at this age [Fig. 1D (a and b), black arrowheads]. These results indicate that the hyperplasia at 2 weeks in MYCN Tg mice is different from that at day 0 in both WT and MYCN Tg mice. Thus, NeuroD1 expression is switched on in hyperplastic regions in MYCN Tg mice during the first 2 weeks after birth.

The in vivo expression profile suggests a possible involvement of NeuroD1 in NB tumorigenesis or development. To examine this hypothesis, we investigated the role of NeuroD1 in NB cells. In the MYCN amplified cell line IMR32, NeuroD1 expression was significantly suppressed by 2 shRNAs acting against human NeuroD1 that were transiently expressed via a lentivirus expression system (Fig. 2A). Although there was no difference in cell growth between the control shRNA and NeuroD1 shRNA-treated cells, we found that NeuroD1 shRNAs inhibited cell motility in a scratch assay (Fig. 2A). Furthermore, we found that SH-SY5Y cells, a human NB cell line without MYCN amplification, showed a high level of NeuroD1 expression (Fig. 2B; Supplementary Fig. S1). In addition, NeuroD1 shRNA effectively suppressed NeuroD1 expression (Fig. 2B) and inhibited cell motility in a scratch assay [Fig. 2C (a and b)]. In support of these data, we report that a Boyden chamber cell migration assay showed NeuroD1 shRNA-mediated inhibition [Fig. 2C (c)].

Next, we established cells in which NeuroD1 shRNA or control shRNA was stably expressed, such that we could perform rescue experiments and analyze the underlying molecular mechanism. The established NeuroD1-knockdown cells showed a significantly suppressed expression of NeuroD1 (Fig. 3A). The NeuroD1 shRNA-treated cells became round, which was strikingly different from that of the shape of parent cells (Fig. 3B). However, data obtained with regard to cell growth and cell-cycle analyses showed that cell growth was comparable between the control shRNA cells and NeuroD1 shRNA-treated cells (Supplementary Fig. S2). As expected, a scratch assay revealed that NeuroD1 shRNA-treated cells showed less motility (Fig. 3C). This inhibition was observed up to 72 hours after scratch (Supplementary Fig. S3). This phenomenon was attributed to the knockdown of NeuroD1 expression, because the shRNA-resistant NeuroD1 expression reversed the NeuroD1 shRNA-induced suppression of cell motility (Fig. 3D).

What is the underlying mechanism? Coincidentally, we found that Slit2 mRNA expression was induced by knockdown of NeuroD1 expression in both IMR32 cells and SH-SY5Y cells (Fig. 4A). Consistent with this finding, SH-SY5Y cells stably expressing NeuroD1 shRNA were found to express the Slit2 protein at a much higher level than cells expressing the control shRNA (Fig. 4B). It is of importance that shRNA-resistant NeuroD1 expression in the NeuroD1 shRNA-treated cells significantly suppressed Slit2 protein expression (Fig. 4B).

It is known that NeuroD1 is a transcription factor that binds to the E-box in the regulatory region of target genes (22). Therefore, in the next step, we explored whether NeuroD1 was recruited to the regulatory region of the human Slit2 gene. There are 6 E-box stretches surrounding the Slit2 gene (Fig. 4C). A ChIP assay revealed that NeuroD1 directly bound to the first and second E-box but not to the other E-boxes (Fig. 4D).

NeuroD1 knockdown led to the inhibition of cell motility and an increase in Slit2 expression and therefore Slit2 might possibly be involved in cell motility. To test this hypothesis, we knocked down Slit2 expression in the cells expressing NeuroD1 shRNA. shRNAs functioning against human Slit2 efficiently suppressed Slit2 expression in NeuroD1 shRNA-treated cells (Fig. 5A). These double-knockdown cells restored cell motility (Fig. 5B and C), and this supports the hypothesis that NeuroD1 promotes cell motility by suppressing Slit2 expression. When we knocked down only Slit2 expression in SH-SY5Y cells, cell motility was enhanced (Supplementary Fig. S4).

To further examine the functional relevance of NeuroD1 to tumorigenesis, we knocked down NeuroD1 expression in tumor spheres established from NB tumor tissue derived from MYCN Tg mice. We examined 2 shRNAs against mouse NeuroD1. They efficiently suppressed NeuroD1 expression in tumor spheres after infection via a lentiviral system (Fig. 6A). Control shRNA-expressing tumor spheres showed a tightly packed shape and thus the interface between cells was nearly invisible (Fig. 6A). In contrast, cell–cell contact in the NeuroD1-knockdown tumor spheres appeared loose (Fig. 6A). Furthermore, NeuroD1 shRNAs strongly suppressed the growth of tumor spheres (Fig. 6B). Effects of NeuroD1 on sphere self-renewal ability are shown in Figure 6C. Primary and secondary sphere-formation ability was impaired after NeuroD1 knockdown when evaluated by sphere number (Fig. 6C). In addition, NeuroD1 knockdown spheres were smaller (Fig. 6C).

The NeuroD1 functions, that is, cell motility and tumor sphere growth, revealed during this study further support the idea that NeuroD1 might be involved in NB tumor development. To examine this idea, we injected 1 × 10⁴ cells from tumor spheres, which had been infected with lentivirus expressing either control shRNA or NeuroD1 shRNA, into 129/SVJ WT mice. Because control shRNA-treated mice died from their tumors at approximately 4 weeks, the deadline was set at 3 weeks in this experiment. We found that tumor sphere cells expressing NeuroD1 shRNA showed significantly less tumorigenic ability (Fig. 6D).

Finally, we considered whether the data obtained were relevant to the clinical features of human NB. To examine this, we utilized the data of the AMC cohort study (23) and analyzed the overall and relapse-free survival probabilities with regard to the expression levels of NeuroD1 (2 probes). As shown in Figure 7A and B, high NeuroD1 expression was closely associated with poor prognosis of human NB in both of the probes tested.

Furthermore, we stratified the samples by MYCN amplification and age because these are important prognosis factors for NB. With regard to the relationship between NeuroD1 expression and MYCN status on survival probability of patients, we could not draw a conclusion, probably because of the limited number of patients with amplified MYCN (data not shown). On the contrary, when we stratified the samples by age, we obtained a significant effect of the NeuroD1 expression on survival of patients older than 1 year with one probe (206282_at; Fig. 7C). For another probe (1556057_s_at), we observed the same tendency as noted for the first probe, although larger and more balanced samples are needed to obtain firm statistical results: the P values were 0.041, in overall survival probability, and 0.063, in relapse-free survival probability (Fig. 7D).

In this study, we have demonstrated that NeuroD1 is highly expressed in the neuroblast hyperplastic region, which is hypothesized to be a precancerous lesion of NB, suggesting that NeuroD1 is involved in NB tumorigenesis. Evidence in support of this hypothesis includes the fact that NeuroD1 knockdown led to suppression of NB cell motility and NB tumor sphere growth. Consequently, NeuroD1 knockdown suppressed in vivo tumor growth derived from tumor sphere cells. Consistent with these data, high expression of NeuroD1 was closely associated with poor prognosis of NB patients. Taken together, our data suggest that NeuroD1 plays a critical role in NB tumorigenesis.

Our results concur with the report by Revet and colleagues that inhibition of cell proliferation and anchorage-independent growth of a NB cell line by the Phox2B downstream target Msx1 is accompanied by downregulation of NeuroD1 (14). However, they found no significant difference in NeuroD1 expressions in NB and ganglioneuroma (differentiated benign tumor); this finding does not necessarily fit in with our results. Therefore, further studies are needed for understanding the roles of NeuroD1, particularly in terms of neural differentiation and NB tumorigenesis. In this context, it may be important to discuss the stemness of the neural progenitors and NB cells. NeuroD1 is expressed in proliferating neuronal progenitor cells not only during embryogenesis but also during adult neurogenesis. Its loss leads to a reduction in the number of differentiated neurons, such as cerebellar granule cells, the dentate gyrus cells, inner ear sensory neurons, amacrine cells, and newborn neurons in the adult hippocampus and olfactory bulb (4–10). In adult neurogenesis, the Wnt pathway activates β-catenin, which accumulates in the nucleus where it forms an activating complex with T-cell factor/lymphoid enhancer factor (TCF/LEF). The complex binds to the Sox/LEF element located in the promoter region of the NeuroD1 gene and activates the transcription of NeuroD1. Thereafter, the neurogenesis process begins in the neural progenitor cells (9, 10). NeuroD1 may preserve the stemness of the neural progenitor cells, that is, it may promote self-renewal and differentiation. Furthermore, our findings showing its involvement in NB tumorigenesis support this idea. Interestingly, NeuroD1 was heterogeneously expressed, that is, it was not expressed in all the cells, and it was expressed in hyperplastic regions of 2-week-old MYCN Tg mice and NB tumors of MYCN Tg mice. Furthermore, NeuroD1 knockdown led to growth suppression of tumor spheres and in vivo tumors derived from tumor sphere cells. Considering that tumor-initiating cells are enriched in tumor spheres (24), it is conceivable that NeuroD1 is involved in the stemness property of NB cells. In line with this idea, NeuroD1 expression was induced in hyperplatic regions of MYCN Tg mice during the first 2 weeks after birth, thus suggesting that dynamic molecular events occur in these regions and lead to NB development.

Slit proteins (Slit1–Slit3) are secreted as extracellular matrix-associated glycoproteins and function as ligands for the repulsive guidance receptor Roundabout (Robo1–Robo4) family. Slit2 is known to bind with the Robo 1 receptor and regulates the migration of neurons during development (25). In the present study, we found that NeuroD1 downregulated Slit2 expression and promoted cell motility whereas NeuroD1 knockdown upregulated Slit2 expression and inhibited cell motility. This is the first study to illustrate an inverse relationship between NeuroD1 and Slit2. Significantly, Slit2 knockdown in NeuroD1-knockdown cells restored cell motility. Therefore, Slit2 is, at least partly, involved in the NeuroD1-mediated regulation of cell motility. Cell motility is closely linked to tumor invasion and metastasis. Indeed, in our allograft NB model, we can find metastasized small tumors in the ipsilateral axillary lymph nodes usually at approximately 4 weeks after sphere cell inoculation (data not shown). Evidence in support of our findings includes the fact that Slit2 is considered a candidate tumor suppressor gene; it is frequently inactivated in various cancers due to hypermethylation of its promoter region and allelic loss. Slit2 has a close relationship with cell migration and growth in tumor development (26–29). However, we did not observe a significant relationship between Slit2 expression and NB patient survival (data not shown). This may suggest that Slit2 does not entirely account for the role of NeuroD1 in NB development, although it is an important player downstream from NeuroD1.

In summary, we found an inverse relationship between NeuroD1 and Slit2 expression. NeuroD1 promotes cell motility and growth of tumor spheres as well as in vivo tumor growth. NeuroD1-mediated cell motility is at least in part due to Slit2 suppression. Our findings highlight the close relation of early neurogenesis and NB tumorigenesis and provide candidate molecular targets for the development of NB therapy.

No potential conflicts of interest were disclosed.

We thank Motoshi Suzuki (Nagoya University) for instruction us in FACS analysis, Didier Trono from the Swiss Federal Institute of Technology (EPFL, Switzerland) for providing the Lentivirus Packaging and Production plasmids psPAX2 and pMD2.G, and Hiroyuki Miyoshi from the Subteam for Manipulation of Cell Fate, RIKEN BioResource Center (Japan), and Atsushi Miyawaki from RIKEN Brain Science Institute (Japan) for providing the cDNA expression vector CSII-CMV-RfA-IRES2-Venus and control vector CSII-CMV-Venus. We also thank Tomoko Masuoka and Misako Tanase for their technical assistance.

This work was supported in part by a Grant-in-Aid for Cancer Research (20-13) from the Ministry of Health, Labour and Welfare, Japan, to K. Kadomatsu, by Grants-in-Aid (22790311 to S. Kishida; 21790317 to Y. Murakami-Tonami) from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan, by a Grant-in-Aid from the Ministry of Agriculture, Forestry and Fisheries, Japan, to K. Kadomatsu and A. Onishi, and by funds from the Global COE program of MEXT, Japan, to Nagoya University.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.