The turtle shell and the relationship of the shoulder girdle inside or ‘deep’ to the ribcage have puzzled neontologists and developmental biologists for more than a century. Recent developmental and fossil data indicate that the shoulder girdle indeed lies inside the shell, but anterior to the ribcage. Developmental biologists compare this orientation to that found in the model organisms mice and chickens, whose scapula lies laterally on top of the ribcage. We analyse the topological relationship of the shoulder girdle relative to the ribcage within a broader phylogenetic context and determine that the condition found in turtles is also found in amphibians, monotreme mammals and lepidosaurs. A vertical scapula anterior to the thoracic ribcage is therefore inferred to be the basal amniote condition and indicates that the condition found in therian mammals and archosaurs (which includes both developmental model organisms: chickens and mice) is derived and not appropriate for studying the developmental origin of the turtle shell. Instead, among amniotes, either monotreme mammals or lepidosaurs should be used.
The turtle shell is often regarded as an evolutionary innovation (sensu ) and the two most commonly cited features unparalleled among other tetrapods are the turtle shell [2–4] and the relationship of the scapula relative to the dorsal ribs . The turtle scapula is regularly referenced as being uniquely located ‘inside’ the dorsal ribcage [2,5,6], or at least ‘deep’ to the ribcage [7–9]. This topological relationship, however, is confounded by the fact that the scapula is located inside the turtle shell, of which only a portion is formed by the turtle's ribcage, whereas the rest is composed of intramembranously ossified dermis . Indeed, the dorsal ribcage is often equated with the shell [3, p. 3] and the putatively unique relationship between the scapula and the ribcage is used as evidence for turtles appearing to have broken basic rules of the vertebrate body plan . The turtle shell is therefore regarded as an evolutionary innovation: neither arising from conspicuous changes to the reptile body plan [7,8,10–12] nor from a hopeful monster [9,13].
Recent developmental and fossil evidence have shed considerable light on the scapula's relationship relative to the dorsal ribcage. As was initially noted by Ruckes  in 1929 and expanded upon by more recent developmental biologists [2,6–8,15], the turtle scapula does not move posteriorly during development to a position within or deep to the ribcage (contra ), but rather the dorsal ribs have a unique dorsolateral growth trajectory into a more superficial position within the dermis. The ribs later spread apart from one another distally to obtain a fan-like arrangement. The dermis finally encapsulates the ribs together with parts of the shoulder girdle . Of note, the fan-like arrangement is exaggerated in the highly derived softshell turtle Pelodiscus sinensis, which is used as the primary model organism for understanding the development of the turtle shell [2,15–17]. As in other amniotes, the turtle shoulder girdle maintains its position relative to the muscle plate throughout development via a unique muscle folding mechanism, in some cases maintaining ancestral connectivities and in others creating novel ones . However, the entire pectoral girdle remains anterior to the cartilaginous first dorsal rib throughout ontogeny, and only becomes encapsulated within the shell upon intramembranous ossification of the nuchal bone, the dorsal ribs and neural spines of the carapace . Indeed, as has long been noted by skilled anatomists , the scapula remains anterior to the first dorsal rib even in adult turtles. The newly discovered stem turtle with a proto-shell, Odontochelys semitestacea from the early Late Triassic of China, also has a scapula that lies anterior to the ribcage, further supporting the idea that the turtle scapula does not lie within the ribcage [3,20].
While it is clear that the scapula of turtles does not lie within the ribcage, its vertical orientation anterior to the ribcage still appears unique. Kuratani et al.  argue that two factors contribute to the placement of the scapula ‘beneath’ the ribcage: (i) axial arrest of the ribs, and (ii) the inability of the scapula to form a ‘scapular blade’ (i.e. a blade-like process that lies on top of the ribcage). While the axial arrest of the dorsal ribs is undoubtedly a key transition in the origin of the turtle carapace [2,3,6–8,16], the putatively unique topological position of the turtle scapula has not been critically examined. The slender, rod-like morphology of the turtle scapula is clearly derived, relative to the massive, rectangular shape found in basal reptiles (e.g. Paleothyris acadiana, Petrolacosaurus kansensis, etc.), but it is unclear whether the topological position anterior to the dorsal ribcage is unique and whether it represents an important transition in the origin of the turtle bauplan [3,7,8]. We analyse the topological position of the scapula in a representative sample of extant and fossil amniotes to determine the basal condition for Amniota. Given the central role that the topological position of the scapula plays in the evolutionary model for the origin of the turtle carapace [3,15], it is important to critically analyse this feature within a broader phylogenetic context.
2. Material and methods
A large sample of articulated skeletons of extant and extinct amniotes was studied and photographed. Sampling was undertaken to cover the greatest possible tree space in order to allow for reconstruction of ancestral conditions for each major amniote clade (Monotremata and Theria for Mammalia, Cryptodira and Pleurodira for Testudines; Sphenodon and Squamata for Lepidosauria; Aves and Crocodylia for Archosauria). Species are included only if the shoulder girdle is preserved in articulation, thereby reflecting as closely as possible the topological position of all elements in vivo. A complete list of species used is provided in the electronic supplementary material.
3. Results and discussion
The development of extant turtles, as well as the gross morphology of extant and fossil turtles, indicate that the scapula is universally oriented vertically and situated anterior to the ribcage. This position appears to be ancestral for amniotes, as among extant taxa it is found in amphibians, monotreme mammals and lepidosaurs (figure 1), and among fossil taxa is found in early tetrapods (e.g. Ariekanerpeton sigalovi), stem amniotes (e.g. Limnoscelis paludis) and early reptiles (e.g. Aerosaurus wellesi) [27,28]. Scapular blades that cover the dorsal ribs, by contrast, occur convergently in therian mammals and archosaurs and the scapular blades are convergently derived from axial somites [29,30]. Ancestrally, the scapulae of tetrapods are braced by dermal bones: the clavicles, interclavicle and cleithra. This plesiomorphically remains the case in turtles, where the scapula is surrounded by the nuchal, the entoplastron (interclavicle) and the epiplastra (clavicles). In monotreme mammals, the scapulae are strongly braced ventrally by the pre- and post-coracoids, the clavicles and the interclavicle. In therian mammals, by contrast, which typically lack all dermal components of the shoulder girdle, the scapular blade extends onto the ribcage and is stabilized by skeletal muscles (m. serratus anterior (AS), m. rhomboidei and m. levator scapulae ). A similar condition is found in archosaurs, where the scapula extends onto the ribcage and is affixed to the ribcage via the AS and the m. levator scapulae-rhomboid complex (sensu ).
Other ‘shelled’ amniotes exhibit both topological relationships of the scapula relative to the ribcage, indicating that this character is not correlated with the presence of a shell. Shelled sauropterygian reptiles from the Triassic, such as Henodus chelyops or Sinosaurosphargis yunguiensis, resemble turtles by having a vertically oriented scapula situated inside the shell, but placed anterior to the ribcage [31,32]. By contrast, therian armadillos or archosaurian ankylosaurs exhibit a scapular blade that is situated on top of the ribcage but underneath the shell.
The realization that a vertically oriented scapula which is situated anterior to the ribcage and that does not overlap the trunk ribs is plesiomorphic for amniotes has important implications for the evolutionary model regarding the origin of the turtle shell. Current hypotheses for the developmental evolution of the turtle shell are largely based on a single model organism, the trionychid Pelodiscus sinensis, and comparisons are made only to the development of two other model amniotes: the domestic chicken (an archosaur) and the house mouse (a therian mammal). Both species have long, posterodorsally directed scapular blades that are located on top of the ribcage. Thus, the position of the turtle scapula is seen as a uniquely derived feature, when in fact this topological relationship represents the basal amniote condition from which the conditions in the mouse and chicken later evolved. Although issues of sampling must be taken into consideration, it is apparent that the derived condition found in therian mammals and archosaurs (which includes both developmental model organisms: chickens and mice) is not an appropriate outgroup for studying the developmental origin of the turtle shell. Instead, among amniotes either monotreme mammals or lepidosaurs should be used.
We thank Greg Watkins-Caldwell for facilitating access to the Yale Peabody Museum of Vertebrate Zoology's collection.
- Received May 14, 2012.
- Accepted June 21, 2012.
- This journal is © 2012 The Royal Society