Neurulation is defined as a process of neural tube closure. Recent reports suggested that upon completion of this process the major factors of neurulation remain in force at least until the central canal of the neural tube is formed. Hence, an idea has been put forward to define the two periods of neurulation: early neurulation corresponds to the period of neural tube closure and late neurulation corresponds to the period of formation of the central canal. These ideas are discussed in a context of neural tube defects that may affect late neurulation and result in distention of the central canal.
Birth defects are the main cause of child death prior to 1 year of age (Petrini et al. 2002). Neural tube defects (NTD) are among the most common inherited defects, although in different countries these numbers vary (0.2–10 per 1000 live births) depending on specific geographic location (Copp et al. 2003). Most NTDs stem from abnormal neurulation. This is why the two types of neurulation – primary, which takes place in the anterior neural tube of terrestrial vertebrates, and secondary, which is characteristic for the posterior neural tube of terrestrial vertebrates and the whole neural tube of teleosts, have been intensely studied in various animal and human models (reviewed in Schoenwolf 1984; Dettlaff et al. 1993; Strahle & Blader 1994; Colas & Schoenwolf 2001; O'rahilly & Muller 2002; Copp et al. 2003; Geldmacher-Voss et al. 2003; Lowery & Sive 2004; Solnica-Krezel 2006; Padmanabhan 2006; Wada & Okamoto 2009; Obladen 2011; Wallingford et al. 2013). Usually the NTD statistics take into consideration the most obvious cases, including anencephaly, craniorachischisis and spina bifida. Many genes are linked to these conditions in mice and their number is on the rise (Harris & Juriloff 2010). As to the less obvious abnormalities of the neural tube characterized by distention of the central canal of the spinal cord, often defined as syringomyelia and hydromyelia, the nature of underlying molecular defects is much less clear. Yet these defects also affect quality of life, albeit to a lesser degree. Prior to the introduction of magnetic resonance imaging (MRI), most of these cases in humans remained undetected. Some of these may represent the weak phenotypes of known NTDs and analysis of the molecular determinants involved may help to better understand the molecular mechanisms of neurulation.
Factors of neurulation
It has been discussed that organogenesis and neurulation in particular depend on generic biophysical determinants of form acting in epithelial rudiments, such as cell-adhesion-generated tissue surface tensions, gravitational effects, viscosity and elasticity (Newman & Comper 1990; Foty et al. 1996; Forgacs et al. 1998; Dias et al. 2014; reviewed by Beloussov & Lakirev 1991; Colas & Schoenwolf 2001; Kondo 2014). At the cellular level neurulation depends on a combination of internal and external factors that shape the neural plate. Here, external pressure exerted by cell proliferation in mesoderm acts below and lateral to the neural plate. In this context, the attempts to connect neurulation and deficiency of proliferation of mesodermal cells and mesoderm morphogenesis are of interest. Malformation Chiari type 1 (CM1) is a hereditary NTD characterized by distention of central canal typical of syringomyelia (Gardner & Goodall 1950; Barry et al. 1957; Milhorat 2000; Sarnat 2008). Recently a connection between CM1 and 58 genes involved in formation of paraxial mesoderm has been analyzed and mutations in three of these genes (CDX1, FLT1 and ALDH1A2) were linked to this human malformation (Urbizu et al. 2013). In particular, ALDH1A2 is involved in metabolism of retinoic acid (RA), which plays a role in convergence-extension (CE) during gastrulation (Kam et al. 2013). CDX1 interacts with PК (Lengerke et al. 2011); FLT1 regulates cell polarity and migration by activating signal transduction pathways of PI3K, MAPK/ERK, ACT1 (Wang et al. 2011).
Similarly important are the internal factors acting within the neural plate, including the apical constriction, cell proliferation, and axonal buildup (Bohme 1988; Copp et al. 2003). The apical constriction driven by the sub-apical belt of F-actin (reviewed in Sawyer et al. 2010) results in an excess of apical membrane. An excessive membrane may hamper further constriction, so to avoid that in Xenopus development it is removed by endocytosis (Lee & Harland 2010). All these internal factors remain in action even after neural tube closure/formation in zebrafish and rodents (Li et al. 2009; Kondrychyn et al. 2013). A number of molecular factors involved in apical constriction, cell proliferation and axonal guidance were characterized in model animals, including Shroom3, Abl/Arg, Rap1, Rho kinase (Rock), Mena/Vasp, MARCKS, myosin IIB, Mid, which expectedly linked neurulation with microtubule- and actin-based cytoskeleton (Koleske et al. 1998; Hildebrand & Soriano 1999; Haigo et al. 2003; Menzies et al. 2004; Hildebrand 2005; Kinoshita et al. 2008; Nishimura & Takeichi 2008; Suzuki et al. 2012; reviewed in Sawyer et al. 2010). Some of these factors remain in action during conversion of the primitive lumen into the central canal, which takes place in parallel with stretching morphogenesis of the two morphogenetic centers of the neural tube – the roof and floor plates (Sevc et al. 2009; Kondrychyn et al. 2013). Since the factors of neurulation remain in operation at a point of neural tube closure, this suggests that neurulation per se is not over at this point either.
As a result of rapid and extensive initial stretching that parallels a reduction of the primitive lumen in zebrafish, the roof plate rapidly extends along 2/3 of the spinal cord diameter during a period of 52–60 h post fertilization (hpf). During later development, the roof plate slowly extends even more. In contrast, similar to a narrow central canal, the floor plate remains relatively short, i.e. 1/6 of the spinal cord diameter, which is comparable to central canal (Figs 1, 2).
What is the importance of asymmetrical extension of the midline structures of the spinal cord? The cross-sections of the human spinal cord reveal the characteristic pattern of distribution of the grey and white matter. The grey matter (i.e., neuronal bodies) is shaped as a butterfly with a thin gray commissure in the middle. Even more interesting is a distribution of the white matter (i.e., axonal tracks and glia), which surrounds the grey matter. Noteworthy, dorsally the white matter is represented by a large block of tissue, including the dorsal median septum (i.e., roof plate). In contrast the ventral white matter is split into two halves by the long ventral median fissure. As a result, the gray commissure ventrally is covered by a relatively thin white commissure (i.e., floor plate; see for example Overney 2003). A comparison of analysis of stretching morphogenesis of the roof plate (in absence of extension of the floor plate) and neuroanatomy of the adult spinal cord allows formulation of the first hypothesis: an asymmetric stretching of the midline structures as a result of conversion of the primitive lumen into the central canal determines the anatomical organization of the adult spinal cord characterized by the butterfly-shaped grey matter and asymmetry of the dorsal and ventral white matter.
What are the molecular determinants of asymmetrical extension of the midline structures in the spinal cord? These define a difference in mechanical properties of these cells, which in turn depends on properties of cytoskeleton and membrane. Collagen plays an important role in mechanical properties of the cell membrane. In vertebrates, 47 mammalian genes encode collagen α chains and more than 30 genes encode collagen-like molecules containing collagenous domains (i.e., Gly-X-Y repeats), but are named because of their other domains or functions (reviewed in Fox 2008). In zebrafish, several of the collagen genes are expressed relatively ubiquitously and some others are tissue-specific. Perhaps, the latter group may be involved in establishing a difference in mechanical properties of the midline structures. In the floor plate, expression of at least six such genes encoding α subunits of collagen was detected (col2α1a, col8α1a, col9α2, col11α1a, col14α1a, col27α1a; Zebrafish Information Network, ZFIN; Mudumana 2003 and unpublished). col2α1a, col11α1a, col14α1a, col27α1a encode structural collagens known for their ability to assemble into elongated fibers and add tensile strength to tissue (Fox 2008 and references therein). In contrast, in the dorsal spinal cord only two collagen-encoding genes (col15α1b and collagen-like col7αl1) were expressed (Zebrafish Information Network, ZFIN). Not only these do not encode structural collagens, it is not clear whether these genes are expressed in the roof plate or in other lineages of the dorsal spinal cord. Hence, at least as our knowledge about expression of structural collagens in the midline structures of the spinal cord is concerned and, in particular, expression of genes encoding structural collagens in the floor plate, the mechanical properties of this structure could be superior compared to that of the roof plate. This may explain an asymmetric stretching of the roof and floor plates in the spinal cord. This situation changes drastically more anteriorly, where the roof plate is either converted into other structures (e.g., choroid plexus) or is relatively short (midbrain), whereas the floor plate becomes very extended (Fig. 1; Kondrychyn et al. 2013). The main factor causing this extension could be the much more significant buildup of axonal pathways in the ventral aspect of the anterior neural tube comparing to that in the spinal cord.
The experimental manipulation of genes expressed in the roof plate of zebrafish causes a break of a contact of the roof plate cells with the central canal resulting in dissociation and deformation of these structures (Kondrychyn et al. 2013; Fig. 2). This may have several consequences, including, but not limited to: (i) breakdown of a transfer of morphogenetic signals (BMP, Wnt) from the roof plate that neurodifferentiation depends upon; (ii) change in a trajectory and diameter of the central canal resulting in abnormal circulation of cerebrospinal fluid (CSF) and neurodegeneration disease (Johanson et al. 2008); and (iii) pressure onto axonal pathways causing abnormal transmission of sensory and motor inputs.
Endpoint of neurulation
A simple definition of neurulation by Karfunkel (1974) suggests that neurulation is completed when the neural tube is formed. Hence, to establish when the neural tube is formed is of importance. There are two time points that need to be considered critically in this context. The first one is represented by the moment of fusion of the neural folds, which in species with secondary neurulation corresponds to cavitation of the neural tube. The second one is reached, when a primitive lumen is converted into a central canal. It is important to note that upon closure of the neural tube the factors of neurulation remain in place. This is illustrated by continuation of the process of F-actin/intermediate filament-dependent apical constriction, neural proliferation, build-up of axonal tracks that drove it (Bohme 1988; Copp et al. 2003; Sawyer et al. 2010). These factors drive conversion of the primitive lumen into the central canal in parallel with continuation of morphogenesis of the roof and floor plates (Snow et al. 1990; Sevc et al. 2009; Kondrychyn et al. 2013). Coincidentally, in the zebrafish a beginning of the roof plate extension correlates with hatching, which manifests an end of embryogenesis, whereas completion of morphogenesis of the roof plate and central canal extends into the early postnatal period in teleosts and mammals both (Li et al. 2009; Kondrychyn et al. 2013). Hence, it seems reasonable to suggest that the endpoint of neurulation is better represented not by a closure of the neural tube, when the processes associated with neurulation still operate, but by formation of a central canal, when an active period of morphogenesis of the neural tube is completed and a definitive neuroanatomical organization of midline structures is established.
Thus, neurulation could be divided into two main periods: early neurulation, which is taking place prior to closure/cavitation of the neural tube, and late neurulation, which is taking place after closure/cavitation of the neural tube. During late neurulation the roof plate cells undergo an active phase of stretching morphogenesis (Fig. 2). After the central canal is formed and late neurulation is completed the stretching of the roof plate cells slows down significantly. The slow phase of roof plate stretching continues as long as the organism grows and the neural tube increases in diameter. Whereas in mammals this process ends by adulthood, in fish, the body grows continuously so the neural tube may grow too.
Neural differentiation in view of the roof plate extension
The Wnt-β-catenin canonical pathway plays an important role in patterning the dorso-ventral axis of the neural tube of vertebrates. Based on analysis of interaction of downstream components of the Shh and Wnt signaling it was postulated that the Gli3-independent Wnt signaling specifies the dorsal aspects of the neural tube, whereas both Gli3-dependent and -independent mechanisms mediate Wnt action at intermediate and ventral levels. It was also suggested that Wnts and Bmps acting as a team control regulators of transcription such as proneural homeodomain and basic helix-loop-helix proteins (Yu et al. 2008; Ulloa & Marti 2010). Despite this progress a comprehensive understanding of mechanisms by which Wnts pattern the dorsal neural tube has not been achieved.
Whereas in the dorsal neural tube Wnts may spread by diffusion, they are known to diffuse only at a short distance, since unlike some other morphogens the hydrophobic Wnts do not diffuse efficiently (Logan & Nusse 2004; Alexandre et al. 2014). Hence, to reach the ventral regions the dorsal-derived Wnts need to be transported by some additional means, which are not fully understood. This leaves a model explaining Wnt activity in the intermediate and ventral neural tube incomplete. It has been suggested that in Drosophila the long-distance transfer of Wnt is due to specialized cell extensions (cytonemes, Roy et al. 2011), or transcytosis along basolateral membranes (Gallet et al. 2008; Strigini & Cohen 2000). Highly polarized cells such as those in the roof plate often act as a source of morphogens. Given an extension of the roof plate cells along 2/3 of the spinal cord diameter and their attachment to the central canal in zebrafish (Kondrychyn et al. 2013), these processes may deliver the dorsal signals such as dorsal Wnts into the ventral neural tube. The roof plate expresses several Zic genes known to regulate many downstream targets, including several genes of the Wnt signaling pathway. Genes implicated in activity of the non-canonical Wnt signaling may act in a cell-autonomous manner during polarization of the roof plate. In zebrafish, Zic3 is known to downregulate several proneural bHLH genes probably to prevent differentiation of the roof plate cells and maintain these as the signaling glia (Winata et al. 2013; Winata et al., in prep). Since Zic3 has been shown to play a role in maintenance of stem cells in mouse (Lim et al. 2007, 2010), it may contribute in the development of the ventral neural tube by maintaining the progenitor niche. Importantly, the ventral extent of the roof plate processes in zebrafish keeps an apical identity and is closely associated with a cluster of cells expressing a stem cell marker (Fig. 2; Kondrychyn et al. 2013).
It seems that the long-distance Wnt signaling represents just one of the flavors of this morphogen. In the embryo there are in-built transcriptionally regulated molecular systems that prevent Wnt spread. For example, it has been shown that the secreted Frizzled–related proteins enhance the diffusion of Wnt ligands and expand their signaling range (Mii & Taira 2009). Quite unexpectedly it was found that Zic3 negatively regulates sfrp1a in the roof plate (Winata et al. 2013). Perhaps, it could be a mechanism to restrict a spread of Wnt signaling to a vicinity of the roof plate. Given the well-known role of Wnts as oncogenes, the means of regulation of Zic3 and its targets could be explored further in a search for anti-cancer therapy. Hence the second hypothesis: despite some inherent limitations of the short distance Wnt signaling, the long roof plate cell processes provide an efficient way to deliver Wnts into various distant locations within a neural tube.
Pathology of neurulation in model animals
An analysis of abnormal early development of the neural tube led to emergence of two theories explaining causes of the NTD. One of these theories dates back even prior to description of the neurulation. Morgagni suggested that the closed neural tube might be reopened due to an excess of pressure of cerebrospinal fluid (Morgagni 1769). Consequently, a description of neurulation by Köelliker and His (Kölliker 1861; His 1874), an idea of arrest of closure of the neural tube, has been proposed (von Recklinghausen 1886; reviewed in Padmanabhan 1984; van Allen et al. 1993). The same ideas were then used to explain the developmental time course of closure of neural tube in various species, including humans. To date, more than 250 genes causing neural tube defects (NTD) were identified in mice (Harris & Juriloff 2010) although not all of these were linked to NTD in human in illustration of species-specific variation in the molecular program regulating neurulation.
The caudal region of mammalian neural tube, where secondary neurulation takes place, is prone to spina bifida. The neurulation events in this area were studied using chick and mice, but the underlying mechanism is still understood insufficiently. In particular, the genetic control of this process needs to be studied further (Gofflot et al. 1997; Muller & O'rahilly 2004). In this context zebrafish could be used as a model of neurulation in the caudal region of mammalian neural tube. Mutations of genes involved in several evolutionary conserved signaling pathways (example, non-canonical Wnt signaling or ciliogenesis) were linked to defects of neurulation in zebrafish (reviewed in Wada & Okamoto 2009; Wallingford et al. 2013). In zebrafish an abnormality of neurulation has been linked with mutations affecting Na/K-ATPase (atp1a1), N-cadherin, Mypt-1/Arnt2, Claudin5a, Fibronectin, Pard6γb, Zic genes, etc. (Elsen et al. 2008; Munson et al. 2008; Nyholm et al. 2009; Gutzman & Sive 2010). Coincidentally, a distension of the central canal has been caused by experimental manipulation of another Zic gene expression (Zic6) and chemical inhibition of Rho-kinase (ROCK; Kondrychyn et al. 2013).
Pathology of central canal
In human, an increased frequency of abnormal development of the neural tube has been linked during pregnancy with certain diseases, such as obesity and diabetes. When the number of such patients grows (Padmanabhan 2006), an increase in the number of NTDs could be expected also. The same effect could be caused by valproic acid or trichostatin taken by epilepsy patients during pregnancy. These drugs regulate histone deacetylases and cause epigenetic changes of chromatin (reviewed in Wallingford et al. 2013). Recently, a reduction in severity and frequency of NTD was noted in newborns compared to that during the early twentieth century (R.J. Berry, personal communication, 2013). It might be brought about by general dietary improvement, including folate fortification of breads and cereals in the USA, Canada, Australia, etc. (Copp et al. 2003) or by systematic prenatal screening of fetuses in Europe. In particular, an exposure of the general public to a folate-enriched diet is not without controversy mainly due to the fact that folate stimulates cell proliferation and its potential long-term side effects are far from being well understood (Wallingford et al. 2013). In parallel, such treatment could eventually lead to an increase of a pool of harmful mutations in the population.
The health status of fetuses rescued from NTD by folate that reached adulthood remains unknown. Folate deficiency in mice causes epigenetic changes up to five generations (Padmanabhan et al. 2013). Folate stimulates cell proliferation (Copp et al. 2003) that may have a long-lasting effect. Indeed, a recent study demonstrating that folate therapy started prior to 8 weeks before pregnancy increases frequency of pre-term delivery (Sengpiel et al. 2013).
Severe abnormality of early neurulation results in well-defined and easily detectable morphology. In contrast, milder pathology could be much less conspicuous and may contribute to a relatively common phenomenon of distention of the central canal. Such distention with dissociated pain and thermal sensory impairment has been termed interchangeably as “hydromyelia” or “syringomyelia”. Without going into a history of this terminological confusion, for the purposes of the current review, it is enough to mention several attempts to define these conditions more strictly (Jinkins & Sener 1999; Milhorat 2000; Roser et al. 2010). In particular, the latter report defines syringomyelia as “malformations of the cranio-cervical junction, for example, Chiari malformation, tethered cord, spina bifida occulta or aperta, subarachnoid cysts, intraspinal tumors, severe scoliosis, spinal trauma in the medical history, status post meningitis, any previous spinal surgery (ventral fusions, dorsal instrumentations) as well as previous peridural anesthesia”. Syringomyelia-associated distention of the central canal is relatively extensive, that is, up to 10 segments in length (Zabbarova et al. 2010). In contrast, “the hydromyelia is predominantly centrally located in the thoracic spine, extending over 3–5 segments and of filiforme shape. No neurological deficits come along with the patients, who present mainly because of diffuse pain, different from neuropathic pain in dissociative syndrome”. Hence, hydromyelia represents a much more benign form of distention of the central canal. It has been suggested that it might be a predisposition for syringomyelia (Roser et al. 2010). In the absence of reliable statistics on frequency of hydromyelia, it is causality linking these two conditions that may be used to provide a rough estimate of hydromyelia in the population. One of the mysteries in the development of syringomyelia is why only 5% of patients with minor spinal cord trauma develop this condition. Could this fraction reflect frequency of hydromyelia in the general population? It seems that this riddle will need to wait until MRI can provide enough data to verify such estimates. At least for now the mechanism of formation of cavities in the spinal cord remains unclear. It has been proposed that these could be hereditary and acquired in origin (Zabbarova et al. 2010).
Hydromyelia is not limited to humans. Due to an increased availability of MRI in veterinary practice, a similar distention of the central canal has been shown in several domestic species – horse, camel, cat and dog (Sponseller et al. 2011 and references therein), where due to an increased prevalence in certain breeds, such as the Cavalier King Charles Spaniel, syringomyelia now is a relatively common neurological diagnosis (Rusbridge et al. 2006). It was detected even in more exotic species such as Reeves' Muntjac (Dutton et al. 2002).
Perhaps, morphological manifestations of defects of late neurulation caused by modulation of activity of Zic6 and ROCK in experiments performed on zebrafish could be considered in view of detailed characterization of hydromyelia and siringomyelia. In particular, an inhibition of activity of Zic6 and non-muscle myosin (Myh6) causes a relatively insignificant distention of the central canal reminiscent of hydromyelia. In contrast, an inhibition of ROCK results in much more significant defects, affects extended areas of the central canal and judging by leakage of the tracer to the surrounding tissues, causes syrinxes (Kondrychyn et al. 2013; Sin et al., unpublished data, 2013–2014). Hence, based upon morphological examination this condition may represent syringomyelia. Hence, a third hypothesis: at a molecular level hydromyelia and syringomyelia could develop due to defects of a mechanism that regulates cell adhesion and cytoskeleton, which play an important role in apical constriction and could be at the core of much more severe forms of NTD.
It is possible that the mild forms of abnormality of the neural tube, such as hydromyelia are caused by relatively benign disturbances of the general determinants of the molecular mechanism of neurulation and/or molecular factors acting during late neurulation. These could be derived for reasons other than those that affect early neurulation and/or reflect relatively weak and regulatory mutations in genes responsible for early neurulation (Patterson et al. 2009).
Perhaps, as the folate therapy will be introduced much more broadly and an overall health status in the population will continue to improve due to a better balanced diet and availability of microelements and vitamins, the benign forms of abnormality of the neural tube will continue to replace the severe forms of NTD. As this trend will develop further, a need for better understanding of an outcome of the folate treatment and whether it results in an increase in frequency of benign forms of abnormality of the neural tube may become more urgent, resulting in a need to study the long-term outcome at a population level.
This work was supported by a grant of the Agency for Science, Technology and Research of Singapore. The author is thankful to Dr Igor Kondrychyn for his contribution into Figures.