Neuroepithelial cells require fucosylated glycans to guide the migration of vagus motor neuron progenitors in the developing zebrafish hindbrain

During development of the nervous system, neural migration plays a crucial role in expanding the size of the brain and ensuring correct arrangement of the neural layers and nuclei. The final location of neurons is especially crucial, because neural function depends on precise connections made by neurons and their targets. Defects in neural migration result in neurological diseases; for example, lissencephaly with cerebellar hypoplasia is caused by disruption of the glycoprotein Reelin(Gressens, 2006).

Many studies have identified molecules involved in the control of neural migration. These studies have revealed that proper neural migration requires various genes that function not only in migrating neurons, but also in the cells that support them. For example, we have reported that the planar cell polarity genes frizzled3a and celsr2 function in neuroepithelial cells to guide the caudal migration of facial motor neurons(Wada et al., 2006). Despite this, there is still very little known about the molecular mechanisms that regulate termination of neural migration and the accumulation of neural progenitors to form the proper layers or nuclei.

The vagus nerve is the tenth cranial nerve, located in rhombomere (r) 8 in the caudal hindbrain. It controls various body functions, including heart beat and gastrointestinal movement (Gilland and Baker, 2005; Taylor et al.,1999). Correct arrangement of vagus motor nuclei following neural migration and accumulation may be crucial for the normal functioning of vagus motor neurons. For example, although sudden infant death syndrome (SIDS), the most common cause of postnatal infant death, is a complex and multifactorial disorder (Moon et al., 2007),it has been suggested that the reduced number of neurons in vagus motor nuclei in victims of SIDS is due to a failure of neuronal migration(Macchi et al., 2002). Thus,there is a need to understand how vagus motor nuclei are formed by neural migration, and the molecular mechanisms that govern this process.

In the present study, we used vagus motor nuclei as a new model system in which to study the molecular mechanisms underlying neural migration. Time-lapse observations of hindbrain explants from the transgenic zebrafish line Tg(CM-isl1:GFP)^rw0, hereafter referred to as isl1:GFP, which expresses green fluorescent protein (GFP) in motor neurons, including vagus motor neurons(Higashijima et al., 2000),demonstrated that vagus motor neuron progenitors are born near the ventral midline, migrate in a dorsolateral direction and accumulate to form the vagus motor nuclei. An N-ethyl-N-nitrosourea (ENU)-based mutant screen using the isl1:GFP transgenic zebrafish isolated towhead (twd^rw685) mutant embryos, in which the vagus motor neuron progenitors migrate beyond their proper position to the dorsal roof of the hindbrain. The twd^rw685 locus encodes GMDS, a key enzyme in the fucosylation pathway(Staudacher et al., 1999).

Fucosylated glycans are involved in a variety of biological and pathological processes, including cell migration(Ma et al., 2006). For example, sialyl Lewis X (sLe^x), a fucosylated carbohydrate structure, is the core recognition epitope that mediates lymphocyte homing and initial leukocyte-endothelial cell adhesion(Ma et al., 2006). The donor substrate for fucosylation, GDP-fucose, is synthesized in the cytoplasm mainly through a de novo synthesis pathway, which includes an enzymatic reaction catalyzed by GDP-mannose 4,6-dehydrase (GMDS)(Fig. 1). Subsequently, a GDP-fucose transporter at the Golgi membrane transports GDP-fucose from the cytoplasm to the Golgi lumen, and fucosyltransferases (FUTs) transfer GDP-fucose to the acceptor molecules. Leukocyte adhesion deficiency II (LAD II), also known as congenital disorder of glycosylation IIc (CDG IIc), results in mental retardation, short limbs and stature, and a flat face with a broad,depressed nasal bridge (Freeze,2001). These features are caused by a defect in the Golgi-localized GDP-fucose transporter, resulting in impaired expression of the fucosylated glycans (Lubke et al.,2001; Lubke et al.,1999; Luhn et al.,2001). Although individuals with LAD II/CDG IIc exhibit neurological abnormalities, how fucosylated glycans regulate neural development is not yet understood. The findings of the present study reveal a novel molecular mechanism in the establishment of the structure of the vagus motor nuclei following neural migration, whereby fucosylation of the substrate plays a key role in preventing vagus motor neuron progenitors from migrating along an aberrant pathway.

Maintenance of zebrafish, ENU-based mutagenesis and genetic mapping of mutant loci were performed as described previously(Tanaka et al., 2007; Wada et al., 2005; Wada et al., 2006; Westerfield, 2007). The simple sequence length polymorphism (SSLP) marker BX510653-1 was generated based on the sequence of the eighth intron of the gmds gene. The isl1:GFP fish and the twd^rw685 fish are available from the National BioResource Project of Japan(http://www.shigen.nig.ac.jp/zebra/index_en.html).

The procedures for time-lapse imaging of the hindbrain were essentially the same as those described previously (Tanaka et al., 2007). The hindbrain explant in the culture chamber was observed under a confocal microscope (LSM510; Carl Zeiss) every 15 minutes from 24 to 48 hours post-fertilization (hpf). Labeling of wild-type cells by rhodamine-dextran and transplantation were performed according to standard protocols (Westerfield,2007).

cDNA cloning, reverse transcription-polymerase chain reaction (RT-PCR) and synthesis of sense-capped mRNAs were performed as described previously(Wada et al., 2006). Details of primers and the antisense morpholino oligonucleotides (MOs; Gene Tools)used in this study can be provided on request. For the injection of mRNA or MO solution, approximately 1 nl of the solution at the concentrations indicated in Table 1 was injected into one-cell stage embryos. The GenBank Accession Numbers for zebrafish L-gmds and S-gmds are AB290320 and AB290319, respectively. Amino acid sequence similarity was analyzed using the CLASTAL W program. The following amino acid sequences of GMDS were used for sequence comparisons: Homo sapiens (GenBank accession number NM_001500), Mus musclus (BC093502), Xenopus laevis (BC111472), Drosophila melanogaster (NM_135044), Caenorhabditis elegans (NM_069162) and Escherichia coli (NZ_AAJT01000120). Embryo genotypes were identified by direct sequencing of the mutation site.

Embryos were fixed at the indicated developmental stage with 4%paraformaldehyde (Wako) at 4°C overnight. In situ hybridization,whole-mount staining, cryosection staining and retrograde labeling of the reticulospinal neurons were performed essentially as described previously(Aizawa et al., 2007; Shepard et al., 2004; Wada et al., 2005; Westerfield, 2007). Details of primary antibodies and biotinylated lectins used in the present study can be provided on request. These probes were detected by fluorescein-conjugated secondary antibodies (Invitrogen; 1:500 dilution for each) or avidin(Chemicon; 1:500 dilution). For immunoblotting, 25 embryos of wild-type and the twd^rw685 mutants were pooled and lysed in Laemmli sample buffer. The remaining steps in the immunoblotting procedure were as described previously (Ohata et al.,2009).

The isl1:GFP transgenic fish expresses GFP in several cranial and spinal motor neurons, including the vagus neurons(Higashijima et al., 2000). Like mammals (Gilland and Baker,2005; Taylor et al.,1999), zebrafish form two vagus motor nuclei in r8, the most caudal part of the hindbrain. One is the dorsolateral motor nucleus of the vagus (dlX), whereas the other is the medial motor nucleus of the vagus (mmX)(Fig. 2A, parts a,b). In the present study, we used time-lapse imaging of hindbrain explants to observe the process of vagus motor nuclei formation(Fig. 2B). The vagus motor neuron progenitors were born near the ventral midline from 24 to 48 hpf(Fig. 2B; see Movie 1 in the supplementary material). The dlX was formed first by the dorsolateral migration and accumulation of vagus motor neuron progenitors from 24 to 36 hpf(Fig. 2B). Then, mmX was formed in the ventromedial region of r8 by 48 hpf(Fig. 2B).

To dissect genetically the process of the formation of vagus motor nuclei,we performed ENU-based mutant screening and isolated the twd^rw685 mutant. In the twd^rw685embryos, the vagus motor nuclei were fused across the midline(Fig. 3A). In addition, the distribution of anterior and posterior trigeminal (Va and Vp) and facial (VII)motor neurons was perturbed in the twd^rw685 mutant (33/33(100%) (Fig. 3A), presumably because of defects in the lateral migration of the trigeminal motor neurons and the posterior migration of the facial motor neurons. The morphology of the twd^rw685 embryos appeared to be normal at 2 days post-fertilization (dpf), except for their `curled up' tails(Fig. 3B). The twd^rw685 mutants died between 6 and 7 dpf.

In the twd^rw685 embryos, ectopic neurons labeled by isl1:GFP are visible as distinctive nuclear structures at 36 hpf. They are widely distributed throughout the dorsomedial hindbrain region between the areas where the dlX is normally formed. By 48 hpf, these ectopic neurons are seen to reach the dorsal roof of the hindbrain (`overshooting')(Fig. 3C; see Movie 2 in the supplementary material).

To distinguish between dlX and mmX neurons, we took advantage of the differences in time when these neurons are born. Specifically, wild-type and twd^rw685 embryos were labeled with BrdU for 30 minutes at 38 hpf and were fixed at 3 dpf for staining with anti-BrdU and anti-GFP antibodies. In wild-type embryos, most dlX neurons were BrdU negative(Fig. 3D, part a), whereas all mmX neurons were BrdU positive. Most of the ectopic neurons in twd^rw685 embryos were also BrdU negative(Fig. 3D, part b), indicating a common origin with dlX neurons. In addition, most BrdU and GFP double-positive cells from isl1:GFP transgenic embryos were localized ectopically in twd^rw685 mutants. This indicates that the twd^rw685 mutation may affect the migration of both dlX and mmX neurons.

To clarify the effect of the twd^rw685 mutation on the migration of dlX progenitors, we observed the embryos using time-lapse imaging. In wild-type embryos, the dlX progenitors migrate in the dorsolateral direction soon after they are born (Fig. 3E; see Movie 1 in the supplementary material). However, in twd^rw685 embryos, the dlX progenitors migrate in more dorsal directions, straying away from the normal migratory pathway(Fig. 3E; see Movie 2 in the supplementary material).

Various markers were used in twd^rw685 mutants to determine the effects of the twd^rw685 mutation. The zn-5 antibody specifically labels segmentally repeated commissural axons in the zebrafish hindbrain (Trevarrow et al.,1990). The formation of zn-5-immunoreactive axons appeared normal in twd^rw685 embryos(Fig. 4A,B). Furthermore,labeling of reticulospinal neurons by injection of a tracer dye into the spinal cord revealed that the anterior-posterior patterning of the reticulospinal neurons in twd^rw685 embryos was identical to that in wild-type embryos (Fig. 4C,D). The LIM homeodomain proteins Lhx2 and Lhx9, as well as glutamate decarboxylases Gad1 and Gad2, are segmentally expressed in the hindbrain (Ando et al., 2005; Sassa et al., 2007). In the present study, the location of both Lhx2/9- and Gad1/2-immunoreactive neurons was almost normal in twd^rw685 embryos(Fig. 4E-H). These results suggest that the overall patterning and differentiation of neurons other than the vagus motor neurons in the posterior hindbrain were unaffected by the twd^rw685 mutation.

To investigate neuroepithelial apicobasal polarity and tight junctions in twd^rw685 embryos, the twd^rw685 mutants were stained with anti-atypical protein kinase C (aPKC) and anti-zonula occludens-1 (ZO-1) antibodies (Fig. 4I-L). The localization of both aPKC and ZO-1 was normal,indicating that neuroepithelial apicobasal polarity and tight junctions were unaffected in twd^rw685 embryos.

The twd^rw685 locus was mapped between SSLP markers z14995 and z14614 on linkage group 20 (LG20) and no recombination was detected at the gene encoding GMDS in 702 meioses(Fig. 5A). The gmdsgene consists of 12 exons and 11 introns(Fig. 5B). cDNA cloning of wild-type embryos identified two splicing variant mRNAs, depending on whether the fourth exon, which consists of 21 bases, is included (L-gmds) or skipped (S-gmds) in the mature mRNA. These two variants share an identical amino acid sequence, except for the seven residues encoded by the fourth exon (Fig. 5B,C). A mutation from T to A was detected in the seventh exon of the gmdsgene from twd^rw685 embryos(Fig. 5B-D), resulting in amino acid substitution from Trp¹⁹³ in L-GMDS (or Trp¹⁸⁶ in S-GMDS) to Arg. Trp¹⁹³ (or Trp¹⁸⁶ in S-GMDS) is conserved in the GMDS from various species, ranging from human to bacteria(Fig. 5C).

To confirm that gmds is the gene responsible for the twd^rw685 phenotype, we performed loss-of-function and gain-of-function analyses of gmds. Injection of an MO designed to inhibit splicing of premature gmds mRNA into one-cell stage wild-type embryos effectively blocked the maturation of gmds mRNA(Fig. 6A). Although injection of the 5-mis-pair control MO had no effect on motor neurons in morphants(0/113; 0%) (Fig. 6B, part a),injection of the splice-blocking MO against gmds produced a fused vagus motor nuclei phenotype (130/133; 98%)(Fig. 6B, part b) that was identical to that observed in the twd^rw685 mutant(Fig. 3A). In addition, the distribution of trigeminal and facial motor neurons was perturbed in gmds morphant embryos (41/41; 100%)(Fig. 6Bb), as in the twd^rw685 mutant embryos(Fig. 3A).

Further experiments were performed in which the effects of injection of either of the in vitro-synthesized sense-capped S- or L-gmdsmRNAs (40 or 80 μg/ml) into one-cell stage embryos derived from heterozygous twd^rw685 parents were investigated. Of the twd^rw685 embryos injected with 80 μg/ml L-gmdsmRNA, 88% (21/24) exhibited a completely rescued phenotype(Table 1; Fig. 6B, part c) and 13%exhibited a partially rescued phenotype, in which some, but not all, vagus motor neurons were found in abnormal locations(Fig. 6B, part c′,arrowheads). When the lower concentration of L-gmds mRNA (40μg/ml) was injected, there was a concomitant decrease in the number of twd^rw685 embryos rescued [14/27 (52%) and 8/27 (30%)embryos completely and partially rescued, respectively] and 19% (5/27) of mutant embryos still showed the twd^rw685 phenotype(Table 1). Injection of 80μg/ml S-gmds mRNA rescued twd^rw685 embryos with a similar efficiency as that seen following injection of the same concentration of L-gmds mRNA [22/30 (73%) and 8/30 (27%) embryos completely and partially rescued, respectively](Table 1; Fig. 6B, part d). By contrast,injection of the mutant-type mRNA did not result in the rescue of any twd^rw685 mutants (Table 1). Overexpression of L- or S-gmds mRNA in wild-type embryos had no effect on the differentiation and migration of vagus motor neurons (Fig. 6B, part e).

In addition, the migration and accumulation of dlX progenitors was investigated using time-lapse imaging. In twd^rw685 embryos injected with 80 μg/ml S- and L-gmds mRNA, the dlX progenitors migrated dorsolaterally and formed the dlX as seen in wild-type embryos (Fig. 6C; see Movie 3 in the supplementary material).

The expression profile of fucosylated glycans in the hindbrain of zebrafish embryos was examined in the present study using biotinylated lectins, because GMDS is a key enzyme in the fucosylation pathway(Sullivan et al., 1998). In wild-type embryos, N-linked fucosylated glycans recognized by Aleuria aurantia lectin (AAL) and Lens culinaris agglutinin(LCA) were expressed ubiquitously throughout the hindbrain(Fig. 7A; see Fig. S1A in the supplementary material) (Varki et al.,1999). However, in twd^rw685 embryos, glycan levels were markedly reduced (Fig. 7B; see Fig. S1B in the supplementary material). The expression of other types of N-linked glycans, namely those recognized by concanavalin A (ConA), wheat-germ agglutinin (WGA) and Ricinus communis agglutinin I (RCA-I), as well as the expression of HNK-1 antigen detected using zn12 antibody (Metcalfe et al., 1990), appeared to be normal in mutant embryos(Fig. 7D,E; see Fig. S1C-H in the supplementary material). In wild-type embryos, the fucosylated tetrasaccharide structure sLe^x was expressed in the ventricular zone, but expression of sLe^x was markedly reduced in twd^rw685 mutant embryos(Fig. 7F,G). This reduction in the expression of sLe^x was confirmed by immunoblotting(Fig. 7H). O-Linked glycans recognized by Ulex europaeus agglutinin I (UEA-I) were not detectable in the hindbrain of either wild-type or twd^rw685 embryos (see Fig. S1I,J in the supplementary material). In mutant embryos rescued by injection of either L- or S-gmds variant mRNA, the expression of N-linked fucosylated glycans recognized by AAL was recovered(Fig. 7C). By contrast, levels of AAL-reactive glycans in twd^rw685 mutant embryos were not rescued by injection of either mutant-type L- or S-gmdsvariant mRNA (data not shown). These results indicate that fucosylated glycans are specifically affected by the mutation in the gmds gene.

FUTs are classified according to the site of fucose addition as follows:α1,2 (encoded by FUT1 and FUT2); α1,3/4 (encoded by FUT3); α1,3 (encoded by FUT4-7 and FUT9-11and ft1-2); and α1,6 (encoded by FUT8)(Kageyama et al., 1999; Ma et al., 2006; Mollicone et al., 2008). Protein O-fucosyl transferases (POFUTs) are encoded by pofut1 and pofut2. Of the FUTs, pofut1, pofut2 and FUT8 have already been cloned in zebrafish. In mammals, these genes encode the enzymes catalyzing O-fucosylation in the consensus sequence of the epidermal growth factor (EGF)-like repeat (POFUT1)(Fig. 8A, part a), O-fucosylation in the consensus sequence of the thrombospondin type-1 repeat (POFUT2) (Fig. 8A, part b) and α1,6-fucosylation to the innermost N-acetylglucosamine(GlcNAc) moiety of the core N-linked glycans (FUT8; coreα1,6-fucosylation) (Fig. 8A, part c). Therefore, we knocked down these genes to investigate the importance of O- and core α1,6-fucosylation in the formation of the vagus motor nuclei. Although MOs against pofut1,pofut2 and FUT8 effectively knocked down the translation of EYFP-tagged mRNAs (Fig. 8B,parts a-f), these MOs did not induce fusion of the bilateral vagus motor nuclei (Fig. 8C, parts a-c). Instead, some of the morphants showed aberrant positioning of the vagus motor neurons lateral to the dlX (Fig. 8C, parts a-c), whereas embryos injected with the standard control MO were identical to the wild-type embryos (data not shown). These results indicate that O- and core α1,6-fucosylation do not contribute to the cessation of migration of vagus motor neurons.

α1,3-fucosylation consists of core α1,3-fucosylation to the innermost GlcNAc moiety of the core N-linked glycans and terminal fucosylation, where GDP-fucose is linked to the oligosaccharide chains as a terminal modification that is not elongated further(Fig. 8A, part c)(Ma et al., 2006). Of the terminal α1,3-FUTs, ft1, ft2, FUT7 and FUT9 have been identified. However, knocking down of those genes did not induce fusion of the bilateral vagus motor nuclei [1/30 (3%), 0/26 (0%), 0/38 (0%), 0/33 (0%),respectively, data not shown]. Next, we knocked down FUT10, which encodes a core α1,3-FUT. MO against FUT10 effectively knocked down the translation of EYFP-tagged mRNA(Fig. 8B, parts g,h), and caused fusion of the bilateral vagus motor nuclei (22/35; 63%),(Fig. 8C, part d). However,unlike twd^rw685 mutants, FUT10 morphants did not show the overshooting of the vagus motor neurons, but showed disruption of neuroepithelial apicobasal polarity and adherens junctions (data not shown). These results indicate that α1,3-fucosylation catalyzed by FUT10 affects neural development in a different manner from that of GMDS.

Fucosylation has been shown to regulate Notch signaling, and genetic inactivation of the Notch activator γ-Secretase causes mis-migration of the facial motor neurons (Ishikawa et al.,2005; Louvi et al.,2004; Ma et al.,2006; Sasamura et al.,2007). Although O-fucosylation has an important role in Notch signaling, the expression of the Notch target gene her4(Takke et al., 1999) was not affected in twd^rw685 mutants in the present study(Fig. 9A-C). Because Notch signaling has an antineurogenic effect on neural stem cells, we observed primary motor neurons and Rohan-Beard sensory neurons in the spinal cord(Inoue et al., 1994; Tokumoto et al., 1995). There were no significant differences detected in the spinal cord between wild-type and twd^rw685 mutant embryos(Fig. 9D-G). Because Notch activation is also required for the segregation of the rhombomere boundary(Cheng et al., 2004), we examined the expression of genetic markers for hindbrain segmentation. The expression pattern of radical fringe and wnt1 in twd^rw685 mutant embryos was identical to that in wild-type embryos (Fig. 9H-K), indicating that there is no significant defect in Notch activity in twd^rw685 mutant embryos.

To determine which cells require gmds for the correct formation of vagus motor nuclei, we performed mosaic analysis by transplanting rhodamine-dextran-labeled wild-type cells at the blastoderm stage into gmds morphant host embryos at the shield stage. The progeny of the donor cells that differentiated into the motor neurons could be identified by their GFP and rhodamine signals. The wild-type dlX progenitors in the hindbrain of wild-type localized normally(Fig. 10A, part a), but those in the hindbrain of gmds morphant embryos did not stop at the correct position and were positioned ectopically, close to the midline, similar to the positioning of dlX progenitors in twd^rw685 mutants(Fig. 10A, parts b,c; C, parts a,b) (n=3). Similar results were obtained following transplantation of wild-type dlX progenitors into twd^rw685 mutants (data not shown; n=3). These results indicate that gmds expression in vagus motor neuron progenitors is not essential for correct migration of these progenitor cells.

Neuroepithelial cells are likely to be regulators of the migration of vagus motor neuron progenitors, because they have been shown to support the caudal migration of facial motor neuron progenitors by preventing the integration of migrating facial motor neuron progenitors into the neuroepithelial layer(Wada et al., 2006) and gmds mRNA was expressed ubiquitously (data not shown). In the present study, when wild-type neuroepithelial cells were placed into the dorsomedial region of the hindbrain of gmds morphants, the vagus motor progenitors were found to migrate to normal positions (line 2 in Fig. 10B, parts a,f; C, part c) (n=5). Of note, morphant vagus motor neuron progenitors entered the dorsomedial region and positioned more dorsally if wild-type cells were not distributed in the dorsomedial region of the hindbrain (line 1 in Fig. 10B, parts a,e). These results suggest that the recovery of fucosylation in neuroepithelial cells in the dorsomedial region of the hindbrain is sufficient to restore normal migration of mutant vagus motor progenitors.

The fucosylation donor substrate GDP-fucose is generated through a de novo synthesis pathway that converts GDP-mannose to GDP-fucose, and a salvage pathway that converts free cytosolic fucose to GDP-fucose(Fig. 1). In HeLa cells, more than 90% of GDP-fucose is derived from the de novo synthesis pathway(Ma et al., 2006). The twd^rw685 mutation occurs in the gmds gene, which encodes a key enzyme in the de novo synthesis pathway (Figs 1 and 5). As a result, expression of fucosylated glycans recognized by AAL and LCA (core-fucosylation), as well as the anti-sLe^x antibody (terminal fucosylation), was markedly and specifically reduced. Injection of normal gmds mRNA was able to restore the expression of these glycans(Fig. 7; see Fig. S1 in the supplementary material). These results indicate that core and terminal fucosylation greatly depend on the de novo synthesis pathway.

The structure of GMDS is highly conserved among organisms; in bacteria and Arabidopsis thaliana, GMDS works as an oligomer(Mulichak et al., 2002; Somoza et al., 2000; Webb et al., 2004). The twd^rw685 mutation changes the conserved hydrophobic residue Trp¹⁹³ in L-GMDS (Trp¹⁸⁶ in S-GMDS) into a basic residue, Arg, in the region involved in oligomer formation(Fig. 5C). Therefore, this mutation may disrupt oligomerization of GMDS, which may be crucial for its enzymatic activity. However, we cannot rule out the possibility that the twd^rw685 mutation affects the enzymatic activity directly and/or the stability of the GMDS protein.

In the present study, we proposed a model in which fucosylated glycans of surrounding neuroepithelial cells prevent overshooting in the tangential migration of vagus motor neuron progenitors(Fig. 10D). There are at least two ways in which fucosylated glycans may guide migration. First, fucosylated glycans may function to maintain the morphology or structure of neuroepithelial cells, which is necessary for the cells to guide migration. However, taking into consideration the fact that neuroepithelial apicobasal polarities and tight junctions are formed properly in twd^rw685 embryos (Fig. 4I-L), it is likely that the neuroepithelial architecture is normal. Second, fucosylated glycans expressed by neuroepithelial cells may function as guidance molecules that repulse migrating cells. In this case,levels of fucosylated glycans should be enriched in the medial region of the hindbrain, without uniform expression in neuroepithelial cells, because exclusion of vagus motor neuron progenitors occurs only in the medial region. Such a biased distribution of fucosylated glycans is possible, as demonstrated by the greater distribution of sLe^x in the ventricular zone of wild-type embryos and its considerable reduction in twd^rw685 mutant embryos(Fig. 7F,G).

Fucosylation is classified into three types: terminal, core- and O-fucosylation, depending on the site of fucose addition. Terminal fucosylation includes α1,2-(catalyzed by FUT1 and FUT2),α1,3/4-(catalyzed by FUT3) and α1,3-(catalyzed by FUT4-7 and FUT9,and FT1-2) fucosylation (Ma et al.,2006). As a representative structure of terminal fucosylation,sLe^x is the core recognition epitope for members of the Selectin family, which mediate lymphocyte homing and initial leukocyte-endothelial cell adhesion (Ma et al., 2006). Core fucosylation includes core α1,3-(catalyzed by FUT10 and 11) and core α1,6-fucosylation (catalyzed by FUT8), and transfers GDP-fucose to the reducing terminal of the core structure of an asparagine-linked oligosaccharide (Ma et al.,2006). Core α1,6-fucosylation has been reported to be essential for both transforming growth factor-β1 (TGF-β1) and EGF receptor signaling (Wang et al.,2006; Wang et al.,2005). In O-fucosylation, GDP-fucose is linked directly to Ser or Thr residues in the consensus sequence of the EGF-like repeat and in the thrombospondin type-1 repeat by POFUT1 and POFUT2, respectively(Ma et al., 2006). The Notch signaling pathway, one of the pathways in which O-fucosylation is involved, is well characterized (Ma et al., 2006). In the present study, we knocked down FUT8,pofut1 and pofut2 genes to investigate the contribution of coreα1,6- and O-fucosylation in the migration of vagus motor neuron progenitors, but did not observe any morphant with bilaterally fused vagus motor nuclei that were seen in twd^rw685 mutants. Our data suggest that some other glycans than core α1,6- or O-fucosylated glycans are involved in the signal that stops the migration of vagus motor neuron progenitors in the right location. Therefore,fucosylated adhesion molecules, such as Selectin ligands or soluble guidance molecules may be fucosylation targets involved in the regulation of vagus motor neuron migration (Ma et al.,2006). Reelin is a candidate soluble guidance molecule, because it is fucosylated (Botella-Lopez et al.,2006), and repression of Reelin cause mis-formation of the vagus motor nuclei (S. Kikkawa, personal communication).

In the present study, we further knocked down terminal α1,3-FUTs(FT1, FT2, FUT7 and FUT9) and a core α1,3-FUT (FUT10), which have been identified in zebrafish. However, those morphants did not show the same phenotype as the twd^rw685 mutants. Therefore, cloning and knocking down of other FUTs, such as terminalα1,2-FUTs (FUT1, FUT2), a terminalα1,3/4-FUT (FUT3), terminal α1,3-FUTs(FUT4-6) and a core α1,2-FUT (FUT11)will be required for the future study.

The Notch signaling pathway has been shown to be a key regulator of many developmental processes, including somitogenesis, vasculogenesis and neurogenesis (Ma et al.,2006). Notch ligands trigger Notch signaling by binding to the extracellular domain of Notch, and this binding depends on the O-fucosylation of the EGF-like repeats in Notch. However, there was no prominent decrease in Notch signaling observed in twd^rw685 mutants, as determined by the expression level of the Notch target gene her4, hindbrain segmentation and the number of neurons (Fig. 9). These results suggest that O-fucosylation is less sensitive to the loss of GMDS compared with core and terminal fucosylation, which are significantly reduced in the twd^rw685 mutant(Fig. 7; see Fig. S1 in the supplementary material). This difference in sensitivity may be due to a smaller quantitative demand for O-fucosylation or to differences in the properties of POFUT1 and POFUT2 compared with other FUTs.

One possible reason why the Notch signal was not affected in twd^rw685 mutants is that the maternally supplied GMDS is sufficient for normal activation of Notch during the earliest stages of neurogenesis. In a preliminary study, we found that injection of an MO that inhibits the translation of maternal/zygotic gmds mRNA causes severe embryonic malformations and cell death at 24 hpf. Therefore, maternally supplied GMDS may have a crucial role in early development. Another possibility is that the GDP-fucose produced by the salvage pathway is sufficient for the O-fucosylation of Notch. In mammals, this pathway uses free cytosolic fucose as a substrate, which is derived from an extracellular source or from lysosomal degradation, converting it to GDP-fucose via fucokinase and GDP-fucopyrophosphorylase(Fig. 1). Conversely, there are no sequence-encoding enzymes involved in the salvage pathway in the Drosophila genome (Roos et al.,2002). Although it is unclear whether the teleost contains the salvage pathway, a candidate gene that encodes zebrafish fucokinaseis registered in GenBank (XM_001344236)(Fig. 1). Therefore, zebrafish may have the salvage pathway, which could account for the O-fucosylation of the Notch EGF-like repeat in the twd^rw685 mutant. However, in our preliminary study, the twd^rw685 mutants were not rescued after the injection of fucose. This suggests that even if the salvage pathway does exist in zebrafish, it is not sufficient to rescue the twd^rw685mutant because less than 10% of GDP-fucose is derived from this salvage pathway (Ma et al., 2006).

The molecular basis of the immunodeficiency in LAD II/CDG IIc can be explained by the function of fucosylated glycans in cell adhesion, mediated by an interaction between fucosylated oligosaccharides, sLe^x and the Selectin family (Becker and Lowe,2003; Ma et al.,2006). However, it remains unclear why individuals with LAD II/CDG IIc exhibit other defects, including mental retardation. The mouse and fly knock-out strains with global depression of fucosylation have been reported previously (Ishikawa et al.,2005; Smith et al.,2002). The FX knockout mouse is embryonic lethal and shows reduced expression of sialyl Lewis X antigens. Severe Notch-deficiency phenotypes were observed when GDP-fucose transporter was impaired by mutation in Drosophila. However, these studies have not explained all the mechanisms underlying the neuropathology of LAD II/CDG IIc. The twd^rw685 mutant is the first vertebrate model in which the gmds gene is affected and in which a neural defect can be demonstrated to result from aberrant migration of vagus motor neuron progenitors. This mutant provides us with an ideal model system in which to study the effects of fucosylation on neural migration. Using this model,neural migration can be visualized in vivo to evaluate the effects of gain-of-function and loss-of-function following injection of DNA constructs,mRNA or MOs, or in response to pharmacological treatment. Further analysis of the twd^rw685 mutant, the FX knock out mouse and the GDP-fucose transporter knockout fly may contribute to a greater understanding of the mechanisms responsible for the neurological disorders in individuals with LAD II/CDG IIc.