Role of the Isthmus and FGFs in Resolving the Paradox of Neural Crest Plasticity and Prepatterning

See allHide authors and affiliations

Science  15 Feb 2002:
Vol. 295, Issue 5558, pp. 1288-1291
DOI: 10.1126/science.1064540


Cranial neural crest cells generate the distinctive bone and connective tissues in the vertebrate head. Classical models of craniofacial development argue that the neural crest is prepatterned or preprogrammed to make specific head structures before its migration from the neural tube. In contrast, recent studies in several vertebrates have provided evidence for plasticity in patterning neural crest populations. Using tissue transposition and molecular analyses in avian embryos, we reconcile these findings by demonstrating that classical manipulation experiments, which form the basis of the prepatterning model, involved transplantation of a local signaling center, the isthmic organizer. FGF8 signaling from the isthmus altersHoxa2 expression and consequently branchial arch patterning, demonstrating that neural crest cells are patterned by environmental signals.

The cranial neural crest is a pluripotent migratory cell population that plays a critical role in the construction of the vertebrate head, giving rise to the facial and visceral skeleton, most of the skull bones and connective tissue, and the neurons and glia of the peripheral nervous system (1–3). The highly conserved segmental organization of the vertebrate hindbrain into rhombomeres (4, 5) plays a key role in patterning the identity and pathways of neural crest cell migration into the branchial arches (6–12). Currently, there is a fundamental paradox in mechanisms that pattern neural crest cells and their derivatives. Noden grafted first-arch neural crest precursors posteriorly to new locations in avian embryos, and these ectopic crest cells gave rise to duplications of first-arch skeletal derivatives, such as the quadrate and Meckel's cartilage. This landmark transposition study (2) led to the model that cranial neural crest cells are preprogrammed in the neural tube before their migration and that they passively carry positional information necessary for craniofacial morphogenesis from the neural tube to the periphery. This prepatterning model has shaped the way we think about craniofacial development during the past 18 years and has also been used to explain skeletal duplications observed in null mutations of A-P patterning genes, such as Hoxa2 and Hoxa3(13–15). However, recent transposition and lineage tracing experiments contradict the prepatterning model, highlighting the plasticity of rhombomeres and cranial neural crest populations [(11, 12,16–23) and reviewed in (5)]. These studies suggest an alternative dynamic model, in which neural crest patterning relies on a balance of instructive signals from the hindbrain, maintenance signals from the branchial arch environment, and cell community interactions.

In this study, we performed experiments aimed at understanding and resolving the basis for these conflicting models and results. An often-ignored aspect of Noden's analysis is that posterior transplantations of presumptive frontonasal or presumptive first-arch neural crest both produced the same quadrate and Meckel's cartilage duplications. Hence, the same ectopic structures formed irrespective of the axial origin of the neural crest cells. What links these different transplantations is the probable inclusion of the mid/hindbrain isthmus in the grafted tissue. In recent years, it has become apparent that local inductive centers, such as the mid/hindbrain junction (isthmus), play roles in anterior neural patterning (24). Noden used the isthmus as a morphological marker for delineating the neural tissue to be grafted posteriorly (Fig. 1A), and therefore one possible explanation for the conflicting results may relate to the inclusion of a localized signaling center along with neural crest progenitors.

Figure 1

Transposition of the mid/hindbrain isthmus posteriorly in place of r4 reprograms Hoxa2 expression in cranial neural crest cells. (A) Diagram illustrating transposition. Noden grafted the region from the isthmus to the boundary between r2 and r3. Our isthmus (i) transplantations comprised only the FGF8-expressing territory at the mid/hindbrain junction. In both cases after removal of endogenous r4, the anterior territory containing neural crest cell (ncc) progenitors was transposed posteriorly into the caudal hindbrain at the level of r4 (arrows) FB, forebrain; MB, midbrain; ov, otic vesicle; s1, somite 1. (Band C) Fgf8 in situ hybridization in a 2.5-day avian embryo showing high levels of expression in the isthmus and transient domains in branchial arches (black arrowhead) in a control embryo (B) and in a grafted embryo (C) 24 hours after transposition of the isthmus in place of r4 (*). (D and E) In comparison to a control embryo [(D), white arrow], Hoxa2is not expressed in grafted isthmus tissue or in migrating neural crest cells derived from the graft [(E), white arrows]. (F) DiI lineage tracing shows that despite the lack of Hoxa2expression, there is extensive migration of neural crest cells from the graft into the second arch (white arrow). ba2, second branchial arch.

To directly test this idea, we transplanted the isthmus posteriorly in place of rhombomere 4 (r4) in ovo, in stage-matched chick embryos at somite stage 8 to 9 (8-9) (Fig. 1, A through C). The donor isthmus included the mid/hindbrain junction and a small population of cells on both sides of the boundary (Fig. 1A). After 24 to 48 hours of in ovo culture, grafted embryos were assayed for effects on Hoxa2expression (Fig. 1, D and E), which is the primary determinant of the second branchial arch neural crest phenotype (13,14). Hoxa2 expression in the second branchial arch neural crest was inhibited (Fig. 1E), and this was not due to an absence of migrating neural crest cells, because 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbonocyanine perchlorate (DiI) labeling of the transplanted tissue shows that numerous graft-derived neural crest cells emigrated and populated the second branchial arch (Fig. 1F). The inhibition of Hoxa2expression by the isthmus is important because the targeted inactivation of Hoxa2 results in homeotic transformations of second-arch neural crest derivatives into proximal first-arch derivatives, in a manner similar to that seen in the classic Noden transplantations (13, 14).Hoxa2 therefore is essential for regulating proper patterning of neural crest–derived skeletal structures in the second branchial arch (13, 14, 25).

The patterning abilities of the isthmus have been attributed in part to FGF8 (26, 27), which is also transiently expressed in the branchial arches during the early phase of neural crest migration (Fig. 1B). We confirmed that the isthmic territory we transplanted expresses FGF8 (Fig. 1C). Hence, we tested the long-term effects of FGF8 alone on Hoxa2 gene expression in cranial neural crest cells by placing FGF8-soaked beads (1 mg/ml) into the mesenchymal tissue adjacent to r4, in 8-9 somite stage chick embryos (Fig. 2A). After 24 hours of in ovo culture, Hoxa2 expression in r4-derived neural crest cells was repressed throughout the entire second branchial arch (Fig. 2, B and C). DiI lineage tracing confirmed that the absence ofHoxa2 expression was not due to a failure of neural crest cells to migrate into the second arch (Fig. 2D). In contrast to second-arch neural crest cells, FGF8-soaked beads did not affect the levels or segmental boundaries of Hoxa2 expression in the hindbrain relative to the control side (Fig. 2B). Hence, the repression of Hoxa2 expression in the neural crest occurs independently of events in the neural tube (Fig. 2, B and C). This is in agreement with ectopic expression studies indicating that Hoxa2activation in the migrating neural crest, as opposed to the neural tube, is required for generation of skeletal transformations (28, 29). Together these results show thatHoxa2 expression is essential for normal patterning of second-arch neural crest cells and is sensitive to the isthmus and FGF signaling.

Figure 2

Transposition of FGF8 beads adjacent to r4 transiently reprograms Hoxa2 expression in cranial neural crest cells. (A) Diagram showing the strategy of placing an FGF8-soaked bead (red circle) next to r4 to examine its effects on neural crest cells. (B and C) Dorsal view of an embryo (B) 24 hours after transposition with an FGF8 bead (*), showing inhibition of Hoxa2 expression in second branchial arch (ba2) neural crest cells on the graft as compared to the control side (white arrow). The bead does not alter the anterior limit (double black arrow), segmental domains, or relative levels of expression in the hindbrain. Lateral view of the same embryo (C) showing a complete inhibition of Hoxa2 expression in second branchial arch neural crest cells (black arrow). (D) Lineage tracing of r4-derived neural crest cells showing that the bead does not prevent neural crest cell migration into the second branchial arch (white arrowhead). (E and F) Lateral views of a control embryo (E) and one containing an FGF8 grafted bead (F) 36 hours after grafting, showing that Hoxa2 expression is inhibited in the vicinity of the bead [black arrow, (F)] but is reestablished in arch mesenchyme cells distal to the bead [white arrow, (F)].

This suggests that signals from the isthmus, presumably involving FGF8, are capable of inhibiting Hoxa2 expression, allowing second-arch crest cells to adopt a first-arch fate. If the inclusion of the isthmus accounts for the transformations used to support the prepatterning model, then identical grafts excluding this territory should not result in duplications of first-arch skeletal structures. Therefore, we transposed the anterior hindbrain, with or without the isthmus, in place of the r4 territory at 8-9 somite stage (Fig. 3, A and C). After in ovo embryo culture for 8 days, we assayed for the long-term phenotypic effects in skeletal morphology associated with suppression of Hoxa2 expression in second-arch neural crest cells by alcian blue staining (Fig. 3, B and D). In grafts containing the isthmus, there was a loss of normal r4-derived second-arch structures, such as the retroarticular process, and in their place the quadrate and Meckel's cartilage characteristic of the first arch were duplicated (Fig. 3, A and B). The embryos were assayed before bone formation, so we could not assess whether the articluar and squamosal bones were also duplicated. However, ectopic cartilage nodules were observed in these relative locations in addition to the duplicated quadrate and Meckel's cartilages. These homeotic transformations are similar to Noden's observations (2) and phenotypes observed in Hoxa2 mutant embryos (13–15). In contrast, grafts of the anterior hindbrain lacking the isthmus resulted in normal skeletal morphology (Fig. 3, C and D).

Figure 3

Skeletal analysis of grafted embryos containing hindbrain transpositions including or excluding the isthmus. (A) Diagram illustrating the posterior transposition of r1 and the isthmus to r4. (B) The posteriorly transposed isthmus/r1 results in duplication of first-arch skeletal structures such as the quadrate (q*) and Meckel's cartilage (m*) and loss of the retroarticular process (rp). (C) Diagram illustrating the posterior transposition of r1 excluding the isthmus to r4. (D) The first-arch skeletal morphology, including Meckel's cartilage (m) and the quadrate (q), is normal.

These findings demonstrate that the transformation of second-arch crest derivatives into first-arch structures, as described by Noden (2), is dependent on the presence of the isthmus. This suggests that the classical first-arch skeletal duplications arising through transpositions of first-arch and frontonasal neural crest (2) were a consequence of the suppression ofHoxa2 expression in the second arch by the isthmus. Therefore, we tested whether FGF8 alone could substitute for the isthmus. However, embryos in which FGF8-soaked beads were grafted into the mesenchyme adjacent to r4 and cultured for 8 days failed to generate duplicated first-arch skeletal elements. This implies that FGF8 alone in this context is unable to replace the isthmus, suggesting that additional factors may also be involved. This prompted us to examine the temporal effects of FGF8 beads on Hoxa2expression in the second branchial arch. We found that as the arch grows over time (36 to 48 hours), Hoxa2 expression is reestablished in the arch mesenchyme at a distance from the bead, but it continues to be inhibited in cells adjacent to the bead (Fig. 2, E and F). Hence, in contrast to isthmic grafts, FGF8 beads only transiently inhibit Hoxa2 expression in the second arch. This result reflects an important difference in the nature of isthmus/tissue versus bead transplantations. In grafts of the isthmus, the entire endogenous presumptive r4 crest was removed and replaced with the FGF8-expressing isthmus. However, in bead grafts, the endogenous r4 crest was left intact, and migrating crest cells derived from r4 are being challenged to reprogram by signals from the grafted bead. These experimental properties, in combination with the transient inhibition of Hoxa2 expression, could well account for the differences in skeletal phenotypes between isthmus and FGF8 bead grafts. Therefore, the variability in the duplications observed by Noden may also be explained by the variability of local FGF8 concentration present in grafted tissue. In contrast to the isthmic grafts, the duplicated first-arch structures observed in theHoxa2 mutants exhibit a mirror image polarity. This implies that other factors must be involved in patterning different axes of polarity in these duplications. The transposition of a signaling center might disrupt the mechanisms that influence polarity or axis patterning.

These experiments underscore the important role played byHoxa2 in branchial arch identity. Recently, functional inroads have been made into understanding the precise mechanisms by which Hoxa2 influences the morphogenesis of second-arch elements (30). Hoxa2 is widely expressed in the second-arch mesenchyme but is excluded from the chondrogenic domains and acts very early in the chondrogenic pathway upstream ofSox9, Col2a1, and Cbaf1I to repress their expression. During normal development, both endochondral and dermal (intramembranous) ossification occurs in first-arch morphogenesis; however, endochondral ossification primarily occurs in second-arch morphogenesis. Therefore, one of the roles ofHoxa2 in the second branchial arch may be the prevention of dermal bone formation. Overexpression studies of Hoxa2 in chick and zebrafish embryos have now confirmed its role as a true selector gene (28, 29). Therefore,Hoxa2 not only inhibits development of the lower jaw skeleton but is also primarily responsible for specifying second branchial arch fate.

The presence of the isthmus as a mechanistic basis for first-arch duplications also helps resolve two puzzling aspects of Noden's work (2). First, there was considerable variability in the frequency of duplications observed, presumably arising through variable inclusion of the isthmus and, consequently, the local concentration of FGF8. Second, it explains why both first-arch and frontonasal neural crest develop similar duplicated skeletal structures when transposed posteriorly, even though they are derived from different axial levels.

Therefore, rather than providing evidence for the prepatterning of neural crest cells, Noden's experiments (2) highlight the importance and effects of local signaling centers, such as the isthmus, in A-P patterning and regulation of Hox gene expression (26, 27). This study, together with recent evidence from mouse, chick, and zebrafish transplantation studies, argues as a general principle that cranial neural crest cells are not prespecified or irreversibly committed before their emigration from the neural tube (5, 11,16–18). Rather, neural crest patterning is based on plasticity and the ability of neural crest cells to respond to environmental influences in the branchial arches, and future attention will be focused on the nature of these signals.

  • * These authors contributed equally to this paper.

  • Present address: The Sanger Centre, Wellcome Trust Genome Campus, Hinxton, Cambs CB10 1SA, UK.

  • To whom correspondence should be addressed at The Stowers Institute for Medical Research, 1000 East 50th Street, Kansas City, MO 64110, USA. E-mail: rek{at}


View Abstract

Navigate This Article