Main

The ecological expansion of crown Mammalia accelerated in the wake of extinctions of diverse and disparate archaic mammaliamorph groups during the mid-Mesozoic1. Haramiyidans represent one such clade of early mammaliamorphs that show preservation of notable links between nonmammalian and mammalian structure and physiology2,3,4. Their fossils, which were dated to the Late Triassic–Jurassic, were historically limited to isolated teeth and incomplete gnathal remains4, 5, hindering a more complete understanding of their radiation. Recent discoveries of articulated skeletal material of eleutherodont haramiyidans from the Middle–Upper Jurassic of China have shed new light on their diversification6, 7, although these fossils have also sparked new debate over their phylogenetic position4, 6,7,8,9,10,11,12. Haramiyidans have frequently been included with multituberculates in ‘Allotheria’, a disputed taxonomic grouping that was diagnosed by dentition, with molars that have multiple cusp rows and palinal (anteroposterior) motion during mastication4, 6, 9. The limited overlapping morphology that is available in eleutherodonts and earlier haramiyidans (for example, Thomasia and Haramiyavia), coupled with a sparse fossil record, complicates our ability to resolve the group’s global radiation and purported relationships to multituberculates. Here, we report an exceptionally preserved cranium from the basal Cretaceous of Utah, USA, that fills substantial morphologic and spatiotemporal gaps in our understanding of this important group. The specimen provides the basis of a new genus and species.

Mammaliaformes sensu Rowe (1986)13

Haramiyida Hahn, Sigogneau-Russell and Wouters (1989)2

Hahnodontidae Sigogneau-Russell (1991)14

Cifelliodon gen. nov.

Cifelliodon wahkarmoosuch sp. nov. (Fig. 1)

Fig. 1: Cranium of C. wahkarmoosuch.
figure 1

ae, Holotype in dorsal (a), left lateral (b), frontal (c), ventral (d) and occipital (e) views. as, alisphenoid (epipterygoid); bs, basisphenoid; C, upper canine alveolus; f, frontal; f i, incisive foramen; f l, lacrimal foramen; g, squamosal glenoid; I2, first upper incisor alveolus (homologous to second incisor position in earlier haramiyidans); I3, second upper incisor alveolus (homologous to third incisor position); j, jugal; l, lacrimal; m, maxilla; n, nasal; o, occipital; os, orbitosphenoid; p, parietal; p pr, paroccipital process; pal, palatine; PC4, in situ posterior upper postcanine (molar); pe, petrosal; pm, premaxilla; pp, postparietal; pt, pterygoid; sq, squamosal; t, tabular; v, vomer.

Etymology. Cifelli’s tooth (Latin: -odon) of the Yellow Cat (Ute language: yellow, wahkar; cat, moosuch). Genus name honours Richard Cifelli for his contributions to Cretaceous mammal research in the American West.

Holotype. An exceptionally preserved skull, UMNH VP 16771 (Natural History Museum of Utah, Vertebrate Paleontology Collection).

Locality and horizon. The holotype is from the ‘Andrew’s Site’ quarry in the Lower Cretaceous Yellow Cat Member, Cedar Mountain Formation, Grand County, Utah, USA15 (Extended Data Fig. 1). Radiometric dating places the age between approximately 139 and 124 million years old (see Methods and Supplementary Information).

Diagnosis. Medium-to-large Mesozoic mammaliaform with broad, shallow skull and rostrum and a reduced marginal tooth count; dental formula: I2:C1:PC4; ultimate upper molars with high anterobuccal cusp and low, broad posterolingual cusp connected by a low ridge; septomaxilla absent; incisive foramina enlarged and positioned posteriorly on palate behind the level of the last (posterior) incisor pair; massive pterygoid transverse process that extends far ventral to the palatal surface; attenuated lacrimal anterior process with limited nasolacrimal contact; prominent sagittal crest; extensive occipital exposure of parietal and postparietal; plesiomorphic retention of a tabular bone; differs from Hahnodon in its larger size and higher aspect ratio of the rear molar in occlusal view (slightly more triangular than oval, with posterior apex).

Description. The skull of Cifelliodon (Fig. 1 and Extended Data Figs. 25) is large relative to other Mesozoic mammaliaforms (basal length around 70 mm; estimated body mass 0.91–1.27 kg)16, 17 (see Methods and Supplementary Information). The upper dentition is largely missing, but preserves alveoli for two incisors, one canine and four postcanine maxillary teeth, including an in situ, unerupted ultimate molar pair (Fig. 2). The incisor roots were strongly posteriorly reclined and converged medially, suggesting procumbency as in haramiyidans, some multituberculates and gondwanatheres—another enigmatic group of mainly herbivorous mammal relatives that was restricted to Gondwana and has sometimes been classified as ‘allotheres’18, 19. The postcanine alveoli have approximately the same diameter as the canine alveolus and are also ‘undivided’ and reclined more posteriorly. The ultimate molar preserves a distinctive central valley that is closed-off anteriorly by a tall antero-buccal cusp and lingual ridge that connects to a lower posterolingual cusp, a pattern analogous to that of the fruit-eating hammer-headed bat Hypsignathus monstrosus20. Other haramiyidans shared a similar configuration with a chewing cycle that had both orthal and palinal movements4, but their upper molars differed in that the posterior cusp is larger than the anterior. In this regard, the ultimate molar of Cifelliodon is a closer analogue to Hypsignathus than that of other haramiyidans, and, except for its larger size and proportional differences, is nearly identical to an isolated tooth from the Lower Cretaceous of Morocco previously regarded as a multituberculate lower molar (M2): Hahnodon taqueti14, 21. Here, we regard the assignment of Hahnodon to multituberculates as problematic because, similar to Cifelliodon, the molar morphology includes a prominent antero-buccal cusp that is uncharacteristic of plagiaulacidan multituberculate M2s, a central valley that is closed off anteriorly and incompletely divided roots (multituberculates display completely bifurcated roots; Fig. 2c). Other authors have also questioned the multituberculate affinity of hahnodontid teeth, tentatively suggesting that they be moved to Haramiyida5.

Fig. 2: Dentition of Cifelliodon.
figure 2

a, Right posterior molar in buccal (left), occlusal (middle) and lingual (right) views. Arrows denote anterior. b, Reinterpretation of Hahnodon. Roots and posterior molar of Cifelliodon (top) shown for comparison. Note that Hahnodon in b does not match completely divided M2 roots of multituberculates (c) (Paulchoffatia, GUI MAM 8/73). Images are not to scale. Hahnodon redrawn from previous publications14, 21. a-b c, anterobuccal cusp; b f, anterobuccal furrow; I, incisor position (homologous positions to earlier haramiyidans); l f, posterolingual furrow; M2, second lower molar; p-l c, posterolingual cusp; PC4, fourth postcanine (posterior molar).

The skull shares additional features with some haramiyidans: upper incisor alveoli deep with procumbent orientation (Extended Data Figs. 68); lower canine fossa on maxilla-premaxilla absent; upper incisor number reduced to fewer than three; and upper postcanine loci reduced to five or fewer4, 9 (Extended Data Table 1). Moreover, Cifelliodon retains several plesiomorphic traits seen in Triassic mammaliamorphs, but not exhibited widely by previously known Cretaceous mammals, including: parietal lateral margins waisted rather than bulging outward; tabular bone present (also in the gondwanathere Vintana); pterygopalatal ridges present (also in select theriiforms, including cimolodontan multituberculates19); pterygoid flanges massive; secondary palate posterior limit well anterior to level of ultimate molar; maxilloturbinal supporting ridges absent on internal surface of maxilla; palatine anterodorsal expansion modest, such that maxilla and frontal maintain contact on anteromedial wall of orbit; and anterior part of the jugal extends to the facial part of the maxilla and forms a part of the anterior orbit. Nevertheless, the specimen also exhibits crown mammal features, including: a premaxilla facial process that is extensive and borders on the nasal; premaxilla with reduced internarial process; and posterior portion of jugal that terminates well anterior to the squamosal glenoid.

High-resolution X-ray computed tomography reveals an endocranial morphology that is transitional between earlier stem mammals and crown Mammalia (Fig. 3). The cranial vault is relatively modest in size (encephalization quotient 0.25–0.30; following a previously published method22), smaller than in the mammaliaform Hadrocodium (approximately 0.50) and some Cretaceous crown mammals (for example, Vincelestes, 0.37; Pucadelphys, 0.32)22, 23. Much of the enlargement in brain volume was formed by the massive olfactory bulbs and piriform cortex as in other early mammaliaforms. Given the medially waisted parietals and limited endocranial capacity, there was little expansion of the neocortex. The endocast surface is largely lissencephalic, olfactory bulbs massive and a shallow peduncle in place of the circular fissure, thus showing poor gross differentiation of neocortical structures. This contrasts with early therians and other crown mammals more typical of the Cretaceous that show greater differentiation of neocortical structures and their boundaries in endocasts (Fig. 3). Many posterior endocranial structures that are visible on the endocast (perilymphatic foramen cast, parafloccular cast) are located at about the level of the wide, shallow glenoid fossa as in other early mammaliaforms, but not as posteriorized as in large-brained therians. As in other early mammaliaforms, the endocranium of Cifelliodon supports the premise that haramiyidans exhibited a similar degree of encephalization and behavioural complexity that was driven mainly by olfaction22.

Fig. 3: Brain endocast in Cifelliodon and other mammaliamorphs.
figure 3

a, b, Endocast of Cifelliodon in left lateral (a) and posterior oblique (b) views. c, Cynodont Brasilitherium29. d, Mammaliaform Morganucodon22. e, Hahnodontid mammaliaform Cifelliodon. f, Hadrocodium22. g, Monotreme Obdurodon22. h, Multituberculate (based on Chulsanbaatar)22. i, Vincelestes23. j, Monodelphis30. Drawings scaled to their relative encephalization quotients (EQ). Major nodes are: (1) Mammaliaformes, (2) Mammalia, (3) Theriiformes, (4) Cladotheria (therians and close allies). cb, cerebellar hemisphere; f j, jugular foramen cast in perilymphatic fossa; nc, neocortex region; ob, olfactory bulb; pf, paraflocculus cast; II, III, IV, V1, cranial nerve exits near sphenorbital fissure; V3, V2, mandibular (and maxillary?) trigeminal branch roots; IX, X, glossopharyngeal and vagus nerve roots.

We assessed the phylogenetic position of Cifelliodon and Haramiyida using an updated morphological character matrix4, 24. Broad-level relationships support that Haramiyida is monophyletic and represents a close sister group to Mammalia4, 24, rather than a paraphyletic assemblage of crown mammals related to multituberculates (in contrast to Zheng et al.6) (Fig. 4 and Extended Data Fig. 9). We define Haramiyida as all mammaliaforms more closely related to Thomasia and Haramiyavia than to Didelphis. Within Haramiyida, Cifelliodon forms a relationship with Hahnodon in a polytomy with the gondwanathere Vintana18. However, the position of gondwanatheres is only weakly supported; for example, an alternate tree topology constraining Vintana to a sister taxon relationship with multituberculates results in a tree score that is only nine steps longer than our most parsimonious tree (Templeton test, P = 0.139). However, support for a monophyletic Haramiyida outside crown Mammalia is significantly more robust than moving the clade into crown mammals along with Vintana and multituberculates (as in the previously published topology18), which requires 24 additional steps (P = 0.032).

Fig. 4: Time-calibrated phylogeny and biogeography of Mesozoic mammaliamorphs.
figure 4

Bayesian consensus cladogram for the analysis of 125 mammalian and nonmammalian synapsid taxa and 538 morphological characters. Note that haramiyidan stem mammals, previously restricted to Laurasia in the Jurassic period, exhibited distributions into Gondwana by the Cretaceous period. Cretaceous nonmammalian mammaliamorphs were rare, but globally widespread. Values at the nodes represent Bayesian posterior probabilities. Values on the time scale are shown as million years before present. North Atlantic land bridge interval is indicated with a grey bar (maximum duration). Tr/J, Triassic/Jurassic extinction.

Although the haramiyidan affinity of hahnodontids has been tenuous in the absence of more complete fossils, the exceptional preservation of Cifelliodon helps to strengthen the group’s relationship to Haramiyida while revising its biogeographic history. Cifelliodon supports that haramiyidan stem mammals survived into the Early Cretaceous and sustained broad distributions in Laurasia and Gondwana, bridging Jurassic and Cretaceous terrestrial ecosystems across continents. The only previous haramiyidan material of Gondwanan origin included the isolated dentition of Allostaffia from the Jurassic of Tanzania25 and Avashishta—an isolated molar of uncertain affinities from the Upper Cretaceous of India26. The further presence of hahnodontids in North America and Africa mirrors underappreciated Pangaea-wide occurrences of Cretaceous tetrapods, including several shared dinosaurian taxa27, 28 that linked northern and southern continents much later than previously recognized (see Supplementary Information).

Ultimately, the Northern Hemisphere radiation of multituberculates and tribosphenic mammals was offset by losses of many archaic mammaliamorph groups during the mid-Cretaceous, including some of the longest-lived mammaliamorph subclades: tritylodontids, docodonts, haramiyidans, plagiaulacidan multituberculates and reduced numbers of triconodonts and ‘symmetrodonts’1. Cifelliodon supports the premise that large-bodied mammaliamorphs of diverse ecologies were present in North American communities up until this mid-Cretaceous diversity bottleneck. Our discovery of Cifelliodon and ongoing fieldwork in the Lower Cretaceous of Utah strengthen the hypothesis that corridors of Pangaea-wide dispersal were accessible to relictual clades of large and small vertebrates at least as recently as the Early Cretaceous27, 28. Consequently, the enigma of why some groups were better at dispersing than others (for example, the conspicuous absence of African multituberculates, an otherwise prolific clade) is more likely to be a consequence of collecting biases or unknown biotic factors rather than strictly tectonic controls or ocean barriers.

Methods

High-resolution X-ray computed tomography

The specimen was micro-CT scanned at the University of Utah’s Core Small Animal Imaging facility using a Siemens INVEON micro-CT scanner (aluminium filter and energy settings of 180 kV and 500 μA with a voxel size of 94.2 μm3). Additional focused scanning was performed at University of Chicago’s PaleoCT facility on a custom General Electric Phoenix VTOMEX S 240 to visualize the unerupted ultimate molars (energy settings of 180 kV and 500 μA with a voxel size of 25.3 μm3). Resulting DICOMs were extracted as image stacks and rendered as 3D models in Avizo 7.0 (Visage Imaging, San Diego, CA).

Phylogenetic methods

The previously published matrix4, 24 was modified to evaluate the relationships between Cifelliodon and other Mesozoic mammaliamorphs. In total, 538 morphological characters from 125 mammal and nonmammalian synapsid taxa were analysed using Bayesian inference and maximum parsimony. Characters 1–515 correspond to those from the previously published studies4, 24, characters 516–518 are borrowed from previously published data18 and characters 519–538 are new. All characters had equal weight and none were ordered. We performed the Bayesian analysis in MrBayes version 3.2.631 implementing the standard Mk model for morphological evolution32 with variable character rates. A log-normal rate distribution was chosen for the evolutionary model because of evidence, which suggested that log-normal distributions may be preferred for morphological datasets33, 34. We ran the analysis for ten million generations (with the first 25% removed as burn-in) and sampled the posterior distribution every 1,000 generations. To increase the chance of state swapping between chains and to help to reduce the standard deviation of split frequencies (that is, help the runs to converge), the difference in ‘temperature’ between chains was reduced to 0.06 (from the default setting of 0.10). Parsimony analysis was performed in TNT version 1.1 (tree analysis using new technology)35. A heuristic search (random addition sequence with 1,000 replicates) was performed and recovered 1,920 equally most parsimonious trees (tree score = 2,776 steps; consistency index = 0.317; retention index = 0.795). The resulting Bayesian consensus and strict consensus parsimony trees are shown in Extended Data Fig. 9.

Reporting summary

Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

Data availability

The specimen is reposited at the UMNH, Salt Lake City, USA, and additional data are reposited online at https://www.morphosource.org/. CT image stacks are available from the corresponding authors upon request. Life Science Identifier (LSID): the new genus and species are registered with Zoobank (http://zoobank.org): urn:lsid:zoobank.org:act:C9D9F344-E058-4E7F-AD58-BAED111F574E.