Non-technical Summary
Procolophonids were a group of early lizard-like reptile ancestors that lived for over 60 million years. They were one of the few groups that survived one of the Big Five major mass extinction events in Earth’s history (end-Permian mass extinction; 252 million years ago). Following the extinction, procolophonids are thought to have evolved adaptations for eating plants, including large, molar-like teeth. Unfortunately, historical evaluations of procolophonid teeth are limited to descriptive observations. While useful, these qualitative descriptions can vary from researcher to researcher and may not be objective or scalable. More recently, quantitative methods of evaluating tooth shape use sophisticated three-dimensional (3D) methods, but these methods are costly—financially and computationally—and as a result may not be available to all researchers. Here we build a predictive model based on living lizards to quantitatively predict the diets of procolophonids using relatively simple measures of tooth shape and size. In our study, we use a method called linear discriminant analysis to separate differences in our living lizard data (measurements of tooth shape and size) by finding a way to draw a boundary that best divides them by their diet. In addition, we tested to see whether our lizard tooth measures differed significantly by diet. Overall, we report two key findings: (1) procolophonids are largely predicted to eat plants but appear to practice a distinct “type” of plant eating from that of lizards, and (2) simple shape and size measures return relatively similar results to those of complex 3D methods. We hypothesize that the plants that procolophonids ate posed a different processing challenge from modern plants, likely contributing to the uniquely-shaped teeth of procolophonids.
Introduction
During periods of mass extinction ecosystem recovery, lineages diversify to occupy open niches after mass extinction events (Dineen et al., Reference Dineen, Roopnarine and Fraiser2019; Grossnickle et al., Reference Grossnickle, Smith and Wilson2019; Lowery and Fraass, Reference Lowery and Frass2019). As a result, organisms often “experiment” ecologically with newly available food sources and habitats. Changes in niche occupation are characteristic of mass extinction events, and such patterns are identifiable in mass extinction events throughout the fossil record. For example, net diversification rates are shown to increase for some clades following mass extinctions (Jablonski, Reference Jablonski2005). In addition, some authors have found post-extinction radiations, evidenced by morphological divergences, in some clades (Antsey and Pachut, Reference Anstey, Pachut, Erwin and Anstey1995; Foote, Reference Foote1999; Jablonski, Reference Jablonski2005). In general, as niches empty from an extinction event, a new diverse fauna will occupy them in the aftermath.
Survivors of mass extinction events can shed light on the mechanisms and selectivity of an extinction and the “strange world” that appears in the recovery period (Jablonski, Reference Jablonski2001; Hull, Reference Hull2015). Further investigation of these taxa builds our understanding of what contributes to the survival of certain species, and even large clades, and to the extinction of others. For example, survivor taxa are hypothesized to possess advantageous traits favored during extinction environments, and post-extinction events are fertile ground for speciation (Ezcurra and Butler, Reference Ezcurra and Butler2018; MacDougall et al., Reference MacDougall, Brocklehurst and Fröbisch2019; Bertrand et al., Reference Bertrand, Shelley, Williamson, Wible and Chester2022). This strange-world phenomenon is exemplified by the species that appeared following the end-Permian mass extinction, which often exhibited unusual morphologies (e.g., drepanosaurs [Renesto et al., Reference Renesto, Spielmann, Lucas and Spagnoli2010]; azendohsaurs [Flynn et al., Reference Flynn, Nesbitt, Parrish, Ranivoharimanana and Wyss2010]; trilophosaurs [Mellett et al., Reference Mellett, Kligman, Nesbitt and Stocker2023]). Furthermore, faunal turnover between the Permian and Triassic was profound, resulting in vastly different terrestrial ecosystems following the extinction, as well as the appearance of the first members of crown tetrapod clades in the fossil record (Viglietti et al., Reference Viglietti, Benson, Smith, Botha and Kammerer2021). The end-Permian mass extinction’s reputation as the “worst” mass extinction draws primarily from the marine realm, whereas terrestrial evidence is largely restricted to southern Africa (e.g., Wignall, Reference Wignall2007; Viglietti et al., Reference Viglietti, Benson, Smith, Botha and Kammerer2021).
One of the few clades of stem reptiles to persist after the end-Permian mass extinction is Procolophonoidea Romer, Reference Romer1956 (Watson, Reference Watson1942; Bakker, Reference Bakker and Hallam1977; Raup and Sepkoski, Reference Raup and Sepkoski1984; Modesto et al., Reference Modesto, Sues and Damiani2001; Wignall, Reference Wignall2007). Procolophonoidea is a clade of small-bodied stem reptiles that lived during the Permian and Triassic (~265–201 Ma). Procolophonoidea refers to the larger clade encompassing Procolophonidae Lydekker, Reference Lydekker, Nicholson and Lydekker1889, an exclusively Triassic clade, and their Permian relatives. Procolophonids and their Permian counterparts exploited several ecological niches, including insectivory (e.g., Kitchingnathus untabeni Cisneros, Reference Cisneros2008b [Cisneros, Reference Cisneros2008b]), durophagous omnivory (e.g., Tichvinskia vjatkensis Chudinov and Vjushkov, Reference Chudinov and Vjushkov1956 [Cisneros, Reference Cisneros2008a]), and high-fiber herbivory (e.g., Hypsognathus fenneri Gilmore, Reference Gilmore1928 [Colbert, Reference Colbert1946; Sues et al., Reference Sues, Olsen, Scott and Spencer2000]). Because of gross tooth morphology, several studies concluded that post Permian–Triassic procolophonids evolved into herbivorous niches, contrasting with their insectivorous pre-extinction forebears (Gow, Reference Gow1978; Cisneros Reference Cisneros2008a; Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021). However, inferences of procolophonid dietary ecology relied largely on qualitative descriptions of tooth shape and lack a quantitative basis, (e.g., Colbert, Reference Colbert1946; Gow, Reference Gow1977a, Reference Gow1978; Sues et al., Reference Sues, Olsen, Scott and Spencer2000; Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021). Previous work on procolophonid dentition was able to distinguish between dentitions “capable” of high-fiber versus non-high-fiber diets (Cisneros and Ruta, Reference Cisneros and Ruta2010). However, no comprehensive analysis of procolophonid dentition and its relationship to dietary ecology has been conducted, despite the wide variation in tooth morphology present in the clade (e.g., Fig. 1). Because we lack comprehensive analyses of procolophonid dietary ecology, the breadth of procolophonid ecology during the Permo–Triassic transition and the subsequent recovery is unknown.
Figure 1. Line drawings of the last tooth in the tooth row of example procolophonids (top) and lizards (bottom) in occlusal view. Anterior (left). Posterior (right). Not drawn to scale.
Small-bodied tetrapods such as procolophonids represent important components of terrestrial ecosystems and can act as proxies of ecosystem health (Gusmão et al., Reference Gusmão, Tessarolo, Dobrovolski and Gonçalves-Souza2024). Furthermore, an organism’s diet is highly informative of its ecological role (Layman et al., Reference Layman, Giery, Buhler, Rossi and Penland2015), and our ability to categorize ancient ecologies influences our subsequent interpretations of post-extinction recovery and niche occupation. To that effect, procolophonids are excellent proxies for ecosystem recovery in the Triassic, particularly as it applies to the evolution of herbivory. By analyzing the unusual dentition of procolophonids, we can explore the breadth of tetrapod diversification following one of the most devastating mass extinctions in Earth’s history (Benton and Wu, Reference Benton and Wu2022).
Here we aim to quantitatively predict the dietary ecology of procolophonids using relatively simple measures of tooth shape and size. On the basis of our review of the literature, we hypothesize that procolophonids are a largely herbivorous clade with insectivorous origins. In this study, we focus on Procolophonidae, a Triassic clade, but include their close Permian relatives (owenettids and nyctiphruretids) to investigate ecological responses to disturbances on the scale of mass extinction events. To test our hypothesis, we developed a dataset of simple tooth measures for a representative sample of extant species of squamates to build a predictive model. We then use this model to infer diets in our fossil dataset. By doing this, we provide a framework for inferring the diets of extinct reptiles that can be used to interpret ecological breadth at different intervals in Earth’s history and expanded on for future use.
Previous work
Many have inferred the diets of individual species of procolophonid, often as early representations of herbivory among Pan-Reptilia. The labiolingually expanded cheek teeth of Hypsognathus fenneri were associated with some kind of high-fiber herbivory (Colbert, Reference Colbert1946) or durophagy (Sues et al., Reference Sues, Olsen, Scott and Spencer2000). These interpretations are supported by the presence of a prominent coronoid process and enlarged orbitotemporal opening, which accommodate larger jaw muscles consistent with herbivory (Sues et al., Reference Sues, Olsen, Scott and Spencer2000). It was further suggested that Procolophon trigoniceps Owen, Reference Owen1876 consumed a specialized diet of seeds, judging from the presence of heavily worn postcanine teeth that appeared capable of “crushing” (Gow, Reference Gow1978). The early-diverging procolophonid Oryporan insolitus Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021 has dentition associated with an herbivorous or durophagous omnivorous diet, judging from the presence of worn bulbous molar-like teeth with wide occlusal surfaces (Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021).
Evolutionary trends in the diet of procolophonoids are well studied, but there are gaps and uncertainties in the literature. Permian species possess small, peg-like teeth suited for insectivory (Cisneros, Reference Cisneros2008a), and many Triassic species have labiolingually broadened, multicuspid teeth suited for herbivory (Gow, Reference Gow1978). It was suggested that herbivory and durophagous omnivory were practiced by later-diverging members of Procolophonidae and that the advent of these dietary strategies was integral for the clade’s Triassic radiation (Cisneros, Reference Cisneros2008a). Dental morphologies required for herbivory likely appeared before procolophonids began eating fully herbivorous diets (Cisneros, Reference Cisneros2008b). Features typically associated with insectivory and omnivory may have been later co-opted for herbivory by later diverging procolophonids (Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021).
The abundant back and forth regarding the dietary ecology of procolophonids is not necessarily a firm consensus of what food items procolophonids ate or the mechanics needed to process certain foods (e.g., high-fiber plants). This is partially because we do not have a reliable proxy based on modern taxa from which to draw conclusions.
Predicting an organism’s ecology from fragmentary fossils is a central challenge in paleontology. To ground such predictions, relationships between morphology and ecology must first be evaluated in extant taxa. Many aspects of dentition are critical for predicting dietary ecology, and methods have focused on the functionally relevant aspects of dental morphology. In mammals, relatively simple metrics, such as degree of hypsodonty and relative grinding area, have proven to be useful for predicting diet (Janis, Reference Janis, Russell, Santoro and Signogneau-Russell1988; Van Valkenburgh, Reference Van Valkenburgh and Gittleman1989; Williams and Kay, Reference Williams and Kay2001). More complex and computationally intensive ways to quantify tooth shape, such as Orientation Patch Count, Rotated (OPCr; Evans et al., Reference Evans, Wilson, Fortelius and Jernvall2007; Wilson et al., Reference Wilson, Evans, Corfe, Smits, Fortelius and Jernvall2012; Evans and Pineda-Munoz, Reference Evans, Pineda-Munoz, Croft, Su and Simpson2018), Dirichlet Normal Energy (DNE; Bunn et al., Reference Bunn, Boyer, Lipman, St. Clair, Jernvall and Daubechies2011), and Relief Index (RFI; Boyer, Reference Boyer2008), use three-dimensional (3D) models of teeth to generate metrics that describe the complexity, curvature, and relief of a tooth. Those methods are most effective when datasets incorporate the entirety of the dentition, instead of limiting to a specific tooth as is typical in mammal research (Crofts et al., Reference Crofts, Smith and Anderson2020). Their utility is also limited when working with fragmentary fossil dentition. To their benefit, however, those methods focus solely on shape and mechanical function, so they are uninhibited by a need for homologous structures. As a result, they can be used across a variety of taxonomic groups. However, most work in this area focused on mammals due to their complex heterodont dentition, which complicates interpretations of unusual dentition in distantly related amniote clades (e.g., Demar and Bolt, Reference Demar and Bolt1981; Cisneros, Reference Cisneros2008a; Reisz et al., Reference Reisz, Scott and Modesto2022; Mann et al., Reference Mann, Henrici, Sues and Pierce2023).
Compared with the extensive literature on mammals, research that evaluates the relationship between diet and dental morphology in reptiles is less abundant. Recent work in this area relied mainly on 3D topographic measures (e.g., OPCr) or geometric morphometric methods for describing tooth complexity (e.g., Zahradnicek et al., Reference Zahradnicek, Buchtova, Dosedelova and Tucker2014; Melstrom, Reference Melstrom2017; Lafuma et al., Reference Lafuma, Corfe, Clavel and Di-Poï2021; Fischer et al., Reference Fischer, Bennion, Foffa, MacLaren, McCurry, Melstrom and Bardet2022; Hoffman et al., Reference Hoffman, Hancox and Nesbitt2023). Although these methods offered insights into adaptations for dietary ecology in reptiles, they are often less accessible to researchers with limited resources. However, linear measurements are one of the simplest and most accessible means of quantifying tooth shape and were used to provide dietary inferences of both extant and fossil species (e.g., Jones, Reference Jones2009; Lafuma, et al., Reference Lafuma, Corfe, Clavel and Di-Poï2021; Singh, et al., Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021). However, studies such as these pertaining to reptiles are limited, hampering our ability to interpret ancient ecologies.
Methods
Sample and measurements
We selected extant squamate species for inclusion in our dataset to obtain a representative sample of dietary and phylogenetic diversity. Our extant sample includes 91 species of lizards with an average of two individuals per species (minimum 1, maximum 7) and was limited to adult individuals when possible. Because sex information is limited in herpetological collections, especially in historical collections, we did not control for sex in our study, although information on how sex differences impact tooth morphology is limited (Townsend et al., Reference Townsend, Akin, Felgenhauer, Dauphine and Kidder1999). Our fossil sample encompasses 23 members of Procolophonoidea (including 20 species of procolophonids and two species of owenettids [Owenettidae Broom, Reference Broom1939], the sister lineage to Procolophonidae), as well as one nyctiphruretid, a member of Procolophonomorpha Romer, Reference Romer1964 (Modesto et al., Reference Modesto, Damiani, Neveling and Yates2003; Sӓilӓ, Reference Säilä2010).
We measured the labiolingual width (W) and mesiodistal length (L) of the last tooth in the tooth row of both sides of the maxillae and dentaries, when available, from extant specimens housed in the herpetology collections of the Field Museum of Natural History and the Yale Peabody Museum. We measured the last tooth in the tooth row because the last tooth in the tooth row is likely the most recently erupted tooth, and therefore we assume that the morphology of this tooth will not have damage or wear that would potentially bias our quantification of morphology and our inferences for ecology. Furthermore, previous work shows that the last tooth in the tooth row exhibits diet-specific morphology (Jones, Reference Jones2009). Measurements were taken using Rexbeti digital calipers and Precision Measuring digital calipers. We collected data from fossil specimens using a combination of published images from the literature and direct measurements of specimens in the Vertebrate Paleontology collections at the Field Museum of Natural History and the Yale Peabody Museum. Digital measurements were taken from published images using FIJI (Schindelin et al., Reference Schindelin, Arganda-Carreras, Frise, Kaynig and Longair2012).
From these measurements, we computed tooth area (A) and length/width ratio (R). A and R were calculated because they are meaningful indicators of tooth size and tooth shape, respectively. Measures of size (L, W, and A) were log-transformed before being incorporated into analyses.
Body size is also a well-known predictor of diet in both extant and extinct amniote groups (e.g., Pough, Reference Pough1973; Carbone et al., Reference Carbone, Mace, Roberts and MacDonald1999; Pineda-Munoz et al., Reference Pineda-Munoz, Evans and Alroy2016; Miller and Pittman, Reference Miller and Pittman2021). While snout–vent length (SVL) is the standard body-size proxy in reptiles, very little postcranial material is available for procolophonids, prohibiting its use here. Instead, we used head length (HL) as a proxy for body size in the extant and fossil datasets (e.g., Ortiz Rodriguez, Reference Ortiz Rodriguez2012). Where HL data were not available in the literature, it was measured from museum collections, taken from a close relative of similar SVL, or estimated using group-specific regression equations from ElShafie (Reference ElShafie2024). Work on fossil taxa shows that close relatives (at least family level) have similar SVL/HL ratios (ElShafie, Reference ElShafie2024). Source of HL data for each taxon is included in Supplementary Table 1. To ensure that HL is a good predictor of overall body size for our sample, we performed a regression analysis of log-transformed HL on log-transformed SVL (Supplementary Fig. 1). For fossil data, measurements of either HL or tooth-bearing bone length (TL) were collected.
To infer the diets of fossil taxa, we compared their dentitions with those of extant lizards with well-attested diets (Cooper and Vitt, Reference Cooper and Vitt2002; Melstrom, Reference Melstrom2017; Meiri, Reference Meiri2024). Data of extant lizard diet were obtained from SquamBase, a comprehensive public database of key squamate trait data (Meiri, Reference Meiri2024). SquamBase uses a discrete dietary classification system in which taxa are recognized as carnivorous, herbivorous, or omnivorous. To distinguish insectivores (faunivores that specialize in eating insects and small arthropods) from carnivores that specialize in eating small vertebrates and their eggs, we referred to the supplementary dietary metadata information available in SquamBase. When diet information was not available from SquamBase, we gathered diet data directly from the literature. All carnivorous taxa were eventually excluded from our dataset because their body sizes routinely exceed the size range that is relevant for procolophonids.
Analyses
To determine whether the procolophonid diet can be predicted from fossil tooth dimensions, we first determined whether tooth dimensions in extant lizards are predictive of their dietary ecology. To do this, we performed a linear discriminant analysis (LDA). LDA is a commonly used supervised learning technique that predicts group membership from multivariate quantitative data. LDA works by first finding linear combinations of features in a training dataset that yield good prediction of the categorical group membership of observations, before then being applied to observations for which group membership is not known. LDA was conducted using the “lda” function in the “MASS” package (Venables and Ripley, Reference Venables and Ripley2002) for R (R Core Team, 2021). Prior probabilities on group membership were set to the observed proportions of the dataset: herbivorous = 0.30, insectivorous = 0.30, and omnivorous = 0.40. We used a leave-one-out cross-validation (CV) to find the posterior probability for each taxon to engage in each dietary ecology. To determine whether the inclusion of size with dental dimensions improves dietary predictions, we performed a total of two LDAs: (1) diet ~ ratio + area + head length and (2) diet ~ ratio + area.
Dietary ecology of each fossil species was predicted from both discriminant functions using the “predict.lda” function from the “MASS” package. Predicted diets and LD scores from both LDAs were then projected into the LDA space for extant species and mapped onto a time-calibrated phylogeny of Procolophonoidea modified from Mueller et al. (Reference Mueller, Small, Jenkins, Huttenlocker and Chatterjee2024). Fossil occurrence data for time calibration were obtained from the Paleobiology Database (PBDB). We also acknowledge that Permian occurrences of procolophonids are dubious and that revisions to the Karoo Basin reptile dataset are ongoing. These updates may influence our results.
In addition to LDA, we used multivariate analysis of variance (MANOVA) to investigate which variables were guiding the differences between these groups. The influence of shared evolutionary history was investigated by conducting a phylogenetic MANOVA using the “aov.phylo” function from the “Geiger” package in R (Pennell et al., Reference Pennell, Eastman, Slater, Brown and Uyeda2014). Phylogenetic MANOVA recomputes the MANOVA for a specified number of dependent variable datasets (here, n = 999) simulated under a Brownian motion model of evolution (i.e., without an effect of the independent variable) and then asks how frequently the observed effect size is larger than the randomized effect sizes. We used the extant squamate phylogeny from Pyron et al. (Reference Pyron, Burbrink and Wiens2013).
Results
Linear discriminant analysis
We performed two LDAs based on measures of tooth size and shape (R, A) (Fig. 2). The first LDA (LDA1) was performed with a proxy for body size (HL), and the second (LDA2) was performed without this proxy. In both LDAs, a gradient consistently presents between herbivorous, omnivorous, and insectivorous lizards (Fig. 2.1, 2.3). Notably, the correct classification rates for both LDAs are low (<60%) and thus severely limit our ability to interpret the results.
Figure 2. LDAs on tooth dimension data for extant and fossil data. (1) Extant squamate data (LDA1): bivariate plot of LD1 and LD2. Observed diet depicted as green (herbivory), yellow (omnivory), and purple (insectivory). (2) Procolophonoid and nyctiphruretid LDA1 diet predictions: bivariate plot of LD1 and LD2. Predicted diet depicted as green (herbivory), yellow (omnivory), and purple (insectivory). Convex hulls of Plot 1 (extant squamate data—LDA1) are superimposed for comparison. (3) Extant squamate data (LDA2): bivariate plot of LD1 and LD2. (4) Procolophonoid and nyctiphruretid data LDA2 diet predictions: bivariate plot of LD1 and LD2. Convex hulls of Plot 3 (extant squamate data—LDA2) are superimposed for comparison.
However, general patterns are still observable: herbivores tend to have more negative scores on LD1, insectivores tend to have more positive scores, and the LD1 scores of omnivores are intermediate between the two (Fig. 2.1, 2.3). Herbivores occupy a region of the LDA space associated with larger tooth areas, wider teeth, and larger body sizes (Fig. 2.1, 2.3). Conversely, insectivores occupy a region of the LDA space associated with smaller tooth areas, narrower tooth ratios, and smaller body sizes. Again, omnivores are intermediate to these. Overall group scores follow this pattern. However, when we project individual scores into LDA space, there is a great deal of overlap. For example, some herbivorous species fall into the positive space along LDA1 (occupied mainly by insectivorous taxa) (Fig. 2.1, 2.3). The removal of body size as a predictor (LDA2) does not appear to change where different dietary groups sit in LDA space.
Two significant discriminant functions were obtained for LDA1. Discriminant function 1 explained 95.73% of the variance between groups with the following coefficients: log(A) (–0.85), R (0.97), and HL (–0.77). Discriminant function 2 explained 4.27% of the variance between groups with the following coefficients: log(A) (–0.06), R (–2.03), and HL (–0.43). LDA1 correctly classified 57.8% of cases, where “correctly classified” is defined as the true diet having the highest posterior probability of the four categories. Two significant discriminant functions were obtained from LDA2. Discriminant function 1 explained 95.52% of the variance between groups with the following coefficients: log(A) (–1.09) and R (0.94). Discriminant function 2 explained 4.48% of the variance between groups with the following coefficients: log(A) (–0.23) and R (–2.05). LDA2 correctly classified 51.6% of cases.
Fossil data predictions
Procolophonids were overwhelmingly predicted to be herbivorous (16–17/23), and only one taxon (Timanophon raridentatus Novikov, Reference Novikov1991) was classified differently between LDA1 and LDA2 (as an omnivore and an herbivore, respectively) (Table 1). Surprisingly, no taxa were predicted to be insectivorous. In both LDAs, procolophonids and owenettids are generally restricted to the negative space along LDA1 (Fig. 2.2, 2.4). This region of the LDA space is associated with larger tooth areas and wider teeth. Notably, procolophonids and owenettids occupy a region of the LDA space largely unoccupied by extant squamates (Fig. 2.2, 2.4). Procolophonids that do overlap with extant squamates fall mainly within the “herbivore” convex hull (except Kapes bentoni Spencer and Storrs, Reference Spencer and Storrs2002 in LDA1, which sits in a region occupied by both the “herbivore” and “omnivore” convex hulls) (Fig. 2.2, 2.4). However, the nyctiphruretid Nyctiphruretus acudens Efremov, Reference Efremov1938 sits in a region of LDA space in which all three convex hulls (“herbivore,” “omnivore,” and “insectivore”) overlap and that is isolated from the procolophonoid taxa in the dataset (Fig. 2.2, 2.4). Posterior probabilities are generally low, but two general takeaways are possible. First, posterior probabilities are somewhat higher for LDA2 than for LDA1. Second, probabilities associated with predicted herbivorous procolophonids are generally higher than those for predicted omnivores and for any extant squamate, ranging as high as 0.87. In addition, in both LDAs, no procolophonids are predicted to be insectivorous, and the posterior probabilities for insectivory are always low, never rising above 0.04 for procolophonids or 0.19 for owenettids. The LDA1 scores for extant specialized insectivorous taxa tend to be positive and >1, and the taxa in our dataset do not possess LDA1 scores in this range.
Table 1. Diet predictions for procolophonids and close relatives. Diet predictions, along with posterior probabilities, for both LDAs (LDA1 and LDA2) are reported along with diet predictions from the literature
When we plotted the results of our LDA fossil predictions onto a phylogeny of procolophonoids and a close relative, we find the outgroup (Nyctiphruretus) and early diverging procolophonoids are recovered as omnivores, while later-diverging taxa are recovered as herbivores (Fig. 3). Contrary to our initial hypothesis, no taxa in our dataset are predicted to be insectivores. Our results suggest a stepwise evolution toward herbivory within Procolophonidae. Moving from the base of the tree to the more nested clades, LDA1 scores decrease substantially (e.g., Saurodektes kitchingorum Modesto et al., Reference Modesto, Damiani, Neveling and Yates2003 at –0.01/–0.03 versus Hypsognathus fenneri at –4.47/–4.52) (Fig. 3; Table 1). Furthermore, taxa within Leptopleuroninae Ivakhnenko, Reference Ivakhnenko1979 are entirely and consistently recovered as herbivores. In addition, when we plotted procolophonid tooth areas through time (Fig. 4), we found that procolophonid tooth size increased as the Triassic progressed. This is consistent with the hypothesis that dental complexity increased in procolophonids toward the end of the Triassic.
Figure 3. Phylogenetic distribution of predicted diets among Procolophonoidea and their close relatives. Time-calibrated phylogeny modified and inferred from Mueller et al. (Reference Mueller, Small, Jenkins, Huttenlocker and Chatterjee2024). Diet predictions depicted as green circles (herbivory) and yellow squares (omnivory). The inset graph (top right corner) is a biplot of taxon age (in 10 Myr time bins) and LD1 scores. LD1 values from LDA1 and LDA2, respectively, are denoted by pink circles and blue triangles.
Figure 4. Ribbon plot comparing the distribution of procolophonoid tooth area (mm2) by taxon age (in 10 Myr time bins). Median values (solid line) are shown for tooth area with shaded ribbon representing interquartile range (IQR). Line drawings of the last tooth in the tooth row of example procolophonids (Coletta seca Gow, Reference Gow2000 and Hwiccewyrm trispiculum Butler et al., Reference Butler, Meade, Cleary, McWhirter, Brown, Kemp, Benito and Fraser2024) included to demonstrate general difference in average tooth area and morphology. Not drawn to scale.
As mentioned, only one procolophonid, Timanophon, had different dietary classifications in each LDA. Because of this, we cannot distinguish whether it was an omnivore or strictly herbivorous. Curiously, Timanophon is found at a transition point in the clade where procolophonids become nearly exclusively herbivorous, except for Procolophon trigoniceps (Fig. 3).
Phylogenetic MANOVA
Tooth shape varies quantitatively among dietary guilds of extant lizards. Insectivores possess the narrowest (mean R = 1.61) and smallest (mean A = 0.246) teeth, whereas herbivores possess the widest (R = 1.49) teeth with omnivores intermediate (R = 1.40). Omnivorous and herbivorous teeth are, on average, similar in size to one another (A = 0.76 and 0.72) but much larger than those of insectivores (A = 0.246).
MANOVAs were conducted to determine whether there is a significant difference between dietary categories in measures of tooth shape and size, with and without the inclusion of HL as a proxy for body size. When phylogenetic relatedness was accounted for, the effect was non-significant (p = 0.487). A second MANOVA was conducted to assess the effect of dietary ecology on measures of tooth shape and size (R, A) without the inclusion of a body-size proxy. Again, there was a non-significant effect when phylogenetic relatedness was accounted for (p = 0.446).
Discussion
Procolophonids were highly specialized, unique herbivores, distinct from extant plant-eating lizards
By predicting dietary ecology in procolophonoids, we offer insight into post-extinction ecosystem dynamics and the macroevolutionary processes governing trends in ecological niche occupation. Contrary to our initial hypothesis, procolophonids do not have insectivorous origins. Rather, we predict that the earliest diverging procolophonids are omnivores. In times of climatic instability and consequent scarcity, dietary generalists such as omnivores are at an advantage because they can exploit a wider range of resources (e.g., McKinney, Reference McKinney1997; Sahney and Benton, Reference Sahney and Benton2008). However, some dietary specialists, such as extant herbivores, are found to possess the highest risks of extinction (Atwood et al., Reference Atwood, Valentine, Hammill, McCauley and Madin2020). We suggest that exploiting generalist niches may have led to the initial success of procolophonids in the post-extinction landscape. However, there is mixed evidence of this pattern outside of Procolophonidae. Some specialist herbivore clades, such as dicynodonts, do survive the end-Permian extinction and persist into the Triassic (Smith and Botha, Reference Smith and Botha2005; Botha-Brink and Angielczyk, Reference Botha-Brink and Angielczyk2010).
We found that procolophonids are overwhelmingly predicted to be herbivorous, with some omnivorous individuals, although there is very little overlap in the LDA space with any modern lizard taxa. No taxa in our sample were predicted to be insectivorous, contrasting with previous qualitative works (Gow, Reference Gow1977a, Reference Gow1978). Although effect sizes are small and phylogenetic MANOVAs imply limited utility of the simple metrics for discriminating diet in the extant sample, diet estimates for procolophonids suggest that our models can be utilized for dietary prediction in fossil taxa, particularly those that have unique morphologies.
While low LDA correct classification scores hamper our ability to discriminate between diet categories in modern lizards, it is striking that procolophonids occupy a distinct area of the LDA space. Because of this distinct occupation, we hypothesize that procolophonids, while herbivorous, either ate plant matter or mechanically processed food distinctly differently from modern herbivorous lizards. We suggest that unique tooth morphologies permitted procolophonids to engage in a unique form of herbivory well suited for the consumption of the tough, fibrous plant material of the Triassic, which we detail further in the following. In addition, we highlight trends in procolophonid tooth shape throughout the Triassic and its relationship to the clade’s progression toward specialized herbivory.
Multiple Triassic small-bodied reptiles exhibit adaptations for herbivory (e.g., procolophonids [e.g., Cisneros and Ruta, Reference Cisneros and Ruta2010; Zaher et al., Reference Zaher, Coram and Benton2019; Pinheiro et al., Reference Pinheiro, Silva-Neves and Da-Rosa2021], opisthodontian and clevosaurid rhynchocephalians [e.g., Jones, Reference Jones2009; Martínez et al., Reference Martínez, Apaldetti, Colombi, Praderio, Fernandez, Malnis, Correa, Abelin and Alcober2013], rhynchosaurs [e.g., Sethapanichsakul et al., Reference Sethapanichsakul, Coram and Benton2023], and trilophosaurids [e.g., Mellet et al., Reference Mellett, Kligman, Nesbitt and Stocker2023]). Further, some Triassic procolophonids and trilophosaurids share notable similarities in jaw and tooth morphology (e.g., bulbous, multicuspid molariform dentition [Chambi-Trowell et al., Reference Chambi-Trowell, Whiteside, Skinner, Benton and Rayfield2021; Foffa et al., Reference Foffa, Nesbitt, Kligman, Butler and Stocker2022]). Our results may reflect a larger trend toward specialized herbivory among small-bodied reptile groups rather than an isolated instance within Procolophonidae. Herbivorous squamates today feed mainly on the flowers, fruits, and leaves of angiosperm plants (Cooper and Vitt, Reference Cooper and Vitt2002; Soltis et al., Reference Soltis, Soltis, Endress, Chase, Manchester, Judd, Majure and Mavrodiev2018). The biochemical and mechanical properties of these food items fundamentally differ from those of the Triassic when gymnosperms were the dominant plant life—flowers and fruits, highly nutritious and easier-consumed food items, were not yet present (Kustatscher et al., Reference Kustatscher, Ash, Karasev, Pott, Vajda, Yu, McLoughlin and Tanner2018). At this time, herbivorous procolophonids had limited options ranging from eating tough, fibrous cycad leaves, seeds of Triassic plants, and the “bark” of these plants. Such plants were woody with high fiber content and would pose a considerable challenge to consume and digest both mechanically and biochemically (e.g., van Marken Lichtenbelt, Reference van Marken Lichtenbelt1992; Grauvogel-Stamm and Ash, Reference Grauvogel-Stamm and Ash2005; Berthaume, Reference Berthaume2016). In addition, as the Early Triassic progressed to the Middle and Late Triassic, plant life recovered from the end-Permian extinction, and a more mature flora evolved that was able to better support an herbivorous fauna (e.g., Hotton et al., Reference Hotton, Olson, Beerbower, Sumida and Martin1997; Grauvogel-Stamm and Ash, Reference Grauvogel-Stamm and Ash2005; Kustatscher et al., Reference Kustatscher, Ash, Karasev, Pott, Vajda, Yu, McLoughlin and Tanner2018). During this progression, the dentition of procolophonids continued to evolve to larger sizes (Fig. 4), and they were exposed to more dietary options, including gymnosperms such as cycads, ferns, lycopods, and sphenophytes (e.g., Grauvogel-Stamm and Ash, Reference Grauvogel-Stamm and Ash2005; Kustatscher et al., Reference Kustatscher, Ash, Karasev, Pott, Vajda, Yu, McLoughlin and Tanner2018). Digesting plants such as these and extracting sufficient nutrients would require ample oral processing (i.e., chewing). The bulbous, molariform posterior marginal dentition present in many procolophonids would be effective for consuming these types of plants.
Procolophonids lack exact modern analogs because they consumed flora that was fundamentally different from what is around today. However, procolophonids show some general characteristics related to herbivory in modern taxa, but they appear to represent a more distinct ecomorphotype. For example, Procolophonidae possess multicuspid teeth, which are present in many extant herbivorous lizards (e.g., Dipsosaurus dorsalis Baird and Girard in Baird et al., Reference Baird, Girard and Le Conte1852, Iguana iguana Linnaeus, Reference Linnaeus1758, and Saara hardwickii Gray in Hardwicke and Gray, Reference Hardwicke and Gray1827). However, the teeth of procolophonids are not laterally compressed like the teeth of extant iguanids (a prolific clade of herbivorous lizards) but rather wide and molariform, similar to durophagous lizards (e.g., Tupinambis teguixin Linnaeus, Reference Linnaeus1758 and Dracaena guianensis Daudin, Reference Daudin1802). In addition, many procolophonids possess an elaborate palatal dentition, with more teeth and distributed among the palatal bones, which many modern lizards lack (Mahler and Kearney, Reference Mahler and Kearney2006; Zaher et al., Reference Zaher, Coram and Benton2019). While this may be a plesiomorphic trait retained in procolophonids, it is possible that, due to the greater elaboration of the palatal dentition, procolophonids used them actively to process plant matter. However, the function of palatal dentition in tetrapods has seldom been observed directly. Horned procolophonids also have a crown-to-crown occlusion system (e.g., Soturnia caliodon Cisneros and Schultz, Reference Cisneros and Schultz2003, Hypsognathus fenneri), a derived feature among the clade thought to be advantageous for high-fiber herbivory or durophagy (e.g., Colbert, Reference Colbert1946; Cabreira and Cisneros, Reference Cabreira and Cisneros2009; Hamley et al., Reference Hamley, Cisneros and Damiani2021). Occlusion systems are described in extant herbivorous lizards (e.g., Pogona vitticeps Ahl, Reference Ahl1926; Haridy, Reference Haridy2018), but they do not appear to occlude crown-to-crown as they do in the more derived procolophonids.
Alternatively, a major transition occurred in the evolution of insects during the Triassic (e.g., Zheng et al., Reference Zheng, Chang, Wang, Fang, Wang, Feng, Xie, Jarzembowski, Zhang and Wang2018; Gui et al., Reference Gui, Liu and Tian2023; Montagna et al., Reference Montagna, Magoga, Stockar and Magnani2024). Entomofaunas are hypothesized to have modernized as far back as the Triassic, in the aftermath of the end-Permian extinction (e.g., Béthoux et al., Reference Béthoux, Papier and Nel2005; Gui et al., Reference Gui, Gao, Liu, Chu, Dal Corso, Tong and Tian2024). By the Middle and Late Triassic, modern insect faunas had emerged and represented a significant portion of Triassic insect diversity (Gui et al., Reference Gui, Gao, Liu, Chu, Dal Corso, Tong and Tian2024). This modernization hypothesis is in line with a greater hypothesis surrounding the evolution of ecosystems following the Permian–Triassic boundary: that the beginning of the Mesozoic marks the beginning of modern terrestrial ecosystems (e.g., Benton and Wu, Reference Benton and Wu2022). Such a change in the composition of the entomofauna may explain why our diet hypotheses for Early–Middle Triassic procolophonids contradict previous ones. We predict some Early and Middle Triassic-age procolophonids were omnivores, as opposed to insectivores (Table 1). Meaning that Early Triassic tetrapods were potentially consuming insects with different morphologies from those consumed by modern squamates and that modern squamate lineages (and their teeth) evolved alongside an entirely different entomofauna. Changes in insect morphology through time, and the associated mechanical challenges, may explain the difference in our findings. Extensive work by Singh et al. (Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021) investigated the evolution of herbivory at the community level in early Mesozoic ecosystems using a combination of geometric and functional morphometric methods. Where our study focuses on Procolophonoidea and uses simple, linear measurements of dentition, Singh et al. (Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021) analyzed Triassic and Jurassic tetrapods using the mandible. Singh et al. (Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021) also included procolophonoid parareptiles in their dataset but did not include members before the Permian–Triassic boundary. In general, the conclusions of our study agree with those of Singh et al. (Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021), notably, the transition of the clade from “tough generalist[s]” (owenettids, basal procolophonids, and procolophonines) to “crushing biter[s]” (leptopleurines) (Singh et al., Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021). On the basis of modern proxies, we also identify owenettids and basal procolophonoids as generalist omnivores and leptopleurines as durophagous specialists, likely specializing on plant materials. Among the procolophonines, however, we do predict a mix of generalist omnivory and herbivory. This difference is likely due to how our studies describe diet differently. In the present study, we restricted our descriptions of diet to the traditional categories—herbivore, omnivore, and faunivore (insectivore, in the case of procolophonoids)—whereas Singh et al. (Reference Singh, Elsler, Stubbs, Bond, Rayfield and Benton2021) classified diet using a functional approach. Despite differences in methods, both studies support the conclusion that procolophonids were largely siloed into specialized herbivory by the end of the Triassic.
Traditional linear measurements remain an informative part of the ecomorphologist’s toolkit
There is a current trend toward using increasingly complex metrics to quantify dental ecomorphology in the literature, but historically, simple measures based on ratios and linear measurements have proved useful for discriminating broad dietary categories (e.g., Janis, Reference Janis, Russell, Santoro and Signogneau-Russell1988; Van Valkenburgh, Reference Van Valkenburgh and Gittleman1989; Williams and Kay, Reference Williams and Kay2001; Jones, Reference Jones2009). Here we used relatively simple measurements (area and ratio of the last tooth in the tooth row) as measures of tooth shape and size. We found that, in general, herbivorous taxa were more likely to have larger, wider teeth and that insectivorous taxa were more likely to have smaller, narrower teeth, supporting previous findings for lepidosaurs (e.g., Jones, Reference Jones2009; Melstrom, Reference Melstrom2017; Lafuma et al., Reference Lafuma, Corfe, Clavel and Di-Poï2021).
Conservatively, the discriminatory power of linear measurements in our study is limited. Low LDA correct classification scores (<60%) restrict our ability to discern between dietary groups in extant lizards, although patterns typically reported in previous studies of extant reptilian teeth are recovered. Because linear measurements offer a comparatively lower discriminatory power than more complex functional traits, such as OPCr or geometric morphometric descriptors of shape, they are not an effective predictive tool for highlighting more subtle differences in morphology. However, we show that linear measurements prove useful when studying taxa with extreme morphologies. Here these measures act as a quantitative basis for confirming the unique and derived dental morphology of procolophonids when compared with extant lizards with varying diets. Despite low correct classification scores, the distinct, separate location of fossil data in LDA space makes relative comparisons possible.
In support of previous findings, we demonstrated that there is an abundance of diversity in reptilian teeth that directly relates to dietary ecology (Jones, Reference Jones2009; Melstrom, Reference Melstrom2017; Lafuma et al., Reference Lafuma, Corfe, Clavel and Di-Poï2021). Furthermore, across different approaches of quantifying and describing diversity, we arrive at the same general patterns—our work using simple linear measurements of tooth shape recovered similar results to work that employed 3D methods to quantify tooth shape (Melstrom, Reference Melstrom2017). Analyses that rely on 3D data are computationally and time-intensive to collect and are highly sensitive to data collection and cleaning methods (Spradley et al., Reference Spradley, Pampush, Morse and Kay2017; Berthaume et al., Reference Berthaume, Winchester and Kupczik2019). Although our methods for evaluating the relationship between tooth morphology and diet lack the level of confidence that 3D methods may provide, we found the same ecomorphological relationships previously recovered. To that end, simple linear measurements offer a more accessible means of testing hypotheses related to diet and dentition in the fossil record, and they may provide an informative first approach in data exploration. Our findings also mirrored the relationships between tooth shape and diet in mammals: we found a similar pattern of labiolingually broadened, larger teeth belonging to herbivores and smaller, narrower teeth belonging to insectivores; with omnivorous tooth shapes as intermediate (Van Valkenburgh, Reference Van Valkenburgh and Gittleman1989; Evans et al., Reference Evans, Wilson, Fortelius and Jernvall2007; Cuozzo et al., Reference Cuozzo, Ungar and Sauther2012; Wisniewski et al., Reference Wisniewski, Nations and Slater2023). Further work exploring whether modern mammals can be used as dietary proxies for disparate clades, such as Triassic reptiles, may elucidate the dietary breadth of extinct clades with unusual tooth morphologies.
Increasing values in every tooth-related metric generally correlate with herbivory. More complex topography (OPCr), larger occlusal surface area, or higher cusp count were identified as indicative of plant-eaters, broadly across Tetrapoda (Evans et al., Reference Evans, Wilson, Fortelius and Jernvall2007; Jones et al., Reference Jones2009; Melstrom, Reference Melstrom2017; Lafuma et al., Reference Lafuma, Corfe, Clavel and Di-Poï2021; Wisniewski et al., Reference Wisniewski, Nations and Slater2023). Combining traits that predict herbivory may best emphasize the different facets of tooth morphology that contribute to plant-eating. Using both relatively simple measurements and more complex methods increases the amount of information with which we can make dietary predictions. This may be particularly useful for making future predictions about the types of plants being eaten (e.g., leaves versus roots) and the adaptations that facilitate eating different foodstuffs.
Given our results, we recommend that future analyses combine simple and complex methods when predicting dietary ecology, especially for extinct taxa. Different types of methods are especially helpful for accommodating the fragmentary fossil record and capturing different features of tooth shape. We can further investigate the relationship between morphology and diet by incorporating aspects of morphology other than teeth. Clearly, there is far more to food processing than the marginal (maxillary and mandibular) dentition, but analyses predicting diet tend to be limited to a few teeth or to specific tooth-bearing bones. In many amniotes, palatal rugosities, pharyngeal spines, and palatal dentitions are known or hypothesized to contribute to food ingestion and processing (Mahler and Kearney, Reference Mahler and Kearney2006; Menegaz et al., Reference Menegaz, Sublett, Figueroa, Hoffman and Ravosa2008; Fraher et al., Reference Fraher, Davenport, Fitzgerald, McLaughlin, Doyle, Harman and Cuffe2010). Incorporating such aspects of morphology into studies on diet in extant and extinct taxa will hopefully offer greater predictive accuracy and a more nuanced picture of the cohesive nature and function of food-processing systems.
Acknowledgments
We thank G. Slater and K. Angielczyk for numerous fruitful discussions that influenced the trajectory of this study, and we additionally thank them and B.-A.S. Bhullar for feedback on an early version of this manuscript. Additional discussion was had with D. Rhoda, R. Ng, and S.S. Strassberg. We thank V. Rhue (YPM), G. Watkins-Colwell (YPM), W. Simpson (FMNH), and R. Kamei (FMNH) for specimen access. S.A.M. was supported by the University of Chicago Committee on Evolutionary Biology, NSF GRFP, Steve Jobs Archive Fellowship, the Von Damm Undergraduate Research Fellowship in Geology and Geophysics, and Yale College Dean’s Research Fellowship; and K.M.J. was supported by Yale Department of Earth and Planetary Sciences and the Smithsonian National Museum of Natural History Peter Buck Fellowship during the duration of this study. We also thank editor J. Calede and X. Jenkins as well as one anonymous reviewer for greatly improving the final manuscript.
Competing interests
The authors of this manuscript declare they have no conflicts of interest in submitting this manuscript.
Data availability statement
Data and all supplemental materials are available from the Data Dryad Digital Repository: https://doi.org/10.5061/dryad.9kd51c5x8.
Open access
