The neural crest is a group of embryonic precursor cells that arise from the edge of the neural ectoderm during development, migrate throughout the embryo, and give rise to a wide array of cell types, including most neurons and glia of the peripheral nervous system (PNS), craniofacial cartilage and bone, pigment cells, and smooth muscle cells (LaBonne and Bronner-Fraser, 1999; Le Dourain and Kalcheim, 1999; Christiansen et al., 2000). For many years, the development of the neural crest has been an area of intense study by scientists aspiring to understand its remarkable abilities to migrate long distances along specific routes and to generate such a vast diversity of cell types. There are multiple steps involved in neural crest cell (NCC) development. First, cells at the border between neural ectoderm and non-neural ectoderm are induced to become neural crest. Second, NCCs delaminate or segregate from the neuroepithelium of the developing neural tube and become migratory cells. Third, NCCs disperse and migrate along specific pathways to reach their final destination. Fourth, NCCs proliferate and differentiate into their specific final cell phenotype. Differentiation involves many steps, some of which occur before NCCs delaminate, while others occur in the environment of the NCC's final destination.
Our current understanding of neural crest development has come from studies of several different vertebrates, including avian, amphibian, rodent, and zebrafish embryos (reviewed in LaBonne and Bronner-Fraser, 1999; Le Dourain and Kalcheim, 1999; Christiansen et al., 2000). Zebrafish, Danio rerio, have become a popular model system to study neural crest development because of the accessibility and optical clarity of their embryos at early stages, as well as the rapidly advancing genetic and molecular tools. NCC lineages and migration pathways have been well characterized in zebrafish (Raible et al., 1992; Eisen and Weston, 1993; Raible and Eisen, 1994; Schilling and Kimmel, 1994; Vaglia and Hall, 2000; Li et al., 2003; Sato and Yost, 2003), and numerous studies have been directed at exploring mechanisms of neural crest induction and differentiation and the development of neural crest–derived structures. Recent reviews have highlighted the advantages of zebrafish genetic screens to analyze craniofacial development (Yelick and Schilling, 2002) and induction or differentiation of neural crest (Kelsh and Raible, 2002). In this review, we will focus on another essential characteristic of NCCs, i.e., their ability to migrate. A great deal of research has been aimed at understanding how NCCs break away from the neuroepithelium, take on migratory capabilities and migrate along their specific pathways, although our understanding is still far from complete. We will discuss current knowledge, obtained from studies using chick, frog, rodent, and zebrafish, of mechanisms underlying the processes of NCC delamination, migration, and pathway choice. In particular, we will highlight the promise of the zebrafish model system for future studies on mechanisms of NCC motility and migration.
DELAMINATION OF NCCs
NCCs are specified at the border of the neural and non-neural ectoderm before closure of the neural tube. This specification is the result of two phases of induction. First, there is an initial early phase that is dependent on a combination of signaling by members of the bone morphogenetic protein (BMP) family from the lateral non-neural epithelium and signals such as Wnt, fibroblast growth factor, and retinoic acid. Second, there is a maintenance phase of induction that is dependent on the expression of BMP and Wnt in the dorsal neural tube and overlying epidermis (reviewed in Aybar et al., 2002; Knecht and Bronner-Fraser, 2002). During neurulation, cells fated to become neural crest undergo an epithelial-to-mesenchymal transition (EMT), a dramatic change in which individual NCCs delaminate from the neuroepithelium, take on mesenchymal characteristics, and migrate into surrounding tissues (Fig. 1). The EMT is a complex process involving disruption of the basal lamina, dissociation of cell adhesions, changes in extracellular matrix, extension of filopodia and pseudopodia, translocation of the cell body, and directed migration (Erickson and Perris, 1993; Hay, 1995; Savagner, 2001). The process of EMT also occurs in several other developmental events besides NCC delamination and is a critical step in carcinoma metastasis. Thus, insight into the mechanisms of EMT will aid in understanding these processes as well as that of NCC development. The complexity of the EMT suggests a complexity of molecular mechanisms mediating the changes from specified premigratory neural crest to the definitive NCC as defined by its mesenchymal morphology and behavior. We will discuss first the signals thought to initiate the process of delamination, and second the regulators of cell–cell adhesion and cytoskeletal motility underlying the EMT.
Initial Signals for Delamination
One of the earliest steps in neural crest specification is the expression of members of the Slug/Snail family of transcription factors, which are thought to play a role in NCC formation (Nieto, 2001). Several studies have supported the idea that Slug and/or Snail expression may be the initial factor that triggers NCC delamination. The first indication of such a role came from antisense inhibition experiments. When slug expression was inhibited in chick, NCCs expressing the marker HNK-1 were present but failed to migrate out of the neural tube (Nieto et al., 1994). This result suggests that Slug is required for NCC delamination. Furthermore, in Xenopus, when cells from the neural folds of slug antisense-injected donors were transplanted into wild-type hosts they did not migrate, suggesting that cell autonomous Slug expression is important for the EMT (Carl et al., 1999). Additional evidence came from studies in which a dominant negative form of Slug was expressed under hormone-inducible control. Early inhibition of Slug activity blocked the formation of NCCs; however, when Slug activity was inhibited later, after NCC specification, the emigration of NCCs from the neural tube was blocked (LaBonne and Bronner-Fraser, 2000). These studies suggest that Slug may have a role both in induction and in delamination. Overexpression studies are consistent with this idea. Overexpression of slug caused expansion of NCC markers in chick (del Barrio and Nieto, 2002) and Xenopus (LaBonne and Bronner-Fraser, 1998), suggesting Slug increases NCC induction. In addition, Slug overexpression increased the number of migrating NCCs. The overexpression studies alone do not distinguish between potential roles of Slug in NCC induction versus NCC delamination; however, they support the loss-of-function studies that strongly suggest Slug is the first step initiating the EMT.
Of interest, both chick and mouse snail can mimic the effects of slug overexpression in chick (del Barrio and Nieto, 2002). Moreover, the effects of slug antisense inhibition can be rescued by either slug or snail RNA injection (Carl et al., 1999). These results suggest that Slug and Snail may be functionally redundant and Snail may serve the function of Slug in NCC delamination in some cases. Indeed, in some species slug is not expressed in premigratory NCCs, suggesting it cannot function to initiate delamination. However, in such cases snail genes are expressed. Specifically, slug is expressed in premigratory NCCs in chick (Nieto et al., 1994) and Xenopus (Mayor et al., 1995), whereas it is expressed solely in the migratory NCCs of zebrafish (Locascio et al., 2002) and mouse (Sefton et al., 1998). Conversely, snail is expressed in the premigratory NCCs of mouse (Nieto et al., 1992) and Xenopus (Mayor et al., 1993) and in the migratory NCCs of chick (Sefton et al., 1998) and Xenopus (Mayor et al., 1993). In zebrafish, there are two snail genes: snail1 is expressed in NCCs only after migration is under way (Hammerschmidt and Nusslein-Volhard, 1993; Thisse et al., 1993); however, snail2 is expressed in premigratory and migratory NCCs (Thisse et al., 1995). Although these varied expression patterns may indicate that the specific functions of members of the Snail/Slug family have diverged in different species, in general, the sum of expression patterns of Slug and Snail is evolutionarily conserved, suggesting a conservation of overall function (Locascio et al., 2002).
What are the steps between Slug/Snail expression and the dramatic changes in cell–cell adhesion and cell morphology associated with NCC delamination? Slug has been shown to act as a transcriptional repressor (LaBonne and Bronner-Fraser, 2000), but to date little is known about genes regulated by Slug in NCC development. However, in mammalian epithelial cells, both Slug and Snail bind to the promoter region of the E-cadherin gene and repress its transcription (Batlle et al., 2000; Cano et al., 2000; Bolos et al., 2003). Overexpression of Slug or Snail in epithelial cells causes a loss of E-cadherin–mediated cell–cell adhesion and EMT (Cano et al., 2000; Bolos et al., 2003). Furthermore, in a snail knockout mouse, mesodermal cells retain E-cadherin expression and fail to undergo EMT (Carver et al., 2001). These experiments show that Snail can directly affect E-cadherin and, thus, cell adhesion. Indeed, the regulation of expression of other cadherins performs a vital role in controlling NCC delamination (discussed below). It remains to be seen whether Slug or Snail directly regulates transcription of these and other cell adhesion molecules involved in NCC delamination. It will be necessary to identify all the downstream targets of Slug/Snail to fully understand whether these transcription factors act directly to initiate the changes in cell morphology and behavior underlying NCC delamination or whether there are also indirect effects through a cascade of other signals.
It also appears that Slug/Snail may act in concert with other signaling molecules. One potential extracellular signal involved in NCC delamination is BMP. As mentioned earlier, BMP signaling in the neural plate is crucial for the initial induction of neural crest fate. After the specification of neural crest, the dorsal neural tube and overlying epidermis begin to express BMP and there is some evidence that BMP may play a second role in mediating NCCs delamination. Sela-Donenfeld and Kalcheim (1999) have shown that addition of BMP to chick neural tube explants is sufficient to accelerate delamination of NCCs that already express slug and, therefore, presumably are already induced to be NCCs. Furthermore, after NCCs are induced and expressing slug, ectopic expression of the BMP inhibitor, noggin, leads to a decrease in the number of migratory NCCs in vivo without decreasing slug expression. Overexpression of Xenopus noggin in the mouse hindbrain also causes loss of migratory NCCs, although it is unclear whether this is an effect on NCC induction or delamination (Kanzler et al., 2000). Of interest, although BMP4 and the type I BMP receptor, R1A, are expressed along the entire rostrocaudal axis of the dorsal neural tube in chick, noggin is expressed in a low rostral to high caudal gradient. This gradient is similar to the rostral to caudal wave of NCC emigration in the trunk, suggesting that Noggin could regulate the temporal aspects of delamination (Sela-Donenfeld and Kalcheim, 1999; Sela-Donenfeld and Kalcheim, 2002). Further support for the role of BMP signaling in NCC delamination comes from a recent investigation of mice lacking the gene encoding Smad interacting protein-1, a component of the BMP signaling pathway. In these mice, one population of NCCs, the vagal neural crest, are not induced. In contrast, cranial NCCs appear to be induced normally, but the cells fail to delaminate (Van de Putte et al., 2003). These studies suggest that in some cases, induction can be uncoupled from delamination and that BMP may play roles in both processes.
Thus far, there is little known about the signaling events leading from BMP to changes in cell adhesion and cytoskeletal dynamics. However, in addition to decreasing migratory NCCs, ectopic expression of noggin in chick also caused reductions in the expression of cadherin-6B and the small GTPase rhoB (Sela-Donenfeld and Kalcheim, 1999), both of which appear to be important for NCC delamination (discussed below). Collectively, these results suggest that BMP may have a role in the delamination of NCC and may be required to control levels of Cadherin-6B and RhoB expression. As is the case for the Slug/Snail transcription factors, it will be a challenge to distinguish effects of BMP on NCC induction versus delamination. Analysis of precise timing of action as well as identification of molecular events downstream of BMP receptor binding will again be necessary to determine whether BMPs play a distinct second role in triggering delamination.
Regulators of Cell Adhesion and the Cytoskeleton
What are the regulators of cell contacts and cytoskeletal dynamics that underlie the cell motility changes in NCC delamination? One group of important players in this process is the cadherin family of cell adhesion molecules (reviewed in Pla et al., 2001). Cadherins are transmembrane molecules that act as Ca2+-dependent cell adhesion molecules, and their regulated expression appears to be critically important for NCC delamination. During the neural plate stage, one cadherin, N-cadherin, is expressed throughout the neural epithelium. Throughout neurulation, N-cadherin gradually decreases in the dorsal most region of the neural tube. During NCC delamination in chick, N-cadherin is down-regulated in migrating neural crest (Bronner-Fraser et al., 1992; Nakagawa and Takeichi, 1995; Nakagawa and Takeichi, 1998). Cadherin-6B is also expressed in the dorsal neural tube of chick and is likewise down-regulated in NCCs that have delaminated from the neural tube (Nakagawa and Takeichi, 1995; Nakagawa and Takeichi, 1998). These expression patterns are consistent with the hypothesis that cadherins mediate strong cell contacts in the neuroepithelium and must be down-regulated in order for NCCs to become migratory.
Manipulations of cadherin function have provided some insight into this and other possible mechanisms. Early experiments showed that disruption of Ca2+-dependent adhesions by depletion of extracellular calcium or calcium channel blockade can promote the premature migration of NCCs in quail neural tube explants (Newgreen and Gooday, 1985). This finding suggested that Ca2+-mediated cell adhesion serves to maintain the integrity of the epithelium at the expense of mesenchymal cell formation. This study, however, does not distinguish between the effects of Ca2+-mediated cell adhesion versus other roles for Ca2+, such as vesicle exocytosis or cytoskeletal organization. Further support for a specific role of N-cadherin came from a study in which a function-blocking antibody was injected in vivo in chick. N-cadherin antibodies caused a disruption of neural tube structure and ectopic clusters of NCCs both within and just outside the neural tube (Bronner-Fraser et al., 1992). Surprisingly, however, no defects in NCC development were reported in a mutant mouse lacking N-cadherin (Radice et al., 1997). Similarly, a recent study of the zebrafish N-cadherin mutant, parachute, found no defects in cranial cartilage or pigmentation (Lele et al., 2002), suggesting that N-cadherin may not be required for appropriate NCC development in zebrafish. In contrast to the loss of function studies, it was found that overexpression of N-cadherin in chick prevented NCC delamination and resulted in the presence of aggregated cells in the lumen of the neural tube that were positive for NCC markers (Nakagawa and Takeichi, 1998). This result suggests that increased cell adhesion mediated by N-cadherin may be important for maintaining the epithelial phenotype of premigratory NCCs. Many questions remain concerning the precise role of N-cadherin in NCC delamination and whether N-cadherin is an important player across species. In mouse and zebrafish, where N-cadherin does not appear necessary, it is possible that other cadherins may serve the function of N-cadherin. To date, the potential roles of cadherins in zebrafish NCC delamination have not been explored further.
In contrast to N-cadherin and Cadherin-6B, two other cadherins, Cadherin-7 in chick and Cadherin-11 in mouse and Xenopus, are expressed in migrating NCCs rather than in premigratory NCCs (Hoffmann and Balling, 1995; Kimura et al., 1995; Nakagawa and Takeichi, 1995; Nakagawa and Takeichi, 1998; Vallin et al., 1998). Overexpression of cadherin-7 in chick or cadherin-11 in Xenopus before NCC emigration prevented delamination (Nakagawa and Takeichi, 1998; Borchers et al., 2001). This finding further supports the idea that increased cell adhesion prevents NCC delamination but does not explain why certain cadherins are up-regulated in migrating NCCs. Differential cadherin expression may provide a mechanism for homophilic interactions that are involved in the aggregation of similar NCCs during migration. Also, certain cadherin interactions may be more permissive for motile cell behaviors allowing for the rapid exchange of cell–cell and cell–matrix contacts. Of interest, in one study, murine sarcoma cells transfected with cadherin-7 or N-cadherin were shown to have features consistent with the putative functions of these proteins in NCC delamination. Sarcoma cells expressing cadherin-7 in vitro were more motile, established transient contacts with each other, and showed higher Cadherin-7 and β-catenin turnover rates relative to those expressing similar levels of N-cadherin. Furthermore, when cadherin-7–expressing cells were grafted near the neural tube of chick embryos, they dispersed more readily than did their N-cadherin-expressing counterparts (Dufour et al., 1999). Furthermore, grafted cadherin-7–expressing sarcoma aggregates were found to be permissive to the migration of nearby NCCs, whereas N-cadherin–expressing aggregates were not. Collectively, these findings support a role for cadherin switching during the NCC EMT as a mechanism to regulate NCC migratory state and perhaps to cause like cells to aggregate during migration. To fully understand the role for changing cadherin expression, it will be necessary to know how the presence or absence of each specific cadherin affects NCC motile behavior.
How is cadherin-mediated cell adhesion related to the cytoskeletal changes underlying the EMT? Cadherins form homo- and heteromultimers both within and between cells. The dynamic interactions among cadherins and between cadherins and the cytoskeleton result in changes in cell contacts and morphology. Cadherins bind β-catenin, which links to the cytoskeleton by means of α-catenin. The rearrangement of the cadherin–β-catenin–α-catenin complex is required for the dynamics of cell adhesion (Fukata and Kaibuchi, 2001; Juliano, 2002). This rearrangement may be regulated in part by the Rho family of small GTPases. The activities of both Rho and Rac were shown to be necessary for the maintenance of cadherin cell junctions in cultured keratinocytes (Braga et al., 1997). Conversely, Rho activation in cultured MDCK epithelial cells caused them to undergo EMT (Zondag et al., 2000). Because downstream signaling pathways of individual Rho family members are thought to interact, it is likely that the precise ratio of Rho to Rac activity is important for their influence on cadherins and the cytoskeleton.
Rho GTPases have also been implicated in NCC delamination. One family member, rhoB, is expressed in the dorsal neural tube and in premigratory NCCs in chick (Liu and Jessell, 1998) but only in migratory and not premigratory NCCs in mouse (Henderson et al., 2000). Disruption of Rho activity in chick neural tube explants resulted in fewer migratory NCCs, down-regulation of cadherin-7, and disruption of actin stress fibers without affecting slug expression (Liu and Jessell, 1998). Other GTPases are expressed in Xenopus cranial NCCs, including Rac1 (Lucas et al., 2002) and Rnd1 (Wunnenberg-Stapleton et al., 1999), suggesting that they may also play a role. Although Rho GTPases are likely to be important, their gene expression patterns and relative activity levels during the process of NCC delamination must be determined to fully understand their role.
In addition to the potential regulators of cell contacts we have discussed, several other molecules have implicated roles in NCC delamination. It is clear that the presence of specific extracellular matrix (ECM) components that regulate cell attachment are important for NCCs to undergo delamination. Numerous ECM molecules have been investigated for their contribution to NCC delamination (reviewed in Perris and Perissinotto, 2000). Within the NCCs, protein kinases and phosphotyrosine signaling have been demonstrated to play a role (Monier-Gavelle and Duband, 1995; Newgreen and Minichiello, 1995, 1996; Brennan et al., 1999; Minichiello et al., 1999). Numerous extracellular signals and signal transduction molecules must undoubtedly work together to orchestrate the cell movements underlying the delamination of a NCC. The challenge for the future will be not only to identify all the players but also to determine their precise functions in vivo.
GUIDANCE OF NCC MIGRATION PATHWAY
After NCCs delaminate from the neural tube, they migrate along specific pathways to reach their final destinations. Here, we will discuss migration of three different types of neural crest: (1) the cranial NCCs that form craniofacial cartilage and bone, as well as neurons and glia of the head PNS; (2) the trunk NCCs that form pigment cells and the PNS at trunk levels; and (3) the cardiac NCCs that contribute to heart development. The cranial NCCs that populate the pharyngeal arches arise from particular hindbrain segments (rhombomeres) and migrate ventrally to specific pharyngeal arches. Individual migratory streams of NCCs have specific rostrocaudal identities and for the most part remain separate during migration. In mouse, chick, and zebrafish, the streams are distinctly separate. In Xenopus, the streams are initially contiguous, but they do not intermingle. The trunk NCCs initially choose between two migration pathways: in general, cells destined to become pigment cells migrate along the dorsolateral pathway between the ectoderm and the dermomyotome of the somites, whereas NCCs that will form the sympathetic chain and dorsal root ganglia (DRG) migrate along the ventromedial pathway between the myotome and the neural tube/notochord. Some differences exist between zebrafish and other vertebrates in these pathways. The NCCs entering the two pathways both migrate in a general ventral direction in zebrafish, and the pathways are instead named the lateral and medial pathways (Raible et al., 1992). The two pathways in zebrafish may not be strictly analogous to those in other vertebrates. For example, NCCs migrating along the medial pathway in zebrafish give rise to pigment cells in addition to neurons (Raible and Eisen, 1994). In some species, NCCs of the ventromedial pathway are further restricted to migrate only along the rostral half of each somite, thus establishing the segmental nature of the peripheral ganglia (Rickmann et al., 1985). However, in zebrafish, this restriction does not occur, rather NCCs in the medial pathway migrate along the center of each somite (Raible et al., 1992). Cardiac NCCs arise from neural tube at levels of the caudal hindbrain and first two to three somites of the trunk in chick and mouse as well as more rostral, preotic regions in the zebrafish (Li et al., 2003; Sato and Yost, 2003). These NCCs migrate ventrally into the caudal pharyngeal arches and into the cardiac outflow tract where they are involved in forming the septum that divides the aorta and pulmonary artery (Waldo et al., 1998). In zebrafish, cardiac NCCs also populate the myocardium (Li et al., 2003; Sato and Yost, 2003).
How are NCCs guided along these specific pathways? Several studies carried out over the past 10 to 15 years have identified molecules thought to play a role in controlling NCC migration. Several ECM components have been implicated, including chondroitin sulfate proteoglycans, peanut agglutinin (PNA) binding molecules, F-spondin, versican, aggrecan, fibronectin, and laminin (reviewed in Erickson and Perris, 1993; Perris and Perissinotto, 2000; Krull, 2001). Generally, ECM molecules are thought to provide either permissive or inhibitory substrates for NCC migration and can provide directional information if they have patterned expression. For example, F-spondin (Debby-Brafman et al., 1999), PNA-binding molecules, and chondroitin sulfate proteoglycans (Oakley and Tosney, 1991; Oakley et al., 1994; Krull et al., 1995) are specifically expressed in the caudal half of each somite in chick and can inhibit NCC migration, suggesting they may act to prevent migration of trunk NCCs into this region in vivo. Perhaps more often, directional information is provided by other secreted or cell-surface signaling ligands that bind receptors on NCCs and direct changes in their motility. We will discuss the evidence that two families of known signaling molecules, ephrins and semaphorins, may guide NCC pathway decisions. Because a large body of literature has shown that ephrins and semaphorins affect the direction of extension of axonal growth cones, it is not surprising that they are also implicated in NCC migration, a similar migratory process.
Ephrins and Cranial NCC Migration
Cranial NCCs migrate from specific hindbrain rhombomeres and their ultimate cell fate is specified to some degree by patterning mechanisms, such as specific hox gene expression, within the hindbrain before migration (Trainor and Krumlauf, 2000, 2001; Schilling and Knight, 2001). It is crucial, therefore, that the streams of NCCs emigrating from particular hindbrain rhombomeres remain segregated during migration and migrate along pathways that will bring them to the final location appropriate for their fate. Although there are some species differences, in general, cranial NCCs migrate from the hindbrain in three streams (Fig. 2; Lumsden et al., 1991; Serbedzija et al., 1992; Schilling and Kimmel, 1994). The most anterior stream arises from hindbrain rhombomeres 2 and 3 (r2 and r3) and predominately populates the first (mandibular) pharyngeal arch. The second stream arises from r4 and r5 and populates the second (hyoid) pharyngeal arch. The most posterior stream arises from r5, r6, and more posterior regions, and migrates to the third and more posterior pharyngeal arches. Cranial NCCs need to maintain their relationships with neighboring cells while being guided to the correct arch, and this process appears to be mediated by a combination of interactions between NCCs and interactions between NCCs and surrounding cells or the ECM.
One family of guidance signals thought to play a role in cranial NCC migration is the ephrins and their receptors the Eph receptor tyrosine kinases (Mellitzer et al., 2000; Kullander and Klein, 2002). Ephrins are membrane-bound ligands that fall into two subclasses: the ephrinAs, which are anchored to the membrane with a glycosyl phosphatidylinositol (GPI) link and bind the EphA receptors, and the ephrinBs, which are transmembrane ligands that bind the EphB receptors. Ephrins of one subclass are capable of binding and activating multiple or all Eph receptors of the same subclass. An exception to this classification is EphA4, which can bind ephrins of both A and B subclasses (Gale et al., 1996). In addition to classic forward signaling, whereby the ephrin ligand activates the Eph receptor, which transduces the signal, the EphB receptors can also activate reverse signaling through ephrinBs (Mellitzer et al., 2000; Kullander and Klein, 2002). In this case, the Eph acts as the ligand and the ephrinB acts as receptor. Numerous studies have shown that the ephrins and Ephs are expressed in specific populations of cranial NCCs or in cells in the NCC environment in Xenopus (Winning and Sargent, 1994; Scales et al., 1995; Weinstein et al., 1996), mouse (Bergemann et al., 1995; Flenniken et al., 1996; Gale et al., 1996), and zebrafish (Xu et al., 1995; Cooke et al., 1997; Durbin et al., 1998) embryos. This patterned expression suggests that they may play a role in guiding NCC migration.
The function of ephrins in neural crest development begins within the hindbrain, where ephrinB-EphB interactions have been shown to regulate the formation of boundaries and prevent cell mixing between hindbrain rhombomeres. Several studies have shown that disruption of EphB and ephrinB expression patterns is correlated with loss of rhombomere boundaries in Xenopus and zebrafish (Xu et al., 1995, 1999; Cooke et al., 2001; Cooke and Moens, 2002). In a zebrafish in vitro animal cap assay, Mellitzer et al. (1999) showed that both forward and reverse ephrinB–EphB signaling are involved in preventing intermingling of cell populations.
After cranial NCCs emigrate from the hindbrain, there is evidence that ephrinB-EphB signaling may also prevent intermingling of migrating streams of NCCs. In Xenopus, the receptors EphA4 and EphB1 are expressed in NCCs migrating to the third and fourth pharyngeal arches, and the ligand ephrin-B2 is expressed in the adjacent stream of NCCs that migrate to the second arch. Disruption of signaling with dominant negative EphB1 or EphA4, or with widespread overexpression of ephrin-B2 caused the third-arch NCCs to stray both rostral and caudal to their normal migratory pathway (Smith et al., 1997). This process appears to involve only forward signaling of ephrin-B2, because the ephrin-B2-expressing second-arch NCCs migrated normally. These experiments suggest an inhibitory interaction between streams of migrating NCCs in which ephrin-B2-expressing second-arch NCCs prevent EphA4/EphB1-expressing third-arch NCCs from migrating into the second-arch region. However, this hypothesis does not explain why the third-arch NCCs also stray caudal to their normal pathway where there is no ephrin-B2 expression. There may be other unidentified ephrins in this area whose signaling was also disrupted in these experiments.
Ephrin-B2 may play a slightly different role in mouse cranial NCC migration. In mouse, ephrin-B2 is expressed in the ectoderm of the first two pharyngeal arches, a region normally avoided by NCCs. The receptors EphA4, EphB3, and EphB1 are expressed by migrating NCCs. In ephrin-B2 null mutant mouse embryos, significantly fewer second-arch NCCs migrate and almost none populate the distal portion of the pharyngeal arch. In addition, their migration pattern is disrupted, with cells scattering into adjacent domains they normally avoid (Adams et al., 2001). Again, this effect appears to be mediated by forward signaling only, because the cytoplasmic domain of ephrin-B2 is not required for proper NCC migration. When the endogenous ephrin-B2 gene was replaced with a truncated version lacking the cytoplasmic sequence, cranial NCCs migrated normally (Adams et al., 2001). Like the studies in Xenopus, these experiments also suggest an inhibitory mechanism. In this case, EphB-expressing second-arch NCCs appear to be inhibited by ephrin-B2–expressing cells in the surrounding pharyngeal arch tissue. This hypothesis does not explain why there are fewer second-arch NCCs migrating in the ephrin-B2 mouse mutant. However, ephrinBs are also expressed within the hindbrain (Bergemann et al., 1995; Flenniken et al., 1996), which raises the possibility that ephrinB signaling could play a role in stimulating NCC delamination from the hindbrain.
In addition to ephrinBs, class A ephrins also may be involved in separation of cranial NCC streams in Xenopus. The receptor EphA2 is expressed in second- and third-arch migrating NCCs, as well as in cells along their pathway (Helbling et al., 1998). The ephrinAs, detected by labeling with an EphA2–alkaline phosphatase fusion protein, are expressed in all of the arches. Treatment with dominant negative EphA2 caused third-arch NCCs to migrate more caudally than normal, into the domain of the fourth-arch NCCs (Helbling et al., 1998). This finding suggests that EphA2 normally acts to define the pathway of third-arch NCCs, although it is not clear whether it plays an attractive or inhibitory role. Mouse EphA2 mutants show no discernible phenotype, suggesting there may be species differences in expression pattern or function (Chen et al., 1996). As a whole, the ephrin experiments discussed above provide convincing evidence that they are important for proper migration of cranial NCCs, but leave many questions unanswered. In addition to inhibitory cues that restrict cranial NCCs, there are likely to be attractive cues that direct them along their pathways. A continuing detailed analysis of expression patterns and functions of ephrins and Ephs will be necessary for a complete understanding of their role in cranial NCC migration.
Ephrins and Trunk Neural Crest Migration Pathways
In the trunk axial levels, two populations of NCCs emigrate from the neural tube and begin migration at different stages of development (Raible et al., 1992; Le Dourain and Kalcheim, 1999; Krull, 2001). The early migrating trunk NCCs, which give rise to neurons and glia of the DRG and sympathetic ganglia, make two pathway decisions (Fig. 3). First, they migrate along a ventromedial pathway between the neural tube and the somitic dermomyotome. Second, in chick and mouse, they are further restricted to migrate through the sclerotome of the rostral half of each somite and not the caudal somite halves. The later migrating population of trunk NCCs, which will differentiate into melanocytes, migrates along a dorsolateral pathway between the dermomyotome and the overlying ectoderm.
The first choice faced by emigrating trunk NCCs is whether to migrate along the ventromedial or dorsolateral pathway. One potential means to regulate this choice is a timing mechanism whereby the ventromedial pathway is accessible only at early stages and the dorsolateral pathway becomes accessible at later stages. Heterochronic transplantation studies in which quail early migrating NCCs were transplanted into a host embryo at the later stage when trunk NCCs are migrating along the dorsolateral pathway have shown that early migrating neurogenic NCCs do not have the ability to migrate along the dorsolateral pathway at any stage (Erickson and Goins, 1995). In the reverse experiment, melanogenic, late-migrating NCCs were transplanted into younger embryos and were able to prematurely extend along their correct dorsolateral pathway. Thus, it appears that melanogenic NCCs have the ability to migrate on the dorsolateral pathway at any stage while neurogenic NCCs do not. Evidence against a timing mechanism in the analogous pathway choice in zebrafish is the observation that trunk NCCs continue to migrate along the medial pathway even after migration along the lateral pathway has begun (Raible et al., 1992).
An alternative mechanism to separate the two pathways is the presence of specific molecular cues to which subsets of NCCs respond. Ephrins also may play a role in this guidance step. In chick, several EphBs are expressed in both early- and late-migrating trunk NCCs. The ephrinB ligands are present in cells along the dorsolateral migration pathway during early and late migration stages (Krull et al., 1997; Wang and Anderson, 1997; Santiago and Erickson, 2002). One of these studies has shown that ephrinB–EphB signaling plays a dual attractive/inhibitory role in the choice between dorsolateral and ventromedial pathways (Santiago and Erickson, 2002). In this study, disruption of signaling with dominant negative ephrin-B1 at early stages caused early neurogenic NCCs to migrate along the dorsolateral pathway, indicating they were no longer inhibited from this pathway. In contrast, the same treatment at late stages prevented melanoblast migration into the dorsolateral pathway, suggesting loss of an attractive signal. This study, thus, provides evidence that ephrinBs are attractive to melanogenic NCCs and draw them into the dorsolateral pathway, while they are inhibitory to neurogenic NCCs and prevent them from entering this pathway. It has not yet been tested whether this effect is mediated by forward or reverse ephrinB signaling. The EphBs, in addition to being expressed on the migrating NCCs, are also expressed in cells along the migration pathways, suggesting reverse signaling as a possibility. However, ephrinB expression has not been demonstrated on migrating trunk NCCs, leaving this as an open question. Of interest, in the guidance of retinal axons to their targets in the optic tectum, ephrinBs appear to have an inhibitory role mediated by forward signaling and an attractive role mediated by reverse signaling (Hindges et al., 2002; Mann et al., 2002). It will be interesting to test whether the same type of mechanism is acting on migrating NCCs in the trunk.
Neurogenic NCCs in chick and mammals encounter a second pathway decision after their entry into the ventromedial path. Specifically, they are restricted to the rostral half-somite and avoid the caudal half. Multiple molecular differences exist between rostral and caudal somite halves. As mentioned previously, several ECM molecules are differentially expressed in the caudal half of the somite and some have been shown to affect NCC migration (reviewed in Krull, 2001). In chick and rat, inhibition by ephrinBs expressed in the caudal half-somite also appears to be involved in this guidance step. In vitro assays have shown that ephrinBs are repulsive to NCCs migrating from explants (Krull et al., 1997; Wang and Anderson, 1997). Furthermore, disruption of ephrinB–EphB signaling with dominant negative ephrin-B1 caused NCCs to migrate along both the rostral and caudal halves of the somite, suggesting the loss of an inhibitory cue in the caudal somite (Krull et al., 1997). Additional evidence indicates that the class A ephrins may be involved in the segmentation of chick ventromedial pathway NCCs along the rostral somite half. Ephrin-A2, -A5, and EphA4 are all expressed in migrating trunk NCCs. Overexpression of any of these caused NCCs to migrate over both the rostral and caudal somite halves (McLennan and Krull, 2002). The authors hypothesize that coexpression of EphA4, ephrin-A2, and ephrin-A5 on NCCs mediates an attraction to the rostral somite. Although the ephrinAs are not thought to mediate reverse signaling in the manner of ephrinBs because they are not transmembrane proteins, there is evidence that they do so through a coreceptor (Davy et al., 1999). Furthermore, ephrinAs on growing axons have been purported to act as receptors in an attractive/adhesive interaction with Ephs on their target cells (Knoll and Drescher, 2002). This finding suggests that a similar mechanism might function to guide ephrinA-expressing NCCs.
Another intriguing complication of ephrinA signaling is the finding that ephrinAs can have cis interactions with EphAs expressed on the same cell. Hornberger et al. (1999) demonstrated that ephrinAs on retinal ganglion cell axons can activate and phosphorylate the EphAs expressed on the same cell. This modulated the EphA function and caused the axons to lose responsiveness to ephrinAs in the environment. Whether or not this type of receptor inactivation can occur between coexpressed ephrinBs and EphBs remains to be seen. This process, together with the potential for reverse signaling by the ephrinBs, the large number of different ephrins and Ephs, and the promiscuity of their binding interactions make interpretation of ephrin signaling challenging. In future investigations of ephrin/Eph roles in NCC migration, it will be crucial to analyze in detail the expression patterns of the ephrins and Ephs to distinguish expression in NCCs vs. expression in cells along their pathway. Furthermore, in experiments involving ephrin or Eph manipulations, it will be important to determine the cell autonomy of ephrin/Eph activity to understand specifically where the signals are acting.
Of interest, in some species, the medial trunk NCCs are not restricted to the rostral half of the somite during their migration. In zebrafish for example, these NCCs can enter the pathway at any point along the somite but then converge to the center of the somite while migrating (Raible et al., 1992). Thus, the DRGs and sympathetic ganglia still develop in a segmented manner even though the NCCs giving rise to them migrate along the center of the somite rather than the rostral half. Several molecules show patterned expression in zebrafish somites. Tenascin-C, chondroitin sulfate, and semaphorin 3A2 (formerly semaphorin Z1b) are all expressed only in the caudal somite, whereas the transcription factor engrailed is expressed in the rostral somite (Bernhardt et al., 1998). One possibility is that NCCs are guided to grow along the border of expression of these molecules, although this mechanism has yet to be explored. Sclerotome cells in zebrafish migrate in a ventral to dorsal pathway in the same location as the medial NCCs, suggesting they might influence NCC migration. However, ablation of sclerotome did not affect the development and patterning of the DRGs (Morin-Kensicki and Eisen, 1997). It will be interesting to determine whether medial pathway NCCs are guided by different mechanisms/molecules in different species, and/or whether the relevant guidance molecules have different expression patterns.
Semaphorins in NCC Migration
The semaphorins are another family of molecules implicated in NCC migration. Originally identified as axon guidance cues, the semaphorins are a large diverse gene family encoding transmembrane, GPI-linked, and secreted proteins (reviewed in Mark et al., 1997; Kolodkin, 1998; Raper, 2000; Pasterkamp and Kolodkin, 2003). All semaphorins contain a conserved 500 amino acid extracellular sema domain and differ in their carboxy-terminal domain structure. The family is divided into seven classes based on these structural differences. Classes 1 and 2 are invertebrate, and classes 3–7 are vertebrate semaphorins. The most studied vertebrate semaphorins are those of class 3, which are secreted proteins. Semaphorins were first identified as repellent axon guidance signals that cause collapse of growth cone structure and/or inhibit axon outgrowth. It is now clear that some semaphorins also can be attractive axon guidance cues in some cases. In addition to regulating axon guidance, semaphorins have been shown to play a role in controlling cell migrations. For example, Sema3A inhibits endothelial cell motility (Miao et al., 1999) and migration of neurons in the developing brain (Marin et al., 2001). The receptors for class 3 semaphorins are composed of neuropilins and plexins (Takahashi et al., 1999). Neuropilins are transmembrane proteins with a short cytoplasmic domain that is not thought to participate in transduction of the semaphorin signal. There are two neuropilins that each binds a subset of class 3 semaphorins. The plexins make up a larger family, with at least 9 vertebrate plexins identified to date (Yu and Kolodkin, 1999; Pasterkamp and Kolodkin, 2003). The cytoplasmic domain of plexins has signaling capability and is thought to be the signal transducing component of the receptor. Plexins also confer specificity of the receptor for a particular semaphorin, although details of specific plexin–semaphorin binding pairs are not known.
Although less well established than the ephrins, growing evidence suggests that class 3 semaphorins also act as inhibitory and attractive guidance signals for migrating NCCs. The first indication that semaphorins guide NCC migration came from in vitro experiments showing that Sema3A causes changes in the morphology of migrating NCCs, including collapse of lamellipodial processes. In addition, explanted NCCs given a choice between a Sema3A substrate and a fibronectin substrate avoid growth onto the Sema3A substrate (Eickholt et al., 1999). In chick, Sema3A is expressed in regions bordering migration pathways of both cranial and trunk NCCs, and at least one component of the Sema3A receptor, Neuropilin-1, is expressed in the migrating NCCs. Together these results suggest that Sema3A may regulate NCC migration pathway by inhibiting NCCs from migrating to incorrect locations (Eickholt et al., 1999). Further support of this hypothesis comes from analysis of sema3A and neuropilin-1 knock-out mice (Kawasaki et al., 2002). In these mice, the NCCs that form the sympathetic ganglia migrate beyond their appropriate stopping point and spread to multiple ectopic locations. Moreover, Sema3A suppresses migration of sympathetic ganglion neurons and promotes their aggregation into compact cell masses in vitro. These authors hypothesize that Sema3A's normal function is to signal ventromedial migrating NCCs to stop migration and coalesce into ganglia.
A different class 3 semaphorin, Sema3C, is hypothesized to act as an attractive guidance signal for cardiac NCCs. Sema3C is expressed in the cardiac outflow tract of mice and the semaphorin receptor, PlexinA2, is present in migrating cardiac NCCs. In Sema3C knockout mice, the cardiac NCCs are able to begin migration and arrive at the entrance to the cardiac outflow tract but are unable to invade the outflow tract (Brown et al., 2001; Feiner et al., 2001). These studies provide evidence that Sema3C acts as an attractive cue to direct cardiac NCCs into the cardiac outflow tract.
In some cases, semaphorins are expressed within migrating NCCs. In their analysis of Sema3C expression in mouse, Feiner et al. (2001) suggested the possibility that Sema3C is expressed not only along the pathway of migrating cardiac NCCs but also within the NCCs themselves. Another class 3 semaphorin, Sema3D, is expressed in zebrafish in the dorsal hindbrain in r3–r5, in the second-arch stream of cranial NCCs, as well as in the trunk NCCs that migrate along the medial pathway (Halloran et al., 1999). This specific expression in subsets of migrating NCCs suggests that Sema3D may have a specific function for their development. For example, Sema3D expression in only the middle stream of cranial NCCs could inhibit intermixing of neighboring NCC streams. Preliminary studies showed that knockdown of Sema3D with morpholino antisense results in reduced numbers of migrating NCCs and disorganization of pharyngeal arches, suggesting Sema3D could be involved in NCC delamination or migration (Berndt et al., 2002).
That some semaphorins are expressed in NCCs raises the intriguing possibility that semaphorins may be required cell autonomously within the migrating cells for proper migration. Of interest, two studies have shown that class 3 semaphorin up-regulation within carcinoma cells is correlated with their ability to migrate (Christensen et al., 1998; Martin-Satue and Blanco, 1999), suggesting semaphorins could act in an autocrine manner to affect cell motility. Also, Sema3B and Sema3C have been shown to antagonize the activity of Sema3A by competitively binding to, but not activating, the receptor Neuropilin-1 (Takahashi et al., 1998). Thus, a semaphorin expressed on migrating NCCs could potentially reduce the sensitivity of the NCCs to other semaphorins. This finding is reminiscent of the ability of ephrinAs to reduce the activity of Ephs expressed on the same cell and shows that a similar level of complexity exists for semaphorin signaling. In summary, investigation of semaphorin function in NCC migration is in its infancy; however, it is clear that they play an important role in the migration of subsets of NCC across species. These intriguing findings set the stage for many future experiments.
ZEBRAFISH AS A MODEL TO STUDY NCC MIGRATION
Many of these studies of NCC delamination and migration have been carried out with model systems other than zebrafish and have contributed significantly to our understanding of underlying mechanisms of motility. The picture is just beginning to emerge, however, and there are still large gaps in our understanding of these mechanisms. Many questions remain unanswered. For example, there are certainly unidentified molecular factors in both the process of EMT and the guidance of NCC migration along specific pathways. In addition to identifying players, it is necessary to understand how these molecules function to affect the motility and behavior of NCCs. The zebrafish's unique complement of characteristics makes it an ideal model system to explore these questions. First, several mutagenesis screens have been aimed at discovering genes important for neural crest development. Second, the function of particular known genes can be analyzed by using a variety of available genetic, molecular, and embryologic techniques. Third, the optical clarity and accessibility of the zebrafish embryo allow high-resolution imaging of living NCC motile behaviors over extended periods. This collection of tools promises to yield answers to many of the remaining questions in NCC development.
Mutagenesis Screens in Zebrafish
An extremely powerful approach to identifying new molecular players involved in neural crest development is the use of mutagenesis screens in zebrafish. Large scale forward genetic screens have traditionally been carried out in invertebrate model organisms such as Drosophila or Caenorhabditis elegans. However, because the neural crest is a unique feature of vertebrates, invertebrate models cannot be used to investigate mechanisms of NCC development. Because of the ability to maintain large numbers of zebrafish in a relatively small area and at low cost, the zebrafish is the most practical vertebrate for forward genetic screens. Several groups have carried out such screens with the goal of analyzing particular aspects of neural crest development. For example, several groups have screened mutant larvae by examining craniofacial morphology and have identified numerous mutants with defects in the structure of the jaw, pharyngeal arches, and craniofacial structures (Neuhauss et al., 1996; Piotrowski et al., 1996; Schilling et al., 1996; Golling et al., 2002; reviewed in Yelick and Schilling, 2002). Others have screened for defects in other neural crest derivatives, including pigment cells (Kelsh et al., 1996; Odenthal et al., 1996) and peripheral nervous system components (Henion et al., 1996).
The mutants discovered in the screens for craniofacial structure display varied malformations of the skeletal elements of the head and jaw (Neuhauss et al., 1996; Piotrowski et al., 1996; Schilling et al., 1996; Golling et al., 2002), many of which could be caused by errors in some stage of cranial neural crest development, including induction, delamination, migration, or differentiation. For several of these mutants, the mutated gene has been identified, although none yet appear to be genes involved specifically in NCC delamination or migration. Rather, most of the mutations identified to date are in genes likely to be involved in other aspects of NCC development. For example, mutations in neurogenin-1 (Golling et al., 2002), sox9a (Yan et al., 2002), wnt5 (Rauch et al., 1997), wnt11 (Heisenberg et al., 2000), madh5 (Hild et al., 1999), endothelin-1 (Miller et al., 2000), gli1 (Karlstrom et al., 2003), and the transcription factor ap2a (Knight et al., 2003), all display defects in craniofacial and/or jaw structure. These genes appear to be affecting NCC induction, differentiation, or earlier developmental events that have indirect effects on craniofacial structure. From these screens, there are still many mutants for which mutated genes have not been identified and that are potential candidate NCC migration mutants. The mutants were discovered initially by examination of gross morphology of the jaw and head in larval fish, and characterized with Alcian blue labeling of cranial cartilage structures. Detailed characterization of migrating NCCs at earlier developmental stages has not been reported. Such future analyses may reveal new insight into mechanisms of NCC delamination or migration.
The screens for pigmentation defects (Kelsh et al., 1996; Odenthal et al., 1996) already have revealed two mutants in which migration of neural crest–derived pigment cells is directly affected. The sparse mutant has a reduction in neural crest–derived melanocytes and encodes a type III receptor tyrosine kinase, kit, which is expressed in melanocytes and is required cell autonomously for their dispersal and migration from the neural tube (Parichy et al., 1999). A related receptor tyrosine kinase, fms, is mutated in the panther mutant (Parichy et al., 2000b). This gene is expressed in the NCCs that give rise to another pigment cell type, xanthophores, and has an analogous function in their development. In the panther mutant, xanthophore precursors fail to migrate and instead accumulate near the dorsal neural tube. These studies have revealed previously unknown functions for the kit and fms genes. Furthermore, they demonstrate that specific subsets of NCCs may require different signals to regulate their migration from the neural tube to their final destinations. In addition to sparse and panther, which have NCC migration defects, the pigmentation screens also have yielded mutants in genes important for induction or differentiation of NCCs, including sox10, a transcription factor (Dutton et al., 2001), mib, a ubiquitin ligase important for Notch-Delta signaling (Itoh et al., 2003), nacre, a microphthalmia-related protein (Lister et al., 1999), and rose, an endothelin receptor (Parichy et al., 2000a).
All the neural crest screens have identified many mutants for which the underlying mutated gene is still unknown. Some of these mutants have phenotypes that show promise for mutations affecting NCC migration. For example, in the laughing man mutant (Henion et al., 1996), the DRGs are displaced to a dorsal location, indicating that they did not migrate properly during development. Two classes of pigment mutants with altered pigment pattern (e.g., choker) or with chromatophores in ectopic locations (e.g., parade) also may reflect defects in migration (Kelsh et al., 1996). In the future, detailed characterization of NCCs along migration pathways in these and other uncharacterized mutants as well as the identification of the mutated genes will undoubtedly lead to the discovery of many players and mechanisms involved in NCC motility.
Other Molecular/Genetic and Embryologic Approaches
In addition to mutagenesis screens, there are multiple molecular and genetic tools as well as traditional embryologic techniques that make the zebrafish ideal for analyzing mechanisms of NCC development. For example, known genes can be targeted for loss-of-function with morpholino antisense (Nasevicius and Ekker, 2000; Ekker and Larson, 2001). Morpholino antisense can be used to test the function of candidate genes thought to be involved in NCC migration and for which there are not available mutants. Moreover, morpholino antisense provides a useful method to rapidly analyze the concurrent loss-of-function of two or more genes by injecting more than one morpholino (e.g., Hunter and Prince, 2002). In addition to loss-of-function analyses, genes can be readily overexpressed in zebrafish with either injection of RNA or DNA into newly fertilized embryos, or by generation of transgenic strains containing the gene of interest driven by a specific promoter. The zebrafish hsp70 heat-inducible promoter provides the added benefit of temporal control of gene overexpression (Halloran et al., 2000). This strategy allows investigation of effects of gene overexpression at select developmental stages without disrupting earlier events. hsp70 promoter-dependent transcription of transgenes can also be induced in individual cells by laser activation of the promoter, allowing spatial control over misexpression as well (Halloran et al., 2000), as has been demonstrated in cardiac NCCs (Sato and Yost, 2003).
Many of the remaining problems in understanding NCC delamination and migration mechanisms require knowledge of the cell autonomy of the functions of guidance signals. In addition to the genetic advantages of the zebrafish, it is also possible to perform heterotopic and heterochronic transplants of NCCs (Raible and Eisen, 1996; Schilling et al., 1996; Schilling et al., 2001) as well as cell transplants at the blastula stage to create mosaic embryos. Although these techniques are not unique to zebrafish, when combined with genetic manipulations they provide a powerful means to test cell autonomy.
Live Imaging of Neural Crest
Thus far, we have focused on the exploration of potential molecules regulating NCC delamination and migration. Many of the questions that remain unanswered pertain to understanding how these molecules act to change the motile behavior of the cells. To thoroughly understand mechanisms of cell motility and migration, and how guidance molecules function, it is crucial to have the ability to observe and analyze living cell behaviors. This ability is a key strength of the zebrafish as a model system for such studies. It has been possible to image living DiI-labeled chick NCCs either in ovo or in neural tube explants. In these studies, analysis of time-lapse NCC migration provided new insight into behaviors of cranial and trunk NCCs (Krull et al., 1995; Kulesa and Fraser, 2000). The zebrafish has an added advantage in that it allows higher resolution images of fine cell structures in migrating NCCs. NCCs in zebrafish are larger and fewer in number than in the chick, making them easier to image (Raible et al., 1992). More importantly, the transparency and accessibility of the zebrafish embryo allows visualization of detailed NCC morphology with simple differential interference contrast (DIC) optics (Raible et al., 1992; Jesuthasan, 1996). Fine structure of filopodial and lamellipodial processes extending from NCCs and cell–cell contacts between NCCs and surrounding cells are clearly visible in intact embryos. The first study to image zebrafish NCCs in living embryos laid the groundwork for future studies by defining the characteristic morphology, pathways and timing of migration for trunk NCCs in particular (Raible et al., 1992).
Time-lapse imaging provides a particularly powerful tool for analyzing dynamic NCC behaviors over the course of their migration, including the dynamics of cell–cell contacts. With time-lapse DIC imaging, it is possible to visualize directly the changes in cell motile behaviors during NCC delamination as well as subsequent migration. Figure 4 shows two cranial NCCs making the EMT during delamination from the neural tube (see Supplementary Movie 1, which is available online at http://www.interscience.wiley.com/developmentaldynamics/suppmat/index.html). While still in the neural tube, these cells display excessive membrane blebbing that is followed by the protrusion of lamellipodial processes and movement away from the neural tube. At a slightly later stage of development, when cranial NCC migration is in progress, details of cellular protrusions and changing contacts among NCCs and with neighboring cells are clearly visible (Fig. 5; see Supplemental Movie 2, which is available online). This type of imaging will allow detailed analysis of the behaviors NCCs display during their delamination and migration and will provide baseline information for future studies manipulating molecules that potentially regulate NCC behavior.
Time-lapse imaging of zebrafish trunk NCCs has provided insight into mechanisms underlying the choice between the medial and lateral migration pathways (Jesuthasan, 1996). Early migrating trunk NCCs (those destined to migrate along the medial pathway) extend protrusions that contact the lateral somite surface at the entrance to the lateral migration pathway. This contact causes a rapid loss of cell motility, and the protrusions collapse and retract, preventing advance along the lateral pathway. This repulsive character of the somite surface is gradually lost as development proceeds, so that the NCCs that migrate later along the lateral pathway are not repelled by the somite surface. These results demonstrate the advantages of live imaging and suggest a mechanism whereby medial migrating NCCs are directed to their appropriate pathway by active repulsion along the lateral pathway.
DIC imaging of NCC dynamics can provide a great deal of insight into guidance mechanisms. However, with DIC imaging alone, the molecular identity of cells cannot be determined. Transgenic zebrafish technology provides an added advantage that particular cell types can be labeled in living embryos. The generation of transgenic zebrafish expressing green fluorescent protein (GFP) under control of the foxD3 promoter brings a new dimension to NCC imaging (Gilmour et al., 2002). The foxD3 gene is expressed in premigratory NCCs and is down-regulated in migrating NCCs (Odenthal and Nusslein-Volhard, 1998). In these transgenic fish, a 14-kb segment of the foxD3 regulatory region is sufficient to drive expression of GFP in the premigratory crest, and in addition, the GFP is stable and remains in migrating NCCs. In these foxD3::eGFP transgenics, premigratory cranial NCCs can be identified within the neural tube and their behavior followed through delamination and migration (Fig. 6). Fine processes of migrating NCCs are also visible (inset in Fig. 6). The foxD3 gene is also expressed in a subset of neural crest–derived glial cells along the lateral line, which are GFP labeled in the foxD3::eGFP transgenics. Gilmour et al. (2002) imaged the migration of these glia in mutant embryos that have defects in lateral-line axon pathways. These experiments showed that the neural crest derived glia use lateral-line axons as guides to navigate along their pathway and demonstrated the strength of combining time-lapse imaging with genetic manipulation to understand mechanisms of guidance of migrating cells.
Many questions remain to be answered before we understand the mechanisms by which NCCs go through the dramatic morphologic changes underlying the EMT and then navigate through the embryo to their final locations. Although numerous important molecular players have been discovered, there are certainly others yet to be found. Imaging of NCC behaviors in living zebrafish embryos is an ideal means to understand the functions of molecular regulators of NCC migration. The zebrafish is unique in that it possesses the advantages of genetic manipulability together with the optical clarity and accessibility that allow live imaging of NCCs. In the future, the combination of live imaging with molecular perturbation will provide a wealth of new insight into mechanisms of NCC migration.