Over the past month, I've been explaining a little basic developmental genetics of the fly, describing the genes involved in anterior-posterior patterning in the early embryo. One of my points has been that these genes are universal tools that we find in all metazoans, and that they represent yet more evidence of common descent. Now I'd like to mention a few things that are different about AP patterning in different animals—and we'll start with a significant admission.

Flies are weird.

In some ways, they are an absolutely horrible choice for a model organism to study common mechanisms of development. They are highly derived; that means they have evolved a number of specific, unique traits that are not representative of the ancestral state, or even of the majority of arthropods. Drosophila are specialized for rapid, reliable development, and the basic mechanisms are layered with redundancies and peculiarities that are contingent on unusual properties of fly early development. This is the animal whose early development has been most fully characterized, and really, it's a byzantine tangle of kludged-together gunk, and it takes considerable effort to sort out what's really basal and what is an avant-garde Dipteran innovation.

For example, I started out by discussing bicoid, an early maternal gene that is expressed in a gradient, high at the front end and low at the back end of the embryo. While we vertebrates also have bicoid homologs, we don't use it in the same way. Why? Because the bicoid gradient is expressed in the cytoplasm, and it only works in flies because their early development is not cellular.

In the diagram to the left is a fly egg at different stages of development. At 30 minutes, at the top, is an egg with a single fertilized nucleus, the round dot to the left of center. What the nucleus does is divide repeatedly without cell division—the result is shown at 1 hour, 10 minutes: a yolky egg with many nuclei. The nuclei then migrate to the edge and form a nuclear layer on the periphery, called the syncytial blastoderm. This is where the early maternal genes do their work, in an environment where no membranes interfere with diffusion, where gene products can freely circulate. Eventually, shown at the bottom, membranes will enclose each nucleus and isolate it as a discrete single cell.

One of the consequences of this pattern of development is that molecular information travels freely and rapidly throughout the zygote, and pattern can form very quickly across the length of the animal. It's also particularly efficient for rapid divisions, since the energetically expensive step of building membranes and cleaving the whole mass of yolk can be deferred to later (syncytial cleavage stages have evolved multiple times in different lineages: many insects do it, birds do it, cephalopods do it.)

Another complication: Drosophila belongs to a group of insects known among embryologists as long germ-band insects. Another embryological group are the short (and intermediate) germ-band insects, which include things like beetles and grasshoppers. The difference is that long-germ band insects lay out their entire body plan, from head to tail, nearly all at once. That is, the head of the animal and the tailtip of the animal are developing simultaneously. Short germ-band insects do it progressively. They start out as embryos with just the head and some thoracic segments, and a wave of development proceeds posteriorly, sequentially adding new segments to the back end. The short germ-band is almost certainly primitive. We vertebrates also do the progressive front-to-back development thing, using what is called a somitic clock. Again, flies are the highly derived, specialized group, they are the ones we understand best, and what we'd like to know now is how these novelties evolved, and how segmentation in general evolved.

I've just run across a paper by Peel and Akam (2003) that concisely ties together these different strategies, the all-at-once emergence of the segmental body plan vs. the progressive emergence of segments one-by-one at a growth zone. One key question they raise is whether the ancestral bilaterian was segmented, and how it partitioned it's body plan, and they bring up this phylogeny of segmentation:

Three hypotheses for the evolution of segmentation during bilaterian evolution. (1) A segmented bilaterian common ancestor (red dot) would require segmented body plans to have been lost at least three times, in lineages leading to deuterostome, ecdysozoan and lophotrochozoan phyla. (2) Segmentation could have arisen twice (blue dots), in the lineages leading to protostomes and vertebrates, respectively. This would require segmented body plans to have been lost at least twice in the lineages leading to ecdysozoan and lophotrochozoan phyla. (3) The third hypothesis sees segmentation having arisen independently three times (black dots) in the lineages leading to vertebrates, annelids and arthropods.

The three major groups all have both segmented and unsegmented phyla, which suggests some competing hypotheses: that the segmented phyla evolved their mechanisms independently, or that the common ancestor was segmented and the unsegmented phyla lost these mechanisms independently. The differences between flies and vertebrates, between a segmentation clock and a global gradient, for instance, suggest that the latter could be the case and that it's all an example of convergence on the handy property of building your body from modular segments. The diagram suggests several alternative hypotheses for when segmentation arose, and the article discusses some of the new evidence that suggests that the 'red' hypothesis, that segmentation is a primitive trait of the Urbilateria, might be the correct one (although, reasonably enough, they don't give a definitive answer just yet, and suggest that we need to look at many other groups.)

For those unfamiliar with vertebrate segment formation, it's driven by oscillating waves of gene products, in particular the dynamic expression of deltaC and her-1/her-7, the vertebrate homologs of the insect pair-rule gene, hairy. There is no known comparable delta/hairy oscillator in Drosophila, but get this: spiders do. Spiders build their segmental abdomens, or opisthosoma, from a growth zone, using a delta-1 gene and a hairy homolog that are cycled repeatedly. So far, spider abdominal segmentation looks an awful lot like zebrafish segmentation!

The other interesting observation about spiders is that their anterior segments, the prosoma or appendage-bearing part of the body, appear to use a different and less well characterized mechanism of segmentation, that may be more like what has been described in flies. This raises the possibility that there are two ancestral mechanisms of segmentation in the bilateria, one that is fly-like and used in the anterior regions of short germ-band animals, and one that is clock-like and used in sequential growth zones. Different lineages have just expanded one or the other mechanism to different degrees. The authors model this as a gradual reduction in the role of the segmental clock in the fly lineage, as they switched from a cellular blastoderm to a syncytial blastoderm.

A model for the transition from short germ to long germ modes of segmentation during insect evolution. The number of segments patterned in a growth zone by a putative segmentation clock (blue) decreases as segmentation genes downstream of the clock come under the control of newly recruited gap genes. Coloured blocks represent newly evolved gap gene response elements, similar to those found in the Drosophila hairyand even-skipped genes. It is unclear as to whether the change to patterning in a cell-membrane-free (syncytial) environment would be a prerequisite for this transition, or could have occurred in parallel with it.

In their own words, they've got an evolutionary model of segment formation that can be tested:

A prediction of this model might be that the primary pair-rule genes of Drosophila ancestrally functioned downstream of a clock. This would explain how the complex cis-regulatory elements, controlling striped expression of these genes, evolved—one stripe enhancer may have evolved at a time as the influence of the clock on anterior segments retreated and anterior determinants (the future gap genes) took over. Indeed the primary pair-rule genes even-skipped and runt appear to be expressed in a reiterated pattern in the opisthosoma of the spider.

But of course, there are reasons to be cautious. Work in animals other than Drosophila is in a relatively preliminary state, and there are a number of significant clades that haven't been adequately examined at all.

A clock-like mechanism in arthropod segmentation remains to be demonstrated. Even if one proves to be involved, it would be premature to conclude that Urbilateria was segmented, for Notch-mediated patterning is used in many and diverse biological processes. The similarities between vertebrate somitogenesis and arthropod segmentation might still be analogous, not homologous. To test whether a segmentation clock was an ancestral feature of bilaterians, we must look more closely at the genes and gene interactions involved, and must survey diverse representatives of basally branching clades within the arthropods, annelids and chordates, as well as less obviously segmented phyla, such as the molluscs. Clearly, this will take time. Despite recent advances the debate as to whether Urbiliteria was segmented looks set to continue for some time to come.

Peel A, Akam M (2003) Evolution of segmentation: rolling back the clock. Current Biology 13:708-710.