The search for early molecular determinants of neural crest cell specification led to the identification of a large variety of transcription factors that appeared to be expressed in prospective crest cells at least until early migration (Fig. 1). This suggested that specification of dorsal neural epithelial cells into neural crest progenitors is necessary and sufficient to initiate a linear signaling cascade characterized by a precise sequence of expression of transcription factors and ultimately leading to their delamination and migration, provided cells are confronted with the appropriate environment. However, in the absence of studies at the single cell level, there is no direct proof for a strict correlation between expression of early neural crest markers and the cellular capacity to undergo delamination. Rather, several observations made essentially in chick and Xenopus suggest that neural crest specification, delamination, and migration are causally independent events.
Slug and Snail were the first transcription factors to be identified in the neural crest, about a decade ago.35 They belong to the Snail family of Zn-finger transcription factors and are most commonly used as neural crest markers.36 In the chick, neural crest cells express Slug but not Snail,37 whereas in the mouse and zebrafish,37,38 it is the reverse situation, crest cells express Snail instead of Slug, and, lasdy, in Xenopus, both factors coexist in crest cells but are induced separately.39 Slug and Snail have been shown to play similar roles and be interchangeable in some experimental systems,40 although they may also perform distinct functions in cells.39 There are numerous indications that Snail transcription factors are involved in EMT.36 Beside prospective neural crest cells, they are expressed in multiple embryonic regions known to undergo EMT such as in the primitive streak and mesoderm during gastrulation, during sclerotome dispersal, and during formation of the heart cushions. Snail mutant mice die at gastrulation, most likely due to defective ingression of the mesoderm in the primitive streak.41 In tumors, Snail triggers EMT through direct repression of E-cadherin, and its expression correlates with the invasive phenotype in cell lines as well as in vivo, in chemically-induced skin tumors.42,43 Likewise, overexpression of Slug in cultured epithelial cells causes desmosome dissociation followed by cell dispersion, upregulation of vimentin, and fibronectin redistribution.44 With regard to the neural crest, loss-of-function experiments in chick and Xenopus based on antisense oligonucleotides to Slug or Snail result in a strong deficit in migrating neural crest cells.35,39,45 For all these reasons, Slug has been presented as a major player necessary for neural crest delamination. However, other studies tend to contradict this view and suggest that Slug may be neither sufficient nor necessary for delamination at least at some axial levels. First, with the possible exception of the cranial levels in the mouse,42 Slug expression is often delayed with respect to the exclusion of E-cadherin from the neural tube particularly at truncal levels. It can be argued that Slug might also repress N-cadherin expression, but this has never been described so far. More significandy, in Xenopus, although Slug or Snail overexpression in whole embryos leads to the expansion of prospective neural crest territory and a greater number of melanocytes, their effect is limited to areas contiguous with endogenous neural-crest-forming regions.39,45 In addition, Slug is unable to induce by itself crest formation in ectodermal explants, suggesting that its sole expression is insufficient to direct a program of neural crest ontogeny.45 This is further supported by cell-tracing experiments which revealed that not all Slug/Snail-expressing cells are fated to become migrating neural crest cells.46 In chick, overexpression of Slug in the neural tube using in ovo electroporation increased the number of neural crest cells migrating out of the dorsal side of the neural tube associated with an increase in Rho-B expression, but this occured only at cranial levels.40 In addition, as observed in Xenopus, only cells situated in the most dorsal side of the embryo, i.e., in the neural crest prospective region, were able to emigrate, whereas cells situated immediately more ventrally exhibited no signs of delamination and expressed no neural crest markers.29. Slug overexpression in the trunk region caused only a slight expansion of the prospective crest cell region, but this was not accompanied by greater numbers of migrating cells. Finally, neural crest delamination can be severely affected in trunk of chick embryos, without detectable repression of Slug expression.47 Thus, paradoxically, despite convincing data on Slug/Snail function in numerous examples of EMT, their precise role in neural crest delamination remains elusive: they may promote delamination in a specific cellular context in the dorsal part of the neural tube but may be insufficient to drive by themselves EMT of neural tube cells situated more ventrally.
The Winged Helix-Forkhead Transcription Factor Foxd-3
Foxd-3 is a transcription factor of the winged helix-forkhead family whose temporal expression also closely matches neural crest induction, delamination, and migration. In Xenopus, chick, and mouse, its expression starts early during neural crest induction approximately coincident with or slighdy before that of Slug, but in contrast to the latter, it remains expressed in most neural crest cells throughout migration, except for melanocyte precursors.4 "51 As for Slug/Snail, targeted inactivation of Foxd-3 in the mouse is embryonic lethal at very early stages of development, before implantation52 and, thus, informations about its possible contribution to neural crest specification and delamination come essentially from gain- and loss-of-function analyses in frog and chick. In the chick, forced expression of Foxd-3 in the trunk neural tube was found to suppress interneuron differentiation and induce precocious and robust expression of HNK-1, a marker for migrating neural crest cells, in the whole neural tube by 24 hours post-transfection. Foxd-3 is also able to provoke ectopic delamination of cells but not until 24 hours post-transfection. Delamination is evident only after 36-48 hours and is accompanied by down-regulation of N-cadherin, up-regulation of integrins and cadherin-7, and disruption of the basement membrane lining the neural tube.28,29,49 In Xenopus embryos, when ectopically overexpressed, Foxd-3 is a potent inducer of neural crest markers, including itself and Slug, but it also promotes expression of neural markers.50 Interestingly, in contrast to Slug, Foxd-3 can induce expression of neural crest markers in distant locations from the neural crest region. Conversely, attenuation of Foxd-3 activity by overexpression of a dominant-negative form of the molecule inhibits neural crest differentiation.50 Thus, although these studies did not di-recdy address Foxd-3 function in delamination, they clearly highlight its critical role in neural crest formation and support observations made in the chick. In conclusion, unlike Slug, Foxd-3 seems to be a potent inducer of crest cell specification and delamination, but the delay for its effect indicates that it does so only indirectly, possibly by activating intermediate genes, and in an uncontroled manner as it induces markers of migrating neural crest cells prior to delamination and at the expense of other cell populations of the dorsal neural tube.
Recendy, the Sox-E subgroup of HMG box-containing transcription factors attracted much attention, because in all species and at all axial levels examined, they are specifically expressed in a temporal order in neural crest cells from early determination to late migration.28'29'53 Sox-9 is the first to appear in prospective crest cells and it is closely followed by Sox-10 and, later, by Sox-8, just before neural crest cells exit the neural tube. At the onset of migration, while Sox-10 is retained by migrating cells, Sox-8 and Sox-9 are rapidly down-regulated. Later during differentiation, Sox genes are reexpressed in distinct neural crest subpopulations and have been shown alternatively to contribute the maintenance of neural crest multipotency or to participate to specific differentiation programs. Thus, Sox-E genes, and particularly Sox-10, compose at present the most universal neural crest cell markers. As observed for Foxd-3, production of neural crest cells is strongly agcted in Xenopus embryos upon knockdown or overexpression of Sox-9 and Sox-10. However, while both factors are critical for early crest determination and melanocyte differentiation, they are apparendy relatively dispensable for delamination and subsequent migration.54"56 Likewise, in mouse embryos lacking Sox-9, neural crest cells are specified and able to start migration but they rapidly undergo programmed cell death shortly after.29 Very recendy, implication of Sox-E transcription factors in delamination has been further investigated in detail in chick by two different laboratories, but their results differ in some significant aspects.28'29'53 Both studies found that forced expression in the trunk neural tube of either Sox-8, Sox-9 or Sox-10, but not of Sox-2, a Sox gene from a different subgroup, convert within less than 24 hours neural tube cells into neural crest-like cells expressing the HNK-1 marker. In addition, both studies showed that Sox-E genes induces cadherin-7 expression but no downregulation of N-cadherin and that they rapidly turn off Rho-B expression. However, while Cheung and coworkers did not document any marked ectopic delamination of neural tube in the electroporated side of the neural tube and concluded that Sox-E genes by themselves are not capable of triggering cell delamination, McKeown and colleagues observed at variance extensive migration. This was seen at all levels along the dors-oventral axis, including in the floor plate, about 36 hours after electroporation, i.e., rather late after induction of a neural crest phenotype and, 48 hours after electroporation, the transfected side of the neural tube was almost entirely disrupted and almost all cells were released into the neighboring sclerotome.28 At present, there are no obvious explanations for these discrepancies, but whatever the exact role of Sox-E genes in neural crest development, it appears that, like Foxd-3, they can elicit cell delamination only secondarily, after induction of HNK-1 and cadherin-7, two markers normally expressed after delamination. In addition, delamination induced by Sox-E genes as well as by Foxd-3 is massive, leaving the neural tube as an empty bag or a flat tire from which the whole content has been poured out, a situation which is never observed normally, therefore suggesting that these transcription factors cause delamination by an aberrant and uncontroled sequence of events at the expense of the other neural cell types.
Recendy, beside Sox-E genes, another Sox transcription factor of the Sox-D subgroup, Sox-5, has been characterized at cranial levels in the chick. It is expressed in premigratory crest cells, slighdy later than Slug and is maintained in most neural crest cells during migration as well as in glial cells of cranial ganglia. Misexpression of Sox-5 in the cephalic neural tube leads to an exquisite phenotype contrasting with the massive effects obtained with Sox-E genes. In the dorsal neural tube, it augments both spatially and temporally the production of crest cells, associated with up-regulation of Foxd-3, Slug, Pax-7, Sox-10 and Rho-B, whereas in more ventral regions of the neural tube, it induces Rho-B expression, but not Foxd-3 or Sox-10 and its capacity to induce delamination is only marginal. Thus, like Slug, Sox-5 effect might be dependent on the cellular context within the neural tube.
Neural crest cells have been found to express several additional transcription factors at the time of delamination, among which Pax-3, AP-2, Myc and members of the Zic family are the most remarkable.58 The role of these factors in neural crest delamination has not been addressed directly and their possible implication in this process cannot then be formally excluded. However, it is clear that because they are not restricted to prospective neural crest cells, they cannot pretend to play a major role by themselves. Pax-3, for example, has been shown to be genetically upstream of Foxd-349 and mouse Splotch embryos in which its gene is mutated exhibit strong defects in neural crest cell generation and migration, possibly as a result from decreased cell-cell adhesion due to oversialylation of N-CAM molecules.5 Yet, the precise role of Pax-3 in the control of cell adhesion remains unclear, as it has been also observed in vitro that its forced expression in mesenchymal cells may induce their aggregation of into multi-layered condensed cell clusters with epithelial characteristics.61 The protooncogene Myc has been implicated in Xenopus in crest cell determination independently of its proliferation role62 and, in the chick, it has been shown to stimulate massive crest cell migration followed by their differentiation into neurons.63 Finally, mice deficient in the AP-2 gene show severe defects causing embryonic lethality and affecting primarily development of the neural crest: failure of neural tube closure, craniofacial anomalies and absence of cranial ganglia.64
It is striking that, among the different transcription factors characterized in neural crest cells at the time of their segregation from the neural tube, none of them exhibit expression patterns matching precisely with delamination, suggesting that this step is essentially dependent on transcriptional events occuring during the previous specification step. However, recent studies allowed to pin down factors that mark precisely crest cell delamination more reliably than Slug or Rho-B for example. Ets-1, a member of the Ets family of winged helix-turn-helix transcription factors has been found to be dynamically expressed in delami-nating crest cells at cranial levels (ref. 65 and E. Theveneau, M. Altabef, and J.-L. Duband, unpublished results). At the midbrain level, for example, its expression starts in prospective crest cells just after apposition of neural folds, at the 5-6 somite stage, i.e., about 4-6 hours before onset of migration, and it persists in the dorsal neural tube until cell delamination ceases, i.e., at the 11-somite stage. In addition, migrating neural crest cells almost immediately turn Ets-1 expression off as soon as they become fully segregated from the neural tube and leave its vicinity. Ets-1 has been previously implicated in various EMTs and migratory events during embryonic development and, in contrast to most other transcription factors expressed by crest cells, a detailed list of its potential target genes has been established: these include key molecules for cell locomotion such as integrins, cadherins and MMPs.66"69 Ectopic expression of Ets-1 in the chick neural tube by in ovo electroporation results in delamination of neural tube cells, at both cranial and truncal levels, although Ets-1 is not prominent in trunk crest cells. Interestingly, Ets-1 -induced cell delamination presents unique characteristics that are not observed with forced expression of Foxd-3 or Sox-E genes (E. Thdveneau, M. Altabef, and J.-L. Duband, unpublished results). It is rapid, cells being seen delaminating within 12 hours posttransfection; it occurs primarily at the basal side of the neural tube, but also less frequently at its apical (luminal) side; unlike Foxd-3, Sox-9 or Sox-10, it is not massive but rather progressive, leaving the neural tube intact in a very similar manner to the normal delamination of neural crest cells; cells exiting from the neural tube upon Ets-1 overexpression do not express neural crest cell markers such as HNK-1 or Slug, but show local disruption of the basement membrane, indicating that Ets-1 most likely triggers delamination by activating expression of MMPs; finally, delamination is not followed by migration, cells remaining for a while at the close vicinity of the neural tube, before undergoing apoptosis. Thus, at least at cranial levels, Ets-1 might regulate late cellular events accompanying neural crest cell delamination independently of a neural crest phenotype, thereby illustrating that specification, delamination, and migration are separable events.
• Detain tmhoti
Slug No NC marker/Wo (MamimlbnfNo migration/No apopltnb
JQJJ.JQ ^ NCflBffefS ^ rVbmrnofiun * mij;ra'W'i
Í Vj'rtrrjiFiafjorr + m re rrtiri>N
: 4 migration
DrlamiiwtMn + migration
Figure 2. Roles of the Rho-B GTPase and transcription factors of the Snail, Fox, Sox and Ets families in neural crest EMT as deduced from gain-of-function experiments in the trunk neural tube of the chick embryo. Rho-B and various transcription factors (indicated on the left) were electroporated either alone or in combination in the lateral and ventral sides of neural tube at stages prior to neural crest migration and the consequences of their forced expression on the ectopic expression of neural crest markers, cell delamination, survival and migration, on celllular events, such as disruption of the basement membrane, and on expression patterns of molecules involved in EMT are indicated at the indicated time periods following electroporation. Data were collected from references 28, 29, 49 and from E. Th^veneau, M. Altabef and J.-L. Duband, unpublished.
Cooperative Activity of Transcription Factors during Neural Crest Cell EMT
The above studies reveal that delamination elicited by transcription factors ectopically expressed in the intermediate and ventral neural tube is either partial or disordered, and that none of them is able to induce a complete neural crest phenotype (Fig. 2). Thus, they do not allow to draw a coherent sketch of the transcriptional network controling neural crest delamination. A possible clue is to identify the epistatic relationships between these transcription factors in prospective neural crest cells.58 In Xenopus, gain- and loss-of-functions approaches revealed highly complex crossregulation of Snail, Slug, Sox-9, Sox-10 and Foxd-3 genes which can influence each other via direct transcriptional activation of repression or through secondary factors, thereby excluding any obvious linear hierarchy among these factors.39'5 5 In chick, the situation contrasts with that observed in Xenopus in that Slug, Foxd-3, and Sox-9 signals are apparently independent and display distinct sets of targets.2*"9'40'49 In addition, Foxd-3 and Sox-9 lie upstream the Sox-10 and Sox-8 genes, consistent with their precocious expression in prospective crest cells. In the mouse, lastly, while the Foxd-3 and Snail mutants were not informative because of premature lethality, the Sox9 mutant provided interesting informations about the functional interactions between Foxd-3, Snail, Sox-10 and Sox-9. It appears that Snail, but not Foxd-3 and Sox-10, is dramatically downregulated in premigratory crest cells in mutant embryos compared with their wildtype littermates, indicating that in this species, Snail is downstream of Sox-9.29 Recently, using the chick system, the Briscoe's laboratory further investigated possible cooperative activities between Slug, Foxd-3 and Sox-9 in the control of truncal crest delamination by comparing the effects of these factors individually or in various combinations.29 While forced expression of Slug showed no obvious effect 24 hours after electroporation, Sox-9 induced HNK-1 expression within 12 hours in the intermediate and ventral neural tube but, after 24 hours, it was not capable of triggering significant cell delamination with no disruption of the basement membrane and of N-cadherin junctions and no integrin upregulation. In contrast, Sox-9 and Slug in conjunction induced robust HNK-1 expression, disorganization of the neural epithelium with degradation of the basement membrane, derealization of N-cadherin out of adherens junctions, but no increase in integrins. Thus, confirming previous observations, Slug is effective in inducing cell delamination only if cells are specified as neural crest cells. Foxd-3 by itself was sufficient to induce first Sox-10 after 12 hours followed by HNK-1 expression after 24 hours associated with a decrease in N-cadherin expression. Ultimately, after 36-48 hours, it provoked an increase in integrins, a breakdown of the basement membrane allowing delamination and migration. Finally, combination of all three factors in neural tube cells caused cells to express neural crest markers, to delaminate entirely from the neural tube and to move actively in the surrounding tissues: this was associated with the complete breakdown of the basement membrane, the disappearance of N-cadherin from the cell surface, and the up-regulation of integrins. Thus, Sox-9, Foxd-3 and Slug ectopically expressed in the neural tube can recapitulate most of the events observed during neural crest delamination except that, unlike for endogenous neural crest cells, delamination is massive and uncontroled and leaves the neural tube totally disorganized (Fig. 2).
From these studies, some of the basic traits of the interplay between transcription factors during the transition from neural crest determination to early migration are now taking shape. First, delamination as well as specification and migration require the cooperating activities of at least three members of distinct families of transcription factors, Snail/Slug, Foxd-3 and Sox-9. Second, there is no simple linear hierarchies among these factors, rather a complex network of mutual interactions. Third, deployment of delamination is complete and efficient only if its basic cellular events are properly ordered in connection with crest cell specification and migration. For example, although neural crest specification is not sufficient to induce complete delamination (as suggested by experiments of Sox-9 overexpression) and that, conversely, delamination can be induced independendy of specification (as suggested by overexpression of Rho-B or Ets-1), delamination is followed by active migration only if cells are specified into neural crest cells.
Was this article helpful?