Integration of multi-level dental diversity links macro-evolutionary patterns to ecological strategies across sharks
- ENS de Lyon, Institut de Génomique Fonctionnelle de Lyon, France
- Laboratoire de Biologie du Developpement de Villefranche-sur-Mer (LBDV, UMR 7009), Institut de la Mer de Villefranche (IMEV), Sorbonne Université, CNRS, France
Figures
Figure 1 with 4 supplements
Overview of heterodonty measures.
(A) Heterodonty, a dentition-level disparity measure, can refer to (1) average differences between successive teeth along the jaw: sequential monognathic heterodonty: HMS, (2) averaged differences among all teeth within the same jaw, total heterodonty: HMT, (3) differences between pairs of teeth belonging to opposing jaws, also termed dignathic heterodonty: HDG. Here, these different measures are partially illustrated using differently dashed arrows. While the upper dentition of Chlamydoselachus anguineus shows fairly similar teeth between upper (turquoise) and lower (red) jaw, the lower dentition of Etmopterus molleri displays conspicuous dignathic heterodonty. (B) Heterodonty measures (Figure 1—figure supplement 1) are correlated. Monognathic heterodonties (HMS, HMT), as well as dignathic heterodonty (HDG), are plotted against each other, revealing positive correlations. Colors encode phylogenetic clades (teal: Squalomorphii, red: Galeomorphii). Dashed lines show linear regression; p-values are based on Pearson’s correlation test.
Figure 1—figure supplement 1
Different methods to quantify tooth morphological distance.
Comparing tooth similarity: Lacking a universally accepted metric of phenotypic difference, we deployed 6 different measures to quantify differences between two given tooth outlines: (i) Euclidean mean distance; the average distance between equally distant points along the tooth outline and the closest points of the second outline, respectively. Note that the measure is only shown for distances from given points on the red onto the blue outline, albeit being actually applied both ways. (ii) Homologous outline distance: we average pairwise distances between a sequence of equally spaced points on the two outlines, that is we compare points with the same numbers/indices. (iii) Superimposed area overlap: similarity is calculated as the ratio between overlapping to non-overlapping area. (iv) Discrete Cosine Fourier distance: The discrete cosine Fourier describing outline shapes yields a number of coefficients defining a sequence of cosine functions with increasing harmonics. Euclidean distances between corresponding coefficients are used to quantify shape distance. (v) Outline angle distance: Assuming that surface angles reflect relevant outline features, we calculate the distance between the functions summing outline angles at an intermediate resolution of outline points (i.e. for 100 outline points each) for two outlines. (vi) Angle Function Discrete Cosine Fourier: We apply the discrete cosine Fourier as in (iv) on the functions used in (v), in an analogous manner. Note that superpositions of the tooth outlines required for (i-iii) are anteceded by a partial Procrustes alignment.
Figure 1—figure supplement 2
Overview of tooth-level complexity measures.
(A) Schematics illustrating the different methods. Different background shades separate different groups of measures that tend to be correlated, Figure 1—figure supplement 3: Cx.exc, Cx.four, Cx.ang. (i) OCR: ratio between outline length and centroid size. (ii) OAR: ratio between outline length and total area. For these methods, high complexity is associated with eccentric shapes with large cusps or protrusions. (iii) OIR: ratio between area of incircle and excircle. Excentric and gracile shapes yield high complexities. (iv) DFS: describing more complex and excentric shapes requires more Fourier harmonics. In the example morphospace, more central shapes featuring smaller coefficient values are more similar to a circle. Thus, this measure does not attribute high values to reiterative patterns, such as equal cusps. (v) OPC: outline angles are discretized according to an absolute coordinate system with different partition numbers. Frequent change of discrete angle numbers along the outline is associated with complexity of outline features. (vi) ANS: sum of angles between outline points for a set of resolutions (here exemplified with different colors), attributing complexity to outline feature density. (vii) ASC: this measure quantifies the difference of outline angle sums between different resolutions, that is it yields high values if different outline features exist on different scales. (viii) AND: diversity of outline angles, that is feature diversity. (B) For each complexity measure, examples of tooth shapes with high and low values are displayed.
Figure 1—figure supplement 3
Correlations between measures.
(A) Species-wise correlations between heterodonty measures for sequential (left) and total monognathic heterodonty (right). Outline similarity measures (EMD, HED, SAO) and, especially, angle-based measures (ADD, OAD), show high internal correlations. (B): Pair-wise correlations for all tooth-level complexity measures, based on data for all teeth. Background shades of gray correspond to defined groups of measures (Figure 1—figure supplement 2). Strong correlations exist particularly between angle-based measures, while the Fourier-based method appears only weakly correlated with the remainder, justifying the assumption of complementarity of methods. For all plots, linear correlation R values are represented by intensity of color.
Figure 1—figure supplement 4
Tooth-level complexity measures are complementary.
(A) Species are ranked based on coarse (Cusp1) and fine (Cusp2) cuspidity. Color corresponds to taxonomic order, in the same sequence as in Figure 2. While both cuspidity ranks appear to be generally correlated, some galean species show high fine cusp numbers (typically: serrations) while maintaining an overall low coarse cuspidity. (B) The average tooth-level complexity (based on a combination of Cx.exc, Cx.four, and Cx.ang) is plotted against the average rank difference when ranks for individual complexity measures are compared, one by one. Several species show substantial differences in complexity depending on the measure applied, again suggesting complementarity of methods.
Figure 2 with 2 supplements
Heterodonty is widespread across all shark clades.
We selected 51 species across the entire Selachimorpha, representing most of the extant shark diversity. Displayed branch lengths are proportional to genetic distance (see Materials and methods) and taxonomic orders are distinguished by background (and font) color. The adjacent heatmap shows species-wise measures of trophic level, tooth-level complexity (average of different measures, Figure 1—figure supplement 2), Fourier-based tooth-level complexity, cuspidity (1: coarse, 2: fine), and globally normalized heterodonty measures, Figure 1, Figure 1—figure supplement 1. HMS/HMT: sequential/total monognathic heterodonty, HDG: dignathic heterodonty, HMTx: maximal heterodonty between any two teeth of the same jaw, Cx: tooth-level complexity.
Figure 2—figure supplement 1
Single tooth discrete Fourier PCA.
We compared individual tooth shapes taking advantage of the discrete cosine Fourier analysis. Here, we show calculated theoretical outlines for combinations of the first four PCs superimposed over binned densities of their respective occupancies across all shark teeth used in the analysis. Higher densities correspond to darker 2D bin color.
Figure 2—figure supplement 2
Heterodonty ratios exhibit differences across shark phylogeny.
Ratios between monognathic and dignathic heterodonty (HMT/HDG) and between maximal monognathic heterodonty and sequential monognathic heterodonty (HMX/HMS, a proxy for graduality, with lower values indicating higher graduality) show different values between shark orders. Here, shark orders are lined up along the y-axis and are marked by colors; the x-axes denote heterodonty ratios using log scales. Notably, Squaliformes exhibit high dignatic and low monognathic heterodonty, while galean sharks show overall more gradual shape change than squalean sharks.
Figure 3 with 4 supplements
Differences in heterodonty show no significant increase with genetic distance for low- to intermediate taxonomic levels, unlike tooth-level complexity.
We ordered all pairs of species by normalized genetic distance (dG) and calculated p-values (one-sided Wilcoxon test) for significance of difference of overall tooth-level complexity (orange), Fourier-based complexity (red), total monognathic heterodonty (HMT, blue), dignathic heterodonty (HDG, dark blue), and total phenotypic distance (gray), between two subsets of 100 species pairs each. Total phenotypic distance is based on position-wise tooth shape comparisons. Subsets were defined as containing the nth to the n+100th species pair ordered by dG, for sliding (incrementally increasing) n. The two subsets were (A) subsequent or (B) 200 ranks apart, in order to account for different scales of comparison. Here, the lines connecting p-values for all n are Bezier-smoothened and plotted against dG of the highest-ranked species pair within the respective lower set. A dotted line marks the 0.05-level of statistical significance. Inlets show the relationship between dG and ordered ranks and examples of two pairs of subsets (higher: red, lower: blue). For explanatory purposes, schematic examples of two pairs of sets with low (red) and high (blue) p-values are displayed beside. (C) For orientation, we display the taxonomic compositions of the ordered species pair sets, with black representing the portion of pairs from the same family (only a few), light gray representing pairs from the same superorder, and white pairs stemming from different superorders, with intermediate shades of gray referring to intermediate taxonomic levels. HMS: sequential monognathic heterodonty, HDG: dignathic heterodonty, Cx: tooth-level complexity.
Figure 3—figure supplement 1
Phenotypic distance increases steadily with genetic distance.
We correlate genetic distance and phenotypic distance for all pairs of shark species. Phenotypic distance is calculated as the normalized Euclidean distance of Fourier coefficients between pairs of teeth in corresponding relative positions along the jaw. (A) Correlation plot with linear regression line (orange dotted line). (B) Genetic distances between pairs of species were binned by deciles, collapsing the first two and highest three deciles, respectively, owing to data scarcity. Overall, a steady and almost monotonous increase of phenotypic distance with genotypic distance is observed. Boxplots show quartiles per subsample, with whiskers stretching between adjacent values.
Figure 3—figure supplement 2
Features show different ranges of strong and weak correlations with genetic distance.
(A) Binned genetic distances between all species pairs, as in Figure 3—figure supplement 1. Corresponding differences in tooth complexity and heterodonty measures, respectively, are shown on the y-axis. For most measures excluding tooth-level complexities, no important differences are seen across most of the genetic distance range, suggesting limited depth of phylogenetic signal. The dark line connects averages per bin, with the gray shadow highlighting the range of the central 50% of all pairs per bin. Violin plots analogous to Figure 3—figure supplement 1B. (B) Dissimilarity of heterodonty and tooth complexity values between bins is quantified by Wilcoxon test, with numbers on x- and y-axes corresponding to bins. p-values are represented by grayscale; note that they can refer to both increasing and decreasing trends.
Figure 3—figure supplement 3
Taxon sensitivity of correlations between genetic and phenotypic distances.
In analogy to Figure 3A, we plot the p-values (Wilcoxon test) of the correlations between genetic and trait distances against the normalized genetic distance. Here, we recalculate after removal of each species, individually, represented by a set of thin lines. These lines are mostly dense, suggesting overall robustness of the results, with the exception of the right tail end of the heterodony measures, driven by the two species Hexanchus nakamurai (Hena) and Pristiophorus japonicus (Prja), with added arrowheads pointing to the respective lines. Inlet heatmaps show individual logarithmized p-value differences from the average line per excluded taxon, arranged as in Figure 2.
Figure 3—figure supplement 4
Phylogenetic signal.
Using the commonly used Moran’s I, Abouheif’s C_mean, Pagel’s λ, and Blomberg’s K, (A) phylogenetic signal strengths and (B) the p-values of their respective significances are plotted for ecological traits, tooth complexity, and heterodonty measures. Most trophic guilds, monognathic heterodonty measures, angle-based complexity measures, and cuspidities only show relatively low phylogenetic signal, while some habitat traits, depth, Fourier complexity, and maximal heterodonty exhibit strong phylogenetic signal.
Figure 4 with 2 supplements
Heterodonty and tooth-level complexity measures separate shark superorders.
(A) Canonical correlation analysis (CCA) reveals combinations of heterodonty and tooth-level complexity measures that are specific for the two superorders, squalean (Sqm, teal) and galean (Gam, red) sharks. (B) The two main clades are separated more clearly if ecological traits are included into the canonical analysis. The colors of the displayed species acronyms correspond to the respective orders, as displayed beside. (C) Violin plots contrast specific features and feature combinations in Squalomorphii and Galeomorphii, with p-values plotted above (Wilcoxon test). Monognathic and dignathic heterodonty, the ratio between the two, Fourier-based tooth-level complexity (Cx_Fourier), and heterodonty and cusp ratios, as well as depth, show significant differences between the two clades. Cx_combined is the sum of all tooth-level complexity measures, Cusp1 and Cusp2 are coarse and fine cuspidity, respectively. Boxes show quartiles and whiskers the respective adjacent values; significances: 0.1>p > 0.05: °, 0.05>p > 0.01: *, 0.01>p > 0.001: **, 0.001>p: ***. HMS/HMT: sequential/total monognathic heterodonty, HDG: dignathic heterodonty, HMX: maximal heterodonty between any two teeth of the same jaw, Cx: tooth-level complexity.
Figure 4—figure supplement 1
Depth and size ranges per species.
This graph displays key traits linked to ecology across the scrutinized taxa. (A) Depth: Thick bars span typical depth range per species as gathered across literature. Thin candlestick ends expand to exceptional reports, red dots, and the color code mark averages. (B) Body size (length). The range of body sizes for sexually mature adults is shown. Averages of lower and upper ranges for both sexes are shown as red dots. Colors mark ratios between hatchling/newborn and average mature adult body length.
Figure 4—figure supplement 2
Modularity of heterodonty is correlated with habitat depth.
(A) The ratio of dignathic (HDG) and sequential monognathic heterodonties (HMS) is plotted against average habitat depth. Colors denote superorders, and dot size is proportional to the log10 of average size. The dashed line is the linear regression. (B) Here, we plot modularity between HMS and HDG as the distance from the linear regression in Figure 1(B), a proxy of how much every taxon deviates from a perfect correlation of mono- and dignathic heterodonties.
Figure 5
Combinations of heterodonty and tooth-level complexity measures reveal two distinct strategies.
(A) Many shark species show, roughly, either high single-tooth complexity and low monognathic heterodonty (both Fourier-based), or vice versa, but rarely exhibit high values for both. Thus, two different clusters emerge, here denoted as group 1 (G1) and group 2 (G2). Blue marks the former (G1, n=27), orange the latter group (G2, n=24). Shown outlines display, respectively, mean tooth shapes for both groups. (B) Group-wise enrichments of trophic and habitat features, each ring presents the ratio between the respective percentages of species belonging to G1 divided by the ones belonging to G2 and the expected unbiased ratio. Ring size reflects the number of species per ecological category. p-values (two-sided binomial test, for G1 and G2, respectively) are annotated for the most significant differences. (C) Violin plots visualizing further group-specific characteristics with p-values (Wilcoxon test). Notably, the groups show divergent ratios between heterodonty measures based on outlines (Xoutl) and outline angles (Xang), corresponding to the heterodonty measures EMD, HED, and SAO vs. OAD and ADD (cf. Materials and methods). Cx.exc comprises complexity measures based on excentricity (OCR, OAR, OIR), Cx.ang (ANS, ASC, AND, OPC) measures based on angle complexity. Cx(OIR) is the minimal ratio between the areas of inscribed and escribed circles. Boxes show quartiles and whiskers the respective adjacent values; significances: 0.1>p > 0.05: °, 0.05>p > 0.01: *, 0.01>p > 0.001: **, 0.001>p: ***. HMS/HMT: sequential/total monognathic heterodonty, HDG: dignathic heterodonty, HMTx: maximal heterodonty between any two teeth of the same jaw, Cx: tooth-level complexity, Troph: mean trophic level.
Figure 6 with 4 supplements
Ecological relevance of dental shape descriptors varies across resolution levels.
(A) Different heterodonty and tooth-level complexity measures show specific correlations with ecological features, such as body size, depth, prey guilds, and habitats. Red and teal hues indicate correlation strengths of CCA1 for linear combinations of the predictors (red: negative correlations; teal: positive correlations). Cx.exec: comprises tooth-level complexity measures based on excentricity (OCR, OAR, OIR); Cx.four: Fourier-based complexity (DFS); and Cx.ang: angle-based complexity measures (ANS, ASC, AND, OPC). (B) Box plots summarizing canonical correlation strengths as a function of resolution/scale of the descriptors. Correlation strengths were found highest for monognathic heterodonty, while fine cuspidity yielded the lowest average correlation. As those measures represent, roughly, morphological trait differences (or complexity) on different resolution levels from fine cusps to differences between jaws, they are plotted in ascending order from the finest to coarsest scale. Canonical correlation strength can be used as a proxy for average relevance, suggesting size scale-dependent differences in ecological trait importance. Dot colors denote different types of ecological traits, while the gray boxes contain the combination of all traits, showing quartiles and extremes of the distribution. Displayed p-values were calculated using a Student’s t-test. Underlying shapes are included for illustrative guidance. HMS/HMT: sequential/total monognathic heterodonty, HDG: dignathic heterodonty, HMX: maximal heterodonty between any two teeth of the same jaw, Cx: tooth-level complexity.
Figure 6—figure supplement 1
Linear correlations between tooth complexity descriptors and ecological traits.
The size of Pearson’s R is plotted for all pairs of tooth-level and dentition-level complexity and ecological trait and superorder. Hues of red denote negative, turquoise positive correlations.
Figure 6—figure supplement 2
Similarity of canonical correlation profiles between ecological traits.
Heatmap showing pair-wise Manhattan distances between binarized (1/–1) canonical correlation profiles for ecological traits (Figure 6) with tooth complexity measures (CX1, CX2, CX3, Cusp1, Cusp2) and heterodonties (HMS, HMT, HMTX, HDG) as phenotypic coefficients. Two discernible sub-clusters emerge, pointing to discrete sets of ecological strategies that are reflected by dental patterns. Red: food guilds, turquoise: habitat specifiers.
Figure 6—figure supplement 3
Scale-dependent ecological relevance is robust against moderate data modifications.
(A) Upon systematic removal of shark species from the data set, we recalculated CCA correlations per scale-specific complexity feature (fine, coarse cuspidity, heterodonties). For each modified data set, we plotted the three inner quartiles and connected the scales by vectors. Colors indicate which species were removed. (B) For the same data, we remove species one-by-one (upper panel) and ecological predictors (lower panel) and plot p-values for pair-wise t-tests between scale-specific complexity measures. The color code is logarithmic and significant values (p<0.05) are highlighted in shades of reddish, with the pairs of measures indicated besides. The results displayed in Figure 6B are re-plotted as reference. It appears that the stark differences between monognathic heterodonty and the other measures shown ibidem are robust against moderate alterations to the data.
Figure 6—figure supplement 4
Summary: Correlations of features per resolution scale and macro-phylogeny do not reflect ecological importance.
For the measures separated by resolution as in Figure 6B, we show the respective correlations with macro-phylogeny (superorders) using Pearson’s correlation R, besides the corresponding p-values, CCA correlation strengths, Figure 6A, Figure 6—figure supplement 1, as well as the p-values of a Wilcoxon test, Figure 4A. The significance threshold of 0.05 for p-values is indicated by a dashed line. Particularly strong correlations exist for fine-grained complexity (Cusp2) as well as dignathic heterodonty (HDG), which is in contrast to the scales at which correlations with ecological proxies are most prominent, suggesting relative independence of ecological and macro-phylogenetic patterns.
Additional files
-
Supplementary file 1
List_NCBI_refs.xlsx.
This file lists all NCBI sequence references used to reconstruct the shark phylogeny.
- https://cdn.elifesciences.org/articles/107406/elife-107406-supp1-v2.xlsx
-
Supplementary file 2
Sharks_eco_refs.xlsx.
Overview of ecological information per species and their respective references.
- https://cdn.elifesciences.org/articles/107406/elife-107406-supp2-v2.xlsx
-
MDAR checklist
- https://cdn.elifesciences.org/articles/107406/elife-107406-mdarchecklist1-v2.pdf
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Integration of multi-level dental diversity links macro-evolutionary patterns to ecological strategies across sharks
eLife 14:e107406.
https://doi.org/10.7554/eLife.107406
