The Earliest Ursine Bear Demonstrates the Origin of Plant-Dominated Omnivory in Carnivora.

In Carnivora, increases in body size often lead to dietary specialization toward hypercarnivory. Ursine bears (Tremarctos and Ursus), however, are the only omnivorous Carnivora that evolved large body sizes (i.e., >50 kg). Traits contributing to their gigantism, and how those traits evolved, have never been studied. Here we propose that special dental characters of Ursinae (parallel buccal and lingual ridges) permit a sagittally oriented mastication associated with increasing emphasis on plant foods. This pattern can be traced back to a new early diverging bear of plant-dominated omnivorous diet, Aurorarctos tirawa gen. et sp. nov. from the late Middle Miocene of North America, which was supported as the earliest known ursine bear by phylogenetic analysis. The anatomical transition to increased masticatory efficiency, probably together with the ability to hibernate, helped bears break prior ecological limitations on body size and led to the evolution of a distinctive lineage of herbivorous-omnivorous, large-bodied Carnivora.

Introduction

Adaptation to omnivory, which increases the flexibility of diet choice from a dominantly carnivorous ancestry, repeatedly evolved within Carnivora (Ewer, 1973). Most omnivorous Carnivora lineages, e.g. Melinae and Taxidiinae in Mustelidae (Ginsburg and Morales, 2000); most members of Procyonidae (Baskin, 1998a); Conepatus and Mydaus in Mephitidae (Baskin, 1998b); Nyctereutes and the extinct clades Phlaocyonini and Cynarctina in Canidae (Wang et al., 1999, Tedford et al., 2009); and Paradoxurinae in Viverridae (Ewer, 1973), however, retain small or medium sizes, seldom reaching a body mass larger than 20 kg, suggesting the presence of ecological or morphological constraints on body mass in omnivorous Carnivora. Analyses of fossil Canidae suggested that selection for larger body sizes would lead to dietary specialization for hypercarnivory (Van Valkenburgh et al., 2004). Similarly, all large terrestrial fossil musteloids, i.e. Megalictis (Valenciano et al., 2016), large Guloninae (Harrison, 1981, Valenciano et al., 2018), Mellivorinae (Valenciano et al., 2015), the large ailurid Simocyon (Wang, 1997), and the large procyonid Chapalmalania (Forasiepi et al., 2014), are hypercarnivores (and often with bone-crushing adaptation). This phenomenon is correlated with energetic constraints on body mass or maximal rate of energy expenditure, since terrestrial Carnivora (except bear) that are larger than 21.5 kg (or slightly larger if the animal has lower basal metabolism rate) can only live upon on large vertebrate prey than smaller invertebrate prey (Carbone et al., 1999, McNab, 2000). This body mass threshold on species relay on small prey can be extended to omnivorous diet species that include significant plant items, because carnivorans, including even the most herbivorous member (the giant panda Ailuropoda melanoleuca), do not have a digestive tract, digestive enzymes, or a gut microbiota well adapted to eating plant material (Nie et al., 2019).

The high energetic requirement and low efficiency in digesting plant food may limit omnivorous Carnivora in evolving large body sizes. Bears (Ursidae) are the exception to the general covariance of large size and Carnivory within Carnivora, with omnivorous ursine lineages evolving to very large sizes, often greater than 100 kg. Energetic demands in large-bodied taxa can be reduced by the ability to hibernate during the cold season (Hellgren, 1998). This behavior, however, cannot explain why only bears evolved a large body size, because hibernation also is present in several omnivorous musteloids, e.g. badgers and skunks (Geiser, 2013). We propose that bears initially evolved a unique masticatory pattern to increase digestive efficiency associated with an omnivorous diet with increasing plant components, which in conjunction with evolution of the ability to hibernate ultimately permitted the evolution of large body sizes even in these omnivorous-herbivorous taxa. There are two derived dental characters in Ursini that clearly increase masticatory efficiency. One is anteroposterior elongation of the posterior molars, specifically M2 and m2, and another is a shift of the masticatory pattern from the ancestral, more transverse chewing direction (latero-medial) to a derived, more sagittal direction (antero-posterior). The latter character is rather distinctive among Mammalia, because most species with increasing grinding functions develop transverse ridges (Figures 1A and 1B). How these two characters evolved is uncertain, however, because early fossil ursine bears have been unknown and no phylogeny has been performed regarding the origins of ursine species and clades.

Figure 1

Masticatory Pattern of Various Mammals, as Reflected by M1/m1 Morphology

(A1 and A2) giant tapir Megatapirus augustus (Perissodactyla) AMNH FM1-8754; (B1 and B2) brush-tail rock-wallaby Petrogale penicillata (Marsupialia) AMNH CA2869; (C1 and C2) spectacled bear Tremarctos ornatus (Carnivora) AMNH M67732; (D1 and D2) Asiatic black bear Ursus thibetanus (Carnivora) IVPP V5601.22 (reversed) and IVPP V5711.4. (E1 and E2) Lambdopsalis bulla (Multituberculata) IVPP V20101. Not to scale. Arrows indicate masticatory direction. Note the sagittally oriented ridges of grooves in Ursus thibetanus and to a lesser extent in Tremarctos ornatus (lower m1 with incomplete lingual ridge).

Here we report a substantial sample of an early diverging North American bear, from late Middle Miocene-aged (∼15–12.5 million-year-old) fossil sites in Nebraska (see Supplemental Information for geological background details) housed in UNSM (University of Nebraska State Museum, Lincoln, Nebraska, U.S.A.), which was preliminarily identified as Ursavus cf. brevirhinus by Hunt (1998). Some traits of this bear are unique to Ursinae, and our analyses suggest that this bear is a new taxon distinct from Ursavus and represents the earliest known member of the Ursinae. It also provides new information for reconstructing initial character states relating to evolution of the unique masticatory pattern of later ursine bears.

Materials and methods, abbreviations, etc. are presented at the end of this manuscript. Two living subfamilies, Ailuropodinae and Ursinae, with the latter being further subdivided into the tribes Ursini and Arctotheriini, are recognized (here we follow (Qiu et al., 2014)).

Results

Systematics

Ursidae Fischer [de Waldheim], 1814 Ursinae Fischer [de Waldheim], 1814

Aurorarctos gen. nov.

Etymology: named after “Aurora,” dawn in Latin, and “Arctos,” for bear in Greek.

Type species: Aurorarctos tirawa sp. nov.

Referred species: so far only known from the type species.

Aurorarctos tirawa sp. nov. (Figure 2)

Figure 2

Specimens of Aurorarctos tirawa gen. et sp. nov.

(A1–A3) UNSM45118, partial humerus, anterior (A1), posterior (A2), and distal (A3) views; (B1–B5) UNSM95237 holotype mandible; B1, buccal view; B2, occlusal view; B3, lingual view; B4–5, stereophotograph of occlusal surfaces of the dentition; (C) UNSM95211, P4, occlusal view; (D) UNSM95209, M1, occlusal view; (E) UNSM95206, M2, occlusal view.

Ursavus cf. brevirhinus (Hunt, 1998)

Etymology: species named after Tirawa, the god of creation for the Pawnee people of the Great Plains.

Holotype: UNSM95237, a left mandible with p4–m2 preserved. (UNSM = University of Nebraska State Museum).

Referred material: see Supplemental Information.

Type locality: Stewart Quarry (Cr150) in Cherry County, northern Nebraska, USA.

Chronology: so far only known from the Late Barstovian (∼15–12.5 Ma) North American Land Mammal “Age” (NALMA).

Diagnosis: small-sized species of early diverging Ursinae. Mandible robust and deepens posteriorly. Premolar series complete; p2–p4 double-rooted; mandible lacking premasseteric fossa or marginal process; P4 lacking parastyle and protocone, and inner lobe located on anterior half of tooth; M1 wide with strong postero-lingual, cusp-like cingulum; M2 with short talon, weak antero-medial accessory ridge of paracone and connected protocone and metacone; p4 with lingual ridge and non-subdivided posterior ridge; m1 with postero-lingual ridge of protocone anteriorly curved, with initial posterior ridge, and variable presence of postero-buccal ridge of protocone, and entoconid often weakly or not subdivided; m2 with wide trigonid and separated paraconid.

Differential Diagnosis

Differs from the genotype species for Ursavus, U. brevirhinus, in having a more massive P4 paracone; wider M1 with strong postero-lingual corner; m1 with anteriorly curved RPrd3 (postero-medial ridge of the protoconid, detailed meaning see (Jiangzuo et al., 2019) and presence of RPrd2 (posterior ridge of the protoconid), with a stronger buccal concavity between the trigonid and talonid; m2 with much wider trigonid and separated paraconid; differs from U. primaevus in smaller size; proportionally shorter M2 talon and connected protocone-metacone; m1 with an anteriorly curved RPrd3 and presence of RPrd2; differs from Ursavus pawniensis in smaller size and less elongated m2 with stronger buccal concavity; differs from Ballusia spp. in larger size with longer M2 talon and derived m1 characters (especially protoconid region); differs from Late Miocene Ursavus species, e.g. Ursavus tedfordi, in smaller size, much larger premolars, larger and more anteriorly located inner lobe of P4, shorter M2 talon, and non-subdivided p4 posterior ridge.

For detailed description and comparison of the fossil material pertaining to this new species, see Supplemental Information.

Phylogenetic Analysis

Phylogenetic analyses of fossil and living bears have been conducted separately on Ailuropodinae (Abella et al., 2012) and Ursinae (Wang et al., 2017), but neither of those studies focused on stem group taxa or was inclusive across Ursidae and thus are not suitable as a comprehensive framework for understanding the relationships and significance of the new taxon. We therefore developed a new character-taxon phylogenetic matrix emphasizing inclusion of more taxa from the stem group of Ursinae. Our new matrix contains 130 craniodental characters and one humerus character (which is the only postcranial character that can be applied to the new species described here). A total of 31 taxa were included, including all living species of Ursidae, with the extant wolf Canis lupus as the outgroup. For details of the character matrix, see the Supplemental Information.

In the phylogenetic analyses, Maximum Parsimony and Bayes Inference (with and without topological constraint, see Methods for details) methods yield similar topologies and support Aurorarctos tirawa as a stem ursine (Figures S6–S8). For relationships within the Ursini, we applied a topological constraint for the relationship of living species of Ursus such as Ursus thibetanus, Ursus malayanus, Ursus ursinus, Ursus americanus, Ursus arctos, and Ursus maritimus, because this topology has been independently confirmed by both a genome-wide phylogeny (Kumar et al., 2017) and a phylogeny based on transposable element insertions (Lammers et al., 2017). Our final tree, used for further analyses, is based on the Bayes Inference phylogeny because this method allows a different evolution rate for each traits, which makes sense (Figure 3).

Figure 3

Bayes Inference Phylogenetic Tree of Ursidae, with a Life Reconstruction of Aurorarctos tirawa gen. et sp. nov.

Note that we apply the total group (crown group + stem group) usage of Ailuropodinae, Ursinae, Ursini, and Arctotheriini. Artwork by Yu Chen.

Ursavus is definitely polyphyletic in our phylogenetic analysis, and none of the other previously recognized species of Ursavus link in an exclusive clade with the genotype species, which is a basal ursid far removed from all the other species previously assigned to Ursavus. Thus, we recognize a new genus and species for the basalmost ursine taxon (previously preliminarily recognized as “Ursavus cf. brevirhinus” by Hunt, 1998 and herein named Aurorarctos tirawa), and all other “Ursavus” species besides the genotype have the genus name in quotation marks, reflecting likely nonmonophyly of this genus. Species previously assigned to this genus include the stem ursid Ursavus brevirhinus (the genotype species), stem ailuropodine “Ursavus” primaevus, stem ursine Aurorarctos tirawa, and three other ursine species (“U.” ehrenbergi, “U.” sylvestris, and “U.” tedfordi) that form a polytomy with the Arctotheriini and Ursini clades. “Ursavus” tedfordi may represent a stem species of Ursini; although this is not resolved by our phylogenetic analysis, because of character conflict suggesting potential affinities with either Ursini or Arctotheriini, the presence of a distinct marginal process and subdivided p4 posterior ridge may support its potential Ursini affinity (Jiangzuo et al., 2019). In addition to its basal position within the Ursinae, Aurorarctos tirawa is morphologically very primitive and thus well exemplifies the ancestral craniodental conditions for ursine evolution.

Discussion

Paleoecology of Aurorarctos tirawa

The paleoecology of the new fossil ursine species can be inferred from four lines of evidence: depth of the mandibular ramus, dental features, caries pits, and morphology of the humerus distal articular surface.

Compared with the complex suite of masticatory and neurosensory functions of the cranium, the mandible functioned only in feeding and thus can be especially useful for inferring paleodiet (Herring, 1993, Piras et al., 2013). In general, a deeper mandibular ramus is correlated with increasing loading resistance (Therrien, 2005). Aurorarctos tirawa has a rather deep mandible (the ratio of m1 versus mandible depth behind m1 is 0.64), close to the relative depth in living bears (see Table S2), in contrast to a much more slender mandible in the earlier stem ursid Ballusia elmensis (ratio = 0.92 according to (Dehm, 1950). Geometric morphometric analysis (Figure 4) also suggests that Ballusia falls within the morphospace occupied by the carnivorous/insectivorous-omnivorous group, whereas Aurorarctos tirawa is located within the range of herbivorous/herbivorous-omnivorous extant bears. This suggests that the origin of an herbivorous-omnivorous diet in living bears evolved through a transitional stage, exemplified by Ballusia, of an insectivorous-omnivorous diet, and that the herbivorous-omnivorous specialization of ursines already had been acquired by the appearance of the earliest representative of Ursinae, Aurorarctos tirawa.

Figure 4

Geometric Morphometric Analyses (PCA) of Aurorarctos tirawa and Related Carnivora, and Evidence of Dental Caries in the New Taxon

(A) Geometric morphometric analysis of mandible profile; (B) geometric morphometric analysis of distal humerus; (C) mandible landmarks (in this figure, the semilandmarks were shifted slightly lower than the mandible ventral contour for improved visibility); (D) distal humerus landmarks; (E) dental caries (boundary marked by white arrows) in four molars of Aurorarctos tirawa: E1 UNSM95200, E2 UNSM95237, E3 UNSM45086, E4 UNSM95237.

Dental characters also support an omnivorous diet for Aurorarctos tirawa. The cusps of Aurorarctos tirawa teeth are blunt and low-crowned. It is especially noteworthy that the pair of m1 entoconids commonly seen in Miocene Ursidae tend to be fused in A. tirawa. This character decreases the puncturing ability relative to that seen in insectivorous carnivores, such as mongooses (Herpestidae), badgers (Meles/Arctonyx), and the fossil stem ursid Ballusia. An omnivorous diet is further supported by the presence of dental caries. Large pits are frequently present in the trigon/trigonid or talon/talonid basins, as observed in three (representing four teeth, see Figure 4A) of the total of 14 specimens of this new taxon that preserve teeth. Evidence of dental caries is distinct from the normal wear pattern of the teeth, which begins at the apices of the cusps rather than from within the basin. In all four of these teeth, the apices have much weaker wear than the basin, so basin pitting clearly represents dental caries. Dental caries is common in living bears (Hall, 1945, Jin et al., 2015) and has been observed in several fossil ursine bears (Gross, 1931, Soibelzon et al., 2014, Wang et al., 2017). The extremely high frequency of dental caries (21.4%) in Aurorarctos tirawa, much higher than in extant bears (2% according to Hall (1945)), indicates that this fossil bear probably relied heavily on sugar-rich fruits in its diet.

All extant ursine bears have a relatively robust metapodial system and employ a plantigrade gait. In contrast, most of the common Early and Middle Miocene bears, usually classified as Hemicyoninae, are digitigrade with elongated limbs (Frick, 1926, Ginsburg and Morales, 1998). Little is known of the postcrania of Ballusia or species of Ursavus, because their fossils are almost exclusively represented by isolated teeth or jaw fragments. One exception is the late Early Miocene species Ballusia orientalis from the Shanwang Biota in Shandong Province, eastern China (Qiu et al., 1985). That species has whole skeletons preserved and exhibits slender postcranial bones that differ markedly from living bears but are somewhat similar in proportions to hemicyonids (Qiu et al., 1985). Therefore, there are large morphological gaps between the early stem ursid Ballusia and living ursine and ailuropodine bears. When the body plan typical of extant ursine bears evolved has been unknown. The humerus of Aurorarctos tirawa, although incomplete, provides useful information on the locomotor pattern of this earliest-diverging ursine. Its deltoid tuberosity is greatly extended along most of the humerus, with overlap of its distal part with the lateral epicondyle of the humerus in the shaft, just as in living bears (Davis, 1964), and the epicondyle also is well developed, although not so expanded as in extant ursine bears. The trochlea facet also is similar to living bears, with a weak trochlea ridge. The antero-posterior depth of the humeral condyle is shallower than in extant bears (Figure 2). There are two other major differences between Aurorarctos tirawa and extant bears. The first is the presence of a large entepicondyle foramen. This foramen is absent in Ursini and present but smaller in Arctotheriini (Erdbrink, 1953, Davis, 1964, Hunt, 1998). The second is the overall more slender form of the humerus in A. tirawa, in both the epicondyle and shaft. To investigate the potential locomotor style of A. tirawa, we performed a geometric morphometric analysis of the distal humerus (Figures 4B and 4D). The results clearly suggest that A. tirawa is distinct from living scansorial bears as well as cursorial Carnivora, but falls into the morphospace of small, arboreal-fossorial Carnivora such as Arctictis, Ailurus, Nasua, Meles, and Taxidea, but its epicondyle is definitely weaker than that of fossorial badger. Overall, the anatomy of the humerus (see details in Supplemental Information) suggests that Aurorarctos tirawa most likely was an arboreal member of the Carnivora.

In summary, Aurorarctos tirawa had a unique ecological habitat relative to the stem ursid Ballusia and to extant bears, in its distinctive combination of herbivorous-omnivorous and arboreal adaptations. In contrast, Ballusia, the nearest relative of crown Ursidae probably was more insectivorous-omnivorous and terrestrial. This suggests that the scansorial habitus of living ursines is not directly evolved from a terrestrial ancestor as was estimated by the ancestral state reconstruction for living species (see Figure S16), through a badger-like body plan, but instead stemmed from an arboreally adapted body plan. This fossil-based inference is further supported by the observation that most medium-sized extant bears still retain substantial abilities to climb trees (Fitzgerald and Krausman, 2002).

Origin of the Unique Bear Masticatory Pattern

We propose that in bears the two derived ursine dental characters mentioned in the introduction are correlated, because the shift to a more sagittally oriented masticatory pattern can generate higher masticatory antero-posterior efficiency, which in turn favors selection toward antero-posteriorly elongated molars. This shift of masticatory pattern in ursines can now be reconstructed as having been achieved by sequential opening of the protoconid-metaconid connection of the m1 (Figure 5A), which represents the original anterior boundary of m1 movement during mastication (Figure 5-B1), permitting occlusal motion in an antero-posterior direction unimpeded by a transverse crest. In Aurorarctos tirawa, the postero-medial ridge (RPrd3) is anteriorly curved, opening additional space for antero-posterior movement (Figure 5-B2). Arctotheriini generally retain their m1 morphology at this state (or slightly more derived), but most living Ursini species are more derived in the ultimate loss of RPrd3, permitting full development of the antero-posterior masticatory pattern without any transverse barrier for such movement (Figure 5-B3). The only living Ursus species retaining a connected protoconid-metaconid is the sun bear Ursus malayanus (Figure 5A), which also retains less elongated posterior molars relative to other species of Ursus, indicating that this crest was lost independently several times in the Ursus lineage or that a similar connection evolved again only in U. malayanus after ancestral loss in Ursini. We map the length ratios of M2/M1 and m2/m1 on the phylogenetic tree to trace the elongation of molars in Ursidae (Figures 5C and 5D). In the basal ursine Aurorarctos tirawa, M2/m2 are not elongated, but in other Ursinae, especially Ursini, M2/m2 are distinctly elongated. In contrast, Ailuropodinae only have a weakly or moderately elongated M2/m2, as in Aurorarctos tirawa. Species with a more derived RPrd3 (state 0 is not anteriorly curved, 1 is anteriorly curved, and 2 is loss of RPrd3) also generally have a longer M2/m2 (Figure 5E). Phylogenetic generalized least squares analyses of the ratios of M2/M1 and m2/m1 relative to the m1 RPrd3 character state also support a positive significant correlation between the two former traits. This suggests that the transition from the original posteriorly curved state through the intermediate anteriorly curved state to final loss of the m1 RPrd3 is a key innovation that permitted a shift of masticatory direction more antero-posteriorly, and in turn favored antero-posterior elongation of the posterior molars, permitting more efficient mastication in an increasingly herbivorous-omnivorous lineage and evolution of larger body sizes. In contrast, Ailuropodinae retain the primitive condition for the RPrd3 and a strong postglenoid process (Figure S17) that restricts antero-posterior jaw movements, and this clade of large bears employed a different pathway to increase masticatory efficiency, i.e., transversely widening the molars without changing the jaw biomechanics. This innovation seems to be less efficient than that of Ursinae, as the largest terminal herbivorous taxon Ailuropoda melanoleuca, which is specialized in skull, dental, postcranial, and physiology (Davis, 1964, Endo et al., 1999, Nie et al., 2015), is still not comparable Ursinae in body size (Figure S18). The unique pattern of increased masticatory efficiency in Ursinae documented by the new stem ursine taxon and ancestral state character reconstructions, followed by or together with evolution of the ability to hibernate, helped ursine bears break the energetic and ecological constraint on body size for herbivorous-omnivorous Carnivora and contributed to their later success as distinctive large-bodied herbivores-omnivores in the dominantly carnivorous Carnivora clade.

Figure 5

Dental Evolution in Ursidae

(A) Evolution of the m1 protoconid region.

(B1–B3) P4/m1 occlusion relationships in three taxa, representing the shift of jaw biomechanics-masticatory direction in bears; B1, Ailuropoda spp. (representing the ancestral condition for ursids); B2, Aurorarctos tirawa; B3, Ursus deningeri.

(C) Evolution of ratio of M2 length versus M1 length along the phylogenetic tree (Figure 2).

(D) Evolution of ratio of m2 length versus m1 length along the phylogenetic tree (Figure 2); (E) boxplots of the ratios of M2/M1 length and m2/m1 length, relative to different states of m1 postero-lingual ridge (state 0 is not anteriorly curved, 1 is anteriorly curved, and 2 is loss of RPrd3).

Limitations of the Study

In this study, we performed the phylogenetic analysis mainly based on dental traits, as the new fossil only has these remains. A more detailed functional analysis would be ideal pending on the discovery of more fossil material. Our codings of European species Ursavus brevirhinus and Ursavus primaevus are only based on materials from type localities, and materials from other localities are probably heterogenetic and need a more systematic review. Our explanation about body size evolution and dietary change still lacks a very robust statistical support, as most fossil species only have very limited material, and physiological factors (e.g. hibernation, basal metabolism rate) cannot be tested for fossil species.

Resource Availability

Lead Contact

Qigao Jiangzuo, jiangzuo@ivpp.ac.cn.

Materials Availability

The fossils are housed in the University of Nebraska State Museum, Lincoln, Nebraska, U.S.A.

Data and Code Availability

The data are available in the Supplemental Information.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.