Neurulation in zebrafish

Blogging on Peer-Reviewed Research

Neurulation is a series of cell movements and shape changes, inductive interactions, and changes in gene expression that partitions tissues into a discrete neural tube. It is one of those early and significant morphogenetic events that define an important tissue, in this case the nervous system, and it’s also an event that can easily go wrong, producing relatively common birth defects like holoprosencephaly and spina bifida. Neurulation has been a somewhat messy phenomenon for comparative embryology, too, because there are not only subtle differences between different vertebrate lineages in precisely how they segregate the neural tissue, but there are also differences along the rostrocaudal axis of an individual organism. A recent review by Lowery and Sive, though, tidies up the confusion and pulls disparate stories together.

Anyone who has been exposed to a little developmental biology knows the basics of neurulation: you take a flat sheet of cells and roll it up into a cylinder. The sheet is the early embryonic ectoderm, and the cylinder that results is the neural tube with its central lumen, or opening. This isn’t the only way to do it, though. The neural tube can also form by the condensation of loosely scattered cells, the mesenchyme, diagramed below. Rolling up a sheet of cells is called primary neurulation, while building a new neural epithelium from mesenchyme is called secondary neurulation. Both can occur in the same animal. Secondary neurulation is common in the tail, which develops in a rostrocaudal sequence without the convenience of initial organization into simple sheets.

(click for larger image)

Comparison of primary and secondary neurulation. (a) Primary neurulation involves columnarization of an existing epithelium, and then rolling or folding the epithelium (blue). (b) secondary neurulation is characterized by condensation of mesenchyme (brown) to form a rod, which then undergoes an epithelial transition to form the neural tube.

The ready interchange of epithelium and mesenchyme is something we also see elsewhere, in gastrulation: some organisms form mesoderm by delamination and migration of mesenchyme, others by the more coherent ingression of epithelial sheets. The developmental program that gets cells from one place and one tissue to another seems to work well enough whether the individual cells are holding hands in adherent sheets, or moving independently.

Another subtle distinction that gets at my particular interests in teleost (fish) development is precisely how the neural tube folds up. There are again regional and species differences in the process, and for a long time teleosts were thought to be oddballs that didn’t do primary neurulation at all—fish nervous systems were thought to form by a thickening of the neural plate ectoderm and secondary cavitation to form the lumen, and they even got a special name, the neural keel. Closer examination has revealed that that is not entirely correct. Fish neural tubes do form by a medial migration of neural plate cells that is just like the medial movements of primary neurulation…it’s just that cells along the midline cling tightly together, making a nearly invisible seam rather than a long open ditch.

(click for larger image)

Variations of primary neurulation. The neural tube is shown at the open stage (top row) and after initial closure (bottom row) (see also Fig. 1). The initial flat neuroepithelium may roll smoothly into a tube (a), bend sharply at one (b) or more (c) hinge points, or form a solid rod of cells (d).
Zebrafish trunk neurulation. An initial epithelium columnarizes to form the neural plate, which then forms a solid neural keel and solid tube. The midline of the tube becomes distinct and a lumen opens from ventral to dorsal. Neural plate cells lineage labeled prior to neural keel formation (red and green) maintain their relative positions during neural tube formation, indicating that cells form the keel by cryptic rolling or folding. The derivation of the zebrafish neural tube from an epithelium and the cell movements involved are typical of primary neurulation.

Lowery and Sive make a special effort to endear themselves to me by describing the development of one species in particular, my favorite experimental animal, the zebrafish Danio rerio. Much work on this particular problem has been done in recent years on the zebrafish (including some by my graduate advisor, Chuck Kimmel) using lineage tracer dyes, time-lapse imaging, and confocal microscopy. There are even a few confocal movies of neural tube dynamics online at Mark Cooper’s lab if you want to see a bit of what I’m talking about.

The diagram to the left illustrates different stages in the formation of the zebrafish neural tube. It first forms by a thickening, or columnarization, of the ectodermal epithelium to form a classic neural plate. The structure thickens yet further into a neural keel by a movement of cells in the sheet towards the midline; the only real difference between this and what goes on in a chick or a mouse is that the apical ends of the cells are adherent and closure forms a tight seam rather than throwing up folds around a groove…but they are topologically identical.

There are a few differences. Some teleost cells can cross the midline during neurulation (shown by the red cell lineage in the diagram), suggesting more lability and that the cells of the teleost neural plate may have some mesenchymal character. As I’ve said, and as Lowery and Sive say, epithelia and mesenchyme are part of a continuum, and are not entirely incompatible cell states.

After neural keel formation, the nervous tissue forms a rod-like structure and separates from the surrounding ectodermal epithelium. One of the factors that we know is involved in this process is the expression of different adhesive cell surface molecules in the ectodermal epithelium and the neural tube—for instance, N-cadherin is expressed in the neural cells, and E-cadherin in the skin. Zebrafish do this as well, and we even have mutants (parachute and glass onion) in N-cadherin that cause neural tube defects.

Speaking of mutants, another purpose of the Lowery and Sive paper is to make a case for the zebrafish as a promising and productive model system for studying neurulation. They’ve already made the case that teleost neural tube formation is entirely comparable to the processes going on in other vertebrates, and it has the additional advantage of being a practical system for doing molecular genetics. The mouse may be an even better system now (but we’re going to catch up!), but mouse embryos are unfortunately imbedded in a meaty bloody mother, and the mouse doesn’t afford the simplicity of access that we can get in an externally developing embryo like the fish.

Here’s a handy short list of mutants that disrupt neurulation that are available right now:

Zebrafish mutants that affect neurulation
Mutant Gene Function/Phenotype
Trilobite vangl2 PCP pathway defects in cell movement
Grumpy lamb1 Adhesion (laminin beta 1) notochord and brain defects
Heart and soul PKCλ Epithelial Polarity retina/neural tube defects
Nagie oko nok Epithelial Polarity (MAGUK) retina/ brain defects
Parachute/ Glass onion n-cad Cell Adhesion neural tube/eye defects
Sleepy lamc1 Adhesion (laminin gamma 1) irregular brain shape
Sonic you shh Hedgehog defects in neural patterning
One eyed pinhead oep Nodal (EGF-CFC member and a co-receptor for nodal) cyclopia, mutation in human holoprosencephaly
Masterblind axin1 Neural specification reduced eye/forebrain
Neckless aldh1a2 Neural specification (retinaldehydrogenase type 2) defects in neural tube and paraxial mesoderm
Somitabun smad5 Neural specification expansion of neuroectoderm
Lock jaw/ Mont blanc tfap2a Neural crest development

The paper is actually a platform for a big proposal to use zebrafish to search for more molecules involved in neurulation.

A crucial goal is to perform large scale genetic screens to isolate a large set of neurulation mutants. This could be achieved by chemical or insertional screening. Genetic screens in the zebrafish are feasible for two reasons. First is the set of attributes that make zebrafish a good genetic model: short generation time, large clutch size and inexpensive breeding. Second is the ease with which fish embryos can be viewed: the neural plate is visible in the unmanipulated embryo, or where a neurectoderm-specific promoter drives a lineage marker such as GFP. These attributes are not found in mammalian embryos, and while amphibian embryos are easy to view, the current lack of genetic techniques is a deficit. A second goal is to test the set of genes implicated in mammalian or amphibian neurulation, by defining zebrafish mutants in these genes. While gene ‘knockout’ by homologous recombination in ES cells is not yet available for the zebrafish, PCR-based screening for mutations in specific genes is possible and promising. Molecular inhibition of gene function by injection of antisense morpholino-based oligonucleotides into zebrafish embryos has also proven effective. A third goal that is a logical outcome of the others is to examine interactions between genes, using epistasis analysis and other methods. For example, double mutants can be constructed by standard genetic crosses, or by a combination of molecular and genetic techniques, where antisense oligonucleotides are injected into embryos mutant for a single gene. Definition of a a large, perhaps comprehensive, set of neurulation mutants would be a huge step for the field.

Hey, I think it’s a great idea, and I’m all for it.

Lowery LA, Sive H (2004) Strategies of vertebrate neurulation and a re-evaluation of teleost neural tube formation. Mechanisms of Development 12:1189-1197.


  1. Bill Anderson says

    So how do your critics who say that there’s no science on Pharyngula happen to miss posts like this? Maybe they don’t recognize science when they see it.

  2. says

    OK, this might be a relevant place to raise this question. Derek James just reminded me of an interesting point, expressed thusly by Ken Stanley:

    How is it possible for 30,000 human genes to encode the connectivity of a 100 trillion-connection brain? And how is it possible for evolution to search a space of such astronomical complexity to discover such a structure?

    My thoughts:

    1. This means that development has gotta matter. Genes do not code for positions of neurons in 3D space, but rather patterns of growth and migration, which are in turn influenced by all the other stuff happening at the same time.

    2. Quoting Cosma Shalizi,

    Biological development, whether of the brain or the body, is intricately but flexibly and responsively regulated by genes, and in fact a large part of the genome exists simply to control gene expression and metabolic activity. These mechanisms are very good at effectively producing highly specific functional structures (e.g., ribs, arteries), especially when those take a nice, cleanly recursive form (e.g., ribs, arteries), though they do this by directly producing self-organizing mechanisms sensitive to positional information. Moreover it is important to recognize that the degree to which brains are plastic is almost entirely a separate question from the degree to which some of their aspects are innate, just as, e.g., muscular development due to exercise, or wound healing, is separate from normal developmental formation. One thing [Gary] Marcus rightly emphasizes is that one expects, on general biological grounds, that a single gene will participate in the development of many different structures, and that new functions will be effected through re-using old components. So it is not at all surprising if more recently evolved functions (e.g. language) have a similar architecture and genetic basis to older ones (e.g. control of rapid sequential motion). Similarly, one need not expect modular functions to be implemented by distinct brain regions. (There are interesting parallels here to Soviet neuropsychology, of the Luria-Vygotsky type.)

    3. Because of the point made in #2 about “architecture” and recursion, the evolutionary process is not searching the full design space imagined by Stanley.

  3. Inquisitive says

    This may be an ill informed question, but what is that apparently 4 celled purple structure beneath the developing neural tube in diagrams 1 and 3 (the notochord?)? It’s present on almost all the stages of development except the 1st, 2nd, and 3rd, of the secondary neurulation steps (apparently being formed by a cluster of mesenchymal cells).

  4. Steven Sullivan says

    Yep, that’s the notochord — a structure of mesodermal origin.
    Notochord/neural tube interaction is a fascinating dev bio subject in its own right.

  5. BlueIndependent says

    @ #1:

    Well some of the critics don’t say there’s no science, only that the science topics are trafficked far less than those smacking down creationism. I pointed out that far fewer posters are capable of speaking intelligently to the scientific topics than to the social ones. Nevertheless, I shall try myself, here.

    So that I understand what structures these processes are referring to, these neurulation sequences define the creation of vital CNS components (e.g.: a spinal cord), or nerves themselves?

  6. shiftlessbum says

    A note to anyone looking this citation up; PZ made a typo. The citation is; Lowery LA, Sive H (2004) Strategies of vertebrate neurulation and a re-evaluation of teleost neural tube formation. Mechanisms of Development 10:1189-1197.

    I usually look to the issue first, then the page(s). But most people who look up the volumn and then the page(s) would not have a problem. I only noticed it because of the way I do it.


  7. Turner says

    I worked in Hazel Sive’s lab as an undergrad – how cool to see her featured here.

  8. David Marjanović, OM says

    I had no idea of secondary neurulation… it complicates a few homology criteria…

  9. David Marjanović, OM says

    I had no idea of secondary neurulation… it complicates a few homology criteria…

  10. Inquisitive says

    Thanks #5, I figured just as much, but wasn’t aware that the notochord can be derived from the mesenchymal cells. Do you know any decent papers or well-explained sources where I could read more about the neural tube/notochord interaction?

  11. Laura Anne Lowery says

    Answer to #6- yes, these neurulation sequences show the formation of the “neural tube”- which will go on to form the brain and spinal cord. The cells which make up the neural tube can be considered “neural stem cells”. Just about all CNS tissue originally derives from this tube- Nerves will grow out from this tissue later in development, but the CNS begins as a tube at the very earliest stages of its development.
    I’m currently finishing up my PhD thesis right now in Hazel Sive’s lab, and my current project is writing a review on the next stages of CNS development in the brain- namely, how does the tube turn into a brain? It probably won’t be published until later this year, but for any people who are interested in early CNS development, it might be interesting. There really isn’t anything published which summarizes it right now.
    Anyway, thanks for your interest in my work! My review has gotten a lot of visibility from being here! So thanks!
    -Laura Anne Lowery, grad student in the Sive Lab