Naturwissenschaften (2001) 88:193–200 DOI 10.1007/s001140100228 REVIEW ARTICLE Rafael Zardoya · Axel Meyer The evolutionary position of turtles revised Published online: 17 May 2001 © Springer-Verlag 2001 Abstract Consensus on the evolutionary position of tur- niches (Pough et al. 1998). Based on the mechanism of tles within the amniote phylogeny has eluded evolution- retraction of their heads into the shell, extant turtles are ary biologists for more than a century. This phylogenetic classified into two main suborders, Cryptodira (turtles problem has remained unsolved partly because turtles that bend their neck vertically) and Pleurodira (side- have such a unique morphology that only few characters necked turtles) (Shaffer et al. 1997). Cryptodires (ap- can be used to link them with any other group of amnio- proximately 200 species) have a worldwide distribution tes. Among the many alternative hypotheses that have (though rare in Australia) whereas pleurodires (approxi- been postulated to explain the origin and phylogenetic mately 50 species) are currently found only in South relationships of turtles, a general agreement among pale- America, Africa and Australia. ontologists emerged in favoring the placement of turtles The exact phylogenetic position of turtles within the as the only living survivors of the anapsid reptiles (those Amniota (reptiles, birds, and mammals) is the subject of that lack temporal fenestrae in the skull). However, re- a lively debate that dates back for over a century (Benton cent morphological and molecular studies have radically 1991). The current classification of amniotes is consis- changed our view of amniote phylogenetic relationships, tent with a single key morphological character, the pres- and evidence is accumulating that supports the diapsid ence and type of temporal fenestration of the skull affinities of turtles. Molecular studies favor archosaurs (Fig. 1) (Günther 1867; Willinston 1917). Extant amnio- (crocodiles and birds) as the living sister group of turtles, tes that show a completely roofed skull (i.e., turtles) are whereas morphological studies support lepidosaurs (tua- tara, lizards, and snakes) as the closest living relatives of turtles. Accepting these hypotheses implies that turtles Fig. 1 Temporal fenestration of the amniote skull. Anapsida: cannot be viewed any longer as primitive reptiles, and the skull of early amniotes that they might have lost the temporal holes in the skull shows complete dermal roofing secondarily rather than never having had them. of the temporal region. Turtles Living turtles (order Testudines) are among the most also have no true temporal openings. Diapsida: the skull striking and strangest tetrapods. Early on in their evolu- of crocodiles and the tuatara tionary history, these reptiles evolved distinctive shells has two temporal windows as effective defense (Burke 1989). This innovation is (shown in black) that accom- probably the key to their evolutionary success and per- modate jaw muscles. Birds, liz- ards, and snakes have modified sistence, and has constrained the evolution of the rest of the primitive diapsid pattern by their morphology (Lee 1996). The general Baüplan of the lost of one or more of the turtles has barely changed since the Triassic (200 MYA), bone arches that defined the yet their carapaces show a great variety of shapes that re- original two openings. Synaps- flect adaptations to terrestrial, freshwater and marine ida: the skull of mammals presents a single lower tempo- ral hole (shown in black) R. Zardoya (✉) Museo Nacional de Ciencias Naturales, José Gutiérrez Abascal 2, 28006 Madrid, Spain e-mail: [email protected] Tel.: +34-91-4111328, Fax: +34-91-5645078 A. Meyer Department of Biology, University of Konstanz, 78457 Konstanz, Germany 194 Fig. 2A–F Alternative hypotheses explaining the phylogenetic Poling 1999; Kumazawa and Nishida 1999; Mannen and position of turtles within living amniotes. A Mammals are the Li 1999; Rieppel and Reisz 1999; Cao et al. 2000). Al- most basal living amniotes. Turtles are the only living representa- tives of anapsid reptiles. Living diapsid reptiles include Lepidosa- ternative hypotheses, based either on morphological or uria (the tuatara, snakes, and lizards) and Archosauria (crocodiles on molecular evidence, have placed turtles as the most and birds) (e.g. Laurin and Reisz 1995; Lee 1997). B Turtles re- basal amniotes (Fig. 2B) (Gaffney 1980), as sister group present the sister group of all other extant amniotes (Gaffney of the Archosauria (crocodiles and birds) (Fig. 2C) 1980). C Turtles have diapsid affinities, and are the sister group of Archosauria (e.g. Kumazawa and Nishida 1999; Zardoya and (Hennig 1983; Platz and Conlon 1997; Zardoya and Meyer 2001). D Turtles have diapsid affinities, and are the sister Meyer 1998, 2001; Kumazawa and Nishida 1999), as group of Lepidosauria (e.g. deBraga and Rieppel 1997). E Turtles sister group of the Lepidosauria (tuatara, lizards, and have archosaurian affinities, and are the sister group of crocodiles snakes) (Fig. 2D) (Rieppel and deBraga 1996; deBraga (Hedges and Poling 1999; Mannen and Li 1999; Cao et al. 2000). and Rieppel 1997), as sister group of crocodiles F The Haematothermia hypothesis: birds are the sister group of mammals (e.g. Gardiner 1993) (Fig. 2E) (Hedges and Poling 1999; Mannen and Li 1999; Cao et al. 2000), or basal to crocodiles and the Haematothermia clade (mammals and birds) (Fig. 2F) called anapsids. Those that have two fenestrae in the (Owen 1866; Gardiner 1982, 1993; Løvtrup 1985). temporal region of the skull (tuatara, lizards, snakes, Living turtles are highly derived reptiles which also crocodiles, and birds) are the diapsids. Finally, living retain numerous amniote ancestral characters. It is diffi- amniotes that possess a single lower temporal hole in cult to estimate with confidence their phylogenetic affin- their skulls (mammals) are the synapsids. According to ities based on morphological data due to a general lack the fossil record, the first branching event in amniote of shared derived characters with other amniotes and due evolution appears to have separated the synapsids from to conflicting phylogenetic signal in different data sets anapsids and diapsids (Carroll 1988; Gauthier et al. (Rieppel and deBraga 1996). Moreover, the major lin- 1988; Laurin and Reisz 1995; Caspers et al. 1996; eages of amniotes (including turtles) show significantly Rieppel and deBraga 1996; Lee 1997; Reisz 1997; Cao different rates of molecular evolution, complicating phy- et al. 1998; Zardoya and Meyer 1998). Within reptiles logenetic inference based on molecular data (Kumazawa (Sauropsida), the anapsid condition shown by turtles is and Nishida 1999). regarded by some as primitive with respect to the more In recent years, new morphological and molecular da- advanced condition of diapsids (Fig. 2A) (e.g. Carroll ta sets bearing on the phylogenetic position of turtles, 1988; Laurin and Reisz 1995; Lee 1997, 2001; Reisz and on amniote relationships, have accumulated. Most 1997; Cao et al. 1998). current phylogenetic information seems to be able to dis- However, the traditional phylogenetic position of tur- tinguish between some of the competing hypotheses and tles as the only surviving representatives of anapsids is provides stunning new insights into our view of this highly controversial (Benton 1991; Rieppel and deBraga long-standing phylogenetic issue. 1996; deBraga and Rieppel 1997; Platz and Conlon 1997; Zardoya and Meyer 1998, 2001; Hedges and 195 Morphological evidence The monophyly of turtles is supported by many derived features such as their shell, strengthened skull, absence of teeth, and the position of the limb girdles inside the rib cage (Gaffney and Meylan 1988). Other undisputed monophyletic groups within the amniotes are: mammals (diagnosed by, e.g., possession of hair and mammary glands; Rowe 1988), squamates, i.e., lizards and snakes (diagnosed by, e.g., their skin and paired copulatory or- gans; Estes et al. 1988), crocodilians (diagnosed by, e.g., their posterior skull bones and elongated wrist bone; Benton and Clark 1988), and birds (diagnosed by, e.g., their feathers and furcula; Ostrom 1976). Furthermore, a close relationship of the tuatara (Sphenodon) and squa- mates to form the monophyletic clade Lepidosauria, and a sister group relationship of crocodilians and birds to Fig. 3 Turtles as anapsid reptiles. Traditional phylogenetic hy- form the monophyletic clade Archosauria are accepted potheses based on morphological and paleontological data place generally among paleontologists. The debate about the Synapsida (mammals and extinct relatives) as the most basal Am- niota. Turtles (Testudines) are considered anapsid reptiles and are position of turtles with respect to mammals, lepidosaurs, grouped either with extinct procolophonids (Laurin and Reisz and archosaurs is still ongoing. 1995) or with extinct pareiasaurs (Lee 1997, 2001) in the Pararep- It is now widely accepted that mammals represent the tilia clade. Living taxa are shown in bold sister group of all other living amniotes (Carroll 1988; Gauthier et al. 1988; Laurin and Reisz 1995; Rieppel and deBraga 1996; Lee 1997; Reisz 1997). However, in the captorhinids are commonly placed basal to diapsids in past, two alternative hypotheses have been postulated. the Eureptilia clade, and turtles are grouped either with Gaffney (1980) proposed that mammals and diapsid rep- procolophonids (Laurin and Reisz 1995) or with pare- tiles (lepidosaurs and archosaurs)