It's just a stage. A phylotypic stage. Part II: The flies [The controversy about the existence of the phylotypic stage]( is more than some bickering about whether one blobby, slimy fish-thing looks more like a [Roswell alien]( than another one does. It's about whether the phylotypic stage means something, whether it tells us something important about development and how developmental changes contribute to evolution. To answer such a question, we need more than another set of comparisons of the shape and movements of embryos and their parts. We need a completely different way of looking at the phylotypic stage, to see if something notable is going on under the hood. So vertebrates all look the same at the tailbud stage. What does that mean?

Embryos look the way they do because of the positions and behaviors of the cells that make them up. The cells in an embryo all have the same DNA, and the link between that DNA and those specific cell behaviors is the basic process of gene expression. (This is a [fundamental principle of developmental biology]( And by gene expression, we usually mean the synthesis of messenger RNA under the direction of genes in the DNA. Different cell types express different sets of genes, and the orchestration of the expression of particular genes at particular times is a big part of what makes development happen. When considering the phylotypic stage, then, developmental biologists wondered: is the apparent similarity of embryos at that stage reflected by similarities in gene expression. Or, more specifically, does the hourglass model hold up when we look at gene expression? This was the focus of the two articles in [last Friday's _Nature_]( that inspired the cool cover.

[The first report]( was authored by a collaborative research group headed by [Pavel Tomancak]( at the [Max Planck Institute of Molecular Cell Biology and Genetics in Dresden](; the co-first authors are Alex Kalinka and Karolina Varga, and you can find the paper (and lots more) [at their excellent site]( After a brief review of the hourglass model and rumors of its non-existence, they state their purpose: to address "the extent to which expression divergence underpins the morphological hourglass pattern at the genome-wide level." Their hypothesis, grounded in [evo-devo ideas](, was that gene expression variation would parallel the morphological variation. Put simply, they predicted that gene expression variation would display that same hourglass pattern: lots of variation early and late, with less variation during that curiously conserved phylotypic stage.

The investigators focused their analysis on a set of six species of fruit fly ([the famous genus _Drosophila_]( This gave them two key advantages in their work. First, [the complete genome sequence is known for all six of those species](, and so the team had access to powerful comparative tools in assessing variations in gene expression. Second, the six species [span some 40 million years of evolutionary divergence]( That enabled the researchers to look for variation over a significant evolutionary timescale, and it was a critical aspect of their experimental design. To see why, we need to look briefly at what they actually measured.

The goal was to detect variations in the level of expression of particular genes and to assess how these changes were related both to developmental time (the age or embryonic stage of one species) and to evolutionary time (the evolutionary "distance" between one species and another). So they developed a parameter called the GST, which represents the amount of variation (over time) in expression exhibited by a particular gene. The authors' description of this analysis might be interesting and comprehensible to many readers:

Our measure of temporal divergence is the three-way interaction between genes, species, and time points, the GST values. These values capture the extent to which the temporal dynamics of different species diverge from one another at specific time points given the time-course as a whole. In this sense, GST values provide a measure of the divergence of relationships between time points thereby enabling us to identify periods in the expression profiles where a coherent flow of information from time point to time point is preserved across species. They also allow us to identify periods where there is a relative temporal disconnect between time points allowing different species to modify their expression levels in different ways. GST values achieve this by measuring the extent to which expression in a given species and time point can be explained by lower order effects, and in particular the gene-by-species (GS), and gene-by-time (GT) effects. Thus, whatever cannot be explained by average differences in expression level between species (GS effects) and average differences in expression at a particular time point (GT effect) is apportioned to the GST value.

It's okay if you didn't follow that completely. The point is that the investigators came up with a way to measure, quantitatively, the variation of gene expression over developmental time and evolutionary time, and their experimental system (_Drosophila_ species) gave them what we would call [dynamic range]( They applied their analytical tools to a data set that is dauntingly large today but would have been positively epic just a decade ago: they measured the expression of each of more than 3000 genes in each of the six species at eight 2-hour intervals during development.

First they showed that their parameters faithfully reported on the known evolutionary relationships among the six species. Then they looked at how variation in gene expression is related to developmental time. I hope at this point that you can immediately see the significance of their result, illustrated in the graph below (click to enlarge).


Variation in gene expression (what the authors call 'divergence') is highest early in development. It increases again later in development. And it's lowest in the middle. In fact, gene expression changes among species are the smallest at stage 5 on that graph. Stage 5 is the extended germband stage, long identified as the phylotypic stage in insects (and all arthropods). The red diamonds indicate mean divergence at each stage. Take some time to enjoy the graph, and if nothing else take note that it depicts one-half of an hourglass.

The authors did a lot more: they showed that the hourglass pattern is not just a weird result of overall variation, but is reflected in the expression levels of scores of individual genes. And they looked at what kinds of genes exhibit the hourglass pattern, and what kinds don't. (For example, genes in the so-called [developmental toolkit]( do the hourglass, while genes involved in more specialized work, such as control of the immune system, don't.)

The results are a major indication that the hourglass picture of development with its peculiar phylotypic stage is a biologically informative model, with the potential to guide us to specific evolutionary events and forces, especially those that mold the form of organisms. [Haeckel's famous dictum]( isn't quite right, but he was onto something.


We'll end with some holiday bonuses, both courtesy of Pavel Tomancak. First, Pavel tells me that despite the fact that two articles dealing with these very significant and very old questions -- the existence of the phylotypic stage and the relationship between development and evolutionary history -- were published simultaneously, the two research groups were unaware of each other's work until mere days before publication. I calculate the odds of that to be...oh, never mind. (Pavel provided some further commentary, some of which is quite complimentary to you, our readers. I'll post it separately.) And, have you looked carefully at that interesting cover image? It's a reconstruction of one of Haeckel's drawings, made entirely of images of fly embryos. _Nature_ won't let us see it up close, but Pavel will. Click to enlarge, or visit [Pavel's lab site]( to get the complete experience. Enjoy!

Next up: [the second paper](, looking at gene age (as in evolutionary age) in fish development. \[Cross-posted at [Quintessence of Dust](\]

Kalinka, A., Varga, K., Gerrard, D., Preibisch, S., Corcoran, D., Jarrells, J., Ohler, U., Bergman, C., & Tomancak, P. (2010). Gene expression divergence recapitulates the developmental hourglass model. Nature 468 (7325), 811-814 DOI: 10.1038/nature09634