This week's Nature reports on a new ridiculously primitive turtle, Odontochelys semitestacea, described by Li, Wu, Rieppel, Wang, & Zhao (above). It comes hot off the heels of another basal turtle, Chinlechelys tenertesta, and the two offer competing ideas about how the turtle shell originally formed. First, let's talk about Chinlechelys, because I never got around to posting about it. It's an incredibly fragmentary turtle, but interesting in that one of the fragments shows that the carpace (upper shell) is made of osteoderms which fuse to expanded thoracic ribs. In Chinlechelys, that fusion is incomplete. In modern turtles, the dorsal osteoderms fuse onto the expanded thoracic ribs in the embryo. The reason Chinlechelys is important is because it demonstrates two things: (1) The carpace was originally made up of rows of armor plating; and (2) the thoracic ribs expanded independantly of armor fusion. Which event happened first is hard to say, but Chinlechelys shows that the two are separate occurrances.
The picture above, from Chinlechelys' description, shows a hypothetical sequence of events leading to the modern turtle shell. The two animals on the left are speculative. The third one from the left is Proganochelys, and the far right is Chinlechelys.
Well, leave it to evolution to screw up our understanding of the world. Enter Odontochelys, a turtle more primitive than either Chinlechelys or Proganochelys by a few degrees. It has a fully-developed plastron (lower shell) but no carpace. Did you hear that? It doesn't have any carpace armor! However, it does posess expanded thoracic ribs, so it was a wide-bodied animal with. It was also marine and, better yet, had lots of teeth in all regions of the mouth.
So that throws the Chinlechelys' carpace hypothesis out the window. Odontochelys presents the idea that ancestral turtles were shallow marine animals that developed a plastron to protect their bellies from predators, and only evolved the carpace later, but not initially from osteoderms. The skeletal structure (expanded thoracic ribs) was already in place, and probably evolved originally to join with the plastron. The ribs probably expanded for structural support. So while the armored carpace hypothesis isn't completely wrong, Odontochelys shows that more works needs to be done. The authors suggest that dermal osteoderms may have never been present, and that the full carpace evolved through "intramembraneous ossification within the carapacial disk." Or, more simply, the spaces between the expanded ribs, filled with dermal or cartiligionous tissue, may have simply ossified through evolution. No dermal armor necessary!
This is an unbelievably exciting and important find, and what's more, the authors did a phylogenetic analysis of the animal and found it group with sauropterygians! That would make turtles diapsids, a contention I've never really understood. It's still exciting, though!
UPDATE: In the "News & Views" section of the issue, Reisz & Head suggest that Odontochelys descended from carpaced ancestors, but lost the dermal ossifications secondarily after moving into the sea. Many modern turtles do this, like leatherbacks and soft-shelled turtles. I could really go either way here.