We have identified and characterized the sequence and expression of two Group B Sox genes in the acorn worm, Ptychodera flava. One sequence represents a Group B1 Sox gene and is designated Pf-SoxB1; the other is a Group B2 Sox gene and is designated Pf-SoxB2. Both genes encode polypeptides with an HMG domain in the N-terminal half. Whole-mount in situ hybridization to embryonic and larval stages of P. flava shows that the two genes are expressed in rather similar patterns at these stages. Expression is first detected in the cells of the blastula and subsequently localizes to the ectoderm during gastrulation. As the mouth forms, expression becomes concentrated in the stomodeum region. During morphogenesis of the tornaria larva, expression in the stomodeal ectoderm remains prominent around the mouth and under the oral hood. Later the genes are prominently upregulated in the ciliary bands and the apical organ. These results provide additional evidence that genes playing essential roles in the formation of the chordate dorsal central nervous system function in the development of the ciliary bands and apical organ, neural structures of this non-chordate deuterostome larva.
Many aspects of the evolution of the chordate body plan from primordial bilaterians are yet to be delineated or understood. One of the most interesting questions is the evolutionary foundations of the dorsal nerve chord that has produced the complex central nervous systems of higher vertebrates. Phylogenetically, chordates are deuterostomes, a superphylum that according to present molecular phylogeny consists of echinoderms and hemichordates as well as the chordates. Recent studies using comparison of 18S rDNA sequences strongly support the association of these phyla in the deuterostomes and assign other phyla traditionally linked to deuterostomes to different groups (Wada and Satoh, 1994; Turbeville et al., 1994; Aguinaldo et al., 1997; Wada, 1998; Knoll and Carroll, 1999; Cameron et al., 2000). We have been studying the expression of homologues of vertebrate brain-specific genes in hemichordates and echinoderms to gain clues concerning the evolution of the vertebrate brain (Tagawa et al., 2001).
The first vertebrate brain gene we examined was T-brain, a gene that is involved in the patterning of the vertebrate forebrain (Bulfone et al., 1995; Ryan et al., 1998). The T-brain homologue of the acorn worm Ptychodera flava (PfTbrain) is expressed in the apical organ, suggesting a link between the apical organ of the deuterostome larvae and the forebrain of vertebrates (Tagawa et al., 2000). This result provides the first molecular support for the hypothesis that the apical organ of dipleurula-type larvae is related to structures that provide the evolutionary substrate for the vertebrate forebrain (Lacalli, 1996)
Another brain-related gene that we have studied is otx, a gene essential for the development of the vertebrate anterior brain. Its homologues in P. flava (Pf-otx; Harada et al., 2000) and the sea cucumber Stichopus japonicus (Sj-otx; Shoguchi et al., 2000) are expressed in the both the preoral and postoral ciliary bands on the ventral or oral face of the larvae from both of these species. Shoguchi et al. (2000) followed Sj-otx expression from fertilized eggs through the auricularia larva and doliolaria larva stages to young juveniles. The Sj-otx is expressed in the components of ciliary bands that give rise to the oral side of juvenile holothurians. These results on the expression of vertebrate brain-specific genes in the ciliary bands of hemichordate and echinoderm larva highlight Garstang's idea of more than 70 years ago (Garstang, 1928) that there is an evolutionary relationship between the ciliary bands of dipleurula larvae and the chordate dorsal nerve chord. He proposed that the chordate dorsal nerve chord arose by rolling up of the ciliary bands on the dorsal side of the larva much as occurs during the formation of the dorsal hollow nerve cord of the juvenile acorn worm (Morgan, 1894).
We decided to examine in hemichordates homologues of other genes important in vertebrate brain development. In the present study, we have focused on the Sox (SRY-related box) gene family. Sox transcription factors contain a highly conserved ∼80 amino acid DNA-binding domain (Sox box) related to those of high mobility group (HMG) proteins. Members of the HMG superfamily bind AT-rich motifs and regulate gene transcription (Grosschedl et al., 1994; Pevny and Lovell-Badge, 1997). Members of the Sox transcription factor family play pivotal developmental roles in a number of organ systems (reviewed by Pevny and Lovell-Badge, 1997).
Sox family genes in vertebrates are classified into groups, presently lettered from A to J based on sequence similarity of the encoded HMG domain (Bowles et al., 2000). The Group B Sox genes are further subdivided into Group B1 genes (Sox1, Sox2, and Sox3 of human, mouse and chick) and Group B2 genes (Sox14 and Sox21 of mouse and zf-Sox19 of zebrafish) (Uchikawa et al., 1999). Analyses of early embryos have revealed that Sox2 and Sox3 are widely expressed in the nervous system in the chicken and Xenopus (Uwanogho et al., 1995; Mizuseki et al., 1998). Sox2 gene expression is correlated with neural competence of the ectoderm (Streit et al., 1997; Mizuseki et al., 1998). Sox14 and Sox21 of chick and zfSox19 of zebrafish are also expressed in the nervous system of embryos (Rex et al., 1997; Uchikawa et al., 1999; Vriz et al., 1996). In addition one group B1, SoxNeuro/dr-SoxB1, and one group B2, Dichaete/Fish/Sox70D/dr-SoxB2.1, Sox genes are expressed in the developing Drosophila nervous system (Nambu and Nambu, 1996; Russell et al., 1996; Cremazy et al., 2000).
These accumulated results establish that Group B Sox genes are involved in the specification and development in the vertebrate nervous system and were probably involved in neural specification in a common early ancestor of some or all of the bilaterian animals. We designed studies to identify Group B Sox genes conserved in hemichordates and determine their pattern of expression in embryos and tornaria larva.
MATERIALS AND METHODS
Animals and embryos
Mature adults of Ptychodera flava were collected from areas with shallow sandy bottoms around Oahu in December and maintained for a number of days at the Kewalo Marine Lab. Spawning was induced by temperature shift as described (Tagawa et al., 1998a). Fertilized eggs were allowed to develop to the specified developmental stages at room temperature. The time table of P. flava development is seen in Tagawa et al. (1998a).
About 120 cDNA clones were randomly selected from the gastrulae cDNA library of P. flava constructed with λZAPII (Stratagene) (Tagawa et al., 1998b), converted to plasmids and sequenced for about 500 bp at their 5′ end. The selected cDNA clones with SoxB2 similarity were completely sequenced on both strands with a BigDye Primer Cycle Sequencing Ready Reaction Kit and ABI-PRISM 377 DNA Sequencer (Perkin Elmer). The cDNA library was then screened with a random-labeled with [32P]dCTP probe (Amersham) of the HMG box of SoxB2 clone under reduced stringency conditions. Eleven clones were selected and sequenced as above. Nine clones contained inserts from the same SoxB1 gene. The longest was sequenced completely on both strands.
Sequence comparison and molecular phylogenetic analysis
Amino acid sequences of the HMG domains of Pf-SoxB1 and Pf-SoxB2 gene products were deduced from their cDNA nucleotide sequences. They were aligned with other SoxB1 subfamily members, and their molecular relationships were analyzed phylogenetically using the PHYLIP Neighbor-Joining method (ver. 3.5; Felsenstein, 1993).
Whole-mount in situ hybridization
Whole mount specimens were hybridized in situ essentially as described by Tagawa et al. (1998b). Briefly embryos and larvae were fixed in 4% paraformaldehyde in 0.5 M NaCl, 0.1 M MOPS, pH 7.5 on ice overnight. After a thorough wash with PBST (phosphate-buffered saline containing 0.1% Tween 20), the specimens were partially digested with 2 μg/ml proteinase K (Sigma) in PBST for 20 min at 37°C and were post-fixed with 4% paraformaldehyde in PBST for 1 hr at room temperature. After 1 hr prehybridization at 42°C the specimens were hybridized with digoxigenin (DIG)-labelled antisense probes for about 16 hr at 42°C. The hybridization buffer contained 50% formamide, 5 x Denhardt's solution, 100 mg/ml yeast RNA, 5 x SSC, and 0.1% Tween 20. Antisense and sense probes were synthesized from the full length cDNA insert following the instructions supplied with the kit (Boehringer Mannheim DIG RNA Labelling kit), and the hybridized fragments were reduced to about 200–500 nucleotides by alkaline hydrolysis. The probes were used at 0.5 μg/ml in the hybridization buffer. After hybridization, the specimens were digested with 20 mg/ml RNase A. After final wash in 50% formamide, 0.5 x SSC, 0.1% Tween 20 (20 min, 50°C), specimens were incubated with 0.5% blocking reagent (blocking buffer) in PBST at room temperature, then with 1:2000 Boehringer Mannheim alkaline-phosphatase-conjugated anti-DIG antibody in the blocking buffer, overnight at 4°C. Washes proceeded in PBST (20 min, four times), then in alkaline phosphatase buffer for 10 min twice. A signal was detected using NBT and BCIP following the supplier's instructions. The reactions were stopped in PBST after 1 to 2 hr, then the embryos were observed under a microscope. For double staining, the Pf-HNF3 probe (Taguchi et al., 2000) was labeled with fluorescein and both probes hybridized simultaneously. After developing signal with the anti-DIG reagent, the DIG antibody conjugate was removed at low pH and Pf-HNF3 signal developed with AP coupled anti-fluorescein FAB (Boehringer-Mannheim) and detected with Fast Red TR / Napthol AS-MX using the same staining procedure described above for the NBT and BCIP substrate.
Isolation and characterization of cDNA clones for hemi-chordate Group-B Sox genes
cDNA clones containing sequences of two different Group-B Sox genes were isolated from a P. flava gastrula cDNA library (Tagawa et al., 1998b) by EST sequencing and library screening methods. About 120 cDNA clones were randomly selected for EST sequencing from the 5′ end. Of these, four clones contained HMG boxes. One had sequence similarity to a Sry-related gene, another to an HMG gene, and two were identical and showed sequences suggestive of Group B2 Sox genes. An apparently full length clone was sequenced. The insert was 1,770 bp in length with an open reading frame that encoded a peptide of 246 amino acids (Fig. 1). It was tentatively designated Pf-SoxB2.
We searched further for cDNA clones with Group B1 type Sox gene sequences. The HMG box of Pf-SoxB2 was used as a probe to screen the P. flava gastrula cDNA library under reduced stringency. The screening yielded 109 positive clones. About 500 bp of the 5′ end of 11 clones were sequenced and it was evident that only two distinct mRNA sequences were represented in the positive clones. Two clones were from Pf-SoxB2 and nine appeared to be from a single SoxB1 gene. The longest SoxB1 clone was sequenced over both strands. The insert was 1,815 bp with an open reading frame encoding a polypeptide of 320 amino acids (Fig. 2). It was tentatively designated Pf-SoxB1.
Sequence analyses confirm that Pf-SoxB1 and Pf-SoxB2 are respectively Group B1 and Group B2 Sox genes
The identification of these clones as Group B1 and Group B2 Sox genes is based on comparisons of the amino acid sequences of the HMG domains. These domains constitute 79 amino acids and are found at the positions 42–120 in the open reading frame of Pf-SoxB1 (Fig. 2) and at positions 7–85 of Pf-SoxB2 (Fig. 1). Fig. 3A shows comparison of the HMG domain amino acid residues of Pf-SoxB1 and Pf-SoxB2 with those of SoxB1 and SoxB2 gene products of other animals. In these sequences 61 residues are identical and 13 are similar in the HMG domain in all of the proteins. The two diagnostic residues that distinguish Group B1 and Group B2 HMG domains occur at positions 2 and 78. Position 2 is R in Group B1 and H in Group B2 while position 78 is T in Group B1 and P in Group B2. As marked by double underlines in Fig. 3A, these diagnostic residues are appropriately found in the respective gene products.
In a second approach, we applied the neighbor-joining algorithm (Saitou and Nei, 1987) to perform a molecular phylogenetic sorting of Pf-SoxB1 and Pf-SoxB2 with Group B Sox gene products using the 79 amino acids residues of the HMG-domain. As shown in Fig. 3B, the tree sorted with Pf-SoxB1 among the Group B1 Sox genes while Pf-SoxB2 is included among members of the Group B2 Sox genes. The same groups included sea urchin SpSoxB1 and SpSoxB2, respectively (Fig. 3B; Kenny et al., 1999). These results establish that both Group B1 and Group B2 Sox gene sequences are found in cells of the gastrula stage of hemichordate embryos.
Pf-SoxB1 and Pf-SoxB2 are expressed in the apical organ and ciliary bands of the tornaria larva
The pattern and localization of expression of Pf-SoxB1 and Pf-SoxB2 during embryonic and larval development were examined by whole-mount in situ hybridization with antisense probes. Temporal and spatial patterns of expression of Pf-SoxB1 and Pf-SoxB2 were similar (Fig. 4). The PfSoxB1 mRNAs appear to be generally expressed at higher levels than are Pf-SoxB2 sequences. Equivalent hybridization signals from the two genes are produced when PfSoxB1 hybridized embryos and larvae are incubated with substrate for about 1/3 as long as Pf-SoxB2 hybridized samples. In situ hybridization with sense probes resulted in background signals only.
Results from whole-mount in situ hybridization of PfSoxB1 in early embryos and tornaria larvae are shown in Fig. 4A–E. Weak hybridization signal is first evident in the blastula at about 12 hr after fertilization. The signal occurs in the basal regions of the cells facing the blastocoel and this signal becomes more prominent in all of the cells as gastrulation is initiated (Fig. 4A). The signal in the arch-enteron cells wanes and disappears as gastrulation proceeds. During mouth formation, the signal in the stomodeal ectoderm upregulates while expression in the remainder of the ectodermal cells is extinguished. The upregulated, strong signal in the stomodeal ectodermal cells appears throughout the cell bodies (Fig. 4B, C) and continues under the oral hood and around the mouth of the tornaria. As the tornaria develops, strong signals appear in the apical organ (black arrowhead of Fig. 4D, E) and ciliary bands (white arrowhead of Fig. 4D, E). This pattern of signals showing expression on the ventral surface of the oral hood above the mouth and extending around the mouth as well as in the apical organ and the ciliary bands continues until at least the 8-day larva.
The timing and pattern of embryonic expression of PfSoxB2, as shown in Fig. 4F-L, is not distinguishable from the expression of Pf-SoxB1. Figure 4F shows the result from a whole-mount in situ hybridization of a Pf-SoxB2 antisense probe to a blastula stage embryo fixed at 12 hr after fertilization. A weak signal is exhibited in the basal region of the cells around the blastocoel (Fig. 4F, arrowhead). As shown in Fig. 4G for an 18-hr early gastrula and Fig. 4H, I for a 21-hr middle gastrula, the signal in the base of the cells becomes more prominent in ectodermal cells as gastrulation proceeds. After mid gastrulation, the signals in the base of the ectodermal cells gradually wane. But as with the PfSoxB1 gene, at the late gastrula stages the cells in the region of ectoderm that will form the stomodeum develop a strong signal throughout their cell bodies (Fig. 4J, K). As with Pf-SoxB1, the expression of Pf-SoxB2 in the stomodeal ectoderm cells continues as a zone of expression under the oral hood and around the mouth.
Tornaria larvae reacted to Pf-SoxB2 probe show signals in the ectoderm under the preoral hood and around the mouth (Fig. 4L, arrow region) as well as in the apical organ (Fig. 4L, black arrowhead) and ciliary bands (Fig. 4L, white arrowhead). As with Pf-SoxB1, this expression of Pf-SoxB2 in these tissues continued until at least the 8-day tornaria larva.
In the forming tornaria, the expression of the HNF-3/forkhead gene of P. flava, Pf-HNF3, is ultimately focussed in the endodermal cells that join the ectoderm to form the stomodeum (Taguchi et al., 2000). To delineate the spatial relationships between the endodermal expression of PfHNF3 and the ectodermal expression of Pf-SoxB2 in the stomodeal region, we performed double in situ hybridization with antisense probes from both genes. Archenteron expression of Pf-HNF3, as shown by red staining, is confirmed in late gastrula by this approach as shown in a vegetal view of the gastrula in Fig. 4I. When the double stained late gastrula preparations are viewed laterally or frontally (Fig. 4K, arrowhead), it is clear that the ectodermal expression of Pf-SoxB2, shown by purple staining, appears around the mouth and does not overlap the endodermal expression of Pf-HNF3, as shown by red signal, is in the pharynx (Fig. 4K). These genes appear to delineate the cells of ectodermal and endodermal origin respectively in the region of the mouth and pharynx in the tornaria larvae (Fig. 4L, arrow).
In this work, we add another example to evidence that genes implicated in the development of the vertebrate brain and nerve chord are expressed in the apical sensory organ and ciliary bands of dipleurula-type deuterostome larvae. We have isolated two P. flava homologues of Group B Sox genes, Pf-SoxB1 and Pf-SoxB2, and have shown that these homologues are both expressed in the apical organ and the ciliary bands of the tornaria. Previously, we described expression of T-brain, a mammalian forebrain-specific gene, in the apical organ in a hemichordate tornaria (Tagawa et al., 2000). As well, we found that homologues of otx, a gene important in anterior brain development in vertebrates, are expressed in the ventral or oral ciliary bands of hemichor-date and echinoderm larvae (Harada et al., 2000; Shoguchi et al., 2000). The cases of the Sox genes reported here represent the first examples of nervous system genes expressed simultaneously in both of these hemichordate larval structures.
The Pf-SoxB1 and Pf-SoxB2 sequences show the highest similarity to the sea urchin counterparts SpSoxB1 and SpSoxB2 derived from Strongylocentrotus purpuratus (Kenny et al., 1999). The study of the expression of the sea urchin genes has focussed on the expression of SpSoxB1 during early stages. The gene is expressed maternally and zygotically. The original uniform distribution becomes asymmetric at fourth cleavage when the protein is reduced in the vegetal micromeres.
The expression of the Group B1 genes, Sox2 and Sox3, in early chick embryos is also correlated with the specification of the ectoderm as well as establishment of neural competence in these cells (Streit et al., 1997; Mizuseki et al., 1998). Sox2 (group B1) expression increases dramatically as neural ectoderm is established. The appearance of Sox2 in neural ectoderm represents one of the earliest molecular responses to neural induction documented so far (Rex et al., 1997). Sox3 (also group B1) in the chick embryo is expressed throughout the ectoderm that is competent to form nervous tissue before neural induction. The expression of Sox3 is lost from cells as they undergo gastrulation to form non-ectodermal tissues. The transcription factor, Brachyury, is produced in cells about to undergo gastrulation a short time before Sox3 transcripts are lost. Therefore, Brachyury expression may act functionally upstream of Sox3 down regulation. Sox3 expression is also lost from non-neural ectoderm shortly after the neural plate becomes morphologically apparent.
Sox2 function is required for neural differentiation of early Xenopus ectoderm. Microinjection of dominant-negative forms of Sox2 (dnSox2) mRNA inhibits neural differentiation of animal caps caused by attenuation of BMP signals (Kishi et al., 2000). Expression of dnSox2 in developing embryos suppresses expression of N-CAM and regional neural markers. Analysis of the temporal requirement of Sox2-mediated signaling using an inducible dnSox2 construct fused to the ligand-binding domain of the glucocorticoid receptor shows that attenuation of Sox2 function beginning at late blastula or late gastrula stage onwards causes an inhibition of neural differentiation in animal caps and in whole embryos. Additionally, dnSox2-injected cells that fail to differentiate into neural tissues are not able to adopt epidermal cell fates. These data suggest that Sox2 gene products are essential for early neuroectoderm cells to consolidate their neural identity during secondary steps of neural differentiation.
In the present study, we found that the hemichordate Group B Sox genes, Pf-SoxB1 and Pf-SoxB2, are expressed in the early ectoderm of blastulae and gastrulae. This early ectoderm expression in hemichordate could play a similar role in specification of the neuroectoderm. In addition to ectoderm, Group B Sox genes are dramatically implicated in the specification and differentiation of the brain and central nervous system of the vertebrates (as well as in the development of multiple other structures). Pf-SoxB1 is a hemichordate Group B1 homologue of Sox2 and Sox3 that are both expressed in the immature neural epithelium of the entire CNS of HH stage 10 to 34 chick embryos. These genes are widely expressed as the nervous system develops in the chicken and Xenopus (Uwanogho et al., 1995; Mizuseki et al., 1998).
Pf-SoxB2 is a hemichordate Group B2 orthologue of Sox14 and Sox21. In mouse and chick, Sox14 expression is restricted to a limited population of neurons in the developing brain and spinal cord of both species. Sox14 marks a subset of interneurons at a defined dorsoventral position adjacent to ventral motor neurons in the spinal cord (Hargrave et al., 2000). Sox21 of zebrafish is expressed in the forebrain, midbrain and hindbrain, but maximally at the mid-brain-hindbrain junction (Rimini et al., 1999)
When linked to the GAL4 DNA binding domain, these Group B2 proteins, Sox21 and Sox14 inhibited activation of the d1-crystallin DC5 enhancer by the Group B1 proteins, Sox1 or Sox2 (Uchikawa et al., 1999). This suggests that the Group B2 Sox proteins may be repressors of group B1 Sox genes. Since the activating B1 Sox genes and repressing B2 Sox genes display interesting overlaps of expression domains in developing tissues such as the optic tectum, spinal cord, inner ear, alimentary tract and branchial arches, this interaction could play a variety of important and subtle roles in regulating development (Uchikawa et al., 1999). In the context of our observations in hemichordates, we observe overlapping expression of the B1 and B2 homo-logues and it is interesting to speculate about possible opposing regulatory activities by the proteins from these two classes of genes.
In Drosophila, the gene now named dr-Sox B2.1 (Bowles et al., 2000), and variously designated Dichaete, Sox70D or fish-hook, is essential for central nervous system development. Null dr-Sox B2.1 mutants show the loss of specific neurons in the brain, fusion of adjacent ventral nerve chord ganglia and aberrant axon scaffold organization (Nambu and Nambu, 1996). During development, the gene has a dynamic expression profile in the forming central nervous system and has an especially intriguing pattern of expression associated with the ventral midline (Soriano and Russell, 1998). The Drosophila Group B1 Sox gene first designated SoxNeuro and recently renamed dr-SoxB1 also has a highly dynamic pattern of expression during development (Cremazy et al., 2000). Early in development the gene is expressed in the presumptive ventro-lateral neuroectoderm. During gastrulation expression occurs in the cephalic and ventral neurogenic regions and then remains associated with the developing central nervous system.
We observed a strong in situ hybridization signal in the ectoderm under the oral hood and around the mouth of the hemichordate larvae. To our knowledge, there is no vertebrate correlate to this early expression of the group B Sox genes during mouth formation. Although there is no direct description of the transition of marked structures in this region from tornaria to adult worm, this region of the tornaria gives rise to the preoral ciliary organ of the adult. This organ was first well described by Brambell and Cole (1939) who believed it had a sensory function. Subsequently, this organ has been shown to be highly innervated as would be expected of a structure with a neural function (Knight-Jones, 1952). Thus, the Sox B gene expression under the preoral hood could mark another region developing neural function (Brambell and Cole, 1939).
The present finding supports the notion that genes responsible for the formation of vertebrate CNS are expressed in the ciliary bands and apical organ of tornaria larvae. The apical organ and the ciliary bands are both neural organs of the larval ectoderm. The apical organ contains serotonergic nerve cells and the ciliated epithelial cells of the ciliary bands produce axons that run in fiber tracts along the bases of the epithelial cells of the bands. These cells represent the very primitive nervous system of the larvae and could be considered logical substrates for the evolution of the vertebrate brain (Garstang, 1928; Lacalli, 1996), if it is established that vertebrate ancestors actually included a form related to these swimming larval stages.
With these results, it seems reasonable to propose that genes used for the formation of complex vertebrate CNS are likely to be expressed in the ciliary bands and/or the apical organ of invertebrate larvae. However, this does not necessarily mean that the ciliary bands and the apical organ of hemichordate larvae are orthologues of the chordate larval and adult CNS. Instead, it may be that when the two groups of deuterostomes evolved from their common ancestor, the same genes were used to determine their nervous systems. There may have been strong evolutionary constraint in the usage of developmental genes in the formation of the nerve cells during evolution of the different groups of deuterostome. Because Sox group B genes are used in the formation of Drosophila CNS, this notion may be expanded to the entire metazoans. Therefore, it is an intriguing subject of future research to explore in detail the molecular machinery for the utilization of these genes in the formation of the nervous system in the different animal groups.
There are a number of very basic questions concerning the expression of these brain-specific genes in deuterostome larvae that remain to be examined. The hemichordate tornaria larva swims for an extended period of pelagic life before it metamorphoses in a worm and settles to the sandy bottom. The worm has a both major dorsal nerve cord and a larger ventral nerve cord (e.g. Brusca and Brusca, 1990). It will be very interesting to ascertain the expression of PfOtx, Pf-Tbrain and Pf-SoxB genes during the formation of the adult nervous system as the tornaria larva transitions to the juvenile worm.
We wish to thank Eiichi Shoguchi and Kazuko Hirayama for their help in the preparation of the manuscript. This work was supported in part by a Grant-in-Aid from the Monbusho, Japan to NS. KT was a postdoctoral fellow of JSPS with a research grant (#03725).