The enteric nervous system (ENS) develops primarily from rhombencephalic neural crest cells that, upon entering the pharynx, proliferate, migrate, and colonize the entire length of the gastrointestinal tract (LeDouarin and Teillet, 1973; Tucker et al., 1986; Burns and LeDouarin, 1998; Young and Newgreen, 2001). Unlike the crest cells that form the sympathetic and spinal sensory ganglia, the enteric crest cells continue to migrate caudally for a protracted period. In the chick embryo, rhombencephalic crest cells leave the neural tube, enter the pharynx at embryonic day (E) 3, and reach the cloaca at E6–7 (LeDouarin and Teillet, 1973). In the mouse, enteric crest cells are found in the foregut at E10 and invade the terminal hindgut at E14 (Kapur et al., 1992; Young et al., 1998). Although several studies have advanced our knowledge regarding the role of crest-derived cells in the development of the enteric nervous system (Young and Newgreen, 2001; Young et al., 2002), the spatial relations of pioneering crest cells to each other and to the surrounding gut mesenchymal cells have not been systematically examined as the cells progress throughout the gut.
Extensive analyses have identified critical roles for several ligand–receptor interactions (e.g., GDNF:Ret, ET-3:EtRB) in maintaining the movements and mitotic competence of the enteric crest cell population (Young and Newgreen, 2001, Newgreen and Young, 2002a, b). A key aspect of this process is balancing proliferation of enteric crest cells and the formation of postmitotic neurons and glia. In a recent study, Young et al. (2002) show that neurons appear very close to the migratory wavefront. We have also studied the locations of cells expressing markers for avian neural crest, neurons, and glia during the initial colonization of the gut. Our results reveal that pioneering enteric crest cells form a lattice-like network of interconnected cells in the primitive gizzard. At the leading edge in the midgut, cord-like chains of crest cells extend ahead of this network just beneath the serosa, whereas in the hindgut, crest cells appear both beneath the serosa and in the submucosal mesenchyme. In both cases, neurons and glia appear simultaneously just behind (proximal to) the wavefront of colonizing crest cells. As the leading front of migration passes each region, the chains of crest cells undergo extensive branching and reorganization to form a cellular network. At the intersections of these branches, small clumps of neurons appear, presaging the distribution of enteric ganglia, while crest cells that connect these nodes remain undifferentiated for a period of time.
We examined the pattern of migration of enteric crest-derived cells as colonization progressed from the gizzard to the colon by using antibodies to HNK-1, which marks enteric crest cells, and antiserum to Hu and β (III)-tubulin, which label neurons. At E3.5 in the chick, Hu+ cells in the gizzard are localized to two areas; a caudal group composed of small clumps of Hu+ cells, and a rostral group of cells that appear isolated from one another and are closely associated with β (III)-tubulin+ fibers from the vagus nerve (Fig. 1A1,A2). Most of the Hu+ cells have small β (III)-tubulin processes (data not shown, see Fairman et al., 1995). These observations indicate that neuron differentiation appears in two distinct regions of the gizzard. The Hu+ cells (green label) are found within the extensive network of connected HNK-1+ cells (red label) located just beneath the serosa of the gizzard. Elongated strands of crest cells extend from the caudal edge of the gizzard to enter the intestine (Fig. 1B1). The most caudal of these HNK-1+ cells appear as chains of connected cells. Within these strands, we observed immunoreactivity to the glial cell marker glial fibrillary acidic protein (GFAP; Fig. 1B2,B3). Usually the GFAP immunoreactivity (GFAP-IR) is found in processes that appear a few cells proximal to the most advanced HNK-1+ cells. These processes and occasionally small GFAP+ cell bodies are present throughout the area colonized by the neural crest cells at E3.5 and continue to appear along with the HNK-1+ cells throughout the period of gut colonization. Some of the cells in the strands of HNK-1+ cells extending into the intestine also express the neuronal marker Hu. These are present as single cells or clusters of two or three cells near the advancing front of HNK-1+ crest cell cords (Fig. 1C). Hu+ cells invariably had GFAP+ processes closely associated with them, but these processes were also found in areas of HNK-1+ cells with no Hu-IR.
At E4.5 in the chick, the chains of HNK-1+ cells have advanced to the region of the yolk stalk (Fig. 2A1). Cross-sections taken at and near the wavefront indicate that most of the HNK-1+ cells are found just under the serosa, although an occasional cell appears deeper in the gut wall mesenchyme (Fig. 2A2). The individual chains in the preumbilical gut are often curved and branched, whereas the leading HNK-1+ strands in the postumbilical gut appear straighter along the rostrocaudal axis of the gut and show fewer branches (Fig. 2B). At this time, the HNK-1+ strands in the postumbilical gut also appear to extend out further in advance of the most caudal Hu+ cells.
Between E5 and E6 in the chick, HNK-1+ cells colonize the remainder of the midgut, including the cecal buds. In the caeca and hindgut of the chick embryo, mesenchymal cells also show HNK-1+ immunoreactivity (Luider et al., 1992), which prevents clear visualization of the neural crest–derived cells. Therefore, we used the equivalent stages of quail embryos, which show less mesenchymal staining (Newgreen et al., 1996) to document the terminal migration of the enteric precursor cells. The pattern of migration/colonization was found to be similar in the chick and quail for the gizzard and midgut, supporting the assumption that colonization in the hindgut is also similar in the chick and quail.
Behind the cords at the wavefront, the crest cells realign to form a mesh-like reticular pattern of interconnected strands. Here, Hu+ neurons are consistently present at the nodes (Fig. 2C). Sections behind the wavefront in the quail show HNK-1+ cells deep within the mesenchyme (Fig. 2D1,D2); these cells are likely to be precursors of the submucosal plexus. In contrast, the HNK-1+ cells near the wavefront are mostly confined to the area under the serosa. Isolated HNK-1+ cells are rarely seen out in front of the leading strands. GFAP immunostaining is present in the leading migratory strands as long, thin processes (Fig. 2E) and behind the wavefront as a highly branched network of processes associated with the Hu+ neurons (Fig. 2F). As described above, Hu+ cells had GFAP-IR closely associated with them in all cases examined but GFAP-IR was not always found with Hu-IR.
By E6.5 the neural crest-derived enteric precursor cells in the chick have reached the ileo-caecal junction and have entered the hindgut. The pattern of colonization in the hindgut is markedly different from that described above for the midgut. The differences are manifested in the distribution of crest cells in the myenteric and submucosal regions. In whole-mounts of the quail colon, the network of HNK-1+ cells in the myenteric region is extensive and highly branched (Fig. 3A), similar to that seen in the preumbilical gut. However, in both transverse (Fig. 3D) and longitudinal (Fig. 3B1) sections, the additional presence of a robust submucosal population of HNK-1+ cells is evident. This submucosal population extends more caudally at this stage than the myenteric crest cell population (Fig. 3B). Because of technical considerations we were unable to determine whether the submucosal cells extend as chains like those seen in the midgut. Another difference is that Hu-IR is both more extensive and more caudal in the submucosa than in the myenteric region (Fig. 3B2,B3,D).
An interesting exception to the more advanced migration in the submucosal region is the presence of a single myenteric strand that forms on the side of the gut opposite Remak's nerve and extends beyond the remainder of the myenteric network toward the terminal colon (Fig. 3A). This figure documents the presence of this strand in the quail; however, similar observations were made in chick. We also observed Hu+ cells in the myenteric HNK-1+ network associated with β (III)-tubulin–positive fibers extending from Remak's nerve into the gut (Fig. 3C1). This association is similar to that observed between Hu+ cells and fibers from the vagus in the E3.5 foregut. However, we did not observe Hu+ cell bodies in the branches from Remak's nerve at this stage (E6.5 quail), and it is unlikely that the Hu+ cells are migrating from Remak's into the gut at this stage. By E7 in quail, the crest-derived cells have reached the end of the colorectum and there is contact between the HNK-1+ cells and fibers extending from the nerve of Remak and the pelvic plexus adjacent to the cloaca (Fig. 3E). At this stage, we did see Hu+ cells within the branches from Remak's nerve. This observation is consistent with those of Burns and LeDouarin (1998), who described sacral crest entering the hindgut by means of fibers of Remak's nerve.
The migration of ENS progenitors is rapid, estimated at 40 μm/hr (Allan and Newgreen, 1980), and colonization involves the integration of mitosis, movement, differentiation, and plexus formation. We have studied the pattern of organization of crest-derived cells as they progress from the gizzard to the termination of the hindgut. Our observations indicate that the crest cells colonize the gut as chains of cells that branch and intersect, not as isolated migratory cells. Behind these leading chains, the cells become intertwined to form a fishnet-like pattern, presaging the spatial relations of enteric ganglia seen in the adult. In the midgut, crest cell chains are found just beneath the serosa near the wavefront but appear deeper, proximal to (behind) the leading edge. This finding indicates that the initial branching occurs within a circumferential plane, but that subsequent branching involves a radial component. In contrast, in the hindgut, the crest cells advance in the submucosa before the myenteric region.
Patterns of Crest Cell Movement
We expected to find many single migrating crest cells at the leading edge in the colonizing population and were surprised to find the crest cells arranged as chains at the wavefront. Because the first description of the neural crest by His nearly 150 years ago, the verb migrate has been applied to crest cells with the implication that the cells move as individuals. However, several observations indicate that crest cells may be associated as they advance. For example, scanning electron microscopic examination of crest cell migration in the avian trunk reveals the presence of both isolated cells and cells “aligned end to end” (Tosney, 1978), which resemble the leading edge chains we find in the gut wall. Crest cells moving on the surface of the myocardium are also arranged in cords (Poelmann et al., 1998). A close association among “migrating” crest cells is especially prominent in the head. Individual cells are rarely seen in front of the leading edge (Anderson and Meier, 1981). Moreover, it has been argued that part of the movement of cephalic crest cells involves a mediolateral expansion of the crest cell population together with a coordinated lateral displacement of the overlying surface ectoderm and underlying myogenic mesoderm (Noden, 1984).
We previously reported the presence of chains of enteric crest cells at the wavefront (Epstein et al., 1991), but the current study provides evidence that these chains constitute the primary pattern of organization at the leading edge as crest cells advance down the avian gut. Others have also seen enteric crest cells aligned in cords in chick (Newgreen et al., 1996) and mouse embryos (Young et al., 1998, 1999), although Young reported both single cells and chains of cells in advance of the wavefront, and Newgreen et al. (1996) found strands of crest cells separated from the leading edge in chick. We rarely saw cells separated from the wavefront. It is not clear whether this difference is a technical problem involving immunolabeling methods or a species-specific difference in the ratio of single cells to those in chains. Natarajan et al. (1999) injected small numbers of dissociated crest cells into the mouse gut; these cells migrated and differentiated but were insufficient to form extensive neuronal networks. Because the number of single cells or cells in chains distinct from the wavefront seen by Newgreen et al. (1996) and Young et al. (1998, 1999) is small, it is unlikely that these mediate complete colonization of the gut. We postulate that the main mode of colonization at the wavefront of migration occurs by the extension of cords of cells.
The appearance of the chains of crest cells at the wavefront of migration bears a striking resemblance to endothelial cell cords present during the initial vascularization of the central nervous system. Branches from a vascular plexus located outside the basement membrane of the brain penetrate the neuroepithelium at precise sites and then branch to form an intramedullary plexus (Noden, 1991a). This behavior is in contrast to the invasive movements of single angioblasts that are present in most mesenchymal environments (Noden, 1991b). In these penetrating endothelial cords, the leading cells extend long processes and initiate branching, whereas endothelial cells immediately behind constitute a mitotic cuff from which additional cells are produced. If a similar situation occurs in the gut, we would predict that the position of established nascent ganglia would not change as colonization progresses. It remains to be determined whether a similar situation exists in the enteric crest cells.
Although no direct evidence for cell–cell connections exists, it is likely that some junctional complex exists between the enteric crest cells. Studies of Lo and coworkers (Lo et al., 1997; Xu et al., 2001) suggest that gap junctions are found between migrating neural crest cells in culture. Additional work is necessary to characterize the connections between enteric crest cells in situ.
In the foregut and midgut, the most caudal cells at the wavefront are seen just beneath the serosa, whereas proximal to the wavefront, cells are found both beneath the serosa and within the depths of the mesenchyme. Newgreen et al. (1996) and Burns and LeDouarin (1998) identified the leading edge crest cells below the serosa at the wavefront, with subsequent penetration to the deeper regions of the gut wall. In the hindgut, our observations indicate that the submucosal and myenteric crest cells advance caudally as two separate and distinct networks. Although sectioned material showed the crest cell migration was more advanced in the submucosal region of the hindgut, whole-mounts clearly indicated a branched network and an advanced migratory strand in the myenteric region. By using quail/chick chimeric grafts, Burns and LeDouarin (1998) found that crest cells populated the submucosa before the myenteric region, and suggested that myenteric ganglia are formed from submucosal crest cells migrating peripherally through the circular muscle. With immunostaining, we find this pattern in the quail hindgut as well, but our observations indicate that HNK-1 cells form a network that advances caudally in the myenteric region. It is possible that this network of myenteric HNK-1+ cells arises from submucosal cells migrating to the myenteric ganglia as Burns and LeDouarin (1998) show. However, our observations indicate a substantial myenteric network and only occasional HNK-1+ cells located between the submucosal and myenteric regions just proximal to the wavefront. Many cells would be expected between the two plexuses if the myenteric plexus was being formed from migrating submucosal cells. Therefore, we speculate that cells in both the myenteric and submucosal paths contribute to the formation of the myenteric ganglia.
Burns and LeDouarin (1998) found extensive neurofilament-IR in the myenteric plexus and concluded that myenteric ganglion cells express a neuronal phenotype before submucosal ganglion cells. We find that more Hu+ cells appear in the submucosa and that they appear in advance of those in the myenteric region. At present, we have no explanation for the discrepancy between our results and the results of Burns and LeDouarin. It could be due to the difference in choice of antibodies, and clearly further investigation of this issue is in order.
Although our observations do not exclude the possibility of some cell movement between the deep and superficial populations, the feature of note is the presence of both deep and superficial routes of ingression by crest cells. We have not determined the site and stage when these two routes appear. In contrast, the myenteric region appears to be the main route of migration in the mouse colon (McKeown et al., 2001).
Pattern of Neural Crest Cell Differentiation
Cells expressing neuronal and glial markers appear as early as E3.5 within the chains near the leading edge of the crest population. Later, neurons become preferentially localized at the nodes where branches of HNK-1–labeled crest cells intersect. The early formed neurons may serve as foci around which the network of crest cells becomes organized. The subsequent formation of ganglia at these foci might occur by the movement of additional Hu+ cells to these aggregation centers or by local recruitment of neuroblasts from nearby neural crest cells. Experiments in the developing avian dorsal root ganglion (Marusich et al., 1994) indicate that Hu-IR appears either at or just after the time of cell cycle exit. It is not known whether a similar process occurs in the gut, whether the terminal division is symmetrical (both daughters are Hu+) or whether the Hu-labeled cells continue to divide or to advance. Additional experiments to address these questions will provide a better understanding of the contribution of single Hu+ cells to the final network.
We also observed an apparent association between the appearance of some neurons in the foregut and hindgut and the presence of fibers of the vagus and Remak's nerve, respectively. The significance of this is not clear, because in the present study, we could not distinguish whether neurons differentiate in response to contact with extrinsic fibers, preferentially move to these axons, or if the extrinsic branches grow toward and contact differentiated neurons. We would expect that, if the extrinsic fibers were seeking out differentiated neurons, then their paths would be tortuous and complex; but they are not. In other parts of the embryo, it is known that axons recruit crest cells to the Schwann cell lineage (Noden, 1975). Additionally, the vagus is known to produce a variety of neurotrophic factors (Helke et al., 1998), providing evidence that the extrinsic fibers may trigger the appearance of neurons in these instances, although clearly throughout most of the gut, neurons differentiate in the absence of extrinsic innervation. This finding suggests the possibility of different sets of cues for neuron differentiation within the gut.
Cells expressing the glial marker GFAP also appear just proximal to the wavefront of advancing HNK-1+ cells. GFAP-IR is most apparent in processes within the HNK-1+ strands, while profiles resembling astrocyte-like cells were also observed. Although a recent report (Zhang, 2001) indicates that the GFAP antiserum used here stained epithelial cells from human embryonic stem cells and may recognize other intermediate filament proteins, the immunostained profiles we observe are consistent with those of astrocytes and unlike those observed in epithelial cells. Our study confirms the observations of Balaskas and Gabella (1998), who first reported the presence of GFAP-IR in the chick gut at early stages of development. Others have also observed GFAP-IR at late stages in the development of the chick (Payette et al., 1984) and mouse (Rothman et al., 1986). Recent observations indicate that glial markers may appear earlier in the mouse gut but still well behind the wavefront (H. Young, personal communication). Our study extends the observations of Balaskas and Gabella by showing the early appearing relationship of GFAP-IR to the HNK-1+, and Hu+ cells at the wavefront. We show that the neurons and glia appear at the same developmental stage in neighboring cells within the chains of crest cells. Although it is thought that neurons develop before glia (see review by Zhang, 2001), recent work indicates that oligodendria and neurons are specified at about the same time in development (Chandross et al., 1999; Zhou et al., 2000).
In summary, we have shown that neural crest cells move caudally through the wall of the avian gut not as single cells but as chains of cells. Along most of the gut, this movement occurs just under the serosa, but in the hindgut, an additional and separate submucosal population is present and appears in advance of the myenteric component. The chains in the foregut and midgut branch extensively at the migratory front, leaving behind a fine network of connected cells. Throughout most of the gut, neurons and glia first appear within the migratory strands at the wavefront, although in parts of the foregut and hindgut, neurons additionally appear closely associated with extrinsic fibers. Neurons initially appear as single or multiple cells within the crest cell chains at the leading edge of the wavefront. As the front moves caudally, these neurons become located at branch nodes, presaging the sites of enteric ganglia formation.
Fertlized Leghorn chicken (Gallus gallus) and quail (Coturnix coturnix japonica) eggs were incubated at 38.5°C and 37.5°C for various times, respectively. The gastrointestinal tracts were removed from embryos and fixed in either 4% paraformaldehyde (Epstein et al., 1991) or a 1:1 mixture of acetone:methanol for 12 hr at 4°C. The fixed tissue was placed in phosphate buffered saline (PBS) containing 2% Triton X-100 for roughly 6 hr and then placed in antibody diluent (PBS with 0.3% Triton X-100, 5.0% goat serum and 0.05% sodium azide) containing HNK-1 (ascites produced from cells obtained from American Type Culture Collection ([ATCC], Manassas, VA;1/1,000), human anti-Hu (1/1,000; Fairman et al., 1995), and mouse anti-β (III)-tubulin (1/500; Lee et al., 1990a, b) for 12–24 hr at 4°C. Hu recognizes a family of RNA-binding proteins (Darnell, 1996; Okano and Darnell, 1997), and β (III)-tubulin is an isoform of tubulin unique to neurons (Lee et al., 1990a, b). Tissues were washed and incubated in a mixture of goat anti-mouse immunoglobulin (Ig) M-Texas Red (1/100; μ-specific, Jackson Labs, West Grove, PA), goat anti-mouse CY2 (1/100; γ specific, Jackson Labs), and goat anti-human IgG-Alexa 488 (1/100; Molecular Probes, Eugene, OR). Tissues were also incubated in a combination of rabbit anti-GFAP (1/1000, Dako, Carpinteria, CA) and either HNK-1 or Hu. The GFAP was visualized with biotinylated goat anti-rabbit (1/1,000) and streptavidin CY3 (1/1500, Jackson Labs). Whole-mounts of the gastrointestinal tracts were mounted on slides in a mixture of 50% glycerol and 50% PBS and observed through a Nikon fluorescence microscope equipped with fluorescein isothiocyanate and Texas Red filters. Images were acquired with a SPOT 2 camera (Diagnostic Instruments, Detroit, MI) and processed initially with Metamorph software (Universal Imaging, Downingtown, PA). Figures were prepared with Adobe Photoshop and Adobe Illustrator.
We thank Dr. Edward Schultz for reading the manuscript; Dr. Tony Frankfurter, University of Virginia, for antibodies to tubulin; and Tim Bert and Andy Swetlik for dissection of embryos.