Unravelling the structural variation of lizard osteoderms

Vertebrate skin is a remarkable organ that supports and protects the body. It consists of two layers, the epidermis and the underlying dermis. In some tetrapods, the dermis includes mineralised organs known as osteoderms (OD). Lizards, with over 7,000 species, show the greatest diversity in osteoderm morphology and distribution, yet we barely understand what drives this diversity. This multiscale analysis of five species of lizards, whose lineages diverged ∼100-150 million years ago, compared the micro- and macrostructure, material properties, and bending rigidity of their ODs, and examined the underlying bones of the skull roof and jaw (including teeth when possible). Unsurprisingly, OD shape, taken alone, impacts bending rigidity, with the ODs of Corucia zebrata being most flexible and those of Timon lepidus being most rigid. Macroscopic variation is also reflected in microstructural diversity, with differences in tissue composition and arrangement. However, the properties of the core bony tissues, in both ODs and cranial bones, were found to be similar across taxa, although the hard, capping tissue on the ODs of Heloderma and Pseudopus had material properties similar to those of tooth enamel. The results offer evidence on the functional adaptations of cranial ODs, but questions remain regarding the factors driving their diversity. STATEMENT OF SIGNIFICANCE: Understanding nature has always been a significant source of inspiration for various areas of the physical and biological sciences. Here we unravelled a novel biomineralization, i.e. calcified tissue, osteoderms, forming within the skin of lizards which show significant diversity across the group. A range of techniques were used to provide an insight into these exceptionally diverse natural structures, in an integrated, whole system fashion. Our results offer some suggestions into the functional and biomechanical adaptations of osteoderms and their hierarchical structure. This knowledge can provide a potential source of inspiration for biomimetic and bioinspired designs, applicable to the manufacturing of light-weight, damage-tolerant and multifunctional materials for areas such as tissue engineering.


Introduction
Vertebrate skin is a remarkable organ that helps to support and protect the body, acting as an interface between the organism and its environment.The skin consists of two layers -the epidermis (derived from ectoderm) and the underlying dermis (originating from mesoderm and neural crest), separated by a fibrous basement membrane [1] .The dermis includes cutaneous nerves, blood vessels, and variable amounts of glandular and fatty tissue.It may also include mineralised organs such as the elasmoid scales of teleost fish or the osteoderms (OD) that develop in many tetrapods.Unlike epidermal scales, which are composed primarily of keratin proteins, mineralised skin organs include varying amounts of organic (e.g.collagen) and inorganic (hydroxyapatite) material.Mineralised organs within the skin date back to the Ordovician ( ∼500 Ma), and demonstrate a diversity of different tissue types including bone, dentine, and enamel-like layers [2] .Among tetrapods, the majority of mineralised skin organs are composed primarily of bone, and hence are referred to as osteoderms ( = bone skin ).Less common are the dermal scales of gymnophionan amphibians (caecilians) and the lamina calcarea of some frogs and toads.Whereas gymnophionan dermal scales closely resemble the elasmoid scales of teleost fish, the lamina calcarea -an unusual calcified layer found within the dermis of tadpoles and adults -appears to lack any obvious comparisons with other vertebrate hard tissues [3] .
ODs have been reported from representative members of most major extant tetrapod lineages including amphibians (various frog species), mammals (such as the armadillo), and reptiles (turtles, crocodilians, and many lizards) [4][5][6][7] .Lizards, with over 7,0 0 0 species, show the greatest diversity in OD morphology and distribution, yet our understanding of their structure, development, and function remains limited [8] .In lizards, ODs may be completely absent, present in restricted areas, or found all over the body in varied arrangements: as non-overlapping mineralised clusters; as a continuous covering of overlapping plates; or as spicular mineralisations that thicken with age and show a variety of structural designs [ 3 , 9 ].What drives this extraordinary diversity is still an unsolved puzzle [8] .
Several authors have carried out detailed histological analysis of lizard ODs to explore their developmental origin and biological constituents [ 6 , 10-13 ].Nevertheless, only a few studies have characterised their overall morphology [ 9 , 14 , 15 ].Even fewer studies have characterised their mechanical properties (e.g.elastic modulus) or biomechanics [16][17][18] despite a significant body of literature for other taxa, such as alligators, armadillos, and various fishes [19][20][21][22] .A better understanding of these structures in lizards has the potential to lead to the development of bioinspired and biomimetic structures and devices [ 15 , 23-29 ].
The macro-and microstructure of the ODs were described for each species using microCT and histological analysis, respectively.The mechanical properties of each OD and its component tissues were characterised using nano-indentation.To compare the mechanical properties of ODs with other mineralised tissues from the same species, we also measured the mechanical properties of components of the frontal bone, dentary, and, where possible, teeth, all taken from fresh frozen specimens.However, as skull material of V. komodoensis was not available for sampling, we substituted bone samples from a skull of V. niloticus that was of comparable size to those of the other lizards studied.Finally, given the observed diversity in the morphology and mechanical properties of the ODs, we used finite element analysis to assess the bending rigidity of the ODs along both the antero-posterior and transverse axes, focusing on the impact of shape and size variability, rather than any differences in their component tissues.

Sample collection
Adult lizards were obtained from the Evans Lab (UCL).These specimens were donated by the Pathology Department, Zoological Society of London, and died of natural causes over a prolonged period; as such the precise date and time of death is unknown and varied between animals.The skull, mandible, and skin samples for single OD extraction were taken from animals that had been frozen (ca.-20 °C) upon receipt.Note that: (1) although the varanid skin sample was taken from V. komodoensis , the skull and dentary samples were from V. niloticus as skull bone samples from V. komodoensis were not available for destructive sampling.These samples were then used for morphological, histological, finite element analysis, and material characterisation.V. niloticus has not been recorded as having OD in its skin [9] and we found no trace of them in our CT scans and dissections; and (2) one OD from either the head or the neck region (see Fig. 1 ) of one animal per species was used in each of the following characterisation methods.
The low number of samples per test (mainly due to specimen availability) is a limitation.However, we cross-checked our findings: (1) a single OD from the same anatomical region of the same H. suspectum was indented where the same pattern was observed as presented here [see 18]; (2) ODs from two other H. suspectum, from the head and from the same anatomical region as this study, were sectioned for visualisation purposes and the same patterns were observed; (3) ODs from two other C. zebrata, from the head, were sectioned for visualisation purposes and the same patterns were observed.Hence, we have confidence in the data presented herein, especially the relative differences reported, despite the small sample number.Note that limited sampling is also in line with the principals of animal Replacement, Reduction, and Refinement (3Rs).

Morphological analysis
With the exception of V. komodoensis , the whole head, as well as single OD (one sample from each specimen), were imaged using the micro-computed tomography (microCT) Nikon XT H 225 ST scanner (Nikon Metrology Ltd., UK) at University College London.MicroCT images were imported into the image processing software Avizo (Thermo Fisher Scientific, Mass, USA) and were manually segmented.3D reconstructions of the ODs were used to carry out basic morphological measurements (length, width, height, volume, surface area) as shown in Fig. 2 .

Histological analysis
Dissected skin samples were defrosted at room temperature and fixed in 20x volume of 10% neutral buffered formalin for at least 24 h, then placed in decalcification solution (1.9% glutaraldehyde, 0.15 M EDTA, in 0.06 M sodium cacodylate buffer, adjusted to 7.4 pH) at 4 °C, changing to a fresh solution every 7 days, for 4 weeks.Further decalcification was performed prior to embedding using a Sakura TDE TM   1428).Samples were embedded in paraffin wax and sectioned either coronally or parasagittally, at 5 μm thickness, with an HM Microm 355S automatic rotary microtome (Thermo Fisher Scientific).These sections were then stained with haematoxylin and eosin (H&E) [40] , Masson's trichrome [41] .The slides were imaged using a Leica SCN400 scanner, and visualised and analysed in Apero Imagescape, and Qpath 0.2.3.

Material characterisation
Dissected ODs (one per species) and coronal sections of frontal and dentary bone (cut at the same point in each species, also one specimen per species) were embedded in cold cure epoxy resin (Buehler, Germany).After 24 h, the samples were cut and polished in the coronal plane.Polishing was performed using silicon carbide paper (320,60 0, 120 0, and 250 0 grit), and aluminium oxide slurries (0.3 and 0.05 μm particle size) on neoprene cloth (Buehler, Germany).The polishing protocol led to a final surface roughness of ∼0.2 μm.Nanoindentation was then performed at room temperature using an Anton Paar system (UNHT 3 , Anton Paar GmbH, Switzerland).A Berkovich diamond tip was employed as it has been widely used for indentation of hard calcified tissues [42][43][44] .Various regions were indented under displacementcontrol to a depth of 2.5 μm at 120 mN/min followed by 20 s hold [ 44 , 45 ].A minimum spacing of 10 μm between the indents was ensured [ 42 , 44 ].The elastic modulus was calculated using the standard Oliver-Pharr method [46] .Here, the Poisson's ratio of the indented tissue and indenter tip were assumed to be 0.3 and 0.07, respectively, with the elastic moduli of the indenter tip being 1140 GPa (based on the manufacturer's data).

Finite element analysis
Three-dimensional models of individual OD were developed from microCT data.CT images were reconstructed in Avizo image processing software where a 3D surface model of each OD was transformed into a meshed solid geometry.The meshed geometries were imported into finite element software (ANSYS, PA, USA).The models were meshed using SOLID187 tetrahedral elements (10 node elements with quadratic displacement behaviours), that are well-suited for modelling irregular geometries.Mesh convergence was carried out, with the final models having approximately 1,0 0 0,0 0 0 elements for C. zebrata , H. suspectum , P. apodus , and V. komodoensis , and about 50 0,0 0 0 for T. lepidus .ODs were assigned isotropic material properties with an elastic modulus of 10 GPa and Poisson's ratio of 0.3.The exact values of the aforementioned properties were of less importance given that the relative comparison of mechanical performance between the OD morphologies was the main focus of the FE analysis.Three-point bending was modelled in two different planes (antero-posterior and transverse) as shown in Fig. 7 .Here ODs were constrained at L/8 of the OD while a total load of 1 N (5 × 0.2 N) was applied at L/2 of the OD.Bending rigidity (EI) was then measured using the follow-ing equation: EI = P L 3 / 48 ω 0 where P is the applied load, L is the length between two supports, and ω 0 , is the total deflection at the point of loading.It is worth noting that our fundamental understanding of how ODs are loaded in vivo is limited; here we loaded the ODs in three-point bending as, given their wider arrangement within the skin, we considered that they are likely to be loaded in this mode in vivo .Nonetheless, the main focus of the analysis performed here was to use computational models to isolate morphological variations observed between the ODs and then investigate the impact of this parameter on OD bending rigidity.

Morphological and histological analysis
Morphological analysis revealed considerable diversity in the shape, size, and distribution of ODs across the species investigated ( Fig. 2 ).T. lepidus was the only species in which ODs were limited to the head, whereas the other four species had ODs over most of the head and body.Whereas the ODs of most species are singular elements, those of C. zebrata (as for most scincid lizards) are compound: each OD is composed of several smaller elements or osteodermites that are linked via fibrous joints.As shown in Fig. 2  ODs demonstrate species-specific variability in size, shape, and arrangement.Whereas C. zebrata ODs are broadly plate-like and overlapping, those of T. lepidus more closely resemble juxtaposed polygons, creating a mosaic-like arrangement.ODs in H. suspectum are also non-overlapping, but they are notably smaller than those of T. lepidus , with a roughly circular or hexagonal shape in dorsal profile.The ODs of P. apodus are trapezoidal and overlapping, whereas those of V. komodoensis resemble small cylinders, a shape referred to as vermiform.In addition to shape, ODs also differ in the presence and arrangement of internal cavities ( Fig. 2 B).The histological analyses revealed unexpected microstructural diversity, including variation in tissue composition and arrangement ( Fig. 3 ).As expected, ODs from all species were dominated by bone.However, the fibrillar organisation of the mineralised matrix, along with the inconsistent presence of a denser capping tissue (discussed below), varied between species.
C. zebrata ODs consist of a series of plates (osteodermites), each of which is composed of a combination of woven-fibred, Sharpeyfibred, and lamellar bone.Woven-fibred and Sharpey-fibred bone was generally associated with the periphery of each osteodermite, where adjacent elements were anchored to each other as well as to the surrounding dermis.Lamellar bone was characteristic of more central regions of each osteodermite surrounding the large marrow cavities and secondary osteons.Compared to the ODs of C. zebrata , those of T. lepidus demonstrated a less ordered microstructure, with numerous large penetrating Sharpey's fibres passing deep into each element, surrounded by a woven-fibred matrix.In section, H. suspectum ODs reveal a thick base of bone overlain by a distinct vitreous capping layer.The bony base is dominated by Sharpeyfibred bone, anchoring each element in the dermis.Neurovascular canals that invade the ODs are lined with a thin layer of lamellar bone.Capping the bony base is a thick layer of collagen-poor tissue with scarce cells, known as osteodermine [ 10 , 13 , 18 ].ODs from P. apodus also develop a collagen-poor capping tissue, although it remains unclear if this tissue is homologous with the osteodermine characteristic of H. suspectum (highlighted in Figs. 3 & 4 with ' * ' in the H. suspectum and P. apodus sections).Deep to the capping tissue, P. apodus ODs are composed of Sharpeyfibred and lamellar bone.V. komodoensis ODs lack a capping tissue.In section, they have a concentric organisation with a core of woven-fibred bone, surrounded by parallel-fibred and lamellar bone.
At the scale of the whole sheet of OD level, cross-sectional analysis at the level of the frontal bone ( Fig. 5 ) highlighted that ODs of C. zebrata are loosely connected to the underlying bone in this region.However, ODs of H. suspectum and P. apodus are fully fused to the frontal bone, although the degree of fusion increases with age in Heloderma (they can be detached from the skull with ease in young animals -personal observation).T. lepidus ODs are in the temporal region and across the supraorbital region.Their attachment to the parietal and frontal was not captured in the present study.However, our unpublished data show that there appears to be a thin layer of soft tissue present at the interface between ODs and skull roofing bones.

Material characterisation
Examination of the mechanical properties of H. suspectum and P. apodus ODs revealed that their mineralised capping tissue (highlighted in Fig. 4  from those of the other mineralised tissues in the skull, jaw, or ODs.The capping tissue displayed a higher elastic modulus and was much more similar to tooth enamel ( Figs. 5 & 6 ).The inherent mechanical properties of the remaining regions of the ODs (i.e., except the capping tissue), despite some regional variability, were similar to those of the frontal and dentary bones from the same species ( Table 1 ).An unexpected finding was the distinct difference in the microstructure of the frontal and dentary bones between the species examined ( Figs. 5 & 6 ).Transverse sections of the frontals of C. zebrata and T. lepidus display a sandwich-like structure with compact dorsal and ventral layers separated by cancellous trabecular bone.A similar sandwich-like arrangement is present in H. suspectum and V. niloticus with central cavities but without the trabeculae.Interestingly, the frontal of P. apodus is strikingly different, with a relatively homogenous arrangement of compact bone throughout.Similarly, the dentaries of both C. zebrata and T. lepidus contain a significant number of cavities, compared to the compact bone of H. suspectum , P. apodus , and V. niloticus .The latter species also showed clearly folded plicidentine connecting the teeth to the jaw ( Fig. 6 ).

Finite element analysis
Given the morphological and biomechanical differences between the ODs of the species studied, finite element analysis was carried out ( Fig. 7 ) to assess the impact of morphological variation on the bending rigidity of individual ODs, using a three-point bending test.For this analysis we focussed solely on overall shape, and differences in bone microstructure or material properties were not considered (although they would, of course, be expected to be significant).As expected, the thin plate-like OD of C. zebrata had the lowest bending rigidity but, less predictably, that of a T. lepidus OD had the highest.Whereas the ODs of T. lepidus and P. apodus were of relatively similar size, those of P. apodus had an almost 60% lower bending rigidity.This could be explained by the fact that ODs of P. apodus are of lower height (thickness) and volume compared to T. lepidus ODs, and also because the harder capping tissue [47] was not modelled in this analysis.H. suspectum ODs showed a very similar bending rigidity across both axes (but the absence of the hypermineralised capping tissue in this test will have weakened the OD significantly [18] ), whereas those of V. komodoensis , as expected from their cylindrical structure, showed the greatest difference in terms of the bending rigidity between the longitudinal and transverse axes.For the other three species, bending rigidity was relatively similar across both axes ( Fig. 7 C).It must be noted that bending rigidity is associated with strain distribution in the ODs.This is highlighted in Fig. 7 , which illustrates that ODs with lower bending rigidity show higher levels of von Mises strain.

Discussion
This work provides insights into the remarkable diversity of lizard ODs and raises interesting questions about the factors driving these differences through their evolutionary history.The selective pressures driving the morphological diversity of lizard ODs remains largely unknown.Although OD probably affect the overall strength and mechanical properties of the skin [ 47 , 48 ], their presence, distribution, and diversity are likely to result from a combination of natural selection and developmental constraints.Similarly, the arrangement of internal cavities within the ODs probably reflects differences in vascularity and the development of marrow [ 8 , 13 ], and it may play a role in thermal regulation of the body.
The variation observed in frontal bone microstructure in the species studied is interesting and unpredicted.One potential explanation for the differences in frontal architecture between species might be a difference in feeding strategy or head use, and therefore in the degree and pattern of strain distributions across the skull [49][50][51][52] .From a biomechanical viewpoint, a thin solid frontal has lower bending rigidity than a thicker hollow frontal and is also  2) to reduce the level of mechanical strain experienced by the frontal and other adjacent bones arising from biting [53] , or fighting.A similar explanation has been proposed for the fusion of cranial sutures (in e.g.lizards) where fusion can at least locally reduce the level of strain on the adjacent bones [ 54 , 55 ].This could be another example of a highly optimised structure within the musculoskeletal system [56][57][58] .These findings, despite their preliminary nature, suggest that bone microstructure -of both ODs and skull elements -deserves a more comprehensive analysis as there may be a potential correlation between bite force, diet, and ecology.Computational models highlighted the crucial role of morphology on the bending rigidity of the individual ODs.As expected, the thin plate-like OD of C. zebrata had the lowest bending rigidity and the thick OD of T. lepidus had the highest bending rigidity.Aside from the bending rigidity of the single ODs, it must be noted that the local balance of flexibility, protection, and reinforcement of the adjacent structures could also be achieved by variation in OD density and arrangement across the whole OD sheet, irrespective of the overall size and morphology of individual ODs [49] .Loosely connected ODs increase the overall flexibility of the OD sheet or its underlying structure (perhaps reflecting locomotion) while maintaining a degree of protection, most likely against conspecifics.
This study represents a first step in assessing the structural and biomechanical diversity of lizard ODs.Given that we analysed only five species out of over 30 0 0 OD-bearing lizard taxa, it would be premature to speculate on the evolutionary significance of our findings or to draw general conclusions on the roles of ODs across lizard diversity.Nonetheless, we consider that our findings raise interesting questions and provide a strong foundation for further work on the microstructure and material properties of both ODs and skull bones.In the longer term, this will facilitate the exploration of the possible correlations between the presence, shape, and density of ODs and, for example, bite force, sprint speed, body size, lifestyle, and predation.Equally, although these results offer some insights into the functional and biomechanical adaptations of ODs, the extent to which their diversity is driven by genetic variation and developmental factors remains largely unknown.

Fig. 1 .
Fig. 1.A time-calibrated phylogeny of the species used in this study with approximately scaled silhouettes for representative adult whole body appearance with marked positions of single OD extractions (orange ovals -dorsal neck sampling for species where OD are present at that location, and marked as cranial OD for Timon lepidus ).Photographs are showing external cranial appearance of the lizard species sampled.The phylogeny was created from data in Tonini et al. [59] using R studio and phytools (R Studio Team, 2015) [60] .Silhouettes are credited to the acknowledged artists under CC or Public Domain Dedication licenses, either via phylopic.orgor by one of the authors (CW).Photographs are credited to the acknowledged photographers and used under CC license conditions from https://search.creativecommons.org/: The figure was assembled in Inkscape.

Fig. 2 .
Fig. 2. Morphological variation in the ODs of five lizards.(A, B) show the lateral (A) and dorsal (B) views of a single OD.In (B) one half of each OD, uncoloured, has been rendered transparent to show any internal marrow cavities.(C) shows the various measurements undertaken and a summary of the quantified parameters on a single OD representative of the study species.

Fig. 3 .
Fig. 3. Histological sections of ODs from the represented species with the epidermal surface dorsally ( = uppermost).Left-side columns are stained using Haematoxalin and Eosin, and right-side columns using Massons' trichrome.Images show an overall view of the OD (scale bar 250 μm) and matching higher magnification panel (scale bar 50 μm).Sagittal views are shown for C. zebrata , and P. apodus to highlight the overlapping rows of OD and were also used for H. suspectum and T. lepidus ; the OD of V. komodoensis is shown in transverse section to highlight the different fibre types.The OD is outlined by dashes in the H&E overview, and capping tissues are highlighted with ' * ' in H. suspectum and P. apodus sections.

Fig. 4 .
Fig. 4. Cross-section of ODs highlighting the mechanical properties of different regions of ODs in different species.Areas of nanoindentation are shown (black triangles) and were used to provide a mean ± standard deviation for each circled area of the section.

Fig. 5 .
Fig. 5. Frontal sections of the skull in the five study species highlighting the mechanical properties of different regions of the section.Note: (1) T. lepidus and P. apodus sections were originally made for half of the skull and were virtually reconstructed to illustrate a complete section whereas the other sections are natural transverse sections of the entire frontal bone.(2) Areas of nanoindentation are shown (black triangles) and were used to provide a mean ± standard deviation for each circled area of the section.

Fig. 7 .
Fig. 7.Characterising the bending rigidity of the ODs in two planes (A, B) shows the pattern of von Mises strain during three-point bending.(C) is a summary of the bending rigidity of the ODs.Note that all ODs were rendered with the same material properties so that the focus was on the effect of shape alone.