Hemichordate evo-devo

Every biology student gets introduced to the chordates with a list of their distinctive characteristics: they have a notochord, a dorsal hollow nerve cord, gill slits, and a post-anal tail. The embryonic stage in which we express all of these features is called the pharyngula stage—it's often also the only stage at which we have them. We terrestrial vertebrates seal off those pharyngeal openings as we develop, while sea squirts throw away their brains as an adult.

The chordate phylum has all four of those traits, but there is another extremely interesting phylum that has some of them, the hemichordates. The hemichordates are marine worms that have gill slits and a stub of a tail. They also have a bundle of nerves in the right place to be a dorsal nerve cord, but the latest analyses suggest that it's not discrete enough to count—they have more of a diffuse nerve net than an actual central nervous system. They don't really have a notochord, but they do have a stiff array of cells in their proboscis that vaguely resembles one. They really are "half a chordate" in that they only partially express characters that are defining elements of the chordate body plan. Of course, they also have a unique body plan of their own, and are quite lovely animals in their own right. They are a sister phylum to the chordates, and the similarities and differences between us tell us something about our last common ancestor, the ur-deuterostome.

Analyzing morphology is one approach, but this is the age of molecular biology, so digging deeper and comparing genes gives us a sharper picture of relationships. This is also the early days of evo-devo, and an even more revealing way to examine related phyla is to look at patterns of gene regulation—how those genes are turned on and off in space and time during the development of the organism—and see how those relate. Gerhart, Lowe, and Kirschner have done just that in hemichordates, and have results that strengthen the affinities between chordates and hemichordates. (By the way, Gerhart and Kirschner also have a new book out, The Plausibility of Life ^{(amzn/b&n/abe/pwll)}, which I'll review as soon as I get the time to finish it.)

So what, exactly, is the hemichordate body plan? The pictures below give a quick anatomy lesson of an enteropneust hemichordate, Saccoglossus kowalevskii.

Saccoglossus kowalevskii, a direct-developing enteropneust hemichordate of the US Atlantic coast. On the left, a juvenile, a week after hatching, two weeks after fertilization of
the egg. Length: 1 mm. On the right, the 'notochord', so-called by Bateson (see text), now called the stomochord, located between the proboscis and
collar. Shown in sagittal section.

The central feature of hemichordate anatomy is the division into three parts. There is an anterior proboscis or prosome, and a structure called the collar or mesosome behind it. Behind that is the metasome or body proper, which also has one or more perforations, or gill slits. This animal is a filter feeder, pushing that proboscis through the much of the sea bottom, scooping it up into the mouth just in front of the collar, and passing the inedible debris back out through the gill slit(s). (Which, obviously, are not functioning as gills or respiratory structures, but as part of a feeding apparatus.)

Look a little deeper in the saggital section to the right, and you can see the internal structure. Each of the parts of the body plan has its own, separate, fluid-filled compartment called a coelom (we chordates have just one). The coelom forms the hydrostatic skeleton of the animal. Extending into the proboscis is a rod of cells, the stomochord, which was initially assumed to be homologous to the notochord. Gerhart et al. have examined the pattern of gene expression here, though, and declared that it is not the hemichordate notochord. Our embryonic notochords and precursor cells express a suite of genes—brachyury, chordin, and nodal, to name few—that are not active in the stomochord. However, the stomochord does express the genes goosecoid and otx that are associated with the chordate prechordal endomesoderm, a structure responsible for patterning the head. Gene expression and location suggest that the stomochord is therefore related to prechordal endomesoderm, and that the animal lacks a notochord homolog.

That brings us to the interesting revelations of this paper. We know that hemichordates have homologs of many familiar chordate genes. The sequences of these genes can be compared, and cladograms made on the basis of the those sequences assembled. The worms fit right in as expected, as a sister group to the chordates and echinoderms.

Phylogeny of hemichordates and chordates, from 18S rDNA comparisons. Note that the chordate ancestor, in addition to the ancestor
of hemichordates and echinoderms, descends from the deuterostome ancestor, which descends from the bilateral ancestor.

The next step, though, is to ask where the genes are active in the embryos. The genes responsible for defining the body plan typically have specific patterns of expression required to carry out their jobs. The Hox genes, for instance, are turned on in a very orderly sequence from front to back in the animal, while other genes responsible for inducing specific structures must be turned on in localized regions. How does the pattern of gene expression compare between chordates and hemichordates?

This is where the story gets really interesting. The authors examined 32 different genes that were selected because we already know that they are important in neural patterning in chordates—these are genes that demarcate specific regions in the developing nervous system. Where are they expressed in hemichordates?

The diagram below illustrates a simplified hemichordate and chordate embryo broken up into 4 broad regions, with the assortment of genes active in each listed on top. They line up! The genes turned on in the hemichordate proboscis correspond to the genes active in the forebrain and anterior structures; mesosome genes are hindbrain genes, and their domains are similarly bounded on the posterior end by the position of the first gill slit; genes in the metasome are like the ones in the body of the chordate, and this is also where notochord-homologous genes are expressed; and tail genes also line up. This is a beautifully unambiguous map!

The anteroposterior map of expression domains for genes important in chordate neural patterning, encoding transcription factors (in black) and
signaling proteins (in red). Note the alignment of the bodies: prosome with ventral forebrain; mesosome and anterior metasome with dorsal
forebrain and midbrain; posterior metasome with hindbrain and spinal cord; and post-anal tails together. The gill slits of both chordates and
hemichordates develop at the same domain intersections. Signaling centers (red bars), which are important in patterning the chordate nervous
system, are similar in hemichordates.

Note too the red bars in both animals above at the anterior and posterior termini, the prosome base, and at the first gill slit. These are key boundaries that correspond to critical signalling centers in development, places where gene activity is particularly important in setting up the organization. These also line up between the two phyla. Same genes, similar order, similar boundary-defining functions.

The anterior-posterior axis lines up well, but what about the dorsal-ventral axis? An idea that has been kicking around for about 200 years is that the anatomical differences between protostomes (arthropods, for instance) and deuterostomes (us) suggest that we are inverted relative to one another. Arthropods have a ventral nerve cord and dorsal heart, we have a dorsal nerve cord and ventral heart, for example. When we look at the patterning genes, we see a similar phenomenon: the dorsal-ventral axis is actually an axis of bmp (or dpp in arthropods) vs. chordin (arthropod equivalent: sog) expression, and in us chordin is active dorsally and bmp is a ventralizer…while in arthropods, it is precisely reversed. The same genes are being used to set up dorsal and ventral signaling centers, but the morphological consequences are inverted between the two lineages.

One question has always been how the last common ancestor of protostomes and deuterostomes was oriented: was it lacking a significant dorsal-ventral axis? If it did have a d-v axis, which side was up, bmp or chordin? In other words, whose ancestor had to flip itself over during evolution?

As a sister phylum to the chordates, it would be interesting to know how the hemichordate dorsal-ventral axis is specified. Here's a surprise: it's organized like an arthropod, inverted relative to us, with bmp dorsally and chordin ventrally.

Anatomy of the dorsoventral axis of hemichordates, and, superimposed, the map of expression domains of genes encoding transcription
factors (black) and signaling proteins (red). Ventral is defined by the location of the mouth. The section crosses the pharynx in the metasome
(mt), but dorsoventral domains have been included from the prosome (pro) and mesosome (ms).

That suggests that the arthropod organization, with bmp a dorsal marker, is the primitive state, and that we are upside down! OK, not really—seriously, don't try to invert yourselves over this—but it means that our wormlike ancestor, for whom dorsal and ventral were much more labile and less significant in its relationship to the world, may have swapped axes at some point. Alternatively, the d-v axis was morphologically ambiguous for all of our phyla ancestrally, and as they specialized, they independently and arbitrarily attached a dorsal and ventral pattern to the bmp-chordin axis.

All of this swapping of axes sounds difficult, but it may not have been. One central feature of the d-v axis now is the location of the nervous system, dorsal or ventral, and the evidence from hemichordates suggests that the last common ancestor of protostomes and deuterostomes may have been brainless; the primitive condition was to have a loose neural net of cells scattered throughout the periphery, with a little bundling of axons in transit. Localizing the nervous system to a discrete cord occurred independently in the two lineages, and occurred after animals had evolved rich somatic patterning mechanisms. The authors suggest that the condensation of the chordate nervous system might have been a consequence of the evolution of a dorsal notochord, which is a powerful source of inductive signals. Our current organization is the result of a cascade of events in evolution centered around the formation of a central notochord, a sequence that correlates well with the notochord's developmental significance.

Gerhart J, Lowe C, Kirschner M (2005) Hemichordates and the origin of chordates. Curr Opin Genet Dev. 15(4):461-7.