Comparative analysis of neural crest cell death, migration, and function during vertebrate embryogenesis



Cranial neural crest cells are a multipotent, migratory population that generates most of the cartilage, bone, connective tissue and peripheral nervous system in the vertebrate head. Proper neural crest cell patterning is essential for normal craniofacial morphogenesis and is highly conserved among vertebrates. Neural crest cell patterning is intimately connected to the early segmentation of the neural tube, such that neural crest cells migrate in discrete segregated streams. Recent advances in live embryo imaging have begun to reveal the complex behaviour of neural crest cells which involve intricate cell-cell and cell-environment interactions. Despite the overall similarity in neural crest cell migration between distinct vertebrates species there are important mechanistic differences. Apoptosis for example, is important for neural crest cell patterning in chick embryos but not in mouse, frog or fish embryos. In this paper we highlight the potential evolutionary significance of such interspecies differences in jaw development and evolution. Developmental Dynamics 229:14–29, 2004. © 2003 Wiley-Liss, Inc.


Craniofacial development is intrinsically related to segmentation along the primary body axis and in vertebrates, one of the earliest manifestations of segmental patterning is the generation of neuromeres in the neural tube. The hindbrain, for example, is partitioned into a series of contiguous units termed rhombomeres (Vaage, 1969) as a direct consequence of cell lineage restriction (Fraser et al., 1990) and differential activity of several regulatory genes (Lumsden and Krumlauf, 1996). In vertebrate embryos, segmentally arranged cranial structures such as the branchial arches and nerve ganglia are built upon this primary metamerism. Hence, the segmental organization of the hindbrain is a conserved strategy used by vertebrates for establishing the spatial patterns of cellular differentiation during craniofacial morphogenesis.

Cranial neural crest cells are a multipotent, migratory population that gives rise to the majority of the peripheral nervous system and the distinctive bone, cartilage, and connective tissues of the head. The segmental organization of the hindbrain is particularly important in patterning the neural crest cell migration pathways. Cranial neural crest cells in all vertebrates generally migrate in discrete segregated streams often separated by clear neural crest-free zones (Farlie et al., 1999). Over the past decade, there has been some debate concerning how the distinct migratory neural crest cell streams are established and in particular the importance of apoptotic cell death (Graham et al., 1993, 1994; Trainor et al., 2002). In addition, there are also reported discrepancies in the range of cell fates derived from individual rhombomeric neural crest cell populations as well as the precise pathways they follow (Noden, 1983; Couley et al., 1993; Sechrist et al., 1993; Köntges and Lumsden, 1996).

One tremendous challenge confronting developmental biologists is to study neural crest cell patterning within embryos in their natural settings. In contrast to the classic static approach of collecting data from a timed series of animal stages, newly designed culture and imaging techniques in zebrafish (Koster and Fraser, 2001; Maves et al., 2002), Xenopus (Wu and Cline, 2003), and chick (Kulesa et al., 2000) are providing real-time insights into the morphogenetic processes that occur during embryonic development. As a result, unique in vivo analyses of cell migratory behaviors are revealing the importance of cell–cell and cell–environment interactions during neural crest cell migration. In this study, we provide a comparative analysis of cell migration and patterning in cranial neural crest cell populations during chick, mouse, frog, zebrafish, axolotl, lamprey, and marsupial embryonic development and highlight the possible evolutionary significance of interspecies differences.


Neural crest cells arise uniformly along the vertebrate axis at the junction between the neuroepithelium and the surface ectoderm, a region commonly referred to as the neural plate border. One commonality observed in all higher order vertebrate species analyzed to date is that cranial neural crest cells migrate in distinct, segregated streams lateral to the even-numbered rhombomeres of the hindbrain and into the adjacent branchial arches. Neural crest cell emigration commences from the midbrain region in chick at approximately the five- to six-somite stage (embryonic day [E] 1.5; Tosney, 1982), in mouse at the five-somite stage (E8.25; Nichols, 1986), in Xenopus at around stage 12 (Sadaghiani and Thiebaud, 1987; Mayor et al., 1995), and in zebrafish at 13–14 hr postfertilization (Schilling and Kimmel, 1994). Rhombomere (r)1 and r2 produce numerous neural crest cells, which comigrate ventrolaterally as a collective stream along with midbrain-derived neural crest cells and extensively populate the entire proximodistal extent of the first branchial arch (ba1; Fig. 1; Lumsden, et al., 1991). In comparison to r1 and r2, r3 produces fewer neural crest cells (in the mouse, zebrafish, and Xenopus) and rather than migrating laterally, some r3-derived neural crest cells migrate anteriorly to join the r2 stream (Sechrist et al., 1993; Schilling and Kimmel, 1994; Trainor and Tam, 1995). Over the past decade, discrepancies have been reported in the migratory pathways and final destinations of odd-numbered rhombomeric neural crest cells and in particular whether these cells truly migrate into a branchial arch. The confusion has been brought about primarily by static dorsal versus lateral analyses of lineage-traced neural crest cell populations and the difficulty in establishing the precise proximal boundaries of a branchial arch. Morphologically, the branchial arch ectoderm is contiguous until it abuts the neural tube. It is fair to say that r3-derived neural crest cells in chick contribute to only the most proximal region of ba1, which corresponds to the location of the trigeminal (semilunar) ganglion of the V cranial nerve (Köntges and Lumsden, 1996). However, in mouse, zebrafish, and Xenopus, r3-derived neural crest cells contribute to more distal portions of the mandibular arch (ba1) (Sadaghiani and Thiebaud, 1987; Osumi-Yamashita et al., 1994; Schilling and Kimmel, 1994; Trainor and Tam, 1995). Thus, in chick embryos midbrain, r1- and r2-derived neural crest cells populate ba1, whereas in mouse, zebrafish, and Xenopus, an additional but small neural crest cell contribution is also derived from r3.

Figure 1.

A schematic summary of cranial neural crest cell migratory behaviors and cell trajectories. On the right-hand side of the illustration, circles represent individual neural crest cells. Coloured circles represent individual neural crest cells that emerge from rhombomere 1 (r1; turquoise), r2 (purple), r4 (blue), or r6 (green). Red circles represent neural crest cells that emerge from r3 or r5, and yellow circles represent r7-derived neural crest cells. The left side of the illustration represents the approximate pathways of the cells from a particular rhombomere. In the rostral portion of the hindbrain, neural crest cells from r1–r2 migrate mostly along rostrolateral diagonal pathways (1,2). Cells often form chain-like arrays that stretch from the neural tube midline to the lateral edge of the hindbrain. Cells then tend to disassemble the chain and migrate as individuals, forming a wide scattering of cells lateral to r1–r2. Some cells from r3 move rostrally to join the neighboring r2 cells in a rostrolateral diagonal trajectory toward the first branchial arch (3). Occasionally, cells from r3 will migrate laterally into the exclusion zone adjacent to r3 but will either stop and collapse filopodia (4) or quickly change direction and move toward a neighboring stream of cells (7). Cells from r3 also migrate in the caudal direction either into r4 or along a caudolateral diagonal trajectory to join the r4 cells moving laterally (5). The trajectories of r3 cells to join neighboring migration streams from r2 and r4 resemble freeway on-ramps (left side of illustration in r3). Neural crest cells from r4 follow mostly lateral trajectories (6) into the surrounding tissue in a much more dense stream of cells than in the r1–r2 region. The shape of the migration stream at the stream front resembles a fan shape, tapering back toward the neural tube, then widening at the neural tube because of the merging of cells from r3 and r5. Neural crest cells from r5 follow along trajectories in a similar pattern to r3 cells to join neighboring migration streams. Cells from r5 occasionally emigrate into the region lateral to r5 but, in the same way as r3 cells, either collapse filopodia and stop or change direction and move toward a neighboring migration stream (7). Cells from r5 also follow along caudolateral diagonal trajectories to merge with cells from r6, moving toward the third branchial arch (8). Neural crest cells from r6 move along caudolateral diagonal trajectories in a dense migration stream, which resembles the stream lateral to r4 (9). The stream front has a wide fan shape, which tapers back toward the neural tube. In contrast to closely migrating streams of cells, cells from r7 form chains that stretch from the neural tube to the region lateral to r7 (10). The caudolateral shape of the migration stream from r5–r6 and the lateral trajectories of cells from r7 merge in a region lateral to r7.

The second branchial arch (ba2) is also composed of neural crest cells derived from multiple rhombomeres namely r3, r4, and r5 (Fig. 1). The remaining r3-derived neural crest cells, migrate posteriorly contributing to the geniculate ganglion of the facial nerve (VII) in the proximal region of the ba2 (Sechrist et al., 1993; Köntges and Lumsden, 1996). In chick and mouse embryos, some r3-derived neural crest cells also contribute to more distal regions of ba2 by migrating well beyond the branchial clefts. It is not clear why r3-derived neural crest cells migrate further into ba2 than they do in ba1, but in chick, the migratory differences are reflected in their final differentiated fates. R3-derived neural crest cells that contribute to the proximal most region of ba1 give rise to sensory neurons of the trigeminal ganglion. In contrast, those r3 cells that migrate into the second arch contribute to cartilaginous elements such as the columella and retroarticular process in addition to sensory neurons of the VII cranial nerve (Köntges and Lumsden, 1996).

In mouse, chick, zebrafish, and Xenopus, r4 by far gives rise to the bulk of the neural crest cells that populate ba2 and similar to r1 and r2, r4-derived neural crest cells migrate as a discrete segregated stream colonizing the entire proximodistal extent of the arch (Sadaghiani and Thiebaud, 1987; Lumsden et al., 1991; Osumi-Yamashita et al., 1994; Schilling and Kimmel, 1994; Trainor and Tam, 1995). In comparison to r3, r5-derived neural crest cells are also unable to migrate laterally in mouse and chick embryos; hence, they migrate anteriorly and posteriorly in a manner similar to cells derived from r3, joining the even-numbered neural crest streams, in this case from r4 and r6 (Sechrist et al., 1993; Kulesa and Fraser, 1998; Trainor et al., 2002). Anteriorly migrating r5-derived neural crest cells contribute to the geniculate and vestibulocochlear (acoustic) ganglia of the facial/vestibulocochlear nerve (VII/VIII) in the proximal region of ba2 but also make minor contributions to the more distally derived columella and retroarticular process as well as hyoid (ba2) muscle attachment sites (Köntges and Lumsden, 1996). It is important to note that r5 in mouse, Xenopus, and zebrafish generates significantly more neural crest cells than r3 such that r5 contributes more neural crest cells to ba2 than r3 but still significantly less than r4 (Hörstadius, 1950; Schilling and Kimmel, 1994; Trainor et al., 2002). In contrast, in the chick, r3 contributes more neural crest cells to ba2 than does r5 (Köntges and Lumsden, 1996).

The third branchial arch (ba3) similar to ba1 and ba2 is composed of multiple distinct populations of neural crest cells (Fig. 1). In mouse, Xenopus, and zebrafish embryos, ba3 is populated by r5- and r6-derived neural crest cells, whereas in chick, ba3 is colonized by r6- and r7-derived neural crest cells (Sadaghiani and Thiebaud, 1987; Osumi-Yamashita et al., 1994; Schilling and Kimmel, 1994; Trainor and Tam, 1995; Köntges and Lumsden, 1996). The proximally migrating neural crest cells contribute to the petrosal ganglion of the glossopharyngeal nerve (IX), whereas the more distally migrating neural crest cells contribute to skeletal derivatives and also give rise to connective tissue (Noden, 1983; Couly and LeDouarin, 1988; Köntges and Lumsden, 1996). Neural crest cells derived from r7 contribute to the fourth branchial arch (ba4) along with a minor more caudally migrating population derived from r6.

Together these analyses demonstrate that hindbrain-derived neural crest cells migrate in distinct segregated streams in all vertebrates examined to date. Each rhombomere produces neural crest cells, with the even-numbered segments generating the most. The quantity of neural crest cells produced by even-numbered rhombomeres is followed in the mouse, Xenopus, and zebrafish by r5 and then r3, whereas in the chick r3 generates more neural crest cells than r5. Despite the general similarity, there are also some subtle differences in the overall patterns of neural crest cell formation and migration between the different species. For instance, in chick, zebrafish, and Xenopus, neural crest cells migrate from a closed neural tube or neural keel, respectively, whereas in the mouse neural crest cell migration commences before neural tube closure.


The formation of an individual migration stream arises from a complex set of cell migratory behaviors (Fig. 1). The characteristics of these streams vary spatially between the rostral (midbrain to r2), mid- (r3 to r5), caudal (r6 to r7) hindbrain. Within a migration stream, individual cell movements range from directed motion toward a branchial arch to low directional wandering within a stream and even turning back toward the neural tube and moving in the reverse direction (Kulesa and Fraser, 1998, 2000). From the midbrain to the r2/r3 border, neural crest cells emigrate over a wide expanse of tissue and generally exhibit diagonal trajectories in the rostral and ventrolateral directions with little caudal deviation (Fig. 1). The caudal portion of the stream is made up of cells derived from r3, which migrate rostrally, crossing the r2/r3 boundary before joining cells and migrating ventrolaterally toward the first branchial arch (Figs. 1, 2).

Figure 2.

Cell movements and behaviours. Tracking of fluorescently labeled neural crest cells during time-lapse imaging sessions shows the cell trajectories from the neural tube into the surrounding unlabeled tissue. Each image represents a confocal section in the longitudinal plane of the mid-hindbrain region, including rhombomeres 3 (r3), r4, and r5. The cell tracks shown represent cell trajectories of cells that were tracked for greater than or equal to 90% of the total number of frames in the time-lapse. A–C: A typical time-lapse imaging sequence and cell tracking analysis of cell trajectories from the r3–r5 region of the hindbrain. Cell targets are acquired (denoted by the red marks) at t = 0 hr and followed for 12 hr, as neural crest emigrate from the neural tube into the branchial arches. C: Multicolored cell trajectories show the positions of fluorescently labeled tracked neural crest cells over the 12-hr session of the time-lapse. D–I: Focal labeling of neural crest cells emerging from r3 and r5 show the cell movements over a 6-hr period from two different (D–F and G–I, respectively) typical time-lapse imaging sessions. The cell tracks over the 6 hr show the pathways of cells from r3 and r5 emerging as caudolateral and rostrolateral diagonal trajectories. Scale bar = 50 μm in I (applies to A–I).

In contrast to the wide distribution of cells migrating into the first arch, the neural crest stream that emerges lateral to r4 and extends into the second arch is narrower and far denser (Figs. 1, 2A–C). The leading front of the neural crest stream resembles a fan, where cells actively extend filopodia in many different directions, dynamically sampling a wide swath of the terrain. The stream tapers in width back toward the neural tube, then expands near the lateral edge of the neural tube. Here, the trajectories of r3 and r5 neural crest cells give the appearance of merging freeway on-ramps to the main r4 superhighway (Figs. 1, 2D–I). Cell trajectories near the front have much higher and more complex directionality than following cells and a slower average speed (Fig. 2A–C). This finding may be due in part to the cells at the front having less interaction with neighboring cells and, thus, a better ability to sense the environment and any potential attraction cues toward the branchial arches.

The caudal-most stream of migrating neural crest cells adjacent to r5–r7 exhibit some of the most intriguing spatial behaviours (Figs. 1, 2). In this area, the dense ventral and caudolateral stream of cells emanating from r6 intersects with a scattering of cells from r7. The r5- and r6-derived neural crest cells actively migrate as individuals in the stream that populates the third and fourth branchial arches. Within this stream, similar to cells in the stream lateral to r4, cells may round up and divide, crawl past neighbors, move across the width of the stream, and even reverse direction back toward the neural tube. In contrast, cells that emerge from r7 tend to migrate in chain like arrays that stretch from the lateral edge of the neural tube to the stream of cells that emigrate from r5 and r6. These chains often consist of 3–10 cells, are one cell wide, and may stretch up to 100–150 μm in length. The intersection takes place laterally at a distance between the neural tube and the third and fourth branchial arches. Here, the chains disassemble near the r5/r6 stream, and cells now separate to migrate as individuals. Currently, it is not known whether the cells from r7 are attracted to this area or whether they require the presence of r5 and r6 cells. Interestingly, the chain-like migratory behaviour of r7-derived neural crest cells appears to be a nonautonomous cell behaviour, because when r7 is transplanted in place of r4 in chick embryos, the r7-derived neural crest cells now migrate as individual cells similar to endogenous r4-derived neural crest cells migrating into ba2 (Trainor and Kulesa, unpublished observations).

Cranial neural crest cells derived from r3 and r5 migrate anteriorly and posteriorly to join the even-numbered neural crest streams and exhibit migratory behaviours characteristic of those streams as described above (Figs. 1, 2). Time-lapse analyses, however, have revealed that small numbers of neural crest cells can occasionally exit laterally from r3 and in doing so, display two distinct migratory behaviors (Kulesa and Fraser, 1998, 2000). First, an individual cell may migrate laterally for only a short distance (∼a couple of cell diameters) before stopping, collapsing its filopodia, and remaining in that position (Fig. 2D–F). Second, cells may enter this lateral region but, after migrating a few cell diameters, may dramatically alter their direction and move toward a neighboring migration stream emanating from r2, r4, or r6. The very limited migration of neural crest cells in the restricted areas adjacent to r3 and r5, suggests that exclusion zones are not absolutely inhibitory. This finding is more evident dorsally in regions where r3 and r5 border the even-numbered rhombomeres. Cells from r3 and r5 often migrate along diagonal trajectories, skirting the neighboring rhombomere boundary (Fig. 2D–I).

Are neural crest cells restricted to populating a particular branchial arch once they exit from the hindbrain and incorporate into a distinct stream or can cells divert and intermingle between streams? Confocal time-lapse imaging analyses reveal that neural crest cells are mostly maintained within a particular migration stream (Kulesa and Fraser, 2000). However, during a typical time-lapse imaging session, small numbers of migrating cells are frequently observed diverting in the direction of a neighboring stream (Fig. 2). When these cells contact the neighbouring stream, although it is infrequent, the cells may cross into the neighboring stream and, thus, move into a different branchial arch. This phenomenon is manifested when cells break away from one stream and contact cells from the neighbouring stream that have also diverted into the exclusion zone. The cell–cell contact persists, and it is very infrequent that the cells actually move past each other and relocate in a new stream. Rather, the cells return to their original stream. This behaviour occurs predominantly in the caudal region of the head, lateral to the otic vesicle where branchial arch streams are closer to each other than in the rostral region of the head. This plasticity to change direction becomes strikingly evident when subpopulations of premigratory neural crest cells are ablated. The result is that neighboring neural crest cells dramatically reroute their migratory trajectories and fill in the destination sites of the missing neighbors (Kulesa et al., 2000). This finding indicates that the neural crest cell composition within an individual branchial arch may be more heterogeneous than previously thought from classic static lineage analyses. Collectively, these analyses demonstrate that cranial neural crest cells display a wide variety of cell migratory behaviors, ranging from movements as individuals to collective cell arrangements.

Despite the range of cell movements exhibited by individually migrating neural crest cells, the overall pattern in vertebrates is remarkably similar in that cranial neural crest cells colonise the first three branchial arches in a pattern consistent with their craniocaudal origins in the hindbrain. This finding highlights the importance of hindbrain segmentation in establishing a blueprint for craniofacial morphogenesis. Cranial neural crest cells give rise to a wide variety of cell lineages that are distinct for each branchial arch (Noden, 1983; Couly et al., 1993; Köntges and Lumsden, 1996) and not only have subtle discrepancies been observed in neural crest cell migratory pathways but also in their long-term differentiated fates within each branchial arch. The most comprehensive analysis we have of cranial neural crest cell derivatives has been determined in chick, and these discrepancies have arisen as a consequence of temporal differences in chick–quail chimeras used as lineage tracers for individual rhombomeric neural crest cell population. In one study, transplantations were performed at the five-somite stage, which is well before the formation of definitive rhombomere boundaries (Couly et al., 1998). At this early stage, rhombencephalic cells have been shown not to respect rhombomere boundaries and, thus, mix with their neighboring rhombomeres (Fraser et al., 1990). Cells from the transplanted quail r3 cross into the donor chick r2 and r4 territories and migrate with r2 and r4 neural crest cells into ba1 and ba2, respectively, ultimately giving rise to cell fates and tissues not specific to endogenous r3. In a second study, transplantations were performed between five and nine somites at a time when the rhombomere boundaries are more distinct, resulting in minimal rhombomeric cell mixing before neural crest cell emigration (Köntges and Lumsden, 1996). In this study, the anterior rhombomeres were observed to contribute neural crest to the more distal parts of a branchial arch derivative with the more posterior rhombomeres contributing to more proximal branchial arch derivatives. Both studies revealed, however, that r3-derived neural crest cells do not migrate into ba1 and consequently do not give rise to skeletal branchial arch derivatives. In contrast r3- and r5-derived neural crest cells migrate deep into ba2 and contribute to derivatives such as the retroarticular process and columella cartilages along with r4-derived neural crest cells, although these populations remain segregated from each other (Köntges and Lumsden, 1996; Ellies et al., 2002).

The precise, specific rhombomeric neural crest cell contribution to the various craniofacial structures in mouse remains to be determined. Lineage tracing by focal dye labeling allows individual rhombomere populations or subpopulations to be tracked for a contribution to specific craniofacial primordia. However, this tracking can only be achieved in mouse embryos cultured in vitro, which limits the length of time the cells can be followed to approximately 2–3 days. The use of cre/lox technology to fate map murine neural crest cells has overcome the limitations of embryo culture and allows long-term tracking of neural crest cell migration and differentiation (Chai et al., 2000). However, to date, the cre/lox approach has not been specific enough to lineage label derivatives from only a single rhombomere. Combining rhombomere specific cis-regulatory sequences with cre/lox technology in the future will allow lineage tracing of neural crest cells from an individual rhombomere. This strategy is essential for comparative analyses of rhombomeric neural crest cells between mouse and other vertebrate species and a better understanding of their evolutionary contribution to craniofacial morphogenesis.


During the past century, urodele embryos were the favourite organism for experimental analyses of the role of mesenchymal/neural crest derivatives and their involvement in the development of the neuro- and viscerocranium. Significantly, it was in urodeles that neural crest cells were definitively shown to be the source of cranial and sensory spinal ganglia and that pioneering studies of neural crest-derived pigment cells were initiated (Raven, 1931; Stone, 1932; DuShane, 1935). Neural crest development in urodele amphibians is similar to that seen in frogs. Neural crest cell migration commences at approximately stage 15–16 during axolotl embryogenesis, and initially, the neural crest cells are observed as a relatively wide longitudinal band of cells on the dorsum of the embryo (Landacre, 1921; Hörstadius and Sellman, 1946; Epperlein et al., 2000). As migration proceeds, four distinct narrow streams (originally termed enlargements) of neural crest cells can be observed ventrally in the embryo (Landacre, 1921). More recently, the migration pathways of focal populations of neural crest cells have been lineage traced from seven arbitrary subdivision of the cranial neural plate (Epperlein et al., 2000). Arising from zone 3, the first enlargement of neural crest cells extends caudally to the vertical level of the posterior border of the eye. In contrast, neural crest cells derived from zone 4 at the level of the trigeminal ganglion (gasserian ganglion), which also constitutes part of the first enlargement, migrate into the prospective maxillary and mandibular arch. The second enlargement that is derived from zone 6 at the level of the VII ganglion contributes to the hyoid (second) arch. The third enlargement at the level of the IX ganglion, and the fourth inconspicuous enlargement at the level of the anterior division of the X ganglion, which correspond to zone 7, give rise to neural crest cells that contribute to the bulging gills or third to sixth branchial arches. Neural crest cells do not appear to migrate laterally from zone 5 similar to the lateral restrictions observed from odd-numbered rhombomeres in mouse and chick embryos. Of interest, the distinct neural crest cell streams appear in a spatiotemporal cranial to caudal manner; however, there is significant heterochrony in their subsequent development. For instance, 6 hr after stage 15, first enlargement neural crest from the trigeminal ganglion region has migrated ventrally and caudally to form two extensions, one anterior to the eye and one posterior to the eye entering the mandibular arch. In contrast at this same stage, neural crest from the second and more posterior enlargements have not yet migrated laterally past their respective ganglia. Approximately 16 hr after the onset of neural crest cell migration, the second and fourth streams have migrated ventrally and have moved into their respective gill arches; however, by comparison, third enlargement neural crest cells migration is minimal. Approximately 28 hr after the onset of crest migration, this heterochrony is no longer evident as all four neural crest streams have entered their prospective branchial arches (Landacre, 1921).

One of the most important findings arising from these studies was the observation that migrating cranial neural crest streams split into two distinct pathways during axolotl embryogenesis (Epperlein et al., 2000). Cranial neural crest cells either migrate into the branchial arches subadjacent to the surface ectoderm pathway where they probably generate pigment cells or they migrate along the pharyngeal endoderm where they most likely generate cartilage (Hörstadius and Sellman, 1946). This suggests that migratory restrictions may have accompanied cranial neural crest cell evolution.


In comparisons of neural crest cell migration pathways between basal vertebrates and gnathostomes, the lamprey as perhaps the most basal extant vertebrate is critical to our understanding of the neural crest cell patterning (Kuratani et al., 2001). At stage 17 (embryonic day 5) during lamprey embryogenesis, the neural plate begins to close and form a neural rod. Shortly after at stage 20, neural crest cell migration commences, and by stage 22 (E7), the presumptive branchial arch region is populated with neural crest cells (Horigome et al., 1999). Similar to mouse, chick, frog, and fish, neural crest cells from the midbrain or immediately caudal to the mid-hindbrain boundary invade the first arch (McCauley and Bronner-Fraser, 2003). Transverse sections of stage 23 embryos have revealed that neural crest cells migrating into the first arch do so both subectodermally as well as along a medial pathway to surround the mesoderm. By stage 25, midbrain-derived neural crest cells are localized to both the upper and lower lips as well as medially and laterally to the mesoderm cores.

Neural crest cells derived from the hindbrain near the otic vesicle, which correlates with the axial position of rhombomeres 4 and 5 contribute to the second (hyoid) arch during embryonic stages 23–25 (McCauley and Bronner-Fraser, 2003). Similar to the first arch, second arch neural crest cells migrate laterally and subectodermally away from the neuroepithelium and subsequently envelope the mesoderm of the hyoid arch medially and laterally. Surprisingly, some cells migrate caudally along the entire length of the presumptive pharyngeal region. By stage 23, the mandibular (first) and hyoid (second) regions have formed discrete arches separated by distinct endodermal outpockets, which partition the neural crest into steams. However, at the same stage, more caudal pharyngeal arch regions contain a column of neural crest cells that extended caudally beyond the posterior limit of the developing pharynx. Neural crest cells derived from the caudal hindbrain migrate ventrally to populate the caudal branchial arches and exhibit a variety of migratory patterns en route (McCauley and Bronner-Fraser, 2003). Caudal hindbrain-derived neural crest cells have been observed migrating both rostrally and caudally similar to that observed in odd-numbered rhombomeres in chicks and mice. Interestingly, during the early stages of neural crest migration, a gap in AP-2 expression appears in the neural tube just anterior to the otic placode (Meulemans and Bronner-Fraser, 2002). In gnathostomes, similar gaps correspond to rhombomeres 3 and 5, which are depleted of neural crest cells. The presence of only a single gap similar to that observed in axolotl embryos is suggestive of reduced patterning in the early migrating neural crest of lamprey.

These results demonstrate that cranial neural crest cells originating from the midbrain and rostral hindbrain in lampreys contribute cells to the first and second branchial arches in a manner similar to that observed in urodeles (Epperlein et al., 2000) and anurans (Collazo et al., 1993). One significant difference, however, is the degree of caudal migration exhibited by hindbrain-derived neural crest cells. Only limited migration of neural crest cells from individual rhombomeres to the adjacent rostral or caudal rhombomere has been observed in gnathostomes, whereas this is greatly exaggerated in the lamprey (McCauley and Bronner-Fraser, 2003). Another significant difference between basal vertebrates and gnathostomes is that the lamprey caudal branchial neural crest cell population is initially unsegmented, similar to urodeles and teleosts (Hörstadius and Sellman, 1946; Schilling and Kimmel, 1994), and moves in a sheet-like manner. Cells within this sheet appear to migrate freely along the entire pharyngeal region in both rostral and caudal directions, and this ability may be the result of heterochronic differences in the development of the pharyngeal pouches. The endoderm of the pharyngeal pouches contacts the overlying ectoderm that creates mechanical barriers to migration along the rostrocaudal axis. This barrier prevents further migration and cell mixing and could account for the funneling of neural crest cells into individual arches (McCauley and Bronner-Fraser, 2003). Similar patterns of neural crest migration followed by segregation occur in amphibia such as Rana, where branchial neural crest cells initially migrate ventrally as a single stream that subsequently becomes partitioned into four groups by the formation of the pharyngeal pouches (Stone, 1929). What these results suggest is that caudal hindbrain neural crest cells have not been specified to contribute to a particular branchial arch in the lamprey (McCauley and Bronner-Fraser, 2003) or possibly even in urodeles.

Neural crest cells are required for formation of branchial arch cartilages in the lamprey, and this set of cartilage bars form laterally between the ectoderm and the underlying mesodermal core (Langille and Hall, 1989). In contrast, in gnathostome fish (i.e., zebrafish), the branchial arch cartilages develop medially to the mesoderm, between the mesodermal core of the arch and the endoderm. Of interest, zebrafish neural crest cells initially migrate laterally beneath the ectoderm into the pharyngeal arch region, and then subsequently they migrate medially to form a concentric ring about a core of mesoderm (Miller et al., 2000). This observation suggests developmental differences in branchial arch cartilage formation between agnathans and gnathostomes underlie the differential localization of branchial arch cartilage condensations. Based on these topologic differences, it has been postulated that the spatial organization of cartilage bars is directly related to the migratory properties of the neural crest cells in this region (Kimmel et al., 2001). The “outside-in” hypothesis proposes that the neural crest cells that invade the mandibular arch segment of the lamprey undergo both medial and lateral migration. By extrapolation, the neural crest cells that invade the caudal branchial arches were proposed to migrate only along a lateral pathway in the lamprey, i.e., medial migration of neural crest may not occur in the lamprey and that this may explain the observation that the branchial arch cartilages in gnathostomes are located medial to the mesodermal core, while in agnathans, these cartilages are external, or lateral to the mesoderm. Lineage tracing of neural crest cell migration in lampreys suggest that the localization of cranial neural crest cells in the lamprey does not differ from that seen in the branchial arches of zebrafish (Schilling and Kimmel, 1994; McCauley and Bronner-Fraser, 2003). The neural crest cells completely surround the mesodermal core in each of the branchial arches, indicating that medial migration within the caudal branchial arches is not a new feature of gnathostomes (McCauley and Bronner-Fraser, 2003). Hence, differences in neural crest cell migration are not responsible for the distinct differentiation properties of the branchial arch cartilages in agnathans and gnathostomes.

Collectively, these results highlight that lamprey neural crest cells are somewhat less restricted in their movement than those in higher vertebrates, which could suggest migration constraints were a later evolutionary innovation. Despite more widespread migration, lampreys exhibit fewer neural crest-derived cell types compared with gnathostomes (McCauley and Bronner-Fraser, 2003). For example, no neural crest contribution to cranial sensory ganglia was observed, which suggests sensory ganglia in lampreys may be entirely placode derived. Other major differences in neural crest derivatives exist between agnathans and jawed vertebrates in both the visceroskeletal elements and the peripheral nervous system (Kimmel et al., 2001). Most notably, lamprey and hagfish, the two extant groups of agnathans, lack jaw structures. This finding suggests that agnathan neural crest cells have a lower degree of differentiation potential than their gnathostome neural crest cell counterparts.


Similar to the lamprey, marsupials are also critical to our understanding of heterochrony in neural crest cell development (Clark and Smith, 1993). Although marsupial mammals are born in a highly atricial state, the neonate must be capable of considerable functional independence. In contrast to the other vertebrate species described above, neural crest cell migration in marsupials has only been analysed to date by scanning electron microscopy (SEM). Although it is possible to lineage trace rhombomeric populations of neural crest cells in marsupials, such as the opossum Monodelphis domestica (Smith and Trainor, unpublished observations), improvements in marsupial embryo culture are still required before a thorough comparative analysis of the dynamics of neural crest cell migration can be achieved with this species.

SEM analyses of neural crest cell migration in marsupials suggest that several timing differences exist when compared with other vertebrates (Vaglia and Smith, 2003). In particular, the onset of cranial neural crest cell migration in opossums is very early relative to neural tube and somite development when compared with mouse and chick. Stage 22 in the marsupial Monodelphis domestica corresponds to approximately 10 days post coitum. At this stage, which is before the initiation of somitogenesis, the neural plate is broad, flat, and open; however, preotic (presumptive r2/3 boundary) and otic (presumptive r5/6 boundary) sulci are clearly distinguishable in the neural plate. First arch neural crest cell migration commences during stage 22 in the region immediately anterior to the preotic sulcus (Vaglia and Smith, 2003). By stage 24 (six- to eight-somite stage), a substantial mass of cells has accumulated in the anterior region of the embryo as neural crest cells begin to differentiate into the broad swellings of the first (mandibular) branchial arch. Stage 24 also marks the commencement of second (hyoid) branchial arch neural crest migration, which begins to accumulate at the level of rhombomere 4. Toward the end of stage 25 (12–13 somites), the first and second branchial arches are clearly distinguishable, particularly the mandibular arch which spans the majority of the area ventral to the midbrain and rhombomeres 1 and 2. Neural crest cells continue to migrate into the mandibular and hyoid arch regions and posterior to the otic vesicle neural crest cells begin to condense at the level of the presumptive third branchial arch (Vaglia and Smith, 2003). Branchial arch neural crest cells migrate as a broad diffuse sheet similar to lamprey and axolotl; however, relatively few neural crest cells are generated at the level of rhombomeres 3 and 5 similar to mouse and chick embryos. Thus, in many respects, the basic pattern of differentiation and migration of neural crest cells in Monodelphis domestica exhibits the primitive and highly conserved condition.

These comparative analyses highlight the general conservation in vertebrates of the migration pathways of cranial neural crest cells in discrete, identifiable streams. Even in basal jawless vertebrates such as lampreys, the neural crest cells migrate in streams analogous to the mandibular and hyoid streams. However, there are also interesting and important differences, including the heterochrony of neural crest cell migration observed in marsupials such as Monodelphis domestica and the heterochrony of an individual neural crest cell stream such as in the axolotl. Migration into the presumptive caudal branchial arches of the lamprey involves extensive rostral and caudal movements of neural crest cells that have not been described in gnathostomes, and there is no evidence of a neural crest contribution to cranial sensory ganglia in lampreys. These results suggest that barriers constraining the rostrocaudal movement of cranial neural crest cells and the generation of additional neural crest derivatives may have arisen during the evolutionary transition of agnathan to gnathostomes.


The segregation of migrating neural crest cells into streams in higher vertebrates is generated in part by the general inability of odd-numbered rhombomeric neural crest cells to migrate laterally and the establishment of clear neural crest-free zones adjacent to r3 and r5. This finding suggests that environmental influences adjacent to r3 and r5 may be critical in controlling the pathways of hindbrain neural crest migration (Farlie et al., 1999) and distinct mechanisms may be used by different species to pattern these events. For instance, unlike chick and mouse, cranial neural crest cells in zebrafish arise from the lateral portion of the neural keel, which remains for a limited period as a coherent mass adjacent to the neural keel extending from the optic cup caudally to the otocyst. At 16 hr postfertilization (14 somites), the crest cells become motile and leave this mass (Schilling and Kimmel, 1994). Schilling and Kimmel (1994) studied the migration of zebrafish neural crest cells and concluded that they migrate as segregated streams similar to other vertebrates. In contrast, in Xenopus, neural crest cells appear to migrate laterally out of the dorsal neural tube, creating an uninterrupted mass of cells adjacent to the neural tube (Hörstadius, 1950). These migratory neural crest cells remain as an uninterrupted mass until they reach the level of the otocyst, when their migration becomes restricted to three discrete streams similar to that seen in other vertebrates.

The mechanisms by which the neural crest exclusion zones adjacent to the odd-numbered rhombomeres (r3 and r5) are generated and their function in segregating neural crest cells into distinct streams remains to be resolved. Therefore, it is essential to understand the mechanisms responsible for segregating the branchial arch streams in individual vertebrate embryos and determine whether they are functionally conserved. It is important to note that, in contrast to other vertebrates, in Xenopus embryos, r5 produces equivalent amounts of crest cells when compared with the even-numbered rhombomeres (Smith et al., 1997). The variation between species in the amount and migratory pathways of neural crest cells generated by r5 suggests that the crest-free zones are not an intrinsic property of odd-numbered rhombomeres themselves.

Analyses in avian embryos reported elevated levels of cell death in the premigratory neural crest populations resident in r3 and r5 (Lumsden et al., 1991). Programmed cell death can be observed during embryonic development by using vital dyes such as acridine orange or fixed stains such as terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL). Acridine orange detects chromatin condensations specific to cells undergoing apoptosis, whereas TUNEL detects DNA breaks that are observed in apoptotic cells once the chromatin has condensed; hence, acridine orange is an earlier detector of cell death than TUNEL. In chick embryos, programmed cell death is first evident in r3 (Fig. 3A, F) at the eight-somite stage by using acridine orange, whereas with TUNEL it is first detected at the nine-somite stage. Apoptotic cells are numerous in r3 at the 9-somite stage; however, by the10-somite stage, apoptotic cells are no longer restricted to r3 and are abundantly observed in r2 and at the boundary between r1 and r2 (Fig. 3B,E). This anterior pattern of cell death persists until Hamburger and Hamilton (HH) stage 13, and the posterior pattern of apoptosis in r5 begins at 11 somites (Fig. 3C,H). At HH stage 12, additional apoptotic cells can also be observed in the ba2 and ba3 neural crest streams (Fig. 3D,G). Rhombencephalic apoptosis is no longer detectable using TUNEL after HH stage 14.

Figure 3.

Programmed cell death as depicted in the chick rhombencephalon using acridine orange or terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL). A–D: Flat-mounted chick hindbrains from eight somites (8s) to 15s. Acridine orange staining appears white. A: At the eight-somite stage, apoptosis is seen in rhombomere 3 (r3) only. HH, Hamburger and Hamilton. B: By the10-somite stage, apoptosis is seen more pronounced in r3 and also in the dorsal midline of r2 and at the r1–2 boundary. C: Cell death is first visible in r5 at the 11-somite stage. D: By the 15-somite stage acridine orange–positive cells are also seen in the migrating ba2 and ba3 neural crest streams as well as in r6. E–H: Transverse sections of an 11-somite stage chick rhombencephalon showing TUNEL staining by peroxidase conjugate (brown). E: Section through r2 showing cell death in the dorsal midline. Cells positive for TUNEL also appear to be dorsal to the neuroepithelium. F: Section through r3, also showing cell death in the dorsal midline and in ventrally migrating crest cells. G: Section through r4 showing very little if any positive TUNEL-stained cells. H: A section through r5 showing many TUNEL-positive cells in the neuroepithelium and also dorsal to the neuroepithelium.

Initially in chick, it was reported that apoptotic signals from the even-numbered rhombomeres mediated Bmp4-induced Msx2 expression in the odd-numbered rhombomeres during the period of neural crest cell formation. Consequently, it was hypothesized that inter-rhombomeric signaling led directly to the apoptotic elimination of premigratory neural crest cells in r3 and r5 and establishment of the exclusion zones that segregated the neural crest cells into discrete streams (Graham et al., 1993, 1994; Ellies et al., 2000). It has been thought that neural crest cell production along the hindbrain axis is uninterrupted. However, if one looks at the expression of a gene specific for chick neural crest cells undergoing the epithelial to mesenchymal transition, one would expect to see uninterrupted expression along the chick hindbrain axis. Yet, the expression of Slug, during chick hindbrain development, appears to tell a different story. Slug expression begins to be expressed uninterrupted along the axis of the rhombencephalon at HH stage 8 to 9+ (five to eight somites; Fig. 4A). At HH stage 10 (9–10 somites), Slug expression becomes down-regulated in the odd-numbered rhombomeres (Fig. 4B). This down-regulation correlates closely with the onset of programmed cell death at HH stage 9+ (eight somites) in the odd-numbered rhombomeres (Fig. 3A). At this stage, we also begin to see the expression of Bmp4 and Msx2 in the odd-numbered rhombomeres. Therefore, in the chick, neural crest production is interrupted along the anterior–posterior axis as shown by the down-regulation of Slug expression, which coincides with the onset of r3- and r5-specific apoptosis (Fig. 4C). Of interest, overexpressing Bmp4 or Msx2 in the chick neural tube leads to an overall (including even-numbered rhombomeres) down-regulation in neural crest cell formation and elevated levels of apoptosis (Takahashi et al., 1998; Endo et al., 2002). Conversely, blocking neural crest cell death in the odd-numbered rhombomeres does not disrupt the segregation of migrating neural crest streams (Ellies et al., 2002). This finding suggests that the segregation of migrating neural crest cells into streams requires more than simple odd-numbered rhombomere specific apoptosis.

Figure 4.

Neural crest cell death after crest emigration from rhombomere 3 (r3) and r5. A–C: Slug gene expression in chick flat-mounted hindbrains. A: A nine-somite stage chick hindbrain showing expression of Slug is uninterrupted along the entire anterior–posterior axis. B,C: At the 10-somite stage (B), Slug expression becomes down-regulated in anterior r3 and posterior r5, whereas at the 11-somite stage (C), Slug is down-regulated in all of r5. The down-regulation of Slug strongly correlates to the onset on programmed cell death in the chick rhombencephalon. D: Shows a Hamburger and Hamilton stage 14 embryo with a homotopic graft of a quail r3 into the place of chick r3. The quail cells are shown in yellow (QCPN). The chimera was also stained for terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL), shown in brown. Migratory neural crest cells from r3 in the ba2 stream also undergo apoptosis, as seen by double labeling of QCPN and TUNEL.

Because the patterns of neural crest migration from r3 and r5 are similar between the mouse and other vertebrates, it was important to examine whether there was conservation in the rhombomeric patterns of apoptosis that could account for the diminished production of neural crest cells relative to the other rhombomeres. The advantage of vital dyes, including Nile blue and acridine orange, is that the detection of apoptotic cells can be combined with in vitro embryo culture such that a picture of cell death can be accumulated over several hours during development rather than a single time point as is the case with TUNEL staining. During the period of neural crest cell migration in the mouse, cell death occurs in a temporally and spatially dynamic manner throughout the neural tube (Fig. 5). Cell death can be detected in the neural plate as early as the three-somite stage, which is before the commencement of neural crest cell migration (Fig. 5A). From the five-somite stage, at which neural crest cells are induced, through to the nine-somite stage, during which period neural crest cells migrate, the spatiotemporal patterns of cell death increase in distribution and intensity (Fig. 5A–G,I–L). Although consistent and reproducible cell death is observed in the hindbrain, there was no specific pattern that could be attributed to either odd or even rhombomeres. Even within a single embryo, the unfused neural folds often display completely different patterns of cell death, highlighting the dynamic nature of neural tube apoptosis (Fig. 5A–D). The elevated levels of cell death detected between E8.5 and E9.5 were associated with the normal caudal to rostral progression of neural tube closure, and similar patterns have also been observed in chick embryos before neural tube closure (Fig. 5H). Similar cell death patterns are also observed in TUNEL assays (Fig. 5I–L). Studies in both the Xenopus and zebrafish have also demonstrated an absence of apoptosis specific to r3 and r5 (Ellies et al., 1997; Hensey and Gautier, 1998; Del Pino and Medina, 1998). Therefore, cell death in r3 and r5 cannot be solely responsible for the segregation of the three distinct migratory streams.

Figure 5.

Nile blue and terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) staining for cell death during neural crest formation and migration. A–D: Dorsal views of Nile blue–stained neural plates in three- (3s) to seven-somite (7s) stage cultured mouse embryos exhibiting dynamic spatiotemporal patterns of cell death. E–G: Lateral views of Nile blue–stained six- to nine-somite stage cultured mouse embryos highlighting the absence of odd-numbered rhombomere (r) -specific cell death. H: Open neural plate stage chick embryo stained with Nile blue displaying extensive cell death in the dorsal neural folds. I–L: Brightfield and matching fluorescent TUNEL-stained five- and eight-somite stage embryos demonstrating the lack of rhombomere specific patterns of cell death.

The absence of localised cell death in r3 and r5 prompted the examination of the patterns of Bmp4 and Msx2 gene expression during mouse hindbrain development and neural crest migration. In contrast to the chick, mouse Bmp4 is not expressed in r3 or r5 during this period and in fact, is absent from the dorsal neural tube (Fig. 6A). Similarly, Msx2 expression, which is activated by BMP signaling in apoptosis, also differs markedly from the patterns of its chick homologue, which are confined, to r3 and r5 during hindbrain development (Graham et al., 1993, 1994). In the mouse, Msx2 is expressed uniformly in the dorsal edges of the neural tube along the entire anterior–posterior axis during the period of neural crest formation and migration (Fig. 6B). Therefore, BMP4/Msx2-mediated apoptosis of odd rhombomere neural crest cells is not responsible for patterning the pathways of neural crest migration in the mouse. Similarly, BMP/MSX signaling does not appear to play a role in zebrafish cranial neural crest cell patterning either because there is a lack of odd-numbered rhombomere-specific cell death observed in the hindbrain (Tom Schilling, personal communication). While it is possible that other Bmps may be involved, neither Bmp7 nor Bmp2 are segmentally expressed in odd rhombomeres. Interestingly, Bmp2 is strongly expressed dorsally within the embryo at the junction between the neural plate and surface ectoderm and overlaps substantially with Msx2 (Fig. 6C). In contrast to Slug expression in chick (Fig. 4A–C), in mouse the homologue Snail is uniformly expressed throughout the cranial neural tube during neural crest cell formation and migration (Fig. 6D). Therefore, although a BMP/MSX apoptotic programme may be conserved in mammals, one major difference between mammals and avians is that, in mice, neural crest apoptosis is not restricted to individual rhombomeres, which is similar to that seen in Xenopus and zebrafish.

Figure 6.

A–D: Patterns of gene expression during neural crest cell formation and migration in the mouse. Dorsal expression patterns of Bmp4 (A), Msx2 (B), Bmp2 (C), and Snail (D). Bmp4 is not expressed in the neural plate at the time of neural crest cell formation and migration. In contrast Msx2, Bmp2, and Snail are uniformly expressed along the vertebrate axis at the junction of the neural plate and surface ectoderm. r3, r5, rhombomeres 3 and 5; ht, heart.

The absence of rhombomere-specific cell death in mouse, zebrafish, and Xenopus, raises the possibility that the reduced capacity of the odd-numbered rhombomeres to generate neural crest is an intrinsic property. This issue, however, has been addressed by assaying for the ability of the odd-numbered rhombomeres to generate neural crest cells in novel environments. In mouse, homotopic grafts of r3 cells transplanted back to r3 generate very few neural crest cells. Of interest, mouse r3 and r5 cells transplanted heterotopically into r2 or r4, give rise to large numbers of neural crest cells, which populate the entire proximodistal extent of the first and second branchial arches, respectively. In chick, both r3 and r5 are able to produce, in vitro and in heterotopic grafts, large numbers of migratory neural crest cells (Fig. 4D; Graham et al., 1994; Ellies et al., 2002). Therefore, r3 and r5 have the intrinsic capability to generate neural crest cells in a manner indistinguishable from even-numbered rhombomeres. However, this capacity appears to be restricted during normal embryonic development emphasizing the importance of the microenvironment adjacent to the neural tube in regulating the pathways of neural crest cell migration (Kulesa and Fraser, 1998; Golding et al., 2000; Trainor et al., 2002).

The ability of r4 to generate neural crest cells also depends upon the environment. During normal development, r4 gives rise to numerous neural crest cells, which extensively populate ba2 and the geniculate/vestibulocochlar ganglion (VII/VIII). Similarly, when r4 cells were transposed into r2, they retained the ability to generate numerous neural crest cells, which migrate into ba1. In contrast, however, when r4 cells are transplanted into r3, their capacity to produce migrating neural crest cells is repressed. This altered capacity is not due to a transformation in the identity of the transplanted tissue nor is it due to apoptosis. Similar results are obtained in interspecies transplantations of r3 and r4 into the avian hindbrain. Together with the neural crest-free zones adjacent to r3 in other vertebrates, these results suggest that there is a common mechanism involving interactions between the odd-numbered rhombomeres and arch tissues used to govern the patterning of neural crest derived the odd-numbered rhombomeres.

These results suggest that the local environment adjacent to the odd-numbered rhombomeres prohibits the formation and/or migration of neural crest cells equally well from odd- and even-numbered rhombomeres. Furthermore, 180-degree rotations of the cranial ectoderm and mesoderm adjacent to r3–r4 in chick embryos do not alter the pathways of neural crest migration (Sechrist et al., 1994). This finding suggests that mesoderm and surface ectoderm do not appear to secrete a broadly diffusible signal that dominantly inhibits neural crest migration. Because in both of these types of transpositions, r3 and r4 cells are juxtaposed and capable of interacting with each other, our experiments show that interactions between the arch environment and rhombomeres are a critical determinants in neural crest cell patterning. Hence, the pathways of cranial neural crest migration and the crest-free zones adjacent to odd-numbered rhombomeres are generated in vertebrates by combinatorial signaling events between the hindbrain and the adjacent environment.

There are important differences in the mechanisms governing neural crest migration from r3 and r5. In terms of cell numbers, r5 in Xenopus and zebrafish embryos generates equivalent amounts of neural crest compared with the even rhombomeres. In mouse, however, r5 generates less neural crest cells than even-numbered rhombomeres but considerably more neural crest cells than r3. In contrast in chick, r3 produces more crest cells than r5. In terms of migration pathways, in mouse and chick embryos (Sechrist et al., 1993; Kulesa and Fraser, 1998, 2000; Kulesa et al., 2000), crest from the odd-numbered rhombomeres migrates rostrally and caudally to join adjacent even rhombomere streams (Figs. 1, 2). The otic vesicle is positioned immediately adjacent to r5, which provides a physical barrier rather than an exclusion zone that inhibits lateral migration from r5. This finding is probably due to the physical inhibition of the otic placode that is induced to form from the tight apposition of the paraxial ectoderm and mesoderm that blocks the space through which neural crest cells would normally migrate. In support of this, mouse mutations such as kreisler in which the otic placode is displaced, exhibit ectopic lateral neural crest cell migration from r5 (Manzanares et al., 1999). This finding is similar to Xenopus and zebrafish where the absence of a physical inhibition permits a similar lateral migration of neural crest cells from r5 (Hörstadius and Sellman, 1946; Schilling and Kimmel, 1994). This finding illustrates that there are differences in the ability of the environments adjacent to the odd-numbered rhombomeres to influence neural crest migration.


The surprising variability of cell trajectories and migratory behaviors seen in time-lapse studies of chick neural crest cells highlights the complexity of events that must be coordinated to integrate the intricate choreography of local cell movements with the global craniofacial pattern. That is to say, even though we cannot a priori predict the trajectories of two neighboring neural crest cells, a global conserved pattern still emerges. Is there a common set of guidance molecules, which maintain neural crest cell streams, defend local exclusion zones, and influence the pathfinding of cells to target destinations? Over the past several years, we have learned that there are a few candidate molecules that appear to prevent in-filling and mixing of neural crest or influence the pathfinding of cells to target destinations.

There is evidence that Eph/ephrin signaling is a mechanism that contributes to restrictions in the mixing of branchial arch neural crest populations. In Xenopus embryos, the neural crest delaminates as a contiguous anterior–posterior band and separation is achieved only during the later phases of migration through the differential expression of the Eph receptors and their ephrin ligands (Sadaghiani and Thiebaud, 1987; Smith et al., 1997). Hence, these mechanisms keep neural crest streams segregated and prevent in-filling from adjacent territories and work in concert with restrictions in the lateral migration of r3 neural crest to establish the neural crest-free zones. Another mechanism which may prevent in-filling and mixing of neural crest from adjacent territories to maintain an exclusion zone is the neuregulin receptor ErbB4. In ErbB4 mutants a population of r4-derived neural crest cells acquire the ability to migrate through the dorsal mesenchyme adjacent to r3 (Golding et al., 2000). The aberrant migration arises as a consequence of changes in the paraxial mesenchyme environment and is not autonomous to the neural crest cells. Because ErbB4 is expressed only in r3 and r5, this phenotype reflects defects in signaling between r3 and its adjacent environment (Golding et al., 2000). This mechanism has now also been shown in the chick, whereby ablation of the ErbB4-expressing territory phenocopies the mouse mutant phenotype (Golding et al., 2000). It also highlights the presence of additional mechanisms for restricting the mixing of neural crest cells, because ventrally, the neural crest streams remain segregated as they migrate into the branchial arches in mutant embryos.

To date, few molecules that influence the pathfinding of cranial neural crest cells have been identified. However, emerging evidence indicates that collapsin-1/semaphorin-III might be involved (Eickholt et al., 1999) as is the presence of a neurite outgrowth inhibitor in the mesenchyme adjacent to r3 (Golding et al., 2000). This neurite growth inhibitor in the r3 adjacent mesenchyme, patterns central projections of cranial sensory axons and may be part of the mechanism that restricts lateral neural crest migration from r3 into the adjacent mesenchyme (Golding et al., 2000). This is consistent with time-lapse analyses in the chick showing that crest cells arising in r3 extend filopodia into the adjacent territory but retract these projections and fail to migrate laterally (Kulesa and Fraser, 1998). Positive, emigration-provoking signals, potentially in concert with extracellular matrix differences, may also be present in the paraxial mesoderm, ectoderm, and or endoderm.

Future studies of neural crest cell guidance may benefit from information shared between embryonic and adult systems, particularly because certain cell migratory behaviors appear to be common among cells in both embryonic and adult systems. For example, several model systems share the collective cell migratory behavior of forming chain-like structures. In the brain of adult mice, cells in the subventricular zone migrate up to 5 mm to the olfactory bulb, where they differentiate into neurons. These migrating cells move as chains, which are composed of elongated, apposed neuroblasts connected by membrane specializations, moving with each other and not guided by radial glial or axonal fibers (Lois et al., 1996). Disruption of Eph/ephrin signaling affects neuroblast cell migration (Conover et al., 2000). In another model system, the cellular slime mold Dictyostelium, individual cells form multicell-wide chains en route to assembling a multicellular organism by using a complex relay system of chemotactic signals (Dormann et al., 2002; Meili and Firtel, 2003). Thus, in the future, insights from each of these model systems will provide clues to help dissect the mechanisms of neural crest cell patterning and migration.

Importance of Segregated Neural Crest Streams

Despite incredible ranges of individual cell movements and behaviours, cranial neural crest cell migration pathways are remarkably similar in all vertebrates. Cranial neural crest cells migrate in discrete segregated streams separated by clear neural crest-free exclusion zones. An important question is why is it necessary to establish these exclusion zones and restrict r3 and r5 neural crest migration? Because neural crest cells arising at distinct axial levels contribute to the establishment of the unique anterior–posterior identity of structures in the adjacent branchial arches, the neural crest-free exclusion zones could be critical for branchial arch patterning by preventing or limiting the intermixing of neural crest cell populations with different positional identities. One example of this could be the need to prevent mixing between non–Hox-expressing and Hox-expressing neural crest cells, which populate ba1 and ba2, respectively. A variety of gain- and loss-of-function studies have shown that Hoxa2 is primarily responsible for specifying second branchial arch fate and can also inhibit normal ba1 development (Rijli et al., 1993, 1998; Gendron-Maquire et al., 1993; Couly et al., 1998; Kanzler et al., 1998; Grammatopoulos et al., 2000; Pasqualetti et al., 2000). This is borne out by the fact that overexpression of Hoxa2 in first branchial arch neural crest cells, transforms the identity of the first arch into that of a second arch and subsequently suppresses jaw formation (Grammatopoulos et al., 2000; Pasqualetti et al., 2000). Hoxa2-expressing neural crest cell populations, therefore, must be excluded from the first branchial arch. Conversely, in Hoxa2 null mutant mice, the second arch is transformed into a first arch identity, resulting in duplicated first arch skeletal structures. This finding highlights the incompatibility of Hox expression and jaw development (Creuzet et al., 2002).

The segregation of neural crest streams also prevents cranial ganglionic fusions as is evident in mouse mutations such as Krox20 and ErbB4 where neural crest streams were able to mix (Schneider-Maunoury et al., 1997; Golding et al., 2000). The loss of ErbB4 expression from r3 leads to the aberrant migration of second arch neural crest cells through the exclusion zone, which results in the fusion of the geniculate/vestibulocochlear to the trigeminal ganglia.

Hence, the segregation of neural crest cells into discrete streams is a prerequisite for proper craniofacial development. This segregation is achieved in part by restricting the generation of neural crest from the odd-numbered rhombomeres and also by signaling from the paraxial environment that generates exclusion zones and keeps the distinct streams from intermixing as they migrate into the adjacent branchial arches. Furthermore, the tight contacts and communication between migrating neural crest cells derived from a similar location, may help to regulate the influences of inhibitory signals in the arch environment (Kulesa et al., 2000; Kulesa and Fraser, 2000). The identity and fate of migrating cells is appropriately regulated through a balance between intrinsic cues acquired in the hindbrain during their formation and extrinsic cues such as cell plasticity and cell community effects mediated by the paraxial environment through which they migrate (Trainor and Krumlauf, 2000; Schilling et al., 2001).

Evolution of Jaw Articulation and the Potential Importance of Localized Cell Death

The development of jaws was perhaps one of the greatest advances in evolutionary history, because it precipitated a revolution in feeding habits and lifestyle. Jaw evolution encompasses the origins of a jaw articulation that allowed an upper and a lower jaw to function together. In the primitive form, the palatoquadrate was fused to the cranium (hyosymplectic) and, thus, was immovable. This organization allowed the mastication muscles to insert on Meckel's cartilage and the palatoquadrate (Fig. 7A). In amphibians, the squamosal bone replaced the hyosymplectic, and the palatoquadrate became the quadrate (Fig. 7B). Jaw articulation in the amphibian is also immovable, which is similar to fish and referred to as monimostylic (de Beer, 1985) and importantly, the mastication muscles in amphibian have insertions similar to those found in teleosts (Edgeworth, 1935). In contrast, avians developed a distinct new type of jaw articulation, whereby the quadrate is detached from the upper jaw and reduced in size (Fig. 7C). The quadrate in aves functions as a pivot point between the cranium and Meckel's cartilage, and this new function for the quadrate gives rise to new mastication muscle attached sites, which are normally found inserted into the quadrate (Edgeworth, 1935). The insertions of the mastication muscles are now found on the squamosal bone of the cranium, and the inner palate (Edgeworth, 1935). This type of jaw suspension, with a movable quadrate, is termed streptostylic (de Beer, 1985).

Figure 7.

Jaw articulation evolution. Schematics show the upper and lower jaw (mandibular arch) in red, the second/third arch in blue, and the cranium in mustard-brown. Dorsal is to the top and anterior to the left. The insertion and origin of the mastication muscle, adductor externus, is shown in green. A–D: The different jaw articulations for the teleost (A), amphibian (B), avian (C), and mouse (D). E: Depicts the evolution from basal extant vertebrate to the advanced mammal. Also shown is the specific type of jaw suspension. Fish and amphibians have a monimostylic suspension, whereas avian and mammals (incus is movable) have a streptostylic suspension. The differences between the avian and mammals in jaw suspension rests in the loss of the quadrate, thus making the mammalian suspension more similar to the amphibian and fish suspension. This difference between chick and other species is reflected in presence or absence of programmed cell death in the odd-numbered rhombomeres. art, articular; c, columella; ch, ceratohyal; hh, hypohyal; hs, hyosympletic; m, meckels; p, partoic; q, quadrate; rap, retroarticular process; sh, stylohyoid; sq, squamosal; st, stages.

In mammals, the squamosal became the new articulation site for the lower jaw and the shift from the quadrate articular joint is correlated with an expansion of the dentary bone (Fig. 7D). The expansion of the dentary bone also coincides with the formation of the ramus on which the mastication muscle is inserted (Edgeworth, 1935). When the dentary bone acquired articulation with the squamosal bone, the articular and quadrate bones were freed to perform new functions or to disappear and Reichert suggested that the mammalian ear ossicles were derived from these regions of the jaw apparatus (de Beer, 1985). According to this hypothesis, the articular separated from Meckel's cartilage in the lower jaw and gave rise to the malleus. In contrast, the quadrate disappeared from the upper jaw apparatus in mammals and developed into the incus. Because the quadrate has disappeared in the mammal jaw, the terms monimostylic and streptostylic cannot be used alone. As the quadrate has become the incus, and is clearly free from the squamosal, thus movable, it can be referred to as streptostylic, However, the upper jaw is firmly fused to the cranium and, thus, akinetic. Therefore, in the mammal, the streptostylic quadrate (incus) is associated with an immovable akinetic bony upper jaw. Kinetic classification is based on the relation between the upper jaw movability and the cranium, instead of whether the quadrate is movable in relation to its neighbor (de Beer, 1985).

These musculoskeletal evolutionary comparisons reveal similarities between jaw articulation in the mouse, fish, and frogs as well as clear differences to the chick (Fig. 7A–E). The formation and general migratory patterns of cranial neural crest cells are remarkably similar in all vertebrate species, which raises the issue of how these species-specific differences arise. Comparative analyses of neural crest cell patterning in vertebrates demonstrates that odd-numbered rhombomere cell death is a specific characteristic of chick embryos. Fish, frogs, and mice do not display any specific programmed cell death in odd-numbered rhombomeres. This finding raises the interesting possibility that localized cell death has played an important role in the evolution of jaw articulation. If role this is true, then inhibiting odd-numbered rhombomere-specific cell death in chick embryos should result in alterations to musculoskeletal development.

Recently odd-numbered rhombomere-specific cell death was inhibited in chick embryos by the overexpression of a WNT antagonist sFrp2, which blocks BMP signaling in the hindbrain (Ellies et al., 2000). Although the segregation of neural crest cell streams was unaffected, when apoptosis was arrested in r3 and/or r5 in chick embryos, individual r3- and r5-derived neural crest cell migratory pathways changed (Ellies et al., 2002). Neural crest cells from r3 were observed migrating anteriorly as normal, but they continued well beyond the trigeminal ganglion (V) into the medial region of ba1. Similarly, posteriorly migrating r5-derived neural crest colonised the distal region of ba3 in addition to the petrosal ganglia (IX). The consequence of this extended migration is the formation of ectopic muscle attachment sites, reminiscent of those in other organisms that are not subjected to the same cell death program. Thus, it appears that a WNT-BMP signaling loop mediates apoptosis in chick r3 and r5 and that the removal of these neural crest precursors was instrumental in the evolutionary elimination of unnecessary jaw muscle attachment sites (Ellies et al., 2002). Therefore, odd-numbered rhombomere-specific cell death, which may have been acquired uniquely in the evolutionary aves lineage branch, is necessary for the proper development of species specific jaw articulation in chicks.