Neurulation in the cranial region – normal and abnormal

Abstract

Cranial neurulation is the embryonic process responsible for formation of the brain primordium. In the mouse embryo, cranial neurulation is a piecemeal process with several initiation sites and two neuropores. Variation in the pattern of cranial neurulation occurs in different mouse strains, and a simpler version of this morphogenetic scheme has been described in human embryos. Exencephaly is more common in females than in males, an unexplained phenomenon seen in both mice and humans. As the cranial neural tube closes, a critical morphogenetic event is the formation of dorsolateral bending points near the neural fold tips, which enables subsequent midline fusion of the neural folds. Many mutant and gene-targeted mouse strains develop cranial neural tube defects, and analysis of the underlying molecular defects identifies several requirements for normal dorsolateral bending. These include a functional actin cytoskeleton, emigration of the cranial neural crest, spatio-temporally regulated apoptosis, and a balance between cell proliferation and the onset of neuronal differentiation. A small number of mouse mutants exhibit craniorachischisis, a combined brain and spine neurulation defect. Recent studies show that disturbance of a single molecular signalling cascade, the planar cell polarity pathway, is implicated in mutants with this defect.

Introduction

Cranial neurulation is an important, early step in brain development. The process begins during gastrulation, when the neuroepithelium is induced to differentiate from the dorsal midline ectoderm. A sequence of molecular interactions involving the influence of fibroblast growth factors (FGFs) and antagonism of bone morphogenetic proteins (BMPs) has been shown to mediate neural induction (Stern, 2002). The neural plate thus formed then undergoes a series of co-ordinated cellular behaviours termed convergent extension, in which cells move medially and intercalate in the midline (Keller, 2002; Copp et al. 2003a). This yields a keyhole-shaped neural plate which is relatively broad rostrally in the future brain, and narrow more caudally in the future spine. As convergent extension progresses, neural folds form at the edges of the neural plate, initially at the boundary between future brain and spinal cord, and subsequently in more rostral and caudal parts of the neural plate (Golden & Chernoff, 1993). In the developing brain, the neural folds elevate by a process that involves midline neuroepithelial bending and expansion of the underlying cranial mesenchyme. This creates convex neural folds the morphology of which is subsequently converted to a concave appearance as a result of dorsolateral bending of the neuroepithelium (Morriss-Kay, 1981). As the tips of the neural folds are brought into apposition in the dorsal midline, epithelial fusion occurs to create the neural tube. Epithelial remodelling of the dorsal midline tissues then creates separate outer surface ectoderm (future epidermis) and inner neuroepithelium (nervous system) layers (Copp et al. 1990).

Although closure of the cranial neural tube is not essential for the subsequent differentiation of specialized neuronal and glial cell types, nor for the formation of nerve connections, it is nevertheless essential for brain development as a whole. This is vividly illustrated by examination of the near-term anencephalic brain that results from failure of cranial neural tube closure in the third week post-fertilization. In humans, the exposed skull base in anencephaly is covered by a vascular tissue described by neuropathologists as the ‘area cerebrovasculosa’, from which cerebrospinal fluid leaks (Harding & Copp, 2002). Very little recognizable brain tissue is present, most having degenerated during the gestational period. The anencephalic mouse brain presents a similar appearance at the end of pregnancy (Fig. 1A). Nevertheless, despite this loss of brain tissue, the cranial nerves emerging from the skull are usually well developed, indicating relatively normal neurogenesis earlier in gestation.

Fig. 1.

Newborn (A) and E15.5 fetuses (B,C) showing the main cranial neural tube defects in mice. (A) Anencephaly in a curly tail mutant. Note the lack of skull vault (between small arrows). The skull base is overlaid by the ‘area cerebrovasculosa’, the remnant of the degenerate brain tissue. Open spina bifida (strictly, myelocele) is present in the lumbosacral region (arrowhead). (B) Exencephaly of the midbrain in a curly tail mutant, showing the everted cranial neural folds (between small arrows). Open spina bifida is also present, affecting the lumbosacral region (arrowhead). (C) Craniorachischisis in a Celsr1 mutant, in which the neural tube is open from midbrain to low spine (between small arrows). Note the presence of a curled tail in all cases (large arrows in A–C). B and C are modified from Copp et al. (2003b).

A very different picture emerges from examination of the malformed brain in the period immediately following failure of neural tube closure. At this early stage, the unfused cranial neural folds are everted, creating the appearance of excessive neural tissue, a feature termed ‘exencephaly’ (Fig. 1B). Strikingly, the exteriorized neural tissue, despite its abnormal anatomical position, is healthy and shows signs of neuronal differentiation and the formation of nerve connections. Hence, brain development proceeds relatively normally in the early stages after failed neural tube closure, but secondary degenerative processes curtail this development, leading ultimately to loss of almost the entire brain by the time of birth (Wood & Smith, 1984; Matsumoto et al. 2002).

A similar sequence of degenerative changes occurs in the spinal region after failure of neurulation at that axial level. Studies in experimental animals have shown that covering the lesion with skin and muscle can halt the degeneration of the spinal cord tissue, which in unoperated cases leads to severe loss of neural tissue at the site of the lesion (Meuli et al. 1995). Clinically, in utero surgery for myelomeningocele (open spina bifida) is practised in a few specialized centres. If affected fetuses are operated before 24weeks of gestation, improved motor and sensory function can be achieved, together with minimization of hydrocephalus and the Chiari type II malformation, in which the hindbrain herniates through the foramen magnum (Bruner et al. 1999; Sutton et al. 1999). By contrast, surgery later in pregnancy, when cord degeneration has already begun, does not improve neurological function (Li et al. 2003). Despite these promising advances in fetal surgery for early spinal defects, it seems highly unlikely that an analogous approach could be taken with exencephaly. Hence, prevention of the potentially lethal condition anencephaly will require earlier embryonic intervention to be successful.

Dependence of skull development on cranial neurulation

Closure of the cranial neural tube is essential not only for maintenance of brain development but also for initial formation of much of the skull. Formation of the skull vault, with contributions from both cranial mesenchyme and cranial neural crest (Jiang et al. 2002), occurs using the embryonic brain as a physical ‘template’ around which to become modelled (Thorogood, 1994). In consequence, the absence of dorsal neural tissue in exencephaly results in failure of formation of the dorsal skull elements. Notably, this defective skull development is a primary failure of skeletal tissue formation, not a secondary degenerative process as in the conversion of exencephaly to anencephaly. In contrast to the lack of the skull vault in anencephaly, the skull base is always present, although invariably malformed (Harding & Copp, 2002). This reflects the presence in early development of ventral neural tube tissue that can serve as a physical model around which the skull base forms, before the onset of neural degeneration later in gestation. Hence, both neural and skeletal development are contingent upon normal closure of the cranial neural tube, although for different reasons.

Rostro-caudal events in mouse cranial neurulation

In the mouse, primary neural tube closure is initiated at the hindbrain/cervical boundary at embryonic day (E) 8.5 (3weeks post-fertilization in humans), when the embryo has 6–7 pairs of somites. After this initial closure event (called Closure 1), neurulation proceeds concurrently in the future brain and spinal regions (Fig. 2). Brain formation comprises two further de novo closure events: at the forebrain/midbrain boundary (Closure 2) and at the extreme rostral end of the forebrain (Closure 3). Bidirectional spread of neurulation between Closures 1 and 2, and between Closures 2 and 3, leads to completion of brain formation at two ‘neuropores’, located in the forebrain (the anterior or rostral neuropore) and in the hindbrain (sometimes also called ‘Closure 4’; Golden & Chernoff, 1993). Spread of neurulation from Closure 1 along the spinal axis culminates in closure of the posterior neuropore, at the 29–30-somite stage (E10), marking the end of primary neurulation.

Fig. 2.

Stages of mouse cranial neurulation, as seen diagrammatically in whole embryos at (A) E8.5 and (B) E9 (with extra-embryonic membranes removed). At E8.5 (five-somite stage) the neural folds are approaching one other at the site of Closure 1 (arrow in A), whereas the cranial neural folds are still wide apart. Ten hours later, at E9 (ten-somite stage), Closure 1 has occurred and neural fold fusion has also been initiated at Closures 2 and 3 (arrows in B). These closure initiation events define two neuropores where cranial closure will later be completed: the anterior or rostral neuropore (ANP) and the hindbrain neuropore (HNP). A third neuropore (the posterior neuropore, PNP) is present in the spinal region. Somites are shown in yellow.

Closures 1 and 3 appear invariant among mouse strains whereas Closure 2 is polymorphic between inbred strains (Juriloff et al. 1991). Although most strains exhibit Closure 2 at the midbrain–forebrain boundary (Fig. 3A), a few exhibit this closure rather caudally, within the midbrain (e.g. DBA/2 strain; Fig. 3B). Other strains undergo Closure 2 rostrally, within the forebrain (e.g. NZW strain; Fig. 3C). Interestingly, strains with a caudally positioned Closure 2 are resistant to exencephaly, whereas those with a rostrally located Closure 2 are highly predisposed. When the splotch (Pax3) mutation was bred onto the DBA/2 background, midbrain closure was enhanced and the frequency of exencephaly among homozygous splotch embryos was reduced from 80% to around 40% (Fleming & Copp, 2000). By contrast, breeding to the NZW strain maintained the high exencephaly rate seen in homozygotes on the splotch parent genetic background, which has a rostral Closure 2 position, as in NZW (Fleming & Copp, 2000). In strains such as NZW, Closures 2 and 3 may be located extremely close together, making the two events difficult to distinguish from each other, and in one strain, SELH, this situation appears to be taken to an extreme with the disappearance of Closure 2 entirely (Fig. 3D). One in five SELH embryos develops exencephaly spontaneously (MacDonald et al. 1989). Thus, variable occurrence of Closure 2 can explain the interstrain variation in frequency of exencephaly that has long been observed in different inbred mouse strains (Austin et al. 1982; Finnell et al. 1986).

Fig. 3.

Variation in the position of Closure 2 (arrowhead) among different inbred mouse strains, as shown on diagrams of the cranial region (viewed from the left side). Open neural folds in the midbrain and hindbrain are shown in red; forebrain neural folds are shown in green. Arrows show the direction of closure between initiation sites.(A) In the majority of strains, Closure 2 occurs at the forebrain/midbrain boundary.(B) One variation is for Closure 2 to occur caudally, within the midbrain, as in the DBA/2 strain. This mode of closure supports midbrain neural fold apposition and counteracts any tendency for these neural folds to remain open, thereby diminishing risk of exencephaly.(C) Another variation is for Closure 2 to occur rostral to the forebrain/midbrain boundary, as in the NZW strain. This type of Closure 2 destabilizes elevation and apposition of the midbrain neural folds, and increases the chance of exencephaly.(D) The most extreme situation, as seen in the SELH/Bc strain, is where Closure 2 is absent altogether. Susceptibility to exencephaly is very high in the absence of Closure 2, with 17% of SELH/Bc mice exhibiting exencephaly. Note that human embryos are suggested to complete cranial neurulation, as shown in D (see also Fig. 4). Abbreviations: fb, forebrain; hb, hindbrain; mb, midbrain. Figure modified from Fleming & Copp (2000).

Controversy in human cranial neurulation

In human embryos, neurulation events have been described that correspond to Closures 1 and 3 in the mouse. That is, the human embryo initiates closure in the future occipital region, with bidirectional closure spreading from this (Closure 1) site as well as, independently, at the extreme anterior end of the body axis (‘adjacent to the chiasmatic plate’; O'Rahilly & Müller, 2002) in an event closely resembling mouse Closure 3. Closure spreads in a caudal direction from this Closure 3 site. The existence of a Closure 2-like event in human embryos is controversial. Some authors claim a similar event to that described in the mouse, based either on direct observation of early human embryos (Nakatsu et al. 2000) or on extrapolation from the pattern of fetal neural tube defects (Van Allen et al. 1993). By contrast, other authors report the lack of a Closure 2-like event from their studies of early human embryos (Sulik et al. 1998; O'Rahilly & Müller, 2002). These latter workers consider that brain formation is achieved by neurulation progressing directly between the Closure 1 and Closure 3 initiation sites, with the closure of a single rostral neuropore.

How can we reconcile these differing views of human cranial neurulation? One possibility is that there are genetic differences between human populations in the position of Closure 2, in close parallel with the variations between mouse strains. This could be a contributory factor to the differences in frequency of neural tube defects observed in different human ethnic groups (Carter, 1974). Another possibility is that the human embryo most closely resembles the situation observed in the SELH mouse strain (Fig. 3D), which has Closures 1 and 3, but not Closure 2 (MacDonald et al. 1989). In this mouse strain, 17% of embryos develop midbrain exencephaly whereas, strikingly, the remaining 83% manage to complete cranial closure despite the absence of Closure 2. This suggests that Closure 2, even in the mouse, is not obligatory for brain formation, and that the majority of embryos that lack this event can still achieve brain closure although there is a strong tendency for midbrain closure to fail. Interestingly, the neurulation-stage human embryo has a proportionately smaller brain than the mouse embryo at a corresponding stage (Fig. 4), raising the possibility that completion of brain closure in the absence of a Closure 2 event is more feasible mechanically in the human embryo than in the mouse. The smaller brain of the neurulation-stage human embryo poses a lesser challenge for cranial neural tube closure, necessitating a less complex pattern of closure that, perhaps in evolutionary development, has enabled the primate lineage to dispense with, or never to develop, Closure 2.

Fig. 4.

Mouse (A) and human (B) embryos soon after the completion of cranial neurulation. Note the relatively smaller midbrain region (mb) in the human embryo compared with the mouse. (A) The mouse embryo (E10) has closed its brain by piecemeal neurulation between three initiation points (1, 2 and 3) with completion of closure at two neuropores (anterior neuropore, ANP; hindbrain neuropore, HNP). (B) The human embryo (35days) seems likely to have completed brain closure by caudally directed spread of neurulation from a rostral initiation site (Closure 3) and rostrally directed spread from Closure 1. The relatively smaller brain may make this simpler pattern of cranial neurulation feasible mechanically in human development. Scale bar: 0.5mm in A; also applies to B. A is modified from Van Straaten & Copp (2001).

Sex and predisposition to exencephaly

It has long been recognized that females are disproportionately represented among human anencephalics (Carter, 1974; Janerich, 1975; James, 1979; Seller, 1987). This distortion of the sex ratio is not seen in myelomeningocele (open spina bifida), in which there is a nearly equal sex ratio, or even a male preponderance. Two main types of hypothesis have been proposed to explain the female prevalence in anencephaly. First, it has been suggested that anencephaly is a more severe condition in males than in females with the result that a larger proportion of affected males are lost as early spontaneous abortions (miscarriages) without coming to detailed clinical examination. According to this idea, the female preponderance among anencephalics is not a reflection of increased susceptibility, but rather of increased survival of affected females compared with affected males. Although this hypothesis cannot be entirely discounted in humans, analysis of mouse strains in which there is a similar female preponderance among exencephalic embryos has provided strong evidence against the ‘differential survival’ hypothesis.

An excess of females with exencephaly has been described in several mutant mouse strains, including curly tail (Embury et al. 1979), cranioschisis (Kalter, 1988), exencephaly (Wallace et al. 1978), splotch (Martin et al. 2003), p53 knockout (Armstrong et al. 1995; Sah et al. 1995) and the SELH strain (MacDonald et al. 1989), as well as in exencephaly induced by hyperthermia (Webster & Edwards, 1984). Analysis of early embryos in these mice shows that females are over-represented among exencephalics from the earliest stage at which this cranial neurulation defect can be recognized. Moreover, when considering all embryos in these mutant litters, female-to-male ratios do not markedly deviate from 1 : 1. Hence, it seems very unlikely that putative affected males could have died earlier from more severe neural tube defects than those exhibited by their female litter mates.

The mouse data suggest that females must be at a higher risk of developing exencephaly than males, prompting the question: why? Another hypothesis to explain this higher risk of faulty cranial neurulation in females invokes a differential rate of growth and developmental progression between embryos of the two sexes (Toriello & Higgins, 1985). This idea was supported by the finding that female mouse embryos are typically smaller and at a slightly earlier developmental stage than male litter mates during the period of neurulation (Seller & Perkins-Cole, 1987). It was suggested that females may develop more slowly than males during the critical window for cranial neural tube closure, thereby spending longer at this ‘vulnerable’ stage, and so proving more susceptible to disturbances that lead to exencephaly. However, a longitudinal study of mouse embryonic growth and developmental progression throughout neurulation showed that embryos of the two sexes in fact progress through the stages of neurulation at identical rates (Brook et al. 1994), although females are slightly retarded compared with males. This finding argues that females do not spend longer in neurulation, although they do begin and complete each neurulation phase a few hours later than their male litter mates. In fact, female embryos become retarded relative to males as early as pre-implantation stages when the presence of a paternally derived X chromosome, at a stage prior to X-inactivation, causes developmental retardation of female embryos (Thornhill & Burgoyne, 1993).

Hence, although the phenomenon of female predisposition to cranial neurulation defects is well recognized, both in humans and in mice, the mechanism of this interesting sex difference remains unknown. A resolution of this issue would be important not only for the insight it will yield into the cranial neurulation process, but also for clues that may emerge to the factors that predispose to, and may protect against, exencephaly and anencephaly.

Diversity among cranial neurulation defects

Cranial neural tube defects take two main forms: (1) exencephaly/anencephaly (Fig. 1A,B), in which the open neural tube defect is confined to the developing brain, and does not extend into the spine (although open low spina bifida may coexist with exencephaly, as a separate defect) and (2) craniorachischisis (Fig. 1C), in which the caudal part of the cranial neural tube (mainly midbrain and hindbrain) is open and continuous with an open neural tube lesion extending throughout the spinal region. Both types are well recognized in humans and in mice, but there are striking differences between the two defects. For example, anencephaly comprises around 50% of all human neural tube defects and is present in the majority of the mouse mutant strains that exhibit neurulation defects. Moreover, a wide range of teratogenic agents can induce exencephaly (Copp et al. 1990). Craniorachischisis, by contrast, makes up no more than 10% of human neural tube defects, is exhibited by only a small number of mutant mouse strains and, to date, has been caused by very few teratogenic agents.

Work in mouse models (Copp et al. 2003b) has demonstrated that exencephaly can result from a wide range of apparently unrelated disturbances of cranial neurulation and is therefore a highly heterogeneous condition in terms of aetiology and developmental mechanism. By contrast, the majority of currently known cases of mouse craniorachischisis have been found to result from disturbance of a single molecular signalling cascade, the planar cell polarity pathway. This non-canonical Wnt signalling system appears to be essential for normal convergent extension during gastrulation, a prerequisite for initiation of neural tube closure (Copp et al. 2003a). Hence, unlike exencephaly, craniorachischisis appears to be a relatively specific developmental malformation, with a narrow range of closely related aetiological factors.

Exencephaly: factors contributing to its pathogenesis

Elevation of the cranial neural folds in mammalian embryos begins with an initial phase in which the cranial mesenchyme undergoes marked expansion, with cell proliferation and a striking increase in extracellular space (Morriss & Solursh, 1978). In the midbrain region, particularly, this phase of cranial neurulation produces bi-convex neural folds that bulge outwards (Fig. 5A). In the second phase of cranial neurulation, the edges of the neural folds form dorsolateral bending points, which ‘flip around’ and approach the dorsal midline, with bi-concave morphology, until the neural fold tips meet in the midline (Fig. 5B; Morriss-Kay, 1981).

Fig. 5.

Appearance of the midbrain neural folds before (A) and after (B) the stage of Closure 2 during mouse cranial neurulation.(A) Initially, the midbrain neural folds are convex, with divergence of the fold apices (small arrows).(B) A few hours later, Closure 2 has been initiated at the forebrain/midbrain boundary (large arrow). Dorsolateral bending points have developed, enabling the midbrain neural folds to reverse their curvature and adopt a bi-concave morphology. This allows the folds apices to approach one other in the dorsal midline (small arrows). Note the open anterior neuropore (arrowhead in B). Scale bar: 0.15mm. Figure modified from Fleming & Copp (2000).

The plentiful supply of mouse mutant and knockout models with exencephaly (Copp et al. 2003b) suggests that many different types of developmental disturbance can lead to this ‘final common pathway’ of faulty cranial neural fold closure. Although exencephaly in the Twist and Cart1 knockouts has been suggested to involve faulty expansion of the cranial neural folds (Chen & Behringer, 1995; Zhao et al. 1996), the phase of cranial neurulation that seems most often disturbed in gene knockouts is the formation of dorsolateral bending points, without which midline neural fusion cannot occur. Figure 6 summarizes some of the factors that appear to play an essential role in this second phase of cranial neurulation.

Fig. 6.

Summary of the principal developmental mechanisms that appear essential for successful closure of the cranial neural tube, with particular reference to the formation of dorsolateral bending points. Examples of mouse mutants in which each mechanism appears to be disrupted are shown in parentheses. Mechanisms are depicted unilaterally for the sake of clarity, but operate bilaterally. See text for explanation. Figure modified from Copp et al. (2003b).

Actin microfilaments

A long-standing idea in neurulation studies is that neural tube closure is largely a consequence of ‘purse-string’ contraction of subapically arranged actin-myosin microfilaments within neuroepithelial cells (Karfunkel, 1974; Sadler et al. 1982). Actin microfilaments are certainly present at this location in the bending neural plate and, most strikingly, cranial neural tube closure can be disturbed by applying drugs such as cytochalasins that disassemble actin microfilaments (Morriss-Kay & Tuckett, 1985; Smedley & Stanisstreet, 1986).

Targeting of genes required for normal cytoskeletal function leads to cranial neural tube defects in several instances. Thus, exencephaly is observed in mice lacking the F-actin-associated protein shroom (Hildebrand & Soriano, 1999), the actin-binding protein vinculin (Xu et al. 1998), the protein kinase C target MARCKS, which cross-links actin filaments (Stumpo et al. 1995), and the RhoGAP p190 regulator of actin cytoskeleton (Brouns et al. 2000). Exencephaly is also seen in embryos doubly mutant for the cytoskeleton-related genes Mena and profilin (Lanier et al. 1999), Mena and VASP (vasodilator-stimulated phosphoprotein; Menzies et al. 2004), and the non-receptor tyrosine kinases Abl and Arg, the function of which is actin-related (Koleske et al. 1998).

It is striking that these knockout mice all exhibit exencephaly, whereas the spinal neural tube closes normally. The only exception is shroom, where a small proportion of null embryos have spina bifida (Hildebrand & Soriano, 1999). This is consistent with findings from the use of cytochalasins in chick, mouse and rat embryos: cranial neural tube closure is exquisitely sensitive to microfilament disruption whereas spinal neurulation is relatively resistant (Morriss-Kay & Tuckett, 1985; Ybot-Gonzalez & Copp, 1999). Hence, a microfilament-based mechanism is crucial for closure of the cranial neural tube, but is not obligatory for spinal neurulation.

Future experimental studies are required to determine whether changes in the contractile state of actin microfilaments actually ‘drive’ dorsolateral bending of the neural folds. Alternatively, cytoskeletal rearrangements could perform a permissive function, stabilizing morphogenetic movements caused primarily by other cellular events.

Cranial neural crest emigration

Neural crest migration and neural tube closure are closely related both spatially and temporally, but the relationship is different in the developing brain and spine of mammalian embryos. In the midbrain and hindbrain, neural crest cells begin to detach from the apices of the neural folds and start migration well in advance of neural tube closure (Morriss-Kay & Tan, 1987). By contrast, in the spinal region, neural crest emigration does not begin until several hours after neural tube closure is complete (Erickson & Weston, 1983). It seems most likely therefore that neural crest emigration may play a key role in regulating mammalian cranial neurulation.

Mice with mutations or knockouts of the Cited2, Pax3 (splotch) and Twist genes all exhibit cranial neural crest defects and exencephaly (Copp et al. 2003b). A similar phenotype is seen in mice over-expressing connexin 43 (Ewart et al. 1997), whereas rat embryos treated in culture with chondroitinase ABC, to digest pre-existing chondroitin sulphate, exhibit coincident delay in neural crest emigration and dorsolateral bending of the cranial neural folds (Morriss-Kay & Tuckett, 1989). These findings suggest that neural crest emigration may be necessary for the cranial neural tube to complete closure. On the other hand, the converse is certainly not true: cranial neural tube closure is not required for normal migration of the neural crest, as evidenced by the many exencephalic mutants in which neural crest migration is normal. It should also be noted that in other species, for example the chick, cranial neural tube closure is completed prior to neural crest emigration (Lawson & England, 1998), although inhibition of neural crest migration can disturb brain closure even in the chick (Nieto et al. 1994).

Precisely how emigration of the cranial neural crest might enable dorsolateral bending of the neural fold tips is unclear. A reduction in cell density at this location, as the neural crest cells depart, could play a permissive role by increasing the mechanical flexibility of the dorsolateral neuroepithelium. Conversely, abnormal retention of the presumptive neural crest cells would ‘stiffen’ the neuroepithelium and prevent bending. Experimental studies to test these cause-and-effect ideas are yet to be reported.

Neuroepithelial apoptosis

Dying cells have long been observed within the neuroepithelium during neurulation (Schluter, 1973), and recent studies demonstrate that this cell death is apoptotic (Lawson et al. 1999). Several knockout mouse strains exhibit neural tube closure defects associated with alterations in the extent of neuroepithelial cell death, with many of these knockouts involving apoptosis-related genes. The Apaf-1, caspase 9 and p53 knockouts, and the Ikk1/Ikk2 and Jnk1/Jnk2 double mutants, all exhibit reduced apoptotic cell death in association with the development of exencephaly. By contrast, exencephalic embryos of the AP-2, ApoB, bcl10, mdm4, mTR, Tcof1 and Tulp1 knockouts show an increase in the number of apoptotic cells in the cranial neural folds or neural tube. Hence, it appears equally detrimental for the intensity of apoptosis to be decreased or increased. It is also noticeable that only cranial neurulation is disturbed in these mutants, with the single exception of fog, a spontaneous mutation of Apaf-1, in which spinal closure defects have also been described (Harris et al. 1997).

Apoptotic cells occur mainly at two sites in the cranial neural plate: at the sites of dorsolateral bending of the neural plate, and at the tips of the fusing neural folds (Fig. 6). The finding that both increased and reduced apoptosis can lead to exencephaly suggests that the intensity of cell death at these locations needs to be kept within strict limits for neural tube closure to proceed normally. Dorsolateral cell death may synergize with neural crest cell emigration, to ‘loosen’ the cranial neuroepithelium, thereby allowing the conversion from a biconvex to biconcave neural fold morphology, whereas apoptosis at the tips of the neural folds appears to play a quite different role. After the apposing neural folds have contacted and adhered, midline epithelial remodelling is needed to break the continuity between neuroepithelium and surface ectoderm on each side, enabling separate neuroepithelial and ectodermal continuity across the midline. Inhibition of apoptosis using the peptide Zvad-fmk produced neural tube defects in the chick embryo, probably by preventing this dorsal midline remodelling (Weil et al. 1997). Detailed experimental studies in which apoptosis is increased or reduced at specific locations in the neural plate are needed to determine precisely how apoptosis participates in cranial neurulation.

Balance between cell proliferation and differentiation

The neural plate is entirely proliferative during neurulation, with cells beginning to exit the cell cycle only following the completion of neurulation. Once cells have become post-mitotic, they embark upon the cellular pathway towards neuronal differentiation. In knockouts affecting the Notch and related signalling pathways, failure of cranial neurulation appears to result from the premature onset of neuronal differentiation in the neural plate, prior to neural tube closure. Inactivation of RBP-Jκ, Hes1 and Numb (Zhong et al. 2000; Ishibashi et al. 1995; Oka et al. 1995) all prevent cranial neural tube closure by de-repressing neuronal differentiation, leading to the precocious appearance of markers of neuronal differentiation during neurulation. A more severe phenotype is seen in double mutants for Hes1 and Hes3 (Hirata et al. 2001). Premature differentiation within the neuroepithelium could render the neural plate mechanically inflexible, preventing dorsolateral bending, or it could interfere with the release of neural crest cells, or inhibit the adhesion process necessary for neural fold fusion. Alternatively, because Notch signalling is required to maintain proliferation in the neuroepithelium, it is possible that a diminution in neural plate cell number is responsible for the development of exencephaly in these mutants. Further studies are needed to distinguish between these and other mechanisms of neural tube defects, caused by disruption of the Notch pathway.

Excessive cell proliferation has also been implicated in some mouse mutants with exencephaly. Over-expression of Notch3 produces increased numbers of nestin-positive progenitor cells in association with failure of cranial neurulation (Lardelli et al. 1996). Moreover, the allelic fog and Apaf1 mutants exhibit increased BrdU labelling, in addition to reduced cell death (Honarpour et al. 2001). Homozygotes have ectopic masses of forebrain tissue protruding from the forehead owing to local absence of the skull vault, a defect closely resembling the human condition of frontal encephalocele. Neuroepithelial ‘overgrowth’ is sometimes described in relation to the exencephalic brains of other mutants (Fig. 1B), but this is rarely more than a morphological impression and, where analysed in detail, a normal or even lengthened cell cycle has been observed in exencephalic brains (Wilson, 1974, 1980; Wilson & Center, 1974).

Which cellular processes drive cranial neurulation?

We have seen that disturbance of several quite different cellular mechanisms (e.g. actin cytoskeleton, apoptosis, neural crest emigration) can produce exencephaly, but in each case it seems most likely that there is loss of a permissive influence rather than abolition of the ‘driving force’ of cranial neurulation. In the spinal region, bending at specific hinge points occurs through intrinsic cellular changes within the neuroepithelium that are co-ordinated by signals from adjacent non-neural tissues. Specifically, midline bending is induced by signals from the notochord, whereas dorsolateral bending is induced by contact with the surface ectoderm on the outside of the neural fold (Smith & Schoenwolf, 1989; Davidson et al. 1999; Ybot-Gonzalez et al. 2002). Sonic hedgehog (Shh) emanating from the notochord regulates the pattern of bending by inhibiting dorsolateral bending at upper spinal levels but not in the low spine, where notochordal Shh production is weak during neurulation (Ybot-Gonzalez et al. 2002). An analogous regulatory mechanism may operate in the cranial region, perhaps underlying the formation of dorsolateral bending points, although this idea has not yet been tested in experimental studies.

Craniorachischisis – role of convergent extension cell movements

This most severe type of neural tube defect develops in mice homozygous for the loop-tail (Lp), circletail (Crc), crash (Crsh), protein tyrosine kinase 7 (ptk7) and Dvl1/2 double mutants (Copp et al. 2003a; Lu et al. 2004). The defect arises from failure of Closure 1 and, although the neural folds elevate apparently normally at the hindbrain/cervical boundary, the folds are spaced widely apart and cannot appose in the dorsal midline, preventing closure (Greene et al. 1998). This failure to initiate neural tube closure precludes closure at lower levels of the spinal axis, and also in the hindbrain and parts of the midbrain. By contrast, Closures 2 and 3 occur apparently normally, leading to relatively normal formation of the forebrain and anterior midbrain.

Recent attention has focused on the shaping of the neural plate, which occurs by the process of cellular convergent extension (Keller et al. 2000). A polarized rearrangement of cells occurs within the plane of the neural plate and in the underlying mesoderm with a net cell movement in a medial direction, and intercalation of cells in the midline. This process lengthens the body axis while narrowing the medio-lateral dimension, converting the initially elliptical embryonic disc into the typical narrow, elongated embryonic shape. Experimental disturbance of convergent extension in Xenopus yields a relatively short, broad neural plate that fails, either totally or partially, to complete neural tube closure (Wallingford & Harland, 2002). The Lp and Crc mutations also exhibit a short, broad neural plate, suggesting that a defect in convergent extension may underlie the neural tube defects in these mutants. Significantly, the genes mutated in the Lp, Crc, Crsh and Dvl1/2 mutants (Vangl2, Scrb1, Celsr1 and Dishevelled) are all homologues of Drosophila genes that are involved in establishing cellular polarity, particularly via the so-called ‘planar cell polarity’ (PCP) pathway (Murdoch et al. 2001, 2003; Curtin et al. 2003). Examination of the inner ear of the mutants has shown that Vangl2, Scrb1, Celsr1 and ptk7 are indeed required for the establishment of PCP in the mammalian embryo (Curtin et al. 2003; Montcouquioi et al. 2003; Lu et al. 2004). Moreover, disruption of homologous PCP genes in Xenopus and zebrafish yields defects of convergent extension (Park & Moon, 2001; Wallingford & Harland, 2001; Darken et al. 2002; Goto & Keller, 2002; Jessen et al. 2002). Hence, the PCP signalling pathway is implicated in the early shaping of the neural plate, a process that appears vital for the correct initiation of neural tube closure.

It remains to be determined whether convergent extension also plays a necessary role in later events of mammalian neurulation. Coexistence of a hypomorphic allele of the node-expressed gene cbl together with the Lp mutation causes embryos of genotype Lp/Lp; cbl/cbl to exhibit both exencephaly and craniorachischisis (Carroll et al. 2003). That is, the cranial defect of craniorachischisis extends further rostrally to affect also the forebrain in these double mutants. This contrasts with ‘simple’ craniorachischisis as seen in Lp/Lp embryos, and a complete lack of neural tube defects in cbl/cbl embryos. Hence, the intrinsic defect of convergent extension in Lp may also affect cranial neurulation events subsequent to failure of Closure 1.

Conclusions

Cranial neurulation is a fundamental event of brain and head development. It is a complex ‘piecemeal’ morphogenetic process, with neural tube closure occurring bidirectionally at several sites simultaneously in the embryo. Considerable variation is seen in the pattern of morphogenesis between inbred strains, and between mice and humans. Disturbance of cranial neurulation can occur as a result of failure of the initial closure event at the brain/spine boundary, yielding craniorachischisis. The other main cranial neural tube defect, exencephaly, results from failure of one or more of the events of cranial neurulation. Midbrain closure, which appears to be particularly unfavourable mechanically, is most frequently disturbed. There is a strong female excess in exencephaly, although the reason for this is unknown. In recent years, the characterization of a great many gene knockouts in the mouse has uncovered requirements for several cellular events in cranial neurulation. These include actin microfilament function, precisely co-ordinated cell death, cell proliferation and cytodifferentiation, and emigration of the neural crest. Hence, exencephaly is a heterogeneous condition that can result from diverse embryonic disturbances. By contrast, craniorachischisis appears to be a relatively homogeneous condition in which a single molecular cascade, the planar cell polarity pathway, has been implicated. This review has identified a number of areas of cranial neurulation in which key studies are required to elucidate further the underlying developmental mechanisms.