‘Increasing knowledge leads to triumphant loss of clarity’
‘The study of segmentation: that way leads only to madness’
Alfred Romer (1894 – 1973), Director of the Museum of Comparative Zoology and Professor of Biology, Harvard University
Some problems in biology excite such interest as to become symptomatic of a field. This is true, I think (I hope), of all biology, but it is definitely truer of some fields than others. Evolutionary biology is one such field. And segmentation is one such problem. Since the great pre-Haeckelian 19th century comparative embryology tradition, the developmental biology of organisms has been correctly viewed as the prism through which the evolutionary history of animals must be viewed and understood (not ever since though – developmental and evolutionary biology fell out with each other for most of the middle of the twentieth century). For most of that time, it was also viewed as the primary source of evidence for the actual phylogeny of the animals – our shared family tree over the last 580 million years (I use this date perhaps ill-advisedly – the date of the origin of animals is another of those problems about which people are prepared to loose their temper).
The history of animal history
There are fundamentally two ways to start developing as an animal (assuming you are one of the 99+% that are bilaterally symmetrical). Once you have gone from a ball of cells to a hollow ball of cells with a hole through which cells will pass to make your gut, you have two options: you can make the hole (the blastopore) into your anus, or into your mouth. Once achieved, you then have an anterior-posterior axis, which can be organised in two ways: it can be segmented, or not. (Actually, this is not true; lots of animals are ‘pseudosegmented’, but more of that perhaps in future).
As such, zoologists for most of the last 150 years have assumed that having a segmented body axis was a shared derived feature of particular animal groups: (annelid) worms and insects must be closely related because they are both segmented. At least, they must be more closely related to one another than they are to odd-looking, unsegmented things like clams or penis worms. If everyone agrees on something for over a century, it tends to be very difficult to convince people otherwise. Unless you are on the cusp of a revolution…
In 1997, an ingenius pioneering application of DNA data to reconstructing the animal family tree suggested that in fact insects and worms are very distantly related, and that their shared segmental architecture was in fact nothing to do with their position in the phylogenetic tree. This paper1 not only set the stage for the explosion in genomic approaches to addressing phylogeny that have dominated the big journals ever since, but it reignited one of the oldest controversies in biology: how old is segmentation? Is it a (relatively) recent invention in the lineages in which it is found (vertebrates, arthropods and annelid worms), or does it in fact date back to that pioneering worm crawling around in the mud with it’s newly invented bilateral symmetry about 550 million years ago?
Well, thanks to developmental biology (essentially Christine Nusselhein-Vollhard et al. with flies in the 70s and 80s and Olivier Pourquie et al. with vertebrates in the 90s, 00s, and 10s), we know (or perhaps knew) that flies segment using engrailed, wingless and hedgehog, a transcription factor and a couple of secreted signals respectively, but that vertebrates use a curious ‘clock’ of oscillating Notch signalling as they grow that interacts with an FGF- and Wnt- secreting posterior ‘growth zone’; vertebrates, unlike flies, grow from the back. So, pretty different. In 2003, building on this pioneering work in traditional models, Guillaume Balavoine’s group showed that the upto-that-point-largely-ignored annelid worm perhaps forms segments using the fly system: engrailed and Wingless2. So, the ancestor of worms and flies (which incidentally was the ancestor of all animals who turn the hole into the mouth, rather than the anus – the Protostomes) by extension was segmented using engrailed and wingless. Vertebrates, though, are different: segmentation in them is not homologous to the protostome (‘first mouth’) condition of annelids and flies. Case closed. Interesting stuff: nobody was right, segmentation was very old, but that first bilaterally symmetrical worm in the mud (550 million years ago) perhaps wasn’t split into segments.
However, at about the same time Wim Damen et al. published an astonishing study3 showing that the spider uses Notch oscillations to make its abdominal segments, and in a 2008 paper4 showed that it uses Wnt signalling in a posterior growth zone. What? So, worms are like flies, but fish are like spiders. What the hell is going on? Around this time I was actually trying to become a zoologist, and was totally confused about the state of the art. I was in good company. The majority of zoologists threw up their arms and resigned themselves to agreeing with Romer (see above). Those that didn’t, who generally where those actually working on the problem, started to pick holes in the annelid data, which was admittedly the weakest amongst the three segmented taxa. It is important to underline here that this is absolutely through no fault of the annelid investigators. It is bloody hard to work on non-traditional model organisms (my phd was on one), and anyone who does so has my undying respect and admiration, both for scientific reasons and because of how impressed I am by their workrate and endeavour. In any case, I suspect (I don’t know) that most in the field were happy to accept that the ancestral worm-in-the-mud (called ‘Urbilateria’) was segmented, using the vertebrate/spider system of a Notch clock.
I have never met Wim Damen, though one has to admire the intellectual courage it takes to start working on a spider – a lot of people will have thought that it was a daft idea. However, the credit now passes to Evelyn Schwager, who worked with him on the next batch of surprises, and who pleasingly is now working in Romer’s old department at Harvard. ‘Underlings’ (we know who we are) never get the attention or praise that they (we) deserve and so while I don’t know this to be true, at this point I want to emphasise the reaction when Evelyn presented her beautiful data5 at the European Evolutionary Developmental Biology Conference in 2008 in Ghent in Belgium. She showed that in fact, spiders use the gene Hunchback, which is called a ‘gap’ gene in Drosophila because it acts high up in the segmentation hierarchy (above engrailed, wingless and hedgehog), to accomplish segmentation, but only in the thorax. So, remarkably, the thorax is ‘fly/annelid-like’ and the abdomen is ‘vertebrate-like’. All great scientific findings or breakthroughs that I have seen possess that ability to make an audience of peers gasp. Schwager and Damen in their experiments managed to halve the number of thoracic segments in a spider and film it. A room of arthropod experts see, on the screen, a 4-legged spider running around. A 4-legged spider. Cue gasps.
Fast forward a couple of years to 2010, and the worm guys (I know they hate being called that) produced some delightful further data, filling in the gaps in their engrailed–wingless story to include hedgehog signalling6. Coupled with the spider story, it seems that we have solved segmentation, and it went like this:
- Urbilateria evolves segments using the Notch clock.
- This is a BRILLIANT invention and it takes over the world (its descedants comprise, remember, over 99% of all animals).
- Those that turn that the blastopore into a mouth also evolve the hedgehog–engrailed-wingless system for making segmented structures as well (but why? And how? Lots of work to be done…)
- Some of these lineages (possibly most) loose one or both of these ways of segmenting a structure, because there are many ways to make a living as an animal and lineages are just as likely get simpler as to get more complex (Aristotle was wrong about this).
Intrepid chicken (bits)
Except that vertebrates don’t have to do it using the clock, it now seems. In some beautifully old-fashioned (not in the sense that they are out-dated, but that they possess a lot of explanatory power – this is a compliment) ex vivo culture experiments Claudio Stern and colleagues have just upset the apple cart7 (though by this point it is perhaps more accurate to say that after the apple cart was knocked over, and all the apples were stolen, replaced with oranges, that were again stolen after the cart was knocked over again, Stern and colleagues have made us question whether we actually need apple carts in this day and age). They have shown that it is possible to make somites, the segmented, epithelialized blocks of mesodermal tissue that are the basis of vertebrate segmentation without an oscillating clock of Notch signalling.
In the embryo somites are added two at a time (one on either side of the spinal cord) as the presomitic mesoderm, the tissue undergoing the Notch oscillations, undertakes a mesenchymal-to-epithelial transition. This MET happens as a result of the slow removal of the signals (FGFs and Wnts) that derive from the posterior growth zone of the embryo. Since the embryo is growing, this yields a moving wavefront of signalling; the whole thing is termed the ‘clock and wavefront’ model, and was first posited in the 1970s.
However, the new study shows that if you take presumptive mesoderm from the posterior primitive streak (the name of the growth zone in chicks ie the tissue that will become presomitic mesoderm, but hasn’t yet expressed the Notch oscillations), expose it to the BMP inhibitor Noggin for a few hours to generate a dorsal mesoderm (ie. somite) fate, and then implant it back into the yolk of an egg, but away from the embryo, you generate somites. All at once. Upto 15 of them. And crucially, without Notch oscillations, and nowhere near the wavefront. The generated somites even possess the Hox expression appropriate to the time at which they were dissected from the primitive streak, so they carry patterning information too, though they don’t possess the anterior-posterior polarity of normal somites. Still, astonishing stuff.
So, worms are like flies, but fish are like spiders, which are also like flies, but chicks (which are essentially just highly evolved fish) are not even necessarily like chicks. I know what Romer would have said.
1Aguinaldo, A., Turbeville, J., Linford, L., Rivera, M., Garey, J., Raff, R., & Lake, J. (1997). Evidence for a clade of nematodes, arthropods and other moulting animals Nature, 387 (6632), 489-493 DOI: 10.1038/387489a0
2Prud’homme, B., de Rosa, R., Arendt, D., Julien, J., Pajaziti, R., Dorresteijn, A., Adoutte, A., Wittbrodt, J., & Balavoine, G. (2003). Arthropod-like Expression Patterns of engrailed and wingless in the Annelid Platynereis dumerilii Suggest a Role in Segment Formation Current Biology, 13 (21), 1876-1881 DOI: 10.1016/j.cub.2003.10.006
3Stollewerk, A., Schoppmeier, M., & Damen, W. (2003). Involvement of Notch and Delta genes in spider segmentation Nature, 423 (6942), 863-865 DOI: 10.1038/nature01682
4McGregor, A., Pechmann, M., Schwager, E., Feitosa, N., Kruck, S., Aranda, M., & Damen, W. (2008). Wnt8 Is Required for Growth-Zone Establishment and Development of Opisthosomal Segments in a Spider Current Biology, 18 (20), 1619-1623 DOI: 10.1016/j.cub.2008.08.045
5Schwager, E., Pechmann, M., Feitosa, N., McGregor, A., & Damen, W. (2009). hunchback Functions as a Segmentation Gene in the Spider Achaearanea tepidariorum Current Biology, 19 (16), 1333-1340 DOI: 10.1016/j.cub.2009.06.061
6Dray, N., Tessmar-Raible, K., Le Gouar, M., Vibert, L., Christodoulou, F., Schipany, K., Guillou, A., Zantke, J., Snyman, H., Behague, J., Vervoort, M., Arendt, D., & Balavoine, G. (2010). Hedgehog Signaling Regulates Segment Formation in the Annelid Platynereis Science, 329 (5989), 339-342 DOI: 10.1126/science.1188913
7Dias, A., de Almeida, I., Belmonte, J., Glazier, J., & Stern, C. (2014). Somites Without a Clock Science, 343 (6172), 791-795 DOI: 10.1126/science.1247575