Olivier Rieppel*
Science
Feb 12 1999: 945-946.Textbooks portray turtles as the most primitive group of egg-laying animals (amniotes) in existence and extol their virtues as model organisms for primitive amniote organization and physiology. In their report published on page 998 of this issue, Hedges and Poling (1) present an exhaustive analysis of turtle relationships on the basis of DNA data. Their results, which support other recent analyses of protein (2) and DNA (3, 4) sequences, indicate that instead of being related to the anapsid root of the reptile evolutionary tree, turtles nest in the tree crown, within Diapsida. These molecular data thus are partially congruent with morphological characters that also support diapsid (5), rather than anapsid (6), turtle relationships. However, the molecular data conflict with paleontological data as to where exactly turtles fit within diapsids. The DNA data also support a highly controversial relationship of the Tuatara, and it will be a challenge not only to paleontologists, as suggested by Hedges and Poling, but also to molecular systematists to resolve these conflicts.
The older, Paleozoic reptiles have a skull in which the region behind the eye socket is completely covered by bone, the anapsid condition (see the figure). In the Diapsida two openings, the upper and lower temporal fossa, develop in the skull, presumably to facilitate muscle fiber attachment (see the figure). The skull of the earliest fossil turtle, Proganochelys from the Upper Triassic of Germany (7), shows a closed temporal region, suggesting that turtles are a surviving branch of Anapsida. Nevertheless, some details of bone configuration in the temporal region of the skull of Proganochelys and other turtles do not match the primitive pattern seen in Paleozoic reptiles (5), and a number of authors (8-10) have proposed that the anapsid turtle skull, where present, developed secondarily. The first comprehensive evaluation of turtle relationships (11) compared the bone and muscle characters of a broad range of extinct and living reptiles, and concluded that turtles are related to a herbivorous group of Paleozoic anapsids, the pareiasaurs. As the anapsid status of turtles became entrenched in textbooks, subsequent analyses of turtle relationships also found that they were related to these Paleozoic reptiles. Most recently, a pareiasaur relationship of turtles was supported by modern cladistic analysis (6).
But broadening the basis of anatomical comparison beyond the Paleozoic once again called into question the anapsid status of turtles (5), especially when Sauropterygia was included in the analysis. The Sauropterygia is a group of secondarily marine reptiles from the Mesozoic, commonly known as plesiosaurs and pliosaurs, which were adapted to a pelagic mode of life in the Jurassic and Cretaceous seas. Early representatives of the group from the Triassic lived in nearshore environments, and in many aspects of their anatomy resembled their terrestrial ancestors more closely than their later descendants. Recent reanalysis (12) of the data set (5) with Sauropterygia included increased support for the position of turtles within Diapsida, but also showed that it is the Sauropterygia that pulls the turtles up into the crown of the reptile tree.
The crown-group diapsids (Sauria) subdivide into two major evolutionary lineages, the archosauromorphs (crocodiles and birds and their fossil relatives such as dinosaurs) and lepidosauromorphs (the Tuatara, lizards, and snakes, and their fossil relatives). Morphological data (5, 12) place the turtles as sister-group of the Sauropterygia, both nested at the base of the lepidosauromorph lineage. This contrasts with all available molecular data, which put turtles on the archosauromorph branch. Although the placement of turtles within Diapsida is the most parsimonious solution on the basis of all data at hand, the statistical support for both lepidosauromorph and archosauromorph affinities of turtles, on the basis of anatomical and molecular data, respectively, may be relatively weak in some cases. This reflects a high degree of independent evolution of morphological similarity (convergence) within anapsid and diapsid reptiles on the one hand, and the impact of a long separate evolution of the turtle branch on molecular characters on the other. However, an as yet unpublished morphological data set (13) places the Sauropterygia at the base of the archosauromorph lineage. The effect of inclusion of turtles into this data set has not yet been explored, but if Sauropterygia exert the same pull as they previously did (5, 11), turtles might end up as archosauromorphs on the basis of morphological characters also.
Some authors have explored diapsid affinities of turtles among extant reptiles only and, by using anatomical and physiological characters, have placed them closer to crocodiles and birds than to the Tuatara (Sphenodon) and squamates (lizards and snakes) (14). Although there are some important similarities among turtles, crocodiles, and birds [such as the secondary subclavian artery (15)], the placement of the turtles on the archosauromorph lineage also raises important questions with respect to other characters that turtles share with the lepidosauromorph clade. One is the complex mesotarsal joint, which necessitates ontogenetic restructuring of the proximal tarsus in turtles, Sphenodon, and lizards (7, 16). Yet, given the current support for diapsid affinities of turtles but the conflicting evidence for their position within diapsids, it is likely that future discussion will shift from whether turtles are diapsids, to where exactly they fit within the Diapsida. One conclusion can safely be reached: Turtles can no longer be blithely considered a model for primitive amniote organization and physiology.
References
Relationships among the major groups of living reptiles. (A) The classical phylogeny based on morphology and the fossil record (1, 2). (B) Maximum likelihood phylogeny of combined sequences from 11 nuclear proteins (1943 amino acids). Scale bar indicates amino acid substitutions per site. (C) Consensus phylogeny of combined sequences from four nuclear protein-coding genes for which sequences of tuatara are available (785 amino acids). For the molecular trees, confidence values (%) supporting the nodes are separated by slash marks and based on the following four methods: interior-branch test, and bootstrap analyses of neighbor-joining, maximum likelihood, and maximum parsimony, respectively. In (B) it was determined that myoglobin had the greatest effect in lowering confidence values; removal of that gene did not change significance of turtle-crocodilian node but raised support for bird-turtle-crocodilian node to 94/90/94/88.