Zebrafish arl6ip1 Is Required for Neural Crest Development during Embryogenesis

Background Although the embryonic expression pattern of ADP ribosylation factor-like 6 interacting protein 1 (Arl6ip1) has been reported, its function in neural crest development is unclear. Methods/Principal Findings We found that knockdown of Arl6ip1 caused defective embryonic neural crest derivatives that were particularly severe in craniofacial cartilages. Expressions of the ectodermal patterning factors msxb, dlx3b, and pax3 were normal, but the expressions of the neural crest specifier genes foxd3, snai1b, and sox10 were greatly reduced. These findings suggest that arl6ip1 is essential for specification of neural crest derivatives, but not neural crest induction. Furthermore, we revealed that the streams of crestin- and sox10-expressing neural crest cells, which migrate ventrally from neural tube into trunk, were disrupted in arl6ip1 morphants. This migration defect was not only in the trunk neural crest, but also in the enteric tract where the vagal-derived neural crest cells failed to populate the enteric nervous system. We found that this migration defect was induced by dampened Shh signaling, which may have resulted from defective cilia. These data further suggested that arl6ip1 is required for neural crest migration. Finally, by double-staining of TUNEL and crestin, we confirmed that the loss of neural crest cells could not be attributed to apoptosis. Conclusions/Significance Therefore, we concluded that arl6ip1 is required for neural crest migration and sublineage specification.


Introduction
The neural crest (NC) contains pluripotent cells which are derived from the border of the neural and epidermal ectoderm and migrate through specific migration pathways in vertebrate embryos. NC cells contribute to the development of such cell types as facial cartilage and bones, neurons, pigment cells and glia of the peripheral nervous system, melanocytes of the skin, and smooth muscle of the heart. At the end of gastrulation, the NC arises at the border, which is also known as the neural fold, and flanks the neural plate bilaterally. During neurulation, the neural plate bends and folds into the neural tube and brings the neural folds together at the dorsal midline. During the folding process, NC precursors are contained within the neural fold. Subsequently, these cells undergo an epithelial-mesenchymal transition, which allows them to delaminate from the neuroepithelium and migrate away from the neural tube to various sites of the embryo [1,2]. NC cells from the midbrain and hindbrain regions migrate throughout the head and contribute to most of the cartilage and bones of the head, as well as neurons of cranial ganglia. The caudal hindbrain zone of the NC, also known as the vagal crest, contributes to the heart and the enteric nervous system. The trunk NC cells give rise to sensory and sympathetic ganglia [3].
Because the NC possesses the ability to generate multiple cell lineages, any abnormal development of the NC has serious repercussions for many different organs. These pathologies, known as neurocristopathies, include conditions such as Waardenburg-Shah syndrome (hypopigmentation and aganglionic megacolon), frontonasal dysplasia (multiple craniofacial defects), and DiGeorge syndrome (craniofacial and heart defects). Therefore, discovering the mechanisms of NC formation and diversification represents an important step in understanding the basis of these pathologies [4].
NC development describes the progression of a NC cell over time from the initial commitment of the cell to its specific fate to its fully differentiated function. In this paper, we are particularly concerned with NC induction and sublineage specification. NC induction refers to the induction of ectoderm cells to the NC lineage and the ability to segregate and migrate away from the neuroepithelium. In fact, studies of mouse, frog, chick and zebrafish NC development have implicated multiple signaling molecules, including BMPs, FGFs, and Wnt proteins, as well as retinoic acid and Notch, in this process [5,6,7,8,9,10,11,12,13,14]. In addition, the pre-migratory NC cells are identifiable by the expression of NC specifier genes, including slug, ap-2, foxd3, sox10 and pax3 [7,9,15,16,17,18,19,20]. However, it remains unclear whether the genes encoding the molecular signals that initiate the migration of NC cells through the embryo are the same as those that determine their ultimate fate. Thus, the specification of NC at the neural plate border and the processes involved in determining the fate of individual neural crest progenitors are questions that remain to be elucidated. Nonetheless, it is thought that the downstream effector genes resulting from the interaction between inductive signals and transcriptional regulators govern NC development.
Following this reasoning, ADP-ribosylation factor-like 6 (Arl6), which belongs to the small ADP ribosylation factor GTP-binding proteins, is a major regulator in intracellular traffic, and Arl6ip1 is an interacting protein of Arl6. In mouse, arl6ip1 is expressed in all hematopoietic cell lineages and in the early myeloid progenitor cells. Arl6ip1 protein is also dominantly localized in intracytoplasmic membranes, suggesting that arl6ip1 functions in protein transport, membrane trafficking, or cell signaling during hematopoietic maturation [21]. Lui et al. (2003) [22] demonstrated that ARMER, also named human Arl6ip1, is a novel ER integral membrane protein which plays a possible role in cell survival by protecting cells through inhibition of caspase-9 activity. Mutation of Arl6 causes Bardet-Biedl syndrome (BBS), a multisystemic disorder characterized by obesity, blindness, polydactyly, renal abnormalities, cognitive impairment and Hirschprung disease [23]. Interestingly, analysis of BBS morphants in zebrafish showed that aberrant NC migration underlies the craniofacial and enteric nervous system defects mirroring mammalian mutants [24], suggesting that BBS-associated genes might be involved in NC migration. Recently,   [25] reported that arl6ip1 is a maternal gene and is expressed ubiquitously through 24 hours post-fertilization (hpf). They also showed that knockdown of arl6ip1 by injection of arl6ip1-specific antisense morpholino oligonucleotides (MO; arl6ip1-MO1) resulted in displaying shortened mandibles with loss of pharyngeal arches and reduced pigmentation, suggesting that the NC of arl6ip1-MO1-treated embryos may be defective. By Western blot analysis,   [25] demonstrated that the protein level of Arl6ip1 is greatly reduced in the arl6ip1-MO1-injected embryos, indicating the specificity and reliability of arl6ip1-MO1 by loss-of-function studies. Taken together, this evidence led us to hypothesize that arl6ip1 may play an important role in NC development. Therefore, we investigated the function of zebrafish arl6ip1 during embryogenesis by means of MO to knock down arl6ip1 mRNA translation. Knockdown of Arl6ip1 activity resulted in reduction, or even loss, of NC derivatives, including pharyngeal arches, cranial and trunk ganglia, enteric neurons and pigment cells. While the induction of NC remained unchanged, later specification of sublineages was reduced in the arl6ip1-MO1-injected embryos. This line of evidence further implicated that arl6ip1 plays a role in the specification of NC derivatives. In addition, we also proved that the down-regulation of the pre-migratory NC markers foxd3, sox10, and snai1b could not be attributed to cell death. Rather, it is suspected that a decreased level of these genes at early somatic stage may cause major defects in the specification of NC derivatives at later stages of embryonic development. Moreover, we concluded that arl6ip1 may act parallel to, or downstream of, pax3 and msxb, but upstream of foxd3, sox10 and snai1b. This suggests that Arl6ip1 is not only required for specification of NC derivatives, but migration, as well. Finally, we demonstrated that shh signaling pathway may be disrupted by defective cilia, leading to the perturbation of NC cell migration and left-right laterality. Thus, the etiology underlying various neurocristopathies, as noted above, including craniofacial dysmorphology and Hirschsprung's disease in BBS, could very likely arise from abnormalities in arl6ip1 expression and function, as these ultimately impact NC development.

Knockdown of arl6ip1 resulted in ectoderm-derived defects of embryos
To investigate whether Arl6ip1 is required for early development, we injected embryos with arl6ip1-MO1 which specifically blocked the translation of arl6ip1 mRNA. Many morphological defects were observed in the 4 ng arl6ip1-MO1-injected embryos, compared to the control-MO-injected embryos. For example, at 24 hpf, the yolk extension and tail were foreshortened, pericardial edema was present, lens and otic vesicles were absent or highly reduced in size, and the head became smaller (Figs. 1A-C). However, since apoptosis can be induced by MO-toxicity, p53-MO was reported by Robu et al. (2007) as an excellent tool to attenuate neural cell death, but with no effect on embryonic morphology [26]. Therefore, we co-injected p53-MO to prevent nonspecific defects caused by cell death. Compared to wild-type embryos, both arl6ip1-MO1 and arl6ip1-MO1 combined with p53-MO (arl6ip1-MO1/p53-MO) morphants showed reduced cell numbers and mismigration of melanophores at 48 hpf (Figs. 1D-F). At 96hpf, the iridophore pigment cells of arl6ip1-MO1 morphants, as well as arl6ip1-MO1/p53-MO-injected embryos, were much reduced in comparison to wild-type embryos (Figs. 1G-I, indicated by arrows).. The jaw morphology shown in wild-type embryos at 96 hpf (indicated by double arrowhead in Fig. 1J) did not form in arl6ip1-MO1-injected and arl6ip1-MO1/p53-MOinjected embryos (Figs. 1K and L). The occurrence rates of reduced iridophore cells and melanophores in the 4 ng arl6ip1-MO1-injected embryos and in the 4 ng arl6ip1-MO1/6ng p53-MO-injected embryos were 84.5% (93/110 ) and 79.5% (128/ 161), respectively. The head size and trunk extension were also recovered in arl6ip1-MO1/p53-MO-injected embryos at 24 hpf (data not shown). Since the off-targeting effect was inhibited by p53-MO, no NC derivatives could be rescued in the arl6ip1-MO1/ p53-MO-injected embryos. We also injected 8 ng of arl6ip1-MO2 into embryos, and the resulting defects phenocopied those of arl6ip1-MO1-injected embryos (data not shown) [25]. These findings suggested that knockdown of Arl6ip1 function causes defects related to NC degeneration.
The occurrence rate of defects induced by injection of arl6ip1-MO1 was dose-dependent (Table 1). We observed that higher concentrations of injected arl6ip1-MO1 resulted in a correspondingly higher rate of defects. However, when 4 ng arl6ip1-MO1 were co-injected with 150 and 200 pg of a wobble arl6ip1 mRNA in embryos, the occurrence rates of mild defects were observed to decrease from 48.4% to 32.1% and 17.6%, respectively, while the occurrence rates of the wild-type-like phenotype increased from 16.1% to 56% and 77%, respectively (Table 1). Furthermore, coinjection of 4 ng arl6ip1-MO1/p53-MO with 75 and 150 pg of a wobble arl6ip1 mRNA in embryos decreased the occurrence rates of mild defects from 74% to 34.5% and 17.5%, respectively, but raised the occurrence rates of wild-type-like phenotype from 14.4% to 60.2% and 80.1%, respectively (Table 1, Fig. S1K). In addition, we co-injected a wobble arl6ip1-gfp mRNA with arl6ip1-MO1, and the Arl6ip1-GFP fusion protein was detected in embryos (Figs. S1B and F), indicating that the injected arl6ip1-MO1 failed to inhibit the translation of the introduced wobble arl6ip1-gfp mRNA. Furthermore, either co-injection of arl6ip1-MO1 or arl6ip1-MO1/p53-MO with a wobble arl6ip1 mRNA rescued the defective phenotypes in embryos induced by arl6ip1-MO1 (Figs. S1B and C, Table 1). To further confirm whether phenotypes of these morphants were specifically induced by the absence of Arl6ip1 function, we constructed a plasmid of pCS2-Arl6ip1-GFP, in which the targeting sequence of arl6ip1-MO1 was fused with the GFP cDNA. The GFP was expressed in the embryos injected with 4 ng arl6ip1-MO1 together with 100 pg GFP mRNA (n = 112), whereas the expression of GFP was totally absent in the embryos injected with 4 ng arl6ip1-MO1 together with 100 pg arl6ip1-MO1-target-gfp RNA synthesized from pCS2-Arl6ip1-GFP (n = 115) (Figs. S1D and H). This line of evidence indicated that arl6ip1-MO1 was capable of specifically blocking the translation of arl6ip1 mRNA and that the defects induced by arl6ip1-MO1 were specific and could be rescued by wobble arl6ip1 mRNA.

Pharyngeal arch development is abnormal in arl6ip1 morphants
When Arl6ip1 was knocked down, the mandible was reduced, as mentioned above. The majority of the head skeleton is derived from the cranial neural crest cells which migrate to the branchial arches and frontonasal process. Therefore, we speculate that the shortened mandible of arl6ip1 morphants may have resulted from defective craniofacial cartilages. To test this hypothesis, we performed Alcian blue staining to observe the craniofacial cartilage of the 4-dpf embryos. The craniofacial cartilages, such as mandibular (1st arch), hyoid (2nd arch) and ceratobranchial (3rd-7th arches), which are all derived from three streams of cranial NC [27], were all clearly shown ( Fig. 2A). In contrast, no cartilages for any of the arches were apparent in either the arl6ip1-MO1-injected embryos or arl6ip1-MO1/p53-MO-injected embryos (Figs. 2C and 2D). The occurrence rate of neurocranium defects, either in the 4 ng arl6ip1-MO1-injected embryos or in the 4 ng arl6ip1-MO1/6 ng p53-MOinjected embryos, was 100% (133/133 in arl6ip1-MO1 embryos; 146/146 in arl6ip1-MO1/ p53-MO embryos). Moreover, severe shortening of trabeculae and complete loss of ethmoid plate (Fig. 2C), which are from anterior neurocranial cartilage, were also observed in the arl6ip1-MO1injected embryos. However, wild-type embryos displayed a wellorganized anterior neurocranial cartilage (Fig. 2B). It should be noted that both of these cartilages were derived from NC for the development of neurocranium. Finally, arl6ip1-MO1/p53-MOinjected embryos exhibited the same craniofacial phenotypes as the arl6ip1-MO1-injected embryos (Fig. 2D), except for the larger head size. Furthermore, the severe craniofacial defect, which is shown in Figures 2D, could be rescued in wobble arl6ip1 mRNAinjected embryos (Fig. 2H). Since the co-injection of p53-MO was used to rule out nonspecific effect by MO-toxicity [26], we are satisfied that these arl6ip1-MO1/p53-MO co-injection data indicated that the complete loss of both pharyngeal arches and ethmoid plate with shortened trabeculae could not be attributed to either cell death or the nonspecific effects of arl6ip1-MO1. Instead, Arl6ip1 was clearly demonstrated to function in NC derivative arches. Schilling and Kimmel (1997) [27] demonstrated that dlx2 is required for jaw development and is expressed by NC cells, which contribute to the pharyngeal arches, as well as by migratory archassociated NC cells. While we found a marked increase of dlx2 expression in the migratory NC pharyngeal arch precursors in wild-type zebrafish embryos (Fig. 2E), we also found that dlx2 was markedly reduced in the migratory NC cells destined for the posterior arches of the arl6ip1 and arl6ip1/p53 morphants (Figs. 2F and G), suggesting that reduced dlx2 expression in the morphants is responsible for the defective craniofacial cartilages. Further-more, the expression pattern of dlx2 in the forebrain of arl6ip1-MO1-injected embryos was similar to that of wild-type embryos, indicating that the differentiation in the forebrain was normal.

NC derivatives are reduced in arl6ip1-deficient embryos
The occurrence of defective cartilages in the arl6ip1-MO1injected embryos strongly suggested that arl6ip1 may play a role in NC development. To test this hypothesis, we used an anti-Hu immunofluorescence assay to examine the cranial ganglia Table 1. Injection of wobble arl6ip1 mRNA enables embryos to rescue the defects induced by injection of arl6ip1-MO1/p53-MO. The morphological defects of embryos were observed at 36 hpf. Severe defects are defined as those where the brain did not form sulcus and gyrus in the head and where the axis appeared as a curved shape in the trunk of embryos. Mild defects are defined as those where the embryonic axis was longer than observed in embryos with severe defects, but where head defects were just as severe. The arl6ip1 mRNA used in this study was synthesized as a wobble form. doi:10.1371/journal.pone.0032899.t001 surrounding the otic vesicles or the dorsal root ganglia and enteric neurons in the trunk, since these ganglia have NC origin. Results showed a severe reduction of neurons in epibranchial ganglia, dorsal root ganglia and enteric ganglia, and a mild reduction of neurons in trigeminal ganglia, either in the arl6ip1 or arl6ip1/p53 morphants (Figs. 3B, C, E and F). The occurrence rates of cranial ganglia, dorsal root ganglia and enteric neuron defects in the 4 ng arl6ip1-MO1-injected embryos were 68.3% (71/104), 72.5% (132/ 182) and 73.8% (48/65), respectively. These data indicated that cranial, vagal and trunk neurons, which are derived from the NC, were defective in the arl6ip1-MO1-injected embryos. We also detected a large number of Hu-positive cells in the head of arl6ip1-MO1-injected embryos (Fig. 3B). To explain this observation, we suggest that antibody penetration into the brain of morphants is more efficient, making brain neurons more visible.

Arl6ip1 is required for the specification of NC sublineages
To study the function of arl6ip1 during NC development, we first detected the expressions of the ectodermal patterning factors wnt8, bmp4, pax3, and msxb. The ectodermal patterning factors were all normally expressed in the arl6ip1-MO1injected embryos (Figs. 4A and F; 4B and G; 4C and H; 4D and I). In addition, the expression of the pre-placodal marker dlx3b was also normal in the arl6ip1-MO1injected embryos (Figs. 4E and J, n = 64). Furthermore, since Trainor and Krumlauf (2000) [28] reported that segmentation and patterning of the hindbrain are necessary for cranial NC development; we also examined the expressions of two hindbrain patterning markers, fgf3 and fgf8. Results showed that fgf3 and fgf8 were also expressed normally in the hindbrain of arl6ip1 morphants (Figs. S2A-D, n = 87) and that 83.2% of morphants exhibited a normal hindbrain segmentation at 24 hpf (Figs. S2E-F, n = 101). These findings suggested that the induction of both early NC progenitors and placodal region cells results in normal development of arl6ip1 morphants. The early patterning and later segmentation of the hindbrain in the arl6ip1-MO1injected embryos were also similar to those of wild-type. Thus, even though arl6ip1 may not function in hindbrain development, it is expressed ubiquitously until 24 hpf, and knockdown of Arl6ip1 did cause small head morphology. This line of evidence suggests that neural crest defects in arl6ip1-MO1 embryos did not result from abnormal head development. On the other hand, when we observed the expressions of pre-migratory NC markers in the progenitors of the neural plate border, such as tfap2a, foxd3, snai1b, and sox10, we found that the expressions of foxd3, snai1b and sox10 were greatly reduced in the arl6ip1 morphants at 3-somite (Figs. 5B and F; 5C and G; 5D and H) through 8-somite stages (Figs. 5I and L; 5J and M; 5K and N). Meanwhile, tfap2a expression in NC progenitors was present in both wild-type and arl6ip1-MO1 embryos (Figs. 5A and E). Double-staining of sox2 (labeling neural plate) and dlx3b (labeling the pre-placodal region) exhibited a patch of unlabeled cells between these two expression domains which was presumed to be the NC. Thus, since the induction of presumptive NC was normal in arl6ip1-MO1 embryos (Fig. S3), we concluded that NC induction occurred normally in arl6ip1 morphants. The occurrence rates of down-regulation of foxd3, snai1b and sox10 in the arl6ip1-MO1-injected embryos were 75.6% (31/41), 82.2% (37/45) and 66.7% (30/45), respectively. Again, since down-regulation of these genes could have resulted from early cell death by the off-targeting effect of MO, we injected 6 ng p53-MO together with 4 ng arl6ip1-MO1 to inhibit apoptosis and nonspecific effects. We found that the level of cell death was dramatically reduced in the embryos injected with p53-MO. In addition, Cole and Ross (2001) [29] have detected the temporal and spatial distribution of apoptotic cells during normal develop- ment of the zebrafish embryo from 12 to 96 hpf. According to their data, only a few apoptotic cells could be seen in the optic vesicle at 12hpf, and no apoptosis was observed in the premigratory neural crest cell region. Robu et al. (2007) [26] also demonstrated that about 15-20% of MO used in zebrafish show off-targeting effects, as represented by a signature neural death peaking at the end of segmentation (1 dpf). They also showed that the neural cell death induced by MO was observed as early as 14 hpf. Moreover, Ekker and Larson (2001) [30] indicated that the onset of cell death induced by MO technology was 14 hpf. In our data, we detected pre-migratory neural crest markers at the 3somite stage (about 10.33 hpf), even earlier than 12 hpf; consequently, we did not observe off-targeting effects at the 3somite stage. On the other hand, when p53 was inhibited, cell survival could not rescue expressions of foxd3, snai1b and sox10 at the 8-somite stage (Figs. 5L-N). Therefore, we concluded that the down-regulation of these gene markers in existing NC cells was caused by defects in NC derivate specification, not cell death. Furthermore, such defects subsequently resulted in the reduction of the cell population, especially in pharyngeal arches of cranial NC sublineages. Finally, we found that the reduced expression of foxd3 induced by arl6ip1-MO1 could be restored by injection of wobble arl6ip1 mRNA in arl6ip1-MO1-injected embryos (Figs. S4A-C). These findings implicated that Arl6ip1 plays a key function in specification of NC sublineages, but not induction.
Trunk NC migration is defective, and cranial NC is reduced in the arl6ip1 morphants, as co-tested with myod staining In wild-type zebrafish embryos, the migration of cranial NC has three distinct streams: migration from rhombomeres (r) r2, r4 and r6 to mandibular (1st arch), hyoid (2nd arch) and ceratobranchial (3rd-7th arches), respectively. We were able to detect these migratory NC streams by crestin probe (Fig. 6A). In the arl6ip1-MO1-injected embryos at 24 hpf, the crestin expression in the cranial crest was reduced in the 3rd NC stream, and crestin was almost absent in the 1st and 2nd stream, as shown in Figure 6D. The reduced expression of crestin in cranial NC cells resulted in defective craniofacial cartilages, which also was proved by previous data of dlx2, as shown in Figure 2. The occurrence rate of reduced defects in the crestinlabeled cranial NC was 73.4% (47/64). However, the trunk NC cells in the arl6ip1 morphants appeared to have defects in NC migration. Because severe trunk extension and somite defects could indirectly affect trunk neural development, such as cell migration, we used arl6ip1 and arl6ip1/p53 morphants with only mild defects impeding normal trunk development to analyze the migration of trunk NC. Compared to wild-type embryos, there was a significant delay in the onset of migration of the arl6ip1 morphants, which had only a small number of ctn-expressing cells migrating within the most anterior somites (Figs. 6B and E). Similarly, the sox10 gene, which is also expressed in migrating trunk NC cells [31], showed a delay in NC migration in the arl6ip1-MO1-injected embryos at 24 hpf (Figs. 6C and F). Compared to wild-type embryos, the cell numbers in trunk NC of arl6ip1-MO1-injected embryos were normal, but these cells could not migrate. The occurrence rates of migration defects in the sox10and crestin-labeled trunk NC were 72.7% (40/55) and 75% (45/60), respectively. Co-injection of arl6ip1-MO1 with p53-MO in embryos showed the same migration defects as the arl6ip1-MO1injected embryos (Figs. 6H and I). To quantify the number of streams of crestin-positive NC cells which migrate through the trunk, we observed that the inhibition of migrating streams of NC cells from the dorsal neural tube was exhibited in the arl6ip1 morphants (Fig. S4G). Again, injection of a wobble arl6ip1 mRNA rescued embryos from these defects (Figs. S4D-F and S4G). When we traced the expressions of crestin and sox10 at 30 hpf, we found that the streams of sox10and crestin-labeled cells in the arl6ip1 morphants still could not migrate normally when compared to the wild-type embryos (data not shown). This evidence revealed that trunk NC cells remained localized between the dorsal surface of the neural keel and overlying ectoderm in the arl6ip1-MO1-injected embryos (Figs. 6E and F). In summary, the defects shown in dorsal root ganglia clearly resulted from the inhibition of trunk NC migration.
To further understand why enteric neurons were absent in the arl6ip1-deficient embryos, as shown in Figure 3D, we examined vagal NC migration. We observed embryos at 36 hpf and found that the crestin-labeled NC cells of wild-type embryos migrated out of the vagal region (Fig. 7A). However, the crestin-labeled NC cells of arl6ip1-MO1-injected and arl6ip1-MO1/p53-MO-injected embryos did not migrate normally and showed decreased progenitor cell numbers when compared to the wild-type (Figs. 7B and C). Additionally, we observed that the enteric nervous system (ENS) precursors labeled by phox2b were distributed along the entire length of intestine from 48 to 72 hpf in wild-type embryos (Figs. 7D and G). In contrast, these phox2b-positive ENS precursors only migrated to the anterior part of the intestine in arl6ip1 morphants and arl6ip1/p53 morphants (Figs. 7E and F; 7H and I). The occurrence rate of migration defects in the phox2b-labeled enteric neurons was 88.7% (63/71). We suggested that this migratory delay and partial loss of enteric precursors would affect normal differentiation of enteric neurons and might lead to loss of the ENS.

Disruption of zebrafish Arl6ip1 function resulted in cilia defects and dampened Hedgehog signaling
During migration, Shh directs migration and patterning of NC cells. Tobin et al. (2008) [24] indicated that knockdown of BBS disrupted the Sonic Hedgehog (Shh) signaling pathway and affected NC cell migration. Therefore, we investigated Shh signaling in arl6ip1 morphants. The results showed that shh expression was unaltered in arl6ip1-MO1/p53-MO-injected embryos (Fig. S5B); however, pax6, normally repressed by Shh signaling, was up-regulated and ectopically expressed throughout the neural tube in arl6ip1-MO1/p53-MO-injected embryos (Fig.  S5D). The occurrence rate of overexpressed pax6 in the in the neural tube was 78.4% (69/88). We therefore concluded that the Shh signaling pathway might be perturbed in arl6ip1-MO1/p53-MO-injected embryos. In addition, Shh signaling reveals a cell autonomous requirement for Hh signal transduction and is directly required for normal development of DRG neurons in zebrafish [32]. Thus, the defects shown in dorsal root ganglia (Fig. 3) clearly resulted from disrupted Shh signaling pathway. Previous study demonstrated that cilia are required for normal Hh signaling [33]. Furthermore, overexpression of GDP-or GTPlocked variants of ARL6/BBS3 in a ciliated cell line, namely retinal pigmented epithelial cells (hTERT-RPE), influences primary cilium length and abundance [34]. Therefore, we additionally determined whether cilia of arl6ip1-MO1/p53-MO-  injected embryos were normal. Kupffer's vesicle (KV) contains abundant, evident and aggressive cilia in zebrafish. By confocal immunostaining of acetylated tubulin in 8-somite stage embryos, apical cilia in the KV were shortened and reduced in number in arl6ip1-MO1/p53-MO-injected embryos compared with the wildtype embryos (Figs. S6I and J). The occurrence rate of reduced KV was 51.6% (49/95). Since zebrafish KV plays a role in leftright patterning [35], we tested whether cilia defect in KV resulted in laterality defects by assaying expression of two conserved leftright genes, southpaw (spaw) and pitx2. Significantly, arl6ip1-MO1/ p53-MO-injected embryos showed randomized expression of spaw and pitx2 (Figs. S6A-H and S6K). Finally, following the method reported by Amack and Yost (2004) [35], we injected fluorescencetagged arl6ip1-MO1 into embryos between 2 to 4 hpf and selected embryos whose MOs had accumulated into dorsal forerunner cells (DFC) to determine if Arl6ip1 was required for KV morphogenesis. The occurrence rate of mid-blastula-injected MOs, which were incorporated into DFC, was 42% (n = 435). The results showed that DFC of arl6ip1-MO1-injected embryos displayed reduced KV (65%, n = 133) and even failed to form KV (23%, n = 133). Thus, cilia were disorganized in the KV of DFC of arl6ip1-MO1-injected embryos (Fig. S7). Taken together, we demonstrated that Arl6ip1 is required for KV morphogenesis and that shh signaling may be perturbed by defective cilia, in turn, leading to the perturbation of NC cells migration and left-right laterality.

Loss of Arl6ip1 function causes cell death in embryos, but not in NC at later stage
In arl6ip1-deficient embryos, pharyngeal arches, dorsal root ganglia and enteric neurons at the later stages of embryonic development were severely reduced or absent. The loss of these cells during development might result from cell death. To investigate the role of Arl6ip1 in NC survival, TUNEL assay was performed to examine cell death. At the 3-somite stage, there were no significant TUNEL signal differences in cell death between wild-type and arl6ip1-MO1-injected embryos (Figs. 8A and D). However, we detected a great increase of TUNEL-positive cells around the dorsal ectoderm in arl6ip1-MO1-injected embryos from the 3-to 15-somite stages, compared to wild-type embryos (Figs. 8B and E; 8C and F). The expression of pre-migratory NC markers was also severely reduced in the arl6ip1-MO1 embryos between the 3-and 8-somite stages (Fig. 5). We detected a great increase of TUNEL-positive cells around the dorsal ectoderm in arl6ip1-MO embryos from the 3-to 15-somite stages, and this increase ended at 36 hpf. To confirm whether these dying cells were overlapped with crestin-positive NC cells, we performed double-labeling with crestin and fluorescent TUNEL (Figs. 8G-I; G9-I9). The significant expression of crestin in wild-type embryos indicated that cranial NC appeared at the anterior region of trunk somites at this stage (Figs. 8G and G9, indicated by arrows). When we compared dying cells with crestin-positive NC cells using double-labeling assay in arl6ip1-MO1-injected embryos, apoptotic cells tended to localize at the more anterior and superficial region, while crestin-positive NC cells were inclined to localize at the posterior region (Figs. 8H and H9). Since cell death occurring in arl6ip1-MO1-injected embryos was inhibited by p53-MO, cell death was observed to be correspondingly decreased in the arl6ip1-MO1/p53-MO-injected embryos (Fig. 8I). Compared with wildtype embryos, we also found the number of crestin-positive cells in arl6ip1-MO1-injected embryos and arl6ip1-MO1/p53-MO-injected embryos to be reduced (Figs. 8G9 vs. 8H9; 8G9 vs. 8I9, indicated by brackets). In addition, most crestin-positive cells failed to show cell death signaling in arl6ip1-MO1-injected embryos (Figs. 8H and H9). To account for the overlapping signals of crestinand TUNEL-positive cells under fluorescent stereomicroscopy, we found an average of 9.6, 6.7 and 1.9 out of 10 cells to be positive to both TUNEL and crestin signals in arl6ip-MO1-injected embryos, wild-type embryos, and arl6ip1-MO1/p53-MOinjected embryos, respectively. However, the appearance of TUNEL-positive cells did not correlate with most crestin-labeled cells at this stage, which could be attributed to other defects in the arl6ip1 morphants. Thus, we concluded that cell death does not play a role in the NC survival of arl6ip1-MO1-injected embryos.

Discussion
In this study, we are the first to investigate the novel function of Arl6ip1 in NC development of zebrafish. Specifically, by using arl6ip1-specific MO, we demonstrated that NC induction occurs normally, but Arl6ip1 is required for specification of sublineages within the NC. In the arl6ip1-deficient embryos, we found that the expressions of dlx2 and crestin in three streams of cranial NC are almost absent, resulting in marked craniofacial defects. Moreover, these defects are not confined to the craniofacial region. Rather, vagal-derived NC cells also fail to populate the enteric nervous system of arl6ip1-deficient embryos. We also showed a significant delay in the onset of migration in the trunk, and we demonstrated that such delay causes defects in dorsal root ganglia and synthesis of pigment cells in the arl6ip1 morphants. Finally, the defects of arl6ip1 morphants could be restored by injecting wobble arl6ip1 mRNA. However, the degree of rescue by increasing arl6ip1 mRNA concentration alone is not as effective as injecting arl6ip1 mRNA/p53-MO. While it was concluded that arl6ip1-MO1 might have some off-targeting effects in embryos, arl6ip1-MO1/p53-MO co-injection could prevent these nonspecific effects. Therefore, all MO experiments carried out in this study used arl6ip1-MO1/p53-MO co-injection, except those unaltered genes in arl6ip1-MO1 embryos. Most importantly, our results clearly demonstrated that Arl6ip1 plays a key functional role in NC migration and specification of sublineages, which points to the likelihood of its association with Hirschsprung's disease and with the aberrant craniofacial characteristics of BBS.
In mouse, arl6ip1 is expressed in all hematopoietic cell lineages, and Arl6ip1 protein is localized to intracytoplasmic membranes [21]. By analysis of the cellular location of arl6ip1 expression, Pettersson et al. (2000) [21] suggested that arl6ip1 may play a role in protein transport, membrane trafficking, or cell signaling during hematopoietic maturation. Lui et al. (2003) [22] also demonstrated that ARMER/Arl6ip1 is a novel ER integral membrane protein which protects cells by inhibiting caspase-9 activity, thereby revealing a possible role for ARMER in cell survival. Interestingly, in this study, we demonstrated that zebrafish arl6ip1 is expressed ubiquitously through 24 hpf and plays a critical role in NC migration and specification of derivatives, rather than hematogenesis. By loss-of-function studies, the results presented in this report strongly support the hypothesis that zebrafish Arl6ip1 performs key functions in NC development, including derivates specification and migration, but not induction. Since Arl6 is a major regulator involved in BBS, the implications for human disease become obvious, particularly as they relate to neuropathic disorders.
In a previous report, genetic studies showed that foxd3 is expressed normally in the pre-migratory neural crest cells of the sox10 (colorless) mutant [36], suggesting that foxd3 may function as an upstream effector of sox10. This hypothesis is supported by Kelsh and Eisen (2000) [36], Elworthy et al. (2005) [37] and Stewart et al. (2006) [38], who demonstrated the effect of foxd3 on the sox10-dependent genetic pathway, where foxd3 mutants (sym1 mutants) appeared to show a more severe neural crest phenotype than sox10 mutants, with the exception of chromatophores. Therefore, based on the genetic analysis presented in this study, it is plausible to theorize that arl6ip1 may regulate the specification of NC derivatives through activation of other pre-migratory NCspecific transcription factors, such as sox10 or foxd3.
Another mutant found to affect foxd3 function is mosm188. A complementation analysis with the sym1/foxd3 mutant found that the two mutations affect the same locus. Sequencing the foxd3 coding region in mosm188 mutants did not display any nucleotide changes, suggesting that mosm188 most likely impairs either foxd3 mRNA expression or stability. Furthermore, since mosm188 mutants express foxd3 ectopically in somites or in gastrula, but not in NC cells, it is likely that the mosm188 mutation knocks out an NC-specific regulatory element within the foxd3 locus [39]. This mutant also showed the same depletion of NC derivatives as sym1 mutation, including craniofacial cartilage, sympathetic neurons, dorsal root ganglia, enteric neurons, and pigment cells. Moreover, it was also reported that the foxd3 gene might act to maintain the undifferentiated state of NC progenitors during embryogenesis. In the absence of Foxd3, it is thought that some NC cells prematurely differentiate, thus producing fewer progenitor cells and leading, in turn, to the normal development of fewer derivatives. Although Arl6ip1 may act upstream of Foxd3 and Sox10 (Fig. 6), knockdown of Arl6ip1 still reduces the number of both sox10and foxd3-expressing NC progenitors. Thus, Arl6ip1 knockdown causes more serious developmental defects in the reduction, or absence, of NC derivatives than sox10 and foxd3 mutants.
Furthermore, since Snail family members can affect early NC development in other species [40,41,42], the severe reduction of snai1b expression in arl6ip1 morphants in NC subpopulations may also contribute to the neuronal and glial defects. However, since snai1b function has not yet been uncovered in zebrafish, it remains to be elucidated whether the snai1b gene also differentially regulates NC specification.
We found that both abnormal trunk extension and somite defects were induced by arl6ip1-MO1. Thus, it is possible that arl6ip1-MO1-injected embryos undergo mispatterning of somitic regions, causing a significant loss of cells in these regions that would otherwise provide guidance or survival cues to the trunk migrating NC cells. These defects could also be indirectly affected by trunk neural development, such as cell migration. Furthermore, Robu et al. (2007) [26] stated that p53 knockdown by itself did not induce any significant defects, as p53 is not required for normal development in mammals or fish (see Figures 1B and 2B of Robu et al., 2007). In the present study, p53-MO did not affect the efficacy of gene-specific MOs, as it did not interfere with the penetrance of gene-specific phenotypes. Our findings also clearly showed that head size and trunk extension were recovered in arl6ip1-MO1/ p53-MO-injected embryos at 24 hpf. In addition, we also confirmed that myod was expressed normally, either in wild-type or arl6ip1-MO1/p53-MO-injected embryos, at 24 hpf (data not shown). Thus, we used either arl6ip1 or arl6ip1/p53 morphants with only mild defects impeding normal trunk development and co-tested with myod staining to analyze the migration of trunk NC. Finally, we demonstrated that trunk NC migration, either in arl6ip1 or arl6ip1/p53 morphants, is defective.
The regulation of cell death by Arl6ip1 may be a conserved function. Lui et al. (2003) [22] gave evidence of this essential role in cell survival by induced overexpression of human Arl6ip1 in cell line HT1080. Briefly, their work demonstrated that human Arl6ip1 plays a possible role in cell survival by inhibiting caspase-9 activity. Recent studies have shown that Snail family members also play a role in the survival of progenitor cell populations and are overexpressed in some human cancers [43,44,45]. In our studies, however, double-labeling with crestin and TUNEL in arl6ip1-MO1 embryos showed that just a few of the dying cells were crestin-positive. Thus, zebrafish Arl6ip may play a role in cell death, but not in NC survival. Because trunk NC cells were localized between the dorsal surface of the neural keel and overlying ectoderm in the arl6ip1-MO1 embryos, defects in dorsal root ganglia clearly resulted from the inhibition of trunk NC migration. Similarly, reduced numbers of dorsal root ganglia are to be found in the zebrafish sym1 mutant. Stewart et al. (2006) [38] demonstrated that foxd3 functions to specify neuronal and glial sublineages and is required for the timing of trunk NC migration. Furthermore, decreased expression of sox10, snai1b and ctn was evident in sym1 mutants. It is worthwhile noting that down-regulation of early NC specifiers may also affect the migration of the NC cells at the later embryonic stage. This result is also consistent with recent studies in chick showing that foxd3 function is required for NC migration by regulating the expression of cell-cell adhesion molecules [46,47,48]. Therefore, the results of our zebrafish study are in agreement with those findings in chick and sym1 mutants since we found that foxd3 retains sufficient expression in the somite of zebrafish embryos at 30 hpf to inhibit trunk NC migration (data not shown).
The early NC phenotype reported here is severe at the 3-somite stage (Fig. 5), yet the NC phenotype reported later in development ( Fig. 6) appears much less severe. We do not think that the NC phenotypes, which are much less severe in later development, as shown in Figure 6, are a consequence of developmental delay. All the NC sublineages are derived from the pre-migratory cell pool at the 3-somite stage, but no clear region can be found indicating localization of the pool of pre-migratory cells that might contribute to the subset of NC sublineages. However, based on the defective NC derivatives which occurred in the arl6ip1-MO1-injected embryos, we learned which NC sublineages are affected. Consequently, in this study, we hypothesized that the subset of NC sublineages of trunk was less affected by the knockdown of Arl6ip1 in the pre-migratory NC pool. In contrast, the subsets of vagal and cranial NC cells were greatly reduced in the premigratory NC pool of arl6ip1-MO1-injected embryos. Therefore, at the later developmental stages, development of both vagalderived enteric neurons and cranial-derived pharyngeal arches was highly reduced in cell numbers. On the other hand, the trunk NC cells appeared much less affected in terms of cell number. Knecht and Bronner-Fraser (2002) [49] reported that Wnt, Bmp2/4 and TGFb 1, 2 and 3 signaling pathways induce multipotent NC progenitor cells and that the newly specified premigratory NC progenitors express a unique set of transcription factors, such as Foxd3, Snai1b, Id2, Sox10 and Tcfap2a. Subsequently, the NC cells turn off early markers, like Sox10, up-regulate a new set of genes, and begin to migrate. At this stage, NC cells derived from the hindbrain migrate as a wave in a caudal direction. The first group of cells starting to migrate includes the cells populating in the pharyngeal arches. This first group of cells is followed by a second wave of migration which supplies cells to trunk NC [50]. In Figures 5F-H, we can clearly observe that the pre-migratory NC of midbrain-hindbrain boundary in the arl6ip1-MO1-injected embryos was only slightly affected. On the other hand, the trunk NC cells could not migrate ventrally in the arl6ip1-MO1-injected embryos. However, the proliferation of trunk NC cells may be normal, resulting in non-migrated trunk NC cells populating the dorsal neural tube. This continuous proliferation of trunk NC cells recovers the loss of the NC cells in the early developmental stages. Actually, the subsets of vagal and cranial NC cells do not recover at all. These findings might plausibly explain why the trunk NC cell numbers were less severe.
Arl6 is the gene underlying BBS type 3 [23]. Analysis of BBS morphants in zebrafish showed that aberrant neural crest migration underlies the craniofacial and enteric nervous system defects mirroring mammalian mutants [24], suggesting that BBSassociated genes might be involved in NC migration. In our studies, arl6ip1 morphants exhibit similar defective phenotypes, such as craniofacial dysmorphology and inhibition of trunk NC migration. We also found the loss of enteric neurons in arl6ip1 morphants. Therefore, we concluded that Arl6ip1 might function in NC cell development through interaction with Arl6. In addition, by loss-of-function studies, our results showed that knockdown of arl6ip1 negatively affects NC development, indicating, in turn, that arl6ip1 could be an underlying etiological factor in the onset of BBS and other neurocristopathies.

Fish husbandry and observation
The wild-type AB strain of zebrafish was cultured at 28.5uC. Embryos were staged by hpf [51]. The phenotype of heart formation was observed under a fluorescent stereomicroscope, MZ FLIII (Leica). Images were captured with a Fine pix S2 pro camera (Nikon) using the Camera Shooting software.

MO knockdown
The MOs were specifically designed to inhibit the translation of arl6ip1 mRNA as follows: fluorescence-tagged arl6ip1-MO1 and arl6ip1-MO1, 59-ACTTTTATTGT-CGCCCTCAGCCATG-39, and arl6ip1-MO2, 59-GATGTTACTTGAGAGTTTAGG-TTCC-39. A control MO specific for Arl6ip1 was designed as 59-GTACCGACT-CCCGCTGTTATTTTCA-39, which is an inverted sequence of arl6ip1-MO1. We also designed p53-MO, 59-GCGCCATTGCTTTGCAAGAATTG-39, to suppress the cell death induced by MO off-targeting. The amount of p53-MO injected was 1.5-fold (w/w) higher than the amount of MO previously described [26]. All MOs were prepared at a stock concentration of 1 mM and diluted to the desired concentration for microinjection into each embryo.
Whole-mount in situ hybridization and TUNEL assay Digoxigenin (DIG)-labeled antisense RNA probe used for whole-mount in situ hybridization was generated by using DIG RNA labeling kit (Roche). Embryos at the desired stage were collected and then fixed with 4% paraformaldehyde for 4 h at 25uC. After fixation, embryos were dechrionized, dehydrated in absolute alcohol, and stored at 220uC. Embryos were rehydrated by immersion for 10 min each in 75%, 50%, and 25% ethanol. Rehydrated embryos were transferred to phosphate buffer saline (PBS). Prehybridization, hybridization, and detection procedures were performed with a DIG detection kit II, according to the instructions in the supplier's manual (Boehriner Mannheim). The TUNEL assay was described previously [52] using The Dead-End TM Colorimetric TUNEL System (Promega) or DeadEnd TM Fluorometric TUNEL System. The overlapping signals of crestinand fluorescent TUNEL-positive cells were detected and counted under a fluorescent stereomicroscope, MZ FLIII (Leica).

Plasmid construction
In order to further demonstrate the specificity of MO targeting, we constructed plasmids of pCS2-Arl6ip1-GFP, in which the binding sequence of arl6ip1-MO1 was fused with the GFP reporter cDNA. The NotI-cut pCS2-Arl6ip1-GFP served as a template to synthesize RNA using the mMessage Machine kit (Ambion). The resultant arl6ip1-MO1-target-gfp RNA was injected at 100 pg per embryo or co-injected with 4 ng arl6ip1-MO1 into one-celled zebrafish embryos. Appearance of the GFP signal in the treated embryos was observed at the 24-hpf stage using green fluorescent microscopy.

Rescue experiment
Since the introduced arl6ip1 mRNA was not bound by arl6ip1-MO1 during the rescue experiment, we designed a wobble arl6ip1 and arl6ip1-gfp mRNA. The wobble arl6ip1-gfp mRNA was a wobble arl6ip1 cDNA fused in frame with egfp cDNA, and we changed the nt 125-149 of zebrafish arl6ip1 cDNA (GenBank Accession NO. NM_201112) from 59-ATGGCTGAGGGCGACAATAAAAGTG-39 to 59-ATGGCcGAaGGaGAtAAcAAgAGcG-39, but without altering the amino acid residues. Capped mRNA transcripts of wobble arl6ip1 and wobble arl6ip1-gfp for rescue experiments were synthesized by SP6 in vitro transcription, according to the protocol of the manufacturer (Epicentre). The resultant mRNAs were diluted to the desired concentration (from 100 to 300 pg/ml), and approximately 2.3 nl were microinjected into one-celled stage embryos.

Whole-mount immunofluorescence staining
Anti-Hu antibodies were used to detect cranial ganglia, enteric neurons, and dorsal root ganglia, as described by O'Brien et al. [53]. After embryos were fixed in 0.4% paraformaldehyde at pH 7.2 for 4 hr at room temperature, they were washed three times with PBS for 15 min each, soaked in 100% acetone at 220uC for at least 10 min, rehydrated through graded methanol in PBS and placed in blocking solution at room temperature for 1 hr (5% sheep serum, 1% bovine serum albumin (BSA), 1% dimethylsulfoxide and 0.1% Tween 20 in PBS). After blocking, the embryos were subjected to primary antibody at 4uC overnight. The embryos were then extensively washed with 1% BSA in PBST and incubated at 4uC overnight with gentle mixing in goat antimouse immunoglobulin conjugated with Cy3, 2 mg/ml diluted 1:50 in blocking solution. The embryos were washed extensively in 1% BSA in PBST and finally in PBS.