53
1 Submitted version Resolving time and space constraints during neural crest formation and delamination Jean-Loup Duband 1, 2 , Alwyn Dady 1, 2, * , and Vincent Fleury 3 1. Université Pierre et Marie Curie-Paris 6, Laboratoire de Biologie du Développement, 75005 Paris, France 2. CNRS, Laboratoire de Biologie du Développement, 75005 Paris, France 3 Laboratoire Matière et Systèmes Complexes, CNRS et Université Denis-Diderot-Paris 7, 75013 Paris France * Present address: Children's Hospital of Pittsburgh, Rangos Research Building, Pittsburgh, PA 15201, USA Author for correspondence: Jean-Loup Duband, Laboratoire de Biologie du Développement, Université Pierre et Marie Curie, 9 quai Saint-Bernard, 75005 Paris, France email : [email protected]; tel : 00 33 1 44273645; fax : 00 33 1 44273497 Running title: Time and space during neural crest development Key words: neural crest, induction, delamination, epithelium-to-mesenchyme transition, migration, sorting, cadherins, core EMT regulatory factors

Resolving time and space constraints during neural crest ...vfleury... · cadherins drive the entire process of neural crest segregation from the rest of the neurectoderm by their

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

  • 1

    Submitted version

    Resolving time and space constraints during neural crest formation and

    delamination

    Jean-Loup Duband1, 2, Alwyn Dady1, 2, *, and Vincent Fleury3

    1. Université Pierre et Marie Curie-Paris 6, Laboratoire de Biologie du Développement, 75005

    Paris, France

    2. CNRS, Laboratoire de Biologie du Développement, 75005 Paris, France

    3 Laboratoire Matière et Systèmes Complexes, CNRS et Université Denis-Diderot-Paris 7,

    75013 Paris France

    * Present address: Children's Hospital of Pittsburgh, Rangos Research Building, Pittsburgh, PA

    15201, USA

    Author for correspondence: Jean-Loup Duband, Laboratoire de Biologie du Développement,

    Université Pierre et Marie Curie, 9 quai Saint-Bernard, 75005 Paris, France

    email : [email protected]; tel : 00 33 1 44273645; fax : 00 33 1 44273497

    Running title: Time and space during neural crest development

    Key words: neural crest, induction, delamination, epithelium-to-mesenchyme transition, migration, sorting, cadherins, core EMT regulatory factors

  • 2

    Abstract

    A striking feature of neural crest development in vertebrates is that all the specification,

    delamination, migration, and differentiation steps occur consecutively in distinct areas of the

    embryo and at different timings of development. The significance and consequences of this

    partition into clearly-separated events is not fully understood yet, but it ought to be related to the

    necessity of controling precisely and independently each step, given the wide array of cell types

    and tissues issued from the neural crest throughout the embryo and the long duration of their

    development spanning almost the entire embryonic life. In this review, using the examples of

    early neural crest induction and delamination, we discuss how time and space constraints

    influence their development and describe the molecular and cellular responses that are employed

    by cells to adapt. In the first example, we analyze how nascent neural crest cells that are initially

    mingled with other neurectodermal progenitors after induction in the blastoderm are

    progressively segregated from the other neural tube and ectoderm populations and positioned at

    the apex of the neural tube prior to migration, owing to the interplay between cadherin-mediated

    selective adhesion and global cell movements. In the second example, we examine how

    cadherins drive the entire process of neural crest segregation from the rest of the neurectoderm

    by their dual role in mediating first cell sorting and cohesion during neural crest specification

    and later in promoting their delamination. In the third example, we describe how the expression

    and activity of the transcription factors known to drive epithelium-to-mesenchyme (EMT)

    transition are regulated timely and spatially by the cellular machinery so that they can

    alternatively and successively get involved first in neural crest specification, then in

    delamination. In the last example, we briefly tackle the problem of how EMT-inducing factors

    may elicit different cell responses in neural crest and neural tube progenitors.

  • 3

    The term neural crest stems from the fact that, in many species and particularly in some urodele

    amphibians, its cellular components protrude over the dorsal midline of the embryo after

    delaminating from the closing neural tube (Le Douarin and Kalcheim, 1999; Trainor, 2013); and

    as a matter of fact, it is precisely at that time that this cell population individualizes as a distinct

    entity, fully segregated from the rest of the neurectoderm from which it derives. This

    morphological feature has long served to define the neural crest as a population of migrating

    cells that arise from the dorsal neural tube and populate various sites in the embryo, where they

    undergo terminal differentiation. However, a variety of cell-tracing studies in chick and Xenopus

    clearly established that neural crest cells are not born in the neural epithelium at the time when

    they initiate migration but long before, by a mechanism that is separate from that of induction of

    the neural plate, the primordium of the central nervous system (Aybar and Mayor, 2002; Basch

    et al., 2006; Betancur et al., 2010; Ezin et al., 2009; Milet and Monsoro-Burq, 2012; Prasad et

    al., 2012). Thus, despite its intimate physical association and intricate functional connections

    with the central nervous system, the neural crest must no longer be regarded as a byproduct of

    the neural tissue but as a genuine, distinct embryonic structure which by itself contributed to the

    emergence of the vertebrates during evolution and their adaptation to all the ecological niches on

    Earth.

    In birds, fishes, and amphibians and probably also in mammals, neural crest progenitors

    are specified during gastrulation, in the blastoderm, before its subdivision into neural and non-

    neural epithelium. As neurulation proceeds, these progenitors are progressively positioned, first

    in the elevating neural folds, then in the dorsal aspect of the neural tube. Later on, by a

    mechanism related to epithelium-to-mesenchyme transition (EMT), they delaminate from the

    neural tube, and owing to a complex code involving multiple guidance cues, they disperse

    through restricted pathways. Finally, they colonize many different territories thoughout the body,

    where they eventually give rise to a large collection of cell types. Thus, development of the

  • 4

    neural crest divides into four well-defined steps: the premigration phase, consisting essentially in

    the early induction of progenitors and their maturation as a population of cells endowed with the

    competence to undergo migration and to give rise to multiple lineages; the delamination phase

    which allows their physical separation from the neural tube; the migration phase; and, finally the

    ultimate phase of differentiation in the sites of election. Intriguingly, these steps all occur in

    distinct areas of the embryo and at different timings during development, the course of neural

    crest ontogeny spanning almost the entire duration of embryonic life: specification in the

    blastoderm during gastrulation and early neurulation, delamination from the dorsal neural tube

    along the embryonic midline at the end of neurulation, migration and differentiation throughout

    the course of organogenesis and late embryogenesis in as diverse locations as the vertebral

    column, the skin, the aorta, the periphery of the brain, the face, the heart, or the gut. Although

    not unique, it is not quite common that during embryogenesis, cells execute sequentially their

    main developmental steps in separate locations. Therefore, time and space are both important

    parameters that may underlie the entire process of neural crest development and may impinge on

    the regulatory processes involved; and there are now many situations in which mismatches

    between them may cause severe developmental anomalies. In the present chapter, using selected

    examples taken from specification and delamination, mainly in the chick and frog but also in the

    zebrafish and mouse, we will address the question of how space is integrated during neural crest

    development and how this is regulated over time.

    Integrating space during neural crest induction and specification: roles of cell movements

    and cadherin-mediated cell sorting

    The generation of neural crest progenitors in the blastoderm results from the combined action of

    various signals mediated by diffusible morphogens of the Wnt, bone morphogenetic protein

    (BMP), and fibroblast growth factor (FGF) families as well as by retinoic acid, emanating from

  • 5

    the direct environment and acting sequentially or in synergy on a restricted cell population of the

    neurectoderm. In response to these signals, cells express a gradually more complex and specific

    repertoire of transcription factors assembled into a gene regulatory network, that assigns them

    their neural crest identity and allows them to segregate from the ectoderm and neural plate

    (Betancur et al., 2010; Groves and LaBonne, 2014; Huang and Saint-Jeannet, 2004; Milet and

    Monsoro-Burq, 2012; Prasad et al., 2012; Sauka-Spengler and Bronner-Fraser, 2008).

    The first manifestation of neural crest induction during gastrulation is the collective

    expression of the transcription factors Pax-7 in chick, Pax-3 in Xenopus, Msx-1/2, and Zic-1.

    Expression commences in a territory at the border between the prospective neural plate and

    ectoderm, hence its name neural plate border, and which encompasses progenitors of the future

    neural crest, as well as those of the dorsal neural tube and ectodermal placodes. Although

    individually these factors are neither restricted to the neural plate border nor sufficient to drive

    neural crest specification, their combined effect is sufficient to restrict progressively the potential

    of cells to adopt ectodermal or neural plate fate, while providing them with the competence to

    acquire a neural crest identity. For this reason, these genes have been regrouped under the term

    of border specifiers. The second step, named specification, is aimed at consolidating this identity,

    in particular to acquire unique adhesive, proliferative, migratory, multipotency and survival traits

    that enable neural crest progenitors to undergo EMT and disperse throughout their migratory

    routes. This is achieved by the cumulative expression of the neural crest specifier genes,

    including members of the Snail and SoxE families of transcription factors, as well as a variety of

    other factors such as Foxd-3, Ets-1, AP-2 and Id genes.

    Strikingly, the expression patterns of the various border specifiers that have been resolved

    in different species reveal that at onset of induction, the neural plate border is not a well-defined

    and delineated territory (Basch et al., 2006; Ezin et al., 2009; Khudyakov and Bronner-Fraser,

    2009; Milet and Monsoro-Burq, 2012). By combining the techniques of lipophilic dye tracing

  • 6

    and time-lapse imaging, Ezin et al. (2009) performed fate-mapping studies in chick and were

    able to trace precisely the position of neural crest progenitors over time, from early induction to

    their final position in the dorsal neural tube. They found that at gastrulation, neural crest

    progenitors receiving the induction signal are scattered in the blastoderm in a broad domain

    corresponding to the BMP-4-producing region and are mingled with other neurectodermal

    progenitors. During neurulation, the neural plate border undergoes spectacular morphological

    changes from a short, wide and flat strip of cells with a horse-shoe shape in the anterior

    blastoderm into two long, narrow and protruding neural folds that appose and fuse together along

    the embryonic midline. During this process, ectodermal, placodal, neural crest and neural plate

    progenitors initially dispersed and mixed in the blastoderm become spatially segregated and

    patterned into highly-ordered, compact entities along the rostrocaudal and dorsoventral axes.

    How is this process achieved? Ezin et al. (2009) revealed that the dynamic displacement of the

    neural plate border and sorting of its different cell components occurs as a result of tightly-

    coordinated movements of convergence, extension, reorientations as well as cell mixing,

    coincidently with the movements of folding, rolling and bending of the neural plate into a hollow

    neural tube. These observations therefore uncover the previously-unsuspected importance of cell

    movements and rearrangements in the progressive definition of the neural crest territory (and by

    the way, that of the other cell populations derived from the neurectoderm) during neurulation and

    its confinement to the dorsal neural tube prior to delamination.

    Movements of convergence extension cannot, however, explain by themselves the

    regroupment of neural crest progenitors and their segregation from the other cells with which

    they are mingled. Other cellular events, such as cell cohesion and cell sorting, have then to be

    invoked to account for the whole process. Interestingly, recent studies in the zebrafish on the

    mechanism of specification of the ventral neural tube progenitors by the inductive signaling

    activity of Sonic hedgehog (Shh) (Xiong et al., 2013) provide clues about the sequence of events

  • 7

    that might be responsible for neural crest regroupment after induction. Indeed, in this species,

    instead of being static in a stable epithelium as often assumed in classical models, both the Shh-

    producing cells in the axial mesoderm and the Shh-responding cells in the neural tube are motile

    and capable of exchanging neighbors. As a result, morphogen signaling across the tissue is

    spatially noisy, and neural progenitors being exposed to variable doses of Shh over time and

    space exhibit a large range of responses and distribute first in a salt-and-pepper specification

    pattern. It is only later that the rearrangement of the different neural progenitors of the neural

    tube into sharply-bordered domains is achieved by a cell-sorting mechanism operating in a

    gradually-more static population. These observations indicate that specification and positioning

    are separate in time and that cell sorting acts to refine spatial patterning by inductive signals. The

    fact that the blastoderm is subjected to intense cell movements during induction of the neural

    plate border and that emission of the BMP-4 inductive signal occurs in a broad band covering a

    large portion of the blastoderm (Ezin et al., 2009) suggests that the same paradigm may apply

    during formation of the neural crest.

    The question now is by which mechanisms neural crest cells would be sorted from the

    other populations of the neurectoderm ? Cell sorting results primarily from the selective

    recognition and adhesion of cell populations expressing different repertoires or doses of

    molecules of the cadherin family (Halbleib and Nelson, 2006). In the case of neural crest cells in

    the chick, until recently the prevailing assumption was that, due to their integration into the

    dorsal neural tube, they express primarily N-cadherin, whereas in the overlying ectoderm cells

    instead express E-cadherin. In addition, it was generally believed that such a situation is

    established precociously during early neural induction and persists throughout neurulation until

    delamination (Kalcheim, 2000; Powell et al., 2013; Strobl-Mazzulla and Bronner, 2012a;

    Taneyhill and Schiffmacher, 2013; Theveneau and Mayor, 2012). Therefore, cadherins were

    essentially regarded as cell-cell adhesion molecules maintaining cell cohesion among neural

  • 8

    crest cells until EMT. However, recent observations suggest that their role may extend beyond

    the sole phase of delamination.

    In the chick, our recent data show that there is no mutually-exclusive partition of E- and

    N-cadherins in the prospective ectoderm and neural epithelium until they are entirely segregated

    at completion of neurulation (Dady et al., 2012). Indeed, in a large anterior half of the embryo,

    E-cadherin is at gastrulation uniformly expressed throughout the superficial layer of the

    blastoderm while N-cadherin is restricted to the primitive streak, the mesoderm and the

    notochord. It is only very progressively that E-cadherin is replaced by N-cadherin in the neural

    plate, with a relatively-long time period where both molecules are coexpressed in cells. Thus,

    rather than defining precociously the embryonic territories fated to become neural, N-cadherin

    upregulation occurs secondarily as a result of neural induction and early neurulation. Moreover,

    a detailed survey of the spatiotemporal expression patterns of E- and N-cadherins throughout

    neurulation revealed a much more intricate pattern than previously thought (Dady et al., 2012).

    At the most anterior levels, from the forebrain to the hindbrain, the E- to N-cadherin switch is

    complete at late stages of neurulation even after neural tube closure, while in the trunk, it occurs

    earlier during fold elevation. This indicates that this switch is not correlated with the movements

    of neurulation, and fully support previous genetic studies in the mouse and zebrafish showing

    that in the absence of N-cadherin, onset of formation of the neural tube occurs normally and that

    the first signs of malformations are detected relatively late (Hong and Brewster, 2006; Radice et

    al., 1997). Which morphogenetic events are then driven by the E- to N-cadherin switch? Studies

    in the zebrafish showed that the lack of N-cadherin prevents cells to establish stable protrusive

    activity and causes alterations in their radial intercalation in the neural epithelium (Hong and

    Brewster, 2006). This suggests that N-cadherin plays a role in cell sorting and in tissue

    stabilization, implying that changes in the expression patterns of E- and N-cadherins would

    correlate primary with the segregation of neural and ectodermal progenitors.

  • 9

    Still, neither E-cadherin nor N-cadherin can account for the subdivision of the blastoderm

    into three distinct territories (and even four when also considering the placodes), and the

    implication of a third partner has to be invoked. Interestingly, the expression pattern of cadherin-

    6B in the chick is suggestive of a role for this cadherin in the segregation of the neural crest

    contingent from the rest of the neurectoderm and its regroupment at the interface between the

    ectoderm and neural tube, as originally proposed by Nakagawa and Takeichi (Nakagawa and

    Takeichi, 1995, 1998). Far from being induced late in the dorsal neural tube after neural crest

    specification, cadherin-6B appears very early in the neural plate border slightly after induction of

    the border genes Pax-7 and Msx-1 and before neural crest specifiers. In addition, consistent with

    a role in neural crest cell sorting, it distributes at first as a punctate pattern and becomes

    gradually more intense and compact as neural crest progenitors accumulate at the apex of the

    neural tube (Dady and Duband, in preparation). Thus, structuration of the blastoderm into

    sharply-delineated, contiguous cellular compartments is likely to result from the dynamic

    interplay between E-cadherin, N-cadherin and cadherin-6B over time during neurulation. In this

    regard, it is worth-mentionning that cadherin-6B has been shown to be upregulated by BMP-4

    signaling (Sela-Donenfeld and Kalcheim, 1999), while N-cadherin is in contrast under the

    control of FGF and the Sox-2 transcription factor, an early definitive marker for neural tissues

    (Linker and Stern, 2004; Matsumata et al., 2005; Uchikawa et al., 2011).

    In order to be efficient in a large scale and not limited to a few neighboring cells,

    cadherin-mediated cell sorting must be coordinated with dynamic cell movements so that cell

    populations can first travel and reorganize over long distances to exchange neighbors and then

    become more static and cohesive to stabilize their mutual associations. How are these

    movements driven and regulated during neural crest induction and specification? In chick, a

    possibility is that they are synchronized with the movements of intercalation occurring nearby

    the primitive streak (Voiculescu et al., 2007; Ybot-Gonzalez et al., 2007). However, the limited

  • 10

    range of these movements along the primitive streak cannot account for the large amplitude of

    the so-called polonaise movements thoughout the blastoderm. Biophysical studies based on high

    resolution microscopy and fine particle tracking in chick allowed to decipher cell movements at

    a large scale over the entire blastoderm, from gastrulation to the end of neurulation, and to define

    the physical parameters, such as viscoelasticity and mechanical forces, that drive the shaping of

    the embryo (Fleury, 2012). Prior to gastrulation, the blastula exhibits roughly three rings of cells

    of different sizes, cells being smaller internally and larger at the periphery. At that time, two

    types of movements can be observed, a radial one involving cells at the periphery responsible for

    the expansion of the blastoderm over the yolk, and a rotatory one in the center. This creates large

    vortices forming a quadrupolar flow. As the embryo initiates gastrulation, propagation of the

    mesoderm exerts a traction on the epiblast, stretching and folding the blastoderm

    anteroposteriorly. Importantly, folds form at boundaries between cell types and match to the

    expression patterns of cadherins in the neurectoderm, segregating the tissue into domains of

    different cell sizes and cell-adhesion properties and separated by deep furrows (Fleury,

    unpublished). While E-cadherin is progressively replaced by N-cadherin in small cells in the

    medial part of the blastoderm, cadherin-6B becomes specifically expressed in the folds. Thus,

    folding in the blastoderm produced by differential cell movements may drive the segregation of

    the diverse cell populations issued from the neurectoderm.

    All together the different analyses described above suggest a model in which, rather than

    being de facto a well-defined area of progenitors all sharing the same differentiation potential

    that is passively moved toward the embryonic midline by the movements of neurulation, the

    neural plate border constitutes an unsteady population that is subjected to intense cellular

    reorganizations and is progressively partitioned by the interplay between physical events

    operating in the entire blastoderm and biochemical factors acting locally (Figure 1). It would be

    interesting to analyze how and to which extent this model developed for a flat embryo like in

  • 11

    chick must be modified to adapt to spheric embryos like in frog and fish.

    Coordinating neural crest cell delamination timely and spatially: regulation of cadherin

    activity

    By analogy to the now-classical models of EMT developed for cultured cell lines and cancer

    cells (Peinado et al., 2007; Thiery et al., 2009; Thiery and Sleeman, 2006), a model has been

    proposed for chick neural crest cells in which EMT would occur through repression of cadherins

    (namely N-cadherin and cadherin-6B) by a mechanism involving the so-called core EMT

    regulatory factors, Snail-1/2, Zeb-2/SIP-1 and Twist-1, and elicited by Wnt-1 and Bmp-4 signals

    (Duband, 2010; Kalcheim, 2000; Powell et al., 2013; Strobl-Mazzulla and Bronner, 2012a;

    Taneyhill and Schiffmacher, 2013; Theveneau and Mayor, 2012). However, although widely

    admitted, this model must be reevaluated at the light of the many progresses accomplished

    during the last decade on the comprehension of the intimate molecular processes regulating both

    spatially and temporally the sequence of events leading to EMT.

    An important point concerns how cadherins are regulated during delamination: what is

    the exact timing of their expression and what are their functions in this event? Moreover, a new

    question has emerged gradually during the last years: are cadherin-6B and N-cadherin the only

    cadherins involved in neural crest EMT in chick? Regarding cadherin-6B, the sharp diminution

    of their messages at onset of EMT is indicative of a role in cell delamination (Nakagawa and

    Takeichi, 1995; Taneyhill et al., 2007). In addition, Snail-2 has been shown to bind specifically

    to E boxes in the regulatory sequence of the cadherin-6B gene and to control directly its

    expression (Taneyhill et al., 2007). Finally, functional studies revealed that at cranial levels,

    knockdown of cadherin-6B leads to premature neural crest cell emigration, whereas its

    overexpression blocks their migration (Coles et al., 2007). Thus, all these data concur with the

    idea of transcriptional repression of cadherin-6B being a major triggering event at onset of EMT.

    Interestingly, recent studies suggest that the precise timing when EMT occurs is rather defined

  • 12

    by posttranscriptional regulation of cadherin-6B activity (Fairchild and Gammill, 2013). Indeed,

    tetraspanin-18, a member of the tetraspanin family specific of epithelial cells, is abundantly

    expressed in premigratory but not migratory cranial neural crest cells. Tetraspanin-18 functions

    as a stabilizer of cadherin-6B proteins at the cell surface, and its downregulation under the

    control of Foxd-3 is required for neural crest cells to initiate migration. These data illustrate the

    complexity of the mechanisms controling the timing of neural crest cell delamination: while

    cadherin-6B expression is under the control of Snail-2, tetraspanin-18 acting as a cadherin-6B

    stabilizer is itself under the control of Foxd-3.

    Several observations, however, argue in favor of alternative or additional roles for

    cadherin-6B during delamination. First, its messages persist in the dorsal neural tube after

    completion of neural crest cell emigration, indicating that its function may not be restricted to the

    temporal control of intercellular adhesion during neural crest EMT (Nakagawa and Takeichi,

    1995). In species like mouse and zebrafish, cadherin-6, a close relative to cadherin-6B, is not

    downregulated from the surface of neural crest cells after delamination, and at trunk levels in the

    chick, cadherin-6B proteins remain present during the early steps of migration (Clay and

    Halloran, 2014; Inoue et al., 1997; Park and Gumbiner, 2010). Additionally, functional studies

    suggest that, rather than maintaining epithelial stability among neural crest cells, cadherin-6B in

    chick and cadherin-6 in fish promote their emigration out of the neural tube (Clay and Halloran,

    2014; Park and Gumbiner, 2010). Indeed, trunk neural crest cells expressing cadherin-6B have

    been shown to exhibit a general loss of epithelial junctional polarity and gain motile properties

    prior to delamination, and cadherin-6B was found to be required for deepithelialization of neural

    crest cells. Interestingly, this effect is mediated in chick by non-canonical BMP signaling

    involving downstream the BMP type II receptor, the LIM kinase and its major target cofilin, and

    in fish by controling Rho GTPase distribution in the cytoplasm. Both signaling pathways

    ultimately regulate locally the organization of the actin cytoskeleton and promote polarized

  • 13

    actomyosin contraction necessary for disruption of apical cell junctions (Clay and Halloran,

    2014; Park and Gumbiner, 2010, 2012).

    The discrepancy between the data in favor of a role of cadherin-6B in cell-cell adhesion

    and those supporting intracellular signaling activity regulating actin dynamics has been

    interpreted in terms of differences between cranial and truncal neural crest cells (see e.g.

    Taneyhill and Schiffmacher, 2013); and, it is now clearly demonstrated that both populations

    differ radically in their modes and kinetics of delamination (Duband, 2010). However, there are

    objectively no solid reason for regional differences in cadherin-6B function and regulation, as

    many of its partners (catenins, cytoskeletal elements) and regulators (Snail-2, Foxd-3) are found

    at both cranial and truncal levels. In addition, a question remains in the signaling model as to

    how cell cohesion is maintained among neural crest cells until delamination if the role of

    cadherin-6B is solely restricted to disruption of cell-cell contacts. This is not consistent with its

    precocious expression during neural crest induction. A possible explanation may reside in the

    fact that these experiments differ in an important detail, i.e. the timing when cadherin-6B

    expression was affected, and this might be sufficient to account for the differences observed.

    Indeed, Coles et al. (2007) modified cadherin-6B expression at cranial levels less than 6 hours

    before delamination, while Park and Gumbiner (2010) affected it at trunk levels more than 12

    hours before. Because the adhesive and signaling functions described for cadherin-6B are not

    mutually exclusive, another scenario can be proposed that would reconcile both models (Figure

    2). Both in the head and trunk, cadherin-6B would promote cell sorting during neural crest cell

    specification and maintain their cohesion until delamination while activating signaling pathways

    that would act as a negative feedback loop and set the grounds for future delamination by driving

    polarized changes in the actin cytoskeleton dynamics. Whether both activities may occur

    concurrently or sequentially prior to delamination is a critical point that remains however to be

    clarified. At cranial levels, additional events such as the stabilizing activity of tetraspanin-18

  • 14

    may complement this process to coordinate neural crest cell EMT into a massive emigration.

    Thus, according to this scenario, premature inhibition of cadherin-6B signaling would block the

    cascade of events leading to delamination, while late inhibition of its adhesive function at a time

    when polarized contraction of actomyosin has already been activated would in contrast

    precipitate emigration. Conversely, early overexpression of cadherin-6B in neural crest cells

    would enhance emission of the signals activating cytoskeleton dynamics and therefore promote

    delamination, whereas its late overexpression can oppose the diminution of endogenous

    cadherins and block delamination. Regardless the precise sequence of events controling its

    expression and function, cadherin-6B must therefore be considered as a unique element of the

    developmental program responsible for the segregation of the neural crest population from the

    rest of the neurectoderm by mediating first their sorting during specification and later by

    inducing EMT.

    As far as N-cadherin is concerned, it is generally considered as the major cadherin to be

    repressed at the time of neural crest cell delamination. In support to this assumption, it has been

    demonstrated that at truncal levels in chick, delamination involves cleavage of N-cadherin by the

    metalloprotease ADAM-10 under the control of BMP-4. In addition, the released cytoplasmic

    fragment of the N-cadherin molecule translocates to the nucleus and stimulates cyclin-D1

    transcription and cell delamination (Shoval et al., 2007). This model presents a certain interest in

    that, as for cadherin-6B, it is consistent with both the adhesive and signaling activities of

    cadherins. However, it has not been validated yet for anterior levels, and given the growing

    evidence for the heterogeneity in the modes of neural crest delamination along the embryonic

    axis, it may be true only for the lower trunk of the embryo. Moreover, it should be stressed that

    the previous analyses on N-cadherin distribution on which this model relies either did not

    employ markers for premigratory neural crest cells or were not performed systematically at all

    axial levels and at all stages throughout neural crest development (Akitaya and Bronner-Fraser,

  • 15

    1992; Duband et al., 1988; Hatta et al., 1987; Park and Gumbiner, 2010; Shoval et al., 2007).

    Thus, several recent studies argue that N-cadherin may not be implicated at all or at least very

    secondarily in neural crest cell delamination at least in the rostral half of the embryo. A detailed

    description of the kinetics of neural crest cell delamination at the midbrain level in chick and

    mouse (Lee et al., 2013) revealed that a large proportion of the delaminating cells does not

    express N-cadherin but rather E-cadherin which is gradually repressed as they separate from the

    neurectoderm. Interestingly, these cells are located proximal to the non-neural ectoderm in the

    neural folds and correspond to the mesectodermal contingent of the cranial neural crest (i.e. cells

    at the origin of bones, cartilage and other non-neural derivatives). Beside these cells, a smaller

    contingent situated proximal to the neural part of the neurectoderm expresses both N-cadherin

    and E-cadherin prior to migration and provides the neuronal derivatives of the cranial neural

    crest. These observations suggest that rather than controling the timing of initiation of cell

    delamination, N-cadherin expression would constitute an early neuronal signature in the

    premigratory neural crest population. In a systematic analysis of the repertoire of cadherin

    expression throughout neural crest formation until their complete segregation from the neural

    tube, we found that at least in the head and upper trunk, N-cadherin expression is induced

    progressively in the neural part of the neurectoderm during neurulation but not in neural crest

    progenitors, including those which adopt a neuronal fate (Dady et al., 2012; Dady and Duband,

    in preparation). Significantly, E-cadherin was not confined to the ectoderm but its expression

    was found in premigratory cells until delamination or even later, e.g. at the midbrain level. Thus,

    N-cadherin repression may not constitute a prominent event during neural crest delamination in

    the anterior half of the body. As discussed above, this role is most likely devoted to cadherin-6B

    whose repression occurs sharply. The function of E-cadherin in this process is apparently less

    critical as its repression proceeds at a slow pace and that, depending on the axial level

    considered, it may be expressed on the surface of migrating cells (though at moderate levels) or

  • 16

    downregulated prior to delamination (Dady et al., 2012; Lee et al., 2013; Dady and Duband, in

    preparation). This view is consistent with the expression patterns and functions during

    delamination of Snail-2 and Zeb-2, known as bona fide transcriptional repressors of E-cadherin

    and cadherin-6B (Peinado et al., 2007; Taneyhill et al., 2007) but not of N-cadherin (Dady et al.,

    2012; Dady and Duband, unpublished). Conversely, Sox-2, an activator of N-cadherin

    transcription (Uchikawa et al., 2003), is excluded from the neural plate border where it was

    found to repress neural crest cell identity (Wakamatsu et al., 2004).

    In conclusion, in a new model that differs quite significantly from the previous ones, we

    propose that, in chick, cadherins play dual roles during neural crest specification and

    delamination and constitute a key element in a coherent program aimed at progressively

    segregating these cells from the other neurectodermal populations (ectoderm, neural tube and

    placodes), first locally within the neurectodermal epithelium by promoting cell sorting and then

    out of it by inducing EMT and migration (Figure 3). An important consequence of this program

    is that it restricts spatially and temporally EMT and migration only to neural crest cells without

    interfering with the stability and development of the neighboring neural and ectodermal

    epithelia. This might explain why EMT can be induced in those tissues only if a neural crest

    identity is forced (Barembaum et al., 2000; Cheung et al., 2005; Park et al., 2012; Théveneau et

    al., 2007). It should be noted, however, that this model may not apply to Xenopus as in this

    species, cadherin-6 is not found in premigratory neural crest (David and Wedlich, 2000) and that

    the E- to N-cadherin switch occurs in the neural plate very early at onset of neurulation, with no

    transition during which both molecules are coexpressed (Nandadasa et al., 2009). In this species,

    which cadherin is recruited to mediate neural crest sorting from the rest of the neurectoderm and

    how it is regulated during delamination are not known. The continuous presence of N-cadherin

    over the surface of cranial neural crest cells during migration (Theveneau and Mayor, 2012)

    raises the intriguing question as to whether onset of migration might not involve a true EMT.

  • 17

    Coordinating neural crest cell specification and delamination: spatiotemporal control of

    the core EMT regulatory factors

    Although the moment when neural crest cells exit the neural tube extends generally over less

    than an hour (Ahlstrom and Erickson, 2009; Clay and Halloran, 2010), Snail-1/2, Zeb-2 and in a

    lesser extent, Twist-1 all appear in neural crest progenitors early during specification long before

    onset of EMT. In addition, as discussed above, their major cadherin targets undergo different

    temporal patterns of repression. This situation contrasts strikingly with that found in the

    primitive streak during gastrulation, where ingressing mesodermal cells express Snail-2 in an

    dynamic fashion and immediately execute a complete EMT program with breakdown of the

    basement membrane, loss of cell polarity and a rapid E- to N-cadherin switch (Acloque et al.,

    2011; Dady et al., 2012; Nakaya et al., 2008). This suggests that in the neural crest, the

    transcription-repressing activity of the core EMT regulatory factors is finely tuned over time and

    that prior to EMT, these factors may exert additional functions independently of their role in

    delamination by regulating different sets of "early" and "late" genes. In this respect, based on

    studies in Xenopus, it has been proposed for long that Snail-1/2 plays a role in both specification

    and delamination (LaBonne and Bronner-Fraser, 2000; Mancilla and Mayor, 1996), albeit no

    specific Snail-1/2 target genes during specification have been identified yet. Therefore, knowing

    how Snail, Zeb and Twist expressions and activities are regulated may provide insight into their

    missions throughout neural crest development.

    EMT is a strictly-controled process driven by multiple, closely-interconnected regulatory

    networks centered around Snail, Zeb and Twist and acting at different levels; and in the past few

    years, a plethora of factors and pathways modulating their expression, cellular localization,

    stability and activity have been identified particularly in the case of Snail-1/2 (Figure 4 and De

    Craene and Berx, 2013, for review). The possible implication of a number of them during neural

  • 18

    crest cell delamination has been investigated during the last decade, most often at cranial levels

    in the Xenopus model, and a temporal sequence of their activity progressively emerges that

    might account for the time course of cell specification and delamination.

    Transcriptional and translational control of the expression of the core EMT regulatory factors

    Transcriptional control of expression of the core EMT regulatory factors is the primary level of

    regulation of the EMT program, and this has been particularly investigated for neural crest cells

    owing to the knowledge accumulated on the gene regulatory networks supporting their

    development (Betancur et al., 2010 and Prasad et al., 2012). As in cancer cells (Lim and Thiery,

    2012; Thiery and Sleeman, 2006), EMT being triggered in crest cells by the BMP-4 and Wnt

    signaling pathways, the classical transcription factors recruited by these pathways (e.g. Smads

    for BMPs and Lef/Tcf for Wnts) have been shown to upregulate directly expression of Snail-1/2

    in these cells (Sakai et al., 2005). However, aside from these factors, additional transcription

    factors have been identified as potent inducers of core EMT regulators during neural crest cell

    specification and delamination. Sox-9, for example, is essential for Snail-2 expression and its

    effect is mediated through direct binding to the Snail-2 promoter in synergy with Snail-2 itself,

    able to activate its own expression (Sakai et al., 2006). Other factors such as the border

    specifiers, Pax-3/7 and Msx-1, can also upregulate Snail-2, but whether it is directly or via other

    factors has not been investigated yet. Recently, functional experiments have placed Elk-3, a

    member of the Ets family of transcription factors present in neural crest cells early during

    specification and migration, downstream of border specifiers and upstream of neural crest

    specifiers, including Snail-2, in the genetic cascade accompanying their development (Rogers et

    al., 2013). How Elk-3, known to function as a transcriptional repressor, control these genes is not

    known. On the other hand, c-Myb, a transcription factor with well-known functions in

    hematopoiesis but also expressed in premigratory and early migrating neural crest cells, has been

  • 19

    found to control expression of Snail-2 and Twist-1 (Betancur et al., 2014; Karafiat et al., 2005).

    Whether this cohort of factors is responsible for modulating expression of Snail-1/2 (as well as

    Twist-1 and Zeb-2) between specification and delamination has not been formally established,

    and to our knowledge, no transcription factors whose expression correlates with initiation of

    delamination have been identified so far. The only described exception is Ets-1, the prototypic

    member of the Ets family, that is induced just prior to EMT at cranial levels. Functional studies,

    however, indicate that Ets-1 acts merely as a coordinator of delamination for cranial neural crest

    cells and not as a direct modulator of Snail-2 expression (Théveneau et al., 2007).

    Relatively little is known about the transcriptional repressors of Snail, Twist and Zeb. At

    variance with what was found in neural crest cells, Snail-1 has also been reported to bind to an E

    box in its own promoter and repress its transcription (Peiro et al., 2006). On the other hand, a

    recent study has identified Sox-3 as a potent negative regulator of Snail-2 expression during

    chick gastrulation. Sox-3 expressed in the neural plate represses Snail-2 gene, thus protecting

    them from ingressing, while in the primitive streak Snail-2 conversely represses Sox-3, activating

    EMT and ingression of precursors of the mesoderm (Acloque et al., 2011). Whether this

    antagonistic relationship also applies to Sox-2/3 and Snail-2 in neural crest cells is not known.

    Although highly plausible, because Sox-2, Sox-3 and Snail-2 are mutally exclusive in neural crest

    progenitors and in the neural tube (Uchikawa et al., 2011; Wakamatsu et al., 2004), it remains to

    be demonstrated. Indeed, Sakai et al. (2006) reported that Sox-2 does not affect Snail-2 promoter

    activity in fibroblast cell lines.

    Small non-coding RNAs also emerged as potent modulators of the expression of the core

    EMT regulatory factors. MicroRNAs (miRNAs) typically suppress gene expression by

    interacting with the 3’-untranslated region of target messengers to repress their translation or

    degrade them. A large number of miRNAs have been involved in the control of EMT and,

    among them, miR-34 and the miR-200 family are strongly associated with the epithelial state and

  • 20

    are downregulated upon EMT (De Craene and Berx, 2013, for a review). While miR-34

    represses Snail-1 expression, miR-200s act as silencers for Zeb-1/2. Interestingly, both miRNAs

    contain E boxes in their promoters and are directly regulated by Snail and Zeb, thereby revealing

    intricate reciprocal feedback loops between miRNAs and their targets (Burk et al., 2008;

    Siemens et al., 2011). These miRNAs are, on the other hand, under the positive control of p53

    (Chang et al., 2011), an important player of neural crest development (see below). Contrary to

    miR-200s and miR-34, miR-9, which is upregulated in breast cancer cells and activated by Myc,

    has been found to target directly E-cadherin messages, leading to increased cell motility and

    invasiveness in epithelial cell lines (Ma et al., 2010). At the present time, the presence of these

    various miRNAs during specification and delamination of neural crest cells has not been

    reported. Yet, miRNAs undubiously play an important role in these cells (Mayanil, 2013; Strobl-

    Mazzulla et al., 2012). Indeed, in mouse, specific deletion in neural crest of DICER, the RNase

    III enzyme required for cleavage of precursor miRNAs into mature miRNAs, leads to

    craniofacial and cardiac anomalies (Zehir et al., 2010). Likewise, in Xenopus, loss of DICER or

    of FMR1 (fragile-X mental retardation syndrome-1) and FXR1 (fragile-X related-1), two RNA-

    binding proteins interacting with the miRNA-induced silencing complex RISC, also causes

    craniofacial defects due to strong reduction of neural crest migration (Gessert et al., 2010). On

    the other hand, consistent with its role in repressing Zeb expression and preventing EMT,

    depletion of miR-200b affects development of cranial neural crest-derived structures but,

    unexpectedly, this effect was not found to result from the block in induction and migration.

    Depletion of another miRNA miR-96, in contrast, provoke alterations of both induction and

    migration (Gessert et al., 2010). Clearly, more systematic analyses of the repertoire, function and

    interplay of miRNAs regulating expression of the core EMT factors during neural crest cell

    delamination are needed to define precisely how they contribute to its spatiotemporal control.

  • 21

    Epigenetic control of the expression of the core EMT regulatory factors

    Control of Snail, Zeb and Twist expression may also be part of more general cellular programs

    regulated at the epigenetic level, allowing genome-wide coordinated modulation of gene activity.

    Though still preliminary, recent studies have highlighted the major transcriptional reprograming

    events that accompany neural crest cell specification and delamination and provide insights into

    the synergistic control of neural crest specifiers by the epigenetic machinery (Mayanil, 2013;

    Strobl-Mazzulla et al., 2012).

    DNA methylation by DNA-methyltransferases (DNMTs) is one of the epigenetic

    modifications resulting in transcriptional repression of genes. Only DNMT-3A/3B, the DNMTs

    responsible for de novo methylations, have been studied during neural crest development.

    DNMT-3B is expressed in premigratory and migrating cells but is not restricted to them, and

    genetic studies in mouse demonstrated that it is largely dispensable in cranial and cardiac neural

    crest cells for migration and differentiation, but is however required in their host tissues,

    branchial arches and heart, during the late steps of their development (Jacques-Fricke et al.,

    2012). DNMT-3A, in contrast, shows a more-restricted expression pattern than DNMT-3B as it

    is at first present in the neural plate border and becomes later confined to premigratory and

    migrating neural crest cells. Knockdown of DNMT-3A provokes a severe reduction in

    expression of Snail-2 and of a variety of other neural crest specifiers, accompanied by a

    remarkable increase in Sox-2/3. Importantly, the latter two genes appeared as direct targets of

    DNMT-3A in neural crest cells. These data therefore indicate that DNA methylation acts as a

    major molecular switch to turn off neural tube transcription factors in neural plate border cells

    and promote neural crest cell fate (Hu et al., 2012). The mechanism by which DNMT-3A is

    recruited to the promoters of its target genes in neural crest cells is currently unknown, but it has

    been shown to interact with a multiplicity of transcription factors, including some that are highly

    relevant to neural crest formation, i. e. p53, Ets-1 and AP-2 (Hervouet et al., 2009). It would be

  • 22

    then of interest to establish the complete list of targets of DNMT-3A in neural crest cells and to

    determine by which factors it is recruited to them.

    Beside DNA methylation, histone methylation and acetylation are the major chemical

    modifications which influence chromatin structure and regulate gene expression (Kouzarides,

    2007). Methylation of histone H3 at specific amino-acid residues recruits a variety of modifiers

    of chromatin and transcriptional activators or repressors, resulting in differential effects on gene

    expression. Roughly, trimethylated lysine 4 of histone H3 (H3K4me3) and H3K36me3 are

    associated with active transcription, while H3K27me3 and H3K9me3 are repressive marks.

    Histone methylation are regulated by the interplay between methyltransferases and

    demethylases, notably the Jumonji proteins (Jmj). To date, only Jmj-D2A has been reportedly

    implicated in the control of Snail-2 expression during neural crest development (Strobl-Mazzulla

    et al., 2010). It is expressed initially throughout the neural plate and at the border and gradually

    resolves to the premigratory neural crest. Of interest, its expression is highest during early

    specification and then declines until onset of migration at which stage it is almost no longer

    expressed. Consistently, loss of Jmj-D2A expression causes dramatic downregulation of neural

    crest specifier genes, notably Snail-2, Foxd-3, and Sox-10, but is ineffective on border genes.

    Additionally, the location and abundance of the H3K9me3 and H3K36me3 epigenetic marks, as

    revealed by chromatin immunoprecipitation, show dynamic occupancy of sites in proximity to

    the transcriptional start site of Snail-2 and Sox-10 genes and clearly reflect their transcriptional

    state. At initiation of specification, Sox-10 exhibits essentially H3K9me3 repression marks,

    while Snail-2 shows a more mitigated pattern with some H3K9me3 repression but H3K36m3

    activation marks as well, reflecting an ongoing transcriptional shift from repressed to active.

    Later, by the time of migration, the occupancy of the repressive mark H3K9me3 near the Sox-10

    and Snail-2 genes is clearly reduced, consistent with high expression of these genes by this stage,

    while H3K36me3 marks are unchanged. Correlating with these changes, direct interactions

  • 23

    between Jmj-D2A and the regulatory regions of Sox-10 and Snail-2 genes can be detected at

    early specification stages but not later during migration. Finally, demethylation��� of H3K9me3 on

    the Sox-10 promoter is inhibited upon knockdown of Jmj-D2A in early neural crest cells. These

    results therefore indicate that the relief of transcriptional repression by the Jmj-D2A-mediated

    demethylation of H3K9me3 marks plays an important role in induction of neural crest specifier

    genes.

    Other chromatin remodellers, such as CHD-7, a chromodomain helicase DNA-binding

    protein homologous to the Drosophila trithorax protein Kismet, have also been shown to control

    expression of core EMT regulators during neural crest development (Bajpai et al., 2010). In

    Xenopus, CHD-7 is expressed in the neural ectoderm and in premigratory and migrating neural

    crest cells. Depletion of CHD-7 specifically targets neural crest specifiers, such as Snail-2, Sox-

    9, and Twist-1 and not border specifiers, and it causes major craniofacial defects. In addition,

    CHD-7 associates in the nucleus with the PBAF (polybromo- and BRG1-associated factor)

    complex, and both remodellers bind to a neural crest-specific distal Sox-9 enhancer and a

    conserved genomic element located upstream of the Twist-1 gene. Consistently, CHD-7 and

    PBAF-bound regions were preferentially enriched for H3K4me1, a mark previously associated

    with enhancers. These data therefore indicate that, complementing Jmj-D2A activity, CHD-7 and

    PBAF cooperate to promote expression of neural crest specifiers.

    Control of the stability and intracellular location of the core EMT regulatory factors

    Snail proteins are by nature highly unstable, including in neural crest cells (Vernon and

    LaBonne, 2006). Snail turnover in the cytoplasm is tightly regulated by regulatory mechanisms

    involving posttranslational phosphorylation and ubiquitination (De Craene and Berx, 2013).

    Schematically, these regulatory mechanisms can be classified into glycogen synthase kinase-3β

    (GSK-3β)-dependent and independent processes (Figure 4). In the GSK-3β-dependent process,

  • 24

    cytoplasmic Snail-1 proteins are phosphorylated at serine-threonine residues first by casein

    kinase-1 then by GSK-3β and they are subsequently processed for ubiquitination by the E3

    ubiquitin ligase β-TRCP-1 for degradation (Vinas-Castells et al., 2010). Thus, signaling events

    elicited by Wnt and other growth factors causing inactivation of GSK-3β may repress Snail-1

    degradation and favor transition toward a mesenchymal phenotype. Given that Wnt-1 signals are

    activated in neural crest progenitors prior to delamination, it is tempting to suggest that their

    activity may recruit β-TRCP-1 resulting in increased stability of Snail proteins. Interestingly, at

    least in chick and mouse, Wnt-1 expression is induced late during neural crest cell specification,

    shortly prior to onset of migration (Burstyn-Cohen et al., 2004), suggesting that Snail

    stabilization upon inhibition of GSK-3β activity may constitute a late triggering event of neural

    crest cell delamination. This hypothesis awaits to be demonstrated, however.

    Beside β-TRCP-1, GSK-3β-independent ubiquitin ligases, such as MDM-2 and FBXL-

    14, and the F-box protein partner of paired (PPA) have also been reported to target Snail proteins

    for degradation (Vinas-Castells, 2010; Lander et al., 2011), and at least two of them, MDM-2

    and PPA have been implicated in neural crest development (Daujat et al., 2001; Lander et al.,

    2011; Vernon and LaBonne, 2006). Indeed, it has been found in Xenopus that, contrary to Snail-

    1, Snail-2 does not contain in its sequence the β-TRCP-1 destruction motif, and inhibiting GSK-

    3β causes only marginal increase in Snail-1/2 stability, suggesting that in this species the Wnt-

    1/GSK-3β/β-TRCP-1 axis plays only a minor contribution to Snail regulation (Lander et al.,

    2011; Vernon and LaBonne, 2006). In contrast, PPA does not target only Snail-1 but also Snail-

    2, Twist-1 and Zeb-2, making this pathway a major candidate for regulating the timing of neural

    crest cell delamination (Lander et al., 2011; Vernon and LaBonne, 2006). PPA is induced

    specifically in premigratory neural crest cells just prior to EMT. In addition, morpholino-

    depletion of PPA stabilizes Snail-2 protein, whereas its misexpression promotes its turnover and

    inhibits the formation of neural crest precursors. Importantly, Sox-9 and Foxd-3 have been found

  • 25

    to repress expression of PPA and oppose its Snail-protein degradation activity when

    overexpressed in the neural tube. Thus, prior to EMT, Snail protein levels (and probably those of

    Zeb-2 and Twist-1) would be maintained constant through repression of PPA by the combined

    activities of Sox-9 and Foxd-3. Then, upon EMT, PPA levels would increase, resulting in higher

    Snail-1/2, Zeb-2 and Twist-1 degradation. Consistently, PPA inhibition has been found to cause

    migration defects, indicating that high levels of Snail-2 might be deleterious for efficient neural

    crest migration (Vernon and Labonne, 2006). Given that neither Sox-9 nor Foxd-3 are

    downregulated at onset of migration, which factors relieve PPA repression remains to be found.

    Insights into the role of MDM-2 in neural crest cells come essentially from studies on the

    regulation of the tumor suppressor p53. Indeed, p53 is one of the main targets of the MDM-2

    ubiquitination activity, an important step in the oncogenic process. MDM-2 expression has been

    found in murine and avian cranial neural crest cells prior to delamination and to increase in

    migrating cells (Daujat et al., 2001; Rinon et al., 2011). Conversely, p53 is at a basal level in

    premigratory neural crest and declines after delamination (Rinon et al., 2011). Functional

    experiments in chick reveal that stabilization of the endogenous p53 protein by an inhibitor of

    MDM-2 activity reduces Snail-2 expression and cell proliferation and inhibits neural crest cell

    delamination but not their specification (Rinon et al., 2011). Conversely, loss of p53 resulted in a

    marked increase in cranial neural crest progenitors in the neural tube and increased cell

    proliferation. Nevertheless, neural crest cells fail to leave the neural tube, suggesting that p53

    levels must be finely tuned by MDM-2 activity in neural crest progenitors for delamination and

    migration to occur. Further studies by Wang et al., 2011 showed that Pax-3 plays a role in neural

    crest cell development, not through its DNA-binding and transcription-regulation activities, but

    essentially by blocking p53 function. Pax-3 has no effect on p53 mRNA levels or the rate of p53

    synthesis but it reduces p53 protein stability by interacting physically with p53 and MDM2.

    These data suggest that during delamination, the Pax-3/ MDM-2/p53 signaling cascade is critical

  • 26

    for controling the rate of cell growth and division in neural crest cells. Whether MDM-2

    ubiquitination of Snail-2 is required for this process has not be investigated, however. Moreover,

    the finding that p53 may control cell invasion in cancer cells by inducing the MDM-2 –mediated

    degradation of Snail-2 (Wang et al., 2009) is indicative of complex feedback loops between p53,

    MDM-2 and Snail-2 during EMT.

    As nuclear Snail proteins degrade more slowly than cytoplasmic Snail, there are

    mechanisms that modulate their nuclear trafficking (Figure 4). Thus, two kinases, PAK-1 and

    LATS-2, favor nuclear retention of Snail-1, thereby enhancing its stability and activity (Yang et

    al., 2005; Zhang et al., 2012). On the other hand, the zinc transporter LIV-1, a STAT-3 target,

    has been found to drive nuclear import of Snail-1 in the zebrafish gastrula and promote EMT

    (Yamashita et al., 2004), whereas the protein kinase D1 (PRKD-1) by enhancing its nuclear

    export in contrast restricts EMT (Du et al., 2010). The roles of LATS-2, LIV-1 and PRKD-1 in

    neural crest cell delamination have not been investigated yet, but a recent report indicates that in

    Xenopus, PAK-1 expression coincides temporally with their migration (Bisson et al., 2012),

    suggesting that it may operate even after delamination. Consistent with this restricted pattern,

    expression of a dominant-negative form of PAK-1 blocks migration but does not affect

    specification. Intriguingly, PAK-1 is able to phosphorylate Snail-1 and Twist-1 but not Snail-2,

    indicating that during migration requirement for Snail-1 and Snail-2 activities are different.

    Recently, a novel level of regulation of Snail-1/2 activity has been uncovered in Xenopus,

    implicating Twist-1. Although it is commonly presented as an important player in neural crest

    cell specification and delamination (Betancur et al., 2010; Milet and Monsoro-Burq, 2012;

    Prasad et al., 2012; Sauka-Spengler and Bronner-Fraser, 2008), Twist-1 function has long

    remained elusive, and it is only very recently that it has been reevaluated. Unlike Snail-1/2 and

    Zeb-2, Twist-1 appears relatively late during the process of neural crest specification and it is

    restricted to cranial regions. However, elegant studies by Lander et al. (2013) showed that Twist-

  • 27

    1 misexpression or depletion impacts on the levels of Snail-1/2 messages as well as on other

    neural crest specifiers. Interestingly, this is not merely through regulation of gene expression but

    primarily through direct binding with Snail-1/2 proteins. The C-terminal WR domain of Twist-1

    interacts with the N-terminus of Snail via a cluster of GSK-3β-phoshorylated aminoacids; of

    note, this interaction does not interfere with the ability of Twist-1 to bind DNA (Figure 5).

    Moreover, Snail-1/2 binding stabilizes Twist-1 by competing with PPA association, suggesting

    that Snail-1/2 impacts positively on Twist-1 activity. Surprisingly enough, GSK-3β-

    phosphorylation of Twist-1 titers Snail-1/2 and diminishes their recruitment to E boxes in the

    promoters of their target genes. Thus, although Snail-2 is not directly targeted to degradation by

    GSK-3β-phosphorylation (see above), these findings ascribe to GSK-3β a critical role in the

    control of its activity via Twist-1, and in the future, determining where and when Twist-1 is

    phosphorylated by GSK-3β will be of importance to decipher the mechanisms triggering neural

    crest cell EMT.

    Finally, recent preliminary data on paladin, an antiphosphatase, may provide interesting

    information regarding how the activity of the core EMT regulators can be modulated over time

    in premigratory neural crest cells (Roffers-Agarwal et al., 2012). Antiphosphatases have been

    shown to bind phosphorylated residues on target proteins and protect them from

    dephosphorylation. In chick, paladin is expressed during neural crest development both at cranial

    and truncal levels. Decreasing paladin expression precociously during neural crest induction

    using morpholinos results in a substantial but not complete reduction in Snail-2 expression at

    premigratory stages. Interestingly, only Snail-2 is strongly repressed, the other targets tested

    being affected either moderately (e.g. Sox-10) or not at all (Foxd-3, cadherin-6B, and Rho-B).

    This indicates that paladin may regulate only a limited set of targets during neural crest

    specification. Moreover, cadherin-6B is apparently down-regulated on time in the absence of

    paladin, suggesting that essentially Snail-2 activities relevant to specification were altered. At the

  • 28

    light of the data presented above, it is tempting to speculate that control of the level of

    phosphorylation of Snail-2 by paladin is critical for some of its activities and may vary over time

    during specification and delamination. Further experiments, however, are needed to verify this

    statement and to better define paladin role and targets. For example, determining which

    aminoacids on Snail are targeted by paladin and whether Snail stability and trafficking are

    affected upon manipulation of its expression should help better understanding how maintenance

    of phosphorylation involving paladin contribute to control of neural crest cell EMT.

    Control of the activity of the core EMT regulatory factors

    One of the reasons for the ability of the Snail, Zeb and Twist transcription factors to induce

    massive phenotypic changes during EMT resides in part in their close interaction with epigenetic

    modifiers, allowing genome-wide changes in gene expression. Indeed, their transcriptional

    regulatory activity requires the participation of various interacting proteins which results in a

    strong transcriptional control of EMT, and this has been particularly well studied in the case of

    the E-cadherin gene. For example, DNA methylation of the E-cadherin promoter responsible for

    its silencing has been correlated with EMT in cancer cells (Lombaerts et al., 2006) and

    associated with increased Snail-1 expression (Cheng et al., 2001). More intriguingly though, it

    has also been reported that DNMT-1 can repress E-cadherin expression in the absence of

    noticeable changes in DNA methylation patterns in its promoter and that this effect is mediated

    by direct interaction with Snail-1 (Espada et al., 2011). During neural crest development,

    although DNA methylation by DNMT-3A has been shown to act as a major molecular switch to

    turn off neural tube transcription factors in neural plate border cells and promote neural crest cell

    fate (Hu et al., 2012), it is not known yet whether Snail-1/2 and the other core EMT regulators

    can associate directly or via adaptors with DNMTs to target specific genes.

    Beside DNMTs, Snail-1 has been found to associate with a variety of histone modifiers to

  • 29

    regulate gene expression (Figure 5). In particular, it induces histone deacetylation of the E-

    cadherin gene through the recruitment of Sin-3A in association with histone deacetylases

    (HDAC-1/2) (Peinado et al., 2004). Interestingly, recent studies by Strobl-Mazzulla and Bronner

    (2012b) uncovered the mechanism by which Snail-2 recruits HDAC to repress cadherin-6B

    expression in neural crest cells in chick. They identified PHD-12, an adaptor protein, whose

    expression is markedly increased in cranial crest cells just before EMT. PHD-12 loss of function

    phenocopies Snail-2 knockdown, inhibiting transcriptional repression of cadherin-6B and

    preventing neural crest emigration. PHD-12 and Snail-2 do not bind to each other, but both

    directly interact with Sin-3A, which in turn complexes with HDAC. PHD-12 is recruited to the

    cadherin-6B promoter during neural crest EMT. Consistent with this, lysines on histone H3 at

    the cadherin-6B promoter are hyperacetylated before neural crest emigration, correlating with

    active transcription, but deacetylated during EMT, reflecting a shift to a repressive state. Finally,

    knockdown of either PHD-12 or Snail-2 prevents deacetylation of the cadherin-6B promoter.

    Collectively, these results suggest that repression of the cadherin-6B gene occurs through

    binding of PHD-12 and Snail-2 to their transcription start site and E boxes. This allows Sin-3A

    and HDAC to be recruited and to deacetylate histone H3 at the promoters, resulting in repression

    of transcription. Of interest, the authors mention that this model also applies to E-cadherin

    repression. In an other study, Murko et al. (2013) showed that treatment of early embryos using a

    pharmacological blocker of HDAC causes neural tube defects at trunk levels, increase in

    cadherin-6B expression and a premature loss of epithelial features among neural epithelial cells.

    Among the different adaptors with which Snail proteins interact in neural crest cells to

    recruit HDAC and drive transcriptional repression of genes are the LIM proteins, characterized

    by the presence of two LIM domains in tandem (Figure 5). The first LIM proteins to be

    identified were the Ajuba family which were found to function as a corepressor of Snail-1/2

    (Langer et al., 2008). Ajuba interacts predominately with the SNAG domain of Snail-1 in cells

  • 30

    and accumulates in the nucleus in a SNAG-dependent manner. This interaction with Snail-1

    potentiates Snail-1 binding to the E boxes of the E-cadherin promoter and promotes its

    repressing activity. Interestingly, a recent study by Ochoa et al. (2012) revealed that contrary to

    Snail-1, Snail-2 binds to Ajuba via its Zn-finger motifs. Expression of Ajuba in Xenopus

    embryos enhances neural crest development in a Snail-1/2-dependent manner, while conversely,

    it depletion phenocopies depletion of Snail-1/2. This study therefore assigns to Ajuba LIM

    proteins a critical role in neural crest cell development as Snail-1/2 corepressors. However, their

    presence in these cells has not been assessed in this study. Interestingly, Ajuba LIM proteins are

    also components of adherens junctions and contribute to their assembly or stability (Srichai et

    al., 2004). Their functional interaction with Snail proteins in the nucleus suggests that they are

    also important regulators of the dynamics of epithelial cells, linking surface events with nuclear

    responses. It would then be informative to determine when and where Ajuba LIM proteins are

    expressed in neural crest progenitors to better evaluate their implication in the spatiotemporal

    control of Snail-1/2 activity.

    More recently, studies in Xenopus and chick identified the LIM domain only protein 4

    (LMO-4) as another LIM protein involved in neural crest development (Ferronha et al., 2013;

    Ochoa et al., 2012). In both species, LMO-4 is distributed in the neural crest and neural plate

    progenitors during early specification to become restricted to neural crest cells at the time of

    their delamination and early migration. In addition, functional studies revealed that LMO-4 is a

    Snail-1/2-interacting protein that is essential for neural crest development. Morpholino-mediated

    knockdown of LMO-4 leads in Xenopus to a deficit in the production of neural crest progenitors

    and in chick to a severe reduction in their delamination. Additionally, while misexpression of

    LMO-4 in the trunk neural tube in chick is insufficient to induce expression of neural crest

    specifiers but speeds up neural crest delamination, in Xenopus, excess LMO-4 leads to ectopic

    expression of Snail-1/2, and to a reduction in the expression of Foxd-3, Sox-8/9/10 and Twist-1.

  • 31

    These differences between chick and frog are likely not to result from species specificities but

    rather to differences in the timing when LMO-4 was overexpressed (late during neural crest

    development in chick and early during specification in frog), thereby revealing different

    requirements for LMO-4 throughout neural crest development. Detailed analyses of LMO-4

    binding specificities further showed that it binds directly to Snail-1/2, but not to other neural

    crest specifiers, and this interaction occurs via the SNAG domain of Snail, like Ajuba LIM

    proteins (Figure 5). Intriguingly, Snail-1/2 binding to LMO-4 does not lead to HDAC

    recruitment, thus raising concerns about its mode of action. The fact that LMO-4 competes with

    HDAC for Ajuba binding suggests that it may play a key modulatory role of Snail-1/2

    interactions with its partners, thereby selectively targeting distinct sets of genes. It will be

    essential then to determine whether LMO-4 recruits specific proteins to Snail-regulatory

    complexes on target promoters and to define how Ajuba and LMO-4 are coordinated over time

    during neural crest development.

    Twist-1 has also been found to recruit epigenetic modifiers as partners for regulating E-

    cadherin gene expression. Notably, it has been found to associate with Bmi-1, a member of the

    polycomb repressive complex, to suppress E-cadherin transcription through binding to the E

    boxes situated in its promoter. Interestingly, Bmi-1 is itself under the direct transcriptional

    control of Twist-1 (Yang et al., 2010). Knowing the complex relationships between Twist-1 and

    Snail-1/2 in neural crest cells (see above), determining whether Bmi-1 is expressed during their

    delamination and may contribute to the control of cadherin expression will be valuable to

    decipher these interactions in greater details.

    Although a coherent, complete picture of the regulatory networks controling expression

    and activity of the core EMT regulators Snail-1/2, Zeb-2 and Twist-1 is still lacking (in part

    because studies were performed in different animal models characterized by distinct gene

    regulatory networks), two major conclusions can be drawn for these studies (Figure 6). First, a

  • 32

    multiplicity of transcription factors and chromatin modifiers and remodellers bind to different

    domains in the regulatory sequences of Snail-1/2, Twist-1 and most likely Zeb-2 (although this

    has not been documented yet) and are recruited to relieve repressing epigenetic marks and induce

    their robust expression. In addition, their restricted expression patterns and functions during

    specification clearly link their activity to the acquisition by cells of their neural crest identity

    rather than to their delamination. Second, the intracellular localization of Snail-1/2, its stability,

    and its activity are dictated by a plethora of kinases and nuclear corepressors which are induced

    or activated concurrently just prior to delamination, suggesting that collectively they function as

    a molecular switch to induce neural crest cell EMT and that this event requires high doses of

    Snail proteins to regulate massively within a short time period a large array of effectors. This

    model represents however only a part of the whole picture as it is essentially based on the control

    of the repressing activities of Snail proteins on its cadherin targets; and it is now largely

    demonstrated that they are also capable of activating numerous genes, in particular those related

    to the locomotory machinery.

    Spatial control of the activity of the Zeb-2 transcription factor

    Contrary to the other core EMT regulatory factors, Zeb-2 is not restricted to premigratory neural

    crest cells but is also expressed by the entire neural plate. Zeb-2 expression is induced

    precociously during primary neurulation and correlates with the progressive switch from E- to N-

    cadherin in neural epithelial cells (Dady et al., 2012). Knock-down of Zeb-2 in the neural tube at

    cranial levels in chick or mouse embryos results in persistent expression of E-cadherin in neural

    epithelial cells and also in strong inhibition of neural crest cell delamination and migration

    (Rogers et al., 2013; Van de Putte et al., 2003). Thus, Zeb-2 is clearly an important player in the

    EMT process of neural crest cells by promoting E-cadherin silencing. Yet, unlike neural crest

    cells, neural epithelial cells retain their epithelial morphology and fail to undergo EMT despite of

  • 33

    persistent Zeb-2 expression over long time periods. Why the same molecule mediates radically-

    different behaviors in neighboring cells that share in common almost the same origin and

    numerous cellular features? As discussed above, an explanation resides in the fact that neural

    tube cells do not express neural crest specifier genes and therefore lack the appropriate signaling

    modules that enable them with the ability to undergo full EMT and disperse. However,

    additional reasons may be invoked that could account for the exquisite manner by which neural

    crest cell delamination is regulated. In particular, it has been found that Zeb-2 is evenly

    expressed at moderate levels in the neural tube and neural crest throughout neurulation during

    neural crest specification but that its expression is sharply increased in neural crest cells at the

    time when they undergo EMT, indicating that execution of the EMT program may require higher

    doses of Zeb-2 than the sole control of E- to N-cadherin switch (Dady et al., 2012; Rogers et al.,

    2013; Van de Putte et al., 2007). This assumption is supported by a recent study on the genetic

    network regulating EMT in cancer cells proposing a model accounting for the three distinct

    states in which cells can be alternatively found, i.e. epithelial, mesenchymal and a so-called

    hybrid state, in which they show mixed mesenchymal and epithelial characteristics (Lu et al.,

    2013). This structural organization of cells was proposed to be regulated by the circuit composed

    of the two interconnected modules, the miR-34/Snail-2 and the miR-200/Zeb-2 modules,

    mentionned above. Interestingly, this model suggests that the miR-34/Snail-2 module exists only

    under two mutually-exclusive forms, (high miR-34/low Snail-2) and (low miR-34/high Snail-2),

    and is a central regulatory circuit of EMT, acting as an on/off switch responsible for either the

    epithelial or the mesenchymal states. The miR-200/Zeb-2 module in contrast can be found in

    three forms, (high miR-200/low Zeb-2), (low miR200/high Zeb-2) and (medium miR-

    200/medium Zeb-2), corresponding to the epithelial, mesenchymal and hybrid states,

    respectively, and may function as a more-refined modulator of the physiology of epithelial cells.

    Although the expression levels of miRNAs in the neural tube remain to be established, the

  • 34

    relatively-modest level of Zeb-2 in neural epithelial cells associated with the lack of Snail-2 may

    explain why, in spite of E- to N-cadherin switch, these cells undergo only partial transformation

    into a pseudostratified epithelium characterized by elongated cell shapes and absence of tight

    junctions, but are unable to achieve complete EMT. Thus, the combination of cadherin switches

    in the early blastoderm under the tight spatial and temporal control of the core EMT regulatory

    factors Snail-2 and Zeb-2 may be responsible for its partition into three distinct portions

    exhibiting different epithelial properties and capacities to execute the EMT program.

    Future prospects

    The few examples described above illustrate how time and space are coordinated in an exquisite

    manner to permit a population of progenitor cells to execute sequentially each of the individual

    steps that eventually lead them to populate a great variety of territories in the embryo and

    provide an extremely large array of cell types. This is achieved through an intricate combination

    of genome-wide epigenetic control of gene expression, recruitment of specific transcription

    factors that act in concert and activation of chemical modifications that regulate locally the

    availability and activity of these factors. How these different actors are timely coordinated and

    whether there is an hierarchy among them will certainly constitute an important subject of

    investigation in the near future. Clearly the sole activations of signaling pathways such as BMPs,

    Wnt and FGF are far from sufficient to account for the precision by which these events occur, as

    illustrated in the case of induction of the neural plate border. Physical processes such as

    constraints applied by forces exerted by moving cells or the viscoelastic properties of tissues are

    likely to play instructive roles at the scale of the whole embryo. We are only starting to address

    the question of how these physical processes are translated into biological cascades and vice

    versa how biological mechanisms can modulate and refine ordered and coherent physical events.

  • 35

    Acknolwledgments

    The authors wish to thank Paul Trainor for his kind invitation to write this review and for his

    patience and to express their friendship for all their colleagues working in the field for their

    endless imagination and stimulating contributions.

    References Acloque, H., Ocana, O.H., Matheu, A., Rizzoti, K., Wise, C., Lovell-Badge, R., Nieto, M.A.,

    2011. Reciprocal repression between Sox3 and snail transcription factors defines embryonic

    territories at gastrulation. Dev Cell 21, 546-558.

    Ahlstrom, J.D., Erickson, C.A., 2009. The neural crest epithelial-mesenchymal transition in 4D:

    a 'tail' of multiple non-obligatory cellular mechanisms. Development 136, 1801-1812.

    Akitaya, T., Bronner-Fraser, M., 1992. Expression of cell adhesion molecules during initiation of

    and cessation of neural crest cell migration. Dev Dyn 194, 12-20.

    Aybar, M.J., Mayor, R., 2002. Early induction of neural crest cells: lessons learned from frog,

    fish and chick. Curr Opin Genet Dev 12, 452-458.

    Bajpai, R., Chen, D.A., Rada-Iglesias, A., Zhang, J., Xiong, Y., Helms, J., Chang, C.P., Zhao,

    Y., Swigut, T., Wysocka, J., 2010. CHD7 cooperates with PBAF to control multipotent neural

    crest formation. Nature 463, 958-962.

    Barembaum, M., Moreno, T., LaBonne, C., Sechrist, J., Bronner-Fraser, M., 2000. Noelin-1 is a

    secreted glycoprotein involved in generation of the neural crest. Nature Cell Biol. 2, 219-

    2225.

    Basch, M.L., Bronner-Fraser, M., Garcia-Castro, M.I., 2006. Specification of the neural crest

    occurs during gastrulation and requires Pax7. Nature 441, 218-222.

    Betancur, P., Bronner-Fraser, M., Sauka-Spengler, T., 2010. Assembling neural crest regulatory

    circuits into a gene regulatory network. Annu Rev Cell Dev Biol 26, 581-603.

    Betancur, P., Simoes-Costa, M., Sauka-Spengler, T., Bronner, M.E., 2014. Expression and

    function of transcription factor cMyb during cranial neural crest development. Mech Dev 132,

    38-43.

    Bisson, N., Wedlich, D., Moss, T., 2012. The p21-activated kinase Pak1 regulates induction and

    migration of the neural crest in Xenopus. Cell Cycle 11, 1316-1324.

  • 36

    Burk, U., Schubert, J., Wellner, U., Schmalhofer, O., Vincan, E., Spaderna, S., Brabletz, T.,

    2008. A reciprocal repression between ZEB1 and members of the miR-200 family promotes

    EMT and invasion in cancer cells. EMBO Rep 9, 582-589.

    Burstyn-Cohen, T., Stanleigh, J., Sela-Donenfeld, D., Kalcheim, C., 2004. Canonical Wnt

    activity regulates trunk neural crest delamination linking BMP/noggin signaling with G1/S

    transition. Development 131, 5327-5339.

    Chang, C.J., Chao, C.H., Xia, W., Yang, J.Y., Xiong, Y., Li, C.W., Yu, W.H., Rehman, S.K.,

    Hsu, J.L., Lee, H.H., Liu, M., Chen, C.T., Yu, D., Hung, M.C., 2011. p53 regulates epithelial-

    mesenchymal transition and stem cell properties through modulating miRNAs. Nat Cell Biol

    13, 317-323.

    Cheng, C.W., Wu, P.E., Yu, J.C., Huang, C.S., Yue, C.T., Wu, C.W., Shen, C.Y., 2001.

    Mechanisms of inactivation of E-cadherin in breast carcinoma: modification of the two-hit

    hypothesis of tumor suppressor gene. Oncogene 20, 3814-3823.

    Cheung, M., Chaboissier, M.-C., Mynett, A., Hirst, E., Schedl, A., Briscoe, J., 2005. The

    transcriptional control of trunk neural crest induction, survival, and delamination. Dev. Cell 8,

    179-192.

    Clay, M.R., Halloran, M.C., 2010. Control of neural crest cell behavior and migration: Insights

    from live imaging. Cell Adh Migr 4, 586-594.

    Clay, M.R., Halloran, M.C., 2014. Cadherin 6 promotes neural crest cell detachment via F-actin

    regulation and influences active Rho distribution during epithelial-to-mesenchymal transition.

    Development 141, 2506-2515.

    Coles, E.G., Taneyhill, L.A., Bronner-Fraser, M., 2007. A critical role for Cadherin6B in

    regulating avian neural crest emigration. Dev Biol 312, 533-544.

    Dady, A., Blavet, C., Duband, J.L., 2012. Timing and kinetics of E- to N-cadherin switch during

    neurulation in the avian embryo. Dev Dyn 241, 1333-1349.

    Daujat, S., Neel, H., Piette, J., 2001. Preferential expression of Mdm2 oncogene during the

    development of neural crest and its derivatives in mouse early embryogenesis. Mech Dev 103,

    163-165.

    David, R., Wedlich, D., 2000. Xenopus cadherin-6 is expressed in the central and peripheral

    nervous system and in neurogenic placodes. Mech Dev 97, 187-190.

    De Craene, B., Berx, G., 2013. Regulatory networks defining EMT during cancer initiation and

    progression. Nat Rev Cancer 13, 97-110.

  • 37

    Du, C., Zhang, C., Hassan, S., Biswas, M.H., Balaji, K.C., 2010. Protein kinase D1 suppresses

    epithelial-to-mesenchymal transition through phosphorylation of snail. Cancer Res 70, 7810-

    7819.

    Duband, J.-L., Volberg, T., Sabanay, I., Thiery, J.P., Geiger, B., 1988. Spatial and temporal

    distribution of the adherens-junction-associated adhesion molecule A-CAM during avian

    embryogenesis. Development 103, 325-344.

    Duband, J.L., 2010. Diversity in the molecular and cellular strategies of epithelium-to-

    mesenchyme transitions: Insights from the neural crest. Cell Adh Migr 4, 458-482.

    Espada, J., Peinado, H., Lopez-Serra, L., Setien, F., Lopez-Serra, P., Portela, A., Renart, J.,

    Carrasco, E., Calvo, M., Juarranz, A., Cano, A., Esteller, M., 2011. Regulation of SNAIL1

    and E-cadherin function by DNMT1 in a DNA methylation-independent context. Nucleic

    Acids Res 39, 9194-9205.

    Ezin, A.M., Fraser, S.E., Bronner-Fraser, M., 2009. Fate map and morphogenesis of presumptive

    neural crest and dorsal neural tube. Dev Biol 330, 221-236.

    Fairchild, C.L., Gammill, L.S., 2013. Tetraspanin18 is a FoxD3-responsive antagonist of cranial

    neural crest epithelial-to-mesenchymal transition that maintains cadherin-6B protein. J Cell

    Sci 126, 1464-1476.

    Ferronha, T., Rabadan, M.A., Gil-Guinon, E., Le Dreau, G., de Torres, C., Marti, E., 2013.

    LMO4 is an essential cofactor in the Snail2-mediated epithelial-to-mesenchymal transition of

    neuroblastoma and neural crest cells. J Neurosci 33, 2773-2783.

    Fleury, V., 2012. Clarifying tetrapod embryogenesis by a dorso-ventral analysis of the tissue

    flows during early stages of chicken development. Bio Systems 109, 460-474.

    Gessert, S., Bugner, V., Tecza, A., Pinker, M., Kuhl, M., 2010. FMR1/FXR1 and the miRNA

    pathway are required for eye and neural crest development. Dev Biol 341, 222-235.

    Groves, A.K., LaBonne, C., 2014. Setting appropriate boundaries: fate, patterning and

    competence at the neural plate border. Dev Biol 389, 2-12.

    Halbleib, J.M., Nelson, W.J., 2006. Cadherins in development: Cell adhesion, sorting, and tissue

    morphogenesis. Gen Dev 20, 3199-3214.

    Hamburger, V., Hamilton, H.L., 1951. A series of normal stages in the development of the chick

    embryo. J Morph 88, 49-92.

    Hatta, K., Takagi, S., Fujisawa, H., Takeichi, M., 1987. Spatial and temporal expression pattern

    of N-cadherin cell adhesion molecules correlated with morphogenetic processes of chicken

    embryos. Dev Biol 120, 215-227.