Torsten M. Scheyer, Paläontologisches Institut und Museum, Universität Zürich, Karl Schmid-Strasse 4, CH-8006 Zürich, Switzerland. Tel.: 41 (0)44 634 2322; fax: 41 (0)44 634 4923; e-mail: firstname.lastname@example.org
The extinct parareptilian clade of pareiasaurs was in the past often presented to constitute a morphocline from larger, less armoured forms to smaller, well armoured forms, indicating that the osteoderm cover became an increasingly prominent aspect in the post-cranial skeleton of these animals. Here, we describe microanatomical and microstructural aspects of osteoderms of the three pareiasaur taxa Bradysaurus, Pareiasaurus and Anthodon from the Permian of South Africa. A generalized mode of osteoderm formation, consistent with intramembraneous skeletogenesis, is hypothesized to be present in all pareiasaurs. Few characters are shared between pareiasaur dermal armour and turtle shell bones and osteoderms. Otherwise, there is strong evidence from microanatomy and histology (i.e. absence of structures that formed via metaplasia of dermal tissue) that indicates nonhomology between pareiasaur dermal armour and the armour of living eureptiles. Analysis with bone profiler revealed no clear connection between bone compactness and lifestyle in the amniote osteoderm sample.
The discovery of Odontochelys semitestacea from the Upper Triassic of southwestern China (Li et al., 2008), a new stem turtle that largely lacks a carapace but has a plastron, throws new light on this discussion. Its early Late Triassic age (ca. 235 Ma, early Carnian, Wang et al., 2008; see also Kozur & Bachmann, 2008) makes O. semitestacea the oldest turtle known. Phylogenetic analysis indicates that it is the sister taxon to all other known fossil and recent turtles. Taking the geologic evidence (Wang et al., 2008) at face value, this turtle inhabited marine environments, raising the question if turtles were originally a marine group (Li et al., 2008) or if O. testacea was secondarily adapted to the marine environment, as were so many later turtles (Reisz & Head, 2008). Thus paleoecology was and still is playing a major role when identifying possible sister-group relationships of turtles (e.g. Joyce & Gauthier, 2004; Scheyer & Sander, 2007; Kriloff et al., 2008; Rieppel, 2008).
Pareiasaur osteoderms have been well known since the beginning of last century, and morphological descriptions and figures of pareiasaur armour plates are found for example in Boonstra (1934), Huene (1944, 1956), Findlay (1970) and Lee (1997a). Until recently, however, their phylogenetic value was thought to be limited (e.g. Hill, 2005). It is thus not surprising that the study of bone histology, a powerful tool of gathering data from fossils in its own right, was so far restricted to pareiasaur long bones to address questions about the mode of skeletogenesis, skeletochronology and paleoecology (e.g. Ricqlès, 1978a,b; Kriloff et al., 2008). A femur and rib of Pareiasaurus was analysed by Ricqlès (1978a), and large pareiasaurs like Pareiasaurus and Bradysaurus were noted to show lamellar-zonal bone tissue in their long bones that is well vascularized, with pseudo-laminar or pseudo-plexiform patterns, linked to rather high growth rates (Ricqlès, 1978a,b). By using linear discriminant models that utilize bone compactness profiles of tibial cross-sections, Kriloff et al. (2008: table 5) inferred that Pareiasaurus had an aquatic lifestyle; however, based on its general osteology, and following Piveteau (1955) in this regard, Kriloff et al. (2008) considered Pareiasaurus more tentatively to be amphibious instead of purely aquatic. Kriloff et al. (2008) further note that an amphibious lifestyle is in accordance with the spongy, pachyostotic nature of the cranial bone of Bunostegos akokanensis as described by Sidor et al. (2003).
The prominent postcranial osteoderms of pareiasaurs, on the other hand, were never subject to a comparative histological study. Here, the bone histology of the dermal armour of three common pareiasaurs from the Permian of South Africa is described and compared. The whole taxonomic breadth (Fig. 2) of pareiasaurs is covered by the sampling of Bradysaurus, currently acknowledged to be the sister taxon to all other pareiasaur taxa; of the type genus Pareiasaurus, and of the dwarf pareiasaur Anthodon. The histology of the osteoderms is compared with previously published data for turtle shells and osteoderms (e.g. Scheyer, 2007a; Scheyer & Sander, 2007).
Hypotheses to be tested by the present study thus are: (i) Pareiasaurs share common bone histological characters; indicating that they also share a similar mode of osteoderm skeletogenesis. (ii) If turtles were to be nested deeply within Pareiasauria, i.e. are the sister taxon to the dwarf pareiasaur Anthodon serrarius (Owen, 1876) (Fig. 2b), they ought to share histological and developmental characters in the postcranial dermal armour. (iii) Recently, Pareiasaurus was considered to have an amphibious lifestyle based on long bone histology (Kriloff et al., 2008). If, as exemplified by turtles and placodonts, both armour and long bones are similarly influenced by the environment, pareiasaur osteoderms may show histological adaptations similar to long bones.
Materials and methods
Osteoderms of the genera Bradysaurus, Pareiasaurus and Anthodon from the Permian of South Africa were used in this study. The taxa, accession numbers, and respective localities are listed in Table 1. For comparison, osteoderms of nine amniote taxa, including lepidosaurs, turtles, crocodylomorphs and ankylosaurs were sampled. In the specimen of Alligator mississippiensis, derived from a captive female that as a juvenile was kept under very poor conditions, the microstructure of the osteoderm was known to be affected by several resorptive cycles induced by oviposition (Klein et al., in press). Although the armour plate has a pathologic groove in its keel, we still decided to include this specimen in the analysis, because the life history is documented for this specimen. Bone compactness parameters of all osteoderm samples are compiled in Table 2.
Table 1. Material sectioned for the study including taxa names, accession numbers, short specimen descriptions, and general remarks about the geological age of the specimens.
Osteoderm mostly free of sediment; Late Permian locality of Dunedin, Beaufort West District, South Africa
Osteoderm mostly free of sediment; Late Permian locality of Dunedin, Beaufort West District, South Africa
Table 2. Taxon names, accession numbers, lifestyles and main compactness profile parameters of the specimens studied.
As shown for example by Kriloff et al. (2008: fig. 1a), the parameters S, P, Min and Max are mean values of 60 radial values measured with bone profiler. Note that lifestyles of pareiasaur taxa are marked with a question mark, because their ecology is to be assessed herein.
Pareiasaurus sp. (smaller osteoderm)
Pareiasaurus sp. (larger osteoderm)
Anthodon serrarius (thick osteoderm)
Anthodon serrarius (thin osteoderm)
Pseudopus apodus (recent)
PIMUZ A/III 1280
Tiliqua scincoides (recent)
PIMUZ A/III 1281
cf. Hesperotestudo (flat osteoderm)
cf. Hesperotestudo (spiked osteoderm)
Alligator mississippiensis (recent)
Thin-sections were prepared and documented in the bone histology labs at the Steinmann Institute, University Bonn, and the PIMUZ, University Zurich, following standard petrographic preparation techniques (e.g. Scheyer, 2007a,b; Scheyer & Sánchez-Villagra, 2007). Almost all pareiasaur samples retained a partial sediment cover, so the original surficial bone tissue is pristine, without being affected by preparation in these areas. The amount of vascularization and bone compactness of the pareiasaur osteoderms was assessed qualitatively by direct inference from the histological data. In both pareiasaur and other amniote specimens, bone compactness was analysed quantitatively using the software bone profiler Version 3.20 (Girondot & Laurin, 2003). The program was initially developed to analyse the cross-sections of long bones, which usually are of rounded, oval, or suboval shapes instead of flat dermal armour bones and has been used since on a broad variety of vertebrates (e.g. Laurin et al., 2004; Germain & Laurin, 2005; Canoville & Laurin, 2008; Kriloff et al., 2008). To use the program, schematic black and white drawings showing bone in black and vascular spaces (i.e. medullary cavity, vascular canals, osteons and resorption spaces) in white were produced from the actual thin-sections using photoshop CS3 (Adobe Systems Inc., San Jose, CA, USA). Compactness profiles were calculated using mainly the automatic bone centre function of the program; however in few cases it was necessary to adjust the centre manually to maximize the scanned bone area (e.g. in the Ankylosauridae indet. specimen).
Bone histology and microanatomy
The outer shapes of pareiasaur osteoderms, as seen in thin-section (Fig. 3), can differ strongly (e.g. Findlay, 1970; Lee, 1997a), whereas their bone histology is surprisingly homogeneous. For this reason, the histology of all three taxa is described together. The sampled pareiasaur osteoderms of A. serrarius showed only weakly festooned lateral margins instead of the well developed peg and socket structure of sutures (Fig. 3e,f). The specimens had a flat internal (= visceral) base and an external convex surface sculptured with radial ridges. A boss was present in osteoderms of Pareiasaurus and Anthodon (Fig. 3b–f). Vascular traces on external bone surfaces are random in Pareiasaurus, and organized and radial in A. serrarius. A centrally situated large foramen opening on the dorsal surface of the osteoderm is present in Pareiasaurus and Anthodon.
In all specimens sampled, the cortices of the osteoderms are pachyostotic; especially at the osteoderm margins. External and internal cortices are mainly of similar thickness in A. serrarius (up to 5 mm or more at margins), although the thickness of the external cortex may be greatly reduced locally by extensive vascularization and resorptive action that can reach the external surface of the osteoderm. In the Pareiasaurus and Bradysaurus samples, the internal cortices are generally thicker than the external cortices (Fig. 3).
The external cortex consists of parallel-fibered bone (PFB) with additional fiber bundles, presumably Sharpey’s fibers, which extend perpendicular or at high angles to the external surface of the osteoderm into the cortical bone (Fig. 4a–c). The cortical bone tissue at the (sutured) margins of the osteoderm is usually indistinct from the external cortical bone. Vascularization in the external cortex of the osteoderm is quite extensive with small primary osteons, erosion bays and radially directed primary vascular canals that extend up to the surface of the bone, thus creating a roughened texture on the bone surface. Towards the interior of the osteoderm, the primary canals anastomose frequently, whereas in the superficial layers of the external cortex, the primary canals are generally isolated, unbranching tubes (Fig. 4a). Few scattered secondary osteons appear towards the cancellous bone. They are absent in the external-most and internal-most layers of the external and internal cortices, respectively. The interior cancellous bone consists of trabeculae that mostly resemble cylindrical struts lined with lamellar bone (Fig. 4d). Many trabeculae show several resorption-reconstruction cycles, but interstitial primary bone tissue may be present. The trabeculae are thicker in diameter in flat osteoderms and increase in length and gracility with increasing osteoderm height. The internal cortices consist of layers of PFB (Fig. 4e,f). Growth marks are well developed both in the external and internal cortices. In osteoderm SAM-PK-10074 (A. serrarius), a maximum number of 17 prominent incremental growth marks were counted. The individual zones between the incremental growth marks strongly decrease in thickness towards the outer cortical bone surface (Fig. 4e), but a true external fundamental system that would indicate a reduction or even a cessation of growth correlated with old age (e.g. Horner et al., 1999) is not observable. Coarse Sharpey’s fibers insert at moderate to high angles into the compact bone (Fig. 4f). Few scattered primary osteons and vascular canals as well as isolated large foramina of up to 1.0 or 1.5 mm in diameter are present in the bone tissue of the internal cortices.
Bone compactness profiles using bone profiler
To independently verify our observations of tissue vascularization and bone compactness above, we obtained bone compactness profiles of the pareiasaur osteoderms and other amniote osteoderms (Fig. 5) using bone profiler. The usage of the program to analyse dorsolaterally compressed osteoderms that may exhibit also a strong relief with bosses and ornamentation, large resorption spaces, and foramina of large diameter might seem farfetched at first (Fig. 6a, b); however, bone profiler still provides adequate results for these specimens (Table 2). In many cases although, small alterations of the schematic drawings of the specimens, i.e. adding thin black lines to close off larger canals and foramina linking the main osteoderm cavity with the external bone surface, were necessary prior to analysis. The complex relief of the specimens and the numerous smaller foramina prevent the program from assessing all bone sectors in a sample because measurement is cut off at these incisures (Fig. 6c). The resulting compactness profiles are interpreted to nevertheless give a good approximation of the overall compactness of the specimens. As roughly indicated by the graph and the compactness parameters (Fig. 7) of the thicker osteoderm of A. serrarius, the interior of the osteoderm is spongiose, with about 50 % of the space being occupied by bone tissue and bone vacuities each, whereas the bone compactness of the cortical part of the osteoderm reaches around 95 %. Compactness values were somewhat lower in the other specimen of A. serrarius, increasing from 38 % at the centre to about 91% at the bone surface. The small flat specimen of Pareiasaurus sp. showed the steepest increase in compactness from about 15% to over 93%, whereas the larger specimen showed an increase from 33% at centre to 90% at the bone surface. The osteoderm of Pareiasaurus serridens, on the other hand, is more homogeneous (with values ranging from 60% to 90%). The values of the Bradysaurus osteoderm (56–71%) have to be treated with caution because of its incompleteness. Otherwise, the specimen shows the least compactness in the outer cortical regions of all samples; the only other specimen with such a low compactness value at the bone surface (about 77%) being the specimen of cf. Diplocynodon. In the latter, the low value is intrinsically linked to the large open dorsal pits (Fig. 5c), which had to be manually closed prior to analysis with bone profiler. The highest compactness values were found in the two lepidosaur species, in which no large vascular spaces are present, and in the specimen (overall 97%) of Steneosaurus sp. The specimen of A. mississippiensis exhibited an intermediate compactness (61% at the centre to 100% at bone surface) between cf. Diplocynodon and Steneosaurus sp. The large gap in the specimen of Ankylosauridae indet. (values increasing from 35% to 89%) was closed by a line prior to sampling; however, the smaller separated patch of bone was not included into the analysis. In the specimen of Nodosauridae indet., the compactness values increased from 50% at the centre to 96% at the bone surface. The two turtle specimens were fairly compact with values of 53–99% (spiked specimen) and 76–99% (flat specimen), respectively.
The bone microanatomy and bone microstructure of pareiasaur armour (as described here) are consistent within the group, thus a shared mode of skeletogenesis can be assumed for pareiasaur dermal armour in general. A well preserved growth record (interpreted as consisting of at least 17 annual cycles) and the overall increasingly reduced spaces between the growth increments found in osteoderm SAM-PK10074, confirm that this small specimen did not come from a juvenile A. serrarius (e.g. Findlay, 1970; Lee, 1997a,b; Spencer & Lee, 2000). However, whether osteoderms were already present at juvenile stages in this dwarf pareiasaur species cannot be confirmed. As exemplified by a find of a ‘very young juvenile or “baby”’ of Elginia that lacks osteoderms (Spencer & Lee, 2000), the postcranial armour is interpreted to be absent in early juvenile stages of pareiasaur development. The large foramina opening to the dorsal bone surface in Pareiasaurus and Anthodon seem to provide a central passage for nutrients through the osteoderms, with a possible function being an increased blood supply to overlying dermal and epidermal tissue layers. Similar structures are known for example from armour plates in ankylosaurs (e.g. Scheyer & Sander, 2004; Main et al., 2005). Whether a stronger blood supply in the pareiasaur osteoderms did also have some thermoregulatory function as was hypothesized for stegosaur back plates (e.g. Buffrénil et al., 1986) remains speculative, although.
There are few characters that pareiasaur osteoderms (i.e. presence of parallel-fibered bone tissue and possible sutured margins in A. serrarius) share with the microstructure and anatomical details of turtle shell bones and osteoderms (e.g. Scheyer, 2007a; Scheyer & Sánchez-Villagra, 2007; Scheyer & Sander, 2007). Turtle shell bones do not show ornamental bosses and radial ridges nor radial growth and vascularization patterns, whereas pareiasaur osteoderms lack diploe structures, scute sulci and a clear distinction of internal and external cortical bone tissue; all of which are possible histology-based synapomorphies of turtle shell bones (e.g. Scheyer, 2007a,b). Furthermore, with the possible exception of Sharpey’s fibers, no microstructures could be identified in the pareiasaur samples that hint at metaplastic ossification, a major developmental mode in fossil and extant eureptilian osteoderm formation (e.g. Zylberberg & Castanet, 1985; Levrat-Calviac & Zylberberg, 1986; Hill, 2005; Scheyer et al., 2007; Vickaryous & Hall, 2008). The pareiasaur osteoderms sampled herein are hypothesized to represent ‘de novo’ structures that develop intramembraneously, displacing preformed integumentary (i.e. dermal) tissue structures instead of incorporating them. The absence of metaplastic dermal tissue in these osteoderms thus demands fundamental changes in skeletogenic pathways from turtle shell and other eureptilian osteoderm development. With the possible exception of ‘de novo’ formed armour ossicles in ankylosaurs (Ricqlès et al., 2001), histological data of pareiasaur osteoderms argue against a homology with dermoskeletal bone structures of extant and extinct reptile groups.
bone profiler is able to calculate bone compactness in the specimens sampled, although it might not be feasible in certain cases to analyse the whole bone tissue based on the strong outer relief and foramina exhibited by some specimens. This would mean including larger empty spaces in the cortical bone, which then would entail lower compactness values in the analysis. Still, a fairly good quantitative estimation of bone compactness and overall vascularization is thus available that complements the qualitative assessment of the bone histology of the osteoderms.
In the thick osteoderm of A. serrarius, the cortex locally can take up 40–50 % of the osteoderm radius. In those amniote osteoderms used for comparison that similarly exhibit a well vascularized bone centre, i.e. the spiked turtle osteoderm TMM 30967-1010.2 and specimen TMP 67.10.29 of Nodosauridae indet., the compacta makes up only about 27–33%. In comparison, the cortices of the sampled pareiasaur osteoderms thus locally show pachyostosis. The pachyostotic cortical bone may be well vascularized, especially in the regions of bosses, or it may show signs of osteosclerosis and hyperostosis, usually at the margins of the osteoderms. In this regard, the data are consistent with data on long bone histology (Ricqlès, 1978a,b; Kriloff et al., 2008) and with reports of pachyostosis in the cranial remains of the pareiasaur Bunostegos (Sidor et al., 2003).
Osteoderms of Proganochelys quenstedti, i.e. the oldest stem turtle for which extensive osteoderm material is known (Gaffney, 1990), would have been interesting to compare with pareiasaur osteoderms. However, these dermal bones were not available or suitable for histological study, but fossil osteoderms of the giant tortoise cf. Hesperotestudo from the Pleistocene of North America were included in the study. Previous examination of the shell bone microstructure revealed that P. quenstedti shared most histological characters with fossil and extant land-dwelling turtles, e.g. testudinid turtles (Scheyer & Sander, 2007). Therefore, we assume that osteoderms would also be quite similar in these taxa. As indicated by the specimens of cf. Hesperotestudo, the compactness and vascularization of the osteoderms, presumably belonging to one individual, can vary strongly (Fig. 5; Table 2), which also seems to be correlated with the outer shape and size of the respective osteoderm.
To summarize, pareiasaur bone compactness profiles vary from those of other osteoderm-bearing amniotes (Figs 3 and 5; Table 2). Differences (i.e. compactness of cortical bone; vascularization of interior parts of osteoderms) become apparent in both the visual and the computational analysis of the compactness of the bone samples. The osteoderms with the highest compactness were found in the terrestrial lepidosaur species; however, as these are diminutive bone structures a few millimeters in thickness, they cannot accommodate large vascular spaces either. In the osteoderm of the thalattosuchian Steneosaurus, a taxon inhabiting coastal marine environments (e.g. Hua & Buffrénil, 1996), pachyosteosclerotic trends almost obscured the diploe structure typical for crocodylian osteoderms. The high compactness values found in the present study thus even exceed the value of bone volume (61%) previously reported for an osteoderm of this genus (Hua & Buffrénil, 1996). We confirm the hypothesis that Steneosaurus lived in coastal marine waters and that the dense osteoderms added bone ballast for buoyancy control (Hua & Buffrénil, 1996).
Unfortunately, no direct connection became apparent through the analysis with bone profiler, which would clearly link bone compactness and lifestyle in the amniote osteoderms used for comparison (see Table 2). In pareiasaurs, there is a light trend of increasing compactness from Bradysaurus to Pareiasaurus, and finally, to Anthodon, which may again be attributable to overall differences in size and shape. Alternatively, this trend might be related to a change in ecological preference from less armoured to well armoured pareiasaurs. The only convergence of model-based compactness was found between the A. serrarius specimens (and to a lesser degree P. serridens) and the nodosaurid specimen. Nodosaurid ankylosaurs are interpreted as being terrestrial animals that, according to evidence of ichnofossils, preferred wet, well-vegetated areas, e.g. floodplains (McCrea et al., 2001; see also Vickaryous et al., 2004). Apart from this overlap, pareiasaur osteoderms generally had a lower model-based compactness than all other amniote osteoderms. Our compactness analyses and the presence of pachyostotic trends in the osteoderms might indicate similar preference of wet habitats in pareiasaurs. However, given the small sample size of the present study and the inconclusive association of bone compactness and lifestyle in the amniote osteoderms used for comparison, we were not able to make further inferences on the lifestyle of the pareiasaurs.
Roger Smith and Sheena Kaal (Cape Town), Jennifer Clack (Cambridge), Tim Rowe (Austin), James Gardner (Drumheller) and Edith Müller (Solothurn) are acknowledged for providing the material for histological study. Johannes Müller, Marcelo Sánchez, Robert Reisz and Jörg Fröbisch (Toronto), the latter also hand-carrying the pareiasaur specimens, are thanked for fruitful discussions. Olaf Dülfer and Georg Oleschinski (Bonn) are acknowledged for their help in preparing and photographing thin-sections, respectively. The manuscript also greatly benefited from the constructive comments of Michel Laurin and Michael Lee, the former providing also invaluable hints for using bone profiler. The project was funded by DFG grant #SA 469/15 and ‘Fonds zur Förderung des akademischen Nachwuchses’ of the University Zurich (FAN).