Tooth shape acquisition
We dissected 63 species that cover both the phylogenetic and dietary diversity of Alethinophidians (Supplementary Material 1 & 2). Our sample is composed of adult or sub-adult specimens from museums (AMNH, MNHN, Jerusalem University) and private collections (details in Supplementary Material 1 & 2). While comparisons of the shape of the teeth from different bones show a significant ecological signal in some clades (Fischer et al., 2022) we here decided to focus only on the dentary bone to ensure functional homology. In snakes, the dentary is involved in capturing, restraining, and manipulating the prey; so, it endures loads related to both prey characteristics (e.g., hardness) and feeding behavior (e.g., holding a vigorous prey, extracting a snail from its shell). The maxilla has a similar function, but it is too functionally and morphologically derived in venomous snakes to be able to make a fair comparison. We isolated the right dentary bone of the specimens and CT scanned it using a Phoenix Nanotom S μCT‐scanner (General Electric, Fairfield, CT, USA) at the Institut de Genomique Fonctionelle, Ecole Normale Superieure (Lyon, France). ForCrotalus adamanteus , Eryx jaculus and Dasypeltis scabra , we scanned the left dentary and mirrored the mesh. The bones were scanned at mid-section to ensure homology in position and as close as possible to the teeth to get the highest possible resolution (voxel size between 0.97-7.50μm), with a voltage of 100kV and a current of 70μA. Only the Python regius tooth was not a mid-section but the first tooth of the dentary (but removing it from the sample barely changed the statistical results). The 3D reconstruction was done using Phoenix datos|x2 (v2.3.0, General Electric, Fairfield, CT, USA) and the subsequent segmentation was performed using VGStudioMax (v1.0, Volume Graphics GmbH, Heidelberg, Germany). We segmented teeth that were not broken or in the process of being replaced, which was easily noticeable on the scans through resorption of the bony attachment in most species. We obtained one tooth per specimen, so 63 dentary teeth. Comparisons between single teeth (from different bones) in phylogenetically diverse samples have demonstrated an ecological signal (Fischer et al., 2022), and the interspecific shape variation is larger than intraspecific variation in our sample, validating the use of one tooth per species (see Supplementary Material 2b).
We used 3D geometric morphometrics to quantify the 3D shape of the teeth. We placed 14 anatomical landmarks: 7 on the outer surface of the teeth, 7 on the pulp cavity surface (inner part of the tooth Supplementary Material 3) and 100 curve semi-landmarks (50 on each layer) using the software MorphoDig 1.2 (Lebrun, 2017). Curves correspond to the anterior and posterior edge of both layers and to the limit of the tooth insertion onto the bone. In addition, 42 and 65 surface semi-landmarks were respectively placed on the inner and outer surfaces of the teeth to obtain an accurate 3D representation of the tooth shape (Fig. 1). This template allows us to obtain information both on their shape and on the thickness of the hard tissue material (dentine and enamel). We placed the anatomical landmarks and curve semi-landmarks by hand on each specimen and checked for the repeatability of the positioning by digitizing these six times in five specimens for which the teeth looked similar. A Principal Component Analysis demonstrated the repeatability of the positioning of our 14 anatomical landmarks and the necessity of using curve and surface semi-landmarks (Supplementary Material 4). We used the ‘Morpho’ package (Schlager, 2015) to project and relax the surface semi-landmarks on each specimen and to slide the curve and surface semi-landmarks while minimizing the bending energy between the specimen and the template (Gunz & Mitteroecker, 2013). We then performed a Procrustes superimposition using the functiongpagen of the ‘geomorph’ package (Adams et al., 2020), the resulting Procrustes coordinates were used to test our hypotheses. The projection, relaxation and Procrustes superimposition were performed using R version 3.4.4 (R Core Team, 2018).
As 3D geometric morphometrics removes size information, we also took linear and angle measurements of the teeth to test their potential link with dietary ecology. We measured the length of the curvature (LC) along with maximal and mean curvature (κmean, κmax) of each teeth using the FIJI (v1.53q) plugin ‘Kappa’ (Mary & Brouhard, 2019). To do so, we used a snapshot of the medial side of each tooth, obtained in GeomagicStudio 2013 (3D Systems, Rock Hill). We digitized the midline to obtain its global curvature (see examples in Supplementary Material 5, raw measurements available in Supplementary Material 1). From the mean and max curvature measurements, we calculated the corresponding degree of curvature (DCmean, DCmax) using the following equation:
\(D_{C}=\left(\ \kappa\ \times L_{C}\ \right)\ mod\ 2\pi\ \times\ \frac{180}{\pi}\)