Patterning the nervous system with Bmp

I'm a little surprised at the convergence of interest in this news report of a conserved mechanism of organizing the nervous system—I've gotten a half-dozen requests to explain what it all means. Is there a rising consciousness about evo-devo issues? What's caused the sudden focus on this one paper?

It doesn't really matter, I suppose. It's an interesting observation about how both arthropods and vertebrates seem to partition regions along the dorso-ventral axis of the nervous system using exactly the same set of molecules, a remarkable degree of similarity that supports the idea of a common origin. Gradients of a molecule called Bmp may be the primitive mechanism for establishing dorso-ventral polarity in animals.

The way the article is phrased, it sets up the results as "contrary to the prevailing view," which is sort of true, and sort of not. It's a novel result that provokes some interesting ideas about animal ancestors, but I also think that views on this matter are in quite a bit of flux…so going contrary to any prevailing view doesn't represent too great a shake-up. Some of the confusion about the importance of the work arises, I think, because there two waves of dorso-ventral specification to consider. Early in development, the dorsal and ventral sides of the embryos of different phyla are established in a similar manner. One side of the embryo expresses a signalling molecule called Bmp (bone morphogenetic protein) in vertebrates or Dpp (decapentaplegic) in Drosophila; the other side expresses Chordin (the homolog in invertebrates is sog, short gastrulation). Bmp suppresses nervous system development, while Chordin permits it, so the side without Bmp forms the nerve cord. The similarities in whole body patterning of both vertebrates and arthropods have been well established.

(In vertebrates, the Bmp side is the ventral side, while in invertebrates, the Bmp side is the dorsal side; we have a dorsal nerve cord, they have a ventral neural plate. We're inverted relative to each other, but don't let that confuse you—we're using the same molecular mechanisms to define the dorsal-ventral axis. I have more discussion of this pattern in hemichordates.)

Later in development, in vertebrates, the neural tube forms from the dorsal sheet of neurectoderm. It rolls up into a tube and comes to lie inside the animal. This is where it gets interesting: the neural tube, which arose from a region where Bmp is low, now expresses Bmp dorsally. Ventrally, it expresses another signalling gene, Shh (sonic hedgehog). Basically, it's re-expressing two competing gradients of molecules, but now within just the neural tube. In the diagram below, Bmp is active in the roofplate (in blue), while Shh (in red) is active in the floorplate, establishing a double gradient.

This double gradient is important: it establishes positional information in the dorsoventral axis of the neural tube that is read by the cells within it, and used to determine identity. Ventrally, for instance, cells where Bmp is low and Shh is high will differentiate into motoneurons, while dorsal cells where Bmp is high and Shh is low will become specific classes of interneurons. These competing gradients are what establishes the layered organization of the spinal cord.

Now that is the situation in vertebrates. What about invertebrates? Do they do the same thing, and re-establish a Bmp gradient in their nervous system that patterns the tissue? That has been a difficult process to assess, because of a difference in the organization of the nervous system. In the diagram below, the vertebrate is on the right, and the neural tube has formed as a separate tube of cells within the animal, and is isolated from the body wall (in gray). It's expressing some key genes—Msx, Gsh, Nkx2.1—in a layered fashion. In Drosophila, on the left, the neural sheet is also expressing homologous genes in a similarly patterned sequence, but notice that it hasn't rolled inward. It's contiguous and continuous with the gray body wall, and basically just forms the insect's ventral floor.

This arrangement makes it difficult to assess whether Bmp is playing a role in patterning the expression of neural tissue. Why? Because there are other genes involved in more general D/V patterning of the whole body that are expressed coincidentally with Bmp/Dpp. In particular, there is a gradient of nuclear localization of a gene called dorsal that is a major regulator of dorso-ventral identity everywhere in the animal; how can we separate the effects of the dorsal gene and the Bmp/Dpp gene?

The work of Mizutani et al. gets clever here. They're asking whether the network of genes that define subsets of neurons in the insect are like the network in vertebrates, so what they need to do is remove the confounding expression of the other D/V genes. To do this, they create hopelessly doomed embryos that have the early D/V signals set to uniform levels. The dorsal gene is uniformly localized everywhere, so that every cell in the ectoderm thinks it should make neural tissue, turning the whole embyo into a uniform tube of neurectoderm. Then, they couple the expression of Bmp to a promoter for stripe 2 of the even-skipped gene: this creates a band of Bmp anteriorly, and a gradient of Bmp that runs longitudinally.

This embryo is totally screwed! It is not going to survive, and the various signals for neural development are all running in the wrong directions. The embryo is really just a simplified testbed for determining how neural cells will respond to the Bmp gradient—even more basically, do they respond to a Bmp gradient at all?

You can look at the paper yourself; it's on PLoS biology, which means that everyone can read it online (isn't that wonderful?) What you'll see in the data section is beautiful color photos of fly embryos that look like easter eggs, where colored markers tag patterns of expression of various genes associated with particular neural types. The key result, though, is that there are bands of differing activity in response to a Bmp gradient, exactly as we see in vertebrates!

Here's their summary diagram, illustrating that homologous genes with similar patterns of interactions are expressed in both the Drosophila neurectoderm and the vertebrate neural plate. Similar genes, similar rules, similar patterns: it sure looks like a conserved evolutionary pathway to specify neural identities.

Simplified summary model indicating the proposed similarities in BMP-mediated patterning of the vertebrate and invertebrate neuroectoderm. Two processes collaborate to establish the pattern of neural identity gene expression in Drosophila and vertebrates: graded BMP signaling preferentially represses expression of ventral neural identity genes (left), which then engage in a chain of ventral-dominant repression wherein more ventral genes prevail in repressing the expression of more dorsal genes (right). The indicated inhibition of Msx1/2 by Pax6 remains hypothetical. Not indicated on this scheme are additional levels of cross-inhibition (e.g., Vnd inhibition of msh, late Ind repression of vnd, and Pax6 repression of Nkx2.2), which are likely to help sharpen and refine the pattern created by the core mechanism of threshold-dependent BMP repression coupled to ventral dominance.

The high degree of correspondence in the neural specification pathways of flies and vertebrates strongly suggests that this is a reflection of a conserved function, that the last common ancestor of both phyla used this mechanism to pattern its nervous system. Other components of the patterning mechanisms, like the dorsal gene and Shh, seem to be late additions that are unique to specific lineages, so the suggestion is that the Bmps are the primitive D/V patterning molecules.

Just to throw a spanner into the works, though, here's a problem: the hemichordates. In my previous article and in another paper in PLoS Biology, the hemichordates are described as a sister phylum to the chordates: they are deuterostomes, and thus more closely related to us than to the arthropods. The complication is that they don't have much of a nervous system at all, but instead have a distributed array of nerve cells that aren't localized. Lowe et al. interpret this to mean that the last common ancestor of chordates and hemichordates, and of arthropods, didn't have a neural tube to pattern!

There are a couple of different ways to resolve these interpretations.

• To fit Mizutani et al.'s explanation, the last common ancestor had a discrete nervous system that was patterned with a Bmp gradient, and the hemichordates secondarily lost theirs. The hemichordate condition is derived.
• To fit Lowe et al.'s ideas, the last common ancestor lacked a nervous system, and chordates and arthropods independently evolved them; both co-opted an existing molecular pathway, the Bmp system, to do the same thing. In this case, the patterning system of the nervous system is convergent. The Bmp system could still be the primitive D/V patterning molecules for the whole organism, it has just been independently reused in the CNS of both phyla.

The way to resolve this is, of course, more comparative data. How are the Bmps used in the Lophotrochozoa? How about urochordates and echinoderms (not that the latter have much of a nervous system to pattern)? Are there developmental modules like the Dpp/vnd/ind/msh pathway present in animals that lack distinct nervous systems? Whatever the answer is, it looks like it will be fun to figure out.

Lowe CJ, Terasaki M, Wu M, Freeman RM, Runft L, Kwan K, Haig S, Aronowicz J, Lander E, Gruber C, Smith M, Kirschner M, Gerhart J (2006) Dorsoventral Patterning in Hemichordates: Insights into Early Chordate Evolution. PLoS Biology 4(9):e291.

Mizutani CM, Meyer N, Roelink H, Bier E (2006) Threshold-Dependent BMP-Mediated Repression: A Model for a Conserved Mechanism That Patterns the Neuroectoderm. PLoS Biology 4(10): e313.