The fossil material comprises 26 well-preserved osteoderms attributed to Glyptosaurinae, registered under the references LAV 1258 to LAV 1273 in the paleontological collections of Montpellier II University (16 osteoderms), and SNB 1022–SNB 1031 (10 osteoderms) in the collections of the Muséum national d’histoire naturelle (Paris, France).

The Glyptosaurinae are considered to be Anguidae by most authors (Augé and Sullivan, 2006, Conrad and Norell, 2008 and Hoffstetter, 1955), although Hill (2005) proposed to consider them as the sister group of a clade including the Scincoidea (Scincidae, Cordylidae, Gerrhosauridae) and Anguidae. Among the Anguidae, they occupy a relatively marginal position, being closer to the Anguinae and Gerrhonotinae than to the Diploglossinae, and forming a sister taxon to the clade that includes the Anguinae and Gerrhonotinae. Up to now, some 13 genera and 28 species of glyptosaurine are encountered in Late Cretaceous-Early Oligocene sites in the USA, Europe, and Asia. The size and general aspect of these lizards must have been comparable to that of extant heloderms, i.e., 60 to 80 cm long animals with extensive, robust osteoderm shields covering most of the body, including the skull roof (review in Hoffstetter, 1955).

The specimens used for this study were collected in the 1970s and 1980s in the Phosphorites du Quercy (Crochet et al., 1981, de Bonis et al., 1973 and Rage, 2006), a broad fossiliferous area located in southwestern France, dated Early Eocene to Early Miocene (Rage, 2006). The specimens referenced LAV 1258 – 1273 are from the Bartonian (standard level MP 16, Middle Eocene of Lavergne locality), whereas the specimens referenced SNB 1022 – 1031 are from the Priabonian (standard level MP 18 of Sainte Néboule locality; Late Eocene).

A subsample of 15 complete and perfectly preserved osteoderms from the cranial (six specimens) and postcranial territories was selected in order to be studied by using four distinct morphological approaches.

Morphologically, glyptosaurine osteoderms (Fig. 1A–E) are characterized by a large size (up to 8.2 × 3.5 mm), an important thickness (up to 2 mm, or more), a hexagonal (cranial osteoderms) or rectangular shape, a variably pronounced dorsal keel, and a typical granular ornamentation (Hoffstetter, 1962). This osteoderm morphology is distinctive of the Glyptosaurinae; the osteoderms of another anguid taxon, the Gerrhonotinae, have a similar aspect, but they are proportionally thinner (Augé, 2005). The generic identification of our specimens is problematic because none of them was found in association with taxonomically significant remains (see Sullivan and Augé, 2006). They may belong to two genera that are relatively common in the Phosphorites du Quercy, Placosaurus Gervais, 1848–1852 and Paraplacosauriops Augé and Sullivan, 2006. In the following, they are considered as belonging to the Glyptosaurinae, with no further specification.

Seventeen of them have a roughly rectangular shape, with an area up to 37 mm² and a mean thickness 0.9 to 1.25 mm (Fig. 1A,B). They represent postcranial osteoderms. The other ones are from the cranial region (Fig. 1C,D); they figure irregular hexagons measuring 10–23 mm² for a mean thickness 1.97–2.4 mm.

The superficial ornamentation of all osteoderms consists of numerous tubercles (Fig. 1A,C,E), either blunt or ended by a sharp apex that create a rough surface, typical of the granular ornamentation. Some tubercles are partly superimposed or seem to have fused with each other, especially at the top of osteoderm keel. In a single osteoderm the diameter at the base of the tubercles may vary from 0.3 to 0.5 mm. The tubercles have a whitish, vitreous coloration that clearly contrasts with the brown color acquired by the other regions of the osteoderms during fossilization. The anterior, “gliding surface” of each rectangular osteoderm (Fig. 1A) is smooth (this part is covered by the preceding osteoderm: Hoffstetter, 1962). The superficial side of the osteoderms also displays vascular pits located between the tubercles (Fig. 1A,C). The deep surface bears 6–8 vascular pits the diameters of which range from 60 to 240 μm (Fig. 1B). The peripheral (equatorial) margin of the cranial osteoderms, and the lateral sides of the postcranial ones, display deep folds and indentations indicative of sutural surfaces (Fig. 1D). Vascular pits also perforate these surfaces.

Simple sections or breaks of the osteoderms show that all of them share the same basic diploe architecture (as already mentioned by Keller, 2009), with a central cancellous core bordered by two compact cortices (Fig. 1 and Fig. 2). However, postcranial osteoderms are more compact than cranial ones, and only display small rounded cavities. In all cases, the cortices house vascular canals that are generally more numerous in cranial than in postcranial osteoderms. Vascular canals consist of either primary or secondary osteons that communicate with the large inner cavities, and open at the deep and superficial sides of the osteoderms by pits as mentioned above. Primary, simple vascular canals occur in some specimens, but they are generally sparse. A striking feature of all osteoderms is the conspicuous difference between the tissue composing the superficial, tubercular layer and the subjacent bony layers (Fig. 1F,G). Most of the volume of the osteoderms is occupied by typical bone that colored in brown during the fossilization process. Conversely, the superficial layer is whitish, vitreous, often translucent or even transparent. This layer is entirely compact and no cavity, i.e. small vascular canals or osteocyte lacunae, can be identified, whatever the magnification used.

Cranial and postcranial osteoderms display four, clearly distinct tissues (Fig. 2). The first one consists in woven-fibered bone located in the core of the osteoderms (Fig. 2B). This tissue extends laterally into a broad, but relatively thin layer, the thickness of which is less than one third of the total osteoderm thickness. Vascular canals are mainly located in this region that displays evidence of intense remodeling. This process resulted in extensive Haversian substitution and broad resorption bays bordered by secondary lamellar bone.

In the basal region of the osteoderms the second tissue type consists in a thick layer of typical lamellar bone (Fig. 2C). This tissue contains numerous Sharpey's fibers that create two fiber systems: short (length 30–40 μm), robust (diameter up to 2.5 μm) and vertically-oriented fibers located deep into the lamellar cortex (Fig. 2C,D); and long bundles of fibers oriented obliquely and located at the extremities of the lamellar bone layer. Periodic growth lines occur throughout the basal lamellar layer. The deepest half of the basal region has been subjected to an active resorption process, mainly progressing downward, that created extremely broad lacunae partly reconstructed by secondary deposits (imbalanced remodeling; Fig. 2C).

The basal layer laterally merges with a third type of tissue, a bony tissue particularly rich in Sharpey's fibers. These fibers are long, parallel to each other, but their direction makes an angle of some 30–40° with the fibers of the basal region (Fig. 2D). The bone tissue that houses such fibers is intermediate between woven-fibered and parallel-fibered bone, and displays an irregular birefringence in polarized light. It is located near sutural surfaces, where it forms an equatorial ring. Toward the center of the osteoderm, this tissue merges with the core of woven-fibered bone, while its deep part shows evidence for extensive, imbalanced remodeling.

The fourth tissue type forms the superficial layer and is particularly thick, up to 400 μm, at the tubercle level. Its histological characteristics completely differ from those of the three bone tissues described above (Fig. 2 and Fig. 3). It is indeed composed of a vitreous, avascular and acellular tissue that is totally monorefringent and does not contain differentiated structures, with the exception of poorly-defined periodical growth marks (Fig. 3A,B). The limit between this tissue and the underlying woven-fibered bone is sharp. It is noteworthy that the superficial layer has the same histological aspect in all osteoderms studied, and that it does not display any sign of diagenetic alteration, even when the underlying bone tissue is heavily altered (Fig. 3C–E).

In several osteoderms, the superficial acellular layer comprises two superimposed strata of tubercles, often separated by a thin sheet of woven-fibered bone. This organization means that two successive generations of tubercles occurred at the same location, spaced with a short period, during which bone tissue was deposited (Fig. 3C–E). The superficial layer lacks Haversian systems, but in some osteoderms there is evidence for an erosion process restricted to the apex of the tubercles belonging to the first generation (Fig. 3A,B). Either a second tubercle was formed directly onto the resorbed surface of the previous one, or woven-fibered bone was deposited prior to the formation of a new tubercle. Conversely, the tissue located between the tubercles never displays any sign of resorption. Also, by place some tubercles are covered by woven-fibered bone and they do not protrude at the osteoderm surface (Fig. 3E).

Our histological observations were confirmed by SEM pictures showing a superficial layer devoid of osteocyte lacunae, in contrast to the subjacent woven-fibered bone tissue that displays abundant cell lacunae (Fig. 4A,B). The boundary between the two tissues is not as clear-cut as suggested by histological observations. In particular, tissue indentations create a strong anchorage of the superficial layer into the woven-fibered bone.

Transverse virtual sections obtained at every level of the osteoderms using computerized microtomography demonstrate unambiguously that the tissue composing the tubercles of the superficial layer is much more mineralized than the underlying bone tissue. The typical graph of X-ray opacity in the osteoderms, as measured along a depth-to-surface vector, is presented on Fig. 4C. The region of the graph that topographically corresponds to bone tissue (be it lamellar or remodeled woven-fibered bone) features a plateau at some 50 to 100 Hounsfield units (HU), whereas the region corresponding to the tubercles of the superficial layer displays a steep pick culminating at some 950 HU or more (the range of Hounsfield scale is –1000 to +1000 HU). There is thus a considerable difference in the mineral content of these two regions. The normal X-ray opacity of bone is in the range 400 – 900 HU (Shetty et al., 2010); a mean value of 650 HU can thus be considered representative of bone sensu lato. It therefore seems likely that the osteoderms underwent pronounced demineralization during fossilization. This demineralization process might have been uneven (depending on e.g. the local histological features of the osteoderm), a situation that could interfere with X-ray density measurements and incites to carefulness in considering the absolute values of these measurements. It nevertheless remains that X-ray tomography reveals a considerable difference in mineral content between the superficial layer that displays values reaching the upper limit of Hounsfield scale in spite of possible demineralization, and the underlying osseous regions.

The mapping of calcium (Fig. 4D) and phosphorus (Fig. 4E) densities on osteoderm sectional surfaces using an electronic probe also gave unambiguous results: the two ions are clearly in higher concentration in the superficial, ornamented layer than in the underlying bone strata. For ion concentration, as well as for X-ray opacity, the highest values are strictly restricted to the tubercles themselves, while the bone regions located below and in some places between the tubercles display the same relative values as the deep bony parts of the osteoderms.

A granular ornamentation comparable to that described here on the surface of glyptosaurine osteoderms is not frequent in tetrapods. Among squamates it seems to be restricted to anguid osteoderms, mainly represented by Glyptosaurinae and Gerrhonotinae (Hoffstetter, 1962). In other taxa the osteoderms are either devoid of superficial ornamentation (e.g., the varanid Varanus [Megalania] priscus: Erickson et al., 2003; the gekkonid Tarentola mauritanica: Levrat-Calviac, 1986), or they display a “vermiculate” ornamentation consisting of branching grooves (e.g., the anguine Ophisaurus apodus: Romer, 1997), or pits separated by crests (e.g., the gekkonid Tarentola neglecta: Levrat-Calviac, 1986). Among the Anguidae, the glyptosaurine osteoderms strikingly differ from the slow worm (Anguis fragilis: Anguinae) osteoderms, which display a typically vermiculate ornamentation resulting from a perivascular resorption of the superficial layer (Zylberberg and Castanet, 1985). Therefore, from the broadest morphological level, there is a conspicuous variability between the osteoderms of closely related taxa (see on this topic: Hoffstetter, 1962 and Levrat-Calviac, 1986). A comparison involving the dermal skeleton of the sarcopterygians as a whole confirms the relative sparseness of the granular ornamentation, but also reveals that the scales of some actinistians (Cœlacanthidae) bear an ornamentation pattern closely resembling that of glyptosaurine osteoderms (Ørvig, 1977: pl. 2A,B, 3C). In both cases, the ornamentation of the dermal skeleton consists of individual tubercles 0.4 to 0.7 mm in basal diameter, with an ogive shape ended by a sharp apex, and separated by areas of bare bone.

The histological structure of lizard osteoderms has been described, with variable precision, in 15 species representing 12 genera and six families: Gekkonidae (Levrat-Calviac, 1986 and Levrat-Calviac and Zylberberg, 1986), Anguidae (Moss, 1969 and Zylberberg and Castanet, 1985), Scincidae (Moss, 1969), Gerrhosauridae (Moss, 1969), Helodermatidae (Moss, 1969 and Vickaryous and Sire, 2009), and Varanidae (Auffenberg, 1981 and Erickson et al., 2003). In the most precisely studied taxa (Tarentola, Anguis and, to a lesser extent, Heloderma), the osteoderms display a simple structure, although the description of the composing tissues is sometimes confused due to the use of various terminologies. The osteoderms possess a deep (or basal) layer consisting of well characterized lamellar bone, and a superficial layer that is made of either a tissue poor in collagen fibers (so-called microfibrillar matrix), but containing osteocyte lacunae (e.g., Tarentola: Levrat-Calviac and Zylberberg, 1986), or of woven-fibered bone (e.g., Anguis fragilis: Zylberberg and Castanet, 1985). Moreover, if the mineralization of the basal layer is always of the inotropic type (i.e., apatite crystals located within and between the collagen fibrils), that of the superficial layer was described as being either inotropic as in Anguis or spheritic (i.e., formation of calcified spherules within a microfibrillar matrix), as in Tarentola (Levrat-Calviac et al., 1986). Up to now, no inner remodeling has been observed in lizard osteoderms (e.g., Moss, 1969). However, the ornamentation of the superficial layer was supposed to result from perivascular resorption (e.g., Anguis fragilis and Tarentola neglecta: Levrat-Calviac, 1986 and Zylberberg and Castanet, 1985).

Glyptosaurine osteoderms differ from those of modern lizards by the greater complexity of their structure, the presence of a rich vascular supply, and an intense and extensive perivascular remodeling. In addition to its histological peculiarities, the superficial layer displays a remarkable feature: in numerous osteoderms, it resulted from two successive “waves” of formation separated by a period of woven-fibered bone deposition. To date, such a feature was not observed in other squamate osteoderms, but was described in the scales of some actinopterygians (Meunier, 1980 and Ørvig, 1978) and actinistians (Ørvig, 1977). The possible meaning of this characteristic is further considered below (Section 4.4).

In squamate taxa for which detailed histological studies are available (Gekkonidae, Scincidae, Anguidae, and Helodermatidae), the osteoderms are considered as resulting from metaplasia (sensu Haines and Mohuiddin, 1968), a process known as a local mineralization of the dermis that is transformed into bone tissue although no osteoblast contributed to its formation (Levrat-Calviac, 1986, Moss, 1969 and Moss, 1972). In this interpretation, the deep part of the dermis, rich in collagen fibers (stratum compactum), becomes the lamellar bone tissue that constitutes the basal part of the osteoderms. Conversely, the superficial part of the osteoderms originates from the upper dermal strata that are relatively poor in collagen fibers (stratum laxum). This dual structure is frequently observed in modern squamate osteoderms but the nature and formation of the superficial layer are variable from one taxon to another, and were diversely interpreted in the literature (see synthetic table in Moss, 1969 and Vickaryous and Sire, 2009, p. 443). Moreover, the exclusive contribution of a metaplastic process can be questioned because active osteoblasts have been identified on osteoderm surface in A. fragilis (Zylberberg and Castanet, 1985) and T. mauritanica (Levrat-Calviac and Zylberberg, 1986). In other sauropsid groups, metaplasia would play a prominent role either as the only mechanism of osteoderm formation as, for example, in some sauropod dinosaurs (D’Emic et al., 2009), or as a temporary mechanism, involved only in the earliest stages of osteoderm formation, as in crocodilians (de Buffrénil, 1982 and Vickaryous and Hall, 2008) and stegosaur plates (Cerda and Powell, 2010, de Buffrénil et al., 1986, de Ricqlès et al., 2001 and Hayashi et al., 2009). The most notable exception to this trend is represented by placodont osteoderms that derive from cartilage anlagen (Scheyer, 2007).

The complex structure of glyptosaur osteoderms is poorly compatible with a growth pattern based exclusively on a metaplastic transformation of the dense and loose dermis. This conclusion relies on four observations made in this study, and listed below.

The vitreous, acellular tissue composing the superficial layer and the tubercles of glyptosaurine osteoderms neither exists in A. fragilis nor in Ophisaurus sp, two anguid lizards. The nature of this tissue cannot be settled with certainty in a fossil taxon, because developmental, histochemical and ultrastructural data are unavailable. Histologically, this tissue displays none of the typical characteristics of bone: it lacks altogether vascular canals, osteocyte lacunae, canaliculi or any kind of intra-osseous tubules that could have housed cytoplasmic extensions of osteoblasts. Moreover, its vitreous, transparent aspect on sections, and its completely monorefringent reaction to polarized light, suggest the absence (or very poor development) of a fibrillar, collagenous network. In addition, this tissue is clearly hypermineralized as compared to bone. Consequently, the histology of this tissue is inconsistent with any kind of bone tissue, including acellular bone (a tissue known, among living vertebrates, in acanthomorph actinopterygians only) that has a normal collagen network (Meunier, 1987). Also, this tissue cannot be considered as some kind of atypical osseous formation resulting from metaplasia for the reasons exposed above. A convergent conclusion, based on histochemical and ultrastructural data, had tentatively been approached by Levrat-Calviac (1986) about the superficial layer of Tarentola osteoderms.

At a broad comparative level, this tissue looks similar, at both the histological and microanatomic levels, to various hypermineralized tissues covering the integumentary skeleton of non-tetrapod vertebrates (review in Sire et al., 2009). These tissues include enamel and enameloids (odontodes and dermal denticles), ganoine (scales of the Polypteridae and Lepisosteidae), hyaloine (scutes of armored Siluriformes) and limiting layer (elasmoid scales). In addition, both the organization of the tubercles and the structure of the superficial layer of the glyptosaurine osteoderms are strikingly comparable to various published pictures showing either the superposition and superficial resorption of odontodes on the scales of Polypteridae (Meunier, 1980 and Meunier and Gayet, 1992), or the growth pattern of actinistian odontodes, with bone layers wedged between successive odontode generations (Ørvig, 1977). Such a similarity is further substantiated by the mineralization rate of the superficial layer of glyptosaurine osteoderms that is much more compatible with enamel than with bone, as shown by the quantitative analysis of X-ray opacity on CTscan sections. However, it is noteworthy that the superficial, hypermineralized tissues occurring on the scales of non-tetrapod sarcopterygians are generally associated with dentine, which is obviously not the case in glyptosaurine osteoderms.

Therefore, considering these similarities, the most parcimonious conclusion is to consider that the tissue composing the superficial layer and the tubercles of glyptosaurine osteoderms actually belongs to the broad category of hypermineralized, enamel-like tissues. We call this particular tissue osteodermine.

The structural data presented here for osteodermine strongly suggest a similar mineralization process as previously described for the hypermineralized, enamel-like tissues: deposition of a loose organic matrix involving either dermal and epidermal or only epidermal contributions, mineralization and maturation, an important step during which most of the organic matrix is removed through proteolytic activity.

In non-glyptosaurine squamates, the existence of a similar, non-osseous layer has been considered by Moss, 1969 and Moss, 1972 in Heloderma osteoderms. Vickaryous and Sire (2009) also concluded to the existence of such a tissue on the osteoderms of Tarentola annularis and T. mauritanica, as yet suggested by Levrat-Calviac (1986). Apart from lizards, such a tissue is unknown in other tetrapod osteoderms, whatever the clade to which they belong.

In the sarcopterygian clade, the occurrence of hypermineralized tissues on postcranial elements of the dermal skeleton has been hitherto clearly settled for basal taxa only: actinistians, some fossil dipnomorphs (i.e., dipnoans and porolepiforms: see review in Sire et al., 2009) and piscine stem tetrapodomorphs (the paraphyletic “osteolepiforms”: Vickaryous and Sire, 2009). Such a feature was not observed in any known tetrapod taxon, with a possible exception for some squamates as mentioned above (Witzmann, 2009 and Witzmann and Soler-Giron, 2010 for basal tetrapods).

In teeth and various dermal skeletal elements, the formation of enamel-like tissues results from epidermal-dermal interactions, with an ultimate contribution of the epidermal basal layer cells in the deposition of organic matrix at the surface (Levrat-Calviac and Zylberberg, 1986, Moss, 1969, Moss, 1972 and Sire et al., 2009). According to present documentation, such epidermal-dermal interactions, unexpressed but latent in deeply nested tetrapodan taxa, would be expressed during osteoderm formation in certain squamates. The discontinuous distribution of this character raises the question whether the osteodermine of squamate osteoderms is a neoformation, representing a developmental convergence with the scales of basal sarcopterygians, or a true homology (Vickaryous and Sire, 2009). The latter hypothesis would, of course, involve the persistence, during sarcopterygian evolution, of a potential capacity of epidermal cells to interact with the dermis to produce hypermineralized tissues in other regions than in the oral cavity. Up to now, there is no definite answer to this question, although the most parsimonious interpretation seems to be the second one. A significant contribution to solve this question would be to bring direct evidence for an epidermal participation to osteodermine formation in squamate osteoderms; an objective that could be achieved for instance by looking for remnants of enamel proteins in fossil and modern material or by demonstrating the expression of enamel proteins in epidermal cells located close to the osteoderm surface. Studies on these topics are currently in progress.