Dietary adaptions in the ultrastructure of dinosaur dentine

Abstract

Teeth are key to understanding the feeding ecology of both extant and extinct vertebrates. Recent studies have highlighted the previously unrecognized complexity of dinosaur dentitions and how specific tooth tissues and tooth shapes differ between taxa with different diets. However, it is unknown how the ultrastructure of these tooth tissues contributes to the differences in feeding style between taxa. In this study, we use third harmonic generation microscopy and scanning electron microscopy to examine the ultrastructure of the dentine in herbivorous and carnivorous dinosaurs to understand how the structure of this tissue contributes to the overall utility of the tooth. Morphometric analyses of dentinal tubule diameter, density and branching rates reveal a strong signal for dietary preferences, with herbivorous saurischian and ornithischian dinosaurs consistently having higher dentinal tubule density than their carnivorous relatives. We hypothesize that this relates to the hardness of the dentine, where herbivorous taxa have dentine that is more resistant to breakage and wear at the dentine–enamel junction than carnivorous taxa. This study advocates the detailed study of dentine and the use of advanced microscopy techniques to understand the evolution of dentition and feeding ecology in extinct vertebrates.

1. Introduction

Studies of the dentition of fossilized animals are becoming increasingly common in recent years given the insights that can be gained in understanding food processing [1–3], diversity in ecological guilds [4–7], intra and interspecific communication [8], development [9–11] and large-scale evolutionary patterns [12–17] in extinct groups. A large focus has been on dinosaurian dentitions, given the recent discoveries of unique developmental patterns [11] and arrangements of tooth tissues [1–3,16] that allow for specialized feeding in the group. Most dinosaurs are characterized as polyphyodont [18], and repeatedly develop new teeth as old teeth are shed throughout life. As such, teeth are common in the fossil record, and are especially crucial for evolutionary studies of these ancient tetrapods.

In practice, more attention has been paid to the study of enamel in fossil animals [19–23] rather than dentine because of obstacles faced when imaging microstructures, despite evidence that the structure of dentine contains a functional and ecological signal, especially in dinosaurs [1–3,16,24,25]. Enamel, as the outer layer of tissue in a tooth, often forms complex structures to resist wear when processing food. During tooth mineralization in vertebrates, the hard enamel tissue is formed outward from the dentino–enamel junction (DEJ), while the relatively softer dentine is formed inwards towards the pulp cavity. Dentine is deposited first during tooth development [26] and remains vital throughout the life of the tooth, contrary to the enamel, which contains almost no organic content after mineralization, thereby preserving information on the earliest stages of tooth development, growth, maintenance and function [27]. By investigating in great detail the DEJ and the tissues adjacent to it in different taxa, we can gain new insights into the interactions between teeth and behaviour, and place these new insights into an evolutionary context.

The function of a tooth is related to its durability. The outer portion of a tooth is strongly influenced by the coupled biophysical characteristics and structural features of the dentinal tubules around the DEJ where the enamel, the hardest bio-mineral tissue, meets the relatively soft dentine [28,29]. Dentine structure and hardness can differ depending on where it is deposited within the tooth [2,3], and can also vary in hardness based on the amount of peritubular and intertubular dentine present [30].

Recent analyses of the tissue complexity of dinosaur teeth [1–3,16] reinforce that differences in tooth structure are related to tooth function and diet. Carnivorous dinosaur teeth have a layer of dentine at the DEJ that is sometimes atubular and globular, which is the primitive condition for archosaurs [1,16]. On the other hand, ornithischian dinosaurs lack this distinct globular layer of mantle dentine, and often have extensions of the dentine tubules into the enamel [1,16]. Finite-element analysis (FEA) of carnivorous dinosaur teeth suggest that the globular mantle layer is important for protection from fracture while processing prey [16], while palaeo-tribology studies of herbivorous dinosaur teeth suggest that the dentine at the DEJ is harder than the bulk dentine of the tooth in order to resist wear while grinding plant material [31]. However, the role of the dentine tubule and how it relates to hardness at the DEJ has not been examined in detail.

Recent advances in imaging modalities in fossil material [17] now allow for in-depth examination of the ultrastructure of tooth tissues. Third harmonic generation (THG) microscopy, typically used for non-invasive biomedical imaging [32,33] has been shown to work well in fossil crocodilian teeth [17] because the fossilization process creates a high contrast between the dentine tubules and the intertubular dentine. Additionally, THG microscopy is non-invasive and yields sub-micron-level detail of anatomical structures, making it an excellent method for providing new information on tooth structure.

In this paper, we utilize scanning electron microscopy (SEM) and THG microscopy to examine the ultrastructure of dentine at the DEJ and to quantify differences in dentine structure between dinosaurian taxa using morphometrics for the first time. Our results suggest that herbivores have a higher density of small tubules near the DEJ, while carnivores have larger tubules at a lower density, reflecting the different requirements of tooth structure for processing plants and prey items in each group, respectively. These differences in tooth microstructure may be useful for inferring the diet of some dinosaurs where the diet is unknown.

2. Material and methods

Teeth from a variety of dinosaurian taxa were examined for this study. The saurischian taxa examined include two teeth of the theropod Sinosaurus, RockHound Museum Taiwan (RHM) 030111, 030112, from China (approx. 195 Ma); and one tooth from all remaining taxa: the theropod Carcharodontosaurus, Royal Ontario Museum (ROM) 52037, from the Cenomanian Kem Kem beds of Morocco (approx. 95 Ma); the sauropod Rebbachisaurus, RHM 101034, from the Cenomanian Kem Kem beds of Morocco (approx. 95 Ma); the theropod Troodon formosus, ROM 5089, from the Campanian Oldman Formation, Alberta (approx. 76 Ma); and the theropod Tyrannosaurus sp., RHM 100801, from North America (approx. 67 Ma). The ornithischian taxa examined comprised one tooth from each of an unidentified hadrosaurid, ROM 58205, from the Campanian Oldman Formation, Alberta, Canada (approx. 76 Ma); Triceratops sp, ROM 67669, from the Maastrichtian Hell Creek Formation of Montana, USA (approx. 67 Ma); and Edmontosaurus sp., ROM 67672 from the Maastrichtian Hell Creek Formation of Montana, USA (approx. 67 Ma). Comparative crocodilian material, as described and measured in Chen et al. [17], includes fossil Alligator mississippiensis, ROM 61452, from Florida, USA (approx. 1.5 Ma), extant crocodilian, Alligator mississippiensis, ROM R4402, from Florida, USA, and an unidentified crocodilian, ROM 67512, from the Cenomanian Kem Kem beds of Morocco (approx. 95 Ma).

2.1. Multiple harmonic generation microscopy

The multiple harmonic generation microscopic imaging (second harmonic generation (SHG) and THG, and two-photon fluorescence (TPF)) of the fossil teeth was performed with a custom-built femtosecond Cr:forsterite laser centred at 1230 nm with a 140 fs pulse width at a 110 MHz repetition rate [32,33] The laser was chosen at this wavelength because of much reduced attenuation and invasiveness inside the specimens. Also, the wavelength of the generated THG signals is required to be longer than 400 nm to avoid the high ultraviolet (UV) absorption in most materials; therefore, the original excitation wavelength should be longer than 1200 nm. To investigate tooth structures through natural and cut surfaces, an epi-collection geometry was used to collect higher harmonic signals, as explained in the set-up schematic of the epi-HGM (figure 1). The 1230 nm IR laser beam was first shaped by a telescope and then directed into a modified beam scanning system (Olympus Fluoview 300). After passing through a pair of galvanometer mirrors and a microscope system (Olympus BX-51), the scanning beam was focused on the specimen. For the demonstration of non-invasive three-dimensional dentinal reconstruction, the laser was focused directly on the natural surface (enamel) of the fossil tooth (figure 1b) by a water-immersion objective with a NA of 0.9 (LUMplanFL/IR 60X/NA 0.9). Additionally, DEJ images were obtained by focusing the laser beam on thin sections of various fossil teeth (figure 1). An average illumination power of 50 mW was applied to the sample surface during all measurements. The excited SHG and THG signals were backward-collected by the same objective. Finally, the collected signals were divided into SHG (615 nm) and THG (410 nm) by a beamsplitter and then sent into two photomultipliers (PMT) with 615 and 410 nm narrowband interference filters in front (Semrock Inc. BP615/10 and BP410/70), separately. The thickness of the two-dimensional virtual sections is around 1 µm, which is determined by the axial resolution of the system [33]. For the measurement of TPF, the BP615/10 filter was substituted with a BP645/50 bandpass filter (Semrock Inc.) and measured at the same section as the SHG signal, repeatedly. With ease of computer processing (Flowview300), the SHG, THG, and TPF images (mainly THG in this paper) could be obtained and represented by red, yellow, and blue pseudocolours, respectively.

Figure 1.

Scheme of the imaging system. (a) Schematic diagram of the multiple harmonic generation microscopy system. The centre wavelength of the Cr:forsterite laser was 1230 nm with a pulse width of 140 fs. As most fossils are quite thick and non-transparent, our system was built to receive the backward emitted nonlinear signals from the samples by using the same objective as the one in the illumination path. Nonlinear signals, including third harmonic generation (THG), two-photon excitation fluorescence (TPF) and second harmonic generation (SHG) were further separated into different channels by dichroic beam splitters and directed into individual photomultiplier tubes for non-descanned detection. Bandpass interference filters were inserted to filter out undesired wavelengths. As the laser beam took a raster scan of the specimen point by point, the data acquisition card recorded the excited nonlinear signals to form two-dimensional images via computer mapping, and each channel can be presented with different pseudocolours on screen. (b) Schematic diagram showing the detailed arrangement between the water-immersion objective and the studied fossil tooth. In this study, fossil teeth were either scanned from the intact surface or cut (thin-section) surface for SHG, THG or TPF microscopic examination.

The three-dimensional images were reconstructed by stacking a series of images in ImageJ (1.47v, NIH) obtained at different depths in a tooth of Sinosaurus. The average step size was 0.4 µm and the penetration depth was from 80 to 150 µm depending on the usage of the sample. The image stacks were transformed into a three-dimensional model with the three-dimensional viewer plugin.

For the specific study of dentinal structures, some fossils were sliced and prepared into thin sections following standard palaeohistological procedures [34]. These specimens were embedded in Castolite AP polyester resin, placed under vacuum and left to dry for 24 h before being cut transversely at several places along the height of the tooth using a Buehler Isomet 1000 wafer blade low-speed saw. Cut specimens were mounted to glass or Plexiglas slides using Scotch-Weld SF-100 cyanoacrylate. Specimens were ground down to approximately 180 µm thick using a Hillquist grinding cup, then ground by hand using progressively finer grits of silicon carbide powder. The highest possible polish was achieved using 1 µ grit aluminium oxide powder.

2.2. Scanning electron microscopy

Three dinosaur tooth specimens were examined with SEM: the ornithischian Triceratops sp., ROM 67669, from the Maastrichtian Hell Creek Formation of Montana, USA (approx. 67 Ma); and two saurischians: cf. Gorgosaurus sp., ROM 57981, from the Dinosaur Park Formation of Alberta, Canada (approx. 70 Ma) and Tyrannosaurus rex, ROM 66108, from the Hell Creek Formation of Montana, USA (approx. 67 Ma). The teeth were cut in several planes and etched in 5% HCl for 10 s and were examined with a Jeol Neoscope JCM-5000 SEM at the University of Toronto Mississauga. This SEM does not require sputter coating. Images of the dentine were taken directly below the DEJ.

2.3. Morphometric analysis

The dentinal tubule images collected in this study using THG microscopy were processed using ImageJ. The tubule diameter (2.3.1), branching density (2.3.2) and tubule density (2.3.3) were measured from two-dimensional images (i.e. figure 2).

Figure 2.

Two-dimensional THG microscopic images of the DEJ from thin sections (180 µm thick). (a) Triceratops, (b) Edmontosaurus, (c) unidentified hadrosaurid, (d) Rebbachisaurus, (e) Troodon, (f) Tyrannosaurus sp., (g) Carcharodontosaurus and (h) Sinosaurus. All scale bars, 20 µm.

2.3.1. Tubule diameter measurements

The two-dimensional images were rotated so that the tubules were oriented vertically. To sample the diameter of the tubules, five horizontal lines were drawn on the images evenly. The maximum diameters of dentinal tubules that crossed horizontal lines were measured manually and averaged.

2.3.2. Branching density measurements

To sample branching density, the images from (2.3.1) were overlain with 5 × 5 grid lines (Grid ImageJ plugin, http://fiji.sc/Grid). The number of branching points in the gridded image were counted within the centres of the squares to avoid distortion or fading along the edges of the images. The branching points within each grid were marked with white spots, and then the number of spots within each grid was calculated by using the Analyze Particle function. The density of branching spots was recorded as number of branching points per 100 µm² and averaged.

2.3.3. Tubule density measurements

The dentinal tubules in the two-dimensional images were identified with the Auto Threshold function in ImageJ. After testing all methods in Auto Threshold, the Renyi Entropy method was used for Rebbachisaurus and the Huang method for the rest, due to the contrast and dentinal tubule nature under THG. The images were cropped manually to avoid any parallax in the sample. Tubule density was measured as a percentage of tubule area within the image divided by the total area of the image. Measurements were taken in three imaged areas in one tooth and averaged.

3. Results

3.1. Third harmonic generation microscopy and scanning electron microscopy for visualization of fossilized dentinal microstructures

Our results indicate that, in all specimens, THG microscopy consistently provides highly informative two-dimensional micro-anatomical images of dentine tubules at the DEJ and consistent THG signals with a high lateral resolution around 400 nm (figure 2). The dentine tubules differ between each species, with different sizes, branching and densities (table 1; electronic supplementary material). No THG signals were observed in the enamel due to the lack of enamel rods and the uniform crystalline enamel structure. The multimodal image of Sinosaurus (figure 3) reveals that under any type of signal, no dentine tubules are seen directly below the DEJ, which matches previous descriptions of a distinct atubular, globular layer of mantle dentine observed with conventional light microscopy and synchrotron X-ray transmission microscopy in theropod dinosaurs [1,16]. SHG signals are weakly visible in the fossil Sinosaurus tooth (as compared to teeth from extant animals which are much stronger [17] due to the strain-induced breakage of the 6/m point group symmetry [35]).

Table 1.

Measurements used for morphometric analysis.

species	diameter (µ)	branching (no. per 100 µm2)	density (%/image)
Alligator mississippiensis	0.99	2.10	38.84
Alligator mississippiensis (1.5 Ma)	0.94	2.21	37.35
Kem Kem crocodile	0.93	2.20	35.74
Sinosaurus sp.	0.92	1.13	37.41
Carcharodontosaurus saharicus	0.95	0.80	38.56
Troodon formosus	1.08	0.80	38.99
Tyrannosaurus sp.	1.57	0.91	34.17
Rebbachisaurus sp.	1.04	0.44	43.13
Triceratops sp.	1.01	1.95	45.06
Edmontosaurus sp.	1.02	1.26	42.24
hadrosaurid indet.	0.94	1.11	43.78

Figure 3.

Multimodal imaging under different nonlinear optical processes in a Sinosaurus tooth. The two-dimensional tomographic image of the fossil tooth was acquired under (i) two-photon fluorescence microscopy (TPF) (blue), (ii) second harmonic generation (SHG) microscopy (red), and (iii) third harmonic generation microscopy (THG) (yellow). The SHG is mainly the result of strain effects. (iv) An integrated image of the three channels. All scale bars, 40 µm.

The three-dimensional micro-morphological images of Sinosaurus reveal in detail the morphology of the dentine tubules directly below the DEJ (figure 4), despite the fact that the depth resolution (approx. 1.5 µm) [36] is not as good as the lateral resolution (approx. 0.4 µm) [36,37] in THG microscopy. The three-dimensional reconstruction reveals in detail the curvature of the dentine tubules towards the DEJ, followed by a slight thickening of the tubules and then a narrowing towards the pulp cavity, concomitant with reduced branching (figure 4).

Figure 4.

Non-invasive three-dimensional THG microscopy of a Sinosaurus tooth. (a) Three-dimensional reconstructions of the DEJ obtained from different depths of the cut surface of an intact tooth by stacking 136 THG optical sections (step size, 0.45 µm). (b) Enlarged three-dimensional-stacked image showing dentine tubule morphology with little branching (10 µm depth). (c–e) Three-dimensional histological reconstruction of an intact tooth visualized with an optical penetration depth greater than 150 µm by stacking 402 THG optical sections (step size, 0.40 µm). (c) Oblique view. (d) View looking from enamel towards dentine. (e) Inset of box in (d) showing dentine tubule anatomy with wavy tubule tips at the DEJ. Scale bars (a,c), and (d) 25 µm, (b) and (e) 8 µm.

The SEM images reveal the structure of the dentine below the DEJ (figure 5). All specimens show varying amounts of peritubular and intertubular dentine. Some specimens show a fossilized infill within each tubule (figure 5c), which explains that the bright THG signals in fossilized dentine tubules could only result from the nonlinear susceptibility χ⁽³⁾ of the mineral infillings and the interfaces created by the different properties and sizes of the mineral particles in the tubules [17]. A high amount of branching is present in Triceratops (figure 5b).

Figure 5.

SEM images of dentine tubules below the DEJ. (a) cf. Gorgosaurus sp., empty tubules with peritubular dentine. (b) Triceratops dentine tubules with high amounts of branching, peritubular dentine and some fossilized infill in tubules (arrows). (c) Tyrannosaurus rex dentine showing fossilized infill (arrows) in the dentine tubules. All scale bars, 20 µm.

3.2. Morphometric analysis of dentine tubules

The two-dimensional morphometric data gathered through THG microscopy (figure 2) were used to compare tubule density, size and branching between all taxa examined in this study. Three crocodilian taxa were used as the outgroup to Dinosauria [17]. The bivariate plot of mean tubule branching against tubule density (figure 6) reflects dietary differences among species. The basal carnivorous saurischian theropod Sinosaurus, and its three derived relatives (Tyrannosaurus, Troodon, Carcharodontosaurus) cluster closely together, while the herbivorous saurischian sauropod Rebbachisaurus is distinctive in having an unusually high tubule density. The hadrosaurid ornithischians also cluster closely together, with tubule density highest in the ceratopsid ornithischian Triceratops. All herbivorous dinosaurs appear to have similar tubule densities, which are higher than their carnivorous relatives. The diameter of the circles in the regression plot reflects the average size of tubule diameter, indicating that the highly specialized tyrannosaurid has the largest tubule diameter among all taxa (figure 6a). Overall, dentinal tubule size is conserved within ornithischian dinosaurs, and is more variable in the carnivorous dinosaurs. The three closely related species of crocodilians are similar in branching and density of the dentinal tubules [17], and have higher branching than all other taxa examined.

Figure 6.

Results of morphometric analysis of dentinal tubules. (a) Bivariate plot of dentinal tubule density mean, tubule branching mean, and diameter. The sizes of the coloured circles are representative of tubule diameter. The colours of the circles match the taxa in the phylogeny (b), which contains all taxa examined in this study. (a) Archosauria; (b) Crocodylia; (c) Dinosauria; (d) Ornithischia and (e) Saurischia. Ma, millions of years.

4. Discussion

THG microscopy is a powerful new tool for high resolution and accurate morphometric quantification of dentinal structures in fossil taxa. The discovery of distinct and analysable patterns of variation in the density, branching rate and diameters of dentinal tubules through THG provide the basis for gaining not only new insights into the evolution of particular groups of vertebrates, but also the functional design of teeth in association with distinct feeding behaviours [17].

This study represents the first successful imaging of dentinal tubules in dinosaurs while maintaining sub-micron-level resolution using THG microscopy. The results of Chen et al. [17] and the morphometric analysis presented here show that little modification of dentinal structures takes place during the fossilization process, as the three species of crocodilians spanning 100 million years of time are highly similar (figure 6). Therefore, the structures viewed in this study can be interpreted as the original hard tissue anatomy of the dinosaurs with little distortion occurring during fossilization.

On average, the herbivorous taxa (the ornithischians and the saurischian Rebbachisaurus) have higher densities of narrow dentinal tubules near the DEJ than the carnivorous taxa, which have larger tubules at lower densities (e.g. Tyrannosaurus, figure 6), regardless of phylogenetic relationships. A high density of narrow tubules alters the microanatomy of the dentine, as more peritubular dentine would be present with the presence of more tubules. As peritubular dentine is harder than intertubular dentine [38], an increase in dentine tubules would increase the hardness of the dentine directly below the DEJ. This agrees with reports of elevated hardness in mantle dentine in hadrosaurs [2] and ceratopsians [3], even though mantle dentine is typically softer than primary dentine [39]. Therefore, the derived condition of hard dentine at the DEJ in ornithischians is a combination of the lack of globular mantle dentine and the increased density of narrow dentine tubules. The increase in narrow tubules and likely increase in hardness in the sauropod Rebbachisaurus is surprising, given that a globular layer of dentine has been reported in other sauropods [16]. However, the presence of enamel spindles in sauropods [16] shows that dentine tubules do reach and pass the DEJ. This suggests that a more detailed analysis of sauropod tooth microanatomy may reveal more unique dental adaptations in this dinosaur group, probably related to the high levels of tooth wear that affect this clade [10].

The differences in dentine ultrastructure between herbivorous and carnivorous dinosaurs and differences in hardness near the DEJ are related to the way these animals processed food. The hardness and elasticity of dentine can change depending on the angle of the dentine tubules within the dentinal matrix [30]. Considering the unique wear patterns formed in ornithischian tooth batteries and the increased hardness of dentine at the DEJ, ornithischian teeth were well adapted for grinding plant material [2,3,11]. The elastic properties of mantle dentine [39], absent in ornithischians but present in saurischians, probably protected the teeth of carnivorous theropods from fracture during feeding on prey items. This is supported by FEA modelling of theropod dinosaur teeth with and without mantle dentine [16]. The distribution of globular mantle dentine within a single theropod tooth suggests different hardnesses throughout the tooth, where only the tips of denticles are strengthened against wear by lacking globular dentine and having enamel spindles [1].

The results of this study can be combined with information on tooth shape (e.g. denticle size and density in theropod dinosaurs) and palaeo-tribology studies [31] to infer diet in dinosaurs with contested feeding habits. For example, the saurischian dinosaur Troodon has been postulated to be herbivorous, omnivorous or carnivorous [40,41]. Based on the structure of the dentine near the DEJ observed in this study, Troodon is more similar to the highly carnivorous theropods Carcharodontosaurus and Sinosaurus than to the strict herbivores. However, the overall shape of Troodon teeth and denticles is significantly different from other theropods [41], suggesting a difference in diet and food processing. A consideration of the macro- and microstructure of the dentition indicates that Troodon probably had a diet high in softer prey items and was not processing tough plant material.

5. Conclusion

THG microscopy is a promising new methodology for investigating dentinal ultrastructures in extinct animals. Fossilized teeth are readily studied using THG microscopy because of the nonlinear susceptibility χ⁽³⁾ of the mineral infillings in dentine tubules, thus, the marriage of modern techniques with palaeontological resources is yielding superb results in understanding patterns of dental development and evolution. Future studies will include morphometric analyses of dentinal structures among members of specific taxonomic groups, which can provide valuable new insights into transitions from one feeding strategy into another, or the evolution of new tooth shapes. This methodology can also be applied to different groups of animals to infer dietary preferences. This study reinforces the usefulness of studying dentine to understand the evolution of dentitions.

Data accessibility

For questions on MHGM methodology, please contact C.-K.S. (sun@ntu.edu.tw); for questions on palaeobiology please contact K.S.B. (brinkkir@dentistry.ubc.ca) or R.R.R. (robert.reisz@utoronto.ca).

Authors' contributions

This work was conducted by C.-K.S. and R.R.R. All authors contributed to the research. Optical experiments, plotting and analyses were performed by Y.-C.C., and W.-M.L., under C.-K.S. supervision. D.-B.S., Y.-N.W. and K.S.B. performed the digital morphometric analyses on dental histology. K.S.B. performed the SEM analyses. R.R.R., K.S.B. and T.D.H. procured the specimens and assisted with specimen preparation and processing. T.D.H. first proposed the application of MHGM for fossils. K.S.B., Y.-C.C., R.R.R. and C.-K.S. wrote paper.

Competing interests

The authors declare no competing interests.

Funding

This work is supported by NSC 102-2120-M-006-003, NSC 101-2314-B-006-048–MY3, and University Advancement under Ministry of Education (Taiwan) for D.-B.S.; the Ministry of Education, Ministry of Science and Technology, and National Health Research Institute of Taiwan, under NHRIEX101-9936EI, and NSC 102-3011-P-002-010 for C.-K.S.; Ministry of Education and NCU (Taiwan), NSERC Discovery Grant (Canada) for R.R.R. and K.S.B.