We have the brains of worms

Way back in the early 19th century, Geoffroy St. Hilaire argued for a radical idea, that vertebrates and most invertebrates were inverted copies of each other. Vertebrates have a dorsal nerve cord and ventral heart, while an insect has a ventral nerve cord and dorsal heart. Could it be that there was a common plan, and that one difference is simply that one is upside down relative to the other? It was an interesting idea, but it didn’t hold up at the time; critics could just enumerate the multitude of differences observable between arthropods and vertebrates and drown out an apparent similarity in a flood of documented differences. Picking out a few superficial similarities and proposing that something just looks like it ought to be so is not a persuasive argument in science.

Something has changed in the almost 200 years since Geoffroy made his suggestion, though: there has been a new flood of molecular data that shows that Geoffroy was right. We’re finding that all animals seem to use the same early molecular signals to define the orientation of the body axis, and that the dorsal-ventral axis is defined by a molecule in the Bmp (Bone Morphogenetic Protein) family. In vertebrates, Bmp is high in concentration along the ventral side of the embryo, opposite the developing nervous system. In arthropods, Bmp (the homolog in insects is called decapentaplegic, or dpp) is high on the dorsal side, which is still opposite the nervous system. At this point, the question of whether the dorsal-ventral axis of the vertebrate and invertebrate body plans have a common origin and whether one is inverted relative to the other has been settled, and the answer is yes.

That raises more questions, of course. One is where the nervous system fits into this scheme. Flipping upside down would seem to be a fairly radical change in organization, and it’s difficult to imagine (failures of our imagination are not scientific data, of course, but trying to reconcile conceptual difficulties can lead to new ideas) how inverting everything could be accomplished while still retaining viability. One resolution of that problem that has been proposed is to simplify: maybe the last common ancestor of arthropods and vertebrates was dorso-ventrally ambiguous — it was a worm for which up and down didn’t really matter, making it easy for one branch of its descendants to commit to one orientation, while another branch committed to the reverse orientation. The key to this idea was the suggestion that maybe that last common ancestor didn’t have a central nervous system — it had a distributed nerve net, and only after the vertebrate lineage had separated from the others did the central nervous system (CNS) condense on the side opposite Bmp’s expression. In this case, the CNS of vertebrates and invertebrates would not be homologous; they would have arisen independently. I’ve discussed this idea before, in the context of hemichordate evolution.

One way to resolve the question of whether the CNS is homologous across the metazoa would be to look for more details; the more similarities we can accumulate in the molecular patterning of the nervous system, the less likely they are to be of independent origin. I also reported on some work in Drosophila that showed similarities in how the fly and vertebrate nervous system are patterned — within the CNS, we all use Bmp to generate a gradient that defines zones where particular neuron types differentiate. These data made it increasingly improbable that the metazoan CNS evolved multiple times, but these competing hypotheses of a patterned CNS in the bilaterian last common ancestor vs. no CNS in the LCA and independent origin are still duking it out. I am amused to quote myself; I ended that article on fly CNS patterning with a question:

The way to resolve this is, of course, more comparative data. How are the Bmps used in the Lophotrochozoa?

I am amused because I just got my hands on a paper from the Arendt lab that describes Bmps and CNS patterning in the Lophotrochozoa. If it weren’t such an obvious question to have asked, I’d call myself a prophet.

Here’s an overview of the situation in vertebrates. The neural tube is patterned by a double gradient: Bmp is high dorsally and low ventrally, while Sonic hedgehog (Shh) is high ventrally and low dorsally. Cells in the CNS can read their position in the gradient, and that defines their identity and their properties as neurons. It’s a system that sets up columns of cells in the neural tube with similar functions. For instance, there is a longitudinal column of motor neurons, neurons that will reach out to innervate muscles, near the floor of the tube all along its length.


What the Bmp/Shh double gradient does is activate different transcription factors at different dorsal/ventral levels of the CNS. We can map those out, as diagrammed below.


What you can see is that the gradient of Shh regulates other, downstream transcription factors. Where Shh is high, Nox6.1 and Nox2.2 are active; where Shh is low, Pax7 and Pax6 are active, and so on. Each dorso-ventral layer of the CNS has its own combinatorial code derived from the pattern of transcription factors operating in it.

That map of transcription factor activation sketches out the basic layout of the neural columns in the CNS. It’s fundamental. There are many, many details that get added on top of it, of course, but it really represents a kind of primal framework, the skeleton on which the later, greater elaborations are built.

The paper from the Arendt lab examines these same transcription factors in a polychaete worm, Platynereis (I’ve discussed that worm before, too— it has very cool eyes). There’s a huge amount of very impressive work in that paper — maps of gene expression, experimental manipulation of the Bmp gradient, staining for specific neurotransmitters, etc. — but I’m going to skip it all and jump ahead to the summary diagram. This is amazing. Look at the simpler diagram above; the one below is laid out in roughly the same way, with bars that stretch from the top of the diagram (dorsal in the vertebrate, lateral in the worm) to the bottom (ventral in the vertebrate, medial in the worm), with the domains of expression of various factors indicated by the extent of the bars. The vertebrate pattern is on the right, the worm pattern on the left.

(click for larger image)

Mediolateral Arrangement of
Neurogenic Domains and of Neuron
Types in the Annelid and Vertebrate
Trunk Nervous Systems. The mediolateral extent of the expression of
neural specification genes is represented by
vertical bars. Dashed lines separate neurogenic domains with a distinct combination of
neural specification genes. Colored bars represent the expression regions of neuronal specification genes at predifferentiation stages.
Hatched bars represent genes that are expressed at differentiation stages only.

Whoa. There are obvious differences, but the most striking fact of the two diagrams is their general similarity. Lophotrochozoans and chordates clearly exhibit homologous patterns of organization in their earliest stages. The color coding in this diagram pushes the homology still further; it illustrates the function and transmitter types common in those domains, and they line up fairly well, too.

This next diagram illustrates the position of these transcription factors a little more concretely, in an outline of a cross section of the animal. One thing to keep in mind is the topological arrangement of the neural tube in vertebrates. Our nervous systems fold themselves over and tuck themselves into the body wall in a process called neurulation. The similarities are more obvious if you mentally unfold the vertebrate neural tube, flattening it out into a flat sheet so the red area is at the dorsal midline, and the purple dots are moved out laterally. Then it becomes obvious that the vertebrate is an upside-down copy of Platynereis.

(click for larger image)

(B) Comparison of mediolateral patterning in
Platynereis, vertebrates, Drosophila and enteropneust. The schematic drawings represent
trunk cross-sections of embryos. The medio-lateral extent of expression is shown for nk2.2
(red), pax6 (gray), and msx (cyan). Midline cells
(black), serotonergic neurons (yellow), hb9+
neurons (blue), and ath+ lateral sensory neurons (purple) are indicated as circles.

What the presence of such similar mechanisms of neural tube patterning in vertebrates and polychaete worms suggests is that the last common ancestor of these two deeply divergent lineages, the Urbilaterian, also had a nervous system with a similar columnar organization of its CNS. Drosophila, which has a roughly similar pattern with some simplifications and differences, is the product of fairly extensive evolutionary changes; it’s less primitive in its structure, and more derived. Similarly, other deuterostomes like the enteropneust in the diagram, have also thrown away most of the structure in their nervous systems, and their diffuse, nerve-net style architecture is not primitive either, but derived.

We’re still left with the curious problem of how and why our ancient chordate answer flipped itself upside down and took to swimming about with its former belly upwards. The authors conclude with some speculation based on Dohrn’s annelid theory; perhaps there was a semi-sessile stage in chordate history, where the animal was living partially buried in the substrate, basically lying on its back in the mud, filter feeding. When they made the transition to free-swimming, they simply retained their orientation and made other adaptations to effectively invert themselves, primarily in constructing a new ventral mouth.

Denes AS, Jekely G, Steinmetz PR, Raible F, Snyman H, Prud’homme B, Ferrier DE, Balavoine G, Arendt D (2007) Molecular architecture of annelid nerve cord supports common origin of nervous system centralization in bilateria. Cell 129(2):27


  1. #1 David Marjanovi?
    May 4, 2007

    I still don’t buy it.

    The homologies are undeniable, but the inversion is IMHO not. I thought flatworms had no less than 8 nerve cords, dorsal and ventral and lateral? Why can’t this be the ancestral condition? Has anyone checked if flatworms express BMP/DPP between their nerve cords?

    (The mollusks would then have retained the ventral half of this arrangement: they have 4 nerve cords, all ventral. Hey! I mentioned cephalopods! :o) )

    Also, RogerQ, if you have your head up your blastopore, that means you have a quite horrible case of spina bifida. In AFAIK all chordates, the caudal neuropore (hopefully) closes above the blastopore; we have both a stomodaeum (newly formed mouth) and a proctodaeum — like the arthropods.

    In arthropods, gastrulation happens on the ventral side and produces a midgut that is blind at both ends; then the blastopore closes while the midgut connects at the ends to invaginations from the ectoderm — the foregut and the hindgut are ectodermal and even lined with cuticle.

    Pardon the pun but this may shed some light on the ‘problem’ with the vertebrate eye- blood supply between the retina and the incoming light source. WE then would be the group that ‘flipped’ and, the ecdysozoan bauplan might be closer to the pleisiomorphic condition for animal optical systems.

    Camera eyes are clearly not homologous between vertebrates and cephalopods — the other mollusks and the other deuterostomes lack them. The comparison you need to make is AFAIK between the eyes and the pineal/parietal organs of vertebrates.

  2. #2 David Marjanovi?
    May 6, 2007

    Hemichordates also lack the tidy discrete CNS.

    My point was that the flatworms have a neat, discrete CNS, with 8 cords instead of 4 or 2 or 1, so that we are probably looking at independent simplifications from that. Am I wrong about that?