Evidence of strong stabilizing effects on the evolution of boreoeutherian (Mammalia) dental proportions

Abstract The dentition is an extremely important organ in mammals with variation in timing and sequence of eruption, crown morphology, and tooth size enabling a range of behavioral, dietary, and functional adaptations across the class. Within this suite of variable mammalian dental phenotypes, relative sizes of teeth reflect variation in the underlying genetic and developmental mechanisms. Two ratios of postcanine tooth lengths capture the relative size of premolars to molars (premolar–molar module, PMM), and among the three molars (molar module component, MMC), and are known to be heritable, independent of body size, and to vary significantly across primates. Here, we explore how these dental traits vary across mammals more broadly, focusing on terrestrial taxa in the clade of Boreoeutheria (Euarchontoglires and Laurasiatheria). We measured the postcanine teeth of N = 1,523 boreoeutherian mammals spanning six orders, 14 families, 36 genera, and 49 species to test hypotheses about associations between dental proportions and phylogenetic relatedness, diet, and life history in mammals. Boreoeutherian postcanine dental proportions sampled in this study carry conserved phylogenetic signal and are not associated with variation in diet. The incorporation of paleontological data provides further evidence that dental proportions may be slower to change than is dietary specialization. These results have implications for our understanding of dental variation and dietary adaptation in mammals.

among the three molars (molar module component, MMC), and are known to be heritable, independent of body size, and to vary significantly across primates. Here, we explore how these dental traits vary across mammals more broadly, focusing on terrestrial taxa in the clade of Boreoeutheria (Euarchontoglires and Laurasiatheria).
We measured the postcanine teeth of N = 1,523 boreoeutherian mammals spanning six orders, 14 families, 36 genera, and 49 species to test hypotheses about associations between dental proportions and phylogenetic relatedness, diet, and life history in mammals. Boreoeutherian postcanine dental proportions sampled in this study carry conserved phylogenetic signal and are not associated with variation in diet. The incorporation of paleontological data provides further evidence that dental proportions may be slower to change than is dietary specialization. These results have implications for our understanding of dental variation and dietary adaptation in mammals.
This apparent mismatch is likely driven by the observation that occlusal morphology can reflect adaptation to the most mechanically challenging foods a mammal processes independently of the frequency of that specific food in the diet (Ungar, 2009;Ungar, Healy, Karme, Teaford, & Fortelius, 2018), and that developmental mechanisms can evolve similarly but in response to different selective pressures, especially among closely related taxa (Ungar & Hlusko, 2016).
Dental phenotypes can also vary with life history traits like age at weaning, prenatal growth rates, and gestation length (Monson, Coleman, & Hlusko, 2019;Smith, 1989Smith, , 2018Smith, Crummett, & Brandt, 1994). Other research has reported that some aspects of dental variation are more strongly associated with phylogenetic relatedness in mammals than diet or life history strategies (Gamarra, Delgado, Romero, Galbany, & Pérez-Pérez, 2016;Macholán, 2006;Monson & Hlusko, 2018a, 2018b. These observations accord with results from comparisons between molecular and morphological data demonstrating that certain dental traits can reliably predict phylogenetic relatedness (Cardini & Elton, 2008;Caumul & Polly, 2005). These studies also suggest that the developmental etiology of dental variation may be a stronger evolutionary force than previously recognized. The canalization of development as a consequence of strong integration and genetic pleiotropy can act as a stabilizing selective pressure limiting rapid evolutionary change (Gibson & Wagner, 2000).
Due to recent advances in genotype:phenotype (G:P) mapping of mammalian dental variation, we can now approach the critical question that lies at the heart of comparative anatomy-to what degree is morphology evidence of a fine-tuned response to selection versus a constrained result of stabilizing selection (Hlusko, 2004(Hlusko, , 2016Lovejoy, Cohn, & White, 1999)? G:P mapping of dental patterning over the last two decades has led to a dramatic increase in our understanding of the genetic and developmental mechanisms that underlie mammalian dental variation (Bei, 2009;Hlusko, Sage, & Mahaney, 2011;Hlusko, Schmitt, Monson, Brasil, & Mahaney, 2016;Thesleff, 2006;Thesleff & Hurmerinta, 1981;Thesleff & Sharpe, 1997;Tucker & Sharpe, 2004).
We now know that size variation in the anterior (incisor and canine) and posterior (premolar and molar) teeth is genetically independent in mammals, and as such, these represent two distinct genetic modules (Grieco, Rizk, & Hlusko, 2013;Hlusko et al., 2011). Within the postcanine module, premolars and molars represent two genetic modules that are influenced by different degrees of pleiotropy (Grieco et al., 2013;Hlusko et al., 2011). Within the molar module specifically, mouse development research has revealed that activating and inhibiting signals during development lead to the sequential and integrated development of the first through third molars, a process referred to as the inhibitory cascade (Kavanagh, Evans, & Jernvall, 2007). The inhibitory cascade model of molar size variation describes some mammalian clades better than others with support for this model published for some carnivorans and rodents (Evans & Jernvall, 2009), catarrhine primates (Schroer & Wood, 2015), and K E Y W O R D S Boreoeutheria, dentition, diet, life history, phylogenetic signal fossil mammals (Halliday & Goswami, 2013), whereas several other taxa are reported to deviate significantly from the predictions of the inhibitory cascade including South American ungulates (Wilson, Sánchez-Villagra, Madden, & Kay, 2012), canids (Asahara, 2013), platyrrhine primates (Bernal, Gonzalez, & Perez, 2013), and voles (Renvoisé et al., 2009).
The vast majority of experimental developmental studies have occurred in the highly derived dentitions of mice (Thesleff, 2015(Thesleff, , 2018. While this approach offers valuable insight into genetic mechanisms, it is important to keep in mind that this is evidence of the mechanisms that pattern the murine dentition. The distinct evolutionary history of rodents resulted in highly derived and reduced dentitions, a potentially significant caveat to the developmental genetics of this model system. Research that focuses on more evolutionarily conserved mammalian dentitions (primates, cervids, and equids) will provide essential insight into the more generalized genetic mechanisms that facilitated and constrained the evolution of mammalian dental variation and, consequently, this key mammalian innovation. To date, experimental manipulation of the development of mammalian dentitions with all four classes of teeth has been limited (but see Moustakas, Smith, & Hlusko, 2011). Much of our understanding of the G:P dental variation map for more generalized, evolutionarily conserved mammalian dentitions derives from quantitative genetic analysis of primates (Hlusko, 2004;Hlusko, Lease, & Mahaney, 2006;Hlusko & Mahaney, 2009;Rizk, Amugongo, Mahaney, & Hlusko, 2008).
The quantitative genetic approach to G:P mapping has revealed evidence of two independent genetic patterning mechanisms that influence dental proportions, or the relative sizes of teeth, in the postcanine dentition . Ratios of the mesiodistal dimensions of the fourth mandibular premolar:second mandibular molar (premolar-molar module, PMM) and first molar:third molar (molar module component, MMC) capture the phenotypic effects of these mechanisms . The PMM and MMC in primates are highly heritable, independent of body size, and underlain by as-of-yet uncertain genetic patterning mechanisms. The MMC is likely related to the inhibitory cascade. However, the inhibitory cascade is morphologically described by the two-dimensional area for the first and third molars (mesiodistal length multiplied by buccolingual breadth; Kavanagh et al., 2007). Because earlier quantitative genetic research found that buccolingual breadth has pleiotropic effects with body size in primates (Hlusko et al., 2006), the MMC relies only on mesiodistal length. Therefore, and in light of the caveat we raised about the potentially derived developmental mechanisms of murines, we use a description of the anatomical structure to define this trait rather than referring to it by a presumed but unconfirmed developmental mechanism. Prior research shows that the MMC and PMM vary with strong taxonomic discrimination across extant and extinct primates . Here, we extend this research to test the hypothesis that PMM and MMC will have strong phylogenetic signals across mammals more broadly.
The permanent postcanine dentition (premolars and molars) develops and erupts throughout ontogeny in most boreoeutherian mammals with some species erupting their molars well after reproductive maturity (e.g., humans and suids; Hillson, 2005). Many life history traits emerge from coordinated changes during ontogeny (Stearns, 2000), and some aspects of dental variation have been shown to be associated with life history in mammals (e.g., timing of the eruption of the first molar, Smith et al., 1994) due in part to the slow development of the permanent dentition. However, recent work continues to emphasize the importance of considering phylogenetic relatedness when interpreting dental variation, as dental traits that were previously associated with life history have been shown to vary with conserved phylogenetic signal independent of life history when considered in a broader phylogenetic framework (Monson & Hlusko, 2018a, 2018b. As such, we considered both life history and phylogenetic relatedness in our investigation of postcanine dental proportions in mammals. Here, we utilized a large morphological dataset spanning Boreoeutheria to assess how conserved or labile these two genetic patterning mechanisms (PMM and MMC) are in the evolution of mammalian dental variation. Boreoeutheria is comprised of two of the major extant eutherian mammalian clades that span a wide range of dietary, behavioral, and ecological adaptations and can be found on every major continent as well as in all major oceans: Euarchontoglires (primates and colugos, treeshrews, and rodents and lagomorphs) and Laurasiatheria (cetartiodactyl and perissodactyl ungulates [the former including cetaceans], carnivorans, pangolins, bats and flying foxes, and hedgehogs, shrews, moles, and solenodons; Nowak, 1999). Many species of nonboreoeutherian placental mammals, afrotherians and xenarthrans (e.g., sloths, anteaters, elephants, and armadillos), are characterized by highly derived and even absent dentitions (Hillson, 2005) and were not included in this study, although future studies that include these taxa may provide an illuminating comparison across mammals more broadly. Instead, for this study we focused exclusively on terrestrial boreoeutherian mammals with complete postcanine dentitions. Boreoeutherians are thought to have evolved approximately 100-80 Ma, with the first fossils definitively attributed to this clade dated to 65 Ma (Archibald, 2003;Kemp, 2005;O'Leary et al., 2013). This clade is ideal for our investigation as it encompasses the vast majority of extant eutherian mammals and a diverse array of dental variation and dietary niches. Additionally, as much of the work on mammalian dentition has focused on humans and other primates (Butler, 1963;Hlusko & Mahaney, 2009;Line, 2001;Townsend, Harris, Lesot, Clauss, & Brook, 2009), this study further contextualizes our understanding of the evolution of primate dental variation within boreoeutherian and mammalian evolution more broadly.
We assessed PMM and MMC across a large sample of mammals that includes N = 1,523 individuals spanning 14 families and two of the major eutherian clades: Euarchontoglires and Laurasiatheria (together known as Boreoeutheria). We combined this large dental phenotypic dataset with eight life history variables and adult body mass to test three hypotheses: (H1) Postcanine dental proportions (as captured by the MMC and PMM ratios) vary significantly across mammals; (H2) there is strong phylogenetic signal in postcanine dental proportions (MMC and PMM) across mammals; and (H3) variation in postcanine dental proportions (MMC and PMM) is associated with variation in diet and life history in mammals.

| Materials
The sample for this study includes N = 1,523 mammals spanning six orders, 14 families, 36 genera, and 49 species of Boreoeutheria (Table 1). We focused exclusively on terrestrial taxa with premolars and molars. These data represent the efforts of thousands of hours of data collection by more than a dozen researchers at 13 different museums across six countries (Table S1), and this is the most comprehensive investigation of mammalian dental proportions to date.

| Data collection and analytical methods
We assessed only adult individuals with complete postcanine dentitions (fourth premolars [P4] to third molars [M3]). As MMC and PMM were described for the mandibular dentition of primates (Hlusko et al., 2016), and because many laurasiatherian mammals have third molars in the mandible and not the maxilla, we focused on mandibular dentitions for this study. We took the length of the mandibular premolars and molars of each individual using Mitutoyo calipers according to previously described protocols (Grieco et al., 2013). Length was measured as the mesiodistal length with some variation based on positioning of the teeth. Due to the immensity of the data collection undertaken, multiple researchers took measurements, specializing on subclades of the larger sample. While not all measurements were taken by a single researcher, data collection of dental linear metrics has been common practice for over a century, and all researchers followed standardized protocols (Grieco et al., 2013 We used the left side of the dentition unless measurements could not be taken, in which case we used the right side of the dentition in the analyses. For taxa that were measured by multiple investigators (e.g., Lagostomus; see Table S1 for more details), a subsample of ten specimens was measured three times to confirm that interobserver

TA B L E 1 Boreoeutherian species sampled in this study
unit-free and calculated consistently across all taxa, and as a previous quantitative genetics study found no correlation between body size and these ratios , no other size correction was done for these two traits. Next, we conducted a series of descriptive statistics by order, family, and genus using the describe By function in psych (Revelle, 2015).
We produced bivariate plots comparing MMC and PMM across Boreoeutheria at several taxonomic levels (genus, family, order) using qplot in ggplot2 (Wickham, 2016). To compare MMC and PMM across families (H1), we conducted a phylogenetic ANOVA using the aov.phylo function in geiger and a published mammalian phylogeny (Faurby & Svenning, 2015;Harmon et al., 2016). We trimmed the phylogeny according to the species included in our sample. All species in our sample were represented in the phylogeny except for Equus burchelli, Mazama bricenii, and Odocoileus hemionus, and as such these taxa were excluded from the phylogenetic analyses and included only in the descriptive statistics and bivariate plots.
In order to test for phylogenetic signal in the dental ratios and all life history variables (H2), we conducted tests for Blomberg's K and Pagel's lambda (Kamilar & Cooper, 2013). For Blomberg's K, a value >1 suggests a stronger phylogenetic signal than expected under Brownian motion (BM), while a value equal to 1 suggests that the traits vary along the phylogeny in a manner consistent with BM, and a value <1 suggests that the traits vary along the phylogeny in a manner that is more random than expected under BM and may be the result of selection on those phenotypes (Blomberg, Garland, & Ives, 2003). For Pagel's lambda, a value closer to 1 indicates higher phylogenetic signal, while a value closer to 0 indicates lower phylogenetic signal. Both analyses test for phylogenetic signal but under different frameworks (Blomberg et al., 2003;Pagel, 1999). For Blomberg's K, the variance is partitioned according to clades, where a K > 1 indicates significant variance between clades, and a K < 1 indicates variance within clades (Blomberg et al., 2003;Molina-Venegas & Rodríguez, 2017).
In contrast, Pagel's lambda tests for similarity of covariance among species against covariance expected under a BM model (Molina-Venegas & Rodríguez, 2017;Pagel, 1999). Variation in the availability of life history data results in different species sample sizes for each trait. Additionally, more comprehensive taxonomic sampling across clades will likely improve our understanding of the relationship between postcanine dental variation and phylogenetic relatedness in mammals.
To further interpret phylogenetic signal and better contextualize the evolution of dental proportions (H2), we estimated ancestral mammalian MMC and PMM values, as well as ancestral values for the life history variables, and generated a series of ancestral state reconstructions (ASR) using contMap in phytools (Revell, 2012) which maps continuous variables along a phylogeny of interest. We quantified the estimated values at internal nodes using fastAnc in phytools (Revell, 2012), a function that generates maximum-likelihood ancestral states for continuous traits. Because there is some evidence that ratios can be poorly modeled by Brownian motion (ratios are unlikely to increase linearly with time), we also ran an ancestral state reconstruction using dental lengths and calculated the ancestral MMC and PMM using reconstructed ancestral dental lengths.
To investigate potential correlates with MMC and PMM variation, we collected data on life history variables from the AnAge database, part of the Human Ageing Genomic Resources database (Tacutu et al., 2013). In each case, we used the species average of MMC and PMM. Previous studies have linked life history variables such as longevity and age at weaning to the timing of tooth formation in primates (Smith, 1989;Smith et al., 1994), and recent work hypothesized that variation in prenatal growth rates is associated with tooth number and development of the postcanine dentition . Building from this literature, we compared dental metrics and proportions with a series of life history variables in this study including gestation length (days), birth weight (grams), adult weight (grams), longevity (years), litter size, age at weaning (days), and age at female sexual maturity (days). We also calculated prenatal and postnatal growth rates according to standard protocols, where prenatal growth rate is the ratio of birth weight to gestation length, and postnatal growth rate is the ratio of adult weight to days to sexual maturity (Montgomery & Mundy, 2013). All life history and body size variables were log-transformed for analyses with the exception of litter size. It has been previously hypothesized that slower prenatal growth rates can lead to reduction or complete lack of development of the third molars in primates . The first and second generations of mammalian teeth begin developing in utero and finish erupting well past sexual maturity in many taxa (e.g., humans) and are thereby subject to many stages of life history including gestation, labor and delivery, weaning, and sexual maturity (Smith, 2018;Tucker & Sharpe, 2004). Based on previous work on primate dental proportions Monson et al., 2019), this study predicts a correlation between MMC and/or PMM and life history in boreoeutherian mammals (H3).
In order to assess any correlations between MMC and PMM and diet in our sample, we collected data on diet from the literature (Nowak, 1999). Animals were classified into one of six dietary categories based on their dominant food sources as detailed by a compilation of observational, fecal, and stomach content studies in Nowak (1999): carnivore, folivore, frugivore, granivore, grazer, and omnivore. The original sources referenced in this compilation of studies varied in method used to determine preferred food type, an important caveat when considering relationships between diet and morphology. Information on classification of individual species can be found in the life history and diet dataset, available in Table S2.
To directly compare life history variables with variation in MMC and PMM (H3), we ran a series of phylogenetic independent contrasts using the crunch function in caper (Orme et al., 2013 As tooth length is a long-standing metric for investigating diet and body size, and dental proportions are calculated from tooth lengths, we regressed individual tooth lengths against cube root body mass and compared the residuals for each tooth against diet across the phylogeny in a PGLS analysis to further compare variation in these traits. Cube root body mass was used here to account for scaling and allometric relationships between tooth length and adult body mass (Ungar, 2015).  There is also no clear pattern of discrimination by diet, reinforcing the lack of significant association between these traits in a phylogenetic context in this sample (Figure 1c,d).  Figure S1). Of the extant clades sampled here, the ancestor of all primates, the ancestor of anthropoid primates, and the ancestor of  (Table S5, Figure S2). Interestingly, due to the diver-  Figure S2).

| RE SULTS
Overall, extant African and Asian monkeys (Cercopithecidae) have dental proportions most similar to the ancestral MMC and PMM values predicted by ancestral state reconstruction in this study. This is supported both when ASR is applied to the MMC and PMM ratios and when ASR is applied to raw dental lengths and MMC and PMM are calculated from reconstructed ancestral values (Table S6, Figure   S3). we find that the residuals are significantly associated with diet for all teeth (p < 0.001) with greatest significance in the first molar (Table 5).

| D ISCUSS I ON
Tooth size, dental proportions, and tooth crown morphology have all been used as proxies for the interpretation of diet in the fossil record (Boyer, 2008;Cardini & Elton, 2008;Caumul & Polly, 2005;Fortelius, Made, & Bernor, 1996;Janis, 1984Janis, , 1997Janis et al., 1998;Jernvall et al., 1996;Ungar, 1998Ungar, , 2017Walker, 1981). We analyzed ratios (PMM and MMC) that reflect the phenotypic output of two genetic patterning mechanisms on the mammalian postcanine module. Our data demonstrate that the relative sizes of premolar and molar teeth, as captured by the MMC and PMM ratios of dental length, are significantly different across boreoeutherian mammals and have strong phylogenetic signal. We interpret this association with phylogenetic relatedness to be evidence that tooth proportionality is highly conserved over evolutionary time, and variation in dental proportions, particularly molar proportions, generally reflects variation in phylogeny over variation in diet. This is shown through tests of phylogenetic signal as well as clear taxonomic discrimination at the genus and family levels in bivariate space. In contrast, the MMC and PMM traits do not vary significantly with diet in a phylogenetic context at this broad taxonomic scale. Some previous studies have associated variable proportions of postcanine tooth length to diet in primates (Asahara, 2013;Lucas, Corlett, & Luke, 1986), but our more taxonomically comprehensive study reveals significant phylogenetic signal that is largely independent of variation in diet, although some individual taxa have taxon-specific dental adaptations that contribute to variation in the MMC and PMM (Asahara & Takai, 2017). This suggests that MMC and PMM may evolve in tandem with the morphology of taxon-specific dental adaptations, such as the carnassials of carnivorans, and the reduced third molars of carnivorans and bats.
While variation in dietary strategies within clades is statistically independent of changes in relative postcanine dental proportions (suggesting that dental proportions contribute less to dietary adaptations than do other cranial and dental phenotypes), individual tooth length measurements relative to adult body mass are significantly correlated with diet. This is consistent with previous studies that found significant allometric relationships between tooth length and body mass, and significant associations between diet and individual tooth lengths (Asahara & Takai, 2017;Copes & Schwartz, 2010;Scott, 2011;Scott et al., 2018). This significant relationship is also likely influenced by ancestral dietary "bauplans" of different clades, where individual teeth have evolved unique morphologies as functional adaptations to processing particular foods (Hunter & Jernvall, 1995;Kay, 1975Kay, , 1977Lucas, 1980;Ungar, 2009;Ungar et al., 2018). The length of the mandibular first molar is most significantly correlated with diet in our sample, likely related to the modification of this tooth into a carnassial for the processing of animal tissues in many species of Carnivora (Asahara & Takai, 2017).
Adaptations to increased biomechanical torque and lever forces associated with the enlarged P 4 /M 1 carnassial complex (Van Valkenburgh, 1991) also likely contribute to the dietary trend identified in our data where all carnivorous species sampled have an average MMC that is <1. Omnivorous members of Carnivora, represented here by Ursus americanus, also have an MMC <1, likely retained from their ancestral dental "bauplan" which included carnassials (Butler, 1946). Less is known about the dental proportions of carnivorous mammals in other orders such as the Tasmanian devil (Sarcophilus harrisii), a marsupial that does not technically have carnassials but does have long shearing blades and retains four molars in the adult dentition (Marshall & Corruccini, 1978;de Muizon & Lange-Badré, 1997;Werdelin, 1987Werdelin, , 1988.  species that evolved relatively recently, over the last 700,000 years (Cahill et al., 2013;Edwards et al., 2011;Hailer et al., 2012;Kurtén, 1964;Slater, Figueirido, Louis, Yang, & Valkenburgh, 2010;Talbot & Shields, 1996). While polar bears exhibit reduced surface area of the postcanine dentition, a feature associated with increased carnivory (Slater et al., 2010), the relative proportions of their postcanine teeth are much more similar to those of their omnivorous relatives and distinct from other carnivorous mammals (e.g., would offer further insight into dental evolution in this family.
Additionally, assessing MMC and PMM in a clade with several taxa with highly divergent/specialized diets could give us a better idea of the extent of phylogenetic inertia in these traits and further refine the timeline for significant morphological divergence in dental proportions.
We also found a lack of correlation between postcanine dental proportions and life history characteristics. Some aspects of the dentition, such as rates and timing of enamel deposition, provide essential insight into variation in life history (Smith, 2018). Our results demonstrate that other aspects of the dentition are decoupled from life history as has been seen in other studies (Monson & Hlusko, 2018a, 2018b. Our analyses indicate that life history and diet may be more evolutionarily labile than postcanine dental proportions and, as such, more responsive to selective pressure. In contrast, postcanine dental proportions likely require significant selective pressure over long timescales to diverge from the ancestral condition. Several mammalian lineages are characterized by significant deviation from early mammals suggesting that they experienced bouts of strong evolutionary pressure (e.g., murines).
To better understand the evolution of postcanine dental proportions in mammals, we performed a subsample analysis comparing our ancestral state reconstructions with data from the fossil record, Protungulatum mckeeveri, Archibald, 1982). Of the fossils sampled, the archaic ungulates are the most ancient, dated to the early Paleocene of North America (Archibald, 1982;Archibald, Schoch, & Rigby, 1983). Of the carnivorans, Arctotherium is the most recent, dated to the Pleistocene (Trajano & Ferrarezzi, 1994). Cyonarctos and Ursavus are dated to the Oligocene and Miocene, respectively (de Bonis, 2013;Qiu et al., 2014), and Campylocynodon (alternately classified as Parictis; Clark & Guensburg, 1972) is dated to the Oligocene of Europe and North America (Chaffee, 1954).
The inclusion of fossil data into our plots of MMC and PMM demonstrates that the oldest fossils (archaic ungulates) fall close to the 1:1 axis of MMC and PMM variation (Figure 4) near the predicted ancestral condition for Boreoeutheria (1.1, 1.2; Table S5). This pattern provides further evidence that ancestral mammals had more homogeneous postcanine dental proportions as has been noted in previous studies (Halliday & Goswami, 2013). for being a very "canid-like" early ursid (de Bonis, 2013; Ginsburg & Morales, 1998). Based on the strong phylogenetic signal in dental proportions observed in the current study, and the clear distinction between dental proportions of extant ursids and canids, we suggest that either modern ursid dental proportions evolved relatively recently, or a reconsideration of the phylogenetic affinities of Cyonarctos and possibly also other hemicyonines may be warranted.
Our results also belie necessary caution when interpreting diet of fossil mammals exclusively from postcanine dental proportions (as captured by MMC and PMM), suggesting that other features of the dentition and skull, including individual tooth lengths relative to body size, are likely more useful for reconstructing diet.
However, MMC and PMM dental proportions can play an important role in understanding the phylogenetic relatedness of extinct mammals, as these traits have strong phylogenetic signal in extant mammals. Our initial exploration of the fossil record also suggests that variation in mammalian dental proportions largely reflects bauplans that were established early in mammalian evolution and that are relatively stable over tens of millions of years. The fossil evidence supports our interpretation that there is significant phylogenetic constraint on the evolution of dental proportions within Boreoeutheria, as fossil mammals tend to have dental proportions similar to their extant counterparts. A larger assessment of the pattern of variation in dental proportions beyond Boreoeutheria to establish the "break points" in phylogenetic constraint will likely make MMC and PMM even more useful for assigning fossils to taxonomic groups.
In summary, we find that variation in postcanine dental proportions (as captured by MMC and PMM) accumulates slowly and characterizes mammalian lineages as they diversified from the plesiomorphic/ancestral ratios. Ancestral eutherian mammals had relatively homogeneous postcanine dental proportions, where the fourth premolar and all molars were similar in size (Butler & Clemens, 2001;Halliday & Goswami, 2013;Sloan & Van Valen, 1965;Ungar, 2010). From that homogeneous condition, several mammalian lineages diversified into distinct extant morphospaces that characterize F I G U R E 4 Variation in MMC and PMM visualized at family-level with fossil species marked with a star and indicated by an uppercase letter. Broader taxonomic affiliation and geological ages of fossils are in parentheses following the species name. See figure for legend the evolution of those groups (see Chiroptera, Canidae, Ursidae, and Suidae

CO N FLI C T O F I NTE R E S T
None declared.

DATA ACCE SS I B I LIT Y
All raw data are available in the Supporting Information.